Pacifastacus Leniusculus) and the Implications for Sediment Recruitment to Rivers

Biotic and abiotic controls of burrowing by signal crayfish (Pacifastacus leniusculus) and the implications for sediment recruitment to rivers.

Harry Sanders

A Doctoral thesis submitted in partial fulfilment of the requirements

for the award of

Doctor of Philosophy of Loughborough University

(June 2020)

The signal crayfish (Pacifastacus leniusculus) is a globally invasive species, and since its introduction to Great Britain has developed a behaviour of burrowing into riverbanks, which has not been documented or observed in its native range. These have the potential to recruit sediment into river systems both directly, through burrow construction, and indirectly, by promoting accelerated erosion and the occurrence of mass failure events. This thesis investigates, for the first time, the drivers of crayfish burrowing, and the effects of crayfish burrows on fluvial geomorphology. The research was conducted using two field investigations that considered the spatial extent and temporal dynamics of crayfish burrows and associated sediment dynamics, and three laboratory studies that considered the physical and biotic mechanisms behind the associations and processes observed in the field. Crayfish burrows were distributed throughout Great Britain across all sampled habitats, and were, on average, 205 mm deep (range 20 – 870 mm), and excavated 1.15 kg of sediment per burrow directly into river systems. In rivers where burrows were present, an average burrow density of 0.39 burrows m- 1 of riverbank (maximum = 1.13 burrows m-1) was observed, directly contributing 0.93 t km-1 of fine sediment (maximum = 4.14 t km-1) to the channel. However, accelerated erosion caused by the presence of burrows recruited 29.5 times more sediment than burrowing alone, with burrowing and the associated acceleration of retreat and collapse recruiting an additional 25.4 t km-1 a-1 of sediment at one field site; an estimated 29.8% of total bank sediment yield. This was supported by physical modelling, which showed that the spatial distribution of burrows was important for determining retreat rates and mechanisms. Numerical modelling successfully predicted the presence, density, and geomorphic impacts of crayfish burrows. Burrow presence was attributed to associations between crayfish population densities and sediment grain size distributions, and burrow density was dependent on river flow velocity. This modelling was supported by mesocosm and flume experiments with live crayfish. Further, mesocosm experiments undertaken in the UK and the USA showed that all signal crayfish have the capacity to burrow, suggesting that burrowing and associated geomorphic change may occur in any invasive signal crayfish population. This thesis has shown that signal crayfish burrows can have substantial geomorphic impact, and is the first research to quantify the geomorphic role of signal crayfish burrows, and more broadly to quantify the relative importance of biotic and abiotic forcing of fluvial sediments. Future research should better consider animals as geomorphic agents, both directly, but also indirectly, by considering the facilitative role that zoogeomorphology may have on wider system processes.

Acknowledgements

Thank you to everyone who has been a part of my PhD experience. It would not have been possible without the support, help, and friendship of so many people.

To my supervisors, Steve Rice, thank you for your trust and belief in me, and providing this opportunity that I have enjoyed so much. Thank you for all the time you’ve spent teaching me to write proper, your constant enthusiasm, guidance, and dedication to making this research happen, and supporting me as researcher and friend throughout. To Paul Wood, thank you all the statistics help, for your ever-present support, advice and good jokes at the stressful times, and for your care, friendship, and bad jokes always. You have both made this a thoroughly enjoyable experience; I am forever indebted.

This study required a considerable amount of field and lab work which would not have been possible without the assistance of many people. In particular, special thanks go to Richard Mason and Milly Bulcock for your continued and dedicated field and pastoral support throughout this thesis. To Richard, whatever ridiculous situation I have thrown you in, from eating spaghetti hoops with nothing but a cable tie in the boot of a department car a Scottish layby, to being my low speed getaway driver for sand heists, you have always been there, and without your supportive energy, advice, and mountainous adventures this thesis would not be finished. To Milly, thank you for injecting so much fun and laughter to our months of travels across the country, meeting a whole new cross section of society at Ilham Hall and beyond (Thorpe Cloud still awaits!). Thank you for being my shoulder to cry on when routinely finding field sites to be navigable by shipping liner, or not to exist at all. To you both, I am eternally grateful.

Thank you also to a host of other people who have helped with fieldwork; to Beth Worley, Ciara Dwyer, Daniel Mills, Ellen Goddard, Hazel Wilson, Jess Green, Kate Mathers, Laura Crawford, Rebecca McKenzie, Sam Wilby and Sarah Sanders for assistance in the field. Also, I must thank the catchment managers, land owners and facilitators for taking time to permit and arrange access; my thanks to Clare Warren, Gill and Richard Trace, Nathan Hall, Julian Payne, and all at the Environment Agency, Paul Bradley and all at PBA Applied Ecology, Mr & Mrs Jones, Tony Booker and Paul Sansom-Timms, Willy Yeomans and David McColl, and all at the Clyde River Foundation, and the Clyde angling community.

iii

Thank you also to the gy.labs team, in particular to Richard Harland and Rebecca McKenzie, for your laboratory assistance in constructing flumes, and allowing me to shovel copious amounts of clay in your laboratories for the last three years. Particular thanks go to Bas Bodewes for all your help in constructing, running, analysing, and solving all the problems the Friedkin flume threw at us. Thanks also to Beth Worley, Daniel Mills, Davide Vettori, Guy Tallentire, James Smith, Keechy Akkerman and Richard Mason for laboratory assistance throughout.

Thank you also to Lindsey Albertson, and all at the AlberCross lab for hosting me for a two month stay at Montana State University. My thanks also to Ben Tumolo, Eric Scholl, Holden Reinhart, Mike MacDonald, and in particular to Zach Maguire for crayfish collection, and to Zach Maguire for laboratory assistance in Montana.

I would also like to thank Loughborough University, the British Society for Geomorphology, the Royal Geographical Society and Santander for your financial support of this project.

Lastly, but my no means least, thank you to everyone in the Loughborough Geography community, to all past and present members of Loughborough Student’s Mountaineering Club, Klimmen climbing club, Loughborough Geography cricket team, The Jackals quiz team, and everyone else who has kept me sane throughout this thesis. Here’s to many more years of high- octane geomorphology!

Table of Contents Abstract ii Acknowledgements iii Table of Contents v List of Tables xi List of Figures xvi

Chapter 1 Introduction 1 1.1 Research Context and Development of Research Theme 2 1.2 Aims and Research Objectives 3 1.3 Thesis Structure 4

Chapter 2 Zoogeomorphology: Crayfish and Riverbanks 7 2.1 Chapter Structure 8 2.2 Zoogeomorphology and Ecosystem Engineering 8 2.2.1 Zoogeomorphology: Biotic Considerations 12 2.2.2 Zoogeomorphology: Abiotic Considerations 15 2.2.3 Biogeomorphology and Zoogeomorphology of Riverbanks 16 2.3 Fine Sediment, River Ecology, and River Geomorphology 21 2.3.1 Channel Sources of Fine Sediment 21 2.3.2 Non-Channel Sources of Fine Sediment 22 2.3.3 Deposition and Storage of Fine Sediment 23 2.3.4 Geomorphological Impacts of Fine Sediment 24 2.3.5 Chemical Impacts of Fine Sediment 24 2.3.6 Ecological Impacts of Fine Sediment 25 2.4 Signal Crayfish as Zoogeomorphic Agents 30 2.4.1 Crayfish as Geomorphic Agents 31 2.4.2 The Natural History of Crayfish Burrows 37 2.4.3 Types of Crayfish Burrows 38 2.4.4 Where Crayfish Burrow 39 2.4.5 Why Crayfish Burrow 40 2.4.6 Geomorphic Implications of Crayfish Burrows 41 2.4.7 Modelling Crayfish Burrowing 42 2.5 Justification of Research Objectives 43

Chapter 3 Quantifying the biological, hydrological, and geophysical controls of riverbank burrowing by signal crayfish 45 3.1 Introduction 46 3.2 Aims 46 3.3 Methods 47 3.3.1 Overall description of approach used 47 3.3.2 Field Site Selection 49 3.3.3 Sampling Strategy 49 3.3.4 Crayfish Burrow Survey 50 3.3.5 Candidate variables for characterising crayfish burrows 53 3.3.6 Statistical Analyses 66 3.3.7 Predictive Modelling 69 3.4 Results 77 3.4.1 Quantifying excavated sediment and burrow characteristics 77 3.4.2 Univariate Associations Between Variables and Burrow Characteristics 82 3.4.3 Modelling Burrows and Sediment Input 86 3.5 Discussion 102 3.5.1 The Geomorphological Significance of Burrow Characteristics 102 3.5.2 Modelling Burrow Presence and Absence: Ideal Models 105 3.5.3 Modelling Burrow Density: Ideal Models 109 3.5.4 Ideal Versus Operational Models 111 3.5.5 Model Contributions and Univariate Associations of Discussion Environmental Variables 112 3.5.6 Discussion Summary 124

Chapter 4 The effects of flow velocity on signal crayfish burrowing and associated sediment dynamics 127 4.1 Introduction 128 4.2 Aims 129 4.3 Methods 130 4.3.1 Physical Set Up 130 4.3.2 Mesocosm Treatments 131 4.3.3 Flow Velocity 134

4.3.4 Crayfish 135 4.3.5 Experimental Procedure 135 4.3.6 Quantifying Sediment Erosion 136 4.3.7 Statistical Analyses 139 4.4 Results 141 4.4.1 Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent? 141 4.4.2 How does flow velocity influence crayfish burrowing behaviour? 144 4.4.3 Does crayfish burrowing change the rate of erosion of non-burrowed sediment? 145 4.4.4 Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates? 148 4.4.5 Results Summary 150 4.5 Discussion 150 4.5.1 Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent? 150 4.5.2 How does flow velocity influence crayfish burrowing behaviour? 151 4.5.3 Does crayfish burrowing change the rate of erosion of non-burrowed sediment? 154 4.5.4 Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates? 155 4.5.5 General Discussion 157 4.5.6 Discussion Summary 163

Chapter 5 Quantifying rapid behavioural change in native and invasive populations 165 5.1 Introduction 166 5.2 Aims 166 5.3 Methods 167 5.3.1 Elected Treatments 169 5.3.2 Collection of Animals 171 5.3.3 Physical Setup and Measurements 172 5.3.4 Data Analyses 176 5.3.5 Consideration of Temperature 176

vii

5.4 Results 178 5.4.1 General Results 178 5.4.2 How does shelter availability affect the propensity of crayfish to burrow? 179 5.4.3 How does population density affect the propensity of crayfish to burrow? 183 5.4.4 How does population provenance affect the propensity of crayfish to burrow? 183 5.4.5 Results Summary 188 5.5 Discussion 188 5.5.1 General Results 190 5.5.2 How does shelter availability affect the propensity of crayfish to burrow? 191 5.5.3 How does population density affect the propensity of crayfish to burrow? 194 5.5.4 How does population provenance affect the propensity of crayfish to burrow? 195 5.5.5 Discussion Summary 200

Chapter 6 The impacts of signal crayfish burrows on riverbank retreat: a biophysical sediment budget 203 6.1 Introduction 204 6.2 Aims 205 6.3 Methods 206 6.3.1 Field Site Selection 206 6.3.2 The Use of Erosion Pins 209 6.3.3 Erosion Pin installation 213 6.3.4 Quantifying Burrows 216 6.3.5 Data Collection 217 6.3.6 Data Processing 219 6.3.7 Data Visualisation 220 6.3.8 Statistical Analyses 221 6.3.9 Modelling Sediment Input 223 6.4 Results 225 6.4.1 General Results 225 6.4.2 Do signal crayfish burrows accelerate bank retreat? 225 6.4.3 How do crayfish burrows accelerate bank retreat? 232

viii

6.4.4 How much sediment does accelerated bank retreat associated with crayfish burrows recruit to river systems? 241 6.4.5 Results Summary 242 6.5 Discussion 244 6.5.1 General Results 244 6.5.2 Do signal crayfish burrows accelerate bank retreat? 245 6.5.3 Diffuse Erosion 249 6.5.4 Bank Collapse 250 6.5.5 Bank Shape 257 6.5.6 Modelling Sediment Input 260 6.5.6 Further Discussion 262 6.5.7 How does burrowing impact bank erosion through time? 263 6.5.9 Discussion Summary 265

Chapter 7 The impact of simulated burrows on riverbank retreat: laboratory experiments 266 7.1 Introduction 267 7.2 Aims 267 7.3 Methods 268 7.3.1 Physical Set Up 268 3.2 Treatments 273 7.3.3 Image Analysis 275 7.3.4 Data Analyses 276 7.3.5 Statistical Analyses 291 7.4 Results 282 7.4.1 The effect of burrow density on bank retreat 283 7.4.2 The effect of vertical burrow position and clustering on bank retreat 287 7.5 Discussion 292 7.5.1 The effect of burrow density on bank retreat 293 7.5.2 The impact of vertical burrow position and clustering on bank retreat 303 7.5.3 Discussion Summary 306

Chapter 8 Summary, key themes, future research, and concluding remarks 308 8.1 Fulfilment of thesis objectives 309

8.2 Research themes 314 8.2.1 Biological and geophysical energy should not be considered independently 314 8.2.2 The value of field data 317 8.2.3 The use of the mesocosm 318 8.3 Future research directions 322 8.3.1 Quantifying burrows 322 8.3.2 Temporal persistence of burrows 323 8.3.3 Burrows and hydrology 326 8.3.4 Modelling sediment 326 8.4 Concluding remarks 327

Reference List 328

List of Tables

Table 2.1: Recorded signal crayfish burrow densities in previous studies. 19

Table 3.1: Field sites used in analysis. 48

Table 3.2: Variables measured in the field survey. 55

Table 3.3: Land use categories employed. 56

Table 3.4: Grain size analysis of one riverbank material sample using different sizing methods. 62

Table 3.5: Regression models constructed to predict the presence of burrows, the density of burrows, and the mass of sediment excavated by burrowing activity. 68

Table 3.6: Tested normality of measured and calculated mean variables for the 38 considered rivers. 70

Table 3.7: Table of significant PCs from PCA. 71

Table 3.8: Table of component loadings from PCA. 71

Table 3.9: Univariate logistic regression Wald statistics used in variable redundancy analysis. 73

Table 3.10: Removed PCA selected variables prior to logistic regression analysis. 72

Table 3.11: PCA selected variables carried forward for logistic regression analysis. 74

Table 3.12: Table of component loadings from PCA. 75

Table 3.13: Table of variables removed from PCA selection due to collinearity with greater weighted variables. 76

Table 3.14: Descriptive statistics of burrow dimensions. 78

Table 3.15: Mean and median burrow dimensions by site. 79

Table 3.16: Pairwise (Mann-Whitney with Bonferroni correction) comparisons of burrow depth, entrance area and mass between sites. 80

Table 3.18: t-tests (two tailed) between burrowed and non-burrowed. 83

Table 3.19: Linear associations between tested variables and BD. 87

Table 3.20: Linear associations between tested variables and MD. 88

Table 3.21: Logistic regression models (enter) statistics. 92

Table 3.22: Logistic regression models (enter) variable statistics. 92

Table 3.23: Logistic regression model (stepwise) statistics, considering the reduced river hydrology model. 93

Table 3.24: Logistic regression model (stepwise) variable statistics, considering the reduced river hydrology model. 93

Table 3.25: Differences between global and restricted logistic regression models. 94

Table 3.26: Model confidence intervals for use in Equation 3.15. 94

Table 3.27: The density of crayfish required to initiate burrowing in each river. 95

-1 Table 3.28: Predictive multiple linear regression models (enter) for BD (burrows m riverbank). 97

-1 Table 3.29: Predictive multiple linear regression models (enter) for MD (kg m riverbank). 98

-1 Table 3.30: Predictive multiple linear regression models (stepwise) for BD (burrows m riverbank). 100

-1 Table 3.31: Predictive multiple linear regression models (stepwise) for MD (kg m riverbank). 101

Table 3.32: Change in r2 values between enter and stepwise regression. 102

Table 3.33: Reported burrow densities from previous studies. 104

Table 3.34: Bank scale model statistics. 107

Table 3.35: Bank scale model input variable loadings and statistics. 108

Table 3.36: Retained and removed variables from differing enter and stepwise regression models. 112

Table 3.37: Summary table of best models advised for predicting the probability of burrow occurrence, the threshold crayfish density at which burrowing is initiated,

xii

burrow density (where burrowing occurs), and the mass of sediment excavated by burrows (where burrowing occurs). 125

Table 4.1: Flow velocities used in experiments. 134

-2 Table 4.2: Mass (kg m ) of total sediment eroded (Mtot’) when crayfish were present and absent under each flow velocity. 142

-2 Table 4.3: Observed difference in mass of sediment eroded (Mtot’; kg m ) between Crayfish and Control treatments at different flow velocities. 143

-2 Table 4.4: Mass (kg m ) of sediment recruited from crayfish burrows (MB’) when all runs and only burrowed runs are considered. 145

Table 4.5: Mass (kg m-2) of sediment recruited through mechanisms other than burrowing

(MNB’). 146

-2 Table 4.6: Mass of total sediment eroded (Mtot’, kg m ) between Artificial Burrows and Control treatments at all flow velocities. 149

Table 4.7: Flow required to dislodge crayfish species downstream in previous experiments. 153

Table 4.8: Sites (12) where mean flow velocity was recorded as greater than the medium flow velocity tested. 153

Table 5.1: Treatment matrix. 169

Table 5.2: Crayfish population locations, what the study population represents, and their associated in text abbreviations. 169

Table 5.3: Population details of crayfish collected from the four locations. 172

Table 5.4: Number of replicates undertaken for each treatment. 176

Table 5.5: Mass of sediment excavated (g) and burrow size (g) of crayfish from UKB, UKNB, and both UK populations in a heated and unheated mesocosm with no shelter. 177

Table 5.6: Mass of sediment excavated (g) and burrow size (g) considering all populations and all treatments. 180

Table 5.7: Predatory fish that may consume crayfish found in the rivers where crayfish were collected. 197

xiii

Table 5.8: Associations between damaged crayfish and the presence of burrows. 199

Table 6.1: A selection of previous studies using erosion pins to quantify riverbank erosion. 207

Table 6.2: Table of banks surveyed. 213

Table 6.3: Calculated burrow metrics of burrow density standardised for bank area (BA; -2 -1 burrows m ) and bank length (BL; burrows m ), burrow volume density 3 -2 3 -1 standardised for bank area (VA; cm m ) and bank length (VL; cm m ), and 2 -2 burrow entrance area density standardised for bank area (EA; cm m ) and bank 2 -1 length (EL; cm m ) for all surveyed banks. 216

Table 6.4: Values calculated for negative value methods and the metrics used for (a) do signal crayfish burrows accelerate bank retreat?, (b) diffuse erosion; (c) bank collapse, and (d) bank shape. 221

Table 6.5: Table of results, considering BL, VL, EL, and Bank Retreat. 229

Table 6.6: Correlation values of the association between total retreat and burrow metric values at both sites. 229

Table 6.7: Table of results, considering VL, EL, and Gross Diffuse Erosion, and Only Retreat Diffuse Erosion. 231

Table 6.8: The association between diffuse erosion and crayfish burrow metrics, when negative values were considered as (a) ‘gross retreat’ and (b) ‘only retreat’. 232

Table 6.9: Table of results, considering BA, BL, VA, VL, EA, EL, and Only Collapse Retreat, Percentage of Pins Collapsed, and Percentage of Retreat as Collapse. 235

Table 6.10: The association between crayfish burrow metrics and (a) total collapses, (b) the percentage of pins collapsed; and (c) the percentage of retreat recorded as collapse, all at the River Bain. 236

Table 6.11: The association between crayfish burrow metrics and change in riverbank shape, calculated using total retreat at both sites. 237

Table 6.12: Table of results, considering EA, EL, and Relative Change in Bank Shape. 238

xiv

Table 6.13: The relative inputs of bank erosion as a result of direct sediment input from crayfish burrows, the accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows. 242

Table 6.14: Observed variability in riverbank retreat along rivers where crayfish burrows were not present. 247

Table 6.15: The total and proportional contributions of sediment to the River Bain from accelerated erosion in the event of an increased distribution of crayfish burrow presence. 260

Table 6.16: The association between burrow volume density per meter of riverbank and bank retreat variables since the last recording, and over the total period. 263

Table 6.17: The association between burrow volume density per meter of riverbank and bank retreat variables since the start of the study. 263

Table 7.1: Results of Kruskal-Wallace tests for significant differences between the seven treatments described in Section 7.3.2, for each of the bank retreat metrics described in Section 7.3.4. 282

Table 7.2: Bank retreat metrics (n = 5) associated with increasing burrow density from Friedkin flume experiments. 283

Table 7.3: Bank retreat metrics (n = 5) of different burrow locations from Friedkin flume experiments. 289

Table 7.4: Change in bank surface area and drainage flow path distance of experimental bank in the presence of burrows. 294

Table 7.5: Rate of bank retreat under adjusted preliminary conditions. 298

Table 8.1: Sediment recruited directly through burrowing over time. 323

Table 8.2: Modelled mass of sediment excavated to all rivers considered in Chapter 3. 324

List of Figures

Figure 1.1: Thesis structure. Objectives stated are detailed in section 1.2. 6

Figure 2.1: The geomorphology of salmon redds has been modelled to change river profiles over geological timescales. 9

Figure 2.2: Biological energy driving abiotic processes. 11

Figure 2.3: Zoogeomorphology by low per capita mass taxa in high population densities.13

Figure 2.4: Bivalve surveys on the upper tidal River Thames, UK. 15

Figure 2.5: Burrow induced bank failures. 19

Figure 2.6: Impacts of fine sediment on juvenile salmonid fish. 28

Figure 2.7: Salmonid egg survival rates reduce with increasing fine sediment coverage. Reproduced from Kemp et al. (2011). 30

Figure 2.8: Signal crayfish are widespread, and have the capability to undertake substantial geomorphic work. 32

Figure 2.9: Signal crayfish presence in English waterbodies. 33

Figure 2.10: A signal crayfish inhabiting a burrow and crayfish burrows in riverbanks. 34

Figure 2.11: Main body parts of a crayfish. 35

Figure 2.12: Diel fluxes observed in turbidity time series. 36

Figure 3.1: Spatial distribution of sampled field sites. 50

Figure 3.2: The measured angles of the burrowed banks. 51

Figure 3.3: Schematic of sampled river reach. 54

Figure 3.4: In cases where crayfish were not present in traps, their presence was checked by searching for (a) live specimens, especially juveniles, (b) cadavers, (c) cadaver remnants, and (d) chelae. 64

Figure 3.5: Measured crayfish carapace length, from the tip of the rostrum to the posterior margin of the carapace. 65

Figure 3.6: Frequency distribution of (a) burrow depth, (b) burrow entrance height; (c) burrow entrance width, and (d) mass of excavated sediment. 78

xvi

Figure 3.7: Significant differences between (a) river width, (b) bankfull discharge; (c) conductivity, (d) grains >90mm; (e) bank mass classed as pebbles, and (f) bank mass classed as silt and clay in rivers where burrows are present, and burrows are absent. 85

Figure 3.8: Selected associations between MD (i-l) and BD (m-p) with (i, m) proportion of riverbank mass classified as silt and clay, (j, n) distance of overhanging riparian vegetation; (k, o) time since initial crayfish invasion, and (l, p) mean male crayfish size. 89

Figure 3.9: Association between riverbed grainsize and the strength of the relationship

between riverbed grainsize and (a) BD and (b) MD. 91

Figure 3.10: Logistic regression prediction models of Table 3.21. 91

Figure 3.11: Relationships between computed regression variables and crayfish burrow density and the mass of sediment directly excavated by crayfish burrowing considering the application models. 99

Figure 3.12: Sedimentation of the bed of a channel between the River Gade and the River Chess. 105

Figure 3.13: Proof-of-concept logistic regression model for bank-scale prediction of burrow presence or absence in the crayfish-infested river reaches. 108

Figure 3.14: The frequency and relative sediment contribution of surveyed crayfish burrow density. 110

Figure 3.15: Associations between (a) river altitude and (b) river slope with the mass of sediment excavated by burrowing. 113

Figure 3.16: Hapturnell Burn at (a) base flow and (b) bankfull flow, approximately 16 hours apart following a heavy rain event. Colleague for scale. 115

Figure 3.17: Associations between burrow depth and flow velocity considering (a) the riverbed at base flow and (b) 0.6 depth at bankfull. 115

Figure 3.18: Associations between burrow size and grain stress at (a) base flow and (b) bankfull flow. 119

Figure 3.19: Riparian vegetation coverage was very high at (a) Potwell Dyke, (b) Gaddesby Brook (Gaddesby); (c) River Ouzel (Lower), and (d) River Tove. 121

xvii

Figure 3.20: Associations between burrow density and crayfish population density. 123

Figure 3.21: Association between time since invasion and burrow density. 124

Figure 4.1: A plan view (top) and a cross-sectional view looking downstream (bottom) of the physical set up used in the flume experiments. 132

Figure 4.2: The physical set up of the flume, before (top) and during (bottom) an experimental run. View looking upstream in both cases. 133

Figure 4.3: Schematic of flume from above. 133

Figure 4.4: Identification of burrows and collapses. 137

Figure 4.5: Total sediment eroded when crayfish were absent on a Control bank. 141

Figure 4.6: Total sediment eroded in the presence and absence of two crayfish. 142

Figure 4.7: Net difference upon the introduction of crayfish compared to an absence of crayfish. 143

Figure 4.8: In the high flow velocities, crayfish immediately opted to use the shelter to escape the flow (a, b, c), often cohabiting. 144

Figure 4.9: Mass of sediment recruited from crayfish burrows (MB’) when (a) all runs and (b) only burrowed runs are considered. 145

Figure 4.10: Mass of sediment recruited through mechanisms other than burrowing (MNB’) in the presence and absence of crayfish. 147

Figure 4.11: Association between sediment recruited directly from burrows (MB’) and

sediment recruited from non-burrowed sources (MNB’). 148

Figure 4.12: Mass of total sediment eroded between Artificial Burrows and Control treatments at all considered flow velocities. 149

Figure 4.13: Association between flow velocity and burrow depth in the field (reproduced from Chapter 3). The association is significant. 154

Figure 4.14: Sediment recruited from collapses (MC’) in each considered flow velocity in the Control treatment. 156

Figure 4.15: A crayfish climbing the wire gauze mesocosm walls. 158

Figure 4.16: Observed sediment transport during the current experiments. 160

xviii

Figure 4.17: The measured proportional contribution of interactive effects between biological and geophysical energy (see Figure 4.16). 161

Figure 4.18: Conceptual diagram of the relative and interactive effects of geophysical and biological energy in driving sediment recruitment. 162

Figure 5.1: Shelter availability treatments used in the experiments; (a) no shelter, (b) a single large rock; and (c) a deep silt substrate. 170

Figure 5.2: A holding tank for crayfish between experimental runs in the UK. 173

Figure 5.3: Schematic diagram of mesocosms used in the experiments. 174

Figure 5.4: The effect of temperature on (a) the mass of sediment excavated and (b) burrow size by one crayfish, for both UK populations. 177

Figure 5.5: A comparison of the size of crayfish burrows. 178

Figure 5.6: The effect of shelter availability on (a) the mass of sediment excavated and (b) burrow size in the low population treatment. 179

Figure 5.7: The effect of the presence of a large rock shelter on the total mass of sediment excavated (a and b) and mean burrow size (c and d) under medium (a and c) and high (b and d) population densities. 181

Figure 5.8: The effect of population density on the total mass of sediment excavated (a and b) and mean burrow size (c and d) when a large rock shelter was absent (a and c) and present (b and d). 183

Figure 5.9: Differences between populations considering (a) the total mass of sediment excavated, and (b) burrow size when a shelter was absent in the low population density treatment. 183

Figure 5.10: Differences between populations on the total mass of sediment excavated (a and b) and burrow size (c and d) when a rock was absent (a and c) and present (b and d) in the medium population density treatment. 185

Figure 5.11: Differences between populations on the total mass of sediment excavated (a and b) and burrow size (c and d) when a rock was absent (a and c) and present (b and d) in the high population density treatment. 186

Figure 5.12: Differences between populations on the total mass of sediment excavated (a) and burrow size (b) across all runs where all populations were considered. 186

xix

Figure 5.13: A burrow constructed by a signal crayfish. 188

Figure 5.14: A crayfish that has not burrowed but nevertheless caused bank retreat through walking on the banks and recruiting sediment through bioturbation. 188

Figure 5.15: The photographed crayfish was hand caught from the entrance of a burrow it was inhabiting. 190

Figure 5.16: Differences in burrow morphology between experiments. 191

Figure 5.17: A signal crayfish burying itself in the deep silt substrate. 192

Figure 5.18: Conceptual diagram for the propensity for crayfish to burrow after invasion.

200

Figure 6.1: Studied banks were evenly spatially distributed. 207

Figure 6.2: Banks selected for study were bare of vegetation and on straight sections of channel. 208

Figure 6.3: Schematic of erosion pins. 209

Figure 6.4: Erosion pin monitoring on the River Bain. 214

Figure 6.5: Fieldwork was not always possible at Gaddesby Brook. 217

Figure 6.6: An erosion pin being measured with a tape measure. 217

Figure 6.7: Photograph of a bank collapse on the River Bain (top) and graphical heat map of the bank generated from erosion pin recordings. 225

Figure 6.8: 3D visualisation of all banks at Gaddesby Brook. 226

Figure 6.9: 3D visualisation of all banks at the River Bain. 227

Figure 6.10: 3D visualisation of a single bank at Gaddesby Brook. 228

Figure 6.11: The association between (a) BL, (b) VL, and (c) EL on the gross retreat of riverbanks at both rivers over the full time period. 230

Figure 6.12: The association between (a and c) VL and (b and d) EL and (a and b) gross diffuse erosion and (c and d) only retreat diffuse erosion at both rivers for the full time period. 233

Figure 6.13: The association between BL and collapse retreat at the River Bain. 234

Figure 6.14: The association between EL and relative change in riverbank shape since the start of the study, considering total retreat at Gaddesby Brook. 238

Figure 6.15: Changes in bank profile over time at Gaddesby Brook. 239

Figure 6.16: Changes in bank profile over time at the River Bain. 240

Figure 6.17: Values of potential discrepancy. 246

Figure 6.18: Variation of retreat between (a) averaged banks and (b) pins on single riverbanks at Gaddesby Brook. 248

Figure 6.19: Flow hydrograph compared to sampling dates and observed erosion. 251

Figure 6.20: A small mass failure directly above a crayfish burrow. 252

Figure 6.21: Collapses from three banks on the River Bain. 253

Figure 6.22: The entrances of crayfish burrows could be seen after collapse, showing that deep burrows have the capacity to influence multiple collapses. 254

Figure 6.23: Conceptual model of hydrologically driven geotechnical bank retreat as a result of crayfish burrows. 255

Figure 6.24: Bank collapses at the River Bain. 257

Figure 6.25: An undercut bank on the River Bain. 258

Figure 6.26: The measured angles of the burrowed banks. 259

Figure 7.1: Friedkin flume setup. 269

Figure 7.2: Schematic of Friedkin flume set up. 270

Figure 7.3: Reduction in average block volume for a block of sand undergoing erosion in a Friedkin flume. 271

Figure 7.4: The location of ‘burrow’ positions in experimental treatments. 273

Figure 7.5: Flowchart of image analysis undertaken to record the rate of bank retreat. 276

Figure 7.6: Example images used in the rectification process. 277

Figure 7.7: Bank retreat over time in a Friedkin run. 279

Figure 7.8: Bank retreat over time in a Friedkin run. 280

Figure 7.9: The effect of burrow density on the rate of total bank retreat. 283

xxi

Figure 7.10: The rate of bank retreat over time in Friedkin experiments when (a) zero, (b) four; (c) eight, and (d) twelve burrows were present. 284

Figure 7.11: The effect of burrow density on (a) proportion of sediment recruited via collapse, (b) mean number of collapse events per run; (c) the mean width of collapses, and (d) the mean length of collapses. 285

Figure 7.12: The effect of burrow density on the rate of initial collapse. 286

Figure 7.13: The influence of burrow location on (a) the rate of total bank retreat, and (b) the size of the initial collapse. 287

Figure 7.14: The influence of burrow location on the rate of bank retreat when (a) total retreat, and (b) diffuse erosion are considered. 287

Figure 7.15: The rate of bank retreat over time in Friedkin experiments when burrows were (a) on the top row; (b) on the bottom row, and (c) clustered. 288

Figure 7.16: The effect of burrow location on (a) proportion of sediment recruited via collapse, (b) mean number of collapse events per run; (c) the mean width of collapses, and (d) the mean length of collapses. 290

Figure 7.17: The effect of burrow locations on the rate of initial collapse. 291

Figure 7.18: Conceptual model for expected erosion. 296

Figure 7.19: Observed erosion rate from Friedkin experiments, with reference to conceptual model. 297

Figure 7.20: Erosion through time in the treatment without burrows. 300

Figure 7.21: Oblique photographs of initial collapses in the presence of (a) four, (b) eight; and (c) twelve burrows. 301

Figure 7.22: Processed images of banks after the initial collapse phase in the presence of (a) zero, (b) four; (c) eight, and (d) clustered burrows. 302

Figure 7.23: Crayfish burrows clustered on (a) the River Chess and (b) Gaddesby Brook, and (c) burrows arranged longitudinally at the bottom of the bank at Gaddesby Brook. 304

Figure 8.1: Conceptual models of the interaction between biotic and abiotic forced in driving fluvial sediments. 314

xxii

Figure 8.2: Conceptual framework for the importance of biotic energy in determining total geomorphic outputs. 315

Figure 8.3: A successful preliminary trial at casting burrows in mesocosms. 322

xxiii

Chapter 1

Introduction

1.1 Research Context and Development of Research Theme

The physical impacts of biota on the environment have long been recognised and documented (Darwin 1891) but were not widely considered until the 1990s (Viles 1988; Jones et al. 1994; Butler 1995). Whilst significant research has been undertaken to understand how habitat affects animals, relatively little research has investigated how animals affect habitat. For example, in fluvial geomorphology the deleterious impacts of fine sediment pollution on zooplankton (Bilotta and Brazier 2008), macrophytes (Jones et al. 2012a), macroinvertebrates (Wood and Armitage 1997; Jones et al. 2012b) and fish (Kemp et al. 2011) are widely understood, but comparatively little research has been undertaken investigating biota as a driver of fine sediment dynamics in rivers (Rice et al. 2012; Emery-Butcher et al. 2020). ‘Ecosystem Engineering’ (Jones et al. 1994) and ‘Zoogeomorphology’ (Butler 1995) have primarily been considered by ecologists investigating the influence of biota on habitat functioning, and geomorphologists investigating the influence of biota on sediment transport, respectively. However, these two fields of research developed largely independently (Hannah et al. 2004; Dollar et al. 2007; Vaughan et al. 2009; Rice et al. 2010; Butler and Sawyer 2012; Rice et al. 2012). Interdisciplinarity between ecosystem engineering and zoogeomorphology has been rare (Butler and Sawyer 2012), and there is an important need to combine the disciplines to gain a holistic understanding of the importance of biota for river system dynamics (Reinhardt et al. 2010; Wheaton et al. 2011; Harvey and Bertoldi 2015).

This is of particular concern given both the per capita and community level potential that biota have in driving lotic sediment transport (Statzner 2012) and is of direct relevance to the Anthropocene due to the introduction, establishment, and spread of invasive species (Pinmentel et al. 2001; Early et al. 2016). Whilst many species modify habitat and recruit large quantities of sediment into river systems, their effects are often very localised, as the geomorphic activity of native animals is typically accounted for in system wide processes (Moore 2006). However, invasive species may not be constrained by the characteristics of their native habitats and as a result represent a disturbance (Harvey et al. 2011). Invasive species have the capacity to enact threshold changes within a system and are thus a particular concern within ecosystem engineering and zoogeomorphology research (Crooks 2002; Fei et al. 2014; Emery-Butcher et al. 2020).

Taxa of substantial concern are crayfish. Crayfish are some of the most successful invasive species worldwide (Gherardi 2013; Kouba et al. 2014). Crustaceans account for 53% of

2 invasive species in European freshwater systems (Karatayev et al. 2009), and 46% of all crayfish species are invasive (Vila et al. 2010). Crayfish are also regarded as an influential aquatic geomorphic agent with regards to their potential to entrain excess sediment under normal flow conditions (Statzner 2012), and so invasive crayfish have the potential to facilitate widespread geomorphic change. In particular, signal crayfish (Pacifastacus leniusculus) is the most widespread invasive crayfish throughout Europe and the UK (Kouba et al. 2014), and is now present in 60% of English sub-catchments, and is expanding at a rate of 1.6% per year (Chadwick 2019). Since its introduction to the UK, signal crayfish have developed a behaviour of burrowing into riverbanks, which has not been documented or observed in its native range, or in other invaded territories (Guan 1994; Harvey et al. 2014; Faller et al. 2016). Signal crayfish burrowing is thought to recruit sediment into river systems (Faller et al. 2016; Rice et al. 2016) and is hypothesised to have implications for riverbank retreat and mass failure (Harvey et al. 2019), but there are currently limited empirical data to support these arguments (Harvey et al. 2019). This thesis aims to remedy this shortcoming by investigating, quantifying and explaining the role of signal crayfish in driving sediment recruitment and river bank erosion in rivers. It is underpinned by the more general goal of improving our understanding of how animals, and in particular invasive species, are important drivers of fluvial sediment dynamics.

1.2 Aims and Research Objectives

The overall aim of this thesis is to investigate and quantify the biotic and abiotic drivers of burrowing behaviour and the geomorphological impacts of burrowing on river bank erosion and sediment recruitment to river channels. This is achieved through a series of in situ field studies and ex situ physical modelling and mesocosm experiments in the UK and USA. Specifically, this thesis aims to address the following objectives:

1. To quantify the mass of sediment excavated by signal crayfish burrowing through an extensive field study of 39 rivers across Great Britain. 2. To examine the biotic and abiotic drivers of signal crayfish burrowing behaviour through a field study of 39 rivers across Great Britain complemented by laboratory flume and mesocosm experiments using crayfish populations that are invasive (in the UK) and from their native geographical range (in the USA). 3. To examine the processes by which crayfish burrowing impacts the geomorphology of riverbank erosion, through field monitoring of riverbanks on two infested UK rivers

complemented by laboratory physical modelling experiments in a purpose built Friedkin flume. 4. To quantify the effect of crayfish burrowing on volumes of riverbank erosion and investigate the relative roles of direct sediment input from crayfish burrows, accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows in recruiting sediment to invaded river channels, using field monitoring of two UK rivers and laboratory physical modelling and flume experiments. 5. To construct models to predict the presence, extent, and geomorphic impacts of burrowing on UK rivers.

1.3 Thesis Structure

To address these objectives, five individual studies, each of which feeds information into one or more of the objectives, were conducted and are reported in chapters 3 to 7. The structure of this thesis is outlined in Figure 1.1, which outlines the organisation of the research and the links between chapters and the research questions. Chapter 2 presents a detailed review of the existing literature on lotic zoogeomorphology, riverbank burrowing, fine sediment in rivers, and the geomorphic capabilities of signal crayfish. The specific research objectives are examined in five primary results chapters (Chapters 3, 4, 5, 6, and 7).

Chapter 3 presents a distributed survey of 39 crayfish invaded rivers throughout Great Britain, quantifying the density of crayfish burrows and the mass of sediment excavated, along with a broad range of biotic and abiotic variables hypothesised to influence crayfish burrowing activity. This fully addresses objective 1, and is then used to develop predictive models regarding the presence and extent of crayfish burrowing to partially address objectives 2 and 5. Chapter 4 specifically considers the relationship between flow velocity and signal crayfish burrowing using live crayfish in flume experiments to address aspects of objective 2. This chapter provides an indication of the relative and combined importance of biological and hydrological energy in recruiting fine sediment from riverbanks, and contributes towards addressing objective 5. Chapter 5 examines the importance of shelter availability and population density on determining signal crayfish burrowing behaviour through a separate series of mesocosm experiments, again with live crayfish. These include crayfish provenance as an experimental factor, and were conducted in Loughborough, UK, and Montana, USA, to compare behavioural differences in invasive and native populations. This helps to assess the potential global extent of the geomorphic implications associated with crayfish burrowing, and

4 contributes to addressing objective 2. Chapter 6 examines the implications of crayfish burrowing on riverbank processes using an erosion-pin monitoring study in two crayfish invaded UK catchments (Gaddesby Brook and the River Bain). This includes the construction of bank erosion sediment budgets to quantify the direct and indirect effects of crayfish burrows relative to erosion in the absence of burrows. These calculations provide in situ contributions to objectives 3 and 4, and temporal contributions to objective 5. Chapter 7 uses a highly idealised experimental setup (Friedkin flume) to investigate how burrow density and burrow arrangements may affect bank erosion, and supports the understanding of objective 3. After these five investigations, Chapter 8 provides a summary and synthesis of results, and considers the wider themes arising from the results chapters. It also presents preliminary results of a continuing study in support of the themes discussed throughout the thesis, and suggests important avenues for future research.

Figure 1.1: Thesis structure. Objectives stated are detailed in section 1.2.

Chapter 2

Zoogeomorphology: Crayfish and Riverbanks

2.1 Chapter Structure

This chapter explores existing literature to evaluate the current understanding and research gaps relating to the key themes explored in this thesis. It aims to contextualise crayfish burrowing within wider frameworks, present the need for an understanding of this research, and scrutinise previous work undertaken regarding crayfish burrowing. This chapter broadly consists of three major sections:

1. Zoogeomorphology and ecosystem engineering, with particular emphasis on how animals interact with riverbank processes; 2. The recruitment and effects of fine sediment on the ecology and geomorphology of lotic ecosystems; and 3. Signal crayfish as zoogeomorphic agents

2.2 Zoogeomorphology and Ecosystem Engineering

The physical impacts of biota on the environment have long been recognised and documented, with the importance of soil turnover from the burrowing of earthworms acknowledged by Charles Darwin (1891). In 1519, a warrant was issued for the arrest of a company of European Moles (Talpa europaea) for burrowing under a field of crops in Stelvio, Italy. When the moles failed to attend court, they were sentenced to exile, but with human escort for protection from foxes, and 14 days grace for infants and pregnant females (Evans 1906). Similarly, in 1713, in Piedade no Maranhao, Brazil, a colony of termites were excommunicated from the Catholic church for burrowing under a monastery wall, and thus destabilising it (Evans 1906; Beirne et al. 2018). Their sentence was officially read aloud by the judge to the offending termite mound, who chaperoned them into the neighbouring field.

Despite this early recognition of animals as geomorphic agents, little academic research into animals driving geomorphic processes, beyond charismatic species such as salmon and beaver, was undertaken until the 1990s, when the term of ‘ecosystem engineering’ was first formally introduced (Jones et al. 1994; Hastings et al. 2007). This term was recognised primarily by ecologists and describes the process by which the behaviour of a keystone species dramatically alters the functioning of the environmental system that it inhabits (Heaming 2012), with direct impacts on at least one other species (Jones et al. 1994; Wright and Jones 2006; Moore 2006). The importance of living organisms for geomorphology was formalised by Viles (1988) who

8 coined the term ‘biogeomorphology’ for the effect of biota on environmental processes. Later, Butler (1995) established the term ‘zoogeomorphology’ to indicate a focus on the role of animals in driving geomorphological processes, an area of research that has seen a notable increase in research publications since the 2000s (Coombes 2016). This increase in awareness of the geomorphic effects of fauna is highly important for understanding the evolution of landscapes and landscape processes (Phillips 2009). It has been estimated that the combined annual geomorphic effect of vertebrates, ants, termites, and earthworms is the displacement of a volume of soil equivalent to the maximum rates of tectonic uplift (Wilkinson et al. 2009), with the geomorphic effect of fauna important over geological timescales (Fremier et al. 2018; Figure 2.1). However, despite growing theoretical understanding and increased empirical evidence, the influence of biota on geomorphological processes is still widely disregarded from mainstream conceptual or numerical models considering sediment and landscape dynamics (Rice et al. 2019).

Figure 2.1: The geomorphology of salmon redds has been modelled to change river profiles over geological timescales (Fremier et al. 2018). Examples include (a) Chinook Salmon (Oncorhynchus tshawytscha) redds on the Columbia River, Washington (road at bottom of photo is ~2m wide), and (b) Pink Salmon (O. gorbuscha) redds on the Fraser River, British Colombia (redds in the foreground are approximately 0.5m wide). (Image source: (a) MSA (2014) and (b) Reid (2012) in Johnson et al. 2019).

While the volume and strength of research in both ecosystem engineering and zoogeomorphology has increased, this has happened largely independently within the fields of ecology and geomorphology, respectively. However, it is necessary to consider and integrate understanding from both in order to fully understand and manage the interactions between biological and geomorphological processes at a system wide scale.

Ecosystem engineers can be classed into two categories; allogenic engineers and autogenic engineers. Allogenic engineers alter the environment in which they exist, which creates or destroys the habitat of other organisms. Examples include beaver (Castor spp.), whose dams alter the hydrological and nutrient regimes of rivers and wetlands (Rosell et al. 2005; Puttock et al. 2017; Figure 2.2a), and may increase ecological richness at the landscape scale (Wright et al. 2002), and earthworms (Oligochaeta), which alter soil nutrient cycling dynamics and thus vegetation community composition (Craven et al. 2017). On the other hand, autogenic engineers modify themselves through growth and distribution, which brings about changes to the ecosystem. For example, kelp (Laminariales) create the kelp forest habitats of coastal waters (Steneck et al. 2002), modifying marine current patterns and benthic community assemblages (Hondolero and Edwards 2017), and in pelagic environments, plankton blooms alter water turbidity and thus the depth at which photosynthesis can occur (Berke 2012).

Zoogeomorphology investigates the same biota-environment interactions, but focusses on the geomorphic effects of the interaction, and exclusively on the relationship between sediment and animals (Butler 1995). Unlike ecological engineers, which are keystone species, any species that affects geomorphic processes can be classified as a geomorphic agent. Examples of zoogeomorphic agents can be observed throughout the animal kingdom: for example, predatory stoneflies including the species Megarcys signata facilitate fine sediment winnowing from gravel-bed river substrates while feeding (Zanetell and Peckarsky 1996), Chacma baboons (Papio ursinus) augment rockfall whilst foraging (Mare et al. 2019), and the feeding behaviours of humphead parrotfish (Bolbometopon muricatum) break down ocean rock into fine sand, which can accumulate to form islands (Perry et al. 2015; Figure 2.2b).

Zoogeomorphology can be witnessed over varying timescales, and fossil records have demonstrated how animals have played a key role in the redistribution of sediment for millennia. Fossilised burrows of dicynodontids (Diictodon sp., Permian; Smith 1987), amphibians (Brachydectes elongates, Permian, Hembree et al. 2004); crayfish (Permian and Triassic; Babcock et al. 1998), lungfish (Triassic; Dubiel et al. 1987), dinosaurs (Oryctodromeus cubicularis, Cretaceous, Varricchio et al. 2007; Woodruff and Varricchio 2011), and beaver (Paleocastor fossor, Neogene, Martin and Bennett 1977) demonstrate the importance of biota across the animal kingdom in mobilising sediment in both terrestrial and aquatic environments. Numerical modelling of salmon spawning has demonstrated how over

10 time periods of 103 to 106 years, the presence or absence of salmon has the capacity to alter riverbed elevations by up to 100m (Fremier et al. 2018).

Figure 2.2: Biological energy driving abiotic processes: (a) ecosystem engineering by beaver (Castor sp.) constructing dams, which pool water, alter flow, sediment, and nutrient regimes, create wetlands, and increase biotic diversity, and (b) zoogeomorphology by green humphead parrotfish (Bolbometopon muricatum) feeding on corals, creating fine sands, which accumulate to form tropical islands. (Image source: (a) Land Trust (2019) and (b) Nekrasov (2019)).

Interactions between animals and sediment also has the ability to affect the functioning of geomorphological systems over much shorter time scales. The reintroduction of grey wolves (Canis lupis) to Yellowstone National Park in 1995 saw large ecological and geomorphic changes within the park in just 15 years (Ripple and Beschta 2012). The threat of predation from wolves brought about behavioural changes in the habitat selection of elk (Cervus elaphus), spending more time in aspen stands and less time in the open or along river margins (Fortin et al. 2005; Mao et al. 2005) with the result that riparian vegetation such as cottonwood (Populus sp.) and willow (Salix sp.) re-established along riverbanks. Previously, browsing suppressed riparian vegetation promoted lateral erosion and braided channels (Beschta and Ripple 2012). Following the changes in the behavioural patterns of elk and the reestablishment of riparian vegetation, river channels straightened, increasing gradients and resulting in vertical incision that changed the hydrological carrying capacity of the channel (Beschta and Ripple 2006). The reestablishment of riparian vegetation has also seen an increase in beaver populations, with the construction of dams further altering hydrological regimes (Wolf et al. 2007). There has been some debate about whether the changes in elk populations have also

11 been caused by bear predation (Barber-Meyer et al. 2010), the disease brucellosis (Cotterill et al. 2018), or climate change (Middleton et al. 2013). However, whilst the dynamics of the trophic cascade leading to elk decline have been debated, changes in elk population and behaviour have had significant geomorphic consequences in less than 15 years.

2.2.1 Zoogeomorphology: Biotic Considerations

The relative importance of biological factors, such as per capita mass, population density, population mass, behavioural traits, and biological invasions influence the ability of an organism to recruit sediment and have been thoroughly assessed and reviewed by multiple authors (Crooks 2002; Moore 2006; Statzner 2012; Rice et al. 2012; Fei et al. 2014; Albertson and Allen 2015; Wilkes et al. 2019; Emery-Butcher et al. 2020). There is very strong evidence to support invasive species as key geomorphic drivers, and meta-analyses suggest population density, as opposed to per capita effects are the strongest drivers of geomorphic capability. However, all are important to consider as each is significant in determining total zoogeomorphic activity.

Examples of zoogeomorphological agency can be witnessed across all scales of organism size. Caddisflies of the order Hydroptilidae grow to a maximum length of 5 mm and construct protective cases out of sediment in their final instar of growth (Holzenthal et al. 2010). The movement of hippopotamus, (Hippopotamus amphibious) weighing up to 1,800 kg, promote channel incision and create new avulsion channels in the Okavango Delta, Botswana (McCarthy et al. 1998). Organism size is an important consideration, as the larger the organism, the greater the potential energy that may be used to promote or inhibit geomorphological processes, and in a meta-analysis of 44 experimental studies, biomass per capita was found to be significantly correlated with the geomorphological effect of an organism (Albertson and Allen 2015).

However, whilst body size is important in determining the potential for sediment transport by a single organism at a local scale, population density is the more important variable to consider at the reach scale. Whilst larger burrowing mammals displace more soil by constructing larger burrows, small mammals collectively move more sediment at a site-scale (Haussmann 2017). This is witnessed in aquatic systems. Smaller, more densely distributed organisms such as insects and crayfish have the same or greater effect on sediment transport than larger, less densely populated species, with population density being a more significant driver of

12 geomorphic processes than organism size in meta-analysis (Albertson and Allen 2015). For example, whilst larvae the caddisfly Agapetus fuscipes construct cases of just 0.02 g, they can occur at densities of over 6,700 larvae m-2 of riverbed, and thus use more sediment than larger, less abundant taxa (Mason et al. 2019; Figure 2.3a). Fiddler crabs (Uca pugnax) only remove up to 3.5 g of sediment per day through burrowing (Vu et al. 2017), but their burrows can exist in densities of up to 800 m-2 (Hughes et al. 2009; Figure 2.3b) and have thus been associated with accelerated marsh creek formation (Escapa et al. 2007).

Figure 2.3: Zoogeomorphology by low per capita mass taxa in high population densities: (a) Caddis larvae (Agapetus fuscipes) cases (0.02 g) found at >6,700 m-2 in Wood Brook, UK, and (b) fiddler crabs (Uca sp.) mobilise up to 3.5 g d-1 of sediment through burrowing, but occur at up to 800 burrows m-2 on the Atlantic coast of North America, so their cumulative effect can be high. (Image source: (b) St. Laurent (2016).

Animal behaviour is an important consideration because it affects the ability of an organism to influence sediment dynamics relative to its body mass. For example, reductions in the shear stresses required to entrain gravelly sediment, caused by signal crayfish (Pacifastacus leniusculus) rearranging substrate grains are greater than those caused by feeding fish of a larger biomass (Statzner 2012). Similarly, higher shear stresses are required to entrain sediment following colonisation by silk spinning caddisfly larvae, compared to the stabilising impacts of mussels of a larger per capita biomass (Statzner 2012).

At a local scale in rivers, animal behaviours are important for determining geomorphic processes in three ways. First, biota play a significant role in entraining sediment already in the river channel, such as crayfish foraging and fighting (Rice et al. 2014; Harvey et al. 2014), fish

13 feeding (Breukelaar et al. 1994; Matsuzaki et al. 2009; Pledger et al. 2014; Pledger et al. 2017; Rice et al. 2019); macroinvertebrate feeding (Zanetell and Packarsky 1996; Statzner et al. 1996) and fish spawning (Hassan et al. 2008). Second, biota can also inhibit the movement of sediment, such as caddisfly larvae that bind sediments with silk (Johnson et al. 2009; Albertson et al. 2014a; Albertson et al. 2014b), sediment compaction and protection by freshwater mussels (Box and Mossa 1999) and through the binding of gravels with byssal threads (Mills 2019). Third, biota can recruit further sediment into river channels through behaviours such as bank burrowing (Sofia et al. 2016; Faller et al. 2016). At the landscape scale, animal activities have the potential to create new landforms. For example, lodge building by beavers (Castor spp.) can create new wetlands, with dams up to 850 m long having been discovered (Thie 2016).

Whether a species is invasive is also an important consideration regarding its geomorphic potential (Crooks 2002; Fei et al. 2014; Emery-Butcher et al. 2020). Whilst native species exhibit geomorphic behaviours, the local geomorphology is typically adjusted to that activity. In their new environment, invasive species may not be constrained by the characteristics of their native habitats and the new environment may not be resilient to their activities (Harvey et al. 2011). This allows invasive species to establish population densities not observed for native taxa; for example, invasive filter feeding bivalves in the upper tidal reaches of the River Thames exhibit average population densities more than 30 times greater than native taxa (Pecorelli 2018; Figure 2.4).

Invasive species can therefore cross geomorphic thresholds, and thus change the geomorphology and ecology of a system from one condition to another (Kimmerer et al. 1994; Butler et al. 2018). For example, in their native ranges, redd excavation by salmonid fish mobilises river gravels, and is an important driver of grain mobility (Gottesfeld et al. 2007), with continued redd excavation possibly causing change in riverbed elevations by up to 100 m over periods of 103 to 106 years in the Pacific Northwest of America (Fremier et al. 2016). However, the redd construction of invasive quinnat salmon (Oncorhynchus tshawytscha) populations in New Zealand altered riffle geomorphology, resulting in significant changes to macrophyte, bryophyte, and invertebrate stream communities (Field-Dodgson 1987).

Figure 2.4: Bivalve surveys on the upper tidal River Thames, UK. (a) At spring tide, in conjunction with weir control by the Port of London Authority, the bed of the River Thames is exposed allowing annual bivalve counts. (b) Invasive bivalve populations were >30 times denser than native taxa, with Asiatic Clams (Corbicula fluminea) reaching densities of 200 m-2.

Of the ‘100 of the World’s Worst Invasive Alien Species’, a list compiled by the Global Invasive Species Database (2020), 81% were determined to have potential geomorphic impacts (Fei et al. 2014). This is compounded by the high population densities that invasive species can achieve (Figure 2.4); invasive species are either able to dominate comparatively ‘easy’ conditions compared to their native range, or are purposefully introduced, such as pigs and horses, with populations being held artificially high (Crooks 2002; Bulter 2006). Habitats are being anthropomorphically stressed through increased fragmentation and climate change, and thus have a lower resilience to community changes. Further, increased globalisation and human activity increases the connectivity of previously isolated environments, thus increasing the potential for the global movement of invasive species. Therefore, the frequency and extent of biological invasions are projected to increase (Simberloff et al. 2013; Almeida et al. 2014), and thus investigation into the zoogeomorphic ability of invasive species is an important avenue of research.

2.2.2 Zoogeomorphology: Abiotic Considerations

Abiotic factors determine whether or not a zoogeomorphic agent is effective (Moore 2006). Zoogeomorphic activity is of particular note in riverine environments due to the abundance of biota and dynamic sediment regimes. Rivers are the key driver of the global redistribution of

15 sediment, with 13.5 x 109 tonnes of suspended sediment delivered to oceans annually (Milliman and Meade 1983), and whilst freshwaters cover only 2.3% of the world’s surface, they hold at least 9.5% of all described species (Reid et al. 2019). The interactions between biota and sediment are likely to be greatest in rivers compared to any other global system, due to the very high density of sediment processes and biotic energy. If only a small proportion of river sediments are entrained or inhibited as a result of biotic activity, this could amass to a globally significant quantity.

Lowland rivers are particularly susceptible to the geomorphic behaviours of animals, particularly in low discharge systems (Albertson and Allen 2015). This is because of their low geophysical energy, and cohesive materials. There is a high differential between the shear stress required to entrain (high) and transport (low) cohesive material (Hjulstrom 1935), and so animals that facilitate the entrainment of fine sediments can have a significant effect. This is of particular concern in the UK, where sediment transport in lowland rivers tends to be sediment capacity-limited as opposed to supply-limited (Naden et al. 2016), and so are particularly sensitive to changes in transport capacity.

Freshwater zoogeomorphology has been under-studied (Rice et al. 2012), with most interest limited to charismatic species such as salmon and beaver. Excluding beaver studies, freshwater research citing Jones et al. (1994) accounts for just 26% and 22% compared to marine and terrestrial environments respectively (Emery-Butcher et al. 2020). The importance of biota forcing geomorphic processes in lowland rivers is under researched; the majority of attention has focussed on flagship animals (e.g. beaver and salmon), with functionally similar ‘Cinderella species’, which undertake comparable work but are less charismatic (e.g. coypu and chub), likely being of much greater importance but are persistently overlooked.

2.2.3 Biogeomorphology and Zoogeomorphology of Riverbanks

The interaction between biota and riverbanks is of specific concern due to the important role that riverbanks play in sediment recruitment to river systems, with between 37% and 94% of fine sediment in lowland rivers estimated to be derived from riverbank erosion (Duijsings 1987; Walling et al. 1999; Kronvang et al. 2013). Excepting widespread acknowledgment of the role of riparian vegetation in stabilising river banks, sediment recruitment from riverbanks has largely been considered an abiotic process. However, some recent research has suggested that animal activity, including burrowing, can also be important.

Whilst the role of animals has often been disregarded (Rice et al. 2019), the influence of vegetation on riverbank dynamics has been extensively studied and its importance widely accepted since the 1980s. The presence of aquatic vegetation in the channel next to the bank absorbs energy and the potential shear between the stream and the bank (Miller et al. 2011). In-stream vegetation also protects the bank from wave action, reducing the size of waves and wave-induced erosion (Coops et al. 1996; Bendoni et al. 2019). Vegetation also has stabilising effects by binding sediments together with long, dense roots (Abernethy and Rutherfurd 2000; Docker and Hubble 2008; Krzeminska et al. 2019), and slows down lateral erosion by providing a continual barrier to hydrological flows (Micheli and Kirchner 2002a). The quality of this protection varies with plant growth, form and species. In the Kern River in the Sierra Nevada, the root-area (the ratio of root area to soil area for a planar exposure) of hydric vegetation covered roughly 50% of soils, whereas the roots of xeric vegetation covered less than 5% of soils (Micheli and Kirchner 2002b), resulting in bank retreat six times faster in areas of xeric compared to hydric vegetation (Micheli and Kirchner 2002a). Vegetation can also change the form of the channel, with vegetated channels typically being deeper, narrower and slower to migrate than unvegetated streams (Hickin 1984; Hey and Thorne 1986; Peixoto et al. 2009; Allen et al. 2018). The large-scale implication for this is that vegetation has the capacity to change entire system morphologies, such as stabilising braided channels to single channel systems (Tal and Paola 2007; Tal and Paola 2010; van Dijk et al. 2013).

Vegetation is not exclusively associated with increased bank stability. The presence of riparian trees can add to the mass loading of the bank and can increase the roughness of the channel boundary to flow (Micheli and Kirchner 2002a). Seasonal plant species such as the Himalayan balsam (Impatiens glandulifera) are highly invasive, but are intolerant to cold conditions and undergo rapid dieback in autumn, leaving banks bare and unconsolidated by roots, which is hypothesised to promote soil erosion and bank retreat along invaded channels (Greenwood and Kuhn 2014; Greenwood et al. 2018).

By contrast, the role of animal activity and riverbank processes has been under researched (Harvey et al. 2019). Riverbanks are an important refuge for many animals, with species native to the UK such as common kingfishers (Alcedo atthis) (RSPB 2016) and water voles (Arvicola amphibious; EWT 2016) relying on riverbank burrows. Bank burrowing is also a trait common of many invasive species, such as the Chinese mitten crab (Eriocheir sinensis; Clark and Robbins 2008) and the signal crayfish (Pacifastacus leniusculus; Guan 1994; Harvey et al.

2011; Faller et al. 2016). Burrowing may recruit large quantities of fine sediment directly into river systems (Faller et al. 2016) and it has been suggested that burrows may accelerate bank erosion processes (Guan 1994; Faller et al. 2016) and increase the probability of riverbank collapse (Viero et al. 2013; Orlandini et al. 2016). This was observed on the Secchia River (Orlandini et al. 2016; Figure 2.5a), and the Foenna Stream (Camici et al. 2014; Figure 2.5b; 2.6c), both in Italy, when the collapse of a crested porcupine (Hystrix cristata) burrow led to the failure of flood-defence levees. The failure on the Secchia caused over $500m of damage (Orlandini et al. 2016). In the UK, signal crayfish (Pacifastacus leniusculus) burrowing is thought to have caused the collapse of a wall at Magdalen College, Oxford University on the River Cherwell (Telegraph 2016; Oxford Mail 2017), and damage to the foundations of Palladian Bridge at Prior Park, Bath, costing over £2m in repairs (BBC 2018). European badgers (Meles meles) were blamed for flooding after burrowing through flood defences on the River Yeo in Somerset (Independent 2013) and for building setts into the banks of the River Steeping Lincolnshire (Telegraph 2019).

The impacts on salt-marsh creek development by vertical burrowing of New Zealand isopods and crabs have been well studied (Talley et al. 2001; Perillo and Iribarne 2003; Perillo et al. 2005; Escapa et al. 2007; Hughes et al. 2009; Wilson et al. 2012; Smith and Green 2015). Burrowing crabs drive creek formation through the exclusion of stabilising vegetation, and directly by destabilising sediment (Escapa et al. 2007). However, these taxa are within their native range and have been promoting sediment redistribution for millennia. The associated creek development is hypothesised to mitigate the effects of sea level rise on saltwater incursion (Vu et al. 2017).

Signal crayfish burrows have been recorded at high densities of up to 21 burrows m-1 of riverbank (Guan and Wiles 1997; Table 2.1) and can contribute substantial masses of sediment to river systems. Surveys throughout the River Thames catchment estimated at least 3 t km-1, and up to 15 t km-1, of fine sediment were directly recruited to rivers by signal crayfish burrowing (Faller et al. 2016), and measurements from six lowland rivers in central England estimated burrow construction to directly recruit up to 0.5 t km-1 of fine sediment per year (Rice et al. 2016).

Figure 2.5: Burrow induced bank failures from (a) crested porcupines (Hystrix cristata) on the Secchia River, Italy, and crested porcupines on Foenna Stream, Italy (b and c). (Image source: Camici et al. 2014).

Mean Burrow Density Maximum Burrow Study Location (burrows m-1) Density (burrows m-1)

Faller et al. 2016 R. Thames tributaries 1.5 6.0 Guan 1994 R. Great Ouse 3.2 5.6 Guan and Wiles 1997 R. Great Ouse - 21.0 Stanton 2004 Gaddesby Brook 2.4 14.0 R. Greet 2.1 6.5

Table 2.1: Recorded signal crayfish burrow densities in previous UK studies.

As well as recruiting fine sediment into river systems, crayfish burrows also have geotechnical implications for riverbanks. Burrows constructed by invasive red swamp crayfish

(Procambarus clarkii) burrows in rice paddies in the Iberian Peninsula have caused the collapse of banks and road support pillars (Arce and Dieguez-Uribeondo 2015), with 73% of excavated burrows collapsing within seven days of construction (Barbaresi et al. 2004). Signal crayfish burrows in the UK have been reported to accelerate riverbank retreat by up to 1 m per year (West 2010), although there are currently no empirical data supporting this. Harvey et al. (2019) provide a comprehensive review and hypothesise that riverbank burrowing by invasive species may alter bank erosion processes, citing three likely effects:

(i) Geotechnical and hydrological effects. Burrows curving towards the top of the bank may reduce the length of failure planes, making collapse more likely. Burrows modify the spatial distribution of pore water pressure within banks (Onda and Itakira 1997; Xin et al. 2009), and the phreatic surface within the bank, increasing the hydraulic gradient across the bank. This may cause erosion within burrows and the collapse of overhanging material. Elevated pore pressure may increase the likelihood of slip failures on receding flood limbs. Alternately, burrows may cause banks to drain better at low flow, reducing the susceptibility of collapse at low flow.

(ii) Hydraulic effects. Individual burrow entrances modify near bank flow structures (Ozalp et al. 2010; Jackson et al. 2015), and enhanced turbulence may amplify direct entrainment of particles from the riverbank face. However, the collective effect of multiple burrows may increase surface roughness and therefore reduce overall rates of fluvial erosion.

(iii) Geochemical and biological effects. Biofilm and fungal networks in soils that stabilise banks are likely to be altered by changes in moisture in the bank driven by the presence of burrows. Burrows may also exclude stabilising agents such as vegetation (Hughes and Paramour 2004).

Physical (Viero et al. 2013; Saghaee et al. 2017) and numerical (Camici et al. 2014; Orlandini et al. 2016; Taccari and vaan der Meij 2016a; 2016b; Borgatti et al. 2017) modelling support these hypotheses and have suggested that the presence of animal burrowing can greatly increase the probability of riverbank collapse, particularly from burrows constructed on the waterside of levees (Saghaee et al. 2017). The deleterious interaction between animals and cohesive riverbanks is of particular concern not only due to the potential for problematic riverbank erosion and enhanced flood risk, but also because of the influence of recruited sediment on

20 downstream processes. Whilst interactions between animals and coarse sediments such as gravel can have localised effects, the suspension and transport of fine sediment can have severe implications for downstream processes and communities.

2.3 Fine Sediment, River Ecology, and River Geomorphology

Fine sediment is classed as a diffuse pollutant under the Water Framework Directive (European Environment Agency 2012). Understanding the sources and effects of fine sediment is therefore critical and the potential impacts that animal burrowing may have makes some consideration of fine sediment relevant here.

Fine sediment typically refers to any sediment finer than 2 mm in size, including sand (63 µm < x < 2000 µm), silt (4 µm < x < 63 µm), and clay (< 4 µm) (Blott and Pye 2001). Fine sediment is naturally found in rivers, and derives from processes both inside the channel, such as riverbank erosion (Julian and Torres 2006; Kronvang et al. 2013; Henshaw et al. 2013), processes from outside the channel, both natural and anthropogenic, such as inputs from agricultural land (Walling and Amos 1999; Collins and Walling 2007; Nuara et al. 2016), and from events outside of the catchment, such as windblown volcanic eruptions (Major 2004). Fine sediment can derive from a wide range of sources, and can comprise inorganic and organic material (Collins and Walling 2007; Nuara et al. 2016) including dead vegetation, woody material, litter fall shredded by macroinvertebrates (Griffith and Perry 1993; Raposerio et al. 2018) and macroinvertebrate faecal matter (Ladle and Griffiths 1980). Whilst fine sediment plays an important role in the lifecycle of many species (Wood and Armitage 1997; Zuanon et al. 2006; Kemp et al. 2011), the presence of excess levels of fine sediment has the potential to cause significant impacts to channel morphology (Xu 2003; Hoffman and Gabet 2007) and cause detrimental impacts to the diversity and abundance of biota (Wood and Armitage 1997; Bilotta and Brazier 2008; Kemp et al. 2011; Jones et al. 2012a; Jones et al. 2012b; McKenzie et al. 2020).

2.3.1 Channel Sources of Fine Sediment

Channel sources of sediment supply have typically been attributed to abiotic processes, and are governed largely by channel discharge (Wolman 1959; Knighton 1973; Hooke 1979; Arulanandan et al. 1980; Smith et al. 2003; Magilligan et al. 2015), although subaerial processes also play a significant role (Simon et al. 2000; Yumoto et al. 2006; Couper and

Maddock 2001; Couper 2003). Channel sources of sediment include the banks, the channel bed, and vegetation growing within the channel.

Riverbeds are often heavily armoured and immobile outside of flood conditions (Church 2000), which protects the bed from scouring under base flow conditions. Sediment from the interstices can be remobilised following deposition, but this is typically sediment deposited from another source within the river channel. However, sand requires the lowest levels of shear stress for entrainment, and so bed erosion in sand bed rivers can be a major contributory factor to total sediment loading (Thorne et al. 1985; Schmidt and Graf 1990; Molinas and Wu 2001; Stephens et al. 2017).

The greatest source of in channel fine sediment is typically riverbanks, which contribute continually through diffuse erosion, and sporadically through mass failure events. The mass of sediment supplied to the channel by non-cohesive banks can be calculated as a function of particle size and shear stress (Buffington and Montgomery 1997; Kimiaghalam et al. 2016), but in cohesive banks the electrochemical cohesiveness of particles make the supply of sediment to the channel difficult to estimate due to the role of subaerial processes that aid weathering of the bank (Yumoto et al. 2006; Couper and Maddock 2001; Couper 2003; Julian and Torres 2006; Kimiaghalam et al. 2016). However, erosion pin recordings at the River Odense, Denmark, have estimated riverbanks to supply up to 94% of total sediment recruited.

2.3.2 Non-Channel Sources of Fine Sediment

The majority of fine sediment in rivers is sourced from outside of the channel, most of which is anthropogenically derived; it has been estimated that globally 80-90% of sediment in rivers is of an anthropogenic origins, principally from agricultural runoff and mining (Farnsworth and Milliman 2003; Wilkinson 2005). Sediment records demonstrate that an increase in sediment delivery to river systems coincided with human settlement; for example, since human settlement in the 1830s in the Mississippi catchment, sediment delivery to Lake Pepin has increased tenfold, with 17% of water volume being replaced by sediment accrual (Engstrom et al. 2009). This figure is applicable globally and is observed at local scales. For example, in the UK, approximately 84% of fine sediments in the Dorset Frome and 82% in the Dorset Piddle are sourced from anthropomorphic activities outside of the channel (Collins and Walling 2007), and in tropical regions, such as Indonesia, the input of fine sediment from some urban areas in

Borneo exceeds that derived from the tectonic activity associated with geomorphic processes (Douglas 1996; Caine 2004).

The largest contributor of non-channel sources of fine sediment globally is farming and agriculture (Owens et al. 2000; Wilkinson 2005; Collins and Walling 2007), which is estimated to be the source of 76% of suspended sediment in rivers in England and Wales (Collins et al. 2009). During the winter, agricultural soils are typically bare, and as a result are exposed to high levels of weathering. Eroded material from agricultural sources is also highly mobile due to the production of structures such as gullies and rills produced by flow concentrations on agricultural fields that transport sediment to rivers directly, as opposed to depositing the sediments during the percolation process to the water table. As a result, the rate of erosion from ploughed fields is estimated to be one to two orders of magnitude greater than those covered by natural vegetation (Montgomery 2007). Whilst predicting the rate of soil erosion can yield significantly over-estimated and under-estimated values (Risse et al. 1993) the lowest estimates for agricultural soil erosion in North America (3.9 mm a-1) are still an order of magnitude greater than the maximum estimates for historical soil erosion (0.24 mm a-1; Wilkinson 2005; Montgomery 2007).

The delivery of sediment to the channel from sources outside of the channel, and sources within the channel, are both seasonal. Suspended sediment concentrations are typically highest during the winter months, due to increased rainfall, less interception and evapotranspiration, and bare agricultural soils (Kemp et al. 2011). Within the channel, suspended sediment concentrations are typically a function of flow (Knighton 1973; Sherriff et al. 2016); flow is typically higher and flood events are more common during the winter months in high latitude areas, and during the wet season in the tropics. Super-seasonal events such as the El Nio Southern Oscillation can also cause elevated sediment fluxes; cores from South American sand bank deposits have shown sand to be transported and deposited in its greatest volumes during La Nina events (Aalto et al. 2003).

2.3.3 Deposition and Storage of Fine Sediment

Whilst the majority of suspended fine sediment is anthropogenically derived, human impacts such as dams, weirs and reservoirs can also remove large quantities of sediment; it is estimated that dams and reservoirs trap 25-30% of all fine sediment globally (Vorosmarty et al. 2003). As a result, rivers downstream of dams can experience issues such as channel incision and

23 canyoning, channel narrowing, and changes in the slope and stability of bed sediments (Lisle and Church 2002; Choi et al. 2005; Kondolf et al. 2019) which can lead to the undermining of structures such as bridges (Kondolf 1997), and thus promote bridge collapse (Batalla 2003). Dams also alter flow regimes, dampening the effects of floods, during which the majority sediment transport occurs (Tabarestani and Zarrati 2015). This is particularly important for the storage of fine sediment, which can accumulate in the bed without being eroded during peak flows (Kondolf 1997).

2.3.4 Geomorphological Impacts of Fine Sediment

Excess levels of fine sediment can affect channel morphology and the surrounding floodplain (Marston et al. 1995; Sidorchuk and Golosov 2003; Xu 2003; Hoffman and Gabet 2007), and too little fine sediment can promote canyoning, channel narrowing and increased bed slopes (Lisle and Church 2002; Choi et al. 2005; Kondolf et al. 2019). Lowland rivers are particularly at risk to morphological changes; low energy systems have a reduced capacity to recover to natural form, with recovery occurring over a longer time scale than in high energy systems (Brookes 1995).

2.3.5 Chemical Impacts of Fine Sediment

Fine sediment can also have severe impacts on water quality. Fine sediments have a high surface area to mass ratio, and so are able to sorb high volumes of nutrients and pollutants. These include phosphates and nitrates from agricultural land (Haggard et al. 1999; Korppoo et al. 2017; Frazar et al. 2019) contributing to eutrophication problems, heavy metals, such as lead, cadmium and zinc from sites of mining (Besser et al. 2007; Ma et al. 2019), which can be poisonous to fish (Hunter et al. 1987, Vinodhini and Narayanan 2008; Wang et al. 2017); organic compounds, such as herbicides and pesticides from agricultural land, which can be detrimental to macroinvertebrate communities (Magbanua et al. 2013), and bacteria and viruses (Bai and Lung 2005; Hassard et al. 2016).

A build-up of fine sediments can also result in severe oxygen depletion in the water. The biological component of fine sediment is aerobically decomposed, which in high quantities can result in a significantly reduced dissolved oxygen content (Dolloff 1987; Ryan 1991), especially under low flow conditions or in high temperatures. In low flow conditions, hypoxia can develop as a result of aerobic breakdown (Schlosser and Kallenmeyn 2000), which can

24 lead to fish losses (Fox and Keast 1990). This is noticeable behind beaver logjams, with the deposited sediment creating areas of hypoxia and extreme acidity (Schlosser and Kallenmeyn 2000).

2.3.6 Ecological Impacts of Fine Sediment

Suspended fine sediment also has a significant detrimental impact to biota of all sizes, from zooplankton (Bilotta and Brazier 2008), to larger invertebrates, such as crayfish (Bauer 1998; Rosewarne et al. 2014), and vertebrates, such as fish (Wilson et al. 1994; Suttle et al. 2004; Kemp et al. 2011), as well as macrophytes (Jones et al. 2012a). Fine sediment can affect biota physically, including abrasion of plants leaves (Lewis 1973; Jones et al. 2012a) and macroinvertebrate gills (McKenzie et al. 2019), chemically, through the absorption and delivery of pollutants such as phosphorus (Barko and Smart 1980), and biologically, through the transportation of bacteria (Droppo 2001). Riverbed substrates can also become smothered in fine sediments, which may reduce available spawning habitat for salmonids (Soulsby et al. 2001; Sear et al. 2016), reduce survival of salmonid eggs (Jensen et al. 2009), limit vertical migration capacity for invertebrates within the substrate (Gayraud and Philippe 2003; Vadher et al. 2015; Dyer et al. 2015), increase invertebrate drift (Doeg and Milledge 1991; Gomi et al. 2010; O’Callaghan et al. 2015), and reduce available habitat for organisms reliant on gravel interstitial spaces (Richards and Bacon 1994; Gayraud and Philippe 2003).

2.3.6.1 Ecological Impacts of Fine Sediment: Macrophytes

The impact of fine sediment on macrophytes is a particularly notable sediment-biota interaction, as macrophytes form the basis of the aquatic food chain, and also have keystone impacts on fine sediment trapping, storage, and production (Sand-Jensen 1998; Jones et al. 2012a). Macrophytes act as a fine sediment store by cycling stored sediments in the bed, and by retaining fine sediments held in root complexes (Barko et al. 1991; Clarke 2002), but are also producers of fine sediments through death and decay, as well as having influences on channel flow and morphology, and thus sediment erosion and depositional patterns (Gurnell et al. 2006; Bertoldi et al. 2009).

Increased turbidity as a result of fine sediment suspension affects macrophytes both directly and indirectly. Fine sediments can harm macrophytes through leaf abrasion (Lewis 1973) and increases in turbidity reduce the depth to which light can penetrate the water column (Bhargava

25 and Mariam 1991), which in turn affects water temperature (Paaijmans et al. 2008) and photosynthesis (Jewson and Taylor 1978; Anthony and Fabricius 2000). Macrophytes are a key producer in the aquatic food chain, and so any impact to macrophytes is likely to be reflected in the invertebrate and fish communities; in ponds macroinvertebrate richness is significantly higher in vegetated than unvegetated substrates (Engel 1988; Hargeby et al. 1994; Zelnik et al. 2018).

2.3.6.2 Ecological Impacts of Fine Sediment: Macroinvertebrates

Fine sediment pollution also has deleterious implications for macroinvertebrates, and can directly impact their survival, distribution, and physiology (Wood and Armitage 1997; Kaller and Hartman 2004; Jones et al. 2012b; Blettler et al. 2015; Everall et al. 2018; McKenzie et al. 2020). Whilst some fine sediment is important for many macroinvertebrate taxa (Whiles and Dodds 2001; Porinchu and MacDonald 2003; Cover et al. 2006), excess fine sediment is deleterious. Whilst some taxa are able to tolerate poor conditions for prolonged periods of time, such as Chironomidae and Oligochaeta (Scullion et al. 1982; McCulloch 1986; Brown and Brussock 1991; Mathers et al. 2017; Doretto et al. 2018), the majority of macroinvertebrate taxa are sensitive to reduced habitat quality conditions and, even in well oxygenated riffles, are still sensitive to fine sediment deposition (Mathers and Wood 2016). In pools, where substrate conditions are less favourable and exclude sensitive taxa, the presence of fine sediment is still influential in determining invertebrate communities. Pools with a higher gravel content are richer in benthic species than those with a higher silt content (Brown and Brussock 1991). The sensitive taxa Ephemeroptera, Plecoptera and Trichoptera (EPT taxa) are used in biomonitoring because of their sensitivity to poor conditions; a threshold of 0.8-0.9% of sediment finer than 0.25 mm has been suggested for the survivorship of EPT taxa in the Appalachian Mountains (Kaller and Hartman 2004). Fine sediment pollution also disproportionately effects sensitive taxa through promoting increased drift, with invertebrate drift increasing linearly with suspended sediment concentrations (Doeg and Milledge 1991) and EPT taxa are the most susceptible to fine sediment induced drift (Davies and Cook 1993).

Suspended fine sediments bind with the gills of larger biota, reducing the capacity for respiration and thus susceptibility to parasites (Lemly 1982; Rosewarne et al. 2014). An increase in suspended sediment levels has been shown to have negative effects on both the native white clawed crayfish (Austropotamobius pallipes) and signal crayfish (Pacifastacus leniusculus) as a result of gill fouling and thus a reduction in aerobic capacity, with white

26 clawed crayfish being the more severely affected (Rosewarne et al. 2014). Elevated suspended sediment levels also increase the capability of parasites to do harm, with the crayfish worm (Branchiobdella astaci) interacting with the sediment to cause harm to the gills of the crayfish. The presence of attached pollutants can increase these impacts; for example, the gills of salmonids such as rainbow trout (Onchorhynchus mykiss) are damaged by the presence of aluminium (Wilson et al. 1994), and invertebrates by copper and cadmium (Hunter et al. 1987).

2.3.6.3 Ecological Impacts of Fine Sediment: Fish

Fine sediment also plays a key role in fish ecology. Multiple species rely on fine sediment for concealment (Friel 2008), for foraging for food (Zuanon et al. 2006), and for breeding, such as egg burial in silt by annual killifish of the order Cyprinodontiformes (Furness 2015); some species, such as the creek chub (Semotilus astromaculatus) actively seek out areas of high sediment concentrations (Gradall and Sweson 1982). However, excess levels of fine sediment can have severe detrimental impacts on fish, directly by physically damaging organs, indirectly via water quality (Kemp et al. 2011), or by acting as a stressor, increasing the susceptibility of the fish to bacterial and fungal infections (Redding et al. 1987).

Macroinvertebrates constitute the main food supply of fish (Diehl 1992; Williams et al. 2003; Harris et al. 2018; Smith 2019), and so populations of fish species specialising in the predation of benthic riffle-dwelling invertebrates often decline in high sedimentation conditions as a result of a limited food resources (Kemp et al. 2011). In stream cage experiments in the South Fork Eel River, California, the growth rates of rainbow trout (Onchorhynchus mykiss) decreased with fine sedimentation due to the resultant transformation of the macroinvertebrate community (Suttle et al. 2004; Figure 2.6a). Increased sedimentation resulted in a shift from vulnerable riffle-dwelling prey to Chironomids and Oligochaets which burrow and are thus harder to forage (Suttle et al. 2004; Figure 2.6b). An increase in fine sediment also filled the interstitial spaces between cobbles, resulting in fish being forced to swim more due to a reduced number of resting sites available to the fish (Figure 2.6c). In Missouri, the spatial distribution of fish populations was best described by the functional feeding traits of the fish in response to sedimentation, suggesting that food availability is perhaps the strongest driver of change to a fish community in an area of high sedimentation (Rabeni and Smale 1995).

Figure 2.6: Impacts of fine sediment on juvenile salmonid fish: (a) reduced fish growth rate, (b) change in prey availability; and (c) proportion of time spent swimming with increasing fine sediment ingress. Reproduced from Suttle et al. (2004).

Feeding can also be impaired at high suspended sediment concentrations as a result of increased turbidity and reduced vision; the feeding efficiency of bluegills (Lepomis macrochirus) halved in turbidity levels of 190 NTU (nephelometric turbidity units; Gardner 1981), banded kokopu (Galaxias fasciatus) had significantly reduced feeding ability at 20 NTU (Rowe and Dean 1998); a 50% in reduction of feeding efficiency at 9.2 NTU was observed in summer conditions for the rosyside dace (Clinostomus funduloides; Zamor and Grossman 2007); the distance at which smallmouth bass (Micropterus dolomieu) reacted to prey reduced from 65 cm in clear water to 10 cm at 40 NTU (Sweka and Hartman 2003), roach (Rutilus rutilus) feeding efficiency reduced from 60% to 0% from 8.5 NTU to 45.1 NTU (Smith 2019), and the distance at which brook trout (Salvelinus fontinalis) reacted to prey reduced from 80 cm in clear water to 12 cm in turbidity above 30 NTU (Sweka and Hartman 2001). Mesocosm experiments on the gulf killifish (Fundulus grandis) showed that feeding efficiency was lower in water with high sediment concentrations than in shaded water with the same light availability, suggesting that light intensity alone is not the key driver, but a scattering of light by the suspended particles is likely the limiting factor (Benfield and Minello 1996).

Suspended fine sediment physically impacts the gills of fish, which has consequences for respiration and for osmoregulation. Suspended sediments bind to the gill epithelium, reducing the gill surface area and thus the ability of the gills to function. Gills are used for gas exchanges, the absorption of salts from the water, and the expulsion of ammonia via chloride cells (Evans et al. 2005). In extreme conditions, continued abrasion on the gills from suspended sediments can induce gill lamellae thickening (Sutherland and Meyer 2007). Two species of Appalachian minnows, Cyrinella galactura and Erimonax monachus, were exposed to increasing

28 concentrations of suspended sediment in mesocosm experiments, in which the gill lamellae thickened at higher concentrations of suspended sediment. Gill damage was minimal in sediment concentrations lower than 100 mg l-1, but at the highest treatments of 500 mg l-1, gill lamellae damage was significant. Growth rates were inversely correlated against gill lamellae thickness, suggesting that respiratory and osmoregulatory limitations are severe for growth and survivorship of young fish. Increasing suspended sediment levels in the Fraser River were also found to be high enough to cause gill trauma to sockeye salmon (Oncorhynchus nerka; Servizi and Martens 1987). The effects on fish gills vary with flow and particle size, with small, angular particles doing more damage to gills than rounded particles (Lake and Hinch 1999). These effects are further exaggerated due to the high oxygen demand of the organic component of fine sediments (Ryan 1991).

Fish eggs and embryos can also be affected by high fine sediment deposition (Figure 2.7). The infiltration of fine sediment into the interstices of the substrate can prevent flow from removing waste products, deplete oxygen levels, and introduce bacteria and fungus to the eggs. High levels of suspended sediment or sedimentation reduces egg and fry survival; the mortality of dace (Leuciscus leuciscus) eggs is dependent on sedimentation and associated reductions in gaseous exchange (Cowx 1990); the weight gain of brook trout (Salvelinus fontinalis) slowed with increasing fine sediment concentrations in mesocosms (Argent and Flebbe 1999); the threshold for survivorship of chinook salmon (Oncorhynchus tshawytscha), coho salmon (Oncorhynchus kisutch), chum salmon (Oncorhynchus keta) and rainbow trout (Oncorhynchus mykiss) eggs was estimated at 10% for sediment finer than 0.85 mm in a meta-analysis of laboratory and field studies (Jensen et al. 2009), and the survivorship of Atlantic salmon eggs decreased with increasing fine sediment levels in a meta-analysis of stream based studies (Kemp et al. 2011). These effects of fine sediment on salmonid eggs are amplified due to peak suspended sediment loads occurring at the same time of year as fish breeding (Kemp et al. 2011).

Figure 2.7: Salmonid egg survival rates reduce with increasing fine sediment coverage. Reproduced from Kemp et al. (2011).

2.3.7 Conservation and Management of Rivers

The ability of aquatic communities to respond to disturbances depends greatly on the nature and severity of the disturbance. Macroinvertebrates respond quickly to disturbances such as channel modification (Newson and Newson 2000), and to short term pulse disturbances, such as sediment released from reservoir cleaning (Gray and Ward 1982), and weir desilting (Doeg and Koehn 1994). However, press disturbances, such as an increase in sediment delivery from increased biotic sediment recruitment, may result in morphological changes to the channel, and thus long term changes to the macroinvertebrate community and fish community structure, with human intervention necessary to restore the community back to its natural state (Wood and Armitage 1997). In order to be able to evaluate the quality and quantity of disruption caused to a stream, it is vitally important to quantify fine sediment recruitment to river systems. However, the biotic contribution to fine sediment loading is largely unknown, and understanding this is therefore key to be able to fully quantify fine sediment, and thus direct appropriate management for rivers.

2.4 Signal Crayfish as Zoogeomorphic Agents

Crayfish are some of the most successful invasive species worldwide (Gherardi 2013; Kouba et al. 2014; Figure 2.8a), and are a highly influential aquatic geomorphic agent with regards to their potential to entrain excess sediment under normal flow conditions (Statzner 2012; Figure 2.8b). In the UK, signal crayfish are considered as a particular geomorphic threat to river

30 systems (Harvey et al. 2011). Signal crayfish are native to the Pacific coast of America, with their range stretching from California to southern Canada (Johnsen and Taugbol 2010; Larson and Olden 2011) but were introduced to Europe in the 1960s for aquaculture and became established as a result of escaping overland from farms, aided through deliberate introduction. Signal crayfish are now present in at least 29 territories (Kouba et al. 2014; Petrusek et al. 2017) including Great Britain (Almeida et al. 2013), where they are currently present in at least 60% of all English sub-catchments and are expanding their range at a rate of 1.6% per year (Chadwick 2019; Figure 2.9).

Growing to 16 cm long, signal crayfish are the largest freshwater macroinvertebrate in the UK and exist in densities of up to 20 adults m- 2 (Bubb et al. 2004), and up to 120 m-2 when the full population (including juveniles) is considered (Chadwick 2019). They are typically associated with small lowland streams (Guan 1994; Stanton 2004; Harvey et al. 2011; Harvey et al. 2014; Rice et al. 2014; Cooper et al. 2016; Faller et al. 2016; Rice et al. 2016), many of which have banks made of cohesive sediment and are low energy systems, and so crayfish activity creates a large differential between invaded and non-invaded conditions (Rice et al. 2014; Cooper et al. 2016; Rice et al. 2016). This is due to crayfish bioturbating sediment passively through foraging and fighting, but also actively through their digging behaviour (Harvey et al. 2014; Rice et al. 2014; Rice et al. 2016). Signal crayfish are not known to burrow in their native habitat (Shimizu and Goldman 1893), but burrow extensively in their invaded range in Great Britain (Harvey et al. 2011; Harvey et al. 2014; Faller et al. 2016; Rice et al. 2016; Figure 2.10). Due to their size, population density, behaviours, invasiveness and spatial extent signal crayfish are important geomorphic agents and Harvey et al. (2011, 2019) call for further interdisciplinary research between freshwater ecology, fluvial geomorphology, and hydraulics in order to quantify their geomorphic importance.

2.4.1 Crayfish as Geomorphic Agents

The interaction between crayfish and sediment already in the river channel has been considered. Signal crayfish play a significant role in altering the geomorphology of gravel bed streams (Statzner et al. 2000; Statzner et al. 2003; Johnson et al. 2011; Rice et al. 2012; Albertson and Daniels 2016; Albertson and Daniels 2018). They are able to move sediments up to six times their own weight, which by modifying bed material textures and structures can promote sediment mobility (Johnson et al. 2010) and result in morphological changes to riffle- pool sequences (Statzner et al. 2000; Statzner et al. 2003). Under laboratory conditions, twice

31 as many grains were mobilised from water-worked gravel beds after exposure to foraging and fighting crayfish, which reduced particle imbrication, which typically stabilises gravel beds, by 37% (Johnson et al. 2011). This behaviour has also been witnessed in experimental outdoor streams, where the presence of rusty crayfish (Orconectes rusticus) increased the mean bed elevation due to reduced gravel consolidation (Statzner and Peltret 2006), and by spiny-cheek crayfish (Faxonius limosus) that moved 11.4% of bed sediments over a 48 hour period in experimental mesocosms (Albertson and Daniels 2018). Signal crayfish reduce the critical shear stress required for incipient motion of gravels by reducing the sand content of gravel beds (Statzner and Sagnes 2008), and by up to 75% through the consumption of algae reducing the skin friction between gravel particles (Statzner et al. 2003), although an increase in sand infiltration to gravel beds during signal crayfish feeding has also been observed (Mathers et al. 2019a).

Figure 2.8: Signal crayfish are widespread, and have the capability to undertake substantial geomorphic work. (a) The distribution of signal crayfish in Europe in 2014 (Kouba et al. 2014), and (b) the hypothetical consequences of animal-induced changes of bed sediments (Statzner 2012). Models on the graph estimate the transport of solids from purely geophysical energy, and taxa on the x axis indicate the discharge required to mobilise sediment that would require a purely physical discharge of 2 m3 s-1 in their absence.

Figure 2.9: Signal crayfish presence in English waterbodies. Red cells indicate signal crayfish presence. Reproduced from Chadwick (2019). WCC represents white clawed crayfish, (debatably) considered native to Great Britain.

Crayfish significantly influence fine sediment dynamics in streams. In exclusion-inclusion experiments, crayfish consistently reduce the volume of fine sediment build up and storage in leaf packs and on the substrate surface. The presence of the New Zealand freshwater crayfish (Paranephrops planifrons) reduced the percentage of surficial fine sediment in mesocosm streams by up to 40% (Parkyn et al. 1997); the presence of P. zealandicus in artificial channels in Otago, New Zealand, increased the rate of leaf litter breakdown by over 500% (Usio 2000) and reduced fine sediment accrual by 75% (Usio and Townsend 2004); the presence of Appalachian brook crayfish (Cambarus bartoni) decreased the volume of fine sediments by 75% (Creed and Reed 2004) in artificial channels in Maryland, USA; the presence of Meek’s crayfish (Orconectes meeki) in exclusion/exposure cages in Arkansas, USA, reduced fine sediment mass by up to 80% (Ludlam and Magoulick 2009); and signal crayfish (Pacifastacus leniusculus) reduced fine sediment layer depth by up to 80% in experimental troughs in British Colombia, Canada (Zhang et al. 2004).

Figure 2.10: (a) a signal crayfish inhabiting a burrow in Gaddesby Brook, Leicestershire, UK, (b) crayfish burrows in riverbanks at the River Bain, Lincolnshire, UK, (c) crayfish burrows in Gaddesby Brook, Leicestershire, UK, and (d) a cast of a crayfish burrow constructed in a mesocosm.

The resuspension of fine sediment by crayfish occurs through multiple mechanisms. As crayfish walk, fine sediment is disturbed by the third and fourth pair of pereopods (Pond 1975). When walking downstream in particular, the tails and chelae are pressed into the substrate for stability (Maude and Williams 1983; Figure 2.11), which results in further bed disruption, and when exhibiting escape behaviours, crayfish move backwards by thrusting their tails, creating thrusts of approximately 1 N m-2 (Webb 1979) that can entrain fine sediment. These activities are exhibited when foraging and fighting, which can lead to significant levels of sediment resuspension (Harvey et al. 2014; Rice et al. 2014; Rice et al. 2016).

Signal crayfish are largely nocturnal (Kozak et al. 2002; Harvey et al. 2014; Johnson et al. 2014), with passive integrated transponder (PIT) tracking in the River Bain, UK, showing that only 6% of their activity occurred during daylight hours (Johnson et al. 2014). As a result, night-time fluxes in fine sediment have been recorded due to the increased levels of crayfish

34 activity, allowing for crayfish bioturbation to be quantified (Harvey et al. 2014; Rice et al. 2014; Cooper et al. 2016; Rice et al. 2016). A diel increase in turbidity of up to six times control conditions was recorded in the presence of crayfish on the River Nene, contributing at least 47% of suspended sediments during base flow (Rice et al. 2014; Rice et al. 2016; Figure 2.12a). Diel increases in turbidity have been consistently recorded in crayfish infested habitats; night- time increases in turbidity and seston at White Clay Creek, Pennsylvania (Richardson et al. 2009), the River Windrush, Oxfordshire (Harvey et al. 2014; Figure 2.12b), and the River Blackwater, Norfolk (Cooper et al. 2016; Figure 2.12c), have all been attributed to crayfish activity. These studies have all demonstrated that crayfish play a significant role in sediment dynamics in lowland rivers.

Figure 2.11: The main body parts of signal crayfish. (Source: Mathers 2017)

Figure 2.12: Diel fluxes observed in turbidity time series attributed to crayfish activity by (a) Rice et al. (2016), (b) Harvey et al. (2014), and (c) Cooper et al. (2016).

2.4.2 The Natural History of Crayfish Burrows

Freshwater crayfish are thought to have evolved from marine lobsters (Nephropoidea) during the late Palaeozoic to early Mesozoic era (c.250mya; Porter et al. 2005) and were perhaps one of the first non-marine organisms to disperse globally. Crayfish then dispersed through freshwater systems on Pangaea prior to the breakup of Laurasia and Gondwana throughout the Triassic (215 to 175 mya), resulting in a division between the two major crayfish families; Astacoidea in the northern hemisphere, and Parastacoidae in the southern hemisphere. Whilst the crayfish fossil record is poor – the oldest discovered fossils of Parastacoidae are in Australia and are approximately 106 to 116 million years old (Martin et al. 2008), and the oldest fossils of Astacoidea have been found in Arizona from the Late Triassic (Miller and Ash 1988) – fossilised burrows can be used to trace the evolution of crayfish. Fossil burrows have been interpreted to be crayfish burrows because of similarities with modern crayfish burrow morphology, and from fossilised burrows containing crayfish fossils (Hasiotis and Mitchell 1993; Babcock et al. 1998). Many paleo crayfish burrows are considerably larger than their modern equivalents, with some fossilised burrows stretching to more than 2m in length with multiple branches (Babcock et al. 1998). However, as burrow chimneys (molehill like structures found at the entrance of terrestrial burrows) have a low preservation potential, only deeper sections of burrows are readily preserved (Martin et al. 2008), and so burrows could have potentially been even larger than currently recorded.

Other structures similar to those exhibited by modern crayfish are also seen in fossils; fossils in mudstone are often found alongside tree root fossils and other woody debris, suggesting that crayfish used rocks and wood debris for cover, burrowing behind these for protection. These fossilised crayfish burrows have been discovered in partly forested flood plains, which were shaped by broad, shallow streams (Babcock et al. 1998), demonstrating that crayfish have inhabited similar environments and exhibited similar burrowing behaviours since their divergence from marine lobsters. Fossil crayfish from Antarctica and crayfish burrows show that they have been burrowing in cold water environments since an early stage of their evolution (Martin et al. 2008). Fossil evidence suggests that the earliest crayfish in the Permian lived in cool lakes fed by glacial melt water (Babcock et al. 1998), and it has been theorised that crayfish burrowed to avoid environmental extremes (Lindqvist et al. 1999). Fossorial evidence therefore shows that crayfish have been constructing burrows in floodplains and

37 riverbanks for a long time, and so their potential geomorphic impact is likely to have affected sediment dynamics over geological timescales.

2.4.3 Types of Crayfish Burrows

There are three distinct categories of burrowing crayfish; primary, secondary and tertiary burrowers (Welch and Eversole 2006). Non-burrowing crayfish have been shown to be more aggressive and more territorial than burrowing crayfish and live a solitary lifestyle (Bergman and Moore 2003) where cannibalism is prevalent (Whitledge and Rabeni 1996; Houghton et al. 2017). In contrast, burrowing crayfish, such as the Australian genus Engaeus, often live in extended family colonies (Punzalan et al. 2001). Primary burrowing crayfish, which are commonly referred to as ‘mudbugs’, are considered terrestrial species, and remain in their burrows almost continuously, using it as a retreat between feeding ventures (Martin et al. 1998). Burrows are often complex and are not necessarily connected to a permanent water supply, where the burrows themselves often serve as a feeding device, with roots within the burrows acting as a permanent food source (Noro and Buckup 2010). External factors tend to play little part in the distribution of primary burrowing crayfish; the influence of soils, seasonality and latitudes are typically absent or very weak (Thoma and Armitage 2008).

Whereas primary burrowers are considered terrestrial species, secondary burrowers live an aquatic lifestyle, only venturing on to land to burrow during the dry season when there is limited or no standing water (Breinholt et al. 2007). This period of burrowing is also when many secondary burrowers carry their eggs or young. During the wet season, secondary burrowers burrow into riverbanks, typically during the day, to use as shelter from predators. Secondary burrowers construct less sophisticated burrows than primary burrowers, although they often have multiple branches and terminals which allow for the habitation of more than one crayfish. Secondary burrowing species can tolerate both aquatic conditions and aerial exposure, and include species such as the red swamp crayfish (Procambarus clarkii), which is globally invasive and famed for its burrowing behaviour.

Tertiary burrowers are fully aquatic crayfish. Their burrows are connected to a permanent water body and are simple in morphology with little branching. Burrowing has been hypothesised to be driven by predator avoidance, for breeding or to secure a source of water during droughts. Tertiary burrowing species require a fully aquatic, open water environment to thrive, such as the common yabby (Cherax destructor), and the signal crayfish (Pacifastacus leniusculus).

2.4.4 Where Crayfish Burrow

Primary burrowing species can be found in a wide range of environments, from damp humid swamps to dry prairies (Welch and Eversole 2006; Reynolds et al. 2013). However, primary burrowing species tend to live in their burrows, using them for protection from predators, protection from extreme conditions, to gain access to water, and for feeding, and so are limited to a slow rate of territory expansion. The terrestrial desert and prairie environments that primary burrowing crayfish inhabit are typically less heterogeneous than aquatic environments (Hynes 1970), and so primary burrowers have reduced options of where to burrow than more mobile aquatic crayfish, especially considering that the range of movement of primary burrowers is a limiting habitat selection factor. Conversely, aquatic crayfish have a broader range of environmental settings available for burrowing. When combined with their greater capacity for movement; for example, red swamp crayfish are able to travel distances of 3 km day-1 (Gherardi and Barbaresi 2000). The mobility of aquatic crayfish is an important factor with regards to their geographic spread and burrowing, as local bank properties have been demonstrated to be important for the distribution of signal crayfish burrows (Faller et al. 2016).

Signal crayfish burrowing in the UK was first recorded in scientific literature by Guan (1994), and has since been recorded in lowland rivers and streams across the UK (Holdich et al. 1994; Ribbens and Graham 2004; Stanton 2004; Harvey et al. 2011; Harvey et al. 2014; Johnson et al. 2014; Rice et al. 2014; Cooper et al. 2016; Faller et al. 2016; Mathers et al. 2016; Rice et al. 2016), but has not been reported in upland catchments thus far. This is of particular concern, as UK lowland rivers are sediment capacity-limited as opposed to supply-limited (Naden et al. 2016) and are thus sensitive to changes in sediment dynamics. While burrowing has been widely reported, burrowing was not the primary research interest in previous publications, with burrows reported as evidence for crayfish bioturbation or to demonstrate population size.

Reported burrow densities in the UK vary between and within sites, which suggests that burrowing is a dynamic response to a range of environmental variables. Guan (1994) and Stanton (2004) both focussed on crayfish burrow densities, but only considered one and two rivers, respectively. Guan (1994) did not report any relationships between burrows and other biotic or abiotic factors, whilst Stanton (2004) reported a significant association between burrow densities and bank silt and clay content, but did not consider any other factors. Faller et al. (2016) considered crayfish burrows across seven rivers of the greater River Thames catchment, and at the bank scale found burrows to be located on steep, vegetation free banks

39 in wider channels. However, only bank scale morphologies were quantified, and the distribution of burrows in relation to channel sediment characteristics, river hydrology, and biological variables were not considered. Therefore, we do not yet have a clear understanding of what differences in morphology, hydrology, or biology between rivers drive the presence or absence of burrows, or how burrow density changes with any of these factors. Quantifying these relations is therefore a key first step to understanding the distribution of crayfish burrows and the potential challenges that they pose.

2.4.5 Why Crayfish Burrow

Many studies have investigated the exaggeration or adaptation of animal behaviours during invasion, such as changes in the size and structure of fire ant (Solenopsis invicta) colonies (reviewed in Holway and Suarez 1999), changing behaviours of birds (Sol and Lefebvre 2000), anti-predator behaviours in guppies (Poecilia reticulata) (Magurran et al. 1992) and the variation in voracity of praying mantis (Jones and DiRienzo 2018). However, signal crayfish burrowing is to our knowledge the first instance of an entirely new behaviour exhibited in invasive populations that is not evident in the native range. It has been theorised that crayfish, as a superfamily, burrow to avoid environmental extremes (Lindqvist et al. 1999), but this has not been addressed considering signal crayfish. Furthermore, it seems unlikely that Great Britain represents a fundamentally more extreme environment than the native range of signal crayfish, and so this seems an unlikely explanation for burrowing in Great Britain. It has been hypothesised that fiddler crabs (Uca sp.) construct burrows for protection against predators, shelter from extreme temperatures, shelter for moulting, and a location for courtship and mating (Montague 1980), but these proposals have not been tested for crayfish. It has been suggested that crayfish burrowing occurs when alternative shelters, especially coarse substrate clasts, are unavailable (Ribbens and Graham 2004), but this has not been empirically tested.

Why crayfish burrow in Great Britain, but not in their native North American range, therefore remains an unanswered question, but an answer could underpin more robust modelling of crayfish burrow distributions, now and in the future. It is therefore a key issue for understanding the extent and severity of geomorphic impacts associated with range expansion across the UK and future invasions elsewhere. Signal crayfish burrows have anecdotally been reported in Switzerland (A. Gouskov pers. comms.) and Czechia (L. Vesely pers. comms.), which suggests that the potential geomorphic impact of signal crayfish invasion extends beyond the UK.

2.4.6 Geomorphic Implications of Crayfish Burrows

Four studies and one anecdotal report have attempted to quantify the mass of sediment excavated by the riverbank burrowing of signal crayfish. The first investigation was undertaken by Guan (1994), who quantified burrow density in a lowland catchment (River Ouse, Buckinghamshire, UK), but not burrow dimensions in situ. Experimental mesocosm studies were undertaken considering the morphology of crayfish burrows, burrow occupancy rate, and differences in burrow construction between crayfish sexes (Guan 1994). No size differences between sexes were reported, but burrow dimensions were not published. Similar work was undertaken by Stanton (2004), who quantified burrow densities in a lowland catchment (Gaddesby Brook, Leicestershire, UK), and considered burrow density and internal burrow characteristics between sites. Significant differences in burrow densities and sizes were found between locations on Gaddesby Brook, suggesting that burrowing is a scaled response to environmental drivers. In the most comprehensive survey to date, Faller et al. (2016), surveyed burrows in seven lowland rivers within the greater River Thames catchment, UK, in relation to local bank conditions and river reach morphological variables. Burrows above the waterline were counted and used to estimate minimum values of excavated sediment assuming that burrows displaced an average sediment mass of 3 kg, which yielded a mean value of 3 t km-1 throughout the studied rivers, and up to 15 t km-1 in the most heavily impacted reaches. However, only burrows above the waterline and visible from bank surveys were considered, and so the estimate was reported as a minimum value. Individual burrow dimensions were not recorded, and so there is uncertainty surrounding the mass of sediment excavated per burrow. In addition, no temporal scale was attached to the production of the counted burrows, and so sediment recruitment rates were not calculated, just total excavated mass. This research provides an important step in the field of crayfish geomorphology as the first extensive burrow survey, but only considers lowland rivers, and does not consider variability in burrow dimensions between sites, as reported in Stanton (2004), as well as burrow density or rates of production.

Rice et al. (2016) investigated a diel turbidity cycle and reported the mass of sediment that signal crayfish bioturbated during baseflow on the River Nene, Northamptonshire, UK. Rice et al. (2016) indicated crayfish burrowing as a potential sediment source, and reported that burrow construction contributed 0.25 to 0.50 t km-1 a-1 of sediment to six rivers in Central England. These estimates were based on an exhaustive burrow surveys along 100 m reaches,

41 considering the dimensions of each burrow and the time that crayfish were known to have been present in the catchment. They therefore provide a minimum value, assuming that the number of burrows present represented all of the burrows constructed during the period of occupation. This is unlikely given that bank erosion can remove burrows: for example, semi-quantitative studies have shown that red swamp crayfish burrows in the Iberian Peninsula collapse frequently (Barbaresi et al. 2000; Arce and Dieguez-Uribeondo 2015).

Signal crayfish burrows have been associated with river bank erosional features (Faller et al. 2016) and it has been suggested that they can promote mass failure and accelerate bank retreat (Harvey et al. 2019). One grey-literature report has attempted to quantify the amount of bank recession associated with burrowing. In relation to an extermination attempt using trapping on the River Lark in Suffolk, some unpublished direct measurements of recession were made as an adjunct part of an MPhil (Stancliffe-Vaughan 2015) and West (2010) reported rates of up to 1 m a-1 of bank retreat, which was attributed to crayfish burrowing activity. Given the extremely limited data on this, Harvey et al. (2019) called for long term, empirical field studies investigating the impacts of burrowing on riverbank processes. This is an important avenue of research because, to date, estimates of sediment recruitment by crayfish have been purely estimated from burrow volumes. This is only the direct contribution that burrowing makes and accelerated bank erosion promoted by burrowing may represent a larger source of crayfish- derived sediment loading to rivers.

2.4.7 Predicting the Drivers and Scale of Crayfish Burrowing

The distribution of crayfish burrows is not consistent between river systems, or within a single river system (Stanton 2004), which suggests that burrowing occurs as a scaled response to environmental drivers. However, little research has been undertaken to model the biotic and abiotic conditions associated with burrow distribution. Our sum knowledge of this stems from Faller et al.’s (2016) study in the River Thames catchment, which considered burrow sizes to be consistent between reaches and considered a limited range of variables. Only morphological characteristics were quantified, with basic categorical data describing flow (‘rippled’ or ‘smooth’), bank material (‘artificial’, ‘non-cohesive’ or ‘cohesive’) and bed sediment characteristics (‘artificial’, ‘cobble’, ‘gravel’, ‘sand’, or ‘cohesive’). Further, there was no consideration of crayfish population dynamics, with crayfish assumed present from local historical surveys. Burrows were associated with wide channels with cohesive bank material, with burrows being constructed most readily on the outside bank of river meanders. Steeper

42 bank profiles were also favoured, especially those with large areas of bare, unvegetated bank face, and the presence of crayfish burrows were significantly associated with erosional features. Logistic regression explained 21% of the variance in crayfish burrow presence, and significantly predicted the presence of burrows at the bank scale, but not between reaches. As well as river morphology, river hydrology (Maude and Williams 1983; DiStefano et al. 2003; Flinders and Magoulick 2005; Clark et al. 2008; Salkonen et al. 2010; Rice et al. 2012; Johnson et al. 2014; Light 2003; Mathers et al. 2020), shelter availability (Ranta and Lindstrom 1992; Bubb et al. 2002; Bubb et al. 2004; Griffiths et al. 2004), water chemistry (Johnson et al. 2014; Gherardi et al. 2013; Welsh and Loughman 2015), and population dynamics (Ranta and Lindstrom 1993; Guan 1994; Bergman and Moore 2003; Daws et al. 2011; Albertson and Allen 2018; Ion et al. 2020) have all been independently associated with crayfish behaviour and distribution, and so are likely to also be important factors to consider to understand crayfish burrowing. There are currently no models available for assessing the likelihood or extent of burrowing beyond the bank scale, and models considering larger scales, such as the river or catchment scale, would be a vital tool for understanding the deleterious impacts associated with continued invasion.

2.5 Justification of Research Objectives

The majority of signal crayfish zoogeomorphology research has focused on gravel bed rivers and the impacts of signal crayfish on sediment already in the river channel. Very few studies have investigated the burrowing behaviour of signal crayfish, and no studies have attempted to quantify burrowing across multiple catchments. Objective 1 (to quantify the mass of sediment excavated by signal crayfish burrowing) will address this research gap (Chapter 3). There has been very limited research relating burrow distribution to local and reach scale biotic and abiotic factors and no complementary experimental testing of hypothesised drivers to unpack and understand associations observed in the field. Objective 2 (to examine the biotic and abiotic drivers of signal crayfish burrowing) will address this research gap and investigate the mechanisms that are currently not understood using field (Chapter 3), mesocosm (Chapter 4) and flume (Chapter 5) measurements.

There has been speculation surrounding the potential geomorphic impacts of riverbank burrowing beyond instantaneous sediment input associated directly with burrow excavation, and there are currently no empirical data to explore suggested mechanisms or quantify bank erosion rates. Objective 3 (to examine the processes by which crayfish burrowing impacts the

43 geomorphology of riverbank dynamics) will use field data (Chapter 6) and observations made during flume experiments (Chapter 4) to fill tis research gap. An additional set of idealised Friedkin flume experiments (Chapter 7), also feed into this question. Objective 4 (to quantify the mass of sediment recruited to rivers via direct and accelerated retreat and mass failure) will use erosion pin data to construct reach scale sediment budgets and so quantify crayfish- induced bank erosion rates and total sediment recruitment for the first time (Chapter 6).

As signal crayfish invasion continues in geographical extent and local severity, it is important to be able to predict the geomorphic implications; for example in terms of the recruitment of fine floodplain sediments into river channels. At present only one regression model exists to predict the presence or absence of burrows within a reach, and no tools are currently available to predict the extent and further impacts of burrowing within or between reaches. Objective 5 (to construct models to predict the presence, extent, and geomorphic impacts of present and future signal crayfish invasion) aims to construct tools to fulfil this research and management gap (Chapters 3 and 6).

Finally, whilst it is becoming increasingly recognised that biota play an important role in determining sediment dynamics, there are currently no empirical comparisons evaluating the relative importance of biotic and geophysical energy in recruiting sediment to river systems. This topic will be addressed in the summary Discussion (Chapter 8).

Further, this research is also justified because it is cool. An animal is introduced, escapes, and develops an entirely new behaviour of destroying riverbanks and damaging infrastructure – and nobody knows when, where, or why this occurs! In recent years, there has been a drive towards science through rationale and publishing metrics (e.g. Steel et al. 2006; Martin 2011; Wilsdon 2016), but science is equally justifiable through curiosity, excitement, and fulfilment (Weber 1946; Blashford-Snell 1994; Lloyd 2003…2020; Naismith 2006…2020; Boese 2008; Abrahams 2012; Munroe 2015; Sanders 2019), and here lies a novel and fascinating problem that I am excited to explore.

Chapter 3

Quantifying the biological, hydrological, and geophysical controls of riverbank burrowing by signal crayfish

3.1 Introduction

The potential impact of signal crayfish on river geomorphology is now well established via field (Harvey et al. 2014; Rice et al. 2014; Cooper et al. 2016; Faller et al. 2016; Rice et al. 2016) and laboratory (Stanton 2004; Harvey et al. 2014) studies. However, the geomorphic impact of signal crayfish is not homogenous. Rather, following invasion, their impact on geomorphic processes is likely to vary as a function of both local environmental considerations and the growth and size of the crayfish population. For example, although numerous studies have documented burrow construction across Great Britain (e.g. Guan 1994; Stanton 2004; Harvey et al. 2011; Johnson et al. 2014; Faller et al. 2016; Rice et al. 2016), burrow distributions vary between rivers (Faller et al. 2016), and there has been limited research characterising or quantifying crayfish burrows, and the biotic and abiotic conditions associated with their distribution. Previous reports of signal crayfish burrow distribution have been limited to lowland rivers and single catchments (e.g. Guan 1994; Harvey et al. 2014; Faller et al. 2016; Rice et al. 2016), and the spatial distribution and extent of their burrowing behaviour is currently not known. Existing work suggests that river size and bank morphology are associated with burrowing in lowland rivers (Faller et al. 2016) but these variables were not significant in determining crayfish burrow densities at the reach scale. In addition to channel morphology, river hydrology (Maude and Williams 1983; DiStefano et al. 2003; Flinders and Magoulick 2005; Clark et al. 2008; Salkonen et al. 2010; Rice et al. 2012; Johnson et al. 2014; Mathers et al. 2020), alternative shelter availability (Bubb et al. 2002; Bubb et al. 2004), water chemistry (Johnson et al. 2014; Gherardi et al. 2013; Welsh and Loughman 2015), and population characteristics (Ranta and Lindstrom 1993; Guan 1994; Bergman and Moore 2003; Daws et al. 2011; Albertson and Allen 2018; Ion et al. 2020) have all been independently associated with crayfish behaviour and distribution, and are likely to be important factors.

3.2 Aims

This chapter aims to provide a national scale field survey of signal crayfish burrowing activities which will fully address objective 1 (to quantify the mass of sediment excavated by signal crayfish burrowing), and address aspects of objectives 2 (to examine the biotic and abiotic drivers of signal crayfish burrowing behaviour) and 5 (to construct models to predict the presence, extent, and geomorphic impacts of burrowing on UK rivers). Specifically, these objectives will be addressed through the following three key questions:

1. How do crayfish burrow characteristics and associated sediment production vary between river systems across Great Britain?

Burrowing has previously only been considered within single catchments but has been recorded throughout Great Britain. To understand the extent and severity of burrowing and its impacts, it is important to quantify the distribution and density of burrows across multiple river systems.

2. What biotic and abiotic factors are associated with differences in crayfish burrowing activity?

The independent factors that drive the presence or absence of burrows, and variations in the density of burrows are currently unknown. Quantifying the associations between potentially important biotic and abiotic variables and burrow characteristics will provide a better empirical understanding of what controls their occurrence and spatial distribution.

3. Is it possible to predict the occurrence, density and excavated sediment yield of crayfish burrowing using local environmental and ecological knowledge?

Using burrow data (dependent) and corresponding abiotic and biotic data (independent) from the sites surveyed to answer questions 1 and 2 presented above, predictive models are sought for burrow occurrence, density and excavated sediment mass.

3.3 Methods

3.3.1 Overall description of approach used

A nationwide field campaign including 39 infested sites across 30 rivers (Table 3.1) surveyed crayfish burrows and a number of physical, hydrological, and biological variables that may influence crayfish burrowing activity. These data were used to document differences in burrow characteristics and excavated sediment yield between rivers, investigate the possible controls on burrowing characteristics, and build predictive models of burrow occurrence, density and sediment production.

Study Site Grid Bankfull Crayfish Area River Location Elevation (m) Underlying Geology Reference Discharge (m3 s-1) Invasion Date Central Wales Mochdre Brook Newtown SO 0882 9001 124 Silurian Siltstone 15.1 1991 East Midlands Eye Brook Tilton on the Hill SK 7411 0424 180 Jurassic Mudstone 2.7 2013 Gaddesby Brook Gaddesby SK 6931 1268 77 Jurassic Mudstone 5.5 1999 Twyford SK 7313 1019 102 Jurassic Mudstone 2.8 1993 Greet Southwell SK 7076 5437 31 Triassic Mudstone 15.3 1984 Nene Hanging Houghton SP 7429 7416 92 Jurassic Mudstone 3.3 1998 Potwell Dyke Southwell SK 7019 5346 33 Triassic Mudstone 3.6 2010 Welland Marston Trussell SP 6967 8638 91 Jurassic Mudstone 8.0 1991 Greater London Chess Croxley TQ 0699 9451 50 Cretaceous Chalk 15.0 1987 Colne Staines TQ 0308 7294 16 Palaeogene Clay 43.3 2004 Staines TQ 0342 7349 16 Palaeogene Clay 27.2 2004 North East Alwent Beck Gainford NZ 1495 1819 89 Carboniferous Mudstone 16.5 2016 Balder Cotherstone NZ 0114 2005 170 Carboniferous Mudstone 19.6 1991 Deepdale Beck Barnard Castle NZ 0445 1665 147 Carboniferous Mudstone 29.6 1991 Lance Beck Cotherstone NZ 0130 1896 198 Carboniferous Mudstone 2.1 1991 Scur Beck Lartington NY 9962 1754 268 Carboniferous Mudstone 1.6 2016 North West Bookill Gill Beck Lower SD 8402 5920 187 Carboniferous Mudstone 2.6 2011 Middle SD 8467 5964 231 Carboniferous Mudstone 9.4 2003 Upper SD 8471 6041 276 Carboniferous Mudstone 3.3 1995 Peak District Bentley Brook Ashbourne SK 1766 4815 132 Carboniferous Mudstone 14.2 2008 Etherow Brookfield SK 0096 9528 121 Carboniferous Mudstone 31.5 1997 Hamps Ford SK 0644 5395 268 Carboniferous Mudstone 35.2 1991 Torkington Brook Offerton Green SJ 9368 8900 78 Carboniferous Mudstone 10.3 1998 Offerton Green SJ 9376 8876 98 Carboniferous Mudstone 21.3 1998 South East Ouzel Monkston SP 8812 3820 63 Jurassic Clay 11.0 2001 Monkston SP 8847 3767 65 Jurassic Clay 21.8 2001 Tove Castlethorpe SP 7924 4374 65 Triassic Mudstone 17.6 1996 South West Broadmead Brook West Kington Wick ST 8237 7687 100 Jurassic Sandstone 16.7 1981 Churn Baunton SP 0206 0483 121 Jurassic Sandstone 4.5 1991 Torridge Black Torrington SS 4665 0614 71 Carboniferous Mudstone 44.4 2017 Black Torrington SS 4690 0616 73 Carboniferous Mudstone 74.8 2017 Windrush Hardwick SP 3828 0593 73 Jurassic Mudstone 16.2 1991 Southwest Scotland Clyde Allershaw Islands NS 9608 1227 279 Ordovician Wacke 14.0 2013 Elvanfoot NS 9603 1612 270 Ordovician Wacke 40.5 1992 Hapturnell Burn NS 9722 0959 299 Ordovician Wacke 16.0 2017 Allershaw Burn Watermeetings NS 9627 1233 336 Ordovician Wacke 3.5 2017 Clyde's Burn Clyde Elvanfoot NS 9698 1604 278 Ordovician Wacke 7.7 2014 Hapturnell Burn Wintercleugh NS 9727 0965 303 Ordovician Wacke 12.7 2017 West Country Thames Cricklade SU 0944 9421 81 Jurassic Clay 24.3 1991

Table 3.1: Field sites used in analyses.

3.3.2 Field Site Selection

In order to capture a full range of crayfish burrow densities, semi-systematic sampling was undertaken to ensure that rivers where burrows were absent, in low abundance, and in high abundance were surveyed. 115 rivers were considered for sampling, based on previous signal crayfish research and personal correspondence with catchment managers that had confirmed the presence of signal crayfish at the site. A total of 39 field sites on 30 rivers across Great Britain (Table 3.1; Figure 3.1) were sampled, with the field sites chosen for reasons of geographical spread, accessibility (e.g. wadable and with landowner permission), crayfish trapping licencing permissions, and to represent a range of hydrological factors (e.g. river discharge and catchment size), geological factors (e.g. underlying geology and elevation), and biological factors (e.g. recorded crayfish invasion date and recorded crayfish population density). The suite of sampled rivers was therefore representative of the broad variability of fluvial form present in the 115 rivers considered for sampling, with rivers of different size, topographic context, geographical location, land use and underlying geology, which likely represent a large proportion of the morphologies and habitats of rivers that signal crayfish are present in throughout the UK.

3.3.3 Sampling Strategy

A 75 m reach of river was randomly selected for survey at each location, and four sets of cross- sectional data were recorded in each reach (at 0 m, 25 m, 50 m, and 75 m). A reduced 50 m reach was surveyed at eight sites (Potwell Dkye, River Hamps, River Etherow, River Chess, River Churn, River Thames, River Tove, and Lance Beck) for accessibility reasons. A 75 m reach was selected as this was the greatest length of river that could be feasibly surveyed in one working day. A single long reach was chosen, as opposed to multiple shorter reaches, considering the daily range of signal crayfish. PIT tags recorded the median and maximum daily movements of signal crayfish as 24 m and 51 m respectively in a lowland river (Johnson et al. 2014), and so burrows recorded in a continuous 75 m reach were more likely to be representative of a response to the measured conditions, whereas three 25 m reaches may be more likely to record burrows constructed in response to external cues beyond the extent of the study reach. The 75 m study reaches also captured multiple morphological features (e.g. riffle and pool sequences) and were large enough to represent the typical morphological form of the sampled river (i.e. features observed in the rivers, such as bars, riffles, and riffles, were typically captured within the 75 m study reaches if present).

Data recorded at cross sections (four per reach; 25 m spacing) were reach averaged to generate single mean values for each variable at each site. Cross sections were not independent and so cannot be analysed independently, but reach averaged values provide a robust estimate of the measured variables. Multiple sites were sampled on some rivers, but the nearest neighbouring study sites were far enough apart to be considered as independent sites. Straight line, between- reach distances, not accounting for channel meandering, were: Bookill Gill Beck = 0.8 km, River Clyde = 2.8 km; River Colne = 0.6 km, Gaddesby Brook = 4.5 km; River Ouzel = 0.1 km, Torkington Brook = 0.3 km; and River Torridge = 0.1 km.

Figure 3.1: Spatial distribution of sampled field sites.

3.3.4 Crayfish Burrow Survey

A survey was undertaken to quantify the extent of burrowing at each site for the full 75 m reach (150 m of riverbank). These were conducted working upstream, to avoid disturbed sediment reducing water clarity and thereby affecting identification of burrows below the water surface. Burrows were identified by their distinctive flattened ‘D’ shape, where the floor of the burrow is flat, and is absent of latrines or feeding platforms that are present at the burrows of rats or water voles.

Burrow depths and the width and height of burrow entrances were measured using a meter rule to the nearest 5 mm. Burrow depths were measured at the centre of the opening, to account for a sloping bank face. The volume of sediment excavated was calculated by treating the burrow shape as an elliptical cylinder (Figure 3.2a) as in Faller et al. (2016):

푉퐵 = 퐴퐸 퐿 (Equation 3.1)

where VB is burrow volume, L is the length of the burrow, and AE is the burrow entrance area:

퐴퐸 = 휋 (푊⁄2 . 퐻⁄2) (Equation 3.2)

where W is the burrow entrance width, and H is the entrance height.

Figure 3.2: The measured dimensions and angles of burrowed banks, considering (a) individual burrows, and (b) riverbank geometry; considering the full height of the riverbank, and the angle of the section of riverbank where crayfish burrows were present.

The horizontal distance to the nearest burrow was measured, and whether the burrow was above or below water was recorded. Excavated sediment yield was estimated by converting burrow volumes to mass assuming a material bulk density of 1.5 g cm-3, as in Faller et al. (2016). These measurements were used to create five reach-averaged burrow variables:

(i) Burrow Density (BD). The number of burrows present per meter of riverbank surveyed (burrows m-1) averaged for the full surveyed reach. (ii) Length. The mean depth of all burrows surveyed (m). (iii) Volume. The mean mass of all burrows surveyed (kg). (iv) Percentage Underwater. The proportion of burrows that were recorded as being underwater at the time of surveying (%).

(v) Mass Density (MD). The total mass of all sediment excavated directly by burrows per meter of riverbank surveyed (kg m-1) averaged for the full surveyed reach.

The density metrics represent densities over the full reach, calculated including lengths of riverbank without burrows. Crayfish burrowing has been shown to be ‘patchy’ (Faller et al. 2016), with local burrow densities that are considerably higher than reach averages. A ‘clustered’ count density and mass density were therefore also created to consider only areas of riverbank where burrows were present. The length of affected reach was calculated as:

퐸푅 = 퐴퐷/퐻퐷 (Equation 3.3)

where ER is the calculated effected reach (%), AD is the average distance between burrows assuming an even distribution (m), and HD is the measured horizontal distance between burrows. AD is calculated as:

퐴퐷 = 2 푅퐿/퐵퐶 (Equation 3.4)

where RL is the studied reach length (m), and BC is the burrow count.

This was used to calculate the density (presence or mass) of burrows in the clustered areas at a local scale:

퐷퐶 = 퐷푅⁄퐸푅 (Equation 3.5)

where DC is clustered density (burrow density or mass density), DR is reach (burrow or mass) density.

3.3.5 Candidate variables for characterising crayfish burrows

Partly based on previous suggestions, 58 variables representing factors that may influence crayfish burrowing behaviour were measured and recorded. These are detailed in Table 3.2, presented as a schematic in Figure 3.3, and the methodology for measuring and calculating these variables, along with the rationale for employing them, are detailed below. All variables were recorded for each cross section unless otherwise stated, and the mean value from the four cross sections was calculated to create reach averaged values, unless otherwise stated.

3.3.5.1 Morphological Variables

Eleven variables relating to the local morphology of the sites were considered.

Altitude (m). Crayfish burrows have typically been investigated (Guan 1994; Stanton 2004; Faller et al. 2016) and reported (Spink and Rowe 1995; Peay 2001; Harvey et al. 2011; Johnson et al. 2014; Rice et al. 2014; Cooper et al. 2016; Rice et al. 2016) in lowland rivers, with no reports of crayfish burrowing activity in upland streams. Streams at high altitudes may be characterised by steeper slopes, coarser grain sizes, and smaller channel widths, which may individually contribute to influence burrowing. The altitude above sea level of the midpoint of each survey reach was calculated using Ordinance Survey data.

River Size. Surveys in the Thames catchment showed burrows to be associated with wider channels (Faller et al. 2016). Larger lowland rivers may promote burrowing because finer bank and bed sediments offer cohesive bank materials that aid construction while coarser bed materials are unavailable as shelter. Conversely, smaller streams have a smaller benthic area and thus a lower population carrying capacity for crayfish before alternative habitats, such as riverbanks, are required for shelter.

Figure 3.3: Schematic of sampled river reach: (a) plan view of 75 m river length, with 4 cross sections of 25 m spacing. River slope calculated along channel thalweg. Wolman sampling undertaken along transects. (b) Plan view of one cross section, with observed river width and bankfull width measurements, locations of measured depth (n > 9), and the area of measured bank angle, and material sample collection. (c) Cross sectional view of one cross section, showing bankfull width and observed river width, observed river depth and bankfull depth measurements, and the location of measured flow velocity. (d) Front on view of river bank, with burrows above and below the waterline, with measured horizontal distance to nearest burrow for effected reach calculations.

Variable Variable Morphological Variables Sediment Mobility Variables -2 Altitude (m) Reach Averaged Boundary Shear Stress, τ 0 (N m ) -2 River Width (m) Grain Stress, τ g (N m ) Bankfull Width (m) Grain Size Entrained in Flow (mm) -2 Average Depth (m) Bankfull Reach Averaged Boundary Shear Stress, τ 0 (N m ) -2 Maximum Depth (m) Bankfull Grain Stress, τ g (N m ) Bankfull Height (m) Bankfull Grain Size Entrained in Flow (mm) Bankfull Depth (m) Hydrological Variables Cross Section Area (m 2 ) Baseflow Velocity Bed (m s -1 ) Bankfull Cross Section Area (m 2 ) Baseflow Velocity Channel (m s -1 ) Slope Maximum Baseflow Velocity (m s -1 ) Land Use Bankfull Flow Velocity (m s -1 ) Riverbank Variables Baseflow Discharge (m 3 s -1 ) Bank Angle ( o ) Bankfull Discharge (m 3 s -1 ) Burrowed Bank Angle ( o ) Water Chemistry Variables Riverbank Organic Content (%) pH Riverbank Pebble Content (%) Conductivity (µS cm -1 ) Riverbank Coarse Sand Content (%) Temperature ( o C) Riverbank Fine Sand Content (%) Vegetation Variables Riverbank Sand Content (%) Macrophyte Presence Riverbank Silt and Clay Content (%) Riparian Vegetation Shading (m) Riverbed Substrate Variables Tree Shading (%) Shelter Availability (%) Crayfish Variables Riverbed D5 Time Since Invasion (a) Riverbed D10 Population Density (CPUE) Riverbed D25 Size, Carapace Length (mm) Riverbed D50 Male Size (mm) Riverbed D65 Female Size (mm) Riverbed D75 Damaged (%) Riverbed D90 Male Damaged (%) Riverbed D95 Female Damaged (%) Grains <90mm (%) Mean Damage Score Grains <128mm (%) Chelae Damaged (%) Grains <180mm (%) Antennae Damaged (%)

Table 3.2: Variables measured in the field survey.

A number of river size metrics were therefore included. The wetted width (w, m) and bankfull width of the river (wbf, m) were recorded using a 30 m tape measure. For river widths greater than 30 m, a staff and automatic level were used. The mean depth of the river (d, m) was calculated by measuring river depth every 0.5 m at each cross section, or at ten positions for rivers less than 5m wide. The maximum depth (dmax, m) on each cross-section was also recorded. Bankfull depth (dbf, m) was calculated as the average water depth plus the distance from the water surface on the day of recording to the estimated bank full water surface 2 2 elevation. Cross-sectional area (CSA, m ) and bankfull cross sectional area (CSAbf m ), were calculated as:

퐶푆퐴 = 푑 푤 (Equation 3.6)

퐶푆퐴푏푓 = 푊푏푓 푑푏푓 (Equation 3.7)

River Slope. In lowland rivers (slope < 0.001), slope has been shown to be a factor determining the location of crayfish burrows (Faller et al. 2016). Therefore, rivers were selected to encompass a range of river slopes. Thalweg long profiles were obtained using a staff and automatic level and bed slope was estimated as the slope of a line fitted using least squares linear regression.

Land Use. The land use adjacent to each bank was classified into one of six categories. These were scored from 1 to 6, with the most ‘natural’ scoring 1, and the most anthropogenically modified scoring 6 (Table 3.3). The left bank and right bank land use score were summed to create the land use score.

Score Feature Score Feature Score Feature 1 Natural or Naturalised, Unmanaged 3 Rural, Managed 5 Urban, Pathway Forest Managed Agriculture Footpath Woodland Arable Land Public Footpath Bedrock Cliff Farmland Road Naturally Vegetated Bank

2 Rural, Unmanaged 4 Urban, Managed 6 Urban, Constructed Pasture Fishery Lake House Field Public Park Industrial Estate Grassland Garden Moorland

Table 3.3: Land use category scores employed for analyses.

Hydrological Variables. Crayfish exhibit lower levels of movement (Maude and Williams 1983; Clark et al. 2008; Salkonen et al. 2010; Johnson et al. 2014), colonisation success (Mathers et al. 2020), and population density (Usio and Townsend 2000) during periods of high flow, and show a preference for areas of low flow velocity in field surveys (DiStefano et al. 2003; Flinders and Magoulick 2005; Johnson et al. 2014) and laboratory experiments (Rice et al. 2012). It has been suggested that signal crayfish use deep pools (Light 2003) and large rocks (Bubb et al. 2002; Bubb et al. 2004) to shelter from high flow velocities in headwater streams, and burrows may be constructed to serve a similar purpose. Higher average flow rates may therefore result in a greater density of burrows where they are needed for protection, or a reduced density of burrows if alternative shelters make burrowing at high flows unnecessary. River flow type (a River Habitat Survey metric; e.g. flow = ‘smooth’ or ‘rippled’) has been associated with crayfish burrows (Faller et al. 2016), but flow velocities were not quantified.

Flow gauges were not available for all reaches, but where available, they were used to determine the quantile of the flow duration curve on the day of sampling (National River Flow Archive 2020). Data were generally collected during base flow conditions (median = Q78), and measured flow velocities and discharges are referred to as ‘base flow’ values. Whilst discharge varies greatly over time, water velocity measured under base flow conditions best represent those of typical ecological equilibrium (Statzner et al. 1998), and so recorded velocities provide relative comparative values between the studied rivers. As burrowing has been hypothesised to shelter crayfish from high flows, bankfull velocities and discharges were calculated for all surveyed reaches.

Average near-bed flow velocity (m s-1) was measured over 30 seconds using a Valeport Flow Meter located 0.05m above the thalweg, and estimated the velocity crayfish would be directly -1 exposed to. Average velocity at 60% of the total flow depth above the thalweg (U60, m s ) was used as an estimate of the mean channel velocity for the cross section (Biggs et al. 1998). Along with the reach averages for these quantities, the maximum velocity (m s-1) measured at any cross-section, either at the bed or for the channel, was also included in subsequent analysis.

The discharge (m3 s-1) at the point of surveying was calculated for each cross section as:

푄 = 퐶푆퐴 푈60 (Equation 3.8)

The flow velocity (m s-1) at bankfull flow was estimated using Manning’s flow resistance equation:

2/3 1/2 푈푏푓 = (1⁄푛) 푅 푆 (Equation 3.9)

where n is Manning’s resistance coefficient (a constant value of 0.045 was used, as this is the average value for gravel bed rivers, and increases the application potential of results, where resultant equations that include values dependent on calculations containing Manning’s n can be applied by river managers without direct physical access to sites to be modelled), S is channel bed slope and R is hydraulic radius (m) calculated as:

푅 = (푤푏푓 푑푏푓)⁄(푤푏푓 + 2 푑푏푓) (Equation 3.10)

Bankfull discharge (m3 s-1) was then:

푄푏푓 = 푈푏푓 퐶푆퐴푏푓 (Equation 3.11)

Riverbed Substrate Variables. Crayfish are benthic animals, and have been recorded using mineral sediment (Peay and Rogers 1999; Bubb et al. 2002; Bubb et al. 2004), large woody material (Walter 2012), and macrophyte stands (Johnson et al. 2014) as a shelter, and so a negative relationship between river substrate shelter availability and crayfish burrowing activity was anticipated.

Shelter Availability (%). Crayfish distributions have been shown to be dependent on physical habitat quality more so than water quality parameters (Welsh and Loughman 2015), and signal crayfish have been observed using large rocks (Peay and Rogers 1999; Bubb et al. 2002; Bubb et al. 2004) and macrophyte stands (Johnson et al. 2014) as shelters. A shortage of these shelters relative to the crayfish population density may drive crayfish to burrow in order to construct an alternative shelter. The percentage of channel width that contained an alternative shelter (e.g. macrophyte stands, coarse woody material, bed clasts greater than 64mm in diameter, riparian root complexes) was used to quantify this factor.

Riverbed Grain Size. An important source of alternative shelter are large bed material clasts, and so the size distribution of riverbed grains was measured. Grain size was sampled by Wolman sampling the bed (Wolman 1954), allowing the grain size of rocks that were submerged, emergent, and exposed to be equally recorded. 100 rocks were sampled at each cross section, totalling 400 for the reach, from which percentiles (D5, D10; D25, D50; D65, D75;

D90, D95) and the percentages larger than three specific sizes (90 mm, 128 mm, 180 mm) were

58 determined. These sizes were chosen because Peay and Rogers (1999) observed that signal crayfish did not use stones finer than 64 mm as a refuge. 400 stones is an appropriate sample size for robust statistics to be calculated (Rice and Church 1996).

Sediment Mobility. If riverbed sediments that are used by crayfish as an alternative shelter are frequently disturbed and redistributed then crayfish may resort to burrowing for more reliable and stable sheltering opportunities. To incorporate bed stability as a consideration, stream discharge, estimates of reach-averaged bed shear stress and grain stress (according to Shields) were included, along with an estimate of the largest entrained particle in each reach (Rice and Toone 2011). These were calculated for bankfull and low (observed) flow, considering the depth measured at the time of survey as:

-2 Total reach averaged boundary shear stress (t 0, N m ) was calculated as:

휏0 = 푝 푔 푑푏푓 푆 (Equation 3.12)

where p is water density (1,000 kg m-3), g is gravitational acceleration -1 (= 9.81 m s ), dbf is mean bank full depth (m), and S is channel bed slope.

-2 Grain stress (tg, N m ) was partitioned according to Wilcock et al. (2009) as:

1/4 3/2 휏푔 = 17 (푆 퐷65) 푈푏푓 (Equation 3.13)

where D65 is the 65th percentile of riverbed grain size distribution (finer than, mm), and -1 Ubf is velocity for bank full flow (m s ).

The maximum grain size entrained by the bankfull flow (Dbf, mm) was then estimated according to Shields as:

퐷푏푓 = 휏푔⁄(휃 (푝푠 − 푝)푔) (Equation 3.14)

where θ is the dimensionless critical shear stress (Shields number) and ps is the average density of grain particles (2,650 kg m3). θ was set to 0.045 throughout as Shields can only be accurately determined through expert opinion and/or complex measurements which were beyond the remit of this study, and 0.045 is the average Shields value of gravel bed rivers. This also increases the application potential of results, where resultant

equations that include values dependent on calculations containing Shields can be applied by river managers without direct physical access to sites to be modelled.

Riverbank Morphology. Faller et al. (2016) suggest that steeper unvegetated banks are more likely to be used for burrows than shallow, vegetated banks and that cohesive banks provide better conditions for maintaining burrow structure (Rudnick et al. 2005). The angle of the riverbank was measured where burrows were present (o) and over the full bank height (o) using a staff and a retractable tape measure (Figure 3.2b).

Arce and Dieguez-Uribeondo (2015) argue that the amount of bank collapse caused by burrowing is dependent on the particle size of the riverbank material, with cohesive sediments more resistant to erosion (Rudnick et al. 2005). Faller et al. (2016) found that crayfish burrows in the River Thames catchment were associated with cohesive banks, but the difference was marginal (99% of burrowed sections were in cohesive material whereas 96% of non-burrowed banks consisted of cohesive material) and banks were qualitatively assigned to be cohesive or non-cohesive without analyses of riverbank materials.

A sample of riverbank sediment was therefore collected, which was a pooled sample of two representative samples from the two riverbanks at the cross section. This was then returned to the laboratory for analysis of organic matter content and grain size distribution. Samples were dried at 105 °C and stored at room temperature prior to analysis.

The organic content of the riverbank (%) was calculated using loss-on-ignition (LOI). Crucibles containing a known weight of sample were reheated to 105 °C to establish the dry weight of the sample, before being heated to 550 °C to burn the organic carbon content (Nelson and Sommers 1996). The samples were then reweighed, and the mass difference between the 550 °C sample and its dry weight was calculated to give the organic carbon content, which was presented as a percentage of the original mass. Whilst this process typically overestimates organic content, as clay minerals may lose structural water and hydroxyl groups at the temperatures used for organic combustion (Boyle 2001; Schumacher 2002), LOI was employed because other methods and pre-treatments are less reliable, cost-effective, and have the potential to be environmentally harmful (Salehi et al. 2011).

To obtain the grain size distribution of the bank sediment, several methods were investigated. First dry sieving was tested whereby samples were dried for 12 hours at 105°C, disaggregated

60 using a pestle and mortar for two minutes, and then sieved using half phi sieves to 63 μm on a sieve shaker for 20 minutes. Despite very small measured losses of sediment (mean = 0.6%; SD = 1.4%), the reported grain size distribution was coarse, and upon inspection with a microscope, it became evident that disaggregation with the pestle and mortar did not break down flocculated clay particles.

Wet sieving was therefore considered. Sediments were dried for 24 hours at 105 °C and then weighed, before being rehydrated for 24 hours in 200 ml of deionised water. After rehydration, sediments were then disaggregated in a sonic bath for 90 minutes. The sonicated sediments were then wet sieved through sieves of 2 mm, 500 μm, and 63 μm, to establish the proportions of sediment that were pebbles (>2 mm), coarse sand (2 mm > x > 500 μm), fine sand (500 μm > x > 63 μm), and silts and clays (63 μm > x) respectively. Contents of the sieves were then dried for 24 hours at 105 °C and weighed. The mass of silts and clays was calculated by subtracting the weighed mass of pebbles and sands from the original dry sediment mass.

Third, sediment was analysed using laser particle sizing. First, the organic compounds were removed by LOI. A Beckman-Coulter LS 230 laser diffraction particle sizer (LPS) was then used to size mineral particles. Sediments were manually sieved to remove grains coarser than 1.4 mm, because 1.4 mm is the maximum grain size that can be passed through the LPS pump. Sediment samples were added to the LPS after (i) no treatment, (ii) 24 hours of rehydration, and (iii) 24 hours of rehydration and 90 minutes in a sonicator bath. Continuous cycling through the LPS pump is another disaggregation method, and so samples were run for 30 minutes, with grain size distribution values being recorded every five minutes. Results demonstrated that the most effective method of disaggregating flocculated clays was sonication for 90 minutes in a sonic water bath (Table 3.4), and so this method was selected for sample disaggregation.

Of the three methods that were considered, appropriate disaggregation of sediment was not achievable using dry sieving, and the LPS was not available for the processing of all sediment, and so wet sieving was employed to particle size every sampled riverbank.

Water Chemistry. Signal crayfish activity varies with temperature (Johnson et al. 2014), and water chemistry is important for determining crayfish distributions (Welsh and Loughman 2015). Water parameters (pH, conductivity (µS cm-1), and water temperature (oC)) were recorded using HANNA HI 98128 and HANNA HI 98312 probes, respectively.

Time (minutes) Treatment D (microns) D (microns) Skew % silt and clay in LPS 50 95 Manual Sieving Dry sieving 198 1020 0.4 13.0 Wet sieving - - - 90.0 Laser Particle Sizer No treatment 5 144 809 1.89 28.3 30 117 731 2.56 34.5 24 hour rehydration 5 95 477 3.02 40.6 30 <63 352 3.89 55.6 90 minutes sonicator 5 <63 411 3.82 57.8 30 <63 333 3.81 60.8

Table 3.4: Grain size analysis of one riverbank material sample using different sizing methods.

Vegetation. Vegetation was surveyed for two reasons. First, in-channel vegetation may provide alternative shelter (Johnson et al. 2014), and so an abundance of vegetation may reduce the necessity to burrow. Second, crayfish activities are also dependent on river temperature (Johnson et al. 2014), and because riparian vegetation can moderate water temperature (Theurer et al. 1985; Wilby et al. 2012; Garner et al. 2017), it may influence crayfish burrowing activity. Three vegetation variables were therefore recorded; (i) macrophyte presence or absence, (ii) riparian vegetation shading (m), the distance that riparian vegetation (excluding trees) overhung the channel, totalled from both the left and right hand banks, and (iii) tree shading (%), the distance that trees overhung the channel, totalled from both the left and right hand banks.

Crayfish Variables. It is hypothesised that crayfish burrowing occurs when population density exceeds that of alternative shelters, and so crayfish population characteristics are a key factor that may drive burrowing.

Crayfish Population Density (CPUE). Guan (1994) found no relationship between crayfish abundance and crayfish burrow abundance working within a single catchment and Faller et al. (2016) used only historical presence/absence data rather than abundance to explain differences in burrow densities. The link between crayfish abundance and crayfish burrow abundance therefore has yet to be fully explored.

Relative crayfish population densities were estimated by trapping, using Swedish ‘trappy’ traps, baited with chicken flavoured cat food. Many crayfish sampling techniques are available (Stebbing et al. 2014) including electrofishing (Peay et al. 2015), night viewing, manual searching (Peay and Rogers 1999), kick sampling (Gladman et al. 2010), artificial refuge trapping (Green 2009; Walter 2012) and quadrat sampling (Larson et al. 2008). However, for the current research, baited traps were used as this is the most commonly used technique in crayfish research (Parkyn 2015), and therefore allows direct comparison with other studies. Whilst trapping is a relatively poor method of estimating total population density (Brown and Brewis 1978; Byrne et al. 1999; Gladman et al. 2010), it allowed for relative comparisons between the sampled rivers. The weakness of trapping is that crayfish traps are designed to harvest adult crayfish from aquaculture, so typically only retain individuals >35 mm in carapace length (Moorhouse and Macdonald 2011a), which leads to underestimation of juvenile numbers. Modifications to improve trapping of juveniles are problematic because of concerns that traps may catch other animals, including water voles, and because crayfish are cannibalistic, so trapped adults prey on trapped juveniles yielding inconsistent data. In this case, because mature crayfish are thought to be primarily responsible for burrow construction, obtaining a reproducible and robust estimate of adult numbers is acceptable.

Three traps were set in pools throughout the 75 m reach for one night. This period of deployment is consistent with other studies. Radio tagging of signal crayfish has demonstrated median and maximum daily movements of 13.5 - 15 m and 407 m respectively in upland rivers (Bubb et al. 2004), and 24m and 51m in a lowland river respectively (Johnson et al. 2014), and so traps were distributed throughout the reach in order to maximise the likelihood of trapping independent areas of crayfish movement.

The average catch per unit effort (CPUE) of the three traps was used as a surrogate measure of a population density. In cases where no crayfish were caught in traps overnight, manual hand searching for live crayfish or remnants of crayfish cadavers (Figure 3.4) was undertaken as a check. Detecting the presence of crayfish requires a minimum population density to have been reached before crayfish are detected through trapping (Gladman et al. 2010). Therefore, in line with standard practice, all CPUE results were (x+1) transformed, because a value of zero did not equal an absence of crayfish in the reach.

Crayfish Size (mm). Guan (1994) found no relationship between crayfish burrow size and crayfish size in mesocosm experiments, and suggest that large adult crayfish do not need to burrow, and instead use their carapace as protection. However, these results have not been confirmed in field studies. It is possible that larger crayfish construct larger burrows, so that burrow characteristics are partly a reflection of crayfish size. The carapace length (CL; Figure 3.5) was measured to the nearest millimetre using callipers for every trapped crayfish, and the mean size of all crayfish, mean size of all male crayfish, and mean size of all female crayfish was calculated.

Figure 3.5: Measured crayfish carapace length, from the tip of the rostrum to the posterior margin of the carapace.

Crayfish Body Damage. Signal crayfish are highly aggressive (Houghton et al. 2017) and are known to defend burrows (Ranta and Lindstrom 1993; Guan 1994; Bergman and Moore 2003). Aggressive interactions commonly result in body damage to one or other individual. If burrows are constructed as an alternative shelter to naturally occurring habitat (e.g. cobbles, macrophyte stands, woody material), suggesting an undersupply of shelters, associations between burrow densities and crayfish body damage may be present.

The percentage of captured crayfish with any physical damage including missing chelae, and missing antennae were recorded, and a ‘damage score’ was created in which one point was assigned for each damaged body part (e.g. one missing claw, one broken antenna, and two missing legs = 4). It is possible that this damage occurred from fighting after capture, but this was considered to be consistent across all traps.

Crayfish Invasion. The length of time that crayfish burrows last and the relation between burrow numbers and the length of time that crayfish have been present are both unknown, but it is reasonable to assume that burrow numbers are in some way related to time since invasion. The date of signal crayfish invasion was therefore obtained for each site, either from published work or by estimating using assumed invasion rates.

Many studies have attempted to quantify crayfish invasion rates, but the presence of crayfish requires a minimum population density to have been reached before crayfish are sufficient to be detected by trapping or sampling (Gladman et al. 2010). It is therefore widely assumed that time since invasion is generally underestimated. Long term monitoring studies tracking crayfish invasion in British rivers have reported similar invasion rates: a ten year study on the River Wharfe using stone turning estimated an invasion rate of 1.2 km a-1 in a downstream direction (Peay and Rogers 1999), and a five year study on the River Lee using trapping established population downstream spread at 1.8 km a-1 (James et al. 2016). James et al. (2016) only detected crayfish at one downstream site 9km from the source population, with no crayfish trapped in the intervening distance. A new population implant by fishermen or others could explain this, but the similarity with other UK studies make the estimated rate plausible. The radio tracking of crayfish in the River Ure and River Wharfe (albeit for maximum mean periods of 33 and 120 days respectively) removed the possibility of human interference and the introduction of new populations, and estimated a population invasion rate of 1.5 km a-1 in a downstream direction, and 0.47 km a-1 in an upstream direction (Bubb et al. 2004).

Where crayfish invasion dates were unavailable in published work about a site, time since invasion was estimated as the distance from the nearest confirmed invasion on the river times spreading rate, where an average value of 1.5 km a-1 was used for spread in a downstream direction and 0.47 km a-1 for spread in an upstream direction. The most likely direction of spread was ascertained from National Biodiversity Network (NBN 2019) sighting records along the river. Wherever possible, estimates of the numbers of years since invasion were corroborated with the NBN (2019), along with media reports and personal communications with catchment managers and angling clubs.

3.3.6 Statistical Analyses

One site (River Windrush) was excluded from analyses, as despite recent records of high crayfish population density (CPUE = 7.3; Moorhouse and Macdonald 2011a; 2011b; Harvey et al. 2014), and the presence of burrows (Harvey et al. 2011; Faller et al. 2016), no burrows or evidence of crayfish presence was found at the site.

All independent variables that were not normally distributed (Table 3.5) were log10 transformed to provide a normal distribution to facilitate the use of parametric statistics. This was

66 appropriate for all considered values. Transformed variables were then used in line with normally distributed non-transformed variables for all statistical procedures. The treatment of dependent variables (burrow metrics) are discussed below in relation to the resultant models. All statistics were performed using SPSS Version 23 (IBM 2015).

Descriptive statistics were calculated for all burrow characteristics, for the dataset as a whole and for each site independently. All burrow characteristics were not normally distributed (Shapiro-Wilk, p < 0.0005 in all cases), and so non-parametric analysis of variance (Kruskal- Wallis (H)) tests were undertaken to compare burrow characteristics between sites.

To assess univariate differences between burrowed and non-burrowed sites, two tailed t-tests (t) were used for pairwise comparisons. Associations between independent variables and burrow characteristics, particularly number and mass burrow density (BD and MD) were investigated using Pearson’s r (r).

Variable W p Variable W p Variable W p Morphological Variables Hydrological Variables Water Chemistry Variables Altitude (m) 0.789 0.084 Baseflow Velocity Bed (m s -1 ) 0.731 0.025 pH 0.977 0.882 River Width (m) 0.939 0.650 Baseflow Velocity Channel (m s -1 ) 0.669 0.005 Conductivity (µS cm -1 ) 0.950 0.719 Bankfull Width (m) 0.918 0.526 Maximum Baseflow Velocity (m s -1 ) 0.675 0.006 Temperature ( o C) 0.792 0.088 Average Depth (m) 0.920 0.538 Bankfull Flow Velocity (m s -1 ) 0.921 0.544 Vegetation Variables Maximum Depth (m) 0.888 0.375 Baseflow Discharge (m 3 s -1 ) 0.645 0.002 Macrophyte Presence - - Bankfull Height (m) 0.943 0.675 Bankfull Discharge (m 3 s -1 ) 0.803 0.108 Riparian Vegetation 0.736 0.029 Bankfull Depth (m) 0.933 0.614 Sediment Mobility Variables Shading (m) Cross Section Area (m 2 ) 0.904 0.452 Reach Averaged Boundary Tree Shading (%) 0.834 0.177 0.888 0.375 2 -2 Bankfull Cross Section Area (m ) 0.967 0.824 Shear Stress, τ 0 (N m ) Crayfish Variables -2 Slope 0.674 0.005 Grain Stress, τ g (N m ) 0.652 0.003 Time Since Invasion (a) 0.851 0.230 Land Use 0.878 0.001 Grain Size Entrained in Flow (mm) 0.652 0.003 Population Density (CPUE) 0.919 0.532 Riverbed Substrate Variables Bankfull Reach Averaged 0.808 0.117 Size, Carapace Length (mm) 0.914 0.504 -2 Shelter Availability (%) 0.894 0.404 Boundary Shear Stress, τ 0 (N m ) Male Size (mm) 0.853 0.235 -2 Riverbed D 5 (mm) 0.646 0.002 Bankfull Grain Stress, τ g (N m ) 0.772 0.060 Female Size (mm) 0.711 0.016

Riverbed D 10 (mm) 0.687 0.008 Bankfull Grain Size Entrained in Flow (mm) 0.772 0.060 Damaged (%) 0.858 0.254

Riverbed D 25 (mm) 0.711 0.015 Riverbank Variables Male Damaged (%) 0.876 0.321 o Riverbed D 50 (mm) 0.767 0.055 Bank Angle ( ) 0.913 0.499 Female Damaged (%) 0.941 0.660 o Riverbed D 65 (mm) 0.809 0.120 Burrowed Bank Angle ( ) 0.843 0.206 Mean Damage Score 0.969 0.838

Riverbed D 75 (mm) 0.855 0.243 Riverbank Orgainc Content (%) 0.659 0.003 Chelae Damaged (%) 0.878 0.330

Riverbed D 90 (mm) 0.891 0.389 Riverbank Pebble Content (%) 0.809 0.119 Antennae Damaged (%) 0.834 0.179

Riverbed D 95 (mm) 0.957 0.760 Riverbank Coarse Sand Content (%) 0.901 0.437 Grains <90mm (%) 0.835 0.182 Riverbank Fine Sand Content (%) 0.954 0.740 Grains <128mm (%) 0.871 0.303 Riverbank Sand Content (%) 0.942 0.666 Grains <180mm (%) 0.862 0.266 Riverbank Silt and Clay Content (%) 0.898 0.419

Table 3.5: Tested normality of measured and calculated mean variables for the 38 considered rivers (Shapiro-Wilk (W) test; df = 38 in all

cases). Variables not normally distributed (shown in bold) were log10 transformed (see section 3.3.6).

3.3.7 Predictive Modelling

As burrow size, and thus mass of sediment excavated from each burrow, was assumed to vary between sites, the density of crayfish burrows (BD), and the mass of sediment excavated by crayfish burrowing (MD) for a given river were modelled independently. Twelve categories of models were constructed (Table 3.6). These comprised of:

(i) Presence / absence of burrows. Logistic regression model using all data to construct a model to predict the presence or absence of burrows in a river. (ii) Density of burrows. Multiple linear regression model using all data to construct a

model, that if burrows were to be present, to predict the density of burrows (BD) in a river. (iii) Mass of sediment excavated by burrows. Multiple linear regression model using all data to construct a model, that if burrows were to be present, to predict the mass of sediment

excavated by burrows (MD) in a river.

Two levels of each of these three models were constructed:

(i) Ideal. To most accurately predict the dependent variable, ‘enter’ regression modelling was used to construct a model that best accounts for the maximum variability of the dataset and so can most accurately predict the dependent variable. (ii) Operational. To create a useable tool for river management, stepwise regression modelling was used to construct a model that best predicts the dependent variable using a reduced number of easily obtained variables.

For each linear regression model (predicting BD and MD), two further levels of model were constructed:

(i) Global model. Models using all available data to best predict the dependent variable. (ii) Restricted model. Models using data limited one set of each measured categorical variables (river hydrology; river morphology; riverbank morphology, river substrate; sediment mobility, crayfish population characteristics; vegetation characteristics, and water chemistry; Table 3.2), to (a) understand which categorical variable group are most strongly associated with crayfish burrowing, and (b) produce further reduced models for use in river management where not all data required for the global model are available.

This resulted in the production of 54 models (Table 3.6).

Predicting: Burrow Presence / Absence Burrow Metric Densities Regression: Logistic Multiple Linear Regression Model Depth Ideal Operational Ideal Operational Regression Method: Enter Stepwise Enter Stepwise

Burrow Metric Used: Burrow Presence / Absence B D M D B D M D Global Model n = 1 n = 1 n = 1 n = 1 n = 1 n = 1

Restricted Model n = 8 n = 8 n = 8 n = 8 n = 8 n = 8

Table 3.6: Regression models constructed to predict the presence / absence of burrows, the density of burrows, and the mass of sediment excavated by burrowing activity.

3.3.7.1 Variable Selection

Variable selection for global models was undertaken using Principal Component Analysis (PCA). For restricted models, variables were selected using Wald statistics and collinearity for logistic regression, and by scrutinising r and collinearity for multiple linear regression.

3.3.7.2 Principal Component Analysis

PCA was undertaken using all variables, to identify redundant variables and determine the dominant (principal) variables. PCA is a common technique used in ecology (Robertson et al. 2009; Huettmann and Diamond 2010; Janzekovic and Novak 2012) and hydrology (Olden and Poff 2003; Monk et al. 2007; Worrall et al. 2014), and is increasingly used in geomorphology (Gurnell et al. 1985; O’Hare et al. 2011; Mathers et al. 2019b).

PCA (with Direct Oblimin rotation) was undertaken using all variables. Only PCs with Eigenvalues >1.00 were selected for consideration. Twelve variables were identified for consideration from the PCA, with the number of variables selected for each axis weighted by the proportion of variance explained by each component relative to the total variation (e.g. PC1 explained 48.8% out of 100% explained variance, and so 6 of the 12 variables were selected from PC1), as in Olden and Poff (2003), Monk et al. (2007) and Worrall et al. (2014).

PCA indicated three significant PCs, which explained 100% of the variance in the data (Table 3.7). Six variables were identified from PC1, five from PC2, and one from PC3 (Table 3.8) which were used as input variables to the global models.

Initial Eigenvalues Extraction Sums of Squared Loadings Component Total % of Variance Cumulative % Total % of Variance Cumulative % 1 29.276 48.793 48.793 29.358 49.759 49.759 2 24.376 40.672 89.420 24.376 40.672 89.420 3 6.348 10.580 100.000 6.348 10.580 100.000

Table 3.7: Table of significant Principal Components from PCA.

Component Variable 1 2 3 PC1

Riverbed D 10 (mm) 1.000 -0.041 0.021 Grains >128mm (%) 0.999 0.066 0.002 -2 Reach Averaged Boundary Shear Stress, τ 0 (N m ) 0.996 -0.099 0.034

Riverbed D 75 (mm) 0.992 -0.027 -0.069 Bank Angle ( °) 0.992 0.130 -0.048 Bankfull Flow Velocity (m s -1 ) 0.991 0.088 0.156 PC2 Flow Velocity Channel at Baseflow (m s -1 ) 0.062 1.022 -0.110 Crayfish Density (CPUE) -0.152 -1.021 0.175 -2 Grain Stress at Baseflow, τg (N m ) 0.325 0.975 -0.129 Flow Velocity Bed at Low Flow (m s -1 ) 0.360 0.954 -0.067 Riverbank Pebble Content (%) 0.439 -0.883 0.535 PC3 Riverbank Fine Sand Content (%) -0.107 0.299 -1.035

Table 3.8: Table of Principal Component loadings from PCA.

3.3.7.3 Logistic Regression

Logistic regression was undertaken to probabilistically classify rivers as either having burrows present or absent. The global model was created using variables identified as PCs in PCA. For restricted models, univariate logistic regression was undertaken on every measured variable for identify candidate variables to be selected for further analysis. Variables significant at the α < 0.25 level (Wald test) were considered for further selection, and variables with p > 0.25 were removed from analysis (Table 3.9). A significance of α = 0.25 was selected as traditional levels of 0.05 and 0.10 often fail to identify important but nuanced variables in logistic regression (Bursac et al. 2008). Collinearity of variables (r > 0.7, p < 0.05) was assessed, and

71 collinear variables with the weaker Wald’s statistic were removed. Qualitative selection was undertaken for similarly scoring collinear variables, with the variable most easily measured and calculated carried forward to allow for the greatest application potential. The remaining variables were entered into a backwards stepwise elimination (likelihood ratio) procedure to select variables for inclusion in the final models.

t0 and riverbank fine sand content were removed due to the Wald statistic exceeding the critical significance level (>0.25) in univariate analysis (β = -0.021, Wald = 0.647, p = 0.421 and

β = 0.029, Wald = 0.849, p = 0.357 respectively). Of the remaining variables, D75, U60 at base flow, and Ubed at base flow were removed due to collinearity (r > 0.7, p < 0.05) with variables of a stronger Wald statistic (Table 3.10). Variables presented in Table 3.11 were then entered into backwards stepwise linear regression.

Removed Correlating Wald (1) p (1) Wald (2) p (2) r p Variable (1) Variable (2)

D 75 D 10 6.293 0.012 8.330 0.004 0.780 <0.0005

τ g U 60 2.892 0.169 2.007 0.157 0.951 <0.0005 U U 1.462 0.227 2.007 0.157 0.908 <0.0005 bed 60

Table 3.10: Removed PCA selected variables prior to logistic regression analysis. Note: τg

and U60 were similarly scoring variables, with τg having a stronger Wald

statistic. However, U60 was carried forward for modelling due to its easier measurement, thus providing a greater application potential for the model.

Variable Wald p Selected? Variable Wald p Selected? Variable Wald p Selected? Morphological Variables Hydrological Variables Water Chemistry Variables Altitude (m) 0.978 0.323 - Baseflow Velocity Bed (m s -1 ) 1.462 0.227 - pH 2.903 0.088** 1 River Width at Low Flow (m) 4.064 0.044*** 1 Baseflow Velocity Channel (m s -1 ) 2.892 0.169* 1 Conductivity (µS cm -1 ) 7.476 0.006*** 1 Bankfull Width (m) 3.834 0.05** 0 Maximum Flow Velocity Low Flow (m s -1 ) 1.852 0.174* 0 Temperature ( o C) 0.698 0.403 - Average Depth at Low Flow (m) 0.001 0.993 - Bankfull Flow Velocity (m s -1 ) 2.265 0.132* 1 Vegetation Variables Maximum Depth at Low Flow (m) 0.082 0.775 - Baseflow Discharge (m 3 s -1 ) 1.707 0.191* 1 Macrophyte Presence 0.414 0.520 - Bankfull Height (m) 0.261 0.609 - Bankfull Discharge (m 3 s -1 ) 5.412 0.020*** 1 Riparian Vegetation Shading (m) 0.056 0.812 - Bankfull Depth (m) 0.252 0.709 - Sediment Mobility Variables Tree Shading (%) 1.498 0.221* 1 Cross Section Area at Low Flow (m 2 ) 1.328 0.249* 0 Reach Averaged Boundary Crayfish Variables 0.647 0.421 - 2 -2 Bankfull Cross Section Area (m ) 2.011 0.156* 0 Shear Stress, τ 0 (N m ) Time Since Invasion (a) 2.556 0.110* 1 -2 Slope 0.662 0.416 - Grain Stress, τ g (N m ) 2.007 0.157* 1 Population Density (CPUE) 3.422 0.064** 1 Land Use 2.608 0.106* 1 Grain Size Entrained in Flow (mm) 2.007 0.157* 0 Size, Carapace Length (mm) 0.538 0.463 - Riverbed Substrate Variables Bankfull Reach Averaged Male Size (mm) 0.119 0.730 - 0.035 0.851 - -2 Shelter Availability (%) 4.955 0.026*** 1 Boundary Shear Stress, τ 0 (Nm ) Female Size (mm) 0.532 0.466 - -2 Riverbed D 5 (mm) 8.983 0.003*** 1 Bankfull Grain Stress, τ g (N m ) 0.508 0.476 - Damaged (%) 0.741 0.389 -

Riverbed D 10 (mm) 8.330 0.004*** 0 Bankfull Grain Size Entrained in Flow (mm) 0.508 0.476 - Male Damaged (%) 0.173 0.677 -

Riverbed D 25 (mm) 5.971 0.015*** 0 Riverbank Variables Female Damaged (%) 0.110 0.740 - o Riverbed D 50 (mm) 7.063 0.008*** 0 Bank Angle ( ) 3.446 0.063** 1 Mean Damage Score 0.532 0.466 - o Riverbed D 65 (mm) 6.801 0.009*** 0 Burrowed Bank Angle ( ) - - - Chelae Damaged (%) 0.397 0.529 -

Riverbed D 75 (mm) 6.293 0.012*** 0 Riverbank Orgainc Content (%) 1.146 0.284 - Antennae Damaged (%) 1.398 0.237* 1

Riverbed D 90 (mm) 6.614 0.010*** 1 Riverbank Pebble Content (%) 4.385 0.036*** 1

Riverbed D 95 (mm) 6.144 0.013*** 0 Riverbank Coarse Sand Content (%) 2.687 0.101* 1 Grains <90mm (%) 6.483 0.011*** 0 Riverbank Fine Sand Content (%) 0.849 0.357 - Grains <128mm (%) 5.418 0.020*** 0 Riverbank Sand Content (%) 0.004 0.947 - Grains <180mm (%) 4.411 0.036*** 0 Riverbank Silt and Clay Content (%) 4.375 0.036*** 1

Table 3.9: Univariate logistic regression Wald statistics used in variable redundancy analysis. Starred values are significant at the (*) α < 0.25, (**) α < 0.10; and (***) α < 0.05 level (see section 3.3.7.1 for selection considerations).

Component Variable 1 2 3 PC1

Riverbed D 10 (mm) 1.000 -0.041 0.021 Grains >128mm (%)* 0.999 0.066 0.002 -2 Bankfull Reach Averaged Boundary Shear Stress at Baseflow, τ 0 (N m ) 0.996 -0.099 0.034

Riverbed D 75 (mm)* 0.992 -0.027 -0.069 Bank Angle ( °) 0.992 0.130 -0.048 Bankfull Flow Velocity (m s -1 )* 0.991 0.088 0.156 PC2 Flow Velocity Channel at Baseflow (m s -1 ) 0.062 1.022 -0.110 Crayfish Density (CPUE) -0.152 -1.021 0.175 -2 Grain Stress at Baseflow, τg (N m )* 0.325 0.975 -0.129 Flow Velocity Bed at Baseflow (m s -1 )* 0.360 0.954 -0.067 Riverbank Pebble Content (%) 0.439 -0.883 0.535 PC3 Riverbank Fine Sand Content (%) -0.107 0.299 -1.035

Table 3.11: PCA selected variables carried forward for logistic regression analysis.

Logistic regression is primarily recommended for large datasets, at least 100 samples required when being used in a data mining context (Motrenko et al. 2014; Bujang et al. 2018; van Smeden et al. 2018), but smaller datasets of less than 10 samples per predictor variable may be acceptable when inputs are drawn from published data and expert opinion (Pavlou et al. 2015). All variables entered into the regression were identified from existing literature, and so with 38 sampled sites, regression models containing up to four predictor variables are considered acceptable.

3.3.7.4 Modelling River Sensitivity to Burrowing

Whilst environmental conditions are likely to remain relatively stable, crayfish populations change in time, and can increase very quickly during the process of invasion. The equations produced from logistic modelling were rearranged to create a tool for calculating the relative crayfish population density required to initiate burrowing in a river given known parameters.

3.3.7.5 Multiple Linear Regression

BD and MD were not normally distributed (Shapiro-Wilk, W38 = 0.752, p < 0.0005 and

W38 = 0.837, p < 0.0005 and respectively), and were positively skewed (1.523 and 2.430, respectively) due to the high number of sites where no burrows were present. As these were

74 true zeros (crayfish were confirmed as present, but no burrows were found), no transformation could normalise the dataset, and so sites where burrows were not present were removed from analysis. Square root transformations were applied to BD and MD values from the remaining sites, which successfully normalised the data (W23 = 0.954, p = 0.349 and W23 = 0.935, p = 0.137 respectively).

Global regression models including the full dataset were constructed for BD and MD using variables informed from PCA for backwards stepwise multiple linear regression. Individual linear relationships between BD and MD with the selected variables were explored using Pearson’s r (r). Highly correlated variables (r > 0.7, p < 0.05) were considered redundant, and the variable with the lowest univariate relationship with the dependent variable was removed from further analyses (Table 3.12). Qualitative selection was undertaken for similarly scoring collinear variables, with the variable most easily measured and calculated carried forward to maximise operational value. This resulted in seven variables being carried forward for further analyses (Table 3.13); previous studies employing this technique have used up to a maximum of six variables to characterise hydrological regimes (Olden and Poff 2003; Monk et al. 2007; Worrall et al. 2014; Mathers et al. 2019b), and so this was acceptable.

Component Variable 1 2 3 PC1

Riverbed D 10 (mm) 1.000 -0.041 0.021 Grains >128mm (%)* 0.999 0.066 0.002 -2 Reach Averaged Boundary Shear Stress at Baseflow, τ 0 (N m ) 0.996 -0.099 0.034

Riverbed D 75 (mm)* 0.992 -0.027 -0.069 Bank Angle ( °) 0.992 0.130 -0.048 Bankfull Flow Velocity (m s -1 )* 0.991 0.088 0.156 PC2 Flow Velocity Channel at Baseflow (m s -1 ) 0.062 1.022 -0.110 Crayfish Density (CPUE) -0.152 -1.021 0.175 -2 Grain Stress at Baseflow, τg (N m )* 0.325 0.975 -0.129 Flow Velocity Bed at Low Flow (m s -1 )* 0.360 0.954 -0.067 Riverbank Pebble Content (%) 0.439 -0.883 0.535 PC3 Riverbank Fine Sand Content (%) -0.107 0.299 -1.035

Table 3.12: Table of Principal Component loadings from PCA. Starred (*) variables were excluded from further analysis due to significant collinearity.

Removed Variable Correlating Variable r p

Grains >128mm D 10 0.820 <0.0005

D 75 D 10 0.851 <0.0005 Bankfull Flow Velocity Reach Averaged Boundary Shear Stress at Baseflow 0.917 <0.0005 Grain Stress at Baseflow Channel Flow Velocity at Baseflow 0.946 <0.0005 Riverbed Flow Velocity at Baseflow Channel Flow Velocity at Baseflow 0.925 <0.0005

Table 3.13: Table of variables removed from PCA selection due to collinearity with greater weighted variables.

Restricted regression models were constructed with the seven categories of variables. All variables with a univariate association of r > 0.3 with BD and MD (independently), that did not display collinearity (r > 0.7, p < 0.05), were used to build stepwise multiple linear regression models.

Stepwise regression is a traditional technique used to create models from complex datasets and is commonly used in ecological and geomorphological research (Winterbottom and Gilvear 1998; Diez et al. 2001; Walters et al. 2003; Thomson et al. 2004; Whittingham et al. 2006; Grauso et al. 2008). However, stepwise regression has been criticised as a method of data mining; large datasets are likely to host nuisance variables that coincidentally correlate with the dependent variable, meaning that stepwise regression models fit very large datasets well, but are not applicable to real world situations out-of-sample (Smith 2018). However, stepwise regression was deemed an acceptable technique for these analyses as:

a) All included variables were selected variables hypothesised to be linked with BD and

MD from previous literature (see section 3.5); b) As stepwise regression was used to create models from categories of variables, the number of variables used was low (maximum = 3), thus reducing the error in the calculated degrees of freedom and thus calculated r2 value, and potential error of the calculated F statistic; c) Collinear variables (r < 0.7, p < 0.05) were removed prior to entry into the stepwise regression procedure, to remove problems associated with collinearity in stepwise regression.

The construction of all models used all of the available data because there were too few cases (logistic regression n = 38, multiple linear regression n = 23) to withhold data for independent testing of the models. The models derived here therefore remain untested on external data but are the first of their kind and are open to evaluation.

3.4 Results

3.4.1 Quantifying excavated sediment and burrow characteristics

A total of 1152 burrows were surveyed from 23 of the 38 surveyed sites. The mean depth of surveyed burrows was 20.5 cm (range 2 - 87 cm), and mean burrow entrances were 5.2 cm high (range 1 - 25cm) and 7.5 cm wide (range 1 - 31cm), which resulted in a mean mass of sediment excavated of 1.15 kg per burrow (range 0.01 kg to 17.62 kg; Table 3.14). All burrow characteristics were positively skewed (Figure 3.6), and 45% of burrows were submerged underwater at the time of surveying. Significant differences were present between sites considering all burrow characteristics (Table 3.15; p < 0.0005 in all cases), and post-hoc testing (Mann-Whitney with Bonferroni correction) indicated significant pairwise differences between the four rivers with the lowest and four rivers with greatest mean burrow depths, the three rivers with the lowest and nine rivers with the greatest mean burrow entrance area; and the four rivers with the lowest and six rivers with the greatest mean burrow mass (Table 3.16).

At sites where burrows were present, reach averaged burrow densities ranged from 0.02 burrows m-1 to 1.13 burrows m-1 (mean = 0.37 burrows m-1), and burrow excavation recruited between 0.01 t km-1 and 4.14 t km-1 of bank material (mean = 0.93 t km-1; Table 3.17). Burrows were heavily clustered into small areas within the studied reaches with burrows occupying 12.0% of available riverbanks, and so local densities of burrows and excavated sediment were substantially higher; local burrow densities ranged from 1.28 burrows m-1 to 7.12 burrows m-1 (mean = 3.77 burrows m-1), which recruited between 0.93 t km-1 and 14.30 t km-1 of bank material (mean = 4.25 t km-1; Table 3.17).

Standard Mean Median Maximum Minimum Deviation Burrow Depth (cm) 20.54 18.0 11.0 87.0 2.0 Burrow Entrance Height (cm) 5.15 5.0 2.4 25.0 1.0 Burrow Entrance Width (cm) 7.53 7.0 3.5 31.0 1.0 Burrow Entrance Area (cm 2 ) 34.83 25.12 32.5 339.12 0.79

Mass of Excavated Sediment (kg) 1.15 0.74 1.4 17.62 0.01

Table 3.14: Descriptive statistics of burrow dimensions when all burrows are considered.

Figure 3.6: Frequency distribution of (a) burrow depth, (b) burrow entrance height; (c) burrow entrance width, and (d) mass of excavated sediment.

Burrow Entrance Burrow Entrance Burrow Entrance Mass of Excavated Burrow Depth (cm) 2 Burrows Height (cm) Width (cm) Area (cm ) Sediment (kg) Underwater (%) Mean Median Mean Median Mean Median Mean Median Mean Median

Bentley Brook 0.0 17.8 17.0 5.7 6.0 7.4 8.0 64.4 28.3 1.083 0.53

Bookill Gill Beck (Upper) 13.6 19.4 18.5 4.1 4.0 6 6.0 20.5 18.8 0.639 0.63

Bookill Gill Beck (Middle) 25.0 26.8 27.0 3.5 3.5 5.8 5.5 15.7 16.1 0.663 0.67

Bookill Gill Beck (Lower) 38.1 16.4 14.0 6.0 4.0 6.2 6.0 26.6 18.8 0.644 0.49

Broadmead Brook 61.4 24.5 23.0 3.4 3.0 5.5 5.0 15.9 12.6 0.672 0.42

Chess 61.0 31.9 26.0 5.7 6.0 7.8 8.0 38 33.0 2.009 1.39

Churn 5.9 18.4 18.0 4.9 5.0 7.7 7.0 31.5 27.5 0.910 0.70

Clyde (Allershaw Islands) 64.5 15.5 14.0 4.8 5.0 6.4 6.0 26.4 19.6 0.674 0.47

Clyde (Elvanfoot) 63.6 20.3 17.5 4.2 4.0 5.8 5.5 21.1 16.1 0.687 0.39

Allershaw Burn 55.6 17.5 14.5 4.5 5.0 6.9 7.0 26 27.5 0.722 0.64

Eye Brook 38.2 22.2 17.0 3.8 5.0 6.5 7.0 20.5 23.6 0.795 0.62

Gaddesby Brook (Gaddesby) 26.8 19 18.0 5.8 6.0 9.1 8.0 43.4 37.7 1.340 1.08

Gaddesby Brook (Twyford) 64.7 18.5 17.0 7.7 7.0 10.6 10.0 68.4 55.0 2.033 1.41

Greet 10.9 26.3 24.0 5 5.0 8.2 8.0 34 31.4 1.556 0.99

Nene 52.2 18.4 17.0 3.8 3.0 5.8 6.0 19 16.5 0.570 0.49

Ouzel (Upper) 41.1 20.4 18.5 8.1 7.5 11.1 9.5 74.6 55.0 2.212 1.81

Ouzel (Lower) 58.3 21.3 19.5 7.4 7.0 9.8 9.5 60.2 52.2 2.042 1.43

Potwell Dyke 7.9 15.9 11.5 4.8 4.5 6.6 6.0 27.6 19.2 0.818 0.36

Thames 100.0 19.8 16.0 4.1 4.0 6.9 7.0 22.7 22.0 0.684 0.79

Torkington (Upper) 0.0 13.6 13.0 4.5 4.0 6.1 6.0 21.5 18.8 0.442 0.38

Torkington (Lower) 0.0 8.7 7.0 3.7 4.0 6 6.0 17.5 18.8 0.239 0.20

Tove 26.9 25.5 24.0 5.3 5.0 6.8 6.0 32.1 20.8 1.402 0.77

Welland 81.7 19.3 21.0 4.8 4.0 7 6.0 27.9 18.8 0.870 0.59

Table 3.15: Mean and median burrow dimensions by site. Only sites where burrows were present (n = 23) are included.

Low Value River High Value River Adjusted U Adjusted p Low Value River High Value River Adjusted U Adjusted p Low Value River High Value River Adjusted U Adjusted p Burrow Depth Burrow Entrance Area Burrow Mass Torkington Brook Broadmead Brook 4.125 0.009 Broadmead Brook Eye Brook -5.429 <0.0005 Torkinton Brook Gaddesby Brook 3.972 0.018 Upper Tove -3.929 0.022 Tove -3.749 0.045 Upper at Gaddesby Greet 4.785 <0.0005 Greet -6.357 <0.0005 Greet 3.852 0.030 Chess 4.981 <0.0005 Bentley Brook 4.526 0.002 Chess 4.407 0.003 Chess -6.454 <0.0005 Gaddesby Brook -2.208 <0.0005 Potwell Dyke Broadmead Brook 5.189 <0.0005 Gaddesby Brook at Twyford -11.266 <0.0005 Tove -4.362 0.003 at Gaddesby Ouzel Lower 5.008 <0.0005 Greet 5.749 <0.0005 Ouzel Lower -10.458 <0.0005 Ouzel Upper 5.468 <0.0005 Chess 5.881 <0.0005 Gaddesby Brook -15.792 <0.0005 at Twyford Nene Gaddesby Brook 7.226 <0.0005 Clyde at Broadmead Brook 4.309 0.004 Ouzel Upper -10.23 <0.0005 at Gaddesby Allershaw Islands Tove -3.833 0.032 Greet 5.891 <0.0005 Greet -4.994 <0.0005 Nene Eye Brook 3.909 0.023 Chess 6.454 <0.0005 Chess 5.176 <0.0005 Greet 5.039 <0.0005 Gaddesby Brook -10.193 <0.0005 Chess 5.271 <0.0005 at Twyford Bookill Gill Beck Greet -3.911 0.023 Gaddesby Brook Ouzel Lower -7.656 <0.0005 9.341 <0.0005 Lower Chess -4.143 0.009 at Gaddesby Ouzel Upper -7.805 <0.0005 Ouzel Lower -9.127 <0.0005 Gaddesby Brook Potwell Dyke Gaddesby Brook 4.514 0.002 -13.548 <0.0005 at Twyford at Gaddesby Ouzel Upper -9.091 <0.0005 Greet 4.166 0.008 Chess 4.81 <0.0005 Clyde at Elvanfoot Chess 3.839 0.031 Gaddesby Brook -6.298 <0.0005 Gaddesby Brook at Twyford -6.049 <0.0005 at Gaddesby Ouzel Lower 5.634 <0.0005 Ouzel Lower -6.858 <0.0005 Ouzel Upper 6.076 <0.0005 Gaddesby Brook -8.763 <0.0005 at Twyford Broadmead Brook Gaddesby Brook -6.817 <0.0005 Ouzel Upper -7.19 <0.0005 at Gaddesby Greet -5.498 <0.0005 Chess -6.099 <0.0005 Gaddesby Brook -9.904 <0.0005 at Twyford Ouzel Lower -7.312 <0.0005 Ouzel Upper -7.491 <0.0005

Table 3.16: Pairwise (Mann-Whitney with Bonferroni correction) comparisons of burrow depth, entrance area and mass between sites. These characteristics were chosen for display due to their potential geomorphic significance (burrow depth in potentially undercutting banks, entrance area in potentially increasing bank face turbulence and mass in potentially altering bank geotechnical properties).

Burrow Density Proportion of Sediment Excavated Local Count Density Local Sediment Excavated (burrows m-1 riverbank) Impacted Bank (%) (t km-1 river) (burrows m-1 riverbank) (kg m-1 riverbank)

Bentley Brook 0.14 - 0.30 - -

Bookill Gill Beck (Upper) 0.29 19.3 0.38 1.51 0.97

Bookill Gill Beck (Middle) 0.03 0.4 0.04 6.36 4.22

Bookill Gill Beck (Lower) 0.14 7.8 0.18 1.79 1.16

Broadmead Brook 1.13 26.1 1.52 4.32 2.90

Chess 0.84 11.8 3.38 7.12 14.30

Churn 0.17 4.0 0.30 4.25 3.87

Clyde (Allershaw Islands) 0.21 7.0 0.28 2.97 2.00

Clyde (Elvanfoot) 0.29 10.9 0.40 2.68 1.84

Allershaw Burn 0.07 5.3 0.10 1.28 0.93

Eye Brook 0.57 19.3 0.98 2.94 2.34

Gaddesby Brook (Gaddesby) 0.82 24.8 2.20 3.31 4.43

Gaddesby Brook (Twyford) 1.02 27.1 4.14 3.76 7.65

Greet 0.37 10.8 1.14 3.41 5.30

Nene 0.89 19.2 1.02 4.65 2.65

Ouzel (Upper) 0.23 5.7 1.00 4.00 8.85

Ouzel (Lower) 0.32 - 1.30 - -

Potwell Dyke 0.38 11.1 0.62 3.42 2.80

Thames 0.09 2.5 0.12 3.60 2.46

Torkington (Upper) 0.11 - 0.10 - -

Torkington (Lower) 0.02 - 0.01 - -

Tove 0.52 8.7 1.46 5.98 8.38

Welland 0.23 5.4 0.38 4.30 3.74

Mean 0.39 11.96 0.93 3.77 4.25

Median 0.29 10.80 0.40 3.60 2.90

Standard Deviation 0.3 8.2 1.1 1.5 3.4

Minimum 0.02 0.44 0.01 1.28 0.93

Maximum 1.13 27.10 4.14 7.12 14.30

Table 3.17: Burrow density (burrows m-1 riverbank), proportion of impacted riverbank (%), total sediment excavated directly by crayfish burrows (t m-1 river), local burrow count density (burrows m-1 riverbank) and local sediment excavated (kg m-1 riverbank). Only sites where burrows were present are included. Too few burrows were present at Bentley Brook, Torkington (Upper) and Torkington (Lower) for the proportion of impacted riverbank to be calculated. Distances between burrows at Ouzel (Lower) were not measured due to fieldwork time constraints.

3.4.2 Univariate Associations Between Variables and Burrow Characteristics

3.4.2.1 Differences between burrowed and non-burrowed rivers

Burrows were recorded from 23 of the 38 surveyed crayfish invaded reaches, and significant differences in measured variables were observed between burrowed and non-burrowed reaches (Table 3.18). Rivers where burrows were present were significantly smaller (smaller width, bankfull width, and bankfull discharge) than rivers where burrows were absent (4.7 m and 8.2 m, t = 2.271, p = 0.029; 6.3 m and 10.1 m, t = 0.708, p = 0.032; and 11.79 m3 s-1 and 25.14 m3 s-1, t = 2.874, p = 0.007 respectively; Figure 3.7a; Figure 3.7b), and burrowed rivers had a higher water conductivity than rivers where burrows were absent (640.4 µS cm-1 and 282.1 µS cm-1, t = 3.271, p = 0.003; Figure 3.7c). Riverbed grain size percentiles were significantly smaller (meaning bed material was significantly finer) at every calculated interval in rivers where burrows were present, and had a lower shelter availability than rivers where burrows were absent (Table 3.18; Figure 3.7d). The greatest difference in substrate metrics between burrowed and non-burrowed rivers was for the percentage of grains greater than 90 mm (8.6% and 26.0% respectively, t = 3.074, p = 0.004). Bank material was also finer, with a significantly lower mass of pebbles and significantly greater mass of silt and clay in reaches where burrows were present compared to where burrows were absent (9.2% and 22.2%, t = 2.429, p = 0.021; and 44.8% and 32.6%, t = 2.334, p = 0.026 respectively; Figure 3.7e; Figure 3.7f). In-channel and riparian vegetation had no significant association with crayfish burrowing, and there were no significant differences in crayfish population characteristics between burrowed and non- burrowed reaches (Table 3.18).

Mean Mean Variable t p Unburrowed Burrowed Morphological Variables Altitude (m) 162.7 132.1 0.984 0.332 River Width (m) 8.2 4.7 2.271 0.029* Bankfull Width (m) 10.1 6.3 0.708 0.032* Average Depth (m) 0.26 0.26 0.032 0.975 Maximum Depth (m) 0.41 0.43 0.279 0.782 Bankfull Height (m) 0.98 0.92 0.502 0.619 Bankfull Depth (m) 1.25 1.16 0.492 0.626 Cross Section Area (m 2 ) 2.44 1.60 1.177 0.247 Bankfull Cross Section Area (m 2 ) 14.00 8.63 1.509 0.140 Slope 0.012 0.014 0.805 0.426 Land Use 6.2 8.3 1.686 0.100 Hydrological Variables Baseflow Velocity Bed (m s -1 ) 0.198 0.145 1.222 0.230 Baseflow Velocity Channel (m s -1 ) 0.345 0.235 1.403 0.169 Maximum Flow Velocity Low Flow (m s -1 ) 0.342 0.237 1.387 0.174 Bankfull Flow Velocity (m s -1 ) 1.95 1.62 1.330 0.192 Baseflow Discharge (m 3 s -1 ) 0.60 0.48 1.552 0.129 Bankfull Discharge (m 3 s -1 ) 25.14 11.79 2.874 0.007** Riverbed Substrate Variables Shelter Availability (%) 65 41 2.483 0.018*

Riverbed D 5 (mm) 11.2 4.2 3.812 0.001**

Riverbed D 10 (mm) 16.0 5.7 3.553 0.001**

Riverbed D 25 (mm) 28.7 11.0 2.789 0.008**

Riverbed D 50 (mm) 52.2 21.9 3.209 0.003**

Riverbed D 65 (mm) 70.5 31.3 3.128 0.003**

Riverbed D 75 (mm) 85.9 40.3 2.959 0.005**

Riverbed D 90 (mm) 140.9 69.2 3.108 0.004**

Riverbed D 95 (mm) 184.8 94.9 2.952 0.006** Grains <90mm (%) 26.0 8.6 3.074 0.004** Grains <128mm (%) 15.1 5.3 2.698 0.011* Grains <180mm (%) 6.7 2.0 2.492 0.017*

Table 3.18: t-tests (two tailed) between burrowed and non-burrowed. Note that means stated are raw values prior to transformation. Continued below.

Mean Mean Variable Unburrowed Burrowed t p Sediment Mobility Variables -2 Reach Averaged Boundary Shear Stress, τ 0 (N m ) 16.7 13.0 0.823 0.416 -2 Grain Stress, τ g (N m ) 3.9 1.7 1.455 0.115 Grain Size Entrained in Flow (mm) 5.4 2.4 1.455 0.115 -2 Bankfull Reach Averaged Boundary Shear Stress, τ 0 (Nm ) 122.0 114.7 0.182 0.856 -2 Bankfull Grain Stress, τ g (N m ) 42.7 33.5 0.705 0.485 Bankfull Grain Size Entrained in Flow (mm) 58.6 46.0 0.705 0.485 Riverbank Variables Bank Angle ( o ) 48.2 62.3 2.014 0.052 Burrowed Bank Angle ( o ) - 69.4 - - Riverbank Orgainc Content (%) 8.2 10.0 1.072 0.292 Riverbank Pebble Content (%) 22.2 9.2 2.429 0.021* Riverbank Coarse Sand Content (%) 11.7 7.6 1.750 0.090 Riverbank Fine Sand Content (%) 34.7 38.5 0.917 0.366 Riverbank Sand Content (%) 46.3 46.1 0.063 0.949 Riverbank Silt and Clay Content (%) 32.6 44.8 2.334 0.026* Water Chemistry Variables pH 8.3 7.9 1.801 0.081 Conductivity (µS cm -1 ) 282.1 640.4 3.271 0.003** Temperature ( o C) 18.0 15.9 0.929 0.359 Vegetation Variables Macrophyte Presence (%) 0.27 0.43 0.641 0.525 Riparian Vegetation Shading (m) 33.2 34.3 0.228 0.822 Tree Shading (%) 47.2 65.1 1.228 0.228 Crayfish Variables Time Since Invasion (a) 12.7 18.6 1.649 0.108 Population Density (CPUE) 2.8 5.2 2.013 0.052 Size, Carapace Length (mm) 47.5 48.8 0.720 0.478 Male Size (mm) 49.5 50.3 0.332 0.743 Female Size (mm) 45.8 47.5 0.716 0.481 Damaged (%) 42.4 34.7 0.861 0.397 Male Damaged (%) 34.4 29.9 0.402 0.691 Female Damaged (%) 43.5 39.8 0.321 0.752 Mean Damage Score 57.5 48.4 0.718 0.479 Chelae Damaged (%) 18.9 23.5 0.616 0.543 Antennae Damaged (%) 27.1 15.2 1.248 0.223

Table 3.18: t-tests (two tailed) between burrowed and non-burrowed. Note that means stated are raw values prior to transformation.

Figure 3.7: Significant differences between (a) river width, (b) bankfull discharge; (c) conductivity, (d) grains >90 mm; (e) bank mass classed as pebbles, and (f) bank mass classed as silt and clay in rivers where burrows are present, and burrows are absent, considering mean values +/-1 standard error (SEM).

3.4.2.2 Associations between measured variables and increasing burrow densities

Significant associations were observed between BD and MD and biological, hydrological, and morphological variables (Table 3.19; Table 3.20). Burrow densities were negatively associated with channel slope (Table 3.19; Figure 3.8a, 3.8e), and with the measured and calculated flow velocities at base flow and bankfull (Figure 3.8b, 3.8f). Riverbed grain size distributions were negatively associated with burrow densities for every measured clast size, with the size of coarse material having significantly stronger associations with MD than fine material (Table 3.20; Figure 3.8c, 3.8g; Figure 3.9). Calculated shear stresses, grain stresses, and the size of sediment entrained in the flow were strongly negatively associated with crayfish burrow densities (Figure 3.8d, 3.8h), with base flow velocities having marginally stronger relationships with burrow densities than calculated bankfull metrics. The only significant association between burrow densities and riverbank morphology was a positive association between

85 riverbank silt and clay content and BD (Table 3.19; Figure 3.8i, 3.8m). The extent of overhanging riparian vegetation was the only vegetation variable significantly associated with burrow densities (Table 3.19; Figure 3.8j, 3.8n). No water chemistry parameters were associated with either BD or MD (Table 3.19; Table 3.20).

Regarding biological parameters, the time since invasion was positively associated with BD and

MD (Table 3.19; Table 3.20; Figure 3.8k, 3.8o), as was the size of male crayfish (Table 3.19; Table 3.20; Figure 3.8l, 3.8p). The size of all crayfish captured was significantly associated with MD (Table 3.20). There was no association for either BD or MD with crayfish population density (CPUE), or with any physical damage / injury observed on the crayfish sampled (Table 3.20; Table 3.20).

3.4.3 Modelling Burrows and Sediment Input

Logistic regression was used to develop a predictive model for the likelihood of crayfish burrow presence or absence in rivers with crayfish, and multiple linear regression was used to -1 develop predictive models of crayfish burrow density (BD, m ) and the mass of sediment -1 excavated by burrowing activities (MD, t km ) where crayfish burrows are present.

3.4.3.1 Logistic Regression: Ideal Models

The global model, constructed using PCA informed variables, was statistically significant (X2 = 25.779, p < 0.0005; Figure 3.10a), and explained 68.3% of the variance in burrow presence between rivers (Nagelkerke R2; Table 3.21). This resulted in a 26.5 point increase in predictive power from the null model. The proportion of riverbed material coarser than 128 mm (%), crayfish population density (CPUE+1), and riverbank angle (°) were the most important predictors of the presence of burrows (Table 3.22).

Restricted models were also successfully developed using water chemistry (Nagelkerke R2 = 50.8%; Figure 3.10b), riverbank morphology (31.5%; Figure 3.10c); riverbed substrate (38.5%; Figure 3.10d), river hydrology (32.2%; Figure 3.10e); and river morphology (33.2%; Figure 3.10f), all significantly predicting the presence or absent of crayfish burrows (Table 3.21). No significant models were developed based on crayfish population characteristics, sediment mobility, or vegetation characteristics (Table 3.21).

Variable r p Variable r p Variable r p Morphological Variables Hydrological Variables Water Chemistry Variables Altitude (m) -0.349 0.051 Baseflow Velocity Bed (m s -1 ) -0.661 0.001** pH -0.201 0.197 River Width at Low Flow (m) 0.062 0.389 Baseflow Velocity Channel (m s -1 ) -0.572 0.002** Conductivity (µS cm -1 ) 0.323 0.082 Bankfull Width (m) -0.012 0.479 Maximum Flow Velocity Low Flow (m s -1 ) -0.565 0.002** Temperature ( o C) -0.048 0.415 Average Depth at Low Flow (m) 0.020 0.465 Bankfull Flow Velocity (m s -1 ) -0.573 0.003** Vegetation Variables Maximum Depth at Low Flow (m) 0.146 0.253 Baseflow Discharge (m 3 s -1 ) -0.149 0.254 Macrophyte Presence - - Bankfull Height (m) 0.078 0.361 Bankfull Discharge (m 3 s -1 ) -0.204 0.175 Riparian Vegetation 0.474 0.032* Bankfull Depth (m) 0.027 0.452 Sediment Mobility Variables Shading (m) Cross Section Areaat Low Flow (m 2 ) -0.043 0.423 Reach Averaged Boundary Tree Shading (%) 0.288 0.092 0.161 0.232 2 -2 Bankfull Cross Section Area (m ) -0.046 0.418 Shear Stress, τ 0 (N m ) Crayfish Variables -2 Slope -0.436 0.021* Grain Stress, τ g (N m ) -0.684 < 0.0005*** Time Since Invasion (a) 0.389 0.033* Land Use 0.015 0.472 Grain Size Entrained in Flow (mm) -0.684 < 0.0005*** Population Density (CPUE) 0.234 0.141 Riverbed Substrate Variables Bankfull Reach Averaged Size, Carapace Length (mm) 0.217 0.172 -0.521 0.006** -2 Shelter Availability (%) -0.449 0.016* Boundary Shear Stress, τ 0 (Nm ) Male Size (mm) 0.589 0.006** -2 Riverbed D 5 (mm) -0.361 0.045* Bankfull Grain Stress, τ g (N m ) -0.574 0.003** Female Size (mm) 0.074 0.379

Riverbed D 10 (mm) -0.386 0.034* Bankfull Grain Size Entrained in Flow (mm) -0.574 0.003** Damaged (%) -0.127 0.292

Riverbed D 25 (mm) -0.428 0.021* Riverbank Variables Male Damaged (%) -0.018 0.473 o Riverbed D 50 (mm) -0.407 0.027* Bank Angle ( ) 0.045 0.419 Female Damaged (%) -0.104 0.332 o Riverbed D 65 (mm) -0.449 0.016* Burrowed Bank Angle ( ) -0.216 0.239 Mean Damage Score -0.158 0.247

Riverbed D 75 (mm) -0.459 0.014* Riverbank Orgainc Content (%) 0.188 0.201 Chelae Damaged (%) -0.214 0.175

Riverbed D 90 (mm) -0.461 0.013* Riverbank Pebble Content (%) -0.343 0.059 Antennae Damaged (%) -0.019 0.468

Riverbed D 95 (mm) -0.398 0.030* Riverbank Coarse Sand Content (%) -0.133 0.278 Grains <90mm (%) -0.309 0.076 Riverbank Fine Sand Content (%) 0.012 0.479 Grains <128mm (%) -0.333 0.060 Riverbank Sand Content (%) -0.068 0.682 Grains <180mm (%) -0.377 0.038* Riverbank Silt and Clay Content (%) 0.361 0.049*

Table 3.19: Linear associations between tested variables and burrow density (BD) for all 23 sites where crayfish were recorded.

Variable r p Variable r p Variable r p Morphological Variables Hydrological Variables Water Chemistry Variables Altitude (m) -0.438 0.018* Baseflow Velocity Bed (m s -1 ) -0.666 < 0.0005*** pH -0.239 0.155 River Width at Low Flow (m) 0.100 0.325 Baseflow Velocity Channel (m s -1 ) -0.581 0.002** Conductivity (µS cm -1 ) 0.379 0.050 Bankfull Width (m) 0.019 0.466 Maximum Flow Velocity Low Flow (m s -1 ) -0.573 0.002** Temperature ( o C) 0.112 0.309 Average Depth at Low Flow (m) 0.179 0.213 Bankfull Flow Velocity (m s -1 ) -0.601 0.002** Vegetation Variables Maximum Depth at Low Flow (m) 0.343 0.055 Baseflow Discharge (m 3 s -1 ) -0.051 0.411 Macrophyte Presence - - Bankfull Height (m) 0.050 0.411 Bankfull Discharge (m 3 s -1 ) -0.218 0.159 Riparian Vegetation 0.413 0.056 Bankfull Depth (m) 0.046 0.418 Sediment Mobility Variables Shading (m) Cross Section Areaat Low Flow (m 2 ) 0.007 0.487 Reach Averaged Boundary Tree Shading (%) 0.274 0.103 0.316 0.071 2 -2 Bankfull Cross Section Area (m ) -0.050 0.411 Shear Stress, τ 0 (N m ) Crayfish Variables -2 Slope -0.502 0.009** Grain Stress, τ g (N m ) -0.689 < 0.0005*** Time Since Invasion (a) 0.355 0.048* Land Use -0.054 0.403 Grain Size Entrained in Flow (mm) -0.689 < 0.0005*** Population Density (CPUE) 0.339 0.057 Riverbed Substrate Variables Bankfull Reach Averaged Size, Carapace Length (mm) 0.459 0.018* -0.529 0.006** -2 Shelter Availability (%) -0.397 0.030* Boundary Shear Stress, τ 0 (Nm ) Male Size (mm) 0.448 0.036* -2 Riverbed D 5 (mm) -0.389 0.033* Bankfull Grain Stress, τ g (N m ) -0.570 0.003** Female Size (mm) 0.321 0.084

Riverbed D 10 (mm) -0.396 0.031* Bankfull Grain Size Entrained in Flow (mm) -0.570 0.003** Damaged (%) -0.200 0.193

Riverbed D 25 (mm) -0.441 0.018* Riverbank Variables Male Damaged (%) 0.176 0.250 o Riverbed D 50 (mm) -0.415 0.024* Bank Angle ( ) -0.036 0.435 Female Damaged (%) -0.375 0.052 o Riverbed D 65 (mm) -0.449 0.016* Burrowed Bank Angle ( ) -0.314 0.015 Mean Damage Score -0.152 0.255

Riverbed D 75 (mm) -0.456 0.014* Riverbank Orgainc Content (%) 0.241 0.140 Chelae Damaged (%) -0.336 0.068

Riverbed D 90 (mm) -0.457 0.014* Riverbank Pebble Content (%) -0.325 0.070 Antennae Damaged (%) -0.027 0.455

Riverbed D 95 (mm) -0.431 0.020* Riverbank Coarse Sand Content (%) 0.028 0.451 Grains <90mm (%) -0.347 0.053 Riverbank Fine Sand Content (%) 0.028 0.451 Grains <128mm (%) -0.387 0.034* Riverbank Sand Content (%) 0.044 0.423 Grains <180mm (%) -0.399 0.030* Riverbank Silt and Clay Content (%) 0.260 0.121

Table 3.20: Linear associations between tested variables and mass density (MD) for all 23 sites where crayfish were recorded..

Figure 3.8: Selected associations between burrow mass density (MD; a-d) and burrow density (BD; e-h) with (a, e) channel slope, (b, f) flow

velocity at the riverbed at base flow; and (d, g) riverbed substrate D90. Continued below.

Figure 3.8: Selected associations between burrow mass density (MD; i-l) and burrow density (BD; m-p) with (i, m) proportion of riverbank mass classified as silt and clay, (j, n) distance of overhanging riparian vegetation; (k, o) time since initial crayfish invasion, and (l, p) mean male crayfish size.

Figure 3.9: Association between riverbed grainsize and the strength of the relationship

between riverbed grainsize and (a) burrow density (BD; r2 = 0.425, t = 2.11, 2 p = 0.079) and (b) burrow mass density (MD; r = 0.582, t = 2.888, p = 0.028).

Figure 3.10: Logistic regression prediction models of Table 3.21 considering (a) PCA informed global model, and restricted models considering (b) water chemistry, (c) river morphology; (d) river substrate, (e) river hydrology; and (f) riverbank morphology.

Predictive Increase Omnibus Tests of Predictions Model Summary Hosmer and Lamashow Test Determined Predictor Variables From Null Model Model Coefficients Correct (%) (=60.0) Nagelkerke R 2 X 2 df p X 2 df p Global, PCA 86.5 26.5 0.683 25.779 3 <0.0005 7.099 7 0.419

Water Chemistry 81.8 21.8 0.508 15.783 2 <0.0005 11.876 8 0.157

Riverbank Morphology 79.4 19.4 0.315 8.832 2 0.012 13.517 8 0.095

Riverbed Substrate 76.3 16.3 0.385 12.696 3 0.005 7.982 8 0.435

River Hydrology 75.7 15.7 0.332 10.463 4 0.033 5.528 7 0.596

River Morphology 71.1 11.1 0.162 4.847 1 0.028 5.855 8 0.663

Crayfish Population Characteristics No significant models found

Sediment Mobility No significant models found Vegetation Characteristics No significant models found

Table 3.21: Logistic regression models (enter) statistics.

Predictor Variables β SE β Wald's X 2 df p eβ (odds ratio)

Predictor: Global, PCA Constant -5.969 2.906 4.220 1 0.04 - Crayfish Population Density 0.826 0.35 5.586 1 0.018 2.284 Riverbank Angle 0.100 0.043 5.371 1 0.02 1.105 Grains >128mm -0.242 0.091 7.017 1 0.008 0.785 Predictor: Water Chemistry Constant 9.330 5.75 2.629 1 0.105 - pH -1.443 0.76 3.628 1 0.057 0.236 Conductivity 0.006 0.00 7.313 1 0.007 1.006 Predictor: River Morphology Constant 1.419 0.60 5.595 1 0.018 - River Width -0.160 0.08 4.064 1 0.044 0.852 Predictor: River Substrate Constant 2.848 0.933 9.324 1 0.002 - Hide Availability -0.083 0.153 0.296 1 0.586 0.920

log 10 D 5 -2.200 1.297 2.876 1 0.090 0.111

D 90 -0.005 0.007 0.456 1 0.499 0.995 Predictor: River Hydrology Constant 2.349 1.463 2.580 1 0.108 -

log 10 U 60 -1.820 1.659 1.203 1 0.273 0.162

U bf -0.259 0.597 0.188 1 0.664 0.772

log 10 Q 1.053 1.105 0.908 1 0.341 2.865

Q bf -0.115 0.060 3.591 1 0.058 0.892 Predictor: Riverbank Morphology Constant -4.015 1.87 4.603 1 0.032 - Bank Angle 0.035 0.02 3.137 1 0.077 1.035 Bank Silt and Clay Content 0.068 0.03 4.672 1 0.031 1.070

Table 3.22: Logistic regression models (enter) variable statistics.

3.4.3.2 Logistic Regression: Operational Models

The construction of restricted models using the stepwise regression procedure produced the same results for all models, except for river hydrology, as the enter regression procedure, considered in section 4.3.1 (Table 3.23; Table 3.24). The restricted stepwise model using only river hydrology variables considered only one, as opposed to four, input variables, and was weaker than the ‘enter’ model, which explained 25.6% (compared to 33.2%; Nagelkerke R2) of variance in burrow presence, and increased the predictive power of the null model by 13.0 points (compared to 15.7 points; Table 3.25).

Predictive Increase Omnibus Tests of Predictions Model Summary Hosmer and Lamashow Test Determined Predictor Variables From Null Model Model Coefficients Correct (%) (=60.0) Nagelkerke R 2 X 2 df p X 2 df p River Hydrology 73.0 13.0 0.256 7.792 1 0.005 6.448 7 0.488

Table 3.23: Logistic regression model (stepwise) statistics, considering the reduced river hydrology model.

Predictor Variables β SE β Wald's X 2 df p eβ (odds ratio) Predictor: River Hydrology Constant 1.667 0.65 6.597 1 0.100 -

Q bf -0.076 0.33 5.34 1 0.021 0.927

Table 3.24: Logistic regression model (stepwise) variable statistics, considering the reduced river hydrology model.

3.4.3.4 Modelling River Sensitivity to Burrowing

Rearranging the equations in the model allowed for a tool to be developed to allow the prediction of the density of crayfish required in a reach before burrowing was initiated, so that the potential susceptibility of uninvaded rivers could be quantified:

푇퐶퐷 = (−100 퐴퐵 + 242 퐺128 + 푀퐶)⁄826 (Equation 3.15)

where TCD is the threshold crayfish density (CPUE+1) for burrows to be constructed,

AB is the angle of the riverbank (°), G128 is the proportion of substrate grains > 128mm (%); and MC is the model confidence interval (see Table 3.26).

Regression Method: Enter Stepwise Prediction Power Change

Global: PCA 26.5 26.5 0.0

Crayfish Population Characteristics - - -

River Hydrology 15.7 13.0 2.7

River Morphology 11.1 11.1 0.0

Riverbank Morphology 19.4 19.4 0.0

Riverbed Substrate 16.3 16.3 0.0

Sediment Mobility - - -

Vegetation Characteristics - - -

Water Chemistry 21.8 21.8 0.0

Table 3.25: Differences between global and restricted logistic regression models.

Model Input Value Confidence

0.00 5969 0.10 6069 0.20 6169 0.30 6269 0.40 6369 0.49 6459 0.51 6479 0.60 6569 0.70 6669 0.80 6769 0.90 6869

1.00 6969

Table 3.26: Model confidence intervals for use in Equation 3.15. Model confidence intervals correspond to the results of logistic regression modelling. Confidence values predict the presence (>0.5) or absence (<0.5) of burrows. A value of 0.00 corresponds to the model predicting an absence of burrows with 100% confidence, and a value of 1.00 corresponds to the model predicting a presence of burrows with 100% confidence.

This was undertaken for all surveyed rivers using MC = 0.51 to predict the population density threshold required for burrowing to occur. The River Windrush, which was excluded from analyses due to the absence of crayfish, was included. Modelling results were compared with the observed population density for each studied river (Table 3.27)

CD TCD CD - TCD Burrows Present Allershaw Burn 4.0 -0.3 4.3 Bentley Brook 5.3 2.8 2.5 Bookill Gill Beck Lower 2.0 5.5 -3.5 Bookill Gill Beck Mille 7.3 4.1 3.2 Bookill Gill Beck Upper 10.7 6.5 4.2 Broadmead Brook 1.7 0.8 0.9 Chess 8.0 -1.8 9.8 Churn 4.3 5.0 -0.7 Clyde at Elvanfoot 2.3 0.6 1.7 Clyde at Allershaw Islands 6.3 4.0 2.3 Eye Brook 7.3 -1.4 8.7 Gaddesby Brook at Gaddesby 12.6 2.7 9.9 Greet 6.3 0.4 5.9 Nene 2.0 0.0 2.0 River Ouzel Upper 2.0 0.4 1.6 River Ouzel Lower 1.7 1.9 -0.2 Potwell Dyke 1.3 -0.3 1.6 Thames 10.0 7.8 2.2 Torkington Brook Lower 1.0 0.1 0.9 Torkington Brook Upper 1.0 1.8 -0.8 Tove 5.3 -0.4 5.7 Gaddesby Brook at Twyford 10.3 2.6 7.7 Welland 6.0 0.1 5.9

Burrows Absent Alwent Beck 8.3 13.0 -4.7 Balder 3.0 15.7 -12.7 Clyde at Hapturnell Burn 1.3 11.3 -10.0 Clyde's Burn Clyde 4.7 5.7 -1.0 Colne Upper 1.0 3.4 -2.4 Colne Lower 1.0 4.9 -3.9 Deepdale Beck 12.7 14.6 -1.9 Etherow 1.7 10.6 -8.9 Hamps 2.7 3.7 -1.0 Hapturnell Burn 1.3 2.1 -0.8 Lance Beck 1.0 5.7 -4.7 Mochdre Brook 1.0 9.1 -8.1 Scur Beck 1.0 2.3 -1.3 Torridge Lower 1.0 -0.3 1.3 Torridge Upper 1.0 0.7 0.3 Windrush 1.0 3.1 -2.1

Table 3.27: The density of crayfish required to initiate burrowing in each surveyed river.

CD is the observed crayfish density (CPUE +1), TCD is the required crayfish

density for burrowing to be initiated, and CD – TCD is the difference between the observed and required values.

3.4.3.3 Linear Regression

Significant models were found for both BD and MD from both enter and stepwise regression techniques. All reported models can be accepted as valid because:

(i) the individual associations between the predictor variables are linear, significant, and greater than 0.3; (ii) the residuals are normally distributed, with no extreme outliers (Mahalanobis Distance probability, p > 0.001); (iii) no multicollinearity was recorded between variables within the model; and (iv) the residuals are homoscedastic.

Considering the ideal models, the global model, constructed using variables informed by PCA, could significantly predict BD and MD (F = 2.798, p = 0.05 and F = 3.767, p = 0.017 respectively), and explained 52.8% of the variance in BD and 65.3% of the variance in MD 2 between rivers (r ; Table 3.28; Table 3.29). Hydrological variables (log10U60) and riverbed substrate variables (log10D10) were the most important contributors to the global models. 2 Significant models were also found for BD and MD considering sediment mobility (r = 0.652 and 0.651), river hydrology (r2 = 0.626 and 0.725); crayfish population characteristics (r2 = 0.388 and 0.283), riverbed substrate (r2 = 0.290 and 0.209); and river morphology (r2 = 0.190 and 0.252 respectively). No significant models were found for riverbank morphology, vegetation characteristics, or water chemistry.

Considering the operational models, the global model, constructed using PCA informed variables, could significantly predict BD and MD (F = 16.817, p < 0.0005 and F = 25.059, p < 0.0005 respectively; Figure 3.11a, 3.11b), and explained 46.8% of the variance in BD and 2 62.3% of the variance in MD between rivers (r ; Table 3.30; Table 3.31). Hydrological variables

(log10U60) and riverbed substrate variables (log10D10) were the only contributors to the global models. This is a reduction of model fit by 0.30 (r2) from the ‘enter’ regression model (Table 3.32).

There was no change in models considering MD for any restricted model. Considering BD, there was no change in models considering river hydrology, river morphology, or sediment mobility. Significant models were developed for crayfish population characteristics (r2 = 0.347) and riverbed substrate (r2 = 0.212), which is a reduction in r2 of 0.041 and 0.078 respectively (Table 3.30).

r 2 Determined Variables Equation F p

√B D = 0.192 - 0.284 log 10 D 10 + 0.006 τ0 + 0.001 A B - 0.366 log 10 U 60 + 0.001 B P + 0.002 B FS

where log 10 D 10 is the 10th percentile of Riverbed Grainsize (mm, finer than, log 10 transformed), τ is Bankfull Reach Averaged Boundary Shear Stress at low flow (N m -2 ); 0.528 Global: PCA 0 2.798 0.050 o -1 A B is the angle of the riverbank ( ), log 10 U 60 is Channel Flow Velocity at low flow (m s ;

log 10 transformed); B P is the mass of bank material classified as pebbles (%), and B FS is the mass of bank material classified as fine sand (%)

√B D = 0.607 – 0.235 log 10 τ g – 0.003 τ g bf -2 0.651 Sediment Mobility where log 10 τ g is Grain Stress at low flow (N m ; log 10 transformed), and τ g bf is Grain 16.817 <0.0005 Stress at Bankfull (N m -2 )

√B D = 0.494 – 0.346 log 10 U bed – 0.176 U bf -1 0.626 River Hydrology where log 10 U 60 is Channel Flow Velocity at low flow (m s ; log 10 transformed), and Ubf 15.934 <0.0005 is Channel Flow Velocity at Bankfull (m s -1 )

√B D = -0.976 + 0.029 S CM + 0.005 TI 0.388 Crayfish Population Characteristics 4.430 0.032 where S CM is Crayfish Size (males; CL; mm) and TI is time since invasion (years)

√B D = 0.781 – 0.001 D 90 - 0.030 HA

0.290 Riverbed Substrate where D90 is the 90th percentile of Riverbed Grainsize (mm, finer than) and HA is Hide 5.666 0.027 Availability (%) √B = 0.118 – 0.194 log S 0.190 River Morphology D 10 4.685 0.043 where log 10 S is Slope (log 10 transformed)

Riverbank Morphology No significant models - -

Vegetation Characteristics No significant models - -

Water Chemistry No significant models - -

-1 Table 3.28: Predictive multiple linear regression models (enter) for BD (burrows m riverbank). All stated models were significant (p < 0.05), and met the assumptions of regression modelling.

r 2 Determined Variables Equation F p

√M D = - 0.117 - 0.365 log 10 D 10 + 0.0.14 τ 0 + 0.000 A B - 0.566 log 10 U 60 + 0.002 C D + 0.004 B P + 0.006 B FS

where log 10 D 10 is the 10th percentile of Riverbed Grainsize (mm, finer than, log 10 -2 transformed), τ 0 is Bankfull Reach Averaged Boundary Shear Stress at low flow (N m ); 0.653 Global: PCA o -1 3.767 0.017 A B is the angle of the riverbank ( ), log 10 U 60 is Channel Flow Velocity at low flow (m s ;

log 10 transformed); C D is crayfish density (CPUE+1), B P is the mass of bank material

classified as pebbles (%); and B FS is the mass of bank material classified as fine sand (%)

√M D = 0.449 - 0.488 log 10 U bed - 0.269 U bf -1 0.725 River Hydrology where U bed is Flow at the Riverbed at Low Flow (m s ), and U bf is Flow Velocity at 25.059 <0.0005 Bankfull (m s -1 )

√M D = 0.613 - 0.301 log 10 tg - 0.003 tgbf

0.651 Sediment Mobility where log 10 τ g is Grain Stress at low flow (N; log 10 transformed), and τ g bf is Grain Stress 16.801 <0.0005 at Bankfull (N)

Crayfish Population √M D = -1.732 + 0.048 S C 0.283 5.922 0.028 Characteristics where S C is Crayfish Size (CL, mm)

√M D = -0.087 - 0.285 log 10 S 0.252 River Morphology 6.739 0.017 where S is Slope

√M D = 0.761 - 0.003 D 90 0.209 Riverbed Substrate 5.536 0.028 where D90 is the 90th percentile of Riverbed Grainsize (mm, finer than)

- Riverbank Morphology No significant models - -

Vegetation - No significant models - - Characteristics

- Water Chemistry No significant models - -

-1 Table 3.29: Predictive multiple linear regression models (enter) for MD (kg m riverbank). All stated models were significant (p < 0.05), and met the assumptions of regression modelling.

Figure 3.11: Relationships between computed regression variables and crayfish burrow density and the mass of sediment directly excavated by crayfish burrowing considering the application models.

r 2 Determined Variables Equation F p

√B D = 0.607 – 0.235 log 10 τ g – 0.003 τ g bf -2 0.651 Sediment Mobility where log 10 τ g is Grain Stress at low flow (N m ; log 10 transformed), and 16.817 <0.0005 -2 τ g bf is Grain Stress at Bankfull (N m )

√B D = 0.494 – 0.346 log 10 U bed – 0.176 U bf -1 0.626 River Hydrology where log 10 U 60 is Channel Flow Velocity at low flow (m s ; log 10 15.934 <0.0005 -1 transformed), and Ubf is Channel Flow Velocity at Bankfull (m s )

√B D = 0.454 - 0.329 log 10 U 60 - 0.280 log 10 D 10

where log 10 U 60 is Channel Flow Velocity at 60% of flow depth in low flow 0.468 Global: PCA -1 8.342 0.003 (m s ; log10 transformed), and log 10 D 10 is the 10th percentile of Riverbed

Grainsize (mm, finer than; log 10 transformed)pebbles (%) Crayfish Population √B = -0.964 + 0.031 S 0.347 D CM 7.964 0.013 Characteristics where S CM is Crayfish Size (males; CL; mm) √B = 0.699 – 0.002 D 0.212 Riverbed Substrate D 90 5.666 0.027 where D90 is the 90th percentile of Riverbed Grainsize (mm, finer than) √B = 0.118 – 0.194 log S 0.190 River Morphology D 10 4.685 0.043 where log 10 S is Slope (log 10 transformed)

- Riverbank Morphology No significant models - -

Vegetation - No significant models - - Characteristics

- Water Chemistry No significant models - -

-1 Table 3.30: Predictive multiple linear regression models (stepwise) for BD (burrows m riverbank). All stated models were significant (p < 0.05), and met the assumptions of regression modelling.

100

r 2 Determined Variables Equation F p

√SB = 0.449 - 0.488 log 10 U bed - 0.269 U bf 0.725 River Hydrology -1 25.059 <0.0005 where Ubed is Flow at the Riverbed at Low Flow (m s ), and Ubf is Flow Velocity at Bankfull (m s -1 )

√SB = 0.419 - 0.094 log 10 D 10 – 0.001 t0 bf + 0.003 C D - 0.403 log 10 U 60 - 0.001 B P

where log 10 D 10 is the 10th percentile of Riverbed Grainsize (mm, finer than,

0.679 PCA log 10 transformed), t0bf is Bankfull Reach Averaged Boundary Shear Stress 6.332 0.002 -1 at low flow, t0 (N); log 10 U 60 is Channel Flow Velocity at low flow (m s ;

log 10 transformed), and B P is the mass of bank material classified as pebbles (%)

√SB = 0.655 - 0.301 log 10 D x - 0.003 tgbf 0.651 Sediment Mobility 16.801 <0.0005 where log 10 D x is Grain Size Entrained in Flow at low flow (mm), and tg bf is Bankfull Grain Stress (N)

√SB = -1.732 + 0.048 S C Crayfish Population 0.283 5.922 0.028 Dynamics where S C is Crayfish Size (CL, mm)

√SB = -0.087 - 0.285 log 10 S 0.252 River Morphology 6.739 0.017 where S is Slope

√SB = 0.761 - 0.003 D90 0.209 Riverbed Substrate 5.536 0.028 where D90 is the 90th percentile of Riverbed Grainsize (mm, finer than)

- Riverbank Morphology No significant models - -

- Vegetation Dynamics No significant models - -

- Water Chemistry No significant models - -

-1 Table 3.31: Predictive multiple linear regression models (stepwise) for MD (kg m riverbank). All stated models were significant (p < 0.05), and met the assumptions of regression modelling.

101

Burrow Metric Used: MD BD

Regression Method: Enter Stepwise Enter Stepwise r 2 r 2 Change in r 2 r 2 r 2 Change in r 2 Global: PCA 0.653 0.623 0.030 0.528 0.468 0.060

Crayfish Population Characteristics 0.283 0.283 0.000 0.388 0.347 0.041

River Hydrology 0.725 0.725 0.000 0.626 0.626 0.000

River Morphology 0.252 0.252 0.000 0.190 0.190 0.000

Riverbank Morphology ------

Riverbed Substrate 0.209 0.209 0.000 0.290 0.212 0.078

Sediment Mobility 0.651 0.651 0.000 0.651 0.651 0.000

Vegetation Characteristics ------

Water Chemistry ------

Table 3.32: Change in r2 values between enter and stepwise regression.

3.5 Discussion

This is the first multi-catchment study of signal crayfish burrows. The results demonstrate the geographical extent of burrowing activity, with burrows recorded in 60% of surveyed reaches, in both upland and lowland rivers across a large part of Great Britain. Significant associations were observed between burrow characteristics and restricted hydrological, morphological, and biological variables, which hint at the factors that drive burrowing behaviour. For the first time, empirical models have been developed that use abiotic and biotic variables to predict the presence and magnitude of burrowing and associated sediment loading by excavation.

3.5.1 The Geomorphological Significance of Burrow Characteristics

3.5.1.1 Differences in Crayfish Burrow Characteristics

Across 23 impacted reaches, 1152 signal crayfish burrows were surveyed and provide the first multi-catchment quantification of burrow dimensions, density, spacing and displaced sediment mass. This sample is representative of a large range of burrowing conditions as reflected in the wide geographical coverage of the survey. Variability in burrow size was high (Table 3.14), with some differences in burrow characteristics between sites (Table 3.15; Table 3.16) indicating that crayfish burrowing is not homogeneous, but an activity that varies in detail from

102 site to site in response to multiple factors (see below). Burrow characteristics cannot be assumed to be similar between rivers (cf. Faller et al. 2016) and variations in burrow dimensions and densities confirm the geomorphological importance of modelling the mass of sediment excavated from burrows independently of count densities.

3.5.1.2 Vertical Distribution of Burrows

Observations about the spatial distribution of burrows at the riverbank scale are also relevant for understanding the geomorphological implications of burrowing. Burrow surveys were conducted during base flow conditions, and 45% of burrows were still submerged. This suggests that burrows were primarily constructed towards the bottom of riverbanks, which has geomorphological and geotechnical consequences, because a greater weight of sediment is above the burrows, and may promote mass failures more readily. Previous research quantifying sediment recruitment in lowland rivers has only considered burrows above the waterline, which was recognised as a conservative approach (Faller et al. 2016). If as many as 45% of burrows were unrecorded, the true mass of sediment excavated from crayfish burrows may be up to twice the 3 t km-1 estimated for the Thames catchment by Faller et al. (2016). The prevalence of burrows low in the bank may be driven by the duration of flows that are maintained at this level compared with reduced flow persistence at higher stages, but this requires quantitative analysis of vertical burrow distribution relative to flow stage, in future work. It is notable that, to date, modelling of burrow-related bank failures has focused on mammal burrows located toward the top of levees and banks (Taccari and van der Meij 2016a; Borgatti et al. 2017; Saghaee et al. 2017). The processes and outcomes might be very different for burrows located lower in the bank such that crayfish burrows will require new and dedicated modelling efforts. Further consideration of this issue is presented in Chapter 7.

3.5.1.3 Sediment Recruitment from Excavated Burrows

Reach-averaged and local densities were consistent with previous single catchment studies (Table 3.33), which suggest that the 23 sampled rivers are representative of the range and extremes found throughout Great Britain, and so the estimates of the mass of sediment recruited through direct sediment excavation from burrows are likely representative of rates observed throughout invaded rivers. At the reach scale, which includes non-burrowed river lengths, burrows recruited between 0.01 t km-1 and 4.14 t km-1 of bank material (mean = 0.93 t km-1). When only burrowed river lengths were considered, the corresponding range was 0.93 t km-1 to 14.30 t km-1 with a mean of 4.25 t km-1 (Table 3.17). These figures are consistent with

103 previous estimates of 3 t km-1 (maximum 15 t km-1; Faller et al. 2016) but increase the robustness of previous estimates due to the wider geographical and morphological range of rivers surveyed. However, without knowing other quantities in the sediment budgets of these rivers, it is not possible to put these figures into context and, for example, express this contribution relative to sediment delivery from upstream. This is discussed further in Chapter 6.5.6.

Mean Burrow Density Maximum Burrow Study Location (burrows m-1) Density (burrows m-1)

Current Study UK-wide 3.8 7.1 Faller et al. 2016 R. Thames tributaries 1.5 6.0 Guan 1994 R. Great Ouse 3.2 5.6 Guan and Wiles 1997 R. Great Ouse - 21.0 Stanton 2004 Gaddesby Brook 2.4 14.0 R. Greet 2.1 6.5

Table 3.33: Reported burrow densities from previous published studies.

3.5.1.4 Fine Sediment Loading by Burrows

Significant positive associations were observed between MD and bed fine sediments (√MD and -1 log10D25 r = -0.441, p = 0.018). At rivers where MD > 0.7 kg m , the D25 did not exceed 2 mm, and all sediment size fractions were finer in burrowed rivers than in rivers where burrows were not present (p < 0.01; Table 3.18). It is possible either that burrowing is a response to high fines content (there are no large clasts to use for shelter), or that burrowing is a cause of fine sediment loading via recruitment of floodplain sediments (Harvey et al. 2011; Faller et al. 2016; Rice et al. 2016). Communication with local stakeholders (landowners, angling clubs, and catchment managers) reported that fine sediment was not present at Gaddesby Brook, the River Chess; Broadmead Brook, and the River Thames prior to crayfish invasion, where it is now substantial (Figure 3.12), so the more likely cause of the association is that burrowing drives high fines content. This is further discussed in Chapter 6.5.6.

104

Figure 3.12: Sedimentation of the bed of a channel between the River Gade and the River Chess. Channel substrate previously consisted of fine gravels, with heavy siltation reported in recent years.

3.5.2 Modelling Burrow Presence and Absence: Ideal Models

Logistic regression was undertaken to model crayfish burrow presence or absence. This was successful, with the ‘enter’ global model explaining 68.3% of the variance in the data (Nagelkerke R2; p < 0.0005), and predicting burrow presence correctly 86.5% of the time, a 26.5% improvement on the null model (Table 3.21). This is a substantial improvement on previous attempts. Faller et al. (2016) employed a similar methodology, using a PCA informed analysis to predict the presence of burrows between reaches. However, no significant differences between burrowed and non-burrowed reaches were recorded, and the logistic regression models did not provide significant outputs. The models presented here therefore provide the first tools that can be used to predict the presence or absence of crayfish burrows in rivers where signal crayfish are present.

Variables included in the predictive model were crayfish population density (β = 0.826), riverbank angle (β = 0.100), and the proportion of substrate grains > 128 mm (β = -0.242). This suggests that burrows are constructed when crayfish population density exceeds the availability of alternative shelters if bank conditions are suitable. Reaches where crayfish density is low, riverbed grainsize is large, or banks are not steep, are less likely to be burrowed. The development of this model confirms that consideration of biological as well as environmental variables was valuable and improved the predictive power of the presence/absence model.

105

Of the three variables in this model, both bank angle and the proportion of grains greater than 128mm are quantitative, and so are consistent between users. However, crayfish trapping is not a robust quantitative measure (Brown and Brewis 1978; Byrne et al. 1999; Gladman et al. 2010), with catch per unit effort (CPUE) varying between season, the style of trap, the number of traps used, and the bait employed. The model currently predicts crayfish burrows to occur at values >0.51. However, due to the known variability in crayfish trapping results, a higher model threshold should be practically applied (e.g. 0.8) to account for this uncertainty.

The presence or absence of burrows was also modelled using different categories of variables (e.g. only hydrological, only substrate, only biological) to create a tool of value to managers where available data are limited or cannot be obtained (e.g. crayfish trapping not permitted). These models also provide some insights into the drivers of crayfish burrowing by systematically isolating factors. Significant models were created for water chemistry (R2 = 0.508), riverbank morphology (R2 = 0.315), riverbed substrate (R2 = 0.385), river hydrology (R2 = 0.332), and river morphology (R2 = 0.162). With the exception of water chemistry, these models were weak and explained less than 40% of the variation in burrow presence, and thus predicted fewer than 80% of cases correctly (Table 3.21). In this case their value as operational tools is limited and inferior to the full global model. Strong associations with water chemistry were unexpected. Previous research has shown water temperature to be significantly related with crayfish behaviour (Johnson et al. 2014), but only conductivity (positive) and pH (negative) significantly contributed to this model. Conductivity is known to affect crayfish distribution (Welsh and Loughman 2015), but signal crayfish were not considered in that study, and there were no consistent trends between species. Relationships between crayfish burrowing and water chemistry are therefore an important line of future research.

Although they produced weak models, with large overestimations of burrow presence (Figure 3.8), consideration of the variables that were included in the final river morphology, hydrology and bank morphology models is instructive. River width was negatively associated with burrow presence. This may be because larger rivers have a greater carrying capacity; that is, a greater surface area for crayfish to disperse over, and so a larger population size of crayfish is required to reach a density that motivates burrowing. The river morphology model predicts that reaches less than 8.9m wide will have burrows. However, smaller upland streams in the sample typically had a greater coarse sediment contribution, which limited the need for burrowing. While this is accounted for in the global model via sediment variables, it is not considered in

106 the morphology-only model, hence its overprediction of burrowing. Likewise, the negative association with river hydrology parameters predicted burrows in rivers of relatively low discharge and flow velocity, but in upland streams coarse sediment limits burrowing so the hydrology-only model over-predicts burrow presence. The bank morphology variables probably over predict burrowing because there are many rivers where banks are susceptible to being burrowed, but drivers including crayfish population density are not in place. As signal crayfish continue to increase their geographical range and populations continue to grow throughout British river systems (Chadwick 2019), this suggests that burrowing is likely to increase.

The global and restricted models consider burrowing at the reach scale. However, burrows were clustered, occupying only 12.0% of bank length in impacted areas. Once it has been established where burrowing is likely to occur, understanding where within a reach burrowing is likely to occur may be important for directing management or predicting precise impacts. Faller et al. (2016) developed a significant logistic regression model for the presence or absence of burrows at the bank scale, which accounted for 21% of the variability in burrow presence using categorical data. It is likely that this could be improved upon using quantitative data but the short distance (25 m) between cross sections used here means that independence cannot be guaranteed, and so these data are not suitable for creating a bank-scale model. However, as a proof of concept a model was created using the same PCA variables used for reach-scale logistic regression modelling (Table 3.11), which was statistically significant (X2 = 76.594; p < 0.0005), explained 63.5% of the variance in burrow presence at the bank scale (Nagelkerke R2), and correctly predicted 83.5% of cases (Table 3.34; Table 3.35: Figure 3.13). This is therefore an important area of future research.

Omnibus Tests of Predictive Increase Model Summary Hosmer and Lamashow Test Determined Predictor Variables From Null Model Model Coefficients (=60.0) Nagelkerke r 2 X 2 df p X 2 df p Bank Scale Model 23.5 0.635 76.594 7 <0.0005 6.999 8 0.537

Table 3.34: Model statistics of the bank scale model presented in section 3.5.2.

107

Predictor Variables β SE β Wald's X 2 df p eβ (odds ratio) Bank Scale Model Constant 0.221 1.12 0.039 1 0.843 -

Log 10 U 60 -0.193 0.66 0.084 1 0.771 0.825 Grains >128mm -0.065 0.03 3.817 1 0.051 0.937

Log 10 D 10 -2.703 0.93 8.476 1 0.004 0.067

U bf -0.654 3.53 3.430 1 0.064 0.520 Bank Pebble Content -0.015 0.02 1.020 1 0.313 0.985 Bank Angle 0.060 0.02 15.045 1 <0.0005 1.061 Crayfish Population Density 0.221 1.12 0.039 1 0.843 1.248

Table 3.35: Model input variable loadings and statistics of the bank scale model presented in section 3.5.2.

Figure 3.13: Proof-of-concept logistic regression model for bank-scale prediction of burrow presence or absence in the crayfish-infested river reaches.

108

3.5.2.1 Modelling River Sensitivity to Burrowing

Logistic regression equations were rearranged to calculate the density of crayfish required to initiate burrowing given known river bank angles and substrate grain size, which was applied to the surveyed rivers (Table 3.27). Results demonstrate that rivers with coarse sediment and shallow bank angles were less vulnerable to burrowing by crayfish, with the River Balder and Deepdale Beck both modelled to require crayfish population densities exceeding 14 CPUE(+1). Other rivers, such as the River Torridge, were modelled to be highly susceptible to burrowing, with burrowing predicted to occur even before crayfish population density reaches a detectable level by trapping. There are currently no known effective control or eradication methods for signal crayfish, but this tool would enable potential eradication or management methods to be prioritised to reaches where burrowing is of greatest potential risk. Modelling these population thresholds throughout British rivers using GIS software is therefore an important avenue of future research.

3.5.3 Modelling Burrow Density: Ideal Models

-1 Multiple linear regression was undertaken to model the density of crayfish burrows (BD, m ) -1 and the mass of sediment excavated by burrowing activities (MD, t km ) if crayfish burrows were present. Of the rivers where burrows were recorded, just 30% of sites exceeded a density of 0.5 burrows m-1, but these recruited 69% of all excavated sediment (Figure 3.14). This underlines the importance of modelling not just the presence or absence of burrows, but also the density of burrows in order to understand resultant sediment dynamics.

Significant global models were constructed for both BD and MD (Table 3.26; Table 3.27), which 2 explained 52.8% and 65.3% of the variance in BD and MD respectively (r ; p = 0.05 and p < 0.0005). This is the first time that densities of burrows, and the mass of sediment excavated by burrowing have been successfully modelled, and provides an important river management tool. The most important contributors to both models were substrate clast size and the flow velocity of the stream. Whilst crayfish population density and the characteristics of the riverbank were important in determining if crayfish burrow, these were less important in determining the extent to which crayfish will burrow. Together these variables only increased 2 the r of the BD and MD models by just 0.060 and 0.030 respectively, and were entirely excluded in the reduced models. Variables included in the reduced models were channel flow velocity at base flow (log10 transformed; -0.329 and -0.504), sediment grain size (log10D10; -0.280 and

-0.338), and bed shear stress at base flow (not considered and 0.013) for BD and MD,

109 respectively. Flow velocity and D10 are easily measured and understood parameters, which increases the accessibility of the models to users without access to pre-existing datasets or specialised equipment.

Figure 3.14: The frequency and relative sediment contribution of surveyed crayfish burrow density. High burrow density was uncommon (30% > 0.5 burrows m-1), but recruited the majority (69%) of sediment.

Significant models were also found when sediment mobility, river hydrology, crayfish population characteristics, river substrate, and river morphology were considered exclusively. Whilst the global selection of variables using PCA analysis yielded the strongest models for determining if crayfish burrow, restricted sediment mobility (r2 = 0.651 and 0.651) and river hydrology (r2 = 0.626 and 0.725) models both explained a greater proportion of the variance in BD and MD, respectively. This demonstrates that whilst PCA is an excellent technique for capturing the variability of the whole system, a smaller set of variables (hydrology) equally well explained the observed variability and allowed the key drivers to be identified. This shows the importance of the approach to analysis taken here, where considering both the whole system and then restricted sets of variables can help identify the drivers of a dependent variable. It may

110 be the case that whilst the PCA attempted to capture the full variability of the system, much of the variation observed is controlled by the hydrological regime. River hydrology is also easier to calculate with a greater degree of accuracy and precision, and provides the best models for predicting BD and MD:

√퐵퐷 = 0.494 − 0.346 log10 푈푏푒푑 − 0.176 푈푏푓 (Equation 3.16)

√푀퐷 = 0.449 − 0.488 log10 푈푏푒푑 − 0.269 푈푏푓 (Equation 3.17)

3.5.4 Ideal Versus Operational Models

As well as ideal models considering all variables, reduced operational models were also created for increased application potential where all variables may not be available or feasibly attainable (Table 3.28; Table 3.29). The reduced models were successful in predicting burrow presence / absence, burrow density, and the mass of sediment excavated, with significant stepwise models found in all cases where ‘enter’ models were significant. These models were the same in most instances, with only 5 of the 18 reduced models being different to the computed ‘enter’ models. The difference in variables included in the five models are presented in Table 3.36. There was little difference in the predictive power between the reduced and full models, with the greatest reduction in performance being a 2.7% drop in predictive power in logistic modelling, and a reduction of 0.078 in r2 value in a linear regression. The maximum number of variables in any reduced model was three, which shows that modelling the distribution of crayfish burrows, and the mass of sediment excavated, is a task that can be undertaken by catchment managers considering few variables which do not require specialist techniques to derive them. The removal of variables from these five ideal models yielded a minimal compromise in model performance (Table 3.36), and so the operational models considered here, and in general the operational methodology employed, should be considered in practical applications.

111

Predictive Power Variables Retrained Variables Removed Enter Stepwise Difference

Burrow Presence: River Hydrology Q bf log 10 U 60 ; U bf ; log 10 Q 15.7 13.0 2.7

M D : Global: PCA log 10 U 60 ; log 10 D 10 ; t0 A B ; C D ; B P ; B FS 0.653 0.623 0.030

B D : Global: PCA log 10 U 60 ; log 10 D 10 τ 0 ; A B ; B P ; B FS 0.528 0.568 0.060

B D : Crayfish Population Characteristics S CM TI 0.388 0.347 0.041

B D : Riverbed Substrate D 90 HA 0.290 0.212 0.078

Table 3.36: Retained and removed variables from differing enter and stepwise regression models. Stated predictive power for burrow presence models are the increase in correct predictions from the null model (60.0%), and predictive power for 2 mass density (MD) and burrow density (BD) models are r values.

3.5.5 Model Contributions and Univariate Associations of Environmental Variables

This chapter has so far considered the geomorphic significance of burrows, and has evaluated the strength, utility and importance of the predictive models. Here, the constituent variables are evaluated as contributors to the models, but also independently regarding their direct associations with crayfish burrow metrics to understand the biotic and abiotic factors associated with differences in crayfish burrowing activity.

3.5.5.1 River Morphology

When considering river morphology variables, river width was the only significant contributor to presence or absence models. As noted above it overpredicted the presence of burrows by 82%, probably because larger rivers have a greater population carrying capacity. This association with width was unexpected, because Faller et al. (2016) found that rivers where burrows were present were wider than those where burrows were absent. In addition, burrows are widespread in lowland rivers (e.g. Guan 1994; Harvey et al. 2011; Rice et al. 2014; Faller et al. 2016) but have not been reported in upland catchments, even though upland streams are expected to be smaller on average. Comparatively small rivers were sampled in the current study due to the requirement of being a wadable depth, but this did not unduly limit the range of widths sampled (River Width: mean = 6.1 m; minimum = 0.95 m, Bookill Gill Beck; maximum = 18.3 m, River Clyde). Burrows were found in both lowland rivers and upland streams, with no difference in slope or altitude between rivers where burrows were present and absent. However, when burrows were present, altitude and channel slope were both negatively

112 associated with the mass of sediment excavated from burrows (Figure 3.15) suggesting that burrowing is less common in streams at higher altitude with steeper slopes.

Figure 3.15: Associations between (a) river altitude and (b) river slope with the mass of sediment excavated by burrowing. Both associations are significant (p = 0.018 and p = 0.009 respectively).

Reach widths increased with decreasing slope and altitude (p < 0.0005), and burrowed reaches were significantly narrower than non-burrowed reaches. This suggests that there is a “goldilocks” sweet spot for burrowing in those small lowland channels dominated by steep cohesive banks, that facilitate burrowing, and have fine substrate material, that limits the number of available shelters, but where width is nevertheless small, so that there is insufficient area for crayfish to avoid each other without seeking shelter in burrows. This is consistent with the small lowland streams where crayfish burrows have been historically reported, such as

Gaddesby Brook (channel width = 2.7 m, D65 = 14.4 mm, bank angle = 49.0°, bank material = 41.0% silt and clay; Stanton 2004), and the River Nene (channel width = 3.1 m,

D65 = 13.0 mm, bank angle = 65.0°, bank material = 61.3% silt and clay; Rice et al. 2014; Rice et al. 2016). This is important, as small tributary streams consist of a greater stream length than large rivers, and so the potential burrowing habitat, and thus potential fine sediment delivery to river systems, is considerably greater than if burrowing were restricted to larger channels.

Whilst the difference was significant only at p = 0.10, it is also interesting that land use scores were higher for reaches with burrows compared to reaches without burrows indicating that greater human modification of land use may promote burrowing. British rivers are heavily channelised (Brookes et al. 1983), and this may increase the risk that crayfish burrowing poses. Anecdotally, landowners at one field site were concerned that building work downstream of

113 the study reach, which had reshaped the bank to create a footpath using earth and sand, may be susceptible to burrowing by signal crayfish. These results support these concerns. The propensity of crayfish to burrow into modified riverbanks or built structures, including flood defences, is an important avenue of future research.

3.5.5.2 River Hydrology

Flow variables were important in determining the density of burrows and mass of sediment excavated in rivers. It was expected that a relationship between crayfish burrowing and flow velocity would be present, because crayfish use deep pools (Light 2003) and large rocks (Bubb et al. 2002; Bubb et al. 2004) to shelter from high flows, and burrows may be constructed to serve a similar purpose. Despite the strong associations between crayfish behaviour and flow characteristics, no research has previously established a link between flow velocity with crayfish burrowing activity. Surface flow types were qualitatively considered (e.g. ‘smooth’ or ‘rippled’) by Faller et al. (2016), which was applied to multivariate modelling, although no univariate relationships between burrowing and flow characteristics were reported.

In the current survey, significant univariate and multivariate relationships with flow velocity were observed. Although bankfull discharge was the only significant flow variable predicting burrow presence and absence, all flow velocities were significantly negatively correlated with both BD and MD (p < 0.01). River hydrology was also the best predictor of BD and MD in multiple linear regression models (r2 = 0.626 and 0.725 respectively), and outperformed the global model in both cases. The variables most important for both models were flow velocity at base flow conditions, and flow velocity at bankfull conditions. From a model application perspective, these are both simple to obtain. Velocity at base flow can be easily measured directly, and when bankfull width, bankfull depth, and slope are known, bankfull velocity can be estimated using Manning’s equation or an equivalent (Equation 3.9). A similar approach could be used to calculate base flow velocities where river access is not possible using gauging station data with the application of depth-discharge rating curves, making it applicable to catchment managers who are unable to physically sample the rivers in question.

Stronger associations were observed between flow velocities and MD than with BD, which suggests that flow velocity is important for controlling the size of burrows that are constructed. As flow velocity varies temporally, particularly in small upland catchments which were considered here (Figure 3.16), and whilst flow velocity at base flow was significantly negatively associated with burrow depth (Figure 3.17a), there was no association between

114 burrow depth and flow velocity at bankfull (Figure 3.17b). Association between burrow metrics and base flow conditions, but not bankfull conditions, suggests that burrowing occurs predominantly during base flow events as opposed to as a reaction to extreme hydrological conditions.

Figure 3.16: Hapturnell Burn at (a) base flow and (b) bankfull flow, approximately 16 hours apart following a heavy rain event. Colleague for scale.

Figure 3.17: Associations between burrow depth and flow velocity considering (a) the riverbed at base flow (p = 0.026) and (b) 0.6 depth at bankfull (p = 0.563).

Crayfish may have faced physical difficulties in constructing burrows during higher flows. Burrowing is an energy-expensive process (Meysman et al. 2006), and this will likely be compounded by the high energy required to maintain position in high flow velocities (Mather and Stein 1993). Burrowing during higher flow velocity events may have stopped at the earliest opportunity to shelter from the high flow. Decreasing burrow size with increased flow velocity

115 may also reflect the value associated with the burrow by the crayfish. The value of a resource to an organism can be determined by the energy required to construct the resource (Parker 1974). Exerting the lowest possible energy to construct a functional, but not luxurious, burrow may reflect the short term nature of the burrow – both in terms of its function to shelter the crayfish from a pulsed flood event, and the increased likelihood of the burrow collapsing from the increased hydraulic stress associated with high flow conditions. This idea is further explored in an experimental setting in Chapter 4.

The suggestion that burrows were predominantly constructed during base flow conditions, which has important implications for sediment dynamics. Burrows were constructed into cohesive banks and recruited a substantial mass of fine sediment directly into the channel; if burrowing occurs predominantly during base flow conditions, it will likely result in the greatest difference in sediment entrainment compared to if burrows are absent. In small lowland channels where burrows were most prolific, the cohesive nature of the sediment combined with the low slope and flow velocity result in very little sediment erosion or entrainment, especially during base flow conditions. This was demonstrated by Rice et al. (2016) at the River Nene, UK, where the activity of crayfish suspended an insignificant mass of sediment during flood periods, but contributed at least 32%, and potentially up to 72.6%, to monthly base flow suspended sediment loads. This sediment was hypothesised to have been recruited through the bioturbation of sediment already in the channel. These results show that burrowing may not only have recruited the sediment into the channel, but could have also played an active role in immediately suspending the sediment during base flow. However, the associations between flow velocity and burrow construction extracted from these analyses are based on uncontrolled data from a field study. Chapter 4 will explore and discuss these associations further in a systematic experimental manner.

3.5.5.3 Riverbed Substrate

2 Riverbed substrate characteristics produced poor predictive models of BD and MD (r = 0.290 and 0.209 respectively). However, in univariate analysis, there were strong associations between grain size and burrow metrics, with sediment significantly finer in rivers where burrows were present, and percentiles were negatively associated with both BD and MD. The greatest difference in substrate metrics between burrowed and non-burrowed rivers was the percentage of grains greater than 90 mm, and coarse sediment grains were more strongly associated with BD and MD than fine grains (Figure 3.9).

116

This is the first quantification of the relationship between crayfish burrowing and riverbed sediment. Substrate has been considered an important factor for crayfish distribution, although results of previous field studies are inconsistent. Field surveys in the River Wharfe, UK, found crayfish using rocks coarser than 64 mm as refuges (Peay and Rogers 1999) but, in the Iberian Peninsula, signal crayfish were found to be negatively correlated with the presence of boulders (Vedia et al. 2017). The results from this field survey show that coarse sediment availability is an important driver in crayfish burrowing behaviour and an important avenue of future research. This is investigated and discussed in detail in Chapter 5.

Crayfish have also been observed using other forms of cover, such as large woody material (Walter 2012) and macrophyte stands (Johnson 2014; Veida et al. 2017), and so a ‘shelter availability’ score was created. This metric represented the width of the channel where an alternative shelter, such as a large rock, large woody material, and macrophyte stands were present. However, this metric had a lower association with crayfish burrowing than considering only sediment grain size. This suggests that mineral sediment, which contributed to the shelter availability score, is the most predominantly used shelter by signal crayfish in the studied rivers. This also shows that the area of riverbed where shelters are available is not as important as the type and quantity of shelters; less burrowing was observed in the presence of more large cobbles, even if the spatial distribution of the shelters was not increased. This was unexpected, as crayfish are highly aggressive (Issa et al. 1999; Goessmann et al. 2000; Kravitz and Huber 2003), and signal crayfish will strongly defend the ownership of a shelter from others (Ranta and Lindstrom 1993; Guan 1994; Bergman and Moore 2003). Under typical conditions, crayfish exhibit social distancing (Daws et al. 2011), however mesocosm experiments demonstrated that narrow-clawed crayfish (Pontastacus leptodactylus) cluster together when stressed by flow, presumably to reduce the impact of shear stresses acting on single crayfish (Ion et al. 2020). Experiments have also shown that exposure to increased thermal conditions reduced crayfish aggression (Gherardi et al. 2013), and this reduction in intraspecific-crayfish competition may enable signal crayfish to utilise shelters at close proximity during periods of environmental stress events.

3.5.5.4 Sediment Mobility

Sediment mobility variables were the most strongly correlated with both BD and MD in univariate analysis, in particular grain stress (r = -0.684 and -0.689 respectively), which was the most strongly correlated variable with both BD and MD of all examined. τ0 was not

117 associated with BD or MD in univariate analysis, but was included as a significant variable in global multivariate analysis. Models consisting exclusively of sediment mobility variables 2 2 strongly predicted both BD (r = 0.651) and MD (r = 0.651), with the restricted sediment mobility model being the strongest of all BD models.

Sediment mobility may be strongly related with burrowing for two reasons. First, sediment mobility is strongly related to both hydrology and sediment grain size, which are both independently important in controlling the presence and extent of crayfish burrows. Sediment mobility may therefore represent both of these variable groups but in a more parsimonious way that the analysis identifies. Secondly, it may be the interaction between flow velocity and sediment that is important to consider. Crayfish burrows were constructed in sites where an abundance of coarse-grained materials were available, but these coarse grains may be mobilised by the flow, and so would not provide ample shelter to crayfish during high flow events. In upland catchments, shelter stability has been identified as a key factor for coping with disturbances (Parvulescu et al. 2016), with crayfish favouring stable shelters. Flow velocity varies spatially, with high hydraulic stress mitigated in streams through the size, arrangement, shape, and sorting of substrate particles that provide refugia for benthos (Biggs et al. 1997, Matthaei et al. 1999; Matthaei et al. 2000). It is possible that when these refugia become unstable, burrowing becomes favourable. Systems where shelter availability exceeded crayfish population density during base flow conditions may therefore be burrowed due to reduced shelter availability during high flow conditions.

There was a marked difference in the size of burrows constructed between rivers with calculated bankfull grain stresses above and below 40 N m-2 (Figure 3.18). The difference in burrow size observed between the high and low grain stress rivers (Figure 3.18b) is likely to follow the same rationale as with flow velocity (section 5.4.2); in high stress environments, burrows were constructed to be functional to shelter the crayfish with immediate effect, which only needed to be large enough to defend the crayfish from the environmental stress, as opposed to from other crayfish, with no energy spent on extending a burrow beyond the immediate necessary requirements.

118

Figure 3.18: Associations between burrow size and grain stress at (a) base flow and (b) bankfull flow. Linear associations are present in both cases (r = -0.664, p = 0.001; and r = -0.437, p = 0.042 respectively), and a difference in burrow size above and below grain stresses of 40 N m-2 at bankfull is evident.

3.5.5.5 Riverbank Morphology

Significant differences in riverbank morphology were found between reaches with and without burrows. Rivers where burrows were present had finer bank material, which consisted of significantly fewer pebbles and a significantly greater mass of silt and clay (Table 3.18). The negative association between bank pebble mass and crayfish burrowing is consistent with burrows constructed by the Chinese Mitten Crab (Eriocheir sinensis), which are constructed in soft bank sediments, but were found to terminate above a gravelly layer in sediments in intertidal streams in San Francisco Bay (Rudnick et al. 2005).

Rivers where burrows were present also had steeper banks (62.3°) than where burrows were absent (48.2°; p = 0.052). Riverbank angle significantly improved the logistic regression models, and was included in both the global model and restricted bank morphology model. The restricted bank morphology model accounted for 16.2% of variance in crayfish burrows presence (Nagelkerke R2), which is similar to Faller et al.’s (2016) model (21%), but is considerably easier to apply in management situations, requiring only two variables compared to 15 for the model proposed by Faller et al. (2016). Bank morphology properties were important in determining if crayfish burrow, but there was only one significant relationship between bank morphology variables and burrow density metrics (silt and clay and MD; r = 0.361, p = 0.049), and no significant multiple linear regression models were found. This

119 supports the argument that as the bank is simply the arena for the burrowing, it will determine if crayfish can burrow, but does not determine the extent to which burrowing occurs.

3.5.5.6 Water Chemistry

Water chemistry variables were not included in the global predictive models and did not produce significant linear models when only water chemistry variables were considered. However, only water chemistry variables contributed to significant models for predicting the presence or absence of burrows. Conductivity was significantly correlated with altitude (r = -0.694, p < 0.0005), but pH was not a strong covariate of any other variable. Conductivity has been shown to be associated with crayfish distribution (Welsh and Loughman 2015) whereas pH has been similar between previously surveyed sites, with no significant associations between pH and crayfish distribution detected (Welsh and Loughman 2015; Loughman et al. 2016). Relationships between crayfish behaviour and water chemistry are therefore an important line of research.

3.5.5.7 Vegetation Characteristics

Vegetation was not an important variable for predicting burrow presence or density. Vegetation variables were not included in global models, and no significant models considering only vegetation variables were found. The distance of shading by riparian vegetation was the only significantly related variable in univariate analysis, which was positively associated with BD. Many streams that had high densities of crayfish burrows were heavily vegetated (Figure 3.19), with burrows being found beneath deep thickets of vegetation (Figure 3.19d), although this may be an association with the small lowland streams where burrows were most prolific. There was no association between in-channel macrophytes and crayfish burrowing. Crayfish have previously been observed using macrophyte stands for shelter (Johnson et al. 2014), however the lack of association here, with only one positive relationship observed with bank face vegetation, suggests that macrophytes do not serve as an alternative shelter to burrowing, and that bank face vegetation does not prevent crayfish from burrowing into the riverbank. These results are in contrast to Faller et al. (2016), who found burrows to be more prevalent on bare banks than vegetated banks.

120

Figure 3.19: Riparian vegetation coverage was very high at (a) Potwell Dyke, (b) Gaddesby Brook (Gaddesby); (c) River Ouzel (Lower), and (d) River Tove. In channel vegetation (Ranunculus aquatilis) was also extensive at the River Ouzel, and burrows were found in abundance underneath the riparian vegetation at all four sites.

3.5.5.8 Crayfish Population Characteristics

In contrast to previous studies, this survey characterised 38 sites across 29 rivers, and thus a broad range of biological variability. Crayfish population density (CPUE+1) ranged from 1.0 at nine sites to 12.7 at Deepdale Beck, and time since invasion varied from 37 years at Broadmead Brook, less than 1 year in the upper Clyde at Allershaw Burn, the River Clyde at Hapturnell Burn, and Hapturnell Burn (W. Yeomans pers. comm.).

Despite the large gradient of crayfish characteristics, crayfish population variables were not different between burrowed and non-burrowed reaches. Crayfish population density was greater in burrowed than non-burrowed rivers (5.2 and 2.8 CPUE (+1) respectively; p = 0.052).

121

Whilst no significant reduced logistic regression model was found, population density was a significant contributor to the global logistic regression model for presence or absence when considered with sediment size. This suggests that rivers have a ‘carrying capacity’ (availability of large substrate grains), and when the population density exceeds this, burrowing occurs to create new shelters to increase the carrying capacity, if the banks are steep enough. This is supported by the calculated differences between observed and threshold crayfish densities required for initiation of burrowing estimated by Equation 3.15. While there were no associations between observed crayfish density and burrow densities (BD r = 0.217, p = 0.172;

MD r = 0.339, p = 0.057), the difference between observed and threshold crayfish densities

(Table 3.33) was positively associated with burrow densities (BD r = 0.505, p = 0.007; MD r = 0.595, p = 0.001; Figure 3.20). This suggests that crayfish excluded from existing shelters construct burrows as alternative shelters, and supports this ‘carrying capacity’ theory. Understanding this interactive effect between population density and shelter availability is therefore a key question of research, and is addressed and discussed in Chapter 5.

The size of male crayfish was positively associated with both BD and MD, but there was no relationship with female crayfish size. This suggests that male crayfish constructed more burrows, which is inconsistent with previous research, and other results from this survey. Mesocosm experiments have shown that male virile crayfish (Faxonius virilis) outcompeted females for burrow occupancy in 73% of cases, which was attributed to their larger size

(Bovbjerg 1953). Although crayfish size was not associated with BD (r = 0.217, p = 0.172), there was a positive association between size and MD (r = 0.459, p = 0.018), suggesting an association between crayfish size and burrow size. Mesocosm experiments by Guan (1994) suggested no relationship between crayfish and burrow size (Guan 1994), but here a positive association was observed (r = 0.627, p = 0.001). Crayfish size was positively associated with river width (r = 0.567, p = 0.004) and negatively associated with the percentage of riverbed grains coarser than 128mm (r = -0.692, p < 0.0005) which were both important in determining the presence and density of crayfish burrows. Independently, crayfish variables were a stronger predictor in modelling burrow density than river substrate or river morphology, but weaker in predicting burrow presence. This therefore suggests that when burrowing occurred, crayfish size was important in determining the size of burrows.

122

Figure 3.20: Associations between burrow density and crayfish population density, -1 considering (a and b) the density of burrows (BD; burrows m ), (c and d) density -1 excavated sediment (MD; kg m ); (a and c) observed crayfish population density (CPUE +1), and (b and d) the difference between observed crayfish population density and modelled threshold crayfish density.

The time since initial invasion was also an important variable to consider, which was positively associated with BD and MD (r = 0.389, p = 0.033; and r = 0.355, p = 0.048 respectively), and significantly improved the restricted linear regression models. There were no associations between the time since initial invasion and any other crayfish variable, and so this association is not considered to be a covariate of another variable. However, whilst the association is positive, there is substantial variation around the trend (Figure 3.21), which demonstrates that the potential geomorphic impacts of crayfish to a river system may be immediate upon invasion, and that burrows may have geomorphic impacts for a considerable period of time. The variation observed may be due to erosional processes; burrows have been associated with accelerated bank erosion and collapse (Sibley 2000; Barbaresi et al. 2004; West 2010; Telegraph 2016; Faller et al. 2016), but the longevity of crayfish burrows has not been quantified, and is an important avenue of future research.

123

Figure 3.21: Association between time since invasion and burrow density.

The associations and models presented here that include crayfish population variables have an underlying assumption that the response of crayfish to environmental variables is consistent between populations. Signal crayfish were introduced to Europe at multiple times from multiple source locations (Kouba et al. 2014), and there is speculation that subspecies may have been introduced, with at least six genetically distinct lines identified in the UK alone (Petrusek et al. 2017). Crayfish are able to learn new behaviours in response to environmental cues (Ion et al. 2020), and it may be that burrowing is a learned behaviour. Establishing whether burrowing differs between populations, and whether it is a learned or an innate behaviour is an important question that is examined and discussed in Chapter 5.

3.5.6 Discussion Summary

Burrows constructed by signal crayfish were found throughout the UK, in both upland and lowland rivers, and greatly expanded on the known range, size and densities of crayfish burrows. Reach scale burrow densities as high as 1.13 burrows m-1 were observed recruiting up to 4.14 kg m-1 of sediment. Equivalent values at the bank scale were 7.12 burrows m-1, and 14.30 kg m-1.Significant differences were observed between rivers from the burrow scale to the reach scale, which demonstrated the geomorphological importance of modelling the mass of sediment excavated from burrows independently of count densities. 45% of burrows were located below the waterline, and so a greater mass of sediment is likely to be recruited directly by burrows, but also through an increased likelihood of mass failure, than previously considered.

124

Models for predicting the presence or absence of burrows and their density were successfully developed. The best of these successfully predicted crayfish burrow presence, and explained 68.3% of the variance in crayfish burrow presence. Rearrangement of these models allowed for the crayfish carrying capacity of a system to be calculated before burrowing was initiated. Further, these models also predicted the density of burrows, and the mass of sediment excavated by burrowing, which accounted for 62.6% and 72.5% of variance respectively, and provide an important and applicable tool for river management. A novel approach considering global and reduced models was employed to understand the major contributors to the models from a complex environment. The most appropriate of these models for considering burrow presence or absence, the population density of crayfish required to initiate burrowing, the count density of burrows and the mass of sediment recruited directly from burrows are summarised in Table 3.37. These could be applied to large geospatial datasets to model crayfish burrowing and its associated impacts on a national scale.

Target Variable Best Model Model Statistics Summary

Probability of P(B) = -5.969 + 0.826 C + 0.100 A - 0.242 G R 2 = 0.683 Burrows D B 128

Threshold Crayfish TC = (-100 A + 242 G + MC) / 826 R 2 = 0.683 Density D B 128

Burrow Density r 2 = 0.626 √B = 0.494 – 0.346 log U – 0.176 U (burrows m -1 ) D 10 bed bf F = 15.934; p < 0.0005

Mass of Sediment r 2 = 0.725 Excavated by √M D = 0.449 – 0.488 log 10 U bed – 0.269 U bf -1 F = 25.059; p < 0.0005 Burrows (kg m )

Table 3.37: Summary table of best models advised for predicting the probability of burrow occurrence, the threshold crayfish density at which burrowing is initiated, burrow density (when burrowing occurs), and the mass of sediment excavated by burrows (when burrowing occurs).

Together, the univariate associations between independent variables and burrow characteristics, along with the analysis of the performance of alternative regression models,

125 suggest that burrowing is driven by population density exceeding the availability of alternative mineral sediment shelters, and if banks are steep and consist of fine sediment. Differences in burrow sizes were attributed to baseflow flow velocities and associated sediment mobility, which suggested the stability of in stream shelters during high flow events were an important driver for crayfish burrowing behaviour.

The greatest burrow densities were associated with small, low energy lowland steams. This is important because biological energy is likely to present the greatest differential between sediment yields compared to uninvaded conditions. Further, as crayfish invasion and extreme flows are projected to increase in severity, geomorphic problems associated with crayfish burrowing may become more pronounced in the future.

126

Chapter 4

The effects of flow velocity on signal crayfish burrowing and associated sediment dynamics

127

4.1 Introduction

Signal crayfish prefer to inhabit areas of low flow velocity in both laboratory (Salkonen et al. 2010) and field (Rice et al. 2012) conditions, and field investigations have shown that discharge and flow depth are the most significant drivers of activity levels in signal crayfish (Johnson et al. 2014). Signal crayfish use deep pools (Light 2003) and large rocks (Bubb et al. 2002; Bubb et al. 2004) to shelter from high flow velocities, and burrows may be constructed to serve a similar purpose. Field surveys (Chapter 3) found significant differences in burrow presence between streams of high and low flow velocity, and a significant negative association between burrow sizes and flow velocity. However, flow velocity may be a proxy for large grain size in upland streams. The presence of large cobbles was shown in Chapter 3 to have a significant effect on the propensity of crayfish to construct burrows, and it may be the presence of a larger sediment size as opposed to the flow velocity which drives crayfish burrowing dynamics.

The erosion of fine sediment from riverbanks can be caused by both biotic and abiotic erosional processes:

(i) Burrowing. The direct (purposeful) excavation of sediment from riverbanks by signal crayfish when constructing burrows (Guan 1994; Harvey et al. 2014). (ii) Bioturbation. The collateral biotic entrainment of sediment by signal crayfish caused by other activities such as foraging and fighting (Harvey et al. 2014; Rice et al. 2014). (iii)Diffuse Erosion. The direct abiotic entrainment of sediment from a riverbank, driven by excess shear stress at the bank face (Buffington and Montgomery 1997; Simon et al. 2000). (iv) Mass Failure. The shearing of comparatively large sections of bank material as one eroded unit, driven by changes in mass balance (Fredlund et al. 1978; Fredlund and Rahardjo 1993).

It has been hypothesised (Guan 1994; Harvey et al. 2011; Harvey et al. 2019), and reported anecdotally (West 2010; Telegraph 2016) and qualitatively (Arce and Dieguez-Uribeondo 2015), that crayfish burrowing leads to accelerated diffuse erosion and mass failure. Harvey et al. (2019) provide a comprehensive review and hypothesise that burrowing may alter these erosional mechanisms through (i) geotechnical and hydrological effects, (ii) hydraulic effects, and (iii) geochemical and biological effects (see Chapter 2.2.3). However, the relative and combined contributions of biological and geophysical energy are poorly understood in fluvial

128 geomorphology (Philips 2009; Rice et al. 2016), and understanding the mechanisms of sediment recruitment under different flow velocities is therefore valuable beyond the specifics of crayfish zoogeomorphology. While geophysical energy (e.g. Hooke 1979; Arulanandan et al. 1980; Rinaldi and Casagli 1999; Simon et al. 2000; Mahalder et al. 2016) and crayfish activity (Harvey et al. 2014; Faller et al. 2016) can both recruit sediment to river systems, their relative contributions in places where crayfish are active, have not been investigated. Cooper et al. (2016) observed a nocturnal increase in suspended sediment concentration of up to 76% at the River Blackwater, Norfolk, and Rice et al. 2014 showed that crayfish activity suspended up to 47% of monthly base flow sediments in the River Nene, UK. However, these studies have reported the biological and geophysical contributions to sediment recruitment independently, without evaluating the role of crayfish in accelerating abiotic rates. Chapter 6 uses field measurements to address this question and demonstrates quantitative associations between crayfish burrows and both diffuse erosion and bank mass failure. However, the nature of the fieldwork meant it was not possible to investigate the mechanisms and controls in detail.

The flume investigation reported in this chapter complements the understanding gained from the field investigation (Chapter 6) by simulating bank erosion by crayfish in a controlled environment. Experiments were undertaken in a laboratory flume to establish how flow velocity influences crayfish burrowing behaviour, and to assess the interactive effects of biological (burrowing, bioturbation) and geophysical (diffuse and mass erosion) processes in the erosion of simulated river banks.

4.2 Aims

This chapter aims to address aspects of objectives 2 (to examine the biotic and abiotic drivers of signal crayfish burrowing behaviour) and 4 (to quantify the effect of crayfish burrowing on volumes of riverbank erosion and investigate the relative roles of direct sediment input from crayfish burrows, accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows in recruiting sediment to invaded river channels) through a series of flume experiments. Specifically, the flume experiments outlined below address two research hypotheses:

(i) that the propensity of crayfish to burrow is reduced with increasing flow velocity; and

129

(ii) that biologically driven erosion and geophysical erosion, which have previously been reported as independent, interact resulting in greater erosion of bank material than either process alone.

The hypotheses were investigated by addressing the following four questions:

(i) Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent? (ii) Does increasing flow velocity reduce crayfish burrowing behaviour? (iii) Does crayfish burrowing change the rate of erosion of non-burrowed sediment? (iv) Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates?

4.3 Methods

Flume experiments were conducted in the presence of crayfish, the presence of crayfish burrows, and the absence of crayfish and crayfish burrows to investigate their relative contribution to erosion rates under increasing flow velocity conditions.

4.3.1 Physical Set Up

An Armfield S6 flume was divided into 14 identical mesocosms (0.4 m x 0.3 m x 0.45 m; Figure 4.1; Figure 4.2) using 25 mm wire gauze. 80mm wide wooden plinths were installed at the top of the wire gauze to prevent crayfish from climbing into neighbouring mesocosms.

Fixed gravel bed boards (D10 7 mm; D50 15 mm; D90 22 mm) were installed throughout the full length of the flume for substrate. An 80mm deep clay bank was created by installing expanded bentonite clay pellets into plastic containers (390 mm x 260 mm x 80 mm), which were affixed onto the left-hand bank of each mesocosm in a vertical position, sensu Harvey et al. (2014). A false wall was constructed upstream and downstream of the experimental sections, on the same side as the clay banks to maintain a consistent channel width. Mesocosms were situated 2.4 m downstream of the flume inlet (3.6 m downstream of the pump outlet), to allow for flow straightening and realistic flow development.

A standard brick (220 mm x 110 mm x 70 mm) was placed on the right-hand side of the channel and elevated at the downstream end by 50 mm, so that it could be used as a shelter for crayfish (Figure 4.1). Bricks, elevated in the same manner, were placed upstream and downstream of

130 the mesocosms to provide flow consistency in each mesocosm. Bricks were used rather than river cobbles because their standard shape provided directly comparable shelter availability and effects on flow in each mesocosm. James et al. (2016) found that crayfish showed no clear preference for different shelter types, including natural versus non-natural shelters. Therefore, the use of bricks over natural shelters is unlikely to affect behaviour and, indeed, crayfish actively used the bricks throughout the experiments.

4.3.2 Mesocosm Treatments

There were three experimental treatments, each of which was replicated fourteen times, at three test flow velocities. Treatments consisted of:

i) Control. A smooth bank with no crayfish present; ii) Artificial Burrows. A smooth bank with an artificial ‘burrow’ constructed into it, with no crayfish present; and iii) Crayfish. A smooth bank with two signal crayfish.

Treatment (i) ‘Control’ measured baseline values for sediment erosion to establish a control under three flow conditions, where erosion was a result of exclusively physical process, and predominantly hydraulic energy (with likely relatively small erosional processes from recirculated sediment), and was not subject to any additional zoogeomorphic erosional forces.

Treatment (ii) ‘Artificial Burrows’ were used to investigate the influence of burrows on bank erosion in the absence of crayfish, under three flow conditions. This treatment allowed an evaluation of how crayfish activity (burrow construction) might affect abiotic erosion. Artificial ‘burrows’ were constructed by removing a cylinder of clay (mean = 70 mm diameter) from the centre of the bank face prior to the start of an experimental run.

Treatment (iii) ‘Crayfish’ was used to investigate how burrowing behaviour and bank erosion changed under three flow conditions. These experimental runs included two crayfish with one available hide, because this resulted in the most consistent burrowing behaviour in the still water mesocosm experiments (Chapter 5). It was hypothesised that an increase in flow velocity would reduce burrowing activity, and so the most consistent and prolific burrowing conditions from still water experiments were selected.

131

Figure 4.1: A plan view (top) and a cross-sectional view looking downstream (bottom) of the physical set up used in the flume experiments. Flow velocities were recorded centrally longitudinally and equidistant between the brick and bank laterally (point A; top), and vertically, a reading was taken at 60% of the flow depth and 0.03 m from the bed (point B and C respectively; bottom).

132

Figure 4.2: The physical set up of the flume, before (top) and during (bottom) an experimental run. View looking upstream in both cases.

Individual treatments were allocated to mesocosms to avoid possible interactions between crayfish in neighbouring mesocosms. Rice et al. (2012) reported that caged signal crayfish spent 20% of the time climbing cage walls, and so it was highly likely that interactions would occur if crayfish were housed in neighbouring mesocosms. Due to size limitations of the flume, data had to be collected over three cycles for each flow velocity to obtain n = 14 pseudo-replicates for each treatment (Figure 4.3).

Figure 4.3: Schematic of flume from above. Three cycles at each flow velocity were undertaken to collect the necessary data.

133

4.3.3 Flow Velocity

Each elected treatment was repeated at three flow velocities (Table 4.1):

i) a low flow velocity, where flow velocity had limited influence on crayfish activity; ii) a medium flow velocity, where crayfish were able to move and function within the flow, but would actively shelter from the flow; and iii) a high flow velocity, where crayfish were not able to successfully move and function within the flow, and would actively shelter from the flow.

30mm Above Bed 60% Flow Depth Flow Velocity Flow Velocity SD (mm s-1) SD (mm s-1) (mm s-1) (mm s-1) Low 76 2.5 86 1.2

Medium 293 5.9 315 5.7

High 566 4.4 590 4.8

Table 4.1: Flow velocities (mm s-1) used in experiments. The locations of recorded velocity recordings are shown in Figure 4.2.

After the banks had been installed, the flume was filled to a depth of 0.21 m (SD 0.013 m), and the flow increased to the experimental conditions over a period of ten minutes. Average velocity over a 30 second period was measured using a Valeport Flow Meter in every mesocosm as indicated in Figure 4.1, at 0.6 depth (representing a mean water column velocity (Biggs et al. 1998)), and 0.03 m above the bed, which represents the flow velocities that the crayfish were directly exposed to.

Neither flow velocity at the river bed or in the channel were normally distributed (Shapiro- Wilk: W = 0.944, p < 0.0005; and W = 0.914, p < 0.0005 respectively), and so non-parametric ANOVA (Kruskal-Wallis (H) tests) were used to confirm a significant difference in flow velocities between the low and medium, and medium and high flow velocities (p < 0.005 in all cases).

134

4.3.4 Crayfish

Twenty-five adult signal crayfish (10 male, 15 female, 47 – 65 mm carapace length (CL), mean 54.2 mm CL) were collected from Glossop Brook, Greater Manchester, UK (SK 0096 9528) for use in the experiments. Whilst there are no burrows present at Glossop Brook, the burrowing behaviour of this population is shown to be consistent with that of crayfish from streams where burrows are common (Chapter 5), and so the crayfish used can be considered representative for these experiments. Crayfish were kept in two holding tanks (1.3 m x 0.7 m x 0.6 m), which were aerated, filtered, and filled to a depth of 0.2 m with 180 l of dechlorinated tap water. Holding tanks were subject to the same lighting regime as the flume and had sufficient refuges (plastic tubes) for all crayfish. Only crayfish larger than 47 mm CL were used in the experiments to ensure that crayfish did not move through the mesh that separated neighbouring mesocosms. For each experimental run, two crayfish were selected at random from the population for use in any ‘crayfish’ treatment mesocosm. Size and sex were recorded, and only crayfish with intact legs, antenna, and chelae were used.

4.3.5 Experimental Procedure

For each mesocosm where the ‘crayfish’ treatment was employed, two crayfish were independently introduced on top of the brick on the right-hand side of the channel. Previous similar experiments have released crayfish 50 mm above the bed (Johnson 2010), however this was not a plausible method due to the high flow velocities used in some experiments. Crayfish were not fed during the experimental runs but were returned to holding tanks for at least 3 days and fed between experimental runs.

Experiments were undertaken from February to April 2019. Each run lasted 24 hours during which crayfish experienced a natural lighting regime: artificial illumination (overhead, lab ceiling, strip lights) for 9 h (08:00 – 17:00) followed by 15 h of darkness (17:00 – 08:00). At the end of each run, the water was drained from the flume channel (at a flow velocity not exceeding the maximum velocity achieved during the experimental run), and the crayfish and banks were removed. Any burrows that had been constructed were characterised and measured in detail, as were any bank collapses.

135

4.3.6 Quantifying Sediment Erosion

For all treatments, the plastic containers holding the clay banks were placed horizontally and filled to the brim with water prior to being installed into the flume. The volume of water needed to fill the containers (Vw_start) was measured using a measuring cylinder, accurate to 20 ml. This was repeated at the end of each experiment (Vw_end), to calculate the total displaced volume of sediment that had occurred throughout the run as:

푉푡표푡 = 푉푤_푠푡푎푟푡 − 푉푤_푒푛푑 (Equation 4.1)

4.3.6.1 Burrows

Burrows constructed by crayfish were visually identified (Figure 4.4a, 4.4b). The volume of sediment excavated in burrows was calculated geometrically, by measuring the width and height of the burrow, both at the entrance and back of the burrow to calculate the volume of sediment removed, and the depth of the burrow. Mean burrow width and height were used to calculate the cross-sectional area of the burrow. Burrows were treated as elliptical cylinders, and the volume of sediment excavated was calculated by:

ACS= π (W̅ /2 H̅/2) (Equation 4.2)

where ACS is burrow cross-sectional area, W is the burrow width, and H is the entrance height.

푉퐵 = 퐴퐶푆 퐿 (Equation 4.3)

where VB is burrow volume, and L is the length of the burrow.

In the event of multiple burrows being constructed, the volumes of the burrows were summed to report the total volume of sediment excavated as burrows in a single run.

4.3.6.2 Bank Collapses

Bank collapses were visually identified as areas of erosion that were inconsistent with the general retreat of the bank face and did not resemble the shape of a crayfish burrow (Figure 4.4c; 4.4d). Bank collapses were measured geometrically, by measuring the dimensions of the collapse at the bank face, and at the back of the collapse using a tape measure. The shape of

136 the collapse was classified (e.g. cuboid, triangular prism), and the volume of sediment lost was calculated as VC.

4.3.6.3 Diffuse Erosion

Diffuse erosion (VD) was calculated as the total sediment loss (Vtot) minus the volume of any burrows and collapses:

푉퐷 = 푉푡표푡 − (푉퐵 + 푉퐶) (Equation 4.4)

The mass of sediment recruited by a method other than burrowing (VNB) is:

푉푁퐵 = 푉퐷 + 푉퐶 (Equation 4.5)

Figure 4.4: Identification of burrows and collapses. Crayfish burrows (a and b) were identified as rounded holes, which were typically seen to be occupied by crayfish prior to the end of the experimental runs. Collapses (c and d) were

137

identified as areas of sediment removed that were inconsistent to the general retreat of the bank face and did not resemble the shape of a crayfish burrow.

4.3.6.4 Conversion from volume to mass

The bentonite clay used in these experiments was calculated to have a bulk density of 1.41 t m-3, and so all volumes were converted to sediment mass (kg) for analysis (Mtot, MB, MR, MC, and

MNB respectively). These mass values were then normalised by the surface area over which -2 they could occur (MB’, MR’, and MC’ (kg m ) respectively); when burrows were artificially created before the start of the experiment, collapses could only occur over the remainder of the surface area, and diffuse erosion could only be measured over the surface area that burrowing or collapse did not affect.

4.3.6.5 Swelling of Clay

Neither the geometric or water displacement method account for the physical expansion of the clay after immersion in water. As a result of clay expansion, some negative values were recorded for artificial burrows and diffuse erosion. Mean values for treatment (i) demonstrated that physical expansion of clay did occur, with negative median recordings for all velocities -2 -2 -2 (MR’ = low: -1.18 kg m ; medium: -0.86 kg m ; high: -2.56 kg m ). The rate of expansion was consistent and was not significantly different across flow velocities (Kruskal-Wallis: H2

2.310, p= 0.315) or cycles (H5 9.441, p= 0.093), and thus will have had a uniform effect across all mesocosms. The rate of clay swelling was not independently tested, and so no uniform value can be applied to all mesocosms to correct for this. However, as expansion recorded in experiments was not different between flow velocities, the uncorrected values presented here allow direct comparison across treatments. As they have not been corrected for clay swelling, they likely underestimate the rate of erosion. However, as these experiments do not seek to present absolute erosion rate estimates, but relative erosion rate estimates between treatments, this is acceptable. Negative readings were converted to a value of zero for statistical analysis, as no net deposition took place, as in Couper et al. (2002).

138

4.3.7 Statistical Analyses

All statistics were performed using SPSS Version 23 (IBM 2015). Tests to assess the distribution of the data using the Shapiro-Wilk test indicated that none were normally distributed (p < 0.005 in all cases). Following Log10(x+1) transformation, the data fulfilled the assumptions required for General Linear Modelling (GLM; Levene’s Test; p > 0.05). GLM was used to analyse the data with Tukey’s HSD post-hoc test for pairwise comparisons.

Spearman’s Rank (rs) was used to examine linear associations between variables. A probability value of α < 0.1 was used for hypothesis testing because:

i) a high variability within runs was expected and observed; ii) non-parametric tests were used; and iii) these experiments are exploring a new area of research and do not seek to establish quantitative relationships, but explore potential underlying processes to help inform field observations.

4.3.7.1 Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent?

GLM was used to examine the effect of flow velocity and crayfish presence on the total mass of sediment excavated (Mtot’ (Log10(x+1))), considering independent main effects and interactions. This was considered for all mesocosms, and also only for mesocosms where burrows were constructed. Considering all mesocosms, only burrowed mesocosms, and only non-burrowed mesocosms separately allowed for the total effects of crayfish to be observed, and to understand whether these effects were dependent on, or could be attributed to, burrow construction.

To ascertain whether the strength of any observed interactions were consistent across flow velocities, a hybrid variable (the net effect of crayfish (CS’)) was determined for each Crayfish mesocosm. For each flow velocity, the median value of total sediment erosion (~Mtot’) measured in the Control treatment was subtracted from the mass of total sediment erosion

(Mtot’) of each Crayfish mesocosm. This resulted in the net difference of sediment recruitment due to the presence of crayfish in each Crayfish mesocosm (CS’):

′ ′ ′ 퐶푆 = 푀푡표푡 (퐶푟푎푦푓𝑖푠ℎ) − ~ 푀푡표푡(퐶표푛푡푟표푙) (Equation 4.6)

139

GLM was used to examine the difference in CS’ between flow velocities with Tukey’s HSD post-hoc test for pairwise comparisons.

4.3.7.2 How does flow velocity influence crayfish burrowing behaviour?

GLM was used to compare the mass of sediment excavated through burrowing (MB’) between the three flow velocities, with Tukey’s HSD post-hoc test for pairwise comparisons. This was considered for all mesocosms, and only mesocosms where burrowing had occurred.

4.3.7.3 Does crayfish burrowing change the rate of erosion of non-burrowed sediment?

GLM, with Tukey’s HSD post-hoc test for pairwise comparisons, was used to examine the effect of flow velocity and the presence of crayfish on the mass of sediment excavated by mechanisms other than burrowing (MNB’ (Log10(x+1))), considering the independent main effects and interactions. This was considered for all mesocosms, for mesocosms where burrows were constructed, and for mesocosms where burrows were not constructed. This permits assessment of the effect of crayfish on collapse and diffuse erosion with and without the presence of burrows to evaluate the effect of burrowing on collapse and diffuse erosion. However, separating mesocosms into those where burrows were constructed and those where burrows were not constructed reduces n in each case. Whilst non-parametric statistics were used, it should be noted that this results has reduced statistical power compared to tests considering all mesocosms.

GLM was used to compare MNB’ between runs where burrowing had occurred, and runs where burrowing had not occurred, at each flow velocity. Spearman’s Rank (rs) was used to correlate

MB’ and MNB’ for all Crayfish mesocosms at each flow velocity, and across all experimental runs to examine linear associations between the mass of sediment excavated directly through burrowing and the mass of sediment derived from non-burrowed sources.

4.3.7.4 Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates?

GLM was used to examine the effect of flow velocity and the presence of artificial burrows on the total mass of sediment excavated (Mtot’ (Log10(x+1))), considering the independent main

140 effects and interactions. GLM was also used to consider the effect of flow velocity in Artificial Burrows mesocosms, with Tukey’s HSD post-hoc test for pairwise comparisons.

4.4 Results

Significant differences were observed in the mass and mechanisms of sediment recruitment between crayfish treatments and between flow velocities. Each of the research questions identified are considered individually below.

4.4.1 Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent?

Flow velocity and the presence of crayfish were both significantly associated with increased sediment recruitment (F2, 78 = 2.537, p = 0.086; Figure 4.5; and F1, 78 = 18.671, p < 0.0005 respectively). There was a significant interaction between the effects of crayfish and flow on the mass of sediment recruited (F2, 78 = 3.357, p = 0.040). The presence of signal crayfish significantly increased mean Mtot’ in the low flow velocity treatment by 307% (F1, 26 = 15.958, p < 0.0005) and by 371% at medium flow velocity (F1, 26 = 11.177, p= 0.003). However, the presence of signal crayfish did not significantly increase Mtot’ in the high flow velocity treatment (F1, 26 = 0.139, p= 0.712; Figure 4.6; Table 4.2).

Figure 4.5: Total sediment eroded (Mtot’) when crayfish were absent on a Control bank, considering mean values, +/- 1 standard error (SEM).

141

Figure 4.6: Total sediment eroded (Mtot’) in the presence and absence of two crayfish, at each considered flow velocity, considering mean values, +/- 1 standard error (SEM), n = 14 for each treatment.

Flow Velocity

Low Medium High

Mean Median Mean Median Mean Median

Control 1.59 0.41 2.89 2.03 6.12 3.53

Crayfish 6.47 4.34 13.60 9.90 6.57 4.75

Difference (C ') 4.88 3.93 10.71 7.87 0.45 1.22 S

-2 Table 4.2: Mass (kg m ) of total sediment eroded (Mtot’) when crayfish were present and absent under each flow velocity.

142

The difference in sediment erosion between Crayfish and Control (CS’) varied across flow velocity treatments (F2, 32 = 3.233, p= 0.053; Figure 4.6; Table 4.2). Tukey’s post-hoc test indicated that more sediment was recruited in the presence of crayfish compared to Control conditions in the medium flow than the high flow velocity treatment (p = 0.068), and that there were no other significant differences in pairwise comparisons. When only burrowed mesocosms were considered, CS’ varied across flow velocity treatments (F2, 19 = 3.930, p = 0.037). Tukey’s test indicated differences in CS’ between both the low (p = 0.050) and medium (p = 0.053) flow velocity, and the high flow velocity treatment, but no difference between the low and medium flow velocity treatment (Figure 4.7; Table 4.3).

Figure 4.7: Net difference upon the introduction of crayfish compared to an absence of

crayfish (CS’) when (a) all runs and (b) only burrowed runs are considered, considering mean values, +/- 1 standard error (SEM).

Flow Velocity

Low Medium High

Mean Median Mean Median Mean Median

All Runs 6.01 3.93 11.56 7.86 3.04 1.22 Only Burrowed Runs 8.85 8.14 8.72 7.46 2.44 1.63

-2 Table 4.3: Observed difference in mass of sediment eroded (Mtot’; kg m ) between Crayfish and Control treatments at different flow velocities.

143

4.4.2 How does flow velocity influence crayfish burrowing behaviour?

Crayfish constructed burrows in more runs at low (eight) and medium flow velocity (nine) than the high flow velocity treatment (five). The dominant crayfish was observed using the brick, with the subordinate crayfish constructing a burrow, or sheltering in the lee of the brick. However, in the high flow velocity treatment, crayfish occasionally cohabited the shelter, or the excluded crayfish sought shelter in the lee of the shelter rather than constructing a burrow (Figure 4.8d).

MB’ decreased with increasing flow velocity when all runs were considered, and burrow size also reduced with flow velocity, but there were no significant differences between flow velocity treatments (All mesocosms: F2, 39 = 1.644, p = 0.206; only burrowed mesocosms: F2, 19 = 0.588, p = 0.565; Figure 4.9; Table 4.4).

Figure 4.8: In the high flow velocities, crayfish immediately opted to use the shelter to escape the flow (a, b, c), often cohabiting. When one crayfish was excluded from the shelter, the subordinate crayfish often sheltered in the lee of the shelter as opposed to constructing a burrow (d).

144

Figure 4.9: Mass of sediment recruited from crayfish burrows (MB’) when (a) all runs and (b) only burrowed runs are considered, considering mean values, +/- 1 standard error (SEM).

Flow Velocity

Low Medium High

Mean Median Mean Median Mean Median

All Runs 2.30 1.33 2.14 1.29 0.94 0.00 Only Burrowed Runs 4.03 3.05 3.33 3.05 2.62 2.25

-2 Table 4.4: Mass (kg m ) of sediment recruited from crayfish burrows (MB’) when all runs and only burrowed runs are considered.

4.4.3 Does crayfish burrowing change the rate of erosion of non-burrowed sediment?

When all runs were considered, the presence of crayfish had a significant effect on mean MNB’ in both low (121%) and medium (246%) flow velocity treatments (F1, 26 = 5.296, p= 0.030 and

F1, 26 = 5.375, p= 0.029 respectively; Figure 4.10a; Table 4.5). However, no significant difference was recorded for the high flow velocity treatment (F1, 26 = 0.250, p= 0.621).

This pattern was also observed when only mesocosms subject to burrowing were considered, and a significant difference was observed for both the low and medium flow velocity treatments

(F1, 20 = 4.805, p= 0.057 and F1, 21 = 6.204, p= 0.021 respectively; Figure 4.10b; Table 4.5);

145 there was no significant difference observed for the high flow velocity treatment (F1, 17 = 0.495, p= 0.491).

When considering only those mesocosms without burrowing by crayfish, the presence of crayfish resulted in an increase in MNB’ for all three flow velocity treatments – at low flow by 57%, at medium flow by 423%, and high flow by 7%. However, none of these differences were significant (F1, 18 = 2.267, p= 0.150; F1, 17 = 2.510, p= 0.132; and F1, 21 = 0.058, p= 0.812 respectively; Figure 4.10c; Table 4.5).

There was no significant correlation between MB’ and MNB’ for all Crayfish mesocosms considering all flow velocities (rs = 0.028, n = 22, p = 0.903; Figure 4.11), and there was no consistent pattern across the three flow velocity treatments (rs = 0.476, n = 8, p = 0.233; rs= -

0.467, n = 9, p = 0.205; and rs= 0.300, n = 5, p = 0.624 respectively). Whilst there were no significant associations, a distinct difference in MNB’ between large and small burrows was observed (Figure 4.11). High variation in non-burrowed sediment was observed when MB’ <4.5 -2 -2 kg m , whereas MB’ >4.5 kg m recruited substantially less sediment from non-burrow sources

(MNB’).

Flow Velocity Low Medium High

Mean Median Mean Median Mean Median

Control 2.06 0.91 3.73 2.03 6.47 3.97 Crayfish (all runs) 4.45 2.61 12.72 8.77 5.63 2.89 Crayfish (only burrowed runs) 5.34 2.81 8.96 7.81 3.34 2.90

Crayfish (only non-burrowed runs) 3.25 2.30 19.50 27.52 6.90 2.44

Table 4.5: Mass (kg m-2) of sediment recruited through mechanisms other than burrowing

(MNB’).

146

Figure 4.10: Mass of sediment recruited through mechanisms other than burrowing (MNB’) in the presence and absence of crayfish under all flow velocities when (a) all runs, (b) only burrowed mesocosms, and (c) only non-burrowed mesocosms were considered, considering mean values, +/- 1 standard error (SEM).

147

Figure 4.11: Association between sediment recruited directly from burrows (MB’) and

sediment recruited from non-burrowed sources (MNB’). Solid line represents a

1:1 ratio, and dashed line shows a distinction in MNB’ between low and high

values of MB’.

4.4.4 Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates?

Artificial burrows were not significantly associated with a change in total sediment recruitment

(Mtot’) under any flow velocity treatment (F2, 78 = 0.087, p = 0.768), although an increase in

Mtot’ was observed when artificial burrows were present for the low and medium flow velocity treatments (Figure 4.12; Table 4.6). Considering only Artificial Burrows, no significant difference was observed in Mtot’ between any flow velocity treatments (F2, 39 = 0.612, p = 0.548). However, there was a significant interaction between artificial burrows and flow velocity (F2, 78 = 2.405, p = 0.097).

148

Flow Velocity

Low Medium High

Mean Median Mean Median Mean Median

Control 1.59 0.41 2.89 2.03 6.12 3.53

Artificial Burrows 2.50 1.90 3.70 4.47 2.81 1.90

-2 Table 4.6: Mass of total sediment eroded (Mtot’, kg m ) between Artificial Burrows and Control treatments at all flow velocities.

-2 Figure 4.12: Mass of total sediment eroded (Mtot’, kg m ) between Artificial Burrows and Control treatments at all considered flow velocities, considering mean values, +/- 1 standard error (SEM).

149

4.4.5 Results Summary

Crayfish burrowed during all flow velocity conditions, but the frequency and size of burrows constructed decreased with increasing flow velocity. For the high flow velocity treatment, utilisation of the shelter was preferred over burrow construction; two crayfish sharing the shelter was observed, which was unexpected. The presence of crayfish significantly increased mean total sediment transport (Mtot’) for the low flow velocity treatment by 307% and by 371% for the medium flow velocity treatment, but not for the high flow velocity treatment. Burrowing increased non-burrowed sediment erosion (MNB’) by 121% and 248% in low and medium flow velocity treatments respectively, but no significant increase in MNB’ was observed for the high flow treatment. Artificial Burrows without crayfish present increased sediment recruitment

(Mtot’) at low and medium flow velocities, but not for the high flow velocity treatment.

4.5 Discussion

During these experiments, signal crayfish presence consistently increased the amount of sediment erosion, confirming crayfishes potential to do geomorphic work, and there were significant effects of flow velocity on their activity. Burrowing was more common and generated more sediment at flows up to approximately 0.30 m s-1 (low and medium treatments), where shelter use by the dominant crayfish led to burrow excavation by the subordinate. However, for the high flow velocity treatment (0.59 m s-1) crayfish burrowed significantly less, excavated smaller burrows and both crayfish sought refuge on the bed, even cohabiting the shelter. Avoiding exposure to the flow was therefore more important than competing for the shelter resource, which is surprising given the reputation of signal crayfish for aggressive interactions.

4.5.1 Is erosion of sediment from simulated riverbanks greater in the presence of crayfish, compared to when crayfish are absent?

The rate of erosion observed in the Control treatments when crayfish were absent increased with flow velocity (Figure 4.5). This was expected because flow velocity is understood to be a dominant control on riverbank erosion (e.g. Lapointe and Carson 1986; Hasegawa 1989), with erosion rates being proportional to near bank excess shear stress, with is proportional to flow velocity. This demonstrates that the processes acting on the simulated riverbanks were consistent with natural systems.

150

At low and medium flow velocities, crayfish significantly increased the mass of sediment recruited (Mtot’; Figure 4.6). This provides further supporting evidence that signal crayfish are an important driver of fine sediment dynamics (Harvey et al. 2011; Rice et al. 2014; Rice et al. 2016; Chapter 6), and in particular in the recruitment of sediment from riverbanks (Harvey et al. 2011; Harvey et al. 2014; Faller et al. 1016; Chapter 6). However, at high flow velocities, signal crayfish did not significantly affect fine sediment recruitment, and burrowed significantly less than at lower flow velocities, with both crayfish seeking shelter to escape exposure to the high flow conditions (Figure 4.6). Cohabitation of the shelter was unexpected. Crayfish are highly aggressive (Issa et al. 1999; Goessmann et al. 2000; Kravitz and Huber 2003), and signal crayfish will strongly defend the ownership of a shelter from other signal crayfish (Ranta and Lindstrom 1993; Guan 1994; Bergman and Moore 2003). This aggressive behaviour was evident and observed during the still-water mesocosm experiments (Chapter 5). However, exposure to unfavourable environmental cues (increased temperature) reduced crayfish aggression in mesocosm experiments conducted by Gherardi et al. (2013), and narrow- clawed crayfish (Pontostacus leptodactylus) have been observed ‘clumping’ communally in increased flow velocities (Ion et al. 2020). The reduced aggression observed here under high flow velocities supports this earlier work, in suggesting that agonistic behaviours are partly dependent on environmental conditions.

4.5.2 How does flow velocity influence crayfish burrowing behaviour?

Significant differences in the mass of burrowed sediment MB’ between the two lower flow velocities and the high flow velocity treatment were observed (Figure 4.9). As well as crayfish burrowing less regularly during the high velocity treatment, burrows were smaller, with MB’ decreasing with increasing flow velocity (Figure 4.9b; Table 4.4). Crayfish have been shown to have lower levels of movement (Maude and Williams 1983; Clark et al. 2008; Salkonen et al. 2010; Johnson et al. 2014), colonisation success (Mathers et al. 2020), and population density (Usio and Townsend 2000) in periods of high flow, and have previously been recorded preferring to occupy low flow areas in field surveys (DiStefano et al. 2003; Flinders and Magoulick 2005) and laboratory experiments (Rice et al. 2012). Crayfish may have faced physical difficulties in constructing burrows during high flow conditions. Burrowing is an energy expensive process (Meysman et al. 2006), so the additional energy required to maintain position during high flow velocities (Mather and Stein 1993) may explain why crayfish avoided burrowing during high flows and instead sought shelter.

151

A negative association between burrow size and flow velocity may also reflect the value associated with the burrow by the crayfish. The value of a resource to an organism can be determined by the energy required to construct the resource (Parker 1974). Exerting the lowest possible energy to construct a functional burrow may reflect the short term nature of the burrow – both in terms of its function to shelter the crayfish from high flows, and the increased likelihood of the burrow collapsing from the increased hydraulic stress associated with those high flow conditions.

Previous studies have found limiting flow velocities for crayfish activity to be lower than the velocities tested here, including for signal crayfish, which are generally more resistant to high flow than other species (Table 4.7). Thus, Salkonen et al. (2010) found that velocities of 0.55 m s-1 were needed to dislodge signal crayfish, lower than the high flow velocity treatment here (0.59 m s-1). Despite this, the tested signal crayfish were still able to construct some burrows in 5 of the 14 high velocity runs. This suggests that the absolute velocity limit for burrowing activity is above 0.59 m s-1. There is some support for this from the field data presented in Chapter 3 where burrows were recorded in streams where higher flow velocities were measured than those tested here (Table 4.8). Of seven field sites with higher flow velocities, burrows were observed at three sites, including the River Clyde (Islands), where a mean flow velocity of 0.74 m s-1 was recorded, with localised recordings exceeding 1.0 m s-1. Of course, it is possible that the snapshot velocities measured at field sites were higher than those during burrow construction, but all field measurements were made during low flow periods.

Flow velocity also varies spatially, with high hydraulic stress mitigated in streams through the size, arrangement, shape, and sorting of substrate, which then provides hydraulic refugia for benthos (Biggs et al. 1997, Matthaei et al. 1999; Matthaei et al. 2000). Whilst the presence of coarse mineral substrate in streams may dampen the effect of flow velocity and hydraulic stress on the behaviour of crayfish, the same negative association between burrow size and flow velocity observed in the current experiments (Figure 4.9b) was observed in the field (Figure 4.13; Chapter 3), with burrow volume and burrow depth both significantly reduced with increasing flow velocities (Figure 4.13). This shows that flow velocity effects crayfish burrowing even in complex environments, and that the processes observed in these experiments are comparable to those in situ.

152

Dislodgement Flow Reference Velocity (mm s-1) Astacus astacus 350 (winter) Salkonen et al. 2010 430 (summer) Salkonen et al. 2010 Cambarus bartonii 390 Maude and Williams 1983 Cambarus robustus 497 Maude and Williams 1983 Creaserinus fodiens 267 Maude and Williams 1983 Faxonius propinquus 347 Maude and Williams 1983 Faxonius rusticus 402 Maude and Williams 1983 Faxonius virilis 284 Maude and Williams 1983 Orconectes immunis 260 Maude and Williams 1983 Orconectes obscurus 298 Maude and Williams 1983 600 Clark et al. 2008 Pacifastacus leniusculus 500 (winter) Salkonen et al. 2010

550 (summer) Salkonen et al. 2010

Table 4.7: Flow required to dislodge crayfish species downstream in previous published experiments.

Velocity at River Velocity at 60% Burrows per 100m Bed (m s-1) Flow Depth (m s-1) Riverbank

Clyde's Burn Clyde 0.39 0.79 0.0 River Clyde (Islands) 0.26 0.74 20.8 River Etherow 0.44 0.73 0.0 River Balder 0.51 0.68 0.0 Hapturnell Burn 0.29 0.66 0.0 River Clyde (Elvanfoot) 0.23 0.62 29.2 Torkington Brook 0.25 0.61 11.2 River Clyde (Hapturnell Burn) 0.16 0.51 0.0 River Welland 0.31 0.51 23.2 River Hamps 0.21 0.46 0.0 River Churn 0.41 0.45 17.0 Bentley Brook 0.31 0.41 14.0

Table 4.8: Sites (12) where mean flow velocity was recorded as greater than the medium flow velocity tested. Seven sites exceeded flow velocities considered here, of which burrows were recorded at three. Burrowed sites are emboldened.

153

Figure 4.13: Association between flow velocity and burrow depth in the field (reproduced from Chapter 3). The association is significant (r = -0.463; p = 0.013).

4.5.3 Does crayfish burrowing change the rate of erosion of non-burrowed sediment?

A significant increase in MNB’ was observed when crayfish were present when all runs were considered (Figure 4.10a), with an increase of 121% and 246% in the presence of crayfish in low and medium flow velocities, respectively. When runs where burrows were constructed and runs where burrows were not constructed were considered separately, the increase of non- burrowed sediment was significant in burrowed runs, but not in runs where crayfish did not construct burrows. This supports the assertion that crayfish burrowing changes the rate of erosion of non-burrowed sediment. Specifically, this suggests that the process of creating and maintaining a burrow drives the recruitment of non-burrowed sediment, and supports the results of Chapter 6, which concluded that burrows are important for driving secondary erosional processes such as mass failures on the riverbank. These results also support the assertion that previous research may have underestimated the mass of sediment that burrows recruit to river systems, which have only quantified sediment excavated directly from burrowing (e.g. Camici et al. 2014; Orlandini et al. 2015; Taccari and van der Meij 2016a; Borgatti et al. 2017; Saghaee et al. 2017).

A difference in MNB’ was observed when burrowed masses were small or large (Figure 4.11).

There was greater variability and higher average amounts of non-burrowed sediment when MB’ -2 -2 <4.5 kg m , but when MB’ >4.5 kg m , substantially less sediment was eroded from non- burrow sources. The likely explanation for this is that successful construction of larger burrows

154 involved crayfish spending more time within the burrow, causing less bioturbation from the bank surface, while smaller burrow volumes tended to be associated with multiple initial attempts at constructing a burrow resulting in greater bioturbation of the bank and the promotion of many small collapses.

Non-burrowed sediments may also be recruited by bioturbation in the absence of any burrow construction. Crayfish recruit sediment through bioturbation associated with foraging, walking and fighting (Angeler et al. 2003; Rodriguez et al. 2003; Statzner and Sagnes 2008; Rice et al. 2014; Harvey et al. 2014). In mesocosms where crayfish burrows were not constructed, the observed difference in non-burrowed sediment erosion may reflect walking on the bank independently of burrowing. The flume used in the current experiments was a recirculating flume, and so turbidity levels were too high to actively monitor crayfish activity; this combined with crayfish burrowing occurring at night meant that crayfish behaviour could not be independently analysed to address such issues. In future, non-recirculating flumes repeating these experiments may be able to isolate the effects of bioturbation and sediment recruited through burrowing to better resolve their roles. Whilst these experiments do not fully show the mechanisms through which non-burrowed sediments were recruited, they explicitly show that burrowing changes the rate of erosion of non-burrowed sediment, and thus prepares the bank for increased levels of erosion by geophysical flows. This provides supporting evidence for the sediment budget presented in Chapter 6, which discusses this association further.

4.5.4 Does the presence of artificial burrows, in the absence of crayfish, change bank erosion rates?

There was no statistical difference in Mtot’ between the Artificial Burrows and Control treatments for any flow velocity treatment (Figure 4.12). Whilst these statistics suggest that the presence of burrows alone does not have a significant influence on sediment recruitment, in the low and medium flow treatments, mean sediment recruitment was greater when artificial burrows were present. It may be that this difference is not significant over a 24 hour period, and when considered over a greater length of time, a significant difference would be observed.

However, for the high flow velocity treatment, Mtot’ was lower in the Artificial Burrows treatment than the Control treatment. This may be because of different processes of erosion acting upon the bank during high flows; more sediment was recruited through the mechanism of collapse with increasing flow velocity (Figure 4.14), which is a stochastic process, and it

155 may be that in the relatively low number of replicates undertaken (n = 14), the number of random collapses has an overriding effect on the result.

Figure 4.14: Sediment recruited from collapses (MC’) in each considered flow velocity in the Control treatment, considering mean values, +/- 1 standard error (SEM).

The experiments undertaken were relatively simple, and only tested for an association between artificial burrows and one variable (flow velocity). Whilst channel sources of cohesive sediment supply are governed largely by channel discharge (Knighton 1973; Hooke 1979; Arulanandan et al. 1980; Smith et al. 2003; Magilligan et al. 2015), other variables that have significant effects on bank erosion, such as water table height (Rinaldi and Casagli 1999), vegetation presence, type and seasonality (Ratliff 1985; Thorne 1990; Micheli and Kirchner 2002a; Micheli and Kirchner 2002b), geochemical bank properties (Mahalder et al. 2016), wetting and drying cycles (Thorne 1982), freeze-thaw cycles (Gatto 1995), and the effects of previous storms (Hooke 1979) were not considered, but may be key in determining the erosion rates of burrowed banks over longer timescales (Harvey et al. 2019). The importance of considering these over long temporal scales is shown in Chapter 6, where significant associations between burrow density and bank retreat were first recorded after two years of monitoring.

156

4.5.5 General Discussion

4.5.5.1 Methodological Approach

The current experiments investigated the response of crayfish burrowing activity to different flow velocity conditions and their effect on sediment erosion. They were conducted in a way that allows the relative contributions of biotic and abiotic erosional mechanisms to be assessed. While the measured values of erosion provide a standardized and robust means of comparison within the remit of the experiments, it is unlikely that the absolute values are representative of natural erosion rates in infested rivers, where additional factors are at play, including bank materials, crayfish densities and the presence of pre-existing burrows. It is also appropriate to note that the behaviours of the crayfish observed here are likely to have been affected by the experimental set up. For example, the small mesocosms were purposefully designed with a single shelter to encourage burrowing, whereas in the field crayfish have the ability to disperse and seek alternative shelters. In addition, wire mesh was used for the mesocosms in order to create as little obstruction to the flow as possible, but presented crayfish with the opportunity to climb. Rice et al. (2012) reported that crayfish spent up to 20% of their time climbing in caged flume experiments, probably because it offered a means of avoiding aggressive interactions. Climbing the mesocosm walls was also observed during the current experiments (Figure 4.15) and the availability of this escape response may have reduced burrowing activity.

4.5.5.2 Biological and Geophysical Interactions

Previously, studies have reported the importance of exclusively geophysical (e.g. Lapointe and Carson 1986; Hasegawa 1989) or exclusively biological energy (e.g. Faller et al. 2016) in forcing sediment dynamics in cohesive river systems. Rice et al. (2016) discuss the potential for interaction between biotic and abiotic sediment forcing, but the relative contributions of biological and geophysical energy were reported independently. The results from Chapter 6 demonstrate the importance of zoogeomorphic processes for promoting abiotic erosion. These interactive or secondary effects were found to be substantial. The strength of these effects is likely to vary with flow velocity assuming this affects the relative contributions of exclusively biological and geophysical energy. Indeed, significant interactions between flow velocity and crayfish presence in driving sediment recruitment were observed in this chapter, which provides the first empirical evidence for such interactions.

157

Figure 4.15: A crayfish climbing the wire gauze mesocosm walls.

Throughout the experiments, the greatest differences in sediment recruitment between the presence and absence of crayfish occurred in medium flow velocity treatments. This is likely because the medium flow provided conditions where both biological energy and geophysical energy could actively erode the bank. An increase in sediment recruitment via abiotic erosion was observed with increasing flow velocity (Figure 4.16a), but the opposite was true of crayfish-driven erosion, with a decrease observed as flow velocities increased (Figure 4.16b). During low flow velocity conditions, erosion was driven by biological processes: crayfish burrowing caused sediment recruitment when little geophysical energy was present to erode the bank. During high flow velocity conditions, erosion was driven primarily by abiotic processes: crayfish spent the majority of time sheltering and did not interact with the bank, but the geophysical effects were greatest because of higher levels of excess shear at the bank face. It therefore seems likely that the medium flow velocity was associated with particularly large crayfish-induced erosion because both crayfish activity and abiotic mechanisms were directly capable of eroding bank materials and because the crayfish activity also facilitated abiotic erosion.

The total sediment recruited in these experiments was different from the simple addition of that portion of the erosion associated with purely biological and purely abiotic processes (Figure

158

4.16c). The difference between the sum of these components and the observed sediment recruitment is the amount of abiotic erosion facilitated by the presence of biological activity, where the biotic component (primarily burrowing) prepares the bank for increased levels of erosion by geophysical flows (diffuse erosion and collapse). Positive effects were observed for the low (13.1% interactive contribution) and medium (54.3% interactive contribution) flow velocity experiments (Figure 4.17).

These results demonstrate the importance of the interactions between biology and geomorphology, and build on the work of Corenbrit et al. (2007), Gurnell et al. (2016), Rice et al. (2016) and Rice et al. (2019). Associations between biota and sediment have typically been considered from a habitat-centric perspective (Rice et al. 2012), but Corenblit et al. (2007), Gurnell et al. (2016), Rice et al. (2016) and Rice et al. (2019) consider biota and sediment interactions as a two way relationship. Corenblit et al. (2007) and Gurnell et al. (2016) propose conceptual models considering the direct interactions of riparian vegetation providing resistance to flow in overbank flood events, and promoting the erosion of material from opposing banks. These propose that the strongest direct biogeomorphic interactions are greatest in high flow events. Field evidence reported by Rice et al. (2016) suggested that crayfish had the strongest influence in recruiting sediment during baseflow conditions, when there was little geophysical energy to facilitate erosion. Rice et al. (2019) discuss the concept that animals can promote or facilitate additional geomorphic activity, such as in the case of benthic foraging by fish, and thus provide a biological surcharge to geomorphic work undertaken. In the current experiments, the changing crayfish activity in response to flow velocity meant that whilst their direct influence on sediment recruitment by burrowing was greatest during low flow conditions, the conditioning effect was greatest at medium flows, and provides the first explicit measurements of this biological surcharge (Figure 4.17).

159

Figure 4.16: Observed sediment transport during the current experiments. (a) is sediment recruited by abiotic processes alone (reproduced from Figure 4.5), (b) is sediment recruited by purely biological activity (reproduced from Figure 4.9b), and (c) is a comparative graph of the contribution of exclusively geophysical energy (blue), as seen in (a), exclusively biological energy (red) as seen in (b), the addition of both exclusively geophysical and biological energy together (purple), and the observed sediment recruitment (green). The difference between the expected (purple) and observed (green) sediment recruitment can be attributed to a conditioning effect whereby biological energy alters the system to promote greater geophysical work; that is, to the promotion of abiotic erosion by biological activity. This proportional contribution is shown in Figure 4.17.

160

Figure 4.17: The measured proportional contribution of conditioning effects whereby expenditure of biological energy facilitates greater abiotic erosion (see Figure 4.16).

Whilst the relationship between biotic and abiotic geomorphic work has been considered under changing energy levels for static vegetation during flood events, (Corenblit et al. 2007; Gurnell et al. 2016), and for large woody material (Piegay and Gurnell 1997; Bertoldi et al. 2015; Pinto et al. 2019), the interaction between vegetation and flow is largely dictated by flow stage (Gurnell et al. 2016), with both live and dead vegetation responding passively to the flow. However, animals are able to change their behaviour in response to changing hydraulics, so the propensity for them to actively affect geomorphic processes can change with increasing flow conditions. A simple conceptual model (Figure 4.18) has been created to represent these relationships. The shaded area represents all conditioning effects from a range of crayfish activities, and is not limited to burrowing; it seeks to create a conceptual understanding of the interaction between these two energy inputs to sediment transport that have previously been considered as independent.

161

Figure 4.18: Conceptual diagram of the relative and interactive effects of geophysical and biological energy in driving sediment recruitment by burrowing in fluvial systems. As flow velocity increases, the geophysical energy (blue) available for sediment transport increases, and biological energy (red) available for sediment transport decreases due to crayfish opting to shelter as opposed to burrowing. If no interactions between biological and geophysical energy occur, the sum of both will result in the total added sediment transport (purple). Facilitations are hypothesised, with sediment recruitment greatest when both biological and geophysical energy are available to combine for sediment transport (green). Any total sediment recruitment above the purple line suggests that facilitation has occurred.

162

The biotic facilitation of abiotic transport increase from low to medium flow velocities is particularly pertinent in the case of signal crayfish. Signal crayfish typically inhabit lowland streams where bank material is cohesive, where relatively high geophysical energy is required to entrain sediments, but relatively low geophysical energy is available to erode material. Signal crayfish initiate material suspension by their burrowing and bioturbating activities, which have been shown to increase suspended sediments by up to 76% (Cooper et al. 2016), and contribute up to 47% of transported material during baseflow conditions (Rice et al. 2014). Results from the current experiments show that up to 73% of this eroded material may derive from secondary erosional processes facilitated by crayfish burrows, which is supported by field evidence at Gaddesby Brook (Chapter 6) where 89.6% of erosion was attributed to secondary erosional processes. Interactive effects mean that only a small increase in either geophysical or biological energy will lead to a disproportionate increase in sediment transport. Signal crayfish invasion and range expansion is predicted to continue to increase into the future (Sibley 2003; Chadwick 2019), and extreme fluvial flows are predicted to become more regular and more severe with climate change (Kay and Jones 2011; Schneider et al. 2013), and so the transition of the lowland systems that signal crayfish inhabit from low to medium flow velocity systems suggest that zoogeomorphic effects could become more pronounced in the future.

4.5.6 Discussion Summary

These experiments demonstrated significant interactions between flow velocity and crayfish presence in driving sediment recruitment. Flow velocity had significant effects on crayfish burrowing. Crayfish burrowed less frequently during high flow velocity treatments, with significantly less burrowed sediment recruited during medium and low velocity treatments. This was attributed to a change in crayfish behaviour in response to flow velocity, which is supported by observations of changed sheltering behaviour at high flows, reduced quality of constructed burrows at high flows and a non-significant trend of increasing erosion when artificial burrows were present but crayfish were not. When crayfish were present, total sediment erosion was highest at the medium flow velocity, with a 371% increase in sediment recruitment compared to control conditions.

These experiments also demonstrate the importance of biology for preparing riverbanks to be eroded by abiotic mechanisms. Considering relative to erosion in the absence of burrowing, when crayfish were present, an increase of 121% and 246% was observed during the low and medium flow velocity treatments, but no change was observed at high flows. This was

163 attributed to the availability of both biological and geophysical energy to undertake work at the medium flow velocity. However, the mass of sediment eroded was greater than the sum of purely biotic or purely abiotic erosion, and so these experiments explicitly quantify how conditioning effects by animals can produce a surcharge in abiotic erosion. A conceptual model of this effect is presented which highlights the importance of considering the direct and indirect effects of biota in zoogeomorphic research. Previous studies considering only the direct effects of animals on sediment transport are therefore likely to have underestimated the total effect of the biotic component.

164

Chapter 5

Quantifying rapid behavioural change in native and invasive populations

165

5.1 Introduction

Animals have the capacity to change behaviour when they become invasive (Reznick and Ghalambor 2001; Reader and Laland 2003; Wright et al. 2010; Sol and Weis 2019), and differences in the strength of a behaviour have been observed between native and invasive populations (e.g. Magurran et al. 1992; Holway and Suarez 1999; Sol and Lefebvre 2000; Jones and DiRienzo 2018). The ability of an animal to change its behaviour upon becoming invasive may be an important mechanism for the invasion of new habitats, such as stronger predator avoidance behaviours to avert novel predators (Levri et al. 2019), stronger feeding responses to consume novel prey items (Martin and Fitzgerald 2005; Green et al. 2011), and increased exploratory tendencies to colonise new habitats (Rahage and Sih 2004; Philips et al. 2008). This is important in the context of signal crayfish, because their burrowing behaviour has not been observed within their native range, or other invasive territories, and thus its expression may not be consistent between populations. Crayfish have shown to have a capacity for learning, with behaviours differing after exposure to flow velocity (Ion et al. 2020) and predator cues (Hazlett et al. 2002; Acquistapace et al. 2003), and so quantifying the response of crayfish from independent populations to external cues is necessary to justify the associations observed in in situ studies, such as Faller et al. (2016) and Chapter 3. Specifically, quantifying the responses of the variables that were significant in constructing predictive models in Chapter 3 (shelter availability and population density) is necessary, to understand if these were direct behavioural drivers or covariates of system wide processes. A series of ex situ experiments were therefore undertaken to establish the associations between crayfish burrowing and population density and substrate size, both independently and conjointly, considering crayfish from native and invasive populations.

5.2 Aims

This chapter aims to address aspects of objective 2 (to examine the biotic and abiotic drivers of signal crayfish burrowing behaviour). Specifically, an experimental approach was used to address three key questions:

1. How does shelter availability affect the propensity of crayfish to burrow?

166

Sediment size was significantly negatively associated with crayfish burrowing in Chapter 3, likely as it provided an alternative shelter. It is therefore hypothesised that crayfish would select to hide in an alternative shelter if available.

2. How does population density affect the propensity of crayfish to burrow?

Population density significantly positively contributed to presence / absence models in Chapter 3, and so it is hypothesised that competition from other crayfish for existing shelters will drive burrowing behaviour.

3. How does population provenance affect the propensity of crayfish to burrow?

Crayfish burrows were present at some, but not all, infested rivers in Chapter 3, and so many generations of crayfish in the unburrowed rivers may never have constructed a burrow, and certainly not in recent years where burrow remnants would be detectable. This reflects differences in biotic and abiotic conditions that promote burrowing between sites, but it may reflect a varying propensity for crayfish from different populations to burrow at all. The behaviour of animals can change upon becoming invasive (Wright et al. 2010; Sol and Weis 2019), and burrowing has not been recorded in the native range of signal crayfish. It is therefore hypothesised that the strength or response will be different between native and invasive populations.

5.3 Methods

An experimental mesocosm study was undertaken to investigate the importance of different types of shelter availability (no shelter availability, presence of a large rock and deep silt substrate), and population density (low, medium and high), on the propensity of signal crayfish to burrow. The experimental design was semi-factorial, outlined in Table 5.1, with each treatment replicated seven times, yielding 49 runs for each population.

Experiments were undertaken in Loughborough, Leicestershire, UK, and in Bozeman, Montana, USA, to allow for experiments to be repeated with crayfish from native and invasive populations. Four populations were chosen for study: two from the UK, where they are invasive, from one river where burrows are present in high densities, and from one river where burrows are absent; and two from Montana, USA, from one river where signal crayfish are native, and from one river where signal crayfish have recently been introduced. Burrows were

167 absent at both Montanan sites. The selected populations represent crayfish from four independent and ecologically distinct backgrounds (Table 5.2):

i) Clark Fork River and Bitterroot River, MT, USA. This is the native North American population of signal crayfish (Larson and Olden 2011). Population density was visibly higher than the East Gallatin River when hand searching, but was not quantified. Testing the native population allowed for the investigation of whether development of the burrowing behaviour had occurred or changed after crayfish had been initially removed from the native population. No burrows were present at the native sites. ii) East Gallatin River, MT, USA. Crayfish are not native to the East Gallatin River, and have been recently introduced, with the first sighting occurring within four years of collection. This population represents a recent invasion, with source individuals likely taken from the native population. Population density is low (catch per unit effort from trapping (CPUE) = 0). Testing this population allowed for the investigation of whether it is being invasive that is important in driving burrowing behaviour, or a specific trait that developed after crayfish were exported from their native range to Britain. No burrows were present at this site. iii) Gaddesby Brook, UK. Crayfish were introduced to the upper parts of Gaddesby Brook in the 1980s, and have now colonised the entire brook. This represents an established invaded population with high population density (CPUE = 9.3). Burrows were abundant (102 burrows per 100 meters of riverbank) at this site. Testing crayfish from this site in comparison with crayfish from the River Etherow allowed for investigating whether burrowing is a learned behaviour. iv) River Etherow, UK. Crayfish were introduced to the adjacent tributary Glossop Brook in 1997, and have spread into the main river channel. The population is establishing, with a medium population density (CPUE = 2.75), Burrows were absent at this site. Testing crayfish from this site in comparison with crayfish from Gaddesby Brook allowed for investigating whether burrowing is a learned behaviour.

168

Hide Treatment No Shelter Silt Rock Low ✔ ✔ ✔ Medium ✔ - ✔

Density

Population Population High ✔ - ✔

Table 5.1: Treatment matrix. Silt treatments were not replicated at medium and high density treatments, due to limited time and funding for completing experiments. As there was no significant difference between silt and no-shelter treatments at low density, silt was discarded as an alternative shelter, with rocks as an alternative shelter prioritised. These seven experimental treatments were undertaken for each four populations.

Years since Estimated Population In text River Location Represents introduction Density (CPUE) abbreviation

Montana, USA Clark Fork River and (46.868, -113.996) US Native Population - - USN Bitterroot River (46.853, -114.099)

Montana, USA East Gallatin River US Introduced Population 4 0 USI (45.782, -111.113)

Leicestershire, UK UK Invasive Population, Gaddesby Brook 17 9.3 UKB (SK 6935 1270) Burrowing

Greater Manchester, UK UK Invasive Population, River Etherow 20 2.75 UKNB (SK 0096 9528) No Burrowing

Table 5.2: Crayfish population locations, what the study population represents, and their associated in text abbreviations.

5.3.1 Elected Treatments

5.3.1.1 How does shelter availability affect the propensity of crayfish to burrow?

To investigate the effect of shelter availability on the propensity of crayfish to burrow, three shelter treatments were selected (no shelter; large rock; deep silt; Figure 5.1), as these represent an array of the alternative shelters commonly available to signal crayfish in British streams,

169 and were significantly associated with burrowing in Chapter 3. Vegetation was also considered for use, as signal crayfish have been reported to use macrophyte stands for shelter (Johnson et al. 2014; Vedia et al. 2017), although this was not observed in Chapter 3. Vegetation was not used because at the mesocosm scale vegetation would likely be consumed, and thus affect the behavioural opportunities for crayfish present between experiments.

For the deep silt substrate treatment, a layer of loose bentonite clay covered the gravels to a depth of 0.05m. This was created by expanding bentonite clay pellets in an excess of water (1:5 pellet to water volume ratio), and manually disaggregating the loose clay structure evenly across the base of the mesocosm. The loose bentonite was allowed to settle to form a uniform bed substrate before any runs were started. Whilst crayfish have previously been recorded using silt as a shelter, loose disaggregated clay provided a functionally similar alternative for these experiments.

Each rock used as an alternative shelter in experiments were rounded or sub-rounded, of medium sphericity, and cobbles (128 mm to 180 mm) on the Wentworth scale. One rock was used per mesocosm, with rocks being deliberately selected to be consistent in both shape and size.

Shelter availability and population density results are reported considering average values of all UKB, UKNB, and USN populations. USI were excluded as not all runs were completed.

Figure 5.1: Shelter availability treatments used in the experiments; (a) no shelter, (b) a single large rock; and (c) a deep silt substrate.

170

5.3.1.2 How does population density affect the propensity of crayfish to burrow?

Experimental population densities were chosen to represent a low (1 crayfish per mesocosm; 5.5 crayfish m-2), medium (2 crayfish per mesocosm; 11 crayfish m-2) and high population density (4 crayfish per mesocosm; 22 crayfish m-2), with the high-density treatment equating to 22 crayfish m-2, a similar density to the highest densities of adult signal crayfish recorded in British streams (Bubb et al. 2004). Higher densities of signal crayfish have been recorded when the presence of juveniles is accounted for, with over 100 crayfish m-2 in Bookill Gill Beck being consistently recorded (D. Chadwick, pers. comm.), but juveniles were not considered in these experiments. Results are reported considering UKB, UKNB, and USN populations. USI were excluded as not all runs were completed.

5.3.1.3 How does population provenance affect the propensity of crayfish to burrow?

To address the effect of population provenance on the propensity of crayfish to burrow, the above treatments were undertaken using four populations of crayfish; two populations of crayfish were sampled from the UK; one from a site where burrows were present, and one from a site where burrows were absent; and two populations from the USA; one from their native range, and one from a range where they have been recently introduced. Crayfish burrows were absent from both USA sites. Details of these sites are detailed in Table 5.2.

5.3.2 Collection of Animals

Signal crayfish were collected from four locations for use in the experiments; two locations from infested rivers in the UK where signal crayfish are invasive, and two populations from Montana, USA, one from their native range, and one where they are invasive.

Crayfish were collected from Gaddesby Brook, Leicestershire, UK (SK 6935 1270), where crayfish burrows are abundant (102 burrows per 100 meters of riverbank), and from the River Etherow, Greater Manchester, UK (SK 0096 9528), where no crayfish burrows are present. Crayfish were also collected from the East Gallatin River, Montana, USA (45° 46’ 55” N, 111° 06’ 47” W), where signal crayfish are invasive based on one reported sighting in the past five years (Montana Field Guide 2019), and from the Clark Fork River (46° 52’ 05” N,

171

113° 59’ 46” W) and Bitterroot River (46° 51’ 11” N, 114° 05’ 56” W), Montana, USA, where signal crayfish are native (Larson and Olden 2011). Crayfish burrows were absent at both USA sites. Population dynamics of the animals collected are detailed in Table 5.3.

Number Carapace Length Mean Carapace Location Male / Female Collected Range (mm) Length (mm)

Gaddesby Brook 37 16 / 21 40-60 49.4

River Etherow 44 18 / 26 37-66 49.3

East Gallatin River 15 10 / 5 30-64 47.4

Clark Fork River and Bitterroot River 31 15 / 16 30-58 38.4

Table 5.3: Population details of crayfish collected from the four locations.

The two populations of crayfish from the UK were kept in two separate circular holding tanks (1.2 m x 1.2 m x 1.4 m), containing 450 l of dechlorinated tap water, which was aerated, filtered, and filled the tanks to a depth of 0.4 m (Figure 5.2). The two populations of crayfish from the USA were kept in two separate holding tanks (1.3 m x 0.7 m x 0.6 m), containing 180 l of dechlorinated tap water, which was aerated, filtered, and filled the tanks to a depth of 0.2 m.

5.3.3 Physical Setup and Measurements

Experiments were first undertaken in the UK between November 2017 and April 2018, and from January to March 2019.

A 0.2 m deep bentonite clay bank, the maximum depth recorded in laboratory conditions for a crayfish burrow (Stanton 2004), was constructed at one end of 14 identical opaque mesocosms (0.53 m x 0.33 m x 0.29 m; Figure 5.3). Bentonite clay pellets were expanded in tap water (1:1.5 pellet to water volume ratio) and compacted into the ends of the mesocosms using a shovel. Banks were constructed to an angle to approximately 70o. Previous studies have used vertical clay banks (Harvey 2014), however Chapter 3 demonstrated that vertical banks were largely absent in studied reaches, and so steep, but not vertical, banks were constructed to best replicate the environments in which burrows are naturally constructed. Banks were smoothed prior to the start of the run.

172

Figure 5.2: A holding tank for crayfish between experimental runs in the UK. Plastic piping was used to provide shelter for crayfish.

A gravel substrate (D10 11.2 mm; D50 13.3 mm; D90 15.9 mm) covered the base of the mesocosm to a depth of 0.03m. The mesocosms were filled with dechlorinated tap water, which was kept at an ambient temperature, of between 14 °C and 18 °C, except for the heated water treatment, which was heated to 21 °C. Mesocosms were illuminated by natural range LED lights (6,500 K white; 1,950 lumens) suspended 0.65 m above the mesocosms with timers that switched the lighting on for 12 h (07:00 – 19:00) then off for 12 h (19:00 – 07:00). Blackout curtains surrounded the experimental setup to exclude light from external sources.

Air stones were used to oxygenate the water. Installing filters in the mesocosms was trialled during preliminary runs, but continual water movement caused by the pumps suspended fine sediment. Filters were therefore not used during experimental runs, to remove the influence of artificial turbidity on crayfish behaviour.

For each run, crayfish were selected at random from the population. Size and sex were recorded, and only crayfish that were not in moult, were sexually mature (larger than 30mm carapace length (CL); Johnson and Taugbol 2010) and had intact legs and chelae were used. The order of treatments was randomised to minimise the effect of potential learning on the outcome of experiments. Crayfish were placed into mesocosms, and experiments ran for 84

173 hours, after which crayfish were removed and returned to holding tanks. Crayfish were fed for at least three days between experimental runs on carrot sticks and sinking catfish pellets but were not fed whilst the experiments were in progress.

Figure 5.3: Schematic diagram of mesocosms used in the experiments.

Burrowing activity was recorded every 12 hours; in the morning after lights were switched on, and in the evening immediately prior to the lights being switched off. This allowed for the speed at which burrows were constructed to be recorded throughout the run. Burrows were deemed ‘usable’ at a depth of 70 mm, as this was the smallest burrow depth recorded which crayfish inhabited and defended a shelter.

Burrow depths and the width and height of burrow entrances were measured using a ruler to the nearest 5 mm. Burrow depths were measured to the centre of the burrow height, due to the sloping bank face. The volume of sediment excavated was calculated by treating the burrow shape as an elliptical cylinder (as in Faller et al. (2016));

퐴퐸 = 휋 (푊⁄2 퐻⁄2) (Equation 5.1)

where AE is burrow entrance area, W is the burrow entrance width, and H is the entrance height.

174

푉퐵 = 퐴퐸 퐿 (Equation 5.2)

where VB is burrow volume, and L is the length of the burrow.

In the event of multiple burrows being excavated, the volumes of the burrows were added together to report the total volume of sediment excavated in a single run, which is herein reported as ‘sediment excavated’. Individual burrows were also analysed between treatments, which are herein reported as ‘burrow size’. This distinction is important to consider, as the same mass of sediment could be excavated by digging one large burrow constructed for homing purposes or multiple smaller burrows designed as bolt holes in use as a defence mechanism.

The bentonite clay used in these experiments had a bulk density of 1.41 g cm-3, and so all volumes were converted to sediment mass for analysis;

푀퐵 = 훾 푉퐵 (Equation 5.3)

where MB is burrow mass, and g is the calculated bulk density of the excavated sediment, which here is 1.41 g cm-3.

Any sediment excavated from the no-shelter and rock treatments was siphoned out of the mesocosms at the 12 hour intervals where burrows where recorded, so as not to recreate the deep silt substrate treatment, and to reduce any potential deoxygenation affects that a large quantity of loose fine clay might have caused.

In each experiment, each treatment was replicated seven times, giving a total of 49 experimental runs for each of the two UK populations.

Experiments were repeated in the USA in 16 near-identical mesocosms (0.47 m x 0.36 m x 0.26 m) with the same lighting set up and regime (USA: 6,500 K white; 300 lumens). Experiments were conducted from October to November 2018, and comprised the same treatments as were applied to the UK populations, excluding the heated water treatment. Again, each experiment was replicated seven times, giving a total of 49 experimental runs for the native US population. However, fewer replications were achieved with the invasive US population, due to lower availability of crayfish, yielding a total of 22 experimental runs (Table 5.4).

175

Treatment and Population Density

Silt No Shelter Rock

Low Low Medium High Low Medium High

UKB 7 7 7 7 7 7 7 UKNB 7 7 7 7 7 7 7 USN 7 7 7 7 7 7 7

Population USI 4 7 - - 7 4 -

Table 5.4: Number of replicates undertaken for each treatment.

5.3.4 Data Analyses

All burrow measurements were tested for normality using the Shapiro-Wilk test, and none were normally distributed (p < 0.001 in all cases). Log and square root data transformations were not possible, due to the presence of a high number of zeros in the dataset, which were true zeros as opposed to an absence of data. Therefore, non-parametric statistical techniques (Mann- Whitney U (U) for differences between pairs, Kruskal-Wallis (H) for differences between groups, and Spearman’s Rank (rs) for correlation) were used for analysis.

5.3.5 Consideration of Temperature

Mesocosms used in the experiments were not temperature controlled, with water temperature remaining ambient (as in Harvey et al. (2014)). Temperature may have an impact on crayfish activity; as crayfish are ectotherms, it is possible that their rate of activity and thus burrowing increases with increased water temperature. Alternately, it has been hypothesised that crayfish burrowing is a mechanism to escape the extreme cold (Babcock et al. 1998), and so burrowing may be negatively correlated with temperature. To ensure that the experimental design was robust to the potential temperature changes experienced within the laboratory environment, which has the potential to be a controlling and unaccounted for factor, a replication was undertaken of a single crayfish in the no-shelter treatment but with water heated to 21 °C for both British crayfish populations.

The median mass of sediment excavated was zero for both the heated and unheated runs, which were not significantly different for either tested population (UKB: U = 20.0; p = 0.620; and UKNB: U = 35.0; p = 0.209; Figure 5.4a; Table 5.5). Burrows constructed were larger in the

176 unheated treatment than the heated treatment (median = 451.6 g and 132.8 g respectively), but there was no significant difference for either tested population (UKB: U = 2.0; p = 0.800; and UKNB: U = 9.5; p = 0.095; Figure 5.4b). The time which it took for crayfish to construct burrows was not different between temperature treatments (UKB: U = 3.5; p = 1.000; and UKNB: U = 8.5; p = 0.190).

Figure 5.4: The effect of temperature on (a) the mass of sediment excavated and (b) burrow size by one crayfish, for both UK populations, considering mean values, +/- 1 standard error (SEM).

Mass of Sediment Excavated (g) Burrow Size (g)

UKB UKNB All Crayfish UKB UKNB All Crayfish

Heated 172.4 56.9 265.0 402.3 93.9 321.1

Mean Unheated 70.4 612.9 635.1 246.6 586.6 683.4

Heated 0.0 0.0 0.0 107.9 199.2 132.8

Median Unheated 0.0 265.6 0.0 246.6 650.8 451.6

Table 5.5: Mass of sediment excavated (g) and burrow size (g) of crayfish from UKB, UKNB, and both UK populations in a heated and unheated mesocosm with no shelter.

177

There was no significant difference between the heated and the unheated treatments in either tested population, but the median mass of sediment excavated by burrowing was higher in the unheated treatment than the heated treatment. A greater difference between median values was observed between temperature treatments for UKNB crayfish. Whilst this was not significant, it should be noted and considered that crayfish from different populations may respond differently to changes in water temperature.

As no significant difference was observed, the effect of temperature was discounted as significantly influencing the results of these experiments. The randomisation of treatment order will also have also dampened any impact that varying temperature may have had on crayfish behaviour, and so within the ambient temperature setting of these experiments, temperature can be discounted as a significantly influential variable.

5.4 Results

5.4.1 General Results

Crayfish burrowed significantly more at night than during the day (U = 5,307.5; p < 0.0005), with 97.2% of sediment excavation occurring when the lights were off. Males constructed larger burrows than females (median = 599.4 g and 438.3 g respectively; Figure 5.5a), but this was not significant (U = 37.0; p = 0.301). There was no relationship between burrow size and crayfish carapace length in males (rs15 = 0.142; p= 0.300), females (rs14 = -0.058; p= 0.425), or when considered together (rs29 = 0.043; p= 0.412; Figure 5.5b).

Figure 5.5: A comparison of the size of crayfish burrows between crayfish (a) sex, , considering mean values, +/- 1 standard error (SEM), and (b) size.

178

5.4.2 How does shelter availability affect the propensity of crayfish to burrow?

In the low density treatment, there was no significant difference in the median mass of sediment excavated between no shelter and the presence of a deep silt substrate in any of the populations, or when the populations were considered together (U = 255.0; p = 0.061; Figure 5.6a; Table 5.6). There was no significant difference between the size of burrows constructed between the deep silt substrate and the absence of a shelter (U = 19.0; p = 0.769; Figure 5.6b), and there was no difference in the time taken to construct burrows between any of the treatments (U = 19.5; p = 0.085;). However, crayfish constructed no burrows in the presence of a large rock, which recruited significantly less sediment (U = 115.5; p < 0.0005; Figure 5.6a). Therefore, the large rock was considered an alternative shelter, but the deep silt was not, and the rock was chosen for use in the medium and high population density runs.

Figure 5.6: The effect of shelter availability on (a) the mass of sediment excavated and (b) burrow size in the low population treatment, considering mean values, +/- 1 standard error (SEM).

179

Mass of Sediment Excavated (g) Burrow Size (g) Shelter Availability Crayfish Density UKB UKNB USI USN All UKB UKNB USI USN All No Shelter Low 70.4 612.9 998.0 204.7 296.0 246.6 858.0 1023.4 358.2 565.1

Medium 97.3 144.4 - 380.1 207.3 170.3 202.1 - 332.6 256.0

High 211.9 1170.7 - 308.8 623.8 296.6 957.5 - 308.8 629.3 Large Rock Low 0.0 0.0 53.7 0.0 0.0 - - 375.9 - -

Mean Medium 151.4 448.6 323.1 375.1 325.1 323.8 314.0 646.3 437.7 358.2

High 530.3 845.0 - 512.9 629.4 464.0 492.9 - 598.4 508.4 Deep Silt Low 132.8 41.7 1394.0 74.6 83.0 929.8 292.2 1394.0 522.0 581.3

All Runs 88.7 275.8 646.8 163.6 436.7 482.7 1029.6 416.4

No Shelter Low 0.0 265.6 946.4 222.7 0.0 246.6 650.8 959.4 312.8 438.3 Medium 0.0 0.0 - 197.6 149.4 138.6 149.4 - 183.5 157.7

High 0.0 534.6 - 411.7 411.7 307.2 531.3 - 365.3 405.1 Large Rock Low 0.0 0.0 0.0 0.0 0.0 - - 375.9 - -

Median Medium 221.4 432.2 0.0 159.4 199.2 323.8 216.1 646.3 169.6 219.2

High 503.6 647.5 - 514.7 514.7 393.8 309.9 - 668.3 350.1 Deep Silt Low 0.0 0.0 1604.4 0.0 0.0 929.8 292.2 1604.4 522.0 522.0

All Runs 204.2 517.0 761.1 328.2 381.0 278.9 972.5 310.4

Table 5.6: Mass of sediment excavated (g) and burrow size (g) considering all populations and all treatments. Total values given for crayfish populations concern only treatments where all four populations were tested (no shelter, low density; large rock, low density; large rock, medium density; deep silt, low density), and total values given for treatments concern only UKB, UKNB, and USN populations, as USI were not tested in every treatment.

180

The presence of a rock resulted in a higher median mass of sediment being excavated than when a rock was absent in the medium and high population densities (Figure 5.7a; Figure 5.7b; Table 5.6), although neither of these were significant (U = 178.0; p = 0.259 and U = 212.5; p = 0.322 respectively). This pattern was also evident in the size of individual burrows that were constructed in the presence and absence of a rock. Larger burrows were constructed when a rock was present than when it was absent in the medium population density treatment (Figure 5.7c), but smaller burrows in the high population density treatment (Figure 5.7d), although none of these differences were significant (U = 206.0; p = 0.165 and U = 194.5; p = 0.727 respectively). There was no significant difference in the time that it took crayfish to construct burrows between the presence and absence of a rock considering the medium (U = 140.0; p = 0.669) or high (U = 215.5; p = 0.152;) population density treatments.

181

5.4.3 How does population density affect the propensity of crayfish to burrow?

When a rock was present, two crayfish excavated significantly more sediment than one crayfish (U = 378.5; p < 0.0005), and four crayfish excavated significantly more sediment than two crayfish (U = 308.0; p = 0.027; Figure 5.8b). The size of the burrows constructed increased with population density (median = low: no burrows; medium: 219.2 g; high: 350.1 g), but the increase from the medium to high population density treatment was not significant (U = 282.0; p = 0.130; Figure 5.8d), although burrows were constructed significantly more quickly in the high population density treatment than the medium density treatment (U = 100.0; p = 0.020).

However, when a rock was absent, the same positive association between sediment recruitment and population density was not apparent, with no significant difference between population densities (H2 1.752; p = 0.416; Figure 5.8a). Burrows that were constructed were also smallest in the medium density treatment (median = low: 438.3 g; medium: 157.7 g; high: 405.1 g), and were significantly smaller than in the low or the high density treatment (H2 8.104; p= 0.017; Figure 5.8). The time to construct these burrows did not differ between population density treatments (H2 3.057; p = 0.217).

5.4.4 How does population provenance affect the propensity of crayfish to burrow?

There was no significant difference between the two British populations of crayfish in the median mass of sediment excavated in the heated treatments (U = 22.0; p = 0.805; Figure 5.4a; Table 5.5), and there was no significant difference between the size of burrows constructed (U = 4.0; p = 1.000; Figure 5.4b).

Considering the low density treatments, when there was an absence of shelters, there was no significant difference in the median mass of sediment excavated between the two British populations and the native population (H2 2.320; p= 0.313; Table 5.6). However, the USI crayfish excavated significantly more sediment than the UKB crayfish (U = 44.5; p = 0.007; Figure 5.9a) and the USN crayfish (U = 5.5; p = 0.011). USI crayfish also constructed the largest burrows and were the only population to construct burrows of a significantly different size to the USN crayfish (U = 1.0; p = 0.019; Figure 5.9b).

182

Figure 5.8: The effect of population density on the total mass of sediment excavated (a and b) and mean burrow size (c and d) when a large rock shelter was absent (a and c) and present (b and d) , considering mean values, +/- 1 standard error (SEM).

Figure 5.9: Differences between populations considering (a) the total mass of sediment excavated, and (b) burrow size when a shelter was absent in the low population density treatment, considering mean values, +/- 1 standard error (SEM).

183

When a rock was present, there was no difference in the median mass of sediment excavated between populations; USI crayfish were the only crayfish to construct burrows, which only occurred on one occasion. When silt was present in the mesocosms, there was no significant difference between the two British and the USN population (H2 0.020; p = 0.990), but the USI crayfish excavated significantly more sediment than crayfish of any other locality (H4 = 13.331; p= 0.004; pairwise: UKB U = 27.0; p = 0.012; UKNB U = 27.0; p = 0.012; USN U = 1.0; p = 0.012; Table 5.6). USI crayfish also constructed the largest burrows, but this was not significantly different from the other populations (H3 1.554; p= 0.670).

Considering the medium density treatment, there was no significant difference between the median mass of sediment excavated when a shelter was absent between both British populations and the USN populations (H2 5.139; p = 0.077; Figure 5.10a), and there was no significant difference between the size of individual burrows constructed (H2 0.958; p = 0.619; Figure 5.10c). There was also no significant difference in the median mass of sediment excavated between any populations when a rock shelter was present (H3 1.993; p = 0.574; Figure 5.10b), and whilst USI crayfish constructed the largest burrows, which were more than twice as large as any other population, the difference was not significant (H3 2.545; p= 0.467; Figure 5.10d).

Considering the high population density treatment, there was no significant difference between the median mass of sediment excavated by the UKB, UKNB, and USN population when a shelter was absent (H2 5.244, p = 0.073; Figure 5.11a), but the size of burrows constructed by UKNB crayfish was significantly larger (531.3 g) than the UKB crayfish (307.2 g; U = 35.0; p = 0.030; Figure 5.11c), but not the USN crayfish (365.3 g; U = 3.5; p = 0.085). When a shelter was present, there was also no difference in median mass of sediment excavated in the high density treatments (H2 0.622; p = 0.733; Figure 5.11b), or in the size of individual burrows constructed (H2 1.486; p = 0.476; Figure 5.11d).

When all runs were considered, without differentiation for crayfish density or shelter provision, there was no significant difference in sediment excavated between either British population, or

USN crayfish (H2 1.709; p = 0.425; Figure 5.12a; Table 5.6). However, USI crayfish (761.1 g) excavated significantly more sediment than UKB crayfish (UKB = 204.2 g; U = 532.0; p = 0.006), constructed burrows more than twice the size of any of the other populations (H3 8.120; p = 0.044; Figure 5.12b) and constructed burrows significantly more quickly than crayfish from any other population (H3 15.621; p= 0.001).

184

185

Figure 5.12: Differences between populations on the total mass of sediment excavated (a) and burrow size (b) across all runs where all populations were considered, considering mean values, +/- 1 standard error (SEM).

186

5.4.5 Results Summary

Crayfish burrowing was almost exclusively nocturnal, and the speed of burrowing, the size of burrows, and the number of burrows constructed were not related to crayfish size, and did not differ between male and female crayfish.

A large rock was considered an appropriate alternative crayfish shelter, with one crayfish burrowing significantly less when a rock was present, whereas there was no difference in burrowing behaviours in the presence and absence of a deep silt substrate. When a rock was present, an increase in population density resulted in a significant increase in total sediment excavated, whereas this pattern was not observed when a rock was absent.

The patterns of burrowing responses were consistent across all tested populations, however there was significant difference in the size of burrows and the total volume of sediment excavated between populations. USI crayfish burrowed significantly more than either USN and UKB crayfish across all tested experiments. UKNB crayfish burrowed more than either the UKB crayfish or the USN crayfish, although this was not significant. There was little difference, which was not significant, between the UKB and USN populations.

5.5 Discussion

Significant differences in signal crayfish burrowing behaviour were recorded as a result of differences in shelter availability and population density, and burrowing was recorded by crayfish from all populations (Figure 5.9; Figure 5.13). Signal crayfish are a highly invasive, widespread organism, and these results demonstrate that they have the potential to cause substantial geomorphic impacts across a global range, not just in the UK where burrowing has so far been recorded. Whilst studies have attempted to quantify the volume of sediment that crayfish burrows recruit into river systems (Faller et al. 2016), the characteristics of burrowed riverbank (Faller et al. 2016), and the geometry and morphology of crayfish burrows (Guan 1994), this is the first study investigating the burrowing responses of crayfish to different biotic and abiotic stimuli. Calculating the mass of sediment excavated by burrowing allowed consistent quantification of behavioural responses.

Bank erosion other than by burrowing was not quantified during this study. However, crayfish were observed walking on the bank, and in medium and high density treatments, fighting between crayfish was also observed on the bank. These were both seen to directly entrain fine

187 sediment into the water column. Crayfish fighting was also observed to entrain fine sediment through the escape behaviours of fighting crayfish, with tail flicks resulting in plumes of suspended sediment. In mesocosms where crayfish did not burrow, loss of bentonite clay mass and riverbank collapse both frequently occurred through these alternative erosive mechanisms (Figure 5.14), demonstrating the wide range of behaviours of signal crayfish that recruit sediment into river systems.

Figure 5.13: A burrow constructed by a signal crayfish in one of the mesocosm experiments.

Figure 5.14: A crayfish that has not burrowed but nevertheless caused bank retreat through walking on the banks and recruiting sediment through bioturbation.

188

5.5.1 General Discussion

There was no observed difference in the size of burrows constructed between males and female crayfish. This supports a previous study that also found no significant difference between the size of burrows constructed by male and female signal crayfish (Guan 1994). However, it has also been found that when offered existing burrows, sexual dimorphism was present, with females electing to occupy smaller burrows, whereas males elected to occupy larger burrows (Ranta and Lindstrom 1993); this is possibly as a result of claw size, because tight fitting burrows are easier to defend. Guan (1994) recorded that up to 40% of crayfish did not construct their own burrow, and simply fought for existing burrows. It may be that crayfish compete for and exchange burrows at such a high rate, that delicacy in constructing their dimensions is not an important factor, because they are likely to be taken over by a competitor, and the speed of construction is a more important construction pressure for reasons of safety.

This theory was supported by the absence of a relationship between crayfish size and burrowing. It has previously been suggested that small adult crayfish are more likely to burrow, as large adult crayfish do not need to burrow, and instead use their carapace as protection (Guan 1994), but this pattern was not witnessed here. Further work to investigate the propensity of smaller crayfish than were used in this experiment would provide additional insight into the propensity of crayfish to burrow throughout their growth. Burrows were regularly found of dimensions smaller than could be constructed by adult crayfish in Chapter 3, and juvenile crayfish of a carapace length <15mm were witnessed occupying small burrows on the River Bain in Chapter 6, which were likely self-constructed (Figure 5.15).

A significant difference was observed between daytime and night-time activity, with 97.2% of burrowing occurring when the mesocosm lights were switched off. Rice et al. (2014) and Harvey et al. (2014) theorise that diel turbidity cycles witnessed on the River Nene and River Windrush respectively are a result of nocturnal crayfish activity. Harvey et al. (2014) conducted mesocosm experiments (n = 6) demonstrating a diel sediment flux was driven by nocturnal crayfish activity, through walking on loose substrate (n = 3) and through burrowing (n = 3). These experiments (n = 183) provide strong supporting evidence that crayfish are likely to be responsible for driving the night-time flux in suspended sediment concentrations observed in infested rivers.

189

Figure 5.15: The photographed crayfish was hand caught from the entrance of a burrow it was inhabiting. The size of the burrow was proportional to its body, and so was likely self-constructed.

A small number of casts were made of completed burrows. In the calculation of burrowed volume, burrows were treated as elliptical cylinders, but the casts reveal that not all burrows were so idealised (Figure 5.16b). All burrows recorded were single entrance burrows, but they often differed in their internal structure. This variation was recognised and documented by Guan (1994), and future work quantifying the ratio of burrow entrances to burrow volumes would increase the accuracy of these results.

5.5.2 How does shelter availability affect the propensity of crayfish to burrow?

Significant differences in crayfish burrowing activity were recorded depending on shelter availability. Most striking were significant differences between the large rock and no shelter treatments, suggesting that large rocks provide alternative shelter that can preclude burrowing; a preference that may reflect an energy saving strategy. Where large rocks are present in rivers, crayfish may therefore be more likely to use them for shelter, making burrowing less likely. Previous field studies are inconsistent: field surveys in the River Wharfe, UK, found crayfish using rocks larger than 64 mm as refuges (Peay and Rogers 1999) but, in the Iberian Peninsula, signal crayfish were found to be negatively correlated with the presence of boulders (Vedia et

190 al. 2017). Chapter 3 found a negative association between the presence of coarse bed material and crayfish burrows, suggesting that should crayfish be present in rivers where large substrate is available, they are likely to reduce any resultant burrowing.

Figure 5.16: Differences in burrow morphology between experiments. (a) A straight burrow with a consistent diameter, and (b) a cast of a burrow sloping upwards with an internal diameter that is not consistent with the burrow entrance at the right of the picture. These differences were observed between many burrows, with (a) representative of many burrows not influenced by mesocosm walls, but the burrow presented in (a) most easily shows this burrow morphology due to its construction next to the translucent mesocosm wall. Scales are approximate.

It has previously been suggested that large males may use a silty substrate as cover in order to conserve energy relative to burrowing, instead lying in the silt, using their hardened carapace as a defence against predators (Guan 1994). Although such behaviour was witnessed here (Figure 5.17), there was no difference in burrowing in the silt substrate and no shelter treatments, suggesting that a deep silt substrate cannot be considered an alternative shelter for signal crayfish use.

The lack of difference in response behaviour between silt and no available shelter suggests that silt availability is unlikely to preclude burrowing, such that burrowing which results in fine sediment deposition is unlikely to dampen further burrowing activity. Whilst there are currently no studies demonstrating a link between crayfish burrowing and fine sediment deposition, this has been widely reported, and an association between the mass of sediment excavated by

191 crayfish burrows and the percentage of fine bed material was found in Chapter 3. It is reasonable to hypothesise that crayfish contribute excavated fine sediment into lowland rivers through their burrowing behaviour. Should this layer of silt act as an alternative shelter for crayfish to use, a negative feedback cycle might be created and so reduce burrowing. However, as the presence of silt did not inhibit crayfish burrowing, crayfish are likely to continue burrowing in rivers in spite of any associated fine sediment deposition.

Figure 5.17: A signal crayfish burying itself in the deep silt substrate during experiments.

Crayfish have been observed using other forms of cover, such as large woody debris (Walter 2012) and macrophyte stands (Johnson 2014; Veida et al. 2017), that were not used in these experiments. Large woody debris could not be used as it would not fit into a mesocosm, and the classification of its functional use would be similar to that of the large rock. Macrophytes may however offer a different type of functional cover that cannot be directly compared to any other shelter availability tested here. Macrophytes were not used because crayfish are known to consume them (e.g. Matsuzaki et al. 2009; Twardoschleb et al. 2013; Wood et al. 2018), and this would add an additional potential behaviour into the mesocosms that could not be standardised across treatments. The use of plastic or silk macrophytes in replacement may offer a suitable comparison but these were not tested here.

192

5.5.3 How does population density affect the propensity of crayfish to burrow?

A significant increase in burrowing activity with increasing population density was observed. Previous studies have not considered population density as a driving factor (Faller et al. 2016) or not found a relationship between burrow densities and crayfish population density (Guan 1994), which has been attributed to the poor relationship between crayfish population density and trapping catch per unit effort. The experiments reported here have demonstrated that population density plays a significant role in driving the burrowing behaviour of signal crayfish, and should be considered as an important factor in future field analyses.

Whilst the availability of rock cover caused a marked reduction in burrowing when only one crayfish was present, this effect was not as strong, or was non-existent, with two or four crayfish present. When a rock was present, a strong linear increase in excavated sediment was observed (Figure 5.17c) with increasing population density, which was not observed when the rock was absent (Figure 5.17a). The rocks used in experiments were suitable shelters for a single crayfish. Increased burrowing with two or more crayfish suggests that the rocks were not adequate for two crayfish and that burrowing occurred when the population density exceeded the available shelter opportunities. This suggests that riverbank burrowing may be directly related to the size of the crayfish population relative to in-stream shelter availability.

Two crayfish burrowed significantly less than one crayfish when no rock was present. Signal crayfish are highly aggressive (Houghton et al. 2017), and so this result may reflect aggressive interactions between the crayfish, where time and energy was spent interacting with other crayfish rather than constructing burrows. This pattern was not observed when a large rock was present. During experiments, competitive interactions were witnessed between crayfish, often over the use of the alternative shelter. This pattern may not have been observed, perhaps because the dominant crayfish commandeered the shelter, leaving the second crayfish to burrow, or perhaps because the large rock broke up the line of sight, reducing the chance of encountering the other crayfish. These results contradict Statzner and Peltret’s (2006) observation that limiting shelter space, which led to more interactions, did not reduce the engineering activity of crayfish but are consistent with other studies that found no relation between increased population density and increased sediment transport (Rice et al. 2012; Albertson and Daniels 2018).

193

These results have implications for the management of signal crayfish. In the low density treatments, the presence of a rock reduced burrowing activity, and so the addition of large rocks into rivers could be a suggested management technique for preventing the geomorphological problems associated with signal crayfish burrowing. However, these results show that both burrow numbers and sizes may increase with increased shelter availability, and so introducing alternative shelters into river systems may have an adverse impact on the system. However, these experiments were only short term, and conclusions on the size of burrows cannot be drawn on a longer time scale from these results.

5.5.4 How does population provenance affect the propensity of crayfish to burrow?

Burrowing was recorded by crayfish in each population examined, showing that all crayfish have the capacity to burrow, even though the field sites from which they were obtained showed significant differences in burrowing, including no evidence of burrowing at all. There was no significant difference observed between the two UK populations (one from a burrowed stream, and one from a river with no evidence of burrowing), suggesting that burrowing therefore may be an innate as opposed to a learned behaviour. This suggests that should signal crayfish be spread to a new river, they have the propensity to burrow regardless of the population that they were introduced from. There were also no significant differences in burrowing between the UK populations and signal crayfish collected from their native range, which is interesting because burrowing has never been recorded in riverbanks across the native range. This suggests that the native population have an innate capacity to burrow and do so in response to drivers, such as a lack of shelter, or an increased population density, as manipulated in these experiments. Lack of burrowing in the field may reflect environmental conditions which limit the need or ability to burrow, including typically coarse bed material size and shallow, rocky banks in their native range.

194

However, signal crayfish from the recently invaded East Gallatin River constructed significantly larger burrows and burrowed significantly more often than any of the three other populations in the mesocosm experiments. This is the case, even though the physical characteristics of the East Gallatin were not substantially different from the other UK rivers surveyed. The riverbank profiles of the East Gallatin were consistent with those found in the UK, consisting of cohesive bank material that formed a steep bank, with many crayfish specimens being collected that were directly walking on the riverbank, but no burrows were found when surveyed. This may be a result of a low population density (the catch per unit effort from attempted trapping was zero). The population in the East Gallatin was only recently introduced, and visual searching and hand searching confirmed the very low population density of signal crayfish relative to the other tested sites. It may be that the population density of signal crayfish at the East Gallatin does not yet exceed the availability of alternative shelters, and so crayfish have not yet had to resort to burrowing as a last resort activity.

It is possible that the burrowing behaviour could be driven by biological factors such as the threat of predation. Signal crayfish alter their behaviour in the presence of predatory fish (Blake and Hart 1993; Blake and Hart 1995; Hirvonen et al. 2007). Whilst crayfish that were naïve to predators did not increase their shelter use (Hirvonen et al. 2007), both visual and chemical stimuli of known predators led to an increase of shelter use (Blake and Hart 1993), and in the absence of alternative reasonable shelters, it could be hypothesised that an increased predatory pressure could lead to increased burrowing as a response.

These experiments investigated adult signal crayfish, and there are very few documented fish species that can predate adult signal crayfish in the UK. Whilst eels (Anguilla anguilla), perch (Perca fluviatilis), brown trout (Salmo trutta), barbel (Barbus barbus), pike (Esox lucius), zander (Stizostedion lucioperca), and chub (Squalius cephalus) have all been documented to consume signal crayfish (Soderback 1994; Blake and Hart 1995; Nystrom et al. 2006; Reynolds 2011; Basic et al. 2015; Wood et al. 2017), evidence focusses largely on the predation pressures on juvenile crayfish, with only pike being recorded as being able to predate upon adult signal crayfish. Avian predators however have a stronger success rate of predating on adult crayfish. In the UK, lesser black-backed gull (Larus fuscus), grey heron (Ardea cinera), great crested grebe (Podiceps cristatus), coot (Fulica atra), and kingfisher (Alcedo atthis) (Bubb et al. 2008; Mortimer et al. 2012) have all be recorded predating on crayfish, with all but kingfishers being recorded predating on adult crayfish. However, grey heron, and many species of functionally

195 similar diving ducks are widely spread throughout Montana and would likely exert the same predatory pressures on signal crayfish within their native and introduced range. However, no burrows were found at the East Gallatin site, and so whilst the East Gallatin crayfish burrowed the most in mesocosm experiments, this was not reflected in the field where predation pressures were direct, and so these data are unable to support this theory.

Signal crayfish are also highly cannibalistic (Houghton et al. 2017), with adults heavily predating upon young, and so if burrowing is a response that helps increase signal crayfish population success, and is a response to predation, this may elicit a positive feedback cycle. American mink (Neovison vison) and Eurasian otter (Lutra lutra) have also been recorded to predate on crayfish (Melero et al. 2014; Britton et al. 2017), but their small range is unlikely to have impacted the burrowing behaviour of crayfish throughout the UK. However, American mink and racoon (Procyon lotor) are both widely distributed throughout Montana and have been identified as natural predators of signal crayfish (Englund and Krupa 2000), and so may exert the same predatory pressures on signal crayfish within their native and introduced range.

The East Gallatin, where burrowing was greatest in the mesocosm experiments, was the only site to also have another species of crayfish recorded to be present (virile crayfish, Orconectes virilis). Signal crayfish that inhabit rivers where no other species of crayfish are present have been shown to be less aggressive in competitive interactions than those from rivers where crayfish are sympatric to other crayfish species (Pintor et al. 2008). If burrowing is a defence behaviour, it could be argued that crayfish from the East Gallatin, which may be less aggressive due to their coexistence with virile crayfish, burrowed more as a result of a defensive behaviour. Crayfish from the East Gallatin also constructed burrows significantly more quickly than crayfish from any other river, which would support the hypothesis that the population from the East Gallatin were burrowing in a defensive manner and using their burrows as an emergency shelter.

Table 5.7 compares the presence of predatory fish species found in the four rivers that crayfish for these experiments were collected from. All fish listed have either been recorded consuming signal crayfish or are functionally very similar to species that have been recorded predating on crayfish. A similar pattern is witnessed between Gaddesby Brook (UKB; lowest), the River Etherow (UKNB; central value), and the East Gallatin (USI; highest) between the volume of sediment that was excavated in the burrowing experiments (Figure 5.9) and the predation pressure from fish. However, crayfish from the Bitterroot and Clark Fork River (USN) were

196 also subject to the same number of predatory species and burrowed significantly less than USI crayfish (East Gallatin River) in the mesocosm experiments. Whilst there is some evidence here to suggest that predation may have an impact on the propensity of crayfish to burrow, this has not been empirically proven, and cannot be asserted.

Bitterroot and Gaddesby Brook River Etherow East Gallatin Clark Fork River

UKB UKNB USI USN

Centrarchidae Smallmouth Bass Micropterus dolomieu - - - Present Largemouth Bass Micropterus salmoides - - - Present Cyprinidae Barbel Barbus barbus - Present - - Chub Squalius cephalus Present Present - - Esocidae Northern Pike Esox lucius Present Present Present Present Lotidae Burbot Lota lota - - Present - Percidae Yellow Perch Perca flavescencs - Present Present Present Walleye Sander vitreus - - Present - Salmonidae Yellowstone Cutthroat Trout Oncorhynchus clarkii bouvieri - - Present Present Westslope Cutthroat Trout Oncorhynchus clarkii lewisi - - Present Present Rainbow Trout Oncorhynchus mykiss - Present Present Present Brown Trout Salmo trutta Present Present Present Present Bull Trout Salvelinus confluentus - - - Present Brook Trout Salvelinus fontinalis - - Present Present Arctic Grayling Thymallus arcticus - Present Present -

Total Predatory Species 3 7 10 10

Table 5.7: Predatory fish that may consume crayfish found in the rivers where crayfish were collected. Distribution data source: Montana Field Guide 2019 and Environment Agency 2016.

The population in the East Gallatin is recently established, with only two recorded signal crayfish sightings in the previous 15 years (Montana Field Guide 2019) and are present in a very short length (circa 100 m) of river. Individuals collected were therefore from the invasion front and are likely to have been bred from a very small genetic pool. It is therefore possible that the expression of the burrowing behaviour could be a result of genetic bottlenecking, the process where the establishment of a small population, such as an invading population, has an initially reduced gene pool. This smaller population then faces higher levels of genetic drift, with random mutations often leading to new genetic expression. Genetic bottlenecking is commonly predicted to be associated with biological invasions (Barrett and Richardson 1986)

197 and may be responsible for influencing behavioural characteristics (Holway and Suarez 1999). Invasions frequently constitute rapid evolution events (Reznick and Ghalambor 2001), with many examples of rapid evolution involving introduced species (Thompson 1998; Whitney and Gabler 2008). It has been suggested that this could be a result of genetic bottlenecking that leads to increased genetic variance, thus promoting rapid evolution (Naciri-Graven and Goudet 2003) – the process where genotypic and phenotypic shifts occur in just a few generations.

Genetic sequencing has indicated that there is a large genetic diversity of signal crayfish throughout Europe (Petrusek et al. 2017), which is likely to represent the result of multiple small implants of signal crayfish into isolated systems, originating from potentially multiple subspecies (Petrusek et al. 2017). Previous studies have demonstrated that the implanting of a small number of individuals can result in rapid physiological adaptations, such as Anole Lizards (Anolis sp.) showing significant physiological adaptation to vegetation traits after just ten to fourteen years following the introduction of either five or ten individuals to a new island in the Bahamas (Losos et al. 1997). These rapid onsets of physiological changes are likely to coincide with rapid change in behavioural adaptation. The genetic bottlenecking of each of the genetically varied European signal crayfish populations may allow for the rapid change in behaviour, and explain why the burrowing behaviour of signal crayfish has been found in some populations, such as those empirically recorded in the UK (e.g. Guan 1994; Faller et al. 2016), and anecdotally reported in Switzerland (A. Gaskov pers. coms.) and Austria (A. Pichler pers. coms.), but not others.

However, the theory of genetic bottlenecking does not explain why signal crayfish populations stop, or slow down, burrowing. If genetic bottlenecking promoted the expression of a burrowing behaviour, a binary response to burrow or not to burrow could be hypothesised, and populations where burrowing is an evidently expressed behaviour, such as the UKB population, would continue burrowing at the same rate post this behavioural shift. However, the median burrow sizes constructed by crayfish in mesocosms decreased with the time since they were introduced to their collected field site (Figure 5.9), with the crayfish from Gaddesby Brook (UKB), which have been present for 17 years at the time of collection, and are the most established of the three populations, being similar to background rates of the native population.

Signal crayfish at the invasion front are less aggressive than crayfish at the population core (Hudina et al. 2015), and it may be that less aggressive individuals are more inclined to burrow than more aggressive individuals. This is supported by the crayfish from the East Gallatin,

198 which could be hypothesised to be less aggressive due to sympatric coexistence with virile crayfish, burrowing significantly more than the other three populations which did not coexist with another crayfish species. This is also supported by evidence from field surveys from Chapter 3; a higher proportion of crayfish were damaged in rivers where burrows were absent than rivers where burrows were present, with the most heavily burrowed rivers having the lowest injury rates of crayfish, although the difference was not significant (Table 5.8).

% Crayfish with Mean Damage % Crayfish % Crayfish with Chelae Antennae Damaged or Score Damaged Damaged or Missing Missing No Burrows 0.728 46.9 33.3 18.5 Low Burrows ( < 0.4 m -1 ) 0.517 34.8 29.1 12.2 High Burrows ( > 0.4 m -1 ) 0.455 31.3 26.1 6.8 F 0.406 0.362 0.789 0.892

p 0.670 0.700 0.465 0.422

Table 5.8: Associations between damaged crayfish and the presence of burrows. Data were collected in Chapter 3, but are not presented. ‘Damage Score’ is calculated by adding one ‘point’ for every damaged leg, chelae, antennae, or broken carapace. No differences between groups are significant.

5.5.5 Discussion Summary

These results clearly indicate that crayfish burrowing activity is associated with the presence of alternative available shelter, with the associated burrowing being mediated by population density. There were also significant differences witnessed between populations, with larger burrow sizes being associated with more strongly established crayfish communities. These results can therefore lead to the hypothesis that burrowing occurs mostly at the invasion front, and then decreases as the population settles to a sustainable level (Figure 5.18).

199

Figure 5.18: Conceptual diagram for the propensity for crayfish to burrow after invasion. At the initial invasion, there is little need for crayfish to burrow, as shelter availability exceeds population density. As population density rapidly increases, the chance of occupying an alternative shelter is rapidly decreased, and so burrows are constructed as an alternative shelter. As the population stabilises, the propensity to burrow is now reduced, as there are a larger number of shelters to compete for, due to the presence of burrows in the river, as well as abiotic shelters. When the population has stabilised to a level equalling shelter availability, the propensity to burrow is reduced, but crayfish are still required to burrow, as bank erosion and collapse continually reduce the number of burrows present in the system.

The letters on the diagram represent the four populations tested in these experiments. (a) represents the Bitterroot and Clark Fork River, where crayfish are native, (b) represents the East Gallatin, where crayfish have recently invaded (4 years); (c) represents the River Etherow (20 years), and (d) represents Gaddesby Brook, where crayfish population densities are now stable, but still maintaining a burrowed bank.

By increasing the number of alternative shelters available, burrowing may increase the carrying capacity of the system, allowing for greater population densities of signal crayfish to be maintained than without.

200

These experiments allow for a better understanding of how behaviours of invasive species change after invasion. Whilst conclusions cannot be drawn on rapid evolution, as evolution implies that the change (a) is genetic, (b) is inherited, (c) is irreversible, and (d) leads to a greater success of breeding and establishment of the population, these experiments do allow for the process of rapid behavioural change to be commented on, and the assertion that signal crayfish have extreme behavioural plasticity, which is present – if not consistent – across all populations. Behavioural plasticity has been demonstrated to be highly important in avian invasions (Sol and Lefebvre 2000), and has been suggested to be an important factor in the success of crustacean invasions (Weis 2010). These results demonstrate the high behavioural plasticity of signal crayfish, which appears to be greatest at the invasion front. Whilst burrowing can be undertaken by the native US population, but is not expressed in natural conditions, this extreme behavioural plasticity after invasion is the first incidence of a completely new behaviour has been quantified.

201

Chapter 6

The impacts of signal crayfish burrows on riverbank retreat: a biophysical sediment budget

202

6.1 Introduction

Burrowing can directly recruit sediment into river systems (Guan 1994; Harvey et al. 2011; Harvey et al. 2014; Rice et al. 2014; Faller et al. 2016; Rice et al. 2016; Haussmann 2017; Chapter 3). However, the measurement of burrows at a point in time does not allow for these figures to be considered temporally. In particular, burrowing delivers sediment directly into the channel, but may also affect sediment loading by causing accelerated bank retreat and mass failure (Harvey et al. 2019), so the total effect of burrowing occurs over a period the exceeds the burrowing activity itself. Evidence that burrowing can accelerate bank erosion comes from ex situ experiments (Onda and Itakura 1994; Vu et al. 2017), and physical (Viero et al. 2013; Saghaee et al. 2017) and numerical modelling (Camici et al. 2014; Orlandini et al. 2016; Borgatti et al. 2017), but there are little ex situ data. Semi-quantitative studies associate the presence of red swamp crayfish (Procambarus clarkii) burrows with bank failure in rice paddies in the Iberian Peninsula (Barbaresi et al. 2004; Arce and Dieguez-Uribeondo 2015), but the influence of signal crayfish burrows on bank retreat and collapse has been reported anecdotally: river bank retreat was reported to have increased by 1 m a-1 after a signal crayfish invasion in the River Lark, Suffolk (West 2010); crayfish burrows were reportedly responsible for the collapse of a college wall in Oxford (Telegraph 2016); and bank collapse at Gaddesby Brook has been attributed to crayfish burrows (Sibley 2000). There is also evidence to suggest associations between crayfish burrows and bank shape, with burrows being associated with undercutting and increased erosion (Faller et al. 2016), and change in bank shape may accelerate sediment recruitment from riverbanks (Simon and Collison 2001; Fox et al. 2007). However, there are no empirical data exploring the link between burrows and bank erosion. Harvey et al. (2019) highlight the lack of quantitative research into the physical damage to riverbanks as a result of burrowing by invasive species and call for long term monitoring projects to assess associated processes beyond the instantaneous input of sediment excavated by burrowing. This chapter explores the effects of burrows and their characteristics on bank retreat along two UK rivers invaded by signal crayfish over a 22 month monitoring period. Erosion pins were used to establish the relative effects of retreat caused by mass failure and diffuse erosion.

203

6.2 Aims

This chapter aims to address aspects of objectives 3 (to examine the processes by which crayfish burrowing impacts the geomorphology of riverbank erosion), 4 (to quantify the effect of crayfish burrowing on volumes of riverbank erosion and investigate the relative roles of direct sediment input from crayfish burrows, accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows in recruiting sediment to invaded river channels) and 5 (to construct models to predict the presence, extent, and geomorphic impacts of burrowing on UK rivers). A field-based monitoring study was undertaken, which considered four key questions:

1. Do signal crayfish burrows accelerate bank retreat?

Whilst there have been reports of signal crayfish burrows accelerating bank retreat, there are currently no data documenting the rate of change from baseline conditions in the absence of crayfish burrows. As signal crayfish burrowing is a behaviour witnessed in rivers across the UK (Chapter 3), and has been recorded from tested native and invasive populations, irrespective of whether burrows were present in rivers where they were collected (Chapter 5), it is important to quantify the impacts that burrowing may have on affected riverbanks.

2. How do signal crayfish burrows accelerate bank retreat?

If signal crayfish burrowing causes banks to retreat, it is important to understand the mechanisms involved and the timescales over which they occur. This study therefore aims to specifically investigate the effect of crayfish burrows on:

(i) Diffuse erosion. Do crayfish burrows accelerate the rate of bank retreat when collapses are absent or excluded from consideration? (ii) Bank collapse. Do crayfish burrows increase the size of individual collapses, or the likelihood of a collapse occurring? (iii) Bank shape. Do crayfish burrows cause bank steepening?

3. How much sediment does accelerated bank retreat associated with crayfish burrows recruit to river systems?

204

If crayfish burrowing does accelerate bank retreat, this may form a substantial mass of sediment in comparison to the initial burrow construction. It is therefore important to use associations observed in question 2 to construct a regression model to estimate the total mass of sediment that crayfish burrows recruit to river systems.

4. What are the relative roles of direct sediment input from crayfish burrows, the accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows in recruiting sediment to invaded river channels?

To contextualise raw mass values of sediment input, it is necessary to understand the relative contribution of these values to assess the importance of their impact on system wide processes.

6.3 Methods

These questions were investigated by monitoring the nature and amount of riverbank retreat at a selection of sites along two infested UK rivers, where erosion pins were installed into riverbanks with varying densities of crayfish burrows.

6.3.1 Field Site Selection

Erosion pins were installed into seventeen banks across two signal crayfish infested rivers, the River Bain, Lincolnshire (TF 2457 7947), and Gaddesby Brook, Leicestershire (SK 7312 1016). The two rivers were selected to capture between-catchment variability across two morphologically and geographically similar (UK lowland) streams. Lowland streams were selected as these are most typically associated with signal crayfish invasion (e.g. Guan and Wiles 1997; Guan and Wiles 1999; Harvey et al. 2014; Rice et al. 2014; Cooper et al. 2016; Houghton et al. 2017). Both rivers have established signal crayfish populations, with crayfish burrows recorded at both sites (Sibley 2000; Peay 2001; Johnson et al. 2014). Reaches were selected that were internally homogenous, to reduce variance within the reach that may influence erosional factors other than those associated with crayfish burrows. In each river a homogenous reach without tributaries was identified, in which broadly consistent bed materials, slope, land use, hydrology and channel geometry isolated variations in burrow density as the primary within-reach variable that could potentially affect bank erosion. The data for each river was analysed, and results are presented, independently throughout the study.

205

Gaddesby Brook is a small, lowland stream, with a bed of alluvial gravel overlain by a deep (up to 0.2 m) layer of clay and silt. The catchment upstream of the study reach is approximately 2 29 km , is predominantly arable and horticultural land cover (51.5%; National River Flow Archive 2020), and lies predominantly over mudstone and siltstone with superficial deposits of alluvium and Pleistocene glaciolacustrine clay and silt deposits, with the study reach being situated approximately 8.5 km from the stream source. The study reach is approximately 100 m long, the mean channel width is 2.7 m, the mean bankfull height is 0.87 m, and the river slope is 0.003. Whilst there is no recorded date of invasion for the Gaddesby Brook study site, crayfish first invaded the stream 7 km upstream of the study site in 1985 (Belchier et al. 1998), and were recorded reaching 7 km downstream in 1998 (Sibley 2000), and so the date of invasion at the study site can be estimated to be circa 1992. This is consistent with Gaddesby Brook population expansion estimates of 1 km a-1 (Harris and Young 1996) and recorded sightings of signal crayfish from the National Biodiversity Network (2019), which first recorded the presence of signal crayfish in the reach in 1993.

The River Bain is a small, lowland river and the study reach consists of a bed of alluvial gravel overlain by a deep (up to 0.5 m) layer of clay and silt. The catchment upstream of the study reach is approximately 66 km2, is predominantly arable and horticultural land cover (76.7%; National River Flow Archive 2020), and lies predominantly over Cretaceous chalk with superficial deposits of Pleistocene till, with the study reach being situated approximately 18.6 km from the river source. The study reach is approximately 1.1 km long, the mean channel width is 3.4 m, and the mean bankfull height is 1.03 m. Crayfish were introduced to a pond in the catchment in the 1970s (Johnson et al. 2014), and were first recorded 7 km upstream in 1993 (Holdich et al. 1994), and so invasion at the study reach can be estimated to circa 1997, using the generic crayfish invasion rate calculations outlined in Chapter 3.

Banks were selected for study that were cohesive, bare of vegetation, and on straight sections of channel (Figure 6.1; Figure 6.2); meander bends were avoided, as they are more erosionally active, and thus would allow for poor comparisons between burrowed banks. Banks were selected to represent a high range of burrow densities, ranging from 0 burrows m-2 at both sites, to a maximum density of 4.8 burrows m-2 at Gaddesby Brook, and 6.15 burrows m-2 at the River Bain.

This study aims to quantify the impact of crayfish burrows on bank retreat, but it is also plausible that erosion observed could be a result of crayfish directly eroding sediments from

206 the bank by their movements and activities outside of burrows. Significant associations between burrow and crayfish population densities have previously been recorded (Arce and Dieguez-Uribeondo 2015; Chapter 3), and so the density of burrows on a bank may simply be a proxy for the density of signal crayfish. However, crayfish have been shown to have wide daily movement patterns, with radio tagging recording median and maximum daily movements of 13.5 - 15 m and 407 m respectively in upland rivers (Bubb et al. 2004), and 24 m and 51 m respectively in a lowland river (Johnson et al. 2014). As a result, the movement of crayfish would likely have a uniform impact on sites throughout the reach. This is particularly true for Gaddesby Brook, where the total reach length considered was approximately 100 m. Therefore, the impact of bioturbation of sediments by activities such as foraging and fighting can be considered to be consistent across sites, and therefore their impact can be discounted, with any associations being attributed to differences in burrow characteristics, as opposed to crayfish activity.

Figure 6.1: Studied banks were evenly spatially distributed throughout the study reaches.

207

Figure 6.2: Banks selected for study were bare of vegetation and on straight sections of channel. Pins were left with 30-50mm exposed for ease of relocation. Examples of banks selected for study on (a) Gaddesby Brook and (b) the River Bain.

6.3.2 The Use of Erosion Pins

Erosion pins are metal rods inserted horizontally into riverbanks at an angle perpendicular to the channel to allow lateral erosion to be quantified. If the pin remains in a known and steady location, retreat around the pin will expose the pin further than at a previous recording, and the difference in exposure then records bank retreat (Figure 6.3).

Erosion pins were chosen for use as they are a simple, affordable mechanism that can be applied to large areas across a wide range of fluvial environments (Table 6.1; Lawler 1993; Lawler et al. 1999; Laubel et al. 2003). Photo-electronic Erosion Pins (PEEP; Lawler 1992) were considered for use, but were not appropriate for this study, as PEEP are expensive, and whilst they give accurate recordings of when erosion events occurred, this study was more concerned with covering a wide area to examine erosion spatially. As well as being affordable when large

208 total areas require monitoring, standard erosion pins are able to detect small amounts of erosion at very local scales (Thorne 1981), and are thus considered the most appropriate technique for short to medium term monitoring studies (Foucher et al. 2017). The application of digital photogrammetry and structure from motion (SfM) was considered, but SfM is unlikely to provide accurate results for partially vegetated banks (Bird et al. 2010; Micheletti et al. 2014), with vegetation having to be removed for accurate recordings (Jugie et al. 2018), and it is not possible to quantify retreat from banks that are submerged or partially submerged underwater through the use of repeat images (Jugie et al. 2018). Erosion pins however, within the bounds of accessibility and visibility, allow erosion to be recorded regardless of the presence of vegetation or the depth of water on the bank, and have been shown to give comparable results to SfM monitoring of the same reach (Jugie et al. 2018) without the complications of image post-processing. Erosion pins were therefore chosen for use for this research.

Figure 6.3: Schematic of erosion pins. Erosion pins are installed into a riverbank (left), and

the exposed length (L1) is measured of each pin. After a period of time, the bank

has retreated (right), exposing more of the erosion pin. The exposed length (L2) is again measured for each pin. Riverbank retreat is calculated by subtracting

L1 from L2 to report erosion for each pin.

209

Study Focus Location Total Pins Used Banks Monitored Pins Per Bank Pin Resolution

Lowland: Gaddesby Brook and River 0.3 m vertically Current Study Impact of crayfish burrows on bank retreat 430 17 28 Bain, UK 0.5 m horizontally

4 PEEP 1 N/A Lawler 1992 PEEP methodology Upland: Arrow and Upper Severn 4 Traditional not stated N/A -

Fouchier et al. 2017 Quantifying long term erosion Lowland: Masniers River, Paris 258 43 6 Not stated

Henshaw et al. 2013 Dominant erosion drivers Upland: Upper Severn headwaters 110 27 4 150 mm vertically

Kronvang et al. 2013 Quantifying erosion rate Lowland: River Odense, Denmark 3000 176 17 Not stated

Palmer et al. 2014 Quantifying severely eroding streambanks Lowland: Walnut Creek, Iowa 10-70 per bank 1.0 m vertically

Upland: Nant Tanllwyth; Wales 103 17 6 0.3 m vertically Couper et al. 2002 Processing of negative recordings Upland: Afon Trannon 60 1 60 1.0 m horizontally Upland: River Arrow 278 4 70 0.5 m vertically Jugie et al. 2018 Comparison with structure from motion Lowland: Meraintaise River, France 65 6 11 2.0 m horizontally Murgatroyd and Ternan 1983 Erosion of forested and non forested stream banks Upland: Narrator Brook, Dartmoor 100 100 1 N/A 0.3 m vertically Gardiner 1983 Impact of frost action on bank erosion Lowlad: River Lagan, Belfast 90 3 30 0.6 m horizontally Lawler et al. 1999 Change in erosion rates Lowland: Swale-Ouse system, UK 421 10 42 (see paper for details)

Simon et al. 2000 Near bank erosion processes Upland: Goodwin Creek, MS, USA 21 7 3 0.6 m vertically

Veihe et al. 2011 Types of erosion occurring Lowland: Harrested, Zealand, Denmark 200 5 40 0.25 m spacing 0.2 5m vertically Laubel et al. 2003 Quantifying erosion rate Lowland: 15 streams, Denmark 1500 100 15 0.5 m horizontally

Table 6.1: A selection of previous published studies using erosion pins to quantify riverbank erosion.

210

Whilst erosion pins can be deployed for a versatile set of environmental conditions, Couper et al. (2002) outlined four limitations to their use:

(i) Pin movement. The use of erosion pins assumes that the pin is stationary in position, and that its position in the bank has not changed between readings. Any pin movement by hydraulic force, the influence of vegetation, or frost action will influence the accuracy of erosion estimates. If the pin is lost altogether, this can leave gaps in the dataset. (ii) Changes in elevation of the bank/slope surface. The advance and retreat of banks may occur independently of erosional and depositional processes, by swelling from changing moisture content, freezing, and thermal expansion. These may exaggerate or mask the extent of recorded erosion. (iii) Influence of pin on erosion. The presence of pins may influence erosional processes (Lawler 1978, 1993), and the act of inserting pins into the bank may affect erosional mechanisms, and most likely accelerate erosion. If pin materials rust, the rust can bind soil to the pin (Bridges and Harding 1971), resulting in localised soil not being eroded at the same rate as the bank that the pin is inserted into, resulting in conservative erosion readings. The insertion of pins into a cohesive bank may also reinforce the bank material, thus potentially decreasing erosion rates (Thorne 1981). (iv) Human interference. The act of measuring pins, may itself increase localised erosions, and vandalism to pins may influence the measured rates of erosion.

These four potential problems were accounted for by: (i) Pin movement. The location of missing pins were recorded and bank erosion features were characterised. If a hole was present where the pin was installed, then the data point was discounted from the dataset (Figure 6.4c). However, if evidence of bank collapse was present (Figure 6.4) it was assumed that the pin had been removed by the bank collapse and the exposed pin length was recorded as the full length of the pin, as in Palmer (2014). This effectively limits the estimate of erosion to the length of the pin, such that large failures might be under-recorded. However, this method provides the most accurate readings attainable, and allows for the underlying processes to be recorded. (ii) Changes in elevation of the bank/slope surface. Rivers were selected that were relatively close to one another (103 km apart), so that environmental conditions

211

were similar, (in lowland England on similar lithologies, in catchments with similar land cover and climates), and so thermal influences would be relatively consistent between sites. (iii) Influence of pin on erosion. Little rusting was observed, and all banks measured were constructed of cohesive material, and so erosion values were treated as conservative estimates. (iv) Human interference. Sites were selected that were rural, not on a public footpath, and not subject to active channelization or vegetation removal regimes in order to reduce the chance of human interference of banks or erosion pins.

6.3.3 Erosion pin installation

Welding rods 350 mm in length and 2.5 mm in diameter were used as erosion pins, as they have been used extensively in other studies (e.g. Lawler et al. 1999; Couper et al. 2002). The use of welding rods minimises the likelihood of anomalous readings associated with pin movement and the influence of the pins on erosion due to their known rusting rates (Couper et al. 2002).

Pins were installed by pushing them into banks perpendicular to the channel, leaving 30 – 50 mm exposed for ease of relocation. Pins were installed in a grid formation, separated by 0.3 m vertically, and 0.5 m horizontally. The lowest row of pins was installed along the low water mark on the bank, and the bank was pinned vertically up to the top of the bank.

317 pins were installed into 11 banks on the River Bain, and 164 pins were installed into six banks at Gaddesby Brook (Table 6.2).

6.3.3.1 Inclusion of Bank Bain 11

The number of burrows, the burrow volume, and the burrow entrance area at Bank 11 on the River Bain were at least twice that of the next most burrowed bank, and 4.26 times that rd considering burrow volume. This bank is statistically outlying (3.83 * IQR > 3 quartile BL rd and 5.82 * IQR > 3 quartile VL). However, this bank is included in analysis as the distribution of crayfish burrows is highly clustered (Stanton 2004; Faller et al. 2016; Chapter 3), and so the spread of banks considered here represent the typical distribution of burrows throughout a

212 reach. Stanton (2004) observed clustering of burrows in Gaddesby Brook, with the modal -1 -1 burrow density being 0 m of riverbank, with some sites exceeding 9 burrows m , and Faller et al. (2016) observed concentrated distributions of burrows, which were recorded in 0.2 - 23.5% (median 3.2%) of the length of reaches surveyed. Field surveys (Chapter 3) also found burrow distribution to be highly clustered throughout reaches. Two sites were surveyed where burrows exceeded a density of 1 burrow per meter of riverbank for the full survey (Gaddesby Brook (1.02 burrows m-1) and Broadmead Brook (1.05 burrows m-1)), but the mean distance between burrows was 0.27 m and 0.25 m respectively, which equates to mean burrows clustering to just 27.1% and 26.1% of available riverbank length, with many areas of very high burrow density being recorded. Therefore, this point represents a typical bank in a river where signal crayfish burrows are present, and is therefore legitimately included in analysis.

Bank Length Bank Area Number of Number of Burrow Volume Burrow Entrance River Bank Number (m) (m2) Pins Burrows (m3 x10-3) Area (m2 x10-3)

River Bain 1 5.5 5.1 34 0 0.00 0.00 2 3.0 3.6 24 0 0.00 0.00 3 3.5 3.6 24 2 0.95 4.71 4 4.5 4.2 28 4 2.27 12.72 5 3.0 3.3 22 4 4.74 13.19 6 3.0 3.5 23 6 2.54 12.56 7 3.5 3.5 23 6 3.25 11.82 8 3.5 3.0 20 7 1.77 9.26 9 3.5 4.5 30 7 2.85 13.27 10 3.5 3.2 21 12 3.73 20.96 11 3.5 3.9 26 24 15.91 51.26

Gaddesby Brook 1 5.0 3.9 26 0 0.00 0.00 2 4.5 4.8 32 5 5.74 85.33 3 4.5 3.6 24 7 6.37 86.74 4 4.0 3.0 20 14 8.35 123.60 5 4.0 3.2 21 17 19.15 62.80 6 4.5 4.8 32 18 14.48 58.80

Table 6.2: Table of banks used in study.

213

Figure 6.4: Erosion pin monitoring on the River Bain: (a) a bank collapse, (b) a bank collapse; (c) the hole where an erosion pin was inserted, but has since been lost, (d) a block of material collapsed into the river with the erosion pin still attached and (e) a bank slump into the river with the erosion pin still attached.

214

6.3.4 Quantifying Burrows

Burrow depths and the width and height of burrow entrances were measured using a meter rule to the nearest 5 mm. Burrow depths were measured at the centre of the opening, to account for a sloping bank face. Volume of sediment excavated was calculated by treating the burrow shape as an elliptical cylinder as in Faller et al. (2016):

퐴퐸 = 휋 (푊⁄2 퐻⁄2) (Equation 6.1)

where AE is burrow entrance area, W is the burrow entrance width, and H is the entrance height.

푉퐵 = 퐴퐸 퐿 (Equation 6.2)

where VB is burrow volume, and L is the length of the burrow.

The number of burrows (nB), total entrance area (∑AE), and total volume of sediment excavated from burrows (∑VB) were calculated for each bank and normalised by respective bank area (A) and bank length (L), yielding six burrow metrics (Table 6.3):

-2 (i) Burrow density per unit bank area (BA = nB / A; burrows m ). (6.3)

-1 (ii) Burrow density per unit length (BL = nB / L; burrows m ). (6.4)

(iii) Burrow entrance area per unit bank area (EA = ∑AE / A; % cover). (6.5)

2 -1 (iv) Burrow entrance area per unit bank length (EL= ∑AE / L; m m ). (6.6)

3 -2 (v) Burrow volume per unit bank area (VA = ∑VB / A; m m ). (6.7)

3 -1 (vi) Burrow volume per unit bank length (VL= ∑AE / L; m m ). (6.8)

Burrow volume was considered because this is likely to be affect bank geotechnical and hydrological properties, and burrow entrance area was considered because this may increase turbulence at the bank face, and thus directly amplify the entrainment of sediment (Harvey et al., 2019). Bank area as well as bank length was used to normalise these quantities in order to include bank height given its role in determining the weight of material above the burrows, and so propensity to fail (Fredlund et al., 1978; Fredlund and Rahardio 1993).

215

-2 -1 3 -2 -4 3 -1 -4 2 -1 -3 River Bank Number B A (burrows m ) B L (burrows m ) V A (m m x10 ) V L (m m x10 ) E A (%) E L (m m x10 ) River Bain 1 0.00 0.00 0.00 0.00 0.00 0.00 2 0.00 0.00 0.00 0.00 0.00 0.00 3 0.56 0.57 2.63 2.71 0.13 1.35 4 0.95 0.89 5.40 5.04 0.30 2.83 5 1.21 1.33 14.36 15.80 0.40 4.40 6 1.74 2.00 7.36 8.47 0.36 4.19 7 1.74 1.71 9.42 9.28 0.34 3.38 8 2.33 2.00 5.91 5.07 0.31 2.65 9 1.56 2.00 6.34 8.15 0.29 3.79 10 3.81 3.43 11.84 10.66 0.67 5.99 11 6.15 6.86 40.81 45.47 1.31 14.65

Gaddesby Brook 1 0.00 0.00 0.00 0.00 0.00 0.00 2 1.04 1.11 11.96 12.76 1.78 18.96 3 1.94 1.56 17.69 14.15 2.41 19.28 4 4.67 3.50 27.85 20.89 4.12 30.89 5 5.40 4.25 60.80 47.88 1.99 15.70 6 3.75 4.00 30.16 32.17 1.22 13.07

Table 6.3: Calculated burrow metrics of burrow density standardised for bank area (BA; -2 -1 burrows m ) and bank length (BL; burrows m ), burrow volume density 3 -2 3 -1 standardised for bank area (VA; cm m ) and bank length (VL; cm m ), and 2 -2 burrow entrance area density standardised for bank area (EA; cm m ) and 2 -1 bank length (EL; cm m ) for all surveyed banks.

6.3.5 Data Collection

Pins were installed into the River Bain in September 2017, and into Gaddesby Brook in October 2017. Data were collected over a 22 month period (647 days at Gaddesby Brook, and 677 days at the River Bain). Pins were scheduled to be recorded approximately every four months; but this was not possible throughout the study period because of inaccessible river conditions, and the ability to locate submerged pins (Figure 6.5). A total of five repeat readings were taken at the River Bain over the 22 month period, and four at Gaddesby Brook. Not all banks were able to be surveyed at every time step as water depth adjacent to individual banks varied longitudinally, so that access varied between sites at high flow. Thirty-eight individual bank recordings were obtained from the River Bain, and 20 from Gaddesby Brook.

216

Figure 6.5: Gaddesby Brook is a flashy system, and so fieldwork was not always possible due to a high water depth which reduced the visibility of submerged pins. Colleague for scale, stood at a 0.1 m deep riffle under high flow conditions.

The length of each individual pin was recorded at each visit, with all pins measured using a standard retractable tape measure (as in Greenwood et al. 2018) to the nearest millimetre, with the length of the pin measured from the bank to the tip of the end of the underside of the pin (Figure 6.6). Erosive features, such as bank collapses, were recorded qualitatively in field notes.

Figure 6.6: An erosion pin being measured with a tape measure.

217

6.3.6 Data Processing

6.3.6.1 Negative Readings

There has been much debate over the treatment of the recording of negative values in erosion pin datasets (reviewed in Couper et al. 2002). Many studies combine positive and negative readings to report morphodynamic activity (e.g. Henshaw et al. 2013; Kronvang et al. 2013), but this does not allow for the nuances in riverbank processes to be explored. Negative pin values were recorded at both Gaddesby Brook and the River Bain on all but one bank. Negative values were retrained in analyses because field observations confirmed that they corresponded to deposition. The methods discussed in Couper et al. (2002) were used to generate two metrics of riverbank retreat:

(i) Gross Retreat: All negative values included in analysis; (ii) Only Retreat: Treat negative values as missing data, allowing for only retreated sections of the bank to be analysed.

Other metrics were considered in Couper et al. (2002), but were not used here;

(i) Activity: The extent of change on the erosion pin, regardless of whether the reading was positive or negative. This was not used, as the resultant data were not appropriate for answering the research questions (see Section 3.8 below) (ii) Replacing all negative values with zero. This was not used, as areas of deposition were observed in the field, and so negative readings were treated as true as opposed to erroneous.

6.3.6.2 Bank Collapses

Bank collapses were visually identified in the field by obvious shear planes and the deposition of coherent blocks or piles of bank material at the bottom of the bank, such as in Figure 6.4. Pins affected by bank collapses were noted in order to monitor the extent and nature of bank collapse at each site.

6.3.6.3 Metrics Used

Six metrics were employed in order to address the research questions:

218

(i) Total Retreat (mm a-1): mean retreat of the bank per year when all pins were considered; (ii) Diffuse Erosion Retreat (mm a-1): mean retreat per year considering only pins that had not been identified as recording a detectable collapse; (iii) Only Collapse Retreat (mm a-1): mean retreat per year considering only pins that had been identified as recording a collapse; (iv) Percentage of Pins Collapsed (% a-1): the percentage of pins from the bank that recorded a collapse per year. (v) Percentage of Retreat as Collapse (%): the percentage of total retreat that could be attributed to collapse. (vi) Relative Change in Riverbank Shape (mm a-1): the difference in retreat recorded between the top and the bottom of the bank, calculated as the mean retreat of the lowest row of pins minus the mean retreat for the highest row of pins. A positive value represents bank steepening, and a negative value represents bank shallowing. This metric was calculated for values of total retreat, and diffuse erosion.

Each of these metrics were calculated for the two methods of considering negative pin readings stated above (see section 3.8 below). Metrics were calculated for each bank as a whole, and for each individual row of pins on each bank, so that the difference between banks, and the difference between the erosion rates vertically on the bank sections could be analysed. Erosion rates were calculated since the previous recording and since the installation of the pins.

6.3.7 Data Visualisation

3D bank profiles were constructed from the erosion pin readings by interpolation onto a 5cm x 5cm grid using a form of nearest neighbour analysis. Known points measured from erosion pins were used to estimate the z coordinate of intermediary cells using a form of nearest neighbour analysis.

푧 = (퐴 + 퐵 + 푅 + 퐿)⁄4 (Equation 6.9)

where z is the calculated z dimension of the cell, A is the above cell, B is the below cell, R is the right hand cell, and L is the left hand cell, through 1,000 iterations.

219

This calculation gives a form of nearest neighbour analysis, where any given grid square is a function of the z coordinate of established measured points (from erosion pins), and is weighted by their distance from the calculated point. Colour ramps were used to visually demonstrate any erosion and accretion that had occurred. This allowed for the change in riverbanks to be visualised both through time and between sites. 2D cross sectional bank profiles were also constructed to visualise differences in vertical retreat. This was undertaken by plotting the mean value of retreat since the start of the monitoring for each row of pins at each time recording.

6.3.8 Statistical Analyses

Erosion estimates for each bank, each time step, and each site were normally distributed (Shapiro-Wilk test; p > 0.10 in all cases), and so parametric tests (Pearson’s r (r) for association and one-tailed Student’s t-test (t) for significance) were used for analysis. A significance value of α < 0.1 was chosen for use, due to the novelty of the research, and as a relatively low number of sites were surveyed for each river (n = 11 and n = 6). A summary of retreat metrics and negative pin recording methods is displayed in Table 6.4. Both rivers were analysed independently throughout. All statistics were performed using SPSS Version 23 (IBM 2015).

6.3.8.1 Do signal crayfish burrows accelerate bank retreat?

To investigate overall bank retreat, the metric of ‘total retreat’ was used in conjunction with the ‘gross retreat’ handling of negative values for each individual bank. The mean values of all pins in a bank were used. The rate of total bank retreat was tested for association with all burrow metrics using Pearson’s r (r) tests. Significance was determined using Student’s t (t) test.

6.3.8.2 How do crayfish burrows accelerate bank retreat?

Diffuse Erosion. To investigate the effect of burrows on processes excluding mass failure, the ‘diffuse erosion’ metric was used in conjunction with the ‘gross retreat’ handling of negative values for each individual bank to investigate the change in retreat across all pins, including deposition; and in conjunction with ‘only retreat’ to investigate the effect of burrows on exclusively eroding areas of riverbank. Whilst crayfish are hypothesised to increase erosion, the consideration of accreting areas is equally important. The mean values of all pins in a bank were used for both data sets. The rate of diffuse erosion was tested for association with all

220 burrow metrics using Pearson’s r (r) tests. Significance was determined using Student’s t (t) test.

Bank Collapse. To investigate the impact of burrows on bank collapse, the ‘only collapse’ metric was used. All collapses recorded were positive, so the treatment of negative values is not applicable for these data, and so ‘gross retreat’ was used for each individual bank. The ‘percentage of pins collapsed’ was also investigated here. As this metric is a binary response (a pin has either recorded collapse or it has not), the treatment of negative values is not applicable for these data, and so ‘gross retreat’ was used. The ‘percentage of retreat as collapsed’ was also analysed. As this was investigating methods of retreat, ‘only retreat’ was used for the treatment of negative values. The mean values of all pins in a bank were used for all data sets. The rate of only collapse erosion, the percentage of pins collapsed, and the percentage of retreat as collapse were tested for association with all burrow metrics using Pearson’s r (r) tests. Significance was determined using Student’s t (t) test.

Change in Bank Shape. To compare the change in bank shape, ‘change in riverbank shape’ was used for each individual bank. The change in riverbank shape was tested for association with all burrow metrics using Pearson’s r (r) tests. Significance was determined using Student’s t (t) test.

Negative Value Methods Gross Retreat Only Retreat Total Retreat a Diffuse Erosion b b Only Collapse Retreat c Percentrage of Retreat as Collapse c Percentage of Pins Collapsed N/A N/A

Metrics Used Metrics Relative Change in Bank Shape d

221

6.3.9 Modelling Sediment Input

A linear regression model was created for each river to estimate the annual total bank retreat and annual mass of sediment input under increasing crayfish burrow densities. The burrow metric with the strongest association with total bank retreat was used to create a model using single linear least squares regression.

Considering question 3, least squares linear regression modelling was undertaken to develop a means of estimating the annual rate of bank retreat R (m a-1) as a function of the burrow metric 3 most strongly and consistently associated with it. Burrow volume per unit bank length VL (m m-1) was the most strongly correlated burrow metric with measured bank retreat R (m a-1) across the full study period and was the only burrow metric significantly correlated with bank retreat on both rivers. Linear regression models for each river were of the form:

푅 = 푎 + 푏 푉퐿 (Equation 6.10)

where, b is the retreat constant (acceleration of retreat due to burrows), VL is the burrow volume (m3 m-1), and a is the baseline retreat rate (the rate of erosion in the absence of -1 burrows, VL = 0; m a ). For the purposes of estimating total reach scale sediment recruitment, (3) was used to calculate an average retreat rate 푅̅, in the presence of

burrows using average burrow volume 푉̅퐿 equal to the sum of all burrow volumes divided by the sum of the all affected bank lengths. The flux of sediment recruited per -1 -1 meter of riverbank with burrows (SR, kg m a ) was then calculated as:

푆푅 = 훾 푅̅ ℎ (Equation 6.11)

where γ is riverbank bulk density (kg m-3), and h is mean riverbank height (m). This -1 -1 rate includes sediment directly excavated by burrowing SB (kg m a ), sediment -1 -1 delivered by erosion in the absence of burrows S0 (kg m a ) and an additional bank erosion surcharge facilitated by burrowing effects S’ (kg m-1 a-1):

푆푅 = 푆퐵 + 푆0 + 푆′ (Equation 6.12)

222

Considering question 4, the constituent fluxes of equation 6.12 can then be independently considered to estimate the masses of sediment recruited to the whole study reach from burrowed banks: (a) from bank erosion in the absence of burrows, (b) directly from crayfish burrow excavation, and (c) from bank erosion facilitated by burrow presence:

푆퐵 = 훾 푉̅퐿 푞 (Equation 6.13)

푆0 = 훾 훼 ℎ (Equation 6.14)

′ 푆 = 푆푅 − 푆0 − 푆퐵 (Equation 6.15)

where q in equation 6.13 is the proportion of observed burrows produced, on average, in a single year. This is not the total number of burrows observed because burrows have a lifespan which exceeds a single year, nor is it number of burrows observed divided by the duration of crayfish occupation because some burrows are lost to erosion. A separate study not presented here (but discussed in Chapter 8), in which >1,500 individual burrows on 5 rivers were monitored for 4 years suggests that the average number of new burrows per year is 56% of the number of burrows present. Here, q is therefore equal 0.56.

These rates can then be converted to mass of sediment recruited per year M (kg a-1) along the entire study reach considering the length L (m) of the reach, the proportion of the reach that exhibits burrowing pB and assuming that both sides of the channel behave similarly:

푀퐵 = 2 푆퐵 퐿 푝퐵 (Equation 6.16)

푀0 = 2 푆0 퐿 (Equation 6.17)

′ ′ 푀 = 2 푆 퐿 푝퐵 (Equation 6.18)

푀푅 = 푀퐵 + 푀0 + 푀′ (Equation 6.19)

From which proportional contributions to the total sediment recruitment can also be determined.

223

Whilst this method provides a means of estimating reach-scale sediment recruitment, it hides substantial within-reach variability in sediment delivery driven by differences of burrow density. R and VL have been measured for each bank along both rivers (Table 6.2; Table 6.3) and can be utilised in bank-scale equivalents of equations 6.11 to 6.19 (with L set to the bank length and pB = 1.0) to obtain estimates of the masses of sediment MB, M0, M’ at that bank scale.

6.4 Results

6.4.1 General Results

Throughout the experiment, erosion pins successfully monitored erosion. The results of data analysis corresponded with visual observations made in the field (Figure 6.7), and erosion was recorded at every riverbank surveyed (Figure 6.8; Figure 6.9).

Erosion rates and patterns varied between banks (Figure 6.8; Figure 6.9), and across the same bank (Figure 6.10) both in space and in time. For bank-averaged values, banks with no crayfish burrows retreated at 0.056 m a-1 at the River Bain and accreted by 0.005 m a-1 at Gaddesby Brook. Across all recordings, riverbank collapse was the dominant method of erosion at the River Bain, with 21.1% of pins recording a collapse per year, which contributed to 57.5% of total erosion recorded. At Gaddesby Brook, only 3.2% of pins recorded a collapse per year, which contributed to 0.6% of total recorded erosion.

6.4.2 Do signal crayfish burrows accelerate bank retreat?

Retreat rates varied between individual banks at each river, and ranged from 0.049 m a-1 (0 burrows m-1) to 0.174 m a-1 (6.9 burrows m-1) at the River Bain, a difference of 253%, and from accreting by 0.005 m a-1 (0 burrows m-1) to retreating 0.025 m a-1 (4 burrows m-1) at Gaddesby Brook (Table 6.5).

Considering the full period of the study, BA, BL, VA and VL showed weak to moderate positive correlations with gross retreat on both rivers (Table 6.5; Table 6.6; Figure 6.11). All burrow metric values were significantly associated with bank retreat at the River Bain, but only VL was significant at Gaddesby Brook. At the River Bain, an increase in one burrow m-1 of riverbank resulted in an increase of 0.017 m a-1, a 30.3% increase in retreat from banks where burrows were absent. The site with the highest burrow density (6.15 burrows m-2 of riverbank)

224 accelerated bank retreat by 181.4%. At Gaddesby Brook, when burrows were absent, accretional processes dominated (-0.005 m a-1), whereas the highest burrow density (4.25 burrows m-1) resulted in erosional processes occurring (0.025 m a-1), with an increase of one burrow m-1 of riverbank increasing bank retreat by 0.004 m a-1. Marginally stronger associations were present between metrics standardised for bank length than for bank area in all but one case.

The value for the River Bain at the highest burrow density is included in analysis, as discussed in section 3.3.1. However, it should be noted that if this point is excluded from analysis, a positive (albeit not significant) trend is still observed (BL: r = 0.412, p = 0.237; VL: r = 0.543, p = 0.105; EL: r = 0.456, p = 0.185).

Figure 6.7: Photograph of a bank collapse on the River Bain (top) and graphical heat map of the bank generated from erosion pin recordings. Red areas mark retreat, and blue areas mark depositional areas of bank slump at the bottom of the bank. The approximate area in the photograph is denoted by the black box.

225

Figure 6.8: 3D visualisation of all banks at Gaddesby Brook after 647 days of monitoring.

226

Figure 6.9: 3D visualisation of all banks at the River Bain after 677 days of monitoring.

227

Figure 6.10: 3D visualisation of a single bank at Gaddesby Brook after 281 (top), 491 (middle) and 647 (bottom) days of monitoring illustrating within-site variations in erosion and accretion.

228

-1 3 -1 -4 2 -1 -3 -1 -1 River Bank Number B L (burrows m ) V L (m m x10 ) E L (cm m x10 ) Retreat (m a ) SEM (m a ) River Bain 1 0.00 0.00 0 0.063 0.01 2 0.00 0.00 0 0.049 0.01 3 0.57 2.71 1.346 0.037 0.01 4 0.89 5.04 2.826 0.040 0.01 5 1.33 15.80 4.396 0.108 0.01 6 2.00 8.47 4.187 0.146 0.01 7 1.71 9.28 3.376 0.118 0.05 8 2.00 5.07 2.647 0.093 0.02 9 2.00 8.15 3.79 0.035 0.01 10 3.43 10.66 5.988 0.079 0.02 11 6.86 45.47 14.646 0.174 0.00

Gaddesby Brook 1 0.00 0.00 0.00 -0.005 0.00 2 1.11 12.76 18.96 -0.002 0.00 3 1.56 14.15 19.28 0.016 0.00 4 3.50 20.89 30.89 -0.004 0.00 5 4.25 47.88 15.70 0.014 0.00 6 4.00 32.17 13.07 0.025 0.00

Table 6.5: Measured values of burrows per meter of riverbank (BL), the volume of burrows

per meter of riverbank (VL), the entrance area of burrows per meter of riverbank

(EL), and the rate of bank retreat and standard error of retreat for each studied riverbank.

Gaddesby Brook River Bain r p r p

B A 0.415 0.207 0.682 *** 0.010

B L 0.560 0.124 0.701 *** 0.008

V A 0.507 0.152 0.744 *** 0.004

V L 0.618 * 0.096 0.738 *** 0.005

E A -0.093 0.430 0.701 *** 0.007

E -0.047 0.465 0.719 *** 0.006 L Table 6.6: Correlation values of the association between total retreat and burrow metric values at both sites. Significant values are denoted by asterisks (* = p < 0.1, ** = p < 0.05; *** = p < 0.01).

229

Figure 6.11: The association between (a) burrows per meter of riverbank (BL), (b) burrow

volume per meter of riverbank (VL), and (c) burrow entrance area per meter of

riverbank (EL) on the gross retreat of riverbanks at both rivers over the full time period.

230

6.4.3 How do crayfish burrows accelerate bank retreat?

6.4.3.1 Diffuse Erosion

Diffuse erosion ranged from 0.016 m a-1 (0.6 burrows m-1) to 0.057 m a-1 (2.0 burrows m-1) at the River Bain, and from -0.004 m a-1 (0 burrows m-1) to 0.025 m a-1 (4.0 burrows m-1) at

Gaddesby Brook (Table 6.7). VL was significantly positively associated with diffuse erosion at both rivers, and VA was significantly associated with diffuse erosion at the River Bain (Figure 6.12; Table 6.8). However, there was a strong difference between the strength of trends between different burrow metrics and diffuse erosion.

Whilst they were not significantly associated, BL and gross diffuse erosion were moderately positively correlated at Gaddesby Brook and weakly positively correlated at the River Bain.

There was no association between EL and gross diffuse erosion. The same trends were observed when ‘only retreat’ was considered, but associations between burrow metrics and retreat were weaker. Stronger associations were present between metrics standardised for bank length than for bank area for both burrow presence and burrow volume. Bain Bank 11 was not included in analysis, as all pins recorded collapse in the final study period, and so there were no diffuse erosion recordings available for analysis.

Gross Diffuse Only Retreat Diffuse River Bank Number V (m3 m-1 x10-4) E (m2 m-1 x10-3) SEM (m a-1) SEM (m a-1) L L Erosion (m a-1) Erosion (m a-1) River Bain 1 0.00 0.00 0.020 0.00 0.024 0.01 2 0.00 0.00 0.040 0.01 0.051 0.01 3 2.71 1.35 0.016 0.00 0.019 0.00 4 5.04 2.83 0.029 0.01 0.032 0.01 5 15.80 4.40 0.069 0.01 0.076 0.01 6 8.47 4.19 0.057 0.02 0.057 0.02 7 9.28 3.38 0.046 0.01 0.046 0.01 8 5.07 2.65 0.040 0.01 0.049 0.01 9 8.15 3.79 0.035 0.01 0.037 0.01 10 10.66 5.99 0.031 0.01 0.048 0.01 11 45.47 14.65 - - - -

Gaddesby Brook 1 0.00 0.00 -0.004 0.00 0.007 0.00 2 12.76 18.96 -0.002 0.00 0.012 0.00 3 14.15 19.28 0.016 0.00 0.017 0.00 4 20.89 30.89 -0.004 0.00 0.004 0.00 5 47.88 15.70 0.014 0.00 0.020 0.00 6 32.17 13.07 0.025 0.00 0.025 0.00

Table 6.7: Measured values of the volume of burrows per meter of riverbank (VL), burrow

entrance area per meter of riverbanks (EL), Gross Diffuse Erosion, and Only Retreat Diffuse Erosion.

231

(a) Gross Retreat Gaddesby Brook River Bain

r p r p

B A 0.412 0.209 0.200 0.290

B L 0.557 0.125 0.276 0.220

V A 0.504 0.154 0.647 ** 0.022

V L 0.616 * 0.097 0.709 ** 0.011

E A -0.098 0.426 0.371 0.146

E L -0.053 0.460 0.468 * 0.086

(b) Only Retreat Gaddesby Brook River Bain

r p r p

B A 0.311 0.274 0.333 0.174

B L 0.478 0.169 0.367 0.149

V A 0.494 0.160 0.679 ** 0.015

V L 0.624 * 0.093 0.702 ** 0.012

E A -0.260 0.309 0.461 * 0.090

E -0.172 0.372 0.509 * 0.067 L

Table 6.8: The association between diffuse erosion and crayfish burrow metrics, when negative values were considered as (a) ‘gross retreat’ and (b) ‘only retreat’. Significant values are denoted by asterisks (* = p < 0.1, ** = p < 0.05; *** = p < 0.01).

232

Figure 6.12: The association between (a and c) burrow volume per meter of riverbank (VL)

and (b and d) burrow entrance area per meter of riverbank (EL); and (a and b) gross diffuse erosion and (c and d) only retreat diffuse erosion at both rivers for the full time period. River Bain Bank 11 is excluded, as it only recorded collapse in the final time period of monitoring.

6.4.3.2 Bank Collapse

Bank collapse was recorded at every site on the River Bain, whereas only five pins recorded bank collapse at Gaddesby Brook (Table 6.9). Over the course of the study, 40.5% of pins recorded collapse at the River Bain, which contributed 57.5% of total retreat recorded, whereas only 3.2% of pins recorded collapse at Gaddesby Brook, recruiting just 0.6% of sediment recorded. At the River Bain, banks ranged from no recordings of collapse (2.0 burrows m-1) to every pin recording collapse (6.86 burrows m-1). Due to the low number (n = 5) of collapses recorded at Gaddesby Brook, only data from one river (River Bain) were considered for quantitative analysis.

233

There was no statistically significant association between any crayfish burrow metrics and the depth of collapse that occurred throughout the study, but all burrow metrics had a strong positive association with the area of bank that collapsed as represented by the percentage of pins collapsed (Table 6.10; Figure 6.13); all associations were significant, but VL and VA had the strongest association with the percentage of pins collapsed, and stronger associations were present between metrics standardised for bank length than for bank area. The bank with the greatest density of burrows (Bank 11) entirely collapsed in the final stage of monitoring following a high flow event. The bank that recorded collapse at Gaddesby Brook was also the bank with the greatest density of burrows. Should Bain Bank 11 be discounted, a positive (albeit not significant) trend is still observed considering all bank metrics.

Figure 6.13: The association between burrows per meter of riverbank (BL) and collapse retreat at the River Bain over the full time period.

234

-2 -1 3 -2 -4 3 -1 -4 2 -1 -3 Only Collapse -1 Pins Collapsed Retreat as River Bank Number B (burrows m ) B (burrows m ) V (m m x10 ) V (m m x10 ) E A (%) E (m m x10 ) SEM (m a ) A L A L L (m a-1) (% a-1) Collapse (%)

River Bain 1 0.00 0.00 0.00 0.00 0.00 0.00 0.162 0.01 17.4 72.67 2 0.00 0.00 0.00 0.00 0.00 0.00 0.287 0.07 11.2 64.32 3 0.56 0.57 2.63 2.71 0.13 1.35 0.139 0.02 11.2 57.13 4 0.95 0.89 5.40 5.04 0.30 2.83 0.064 0.02 5.8 15.08 5 1.21 1.33 14.36 15.80 0.40 4.40 0.144 0.02 19.6 48.38 6 1.74 2.00 7.36 8.47 0.36 4.19 0.209 0.06 44.5 85.37 7 1.74 1.71 9.42 9.28 0.34 3.38 0.242 0.10 21.1 67.26 8 2.33 2.00 5.91 5.07 0.31 2.65 0.135 0.02 35.0 63.25 9 1.56 2.00 6.34 8.15 0.29 3.79 0.000 0.00 0.0 0.00 10 3.81 3.43 11.84 10.66 0.67 5.99 0.131 0.01 20.5 58.53 11 6.15 6.86 40.81 45.47 1.31 14.65 0.191 0.01 53.9 100.00

2 2 Table 6.9: Measured values of burrows per m of riverbank (BA), burrows per meter of riverbanks (BL), burrow volume per m of riverbank 2 (VA), burrow volume per meter of riverbank (VL), burrow entrance area per m of riverbank (EA), burrow entrance area per meter

of riverbank (EL), and Only Collapse Retreat, Percentage of Pins Collapsed, and Percentage of Retreat as Collapse.

235

(a) Only Collapse River Bain

r p

B A -0.014 0.484

B L -0.020 0.476

V A 0.069 0.420

V L 0.061 0.429

E A -0.016 0.481

E L -0.019 0.478

Percentage of Pins River Bain (b) Collapsed r p

B A 0.686 *** 0.010

B L 0.688 *** 0.010

V A 0.669 ** 0.012

V L 0.659 ** 0.014

E A 0.663 ** 0.013

E L 0.663 ** 0.013

Percentage of Retreat River Bain (c) as Collapse r p

B A 0.396 0.114

B L 0.389 0.119

V A 0.419 * 0.100

V L 0.408 0.106

E A 0.370 0.131

E 0.364 0.135 L

236

6.4.3.3 Bank Shape

Across the full period of the study, EA and EL were significantly associated with change in riverbank shape at Gaddesby Brook (Table 6.11; Table 6.12; Figure 6.14). There was no association between any retreat metric and any burrow metric at the River Bain (Table 6.12).

2D cross sectional bank profiles revealed that the differences in bank height retreat were a result of undercutting at Gaddesby Brook (Figure 6.15). Undercutting was associated with all banks where crayfish burrows were present, and greater undercutting was associated with banks of high burrow volume densities. Undercutting also occurred on the River Bain, but no consistent patterns or associations were witnessed in relation to crayfish burrow volume density (Figure 6.16).

Gaddesby Brook River Bain

r p r p

B A 0.465 0.176 0.044 0.448

B L 0.371 0.234 0.098 0.387

V A 0.410 0.210 0.187 0.291

V L 0.349 0.249 0.218 0.260

E A 0.840 ** 0.018 0.122 0.360

E L 0.898 *** 0.008 0.168 0.311

Table 6.11: The association between crayfish burrow metrics and change in riverbank shape, calculated using total retreat at both sites. Significant values are denoted by asterisks (* = p < 0.1, ** = p < 0.05; *** = p < 0.01).

237

Relative Change in River Bank Number E (%) E (m2 m-1 x10-3) A L Bank Shape(m a-1) River Bain 1 0.00 0.00 0.056 2 0.00 0.00 -0.097 3 0.13 1.35 -0.079 4 0.30 2.83 0.025 5 0.40 4.40 0.062 6 0.36 4.19 -0.074 7 0.34 3.38 -0.151 8 0.31 2.65 -0.002 9 0.29 3.79 0.070 10 0.67 5.99 -0.075 11 1.31 14.65 0.023

Gaddesby Brook 1 0.00 0.00 -0.071 2 1.78 18.96 0.006 3 2.41 19.28 -0.003 4 4.12 30.89 0.008 5 1.99 15.70 -0.005 6 1.22 13.07 -0.040

2 Table 6.12: Measured values of burrow entrance area per m of riverbank (EA), burrow

entrance area per meter of riverbank (EL), and Relative Change in Bank Shape.

Figure 6.14: The association between burrow entrance area per meter of riverbank (EL) and relative change in riverbank shape since the start of the study, considering total

238

retreat at Gaddesby Brook. Positive values equate to bank steepening, and negative values equate to bank shallowing.

Figure 6.15: Changes in bank profile over time at Gaddesby Brook. Values in the top left

corner are burrow volume per meter of riverbank (VL), as these had the strongest association of all metrics with change in bank shape. Very little undercutting can be seen at banks without crayfish burrows, whereas undercutting between 0 and 0.6 m in height can be seen on all burrowed banks.

239

Figure 6.16: Changes in bank profile over time at the River Bain. Values in the top left corner

are burrow volume per meter of riverbank (VL), as these had the strongest association of all metrics with change in bank shape. Undercutting was witnessed at some banks, but there was no consistent pattern in change in bank shape in relation to crayfish burrow density.

6.4.4 How much sediment does accelerated bank retreat associated with crayfish burrows recruit to river systems?

VL was the most strongly correlated burrow metric with total retreat across the full study period and was the only burrow metric significantly correlated with bank retreat on both rivers. Linear regression models of the dependence of total retreat on VL were therefore created for each river.

The reach scale regression model for the River Bain was significant (y = 27.3 VL + 0.058; 2 2 r = 0.545, F = 10.769, p < 0.01), but Gaddesby Brook was not (y = 4.67 VL - 0.003; r = 0.382, F = 2.472, p = 0.191).

240

At the reach scale, accelerated erosion facilitated by burrows recruited 24.9 t km-1 a-1 of sediment into the River Bain. At the bank scale, accelerated retreat from crayfish burrows recruited an average of 48.7 kg m-1 a-1, ranging up to 189.1 kg m-1 a-1 of sediment at the River Bain, and an average of 18.2 kg m-1, ranging up to 44.6 kg m-1 a-1 of sediment at Gaddesby Brook (Table 6.13).

6.4.5 Relative importance of biotic, abiotic and interactive sediment recruitment

Accelerated bank erosion caused by crayfish burrows was an important driver of sediment dynamics on both rivers. At the reach scale, accelerated erosion recruited 49.2 times more sediment than burrows alone, and recruited 12.2% of all sediment at the River Bain, including banks where burrows were absent. Considering the measured banks, accelerated retreat recruited 29.8% of all sediment, compared to just 0.6% directly from burrows. This represents an increase of 43.7% of sediment recruitment compared to riverbanks without burrows. At Gaddesby Brook, where the reach scale regression model was not significant, bank scale observations can be applied to equations 6.11 to 6.19, with a being the observed erosion rate at the only bank without burrows (Bank 1). At Gaddesby Brook, sediment deposition occurred in the absence of burrows, but erosion occurred in the presence of burrows. Therefore, burrows (10.4%) and accelerated retreat from burrows (89.6%) jointly recruited 100% of sediment from the studied banks.

6.4.5 Results Summary

Erosion pins successfully monitored bank retreat and recorded significant positive associations between crayfish burrows and increased erosion rates. Burrows were positively associated with all methods of erosion, suggesting that burrows influence all mechanisms of bank retreat. Burrow volume had the strongest associations with retreat metrics, with burrow entrance area having the weakest. The density of burrows standardised for riverbank length, rather than riverbank area, had a stronger association with riverbank retreat in all cases. Burrows were associated with undercutting, and at Gaddesby Brook, an increased density of crayfish burrows significantly steepened the bank. The presence of crayfish burrows was calculated to recruit 24.9 t a-1 at the River Bain, representing a 13.9% increase in sediment delivery, and the presence of burrows caused sediment dynamics at Gaddesby Brook to reverse from depositional to erosional processes.

241

Only Erosion in the Absence Accelerated Bank Retreat Bank Only Crayfish Burrows River of Crayfish Burrows Caused by Crayfish Burrows Number (kg a-1) % (kg a-1) % (kg a-1) % River Bain 1 482.5 100.0 0.0 0.0 0.0 0.0 2 265.9 100.0 0.0 0.0 0.0 0.0 3 313.6 99.7 0.8 0.3 -114.7 -36.5 4 364.1 99.5 1.9 0.5 -117.2 -32.0 5 287.1 53.7 4.0 0.7 243.5 45.6 6 300.2 39.8 2.1 0.3 451.5 59.9 7 301.5 49.2 2.7 0.4 308.9 50.4 8 261.9 62.1 1.5 0.4 158.4 37.6 9 392.8 99.4 2.4 0.6 -155.7 -39.4 10 274.1 73.3 3.1 0.8 96.5 25.8 11 338.0 33.4 13.4 1.3 661.8 65.3

Total Burrowed Banks 3 581.6 69.6 31.9 0.6 1 532.0 29.8 Total Reach 179 220.0 87.6 506.8 0.2 24 942.9 12.2

Gaddesby Brook 1 -26.7 -100.0 0.0 0.0 0.0 0.0 2 -32.9 -216.6 4.8 31.7 14.2 68.3 3 -24.6 -22.7 5.3 4.9 107.1 95.1 4 -20.5 -292.4 7.0 100.0 -4.3 -61.3 5 -21.6 -28.0 16.1 20.8 73.7 79.2 6 -32.9 -16.2 12.2 6.0 200.8 94.0

Total Burrowed Banks -159.3 -36.5 45.4 10.4 391.5 89.6 Total Reach ------

Table 6.13: The relative sediment recruitment from bank erosion as a result of direct sediment input from crayfish burrows, the accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows. Percentages are calculated as a proportion of total sediment recruited; negative values represent deposition, expressed as a proportion of sediment recruited.

242

6.5 Discussion

This study utilised a large number of erosion pins (430) to monitor 17 riverbanks along two rivers affected by signal crayfish burrowing. Pins were deployed at a high resolution in comparison to previous studies (Table 6.1) and monitored for 22 months. The data obtained demonstrate, for the first time, a quantitative association between animal burrows and riverbank retreat, and quantify the association between burrow characteristics and mechanisms of retreat.

Damage to riverbanks beyond the immediate effects of burrow construction has only previously been qualitatively recorded (Sibley 2000; West 2010; Arce and Dieguez-Uribeondo 2015). Harvey et al. (2019) highlighted the lack of quantitative research into the physical damage to riverbanks as a result of burrowing by invasive species, and called for long term monitoring projects to assess associated processes beyond the immediate input of sediment produced by burrowing (e.g. Faller et al. 2016; Sofia et al. 2017). This work provides the first quantitative evidence for both of these cases, demonstrating the impact of burrows on fluvial erosion and on mass failure, across a 22 month study period. Harvey et al. (2019) hypothesise that animal burrows may alter bank erosion processes through (i) geotechnical and hydrological effects, and (ii) hydraulic effects. Significant associations between burrow volume and the prevalence of collapse events, and burrow entrance area and diffuse erosion provide supporting evidence for both hypotheses.

6.5.1 General Results

Erosion pins successfully monitored erosion. The quantitative results of data analysis corresponded with qualitative field observations, in terms of processes that occurred at the bank scale, such as localised collapses (Figure 6.7), and at the reach scale, with pins recording collapse-dominated erosion on the River Bain and diffuse erosion on Gaddesby Brook.

Accretion recorded at Gaddesby Brook typically occurred at the top of banks (Figure 6.8), which were often covered with loose soil. Whilst these banks were completely submerged during high flow events, at low flow events, when the pins were measured, the pins recorded the movement of sediment down the bank as well as bank retreat.

243

A substantial difference was observed between rivers. Riverbank retreat at the River Bain was considerably faster than at Gaddesby Brook at all burrow densities considered, with mean rates being more than ten times faster on the River Bain than at Gaddesby Brook. The nature of retreat was also different between rivers, with the River Bain being a collapse-dominated system, and Gaddesby Brook being a diffuse erosion-dominated system. The two rivers were selected to attempt to remove the influence of external variables other than crayfish burrows that may drive erosion, but within the bounds of typical rivers invaded by crayfish. The catchment size and discharge are of the same order of magnitude for both rivers, both catchments are predominantly agricultural, and both rivers had a similar level of riparian vegetation and patterns of local flow velocities. The grain size of materials collected from exposed river banks was similar on both rivers (41% silt and clay with 59% sand on Gaddesby Brook, and 35% silt and clay with 64% sand at the River Bain), which suggests that the difference in erosion rate is not caused by material properties. However, British Geological Survey borehole data from the floodplains, which may be more representative of general floodplain materials, indicate a predominance of clays at Gaddesby Brook and sandy silts at the River Bain (British Geological Survey, 2020), which is consistent with greater erosion rates and more collapse on the Bain. It is also notable that banks were considerably steeper on the Bain (74° versus 49° on Gaddesby Brook). These factors were not investigated in this study, but may be important for explaining the differences in rates and mechanisms of retreat between the two sites. Despite the difference in retreat rates and mechanisms, crayfish burrow metrics were significantly associated with retreat metrics on both rivers, strongly suggesting that burrows play an important role in the geomorphology of rivers that exhibit different rates of retreat.

6.5.2 Do signal crayfish burrows accelerate bank retreat?

Increased crayfish burrow density was associated with greater total retreat at both study sites, with the presence of each additional burrow per metre increasing retreat by 0.017 m a-1 on the River Bain. Compared to erosion of banks without burrows this represents an average increase of 30.3%, and the bank with the highest burrow density (6.15 burrows m-2) recorded retreat that was 253% greater than the erosion on banks without burrows. This supports observations that signal crayfish accelerate bank erosion (Sibley 2000; West 2010) and confirms Harvey et al.’s (2019) assertion that burrowing can affect riverbank morphology.

244

At Gaddesby Brook, when burrows were absent, accretional processes dominated (-4.6 mm a- 1), whereas the highest burrow density (4.25 burrows m-1) resulted in erosional processes occurring (24.9 mm a-1), with an increase of one burrow m-1 of riverbank increasing bank retreat by 4 mm a-1. This shows that a relatively small increase in local crayfish burrows density to one burrow m-1 of riverbank is modelled to prevent deposition, and alter the morphological functioning of the system from being depositional to a net neutral bank, with any further burrowing changing the direction of the morphological change.

All burrow metrics were significantly correlated with riverbank retreat on the River Bain (Table 6.5), which demonstrates the importance of crayfish burrows driving the geomorphological processes in the system. All burrow metrics were positively correlated with riverbank retreat at Gaddesby Brook, and whilst the trends were visually strong, they were not all significant, due to the low number (six) of individual banks monitored, and the high levels of cohesion in the bank. It is likely that burrows had a lower observable impact at Gaddesby Brook due to the lower erosion rates, and their influence on erosional processes may become apparent over a longer timescale than considered here. It may also be that the higher levels of cohesion meant that banks were more resistant to structural change. This is supported by the lower levels of collapse observed throughout the monitoring at Gaddesby Brook compared to the River Bain, with considerably higher hydraulic forces required to destabilise the more cohesive material compared with the sandy banks of the River Bain, which required comparatively lower levels of stress for erosion to occur.

The significance of the associations observed on the River Bain were dependent on the inclusion of the bank with the highest burrow densities. Whilst the inclusion of this point is legitimate, it is still worthy of note and discussion. Should the point be excluded, it could be concluded that there is no association between burrow density and total bank retreat on the River Bain. This association is dependent on collapse events. When only diffuse erosion was considered, a significant association was observed when not considering Bank 11, but there was no association when considering only collapses. The mass of sediment recruited by collapses was substantially greater than the mass of sediment recruited by diffuse erosion on the River Bain, and so when considering total retreat, the association between burrows and diffuse erosion was dampened by the collapses that occurred.

A large range of bank retreat values were recorded at the density of 2 burrows m-1 on the River Bain (Figure 6.17). This spread in retreat values may be a result of the collapse-dominated

245 nature of retreat on the River Bain. Collapses, and thus recorded retreat, are sporadic events, and so across the study period, some but not all collapses will be recorded, which will add to the variance of the dataset. Longer monitoring periods would therefore likely result in reduced spread among the data, and thus a better understanding of the association between the measured variables.

Figure 6.17: Values of potential discrepancy. Reproduced from Figure 6.11a. A large range of bank retreat values were recorded at the density of 2 burrows m-1 on the River Bain, which may be a result of the collapse dominated nature of retreat on the

River Bain. The circled value has an outlying density of burrows (BL), but the recorded annual retreat is not an outlying value. The validity of these points of discrepancy are discussed in Section 6.5.2.

However, these patterns of variability are also witnessed in diffuse erosion, suggesting that it is not just bank collapse that is driving the variability between banks. Whilst banks were selected to be subject to uniform conditions outside of the density of signal crayfish burrow metrics, there will be random effects occurring to each bank that cannot be controlled for. In field investigations, it is not possible to remove external factors, or to standardise across the variables of interest. However, variability in rates of riverbank retreat has been observed

246 throughout bank retreat studies where crayfish burrows are absent, and so natural variation between morphologically similar banks is to be expected (Table 6.14). The natural variability of factors not controlled for is substantial around a single value of burrow density (e.g. 0.035 m a-1 (2 burrows m-1) and 0.146 m a-1 (2 burrows m-1) on the River Bain (Figure 6.17)), and yet there are still significant associations between bank retreat and crayfish burrowing. This demonstrates the strength of effect that burrowing has and its importance as a driver of erosion in a highly variable system.

Study Observed Variability

Lawler et al. 1999 77.7 (+/-105.6) to 440.1 (+/-181.7) mm a -1

Kronvang et al. 2013 22 (+/-3) to 109 (+/-43) t km -1

Henshaw et al. 2013 7 (+/-4) to 299 (+/-321) mm a -1

Palmer et al. 2014 16 (+/-5.3) to 220 (+/-227) mm a -1

Foucher et al. 2017 -275 to 100 mm a -1

Table 6.14: Observed variability in riverbank retreat along rivers where crayfish burrows were not present in previous published studies.

There was also considerable variation between pins within banks (Figure 6.18). The retreat values for individual pins, rather than site averages, could be used for analysis, and indeed, considering BL and total retreat across the full study period at Gaddesby Brook, treating pins individually yielded significant results (Figure 6.18b). However, bank averages were used for two reasons; (i) this has been the accepted methodology in previous erosion pin research, (e.g. Couper et al. 2002; Henshaw et al. 2013), and (ii) burrows were hypothesised to have geomorphic impacts on bank erosion at the bank scale. McShane et al. (2019) highlight the importance of considering the justification of methods and the understanding of mechanisms, not only statistical significance, when choosing between alternative ways of grouping or averaging data. Whilst McShane et al. (2019) provide general discussion, their arguments are applicable here; whilst treating pins individually had a significant association with burrow metrics, the very low r2 value suggests that burrows are better considered collectively, and may collectively act conjointly to promote erosion at a broader scale than that measured by individual pins.

247

Figure 6.18: Variation of retreat between (a) averaged banks and (b) pins on single riverbanks at Gaddesby Brook. When considering pins independently, the association is significant but the association between variables is very poor (p < 0.05; r2 = 0.026) compared to considering banks as an entire entity (p = 0.124; r2 = 0.314).

6.5.3 Diffuse Erosion

VL was significantly positively associated with diffuse erosion at both rivers, and VA was significantly associated with diffuse erosion at the River Bain (Table 6.7). The same trends were observed when the ‘only retreat’ approach to negative pin values was used, but associations between burrow metrics and retreat were weaker. This suggests that stronger associations were seen in areas of deposition than retreat, and that burrows are important for inhibiting the settling of loose sediment delivered to the channel from other sediment sources.

248

The presence of burrows has been hypothesised to alter the surface roughness of the riverbank, which would likely have implications for the deposition of sediment (Harvey et al. 2019). It was also demonstrated that burrows steepened riverbanks at Gaddesby Brook (Figure 6.14), which may have rendered burrowed slopes too steep to retain sediment settling from above.

Further, burrow entrance area (EA and EL) was significantly associated with bank erosion, which may increase turbulence at the bank face (Harvey et al. 2019), and thus directly amplify the entrainment of particles.

6.5.4 Bank Collapse

Area of bank collapse was significantly associated with all burrow metrics on the River Bain. Burrows and burrow remnants partially remained and were clearly visible after collapse events had occurred, supporting hypotheses that single burrows can affect multiple collapse events (Faller et al., 2016). The presence of crayfish burrows constructed by red swamp crayfish (Procambarus clarkii) has been linked to the destabilisation and collapse of banks in rice paddies in the Iberian Peninsula (Arce and Dieguez-Uribeondo 2015), with 73% of recorded burrows collapsing within nine days of excavation (Barbaresi et al., 2004). These results demonstrate a link between burrowing and riverbank mass failure through in situ monitoring for the first time, supporting the hypotheses of Harvey et al. (2019) that burrows can drive riverbank mass failure.

Physical (Viero et al. 2013; Saghaee et al. 2017) and numerical (Camici et al. 2014; Orlandini et al. 2016; Borgatti et al. 2017) modelling has suggested that the presence of animal burrowing can also greatly increase the probability of riverbank collapse, particularly from burrows constructed on the waterside of levees (Saghaee et al. 2017). This occurs as a result of animal burrows increasing the hydraulic gradient across a levee, making the structure more liable to collapse and represents a significant risk to flood infrastructure. This was observed on the Foenna Stream (Camici et al. 2014) and on the Secchia River (Orlandini et al. 2016), Italy, when the collapse of a crested porcupine (Hystrix cristata) burrow lead to the failure of an entire levee, resulting in over $500m of damage in the latter case (Orlandini et al. 2016). These results provide field evidence to support these modelling studies, and support the conclusions of Camici et al. (2014) and Orlandini et al. (2016) that the porcupine burrows were responsible for the major collapses observed on the Foenna Stream and Secchia River.

249

Cohesive stream banks are typically composed of aggregate material, which is porous and unsaturated under base flow conditions (Rinaldi and Casagli 1999; Simon et al. 2000). A low level of saturation results in negative pore water pressure, and an increased level of cohesion in the bank, whereas a high level of saturation results in positive pore water pressure, which reduces the cohesion of the bank and increases the likelihood of mass failure or bank erosion (Simon and Collison 2001; Fox et al. 2007). The presence of animal burrows in the bank would increase the porosity of the riverbank, thus increasing the pore water pressure, and therefore making it more susceptible to mass failure under high flow events. Whilst there was no association between burrow metrics and the depth of collapses that occurred, the likelihood of pins recording collapse was strongly positively correlated with burrow density (Figure 6.13), showing that the presence of burrows accelerated the probability of the occurrence of mass failures.

Fieldwork was regularly undertaken on the River Bain for a separate project unrelated to this research, and collapse events were observed to occur following high flow events, but this was not quantified. A flood hydrograph was available for the first half of the monitoring period, and erosion was four times higher after the first high flow period than the low flow period, and the percentage of pins that recorded collapse was ten times higher after the first high flow period than after the low flow period (Figure 6.19). However, associations during these time periods with crayfish burrow metrics were not significant, and so no quantitative conclusions can be drawn to support the hypothesis. Equally, mass failure events are typically associated with high flow events, and whilst crayfish burrows may have preconditioned the banks, this change in erosion may be entirely independent of the influence of crayfish burrows.

Statistical analyses were not undertaken for Gaddesby Brook, as only one collapse occurred on the pinned banks throughout the monitoring period. However, the one collapse that did occur was the bank with the highest density of crayfish burrows (37.84 m3 m-2 x 10-4 bank). The largest collapse at the River Bain was also recorded on the bank with the highest density of crayfish burrows. Whilst quantitative analysis could not be carried out for Gaddesby Brook, this evidence suggests that burrows were an important factor for driving collapse on both river systems. Whilst the location of crayfish burrows on the measured banks was not recorded, bank collapses on the River Bain were observed at the location of the burrows. Collapses were often tall, not very wide, and directly above the burrowed sections of the banks (Figure 6.20; Figure 6.21), showing that burrows were responsible for promoting the collapse of material from

250 directly above the burrow at a local scale. This is supported by the linear association observed between burrow densities and the area of bank collapsed, suggesting that burrows are responsible for recruiting sediment within the local reach, with large scale bank collapses occurring in the presence of multiple burrows, where multiple smaller collapses are compounded.

As well as promoting collapses spatially, burrows were important for promoting collapses over a long time period. Burrows and burrow remnants partially remained and were clearly visible after collapse events had occurred (Figure 6.22), showing that single burrows could affect multiple collapse events (Figure 6.23).

Figure 6.19: Flow hydrograph for October 2017-2018 at the River Bain compared to sampling dates and observed erosion.

251

Figure 6.20: A small mass failure directly above a crayfish burrow on the River Bain.

252

Figure 6.21: Collapses from three banks on the River Bain, highlighted in black boxes. Collapses were recorded by erosion pins to typically occur at the location of crayfish burrows, and were taller than they were wide, typically in the shape of columns.

253

Figure 6.22: The entrances of crayfish burrows could be seen after collapse at the River Bain, showing that deep burrows have the capacity to influence multiple collapses.

254

Figure 6.23: Conceptual model of hydrologically driven geotechnical bank retreat as a result of crayfish burrows. (a) A cohesive riverbank with a signal crayfish burrow; (b) fluvial action leads to erosion at the bank toe; (c) burrow volume and fluvial undercutting reduces the cohesive strength of the bank, facilitating mass failure; (d) the cohesive bank post-retreat, but the burrow remnants remain, meaning they can affect multiple collapses, and cause the cycle to repeat (e and f). Dotted lines show erosion / failure extent, and shaded areas demonstrate area of sediment eroded since previous panel.

255

The strength of the association between burrow volume density and bank retreat is particularly noteworthy on the River Bain where mass failure dominates. Riverbank retreat is often mediated in mass failure driven systems through the deposition of submerged slump blocks of deposited material, which can promote erosion by directing flow toward the bank (Hackney et al., 2015) or protect the bank from high flow events by diverting flows away from the bank toe (Wood et al., 2001; Parker et al., 2011). On the River Bain, slumped blocks quickly became vegetated (Figure 6.24), increasing the stability of the slumped material (Abernethy and Rutherfurd 2000), and potentially slowing down lateral erosion to sheltered banks by providing a continual protective barrier (Micheli and Kirchner 2002a; 2002b). There was still a strong positive correlation between burrow metrics and bank retreat, so if the slumped material was protecting the bank, then without it the gross impact of crayfish burrows is likely to be greater. Moreover, the presence of protective slumped material, implies that mass failures on the River Bain were controlled by internal bank processes rather than driven by hydraulic processes and, for example, bank undercutting. This hypothesis is supported by bank profile data that showed no association with crayfish burrows and undercutting on the River Bain (Figure 6.16), at least over the timescales observed here.

6.5.5 Bank Shape

Across the full period of the study, burrow entrance area (EL and EA) were strongly positively correlated with the steepening of riverbanks at Gaddesby Brook (Table 6.11). Undercutting was associated with all banks where crayfish burrows were present on Gaddesby Brook (Figure 6.14; Figure 6.15), and greater undercutting was associated with banks of high burrow volume densities, suggesting that burrows are driving the process of undercutting.

Accretion was most strongly witnessed at the top of banks (Figure 6.8), where loose soil was observed. Whilst it could be argued that the downward movement of loose soil was independent of crayfish burrows, crayfish burrows altered the shape of the banks, which may have promoted or inhibited soil movement. This contributed to the observed difference in sediment dynamics at the top and toe of the bank. The lower rate of collapse at Gaddesby Brook compared to the River Bain may mean that erosion is more localised, and so crayfish burrows may have a more distinct influence on the surrounding bank at a localised scale.

Undercutting was visually observed at the River Bain, but was not empirically recorded (Figure 6.16); erosion pins likely under-recorded undercutting, because the spatial and temporal

256 resolution of pin placement was too coarse (Figure 6.25). Collapses were common on the River Bain, and the cycle of undercutting taking place, leading to collapse, and a fresh, not undercut bank being left, was likely shorter than the time between monitoring periods. This may also be due to the River Bain being a collapse-dominated system, where riverbanks cannot be further steepened before collapse occurs, as they are already at a critical threshold. At the River Bain, crayfish burrows were accelerating the collapse process, but were not altering the nature of erosion in the system.

Figure 6.24: Bank collapses at the River Bain. Collapsed blocks may have shielded the bank from hydraulic erosion after collapse. Exposed bank slumps often became vegetated, which may have aided the protective effect.

The rate of erosion at the River Bain was faster than that in Gaddesby Brook on unburrowed banks, and it may be that the vertical nature of the banks in the River Bain is a result of crayfish burrowing activity, with bank steepening having already occurred, which has seen the change in nature of erosion. With time, this may be witnessed at Gaddesby Brook, as crayfish continue to increase the steepness of the impacted banks. More erosion was recorded at the toe of banks than at the top of banks, which was consistent with previous studies (Laubel et al. 2003; Veihe et al. 2011; Kronvang et al. 2013). However, this effect was accentuated with increasing crayfish burrow densities (Figure 6.14). At the start of surveying, the banks on the River Bain were near vertical (Figure 6.24), whereas the average slope of banks at Gaddesby Brook were surveyed to be 49° across the full height bank, and 54.7° at the toe where crayfish burrows were present (Chapter 3; Figure 6.26). It has previously been suggested that crayfish prefer to construct burrows in steep banks (Faller et al. 2016), but this evidence suggests that burrowing actively steepens banks over time, which may have further implications for the accelerated

257 erosion of banks through modifying geotechnical and hydrological processes (Simon and Collison 2001; Fox et al., 2007).

The association between crayfish burrows and bank undercutting is likely to be underestimated here. These processes were detected by the pins despite processes occurring on a finer scale than the pins were always able to detect. Banks were visually observed to be undercut prior to collapses being recorded, but the coarse resolution of the pins was not able to consistently empirically record these processes (Figure 6.25), suggesting that the actual impact of crayfish burrows on riverbank undercutting are stronger than could be quantified by the coarse resolution of erosion pins employed here.

Figure 6.25: An undercut bank on the River Bain. Banks were visually recorded to be undercut prior to mass failure, however the resolution of the erosion pins (denoted by white circles) was often too coarse to detect the local processes that lead to large scale collapses.

258

Figure 6.26: The measured angles of the burrowed banks, considering the full height of the riverbank, and the angle of the section of riverbank where crayfish burrows were present.

6.5.6 Modelling Sediment Input

Accelerated retreat facilitated by crayfish burrows was calculated to recruit 24.9 t km-1 a-1 of sediment into the River Bain, which represented a 14.2% and 43.7% increase on erosion in the absence of burrows at the reach and bank scale respectively (Table 6.13). This is one to two orders of magnitude greater than previous estimates of sediment production caused by burrow excavation alone: 0.25 - 0.5 t km-1 a-1 (Rice et al., 2016) and 3.0 t km-1 (Faller et al., 2016).

The distribution of crayfish burrows in invaded catchments is patchy (Faller et al., 2016), and only 25% of bank length along the River Bain study reach was affected by crayfish burrows. Distributed field surveys across rivers in the UK (n = 69 and 23) have found maximum impacted bank lengths at the reach scale of 23.5% and 27.1% (Faller et al., 2016; Chapter 3, respectively), and so the proportions at these study sites indicate that these are severely impacted reaches. However, crayfish invasion is progressive, with the proportion of impacted reach significantly positively associated with length of crayfish occupancy since invasion

259

(Chapter 3), and so the proportion of impacted banks may increase through time. Should burrows occupy 100% of riverbanks, accelerated erosion from burrows at the River Bain would be responsible for recruiting 99.8 t km-1 a-1 into the system, which would be a 35.5% total sediment contribution, and an increase of 56.8% over bank erosion in the absence of crayfish burrows (Table 6.15).

Burrow Mass of Sediment Proportional Increase from the Prevalence (%) Recruited (t km-1 a-1) Contribution (%) Absence of Burrows (%) 30 29.9 14.3 17.0 40 39.9 18.1 22.7 50 49.9 21.7 28.4 60 59.9 24.9 34.1 70 39.8 27.9 39.8 80 79.8 30.6 45.4 90 89.8 33.2 51.1 100 99.8 35.5 56.8

Table 6.15: The total and proportional contributions of sediment to the River Bain from accelerated erosion in the event of an increased distribution of crayfish burrow presence (see section 6.3.9 for equations).

These calculations provide robust estimates of the relative contribution of crayfish burrows to the recruitment of riverbank sediment along the two study reaches. It is difficult to extrapolate these observations to other rivers, not least because of the significant variable impact between these two lowland streams. Nevertheless, the data provide an insight into the magnitude of sediment being recruited to rivers as a direct and indirect result of crayfish burrowing relative to bank recession in the absence of burrows. The direct effects of burrowing contribute small amounts of sediment relative to the additional bank erosion facilitated by burrow presence: 49.2 times more on the River Bain, and ten times more at Gaddesby Brook. Table 6.13 shows that bank erosion processes facilitated by burrow presence account for 12.2% of sediment recruitment at the River Bain at the reach scale (accounting for unborrowed stretches), and that equivalent values at individual banks where burrows are present can be substantially higher.

This research quantifies the significant effect that crayfish can have on fine sediment recruitment via bank erosion, with implications for, amongst other things, ecosystem health. In

260 particular, excessive fine sediment can smother riverbed gravels, which can result in reduced fish spawning habitat (Soulsby et al., 2001; Kemp et al., 2011; Sear et al., 2016), and egg and juvenile survival (Suttle et al., 2004; Kemp et al., 2011; Jensen et al., 2009). Significant declines in fish populations have been observed since 1985 following the introduction of signal crayfish the River Bain in 1992 (r = -0.161, p < 0.001), particularly of dace (Leuciscus leuciscus; r = -0.445, p < 0.001) and chub (Leuciscus cephalus; r = -0.403, p < 0.001), for which clean gravels are important for feeding and spawning activities (Pledger et al., 2016; Rice et al., 2019; Environment Agency, 2020). A significant decrease in dace populations has also occurred at Gaddesby Brook following crayfish invasion over the same time period (r = - 0.722, p = 0.004; Environment Agency, 2020). At both the River Bain and Gaddesby Brook, fine sediment accumulation over the riverbed has increased over the last 20 years, coinciding with crayfish invasion. Landowners and Environment Agency staff report clean gravel substrates and an absence of fine sediment accumulations twenty years ago, but in this study accumulations of 0.4 m at locations within both reaches. Other factors could be involved in accelerating fine sediment deposition and impoverishing fish habitat, but our sediment budgeting demonstrates that crayfish can be an effective cause of fine sediment loading that contributes to degradation of benthic habitat.

6.5.6 Further Discussion

The influence of biota on geophysical processes has often been dismissed, typically on the basis that the availability of biological energy is not able to match that of geophysical energy available for mobilising sediment, even though it has been estimated that there is approximately 1.1 x 106 more energy globally from net primary productivity than from purely geophysical sources (Philips 2009). Whilst the vast majority of this energy will not be expended in undertaking geomorphic activity, only a tiny proportion would be required to have a considerably influence in geomorphic processes. This is also on the assumption that output of energy applied to moving sediment, either biological or geophysical, is equal. The results from this monitoring and modelling show how small initial modifications created from biological energy can catalyse change on a greater scale enacted by geophysical energy, with 49 times as much sediment being eroded as a secondary process compared to directly from burrowing along at the River Bain. Whilst crayfish burrows directly recruited 0.6% and 10.4% of the bank scale sediment yield at the River Bain and Gaddesby Brook respectively, secondary erosional processes recruited 29.8% and 89.6% of sediment at both rivers respectively. Table 6.13

261 demonstrates the recorded relative erosion undertaken by purely geophysical processes, purely biological processes, and secondary processes as a result of the biological input. This demonstrates the key role that biological energy has in recruiting sediment to river systems, and driving change at the landscape scale.

6.5.7 How does burrowing impact bank erosion through time?

All results presented thus far have utilised erosion rates over the full study period to gain a long term perspective, because the episodic nature of both fluvial erosion and mass movements makes a longer term estimate of retreat more robust. The significant associations identified over this long term period were not typically detectable over short time periods between individual recordings (Table 6.16). When considered individually, associations between erosion and burrow metric were very weak (r > 0.400 in only 4 of 12 results shown), and there were only significant associations when short term, adjacent recordings were considered (maximum period to 252 days).

However, where the cumulative erosion over time since the start of the study is analysed against the strongest correlating variable, as in Table 6.17, the strength of these associations increased through time. Indeed, no significant associations were recorded after 263 days at the River Bain, with the first recorded significant association occurring after 515 days of monitoring. Whilst previous studies investigating red swamp crayfish (Procambarus clarkii) burrows into rice paddies in Spain observed burrow-driven collapses over just eight days (Arce and Dieguez- Uribeondo 2015), the importance of long term monitoring for both river science (e.g. Parr et al. 2003; Burt et al. 2008; Jaffe et al. 2008) and invasive species (e.g. Blossey 1999) has been well documented. Harvey et al. (2019) propose a selection of invasive species known to create burrows that may be of potential geomorphic threat, and these results demonstrate that monitoring over longer timescales may be required to understand their geomorphic impacts.

262

Section Period 1 Period 2 Period 3 Period 4 Full Study Length Metric Used r p r p r p r p r p Field Site All retreat

Total Retreat (V A ) 0.147 0.728 0.165 0.696 0.517 0.126 0.661** 0.027 0.744*** 0.004 River Bain Diffuse Erosion

Gross retreat (V L ) 0.052 0.903 0.359 0.383 0.248 0.489 0.039 0.916 0.709** 0.011 River Bain Only collapse retreat

Percentage of pins collapsed (V A ) 0.164 0.697 0.453 0.259 0.627* 0.053 0.310 0.353 0.688*** 0.010 River Bain

Table 6.16: The association between burrow volume density per meter of riverbank (VL) and bank retreat variables since the last recording, and over the total period. The most strongly correlated variables from each section are presented. Significant values are denoted by asterisks (* = p < 0.1, ** = p < 0.05; *** = p < 0.01).

Section Period 1 Period 2 Period 3 Period 4 Metric Used r p r p r p r p Field Site All retreat

Total Retreat (V A ) 0.147 0.728 0.161 0.703 0.588* 0.074 0.744*** 0.004 River Bain Diffuse Erosion

Gross retreat (V L ) 0.052 0.903 0.361 0.379 0.107 0.770 0.709** 0.011 River Bain Only collapse retreat

Percentage of pins collapsed (B L ) 0.164 0.697 0.415 0.266 0.647** 0.043 0.688*** 0.010 River Bain

Table 6.17: The association between burrow volume density per meter of riverbank (VL) and bank retreat variables since the start of the study. The most strongly correlated variables from each section are presented. Significant values are denoted by asterisks (* = p < 0.1, ** = p < 0.05, *** = p < 0.01).

263

6.5.9 Discussion Summary

These results have directly quantified the influence of animal burrows on geomorphic processes for the first time. They have quantified the direct and indirect influence of animal burrows on riverbank erosion and shown that the presence of signal crayfish burrows significantly increases rates of bank retreat. Burrowing steepened banks, significantly increased the rate of diffuse erosion, and increased the spatial extent of mass collapses. These results support the hypotheses of Harvey et al. (2019) that animal burrows may alter riverbank erosional processes by promoting geotechnical failures and by altering hydraulics at the bank face. The strength of associations observed throughout this monitoring increased through time. It may therefore be that with continued monitoring, the strength of these associations, and thus the resultant accuracy of prediction may increase. The strength of the associations shown here must therefore be treated as conservative estimates of the true association of the influence of burrows on riverbank retreat.

Sediment budgeting shows that accelerated bank erosion caused by the presence of burrows recruited 49.2 times the mass of sediment from burrowing alone at the River Bain and contributed 12.2% and 29.8% of the total recruited sediment at reach and bank scales, respectively. This demonstrates the importance of burrowing in priming riverbanks to catalyse more substantial, secondary erosion that would otherwise be absent. This is consistent with zoogeomorphic impact of other animals in rivers, including many benthic feeding fish, which impact sediment dynamics primarily by altering the propensity of sediment to move, rather than by directly moving it (Rice et al. 2019).

More generally, this work supports the suggestion that biological processes can be a substantial driver of geomorphological change, supplementing the geophysical processes that geomorphologists tend to focus on. The budgeting of sediment recruitment into biotic (direct burrowing), abiotic (erosion in the absence of burrowing) and interactive (abiotic erosion facilitated by burrowing) components is novel and adds to a small body of work comparing biotic and abiotic contributions to sediment dynamics. These calculations enable the first comparative estimates of the relative importance of biotic, abiotic and interactive processes in driving fine sediment recruitment to river systems and confirm that signal crayfish burrowing can contribute a significant proportion of the fines delivered to infested streams.

264

Chapter 7

The impact of simulated burrows on riverbank retreat: laboratory experiments

265

7.1 Introduction

Chapter 4 and Chapter 6 demonstrated significant associations between signal crayfish burrow metrics and accelerated bank retreat, and suggested that crayfish burrows may alter the dominant processes involved in riverbank erosion. This supports evidence from burrowing taxa in mesocosms (Onda and Itakura 1994; Vu et al. 2017) and field observations of crayfish burrows (Sibley 2000; West 2010; Telegraph 2016) that suggest that burrows have the capacity to drive large scale change on rivers. Whilst Chapter 6 showed a strong crayfish burrow effect, the processes and mechanisms driving change can only be inferred because discrete, time-point field monitoring does not allow direct observation of the mechanisms that cause erosion. Physical laboratory experiments potentially allow the underpinning processes to be investigated through replicated and controlled experimental design. This may include evaluating the role of specific factors in changing erosion rates; for example, different burrow densities and location patterns. Harvey et al. (2019) suggested various mechanisms by which burrowing may alter bank erosion processes and called for quantitative research into the relative importance of burrow properties in driving riverbank retreat. Chapter 6 provides quantitative field evidence to address this challenge, and this chapter uses a replicated laboratory-based experimental approach to further explore these associations.

7.2 Aims

This chapter aims to address aspects of objective 3 (to examine the processes by which crayfish burrowing impacts the geomorphology of riverbank erosion). Specifically, this research aimed to empirically test associations between burrow characteristics, and riverbank retreat rates while observing retreat mechanisms. Specifically, a Friedkin flume was purpose built to test three hypotheses:

1. An increase in burrow density will increase the rate of erosion.

2. The vertical positioning of burrows in the bank will influence the rate of erosion. In the field, crayfish burrows were found distributed throughout the full vertical height of the riverbank, and burrows at the base of the bank are hypothesised to result in greater collapse and erosion than burrows higher up the bank, due to the greater weight of sediment above the burrowed area.

266

3. The spatial organisation of burrows will influence erosion. Clustered burrows are hypothesised to have a cumulative effect in reducing the strength of the bank, and may cause greater accelerated collapse compared to the same number of burrows distributed more evenly.

7.3 Methods

7.3.1 Physical Set Up

Scaled experimental set ups have been commonly used to study natural river systems, including bedrock (e.g. Johnson and Whipple 2007; Grimaud et al. 2016; Baynes et al. 2018a; reviewed in Lamb et al. 2015), alluvial (e.g. Frankel et al. 2007), and cohesive sediment (e.g. Friedkin 1945; Kleinhans et al. 2010; van de Lageweg et al. 2010; van Dijk et al. 2013a; van Dijk et al. 2013b; Kleinhans et al. 2014; Bodewes et al. 2017) systems. The hydraulics of experimental channels have been shown to accurately scale to flows observed in nature (Postma et al. 2008; Baynes et al. 2018b), and the processes observed under laboratory conditions when scaling cohesive sediment reflect those witnessed in the field (Lamb et al. 2015). This allows for events that occur over long timescales, and events that occur infrequently, to be studied in a manner that would not be possible via field monitoring.

To investigate the influence of crayfish burrows on the process of riverbank collapse, a Friedkin flume was constructed (Friedkin, 1945; Figure 7.1; Figure 7.2), following a standardised design used to investigate scaled bank stability. The Friedkin flume consists of a fixed channel of regular design that directs water at a standard discharge toward a simulated riverbank at a standard angle of attack. Erosion is measured as the rate of bank recession made by repeat observations over time. Previous research has used this technique to investigate bank stability under varying conditions of bank material grain size, chemical cohesive agents, and vegetation colonisation (Kleinhans et al. 2010; van de Lageweg et al. 2010; van Dijk et al. 2013a; van Dijk et al. 2013b; Kleinhans et al. 2014; Bodewes et al. 2017). The design and flow characteristics used here were the same as those used in previous studies, to ensure that the results obtained were directly comparable. The key difference in the current experiments was the introduction of simulated burrows into the riverbank, at different (i) densities, (ii) vertical spacings and (iii) spatial organisation, to address the three hypotheses above in replicated

267 experiments. These experiments were carried out in association with Bas Bodewes (PhD Researcher, University of Hull), using sediment mixtures and measurement protocols that his experiments (examining the stabilising effects of vegetation on riverbanks) suggested would be appropriate for evaluating burrow effects.

The Friedkin flume is a simple, heuristic, process-focussed physical model that allows for basic experimentation of how different treatments (simulated burrows in this case) effect bank erosion. The flume is not formally scaled in any way, and does not represent any specific location. This form of scale modelling allows for the investigation of processes and morphological responses to be compared to natural systems (Baynes et al. 2018c), and so whilst no quantitative associations can be applied from this experimental setting, results can provide evidence about how burrows may affect real riverbanks.

The experimental setup consisted of a 1.1 m x 0.0 5m x 0.05 m channel, with a flow rate of 470 l h-1, and a flow depth of 20 mm. At the tail of the channel, a 0.28 m x 0.07 m x 0.03 m block of cohesive sediment was placed at 45° to the water flow (Figure 7.2). The simulated bank was created by mixing 800 g of sand (D10 234.3 µm; D50 359.5 µm; D90 542.1 µm) with 80 ml of water (100 ml kg-1) and 0.8 g of Xanthan Gum (1 g kg-1). Xanthan gum is a binding agent that causes the sand to act in a cohesive manner, to represent cohesive fine sediment. A relatively high concentration (1 g kg-1) of xanthan gum was used as burrows were hypothesised to increase the rate of erosion, and previous experiments had shown that 1 g kg-1 xanthan gum yielded the slowest rate of bank retreat with the mixture of sediment size used (Bodewes et al. 2017; Figure 7.3). The same sand as used in the block was used to line the channel to create hydraulic roughness, which minimizes scaling effects such as water surface tension (Peakall et al. 2007; van Dijk et al. 2012).

268

Figure 7.1: Friedkin flume setup: (a) Plan view schematic of Friedkin flume; (b) the Friedkin flume in the laboratory, showing (1) camera position, (2) surrounding [open] blackout curtains; (3) lamp, (4) sand block; (5) return weir and hose, (6) water sump, containing the return pump; (7) return hose, (8) return outlet; and (9) flow channel; (c) a (cropped) overhead photograph of the setup taken by the GoPro camera; and (d) the ‘riverbank’ cohesive sand block setting in a mould in situ.

269

Figure 7.2: Schematic of Friedkin flume set up. (a) Oblique view schematic of the bank; and (b) oblique view photograph of the experimental bank.

Test blocks were created by mixing all ingredients in a food mixer over a period of two minutes and turning the mixture into an aluminium mould that was then left to set for 165 minutes in situ (Figure 7.3c). ‘Burrows’ were created by using 45 mm long, 6 mm diameter wooden dowels, which were placed in the sand mix before it set, in specified locations within the aluminium mould. The mould had fourteen holes where burrows could be located, distributed across two vertical rows. Dowels on the bottom row were centred 5 mm from the bottom of the bank (covering 2 - 8 mm in the vertical) and dowels on the top row were centred at 15 mm from the bottom of the bank (12 - 18 mm). The aluminium mould and dowels were removed prior to the start of the experimental run, with the tubular hole in the bank replicating a burrow.

270

Figure 7.3: Reduction in average block volume for a block of sand undergoing erosion in a Friedkin flume. Different lines are for the same sand grain mixture, but with increasing cohesion as a result of the properties from the concentration of xanthan gum (XG, red) and carrageenan (CG, blue) used. Black lines are three bare sediment settings as reference. Reproduced from Bodewes et al. (2017) with permission.

The two rows of burrows were both beneath the depth of the flow (20 mm) during experimental runs, and so were subject to fluvial erosion.

The inclusion of burrows above the waterline was also considered because exposed burrows are common during low flow conditions. Exposed burrows could influence erosion by either (a) reducing the tensile and rigid strength of the bank, promoting collapse, or (b) reducing the mass acting upon the bank beneath, reducing bank mass and therefore the chance of collapse. However, it was hypothesised that burrows would have the strongest influence on erosional processes during high flow events when they would be submerged, and so only the impact of submerged burrows was tested.

271

A GoPro Hero 5 camera was mounted at a height of 0.96 m directly above the sand block to record images of the retreating bank every 60 seconds throughout each experimental run. A blackout curtain was suspended around the flume to exclude ambient light, and a desk lamp was positioned to create a high-contrast shadow at the retreating edge of the bank. Additional photographs were also taken from an oblique position to record the nature of collapse. Runs ended when bank erosion reached 40 mm into the bank, to ensure that experimental runs ended prior to reaching the burrow termini which were initially 45 mm deep. Previous Friedkin studies (e.g. Kleinhaus et al. 2014) terminated experimental trials when half of the sediment block (50 mm) was eroded; stopping the runs at 40 mm therefore gives a comparable experimental methodology to previous research.

3.2 Treatments

To test the three hypotheses above, seven treatments were tested (Figure 7.4), with five replicates of each treatment were undertaken, providing a total of 35 experimental runs.

1. No burrows

2. 4 burrows, randomly distributed amongst the 14 possible positions

3. 8 burrows, randomly distributed amongst the 14 possible positions

4. 12 burrows, randomly distributed amongst the 14 possible positions

5. 7 burrows, filling the bottom row

6. 7 burrows, filling the top row

7. 7 burrows, clustered into one localised area

Random positions were assigned using a random number generator and recorded for analysis.

272

Figure 7.4: The location of ‘burrow’ positions in experimental treatments. Five runs per burrow treatment were undertaken. The location of randomly assigned burrows changed between treatments; those shown here are examples.

273

7.3.3 Image Analysis

Image analyses were undertaken using photographs from the GoPro camera mounted above the flume. Images were automatically processed to calculate the rate of bank retreat by tracing the edge of the shadow (at the boundary of the sand block) from the lamp created by the eroding face of the bank. Image analysis was undertaken in Matlab, using a script based on the script of Wout van Dijk (2013b), which was adapted for Friedkin experiments by Bas Bodewes (2017), and adapted by Bas Bodewes and the author for the current experiments.

Image analysis was undertaken according to the procedure illustrated in Figure 7.5:

(i) Each raw image was rectified to account for camera distortion, to ensure that each pixel recorded an equal area of sediment. Distortion was quantified by photographing a checkerboard pattern of known dimensions in multiple (n = 40) locations across the experimental setup (Figure 7.6), and raw images were rectified in Matlab to represent the true area.

A second Matlab script processed rectified images through to completion:

(ii) The images were turned to greyscale by isolating the red channel from the RGB image data. (iii) Images were cropped to the area of interest. The area and rotation of the cropped images from each run was manually validated and adjusted to account for small movements of the location of the bank edge caused by inadvertent flume setup movement between runs and human error in installing the bank. (iv) A threshold was applied to convert each image into only black or white, and thus isolate the shadow cast by the retreating bank face from the top of the bank, which was brightly illuminated by the lamp. (v) Morphological operations were then employed to remove clusters of (a) black pixels that were detached from the block shadow, and (b) white pixels that were embedded within the main shadow block to create one ‘shadow’ area. (vi) The boundary of the ‘shadow’ area was then determined. (vii) The boundary line was then combined with lines imposed marking the upstream, downstream, and rear of the study area of the sand block to create a shape file of the study area. The area within the four lines (rear of the block, upstream extent of

274

the study area, downstream extent of the study area, and the retreating bank face) was quantified. (viii) The area calculated was then compared to the area calculated for the previous image. If the number of pixels identified as part of the sand block exceeded that of the previous image, the threshold value was increased for the conversion to a black and white image (step iv), and the process repeated. This was undertaken to account for changing ambient light conditions. Whilst the blackout curtains excluded the majority of external light, the laboratory lights detectably changed the background lighting, and so the threshold was increased to ensure that the edge of the bank was being detected. This ran for a maximum of 11 cycles. If the issue was not resolved after 11 cycles, the image was flagged for manual verification. If the number of pixels present was less than the previous image, the block area was recorded for the timestep, and the next image considered. (ix) Recorded outputs were then exported for analyses.

7.3.4 Data Analyses

Bank retreat data was analysed to identify stages, rates and features of retreat (Figure 7.7), which could be analysed both conjointly and independently. Bank area values were converted to the percentage of the original block remaining for ease of visualisation and reporting. Several aspects of bank retreat were calculated and examined:

7.3.4.1 Total Retreat

The rate of erosion (mm minute-1) over the entire run was calculated. All datapoints from the first data point after the initial retreat had stabilised until the run had reached its minimum value were used, and the rate of erosion was calculated using linear least squares regression (Figure 7.7).

275

Figure 7.5: Flowchart of image analysis undertaken to record the rate of bank retreat.

276

Figure 7.6: Example images used in the rectification process. Photographs of a checkerboard pattern of known dimensions were taken throughout the experimental setup, which informed a Matlab script to rectify experimental images to account for camera lens distortion.

7.3.4.2 Collapse Retreat

Collapses were identified from bank retreat graphs as step changes in the rate of erosion (Figure 7.7). All suspected collapses were corroborated with photographic evidence such that step changes were classified as collapses only if a notable section of bank could be visually identified as having collapsed between images. The dimensions (length, width, and area) of each collapse were then calculated using the rectified images. These data were used to produce six collapse metrics:

1. Number of Collapses. The total number of collapses that occurred during the run.

2. Retreat as Collapse (%). The proportion of the total retreat caused by collapses during the run.

277

3. Total Collapse Width (mm). The sum of the width of all collapses that occurred during the run.

4. Mean Collapse Width (mm). The mean width of all collapses that occurred during the run.

5. Total Collapse Length (mm). The sum of the length of all collapses that occurred during the run.

6. Mean Collapse Length (mm). The mean length of all collapses that occurred during the run.

7.3.4.3 Diffuse Erosion

The rate of diffuse erosion (mm minute-1) was calculated to quantify the rate of bank retreat caused by erosion other than collapse events. Collapses were removed by increasing the post- collapse values by the size of the collapse to leave a smooth erosion pattern without the collapse steps (Figure 7.8). All datapoints from the first datapoint after the initial retreat had stabilised until the run had reached its minimum value were used, and the rate of erosion was calculated using linear least squares regression (Figure 7.8).

7.3.4.4 Initial Retreat

The initial rate of retreat has previously been observed to be greater than that observed later in the run (van Dijk et al. 2013b; Bodewes et al. 2017; Figure 7.3). The proportion of sediment lost (%) prior to the rate of erosion stabilising was recorded.

278

Figure 7.7: (Top) Bank retreat over time in a Friedkin run. The black dots are observed readings, and the grey line is the applied regression line. The dark grey bar highlights the initial collapse after the flume is started, the lightest grey background highlights periods of standard bank retreat, and medium grey bars highlight step changes that were identified visually from photographs as bank collapses. (Below) a pair of images from before (middle) and after (bottom) the first bank collapse event in the run illustrated above, denoted by a cross, with the blue lines superimposing the line of the bank edge in the opposing image.

279

Figure 7.8: Bank retreat over time in a Friedkin run with (a) collapse and diffuse erosion, and (b) with collapse events removed. The black dots are observed readings, and the grey line is the applied regression line.

7.3.5 Statistical Analyses

All variables other than total retreat were not normally distributed (Shapiro-Wilk p < 0.05) for at least one treatment, and so non-parametric ANOVA tests (Kruskal-Wallis (H)) were used to identify differences in erosion metrics between treatments. A significance value of α < 0.1 was applied because:

iv) There were a low number of replicates (n = 5); v) high variability within runs was expected and observed, as with Bodewes et al. (2017); vi) non-parametric tests, which are of a lower power, were used; and vii) these experiments are exploring a new area of research; they do not seek to prove or disprove quantitative theories, but explore processes underpinning hypotheses and field observations.

280

One outlying experimental run (no burrows) was detected during image analysis (13.01 * IQR > 3rd quartile of total retreat rate) and was subsequently removed prior to statistical analysis.

To investigate the impact of burrow density on erosion, Kruskal-Wallis tests were undertaken for all retreat metrics between runs where 0, 4, 8, and 12 burrows were present. To investigate the impact of the vertical positioning of burrows in the bank on erosion, Kruskal-Wallis tests were undertaken for all retreat metrics between runs where 7 burrows were present at the top of the bank, and where 7 burrows were present at the bottom of the bank. To investigate the impact of the spatial organisation of burrows on erosion, Kruskal-Wallis tests were undertaken for all retreat metrics between runs where 7 burrows were clustered together, compared to 7 burrows positioned at the top or bottom of the bank, and 8 burrows randomly distributed throughout the bank. Mann-Whitney U (U) tests were used for post-hoc pairwise comparisons in all cases. All statistics were performed using SPSS Version 23 (IBM 2015).

7.4 Results

Throughout the experiment, image analysis of the Friedkin flume runs successfully monitored erosion, consistent with the retreat visually observed between images. Collapses were successfully identified and quantified, and erosion rates were calculated for each specified metric. The Friedkin apparatus functioned as in previous published experiments, with accelerated erosion at the beginning of each experimental run (van Dijk et al. 2013b; Bodewes et al. 2017). The use of xanthan gum with sand to create cohesive sediment was successful. Across all runs, the bank maintained a near-vertical profile. Across all runs diffuse erosion dominated, with collapses contributing 11.3% of total retreat. This compares to field measurements at the River Bain (a cohesive sand system) and Gaddesby Brook (a cohesive clay system), where collapses contributed 57.5% and 0.6% to overall retreat, respectively (Chapter 6).

Across all dependent variables there were few significant differences associated with either burrow density or location treatments when all treatments were compared (Table 7.1). The rate of total retreat (H3 = 7.530, p = 0.057), diffuse erosion (H3 = 7.143, p = 0.067), the number of collapses (H3 = 7.225, p = 0.065), and the sum of collapse widths (H3 = 7.086, p = 0.069) were the only variables where significant differences existed between treatments.

281

7.4.1 The effect of burrow density on bank retreat

A significant difference in total bank retreat was observed between the burrowed and non- burrowed runs. Whilst there was no difference between total rate of retreat between burrow density treatments, the total retreat of all burrowed sand blocks was significantly slower than non-burrowed blocks (U = 19.0; p = 0.032 for all treatments; Figure 7.9a; Figure 7.10; Table 7.2), which was not the anticipated result. Greater variability in timeseries of erosion was observed between runs with 8 and 12 burrows than runs with 4 burrows, although there was also a high amount of variation between runs with no burrows. Results were similar when the influence of bank collapses was removed; diffuse erosion rate was significantly lower in each burrowed treatment than the treatment without burrows (4 burrows: U = 18.0; p= 0.063; 8 burrows: U = 18.0; p = 0.063; 12 burrows: U = 20.0; p = 0.016) and there was no significant difference in rates of erosion between the different burrow densities when zero burrows was excluded (H2 0.380; p = 0.827; Figure 7.9b).

Burrow Density Burrow Location

Test Statistic Test Statistic (H = 3 in all cases) Significance (H = 3 in all cases) Significance Total Retreat 7.530 0.057* 4.189 0.242 Diffuse Erosion 7.143 0.067* 3.526 0.317 Number of Collapses 7.225 0.065* 3.373 0.338 Retreat as Collapse (%) 3.332 0.343 2.116 0.549 Total Collapse Depth 7.086 0.069* 2.784 0.426 Mean Collapse Depth 0.086 0.993 1.696 0.638 Total Collapse Length 5.890 0.117 1.903 0.593 Mean Collapse Length 3.414 0.332 0.557 0.906

Initial Retreat 1.846 0.605 5.194 0.158

Table 7.1: Results of Kruskal-Wallace tests for significant differences between the seven treatments described in Section 7.3.2, for each of the bank retreat metrics described in Section 7.3.4. Significant results are denoted by an asterisk (* α < 0.1).

282

Figure 7.9: The effect of burrow density on the rate of total bank retreat when (a) total retreat, and (b) diffuse retreat are considered, considering mean values +/- 1 standard error (SEM).

) )

-1 -1

Total Total Retreat Rate (mm min Rate Diffuse Erosion (mm min of Collapses Number Retreat as Collapse (%) Total Collapse Depth (mm) Collapse Individual (mm) Depth Total Collapse Length (mm) Collapse Individual (mm)Length Initial Collapse(%) Loss

No Burrows 0.142 0.106 2.8 11.8 29.7 11.1 124.5 44.1 22.2

4 Burrows 0.079 0.073 1.0 8.0 11.3 11.2 48.7 45.9 17.6

Mean 8 Burrows 0.085 0.074 1.8 25.4 20.2 12.9 72.0 33.7 18.2

12 Burrows 0.071 0.066 0.6 4.8 7.8 12.1 21.2 33.8 20.7

No Burrows 0.136 0.099 2.5 11.3 29.2 10.9 12.8 48.8 22.4

4 Burrows 0.866 0.077 1.0 5.1 11.1 11.5 33.7 46.1 15.5

Median 8 Burrows 0.091 0.074 1.0 9.1 18.4 10.7 26.1 31.8 19.9

12 Burrows 0.062 0.066 0.0 0.0 0.0 12.1 0.0 33.8 21.4

Table 7.2: Bank retreat metrics (n = 5) associated with increasing burrow density from Friedkin flume experiments.

283

Figure 7.10: The rate of bank retreat over time in Friedkin experiments when (a) zero, (b) four; (c) eight, and (d) twelve burrows were present. Coloured lines show mean values, and grey lines show individual experimental runs. (e) A direct comparison of mean values of retreat under increasing burrow density.

284

Collapses occurred significantly more frequently when burrows were absent (H3 7.225; p = 0.065; Figure 7.11b), but collapses that did occur under higher burrow densities (8 and 12 burrows) were significantly wider (H3 7.086; p = 0.069) than those associated with lower burrow densities (0 and 4 burrows; Figure 7.11c). The absence of burrows also resulted in the highest mean initial collapse at the start of the experimental runs. However, when only burrowed runs were considered, a small increase was observed in the volume of sediment lost in the initial collapse phase as burrow density was increased (median = 4: 15.5%; 8: 19.9%; 12: 21.4%; Figure 7.12). There were no significant differences between treatments when considering the proportion of erosion occurring as collapse (Figure 7.11a) or the length of collapses (Figure 7.11d).

285

Figure 7.12: The effect of burrow density on the rate of initial collapse, considering mean values +/- 1 standard error (SEM).

7.4.2 The effect of vertical burrow position and clustering on bank retreat

The location of burrows had an effect on bank erosion rates. The total rate of retreat was faster for banks where seven burrows were clustered compared to banks with a similar density of burrows (eight) that were placed in random locations (median = 0.139 mm min-1 and 0.091 mm min-1 respectively, U = 22.0; p = 0.056; Figure 7.13a). Compared with eight randomly distributed burrows, clustered burrows were associated with significantly greater initial collapse (median = 19.9% and 26.0% respectively; U = 22.0; p = 0.056; Figure 7.13b).

There was little difference in the total rate of erosion between burrows placed at the bottom of the bank and clustered together (median = 0.163 mm min-1 and 0.139 mm min-1 respectively), but both treatments resulted in a higher total erosion rate than burrows at the top of the bank (0.100 mm min-1; Figure 7.14). Very high variability was observed in all burrow location runs (Figure 7.15). When only diffuse erosion was considered, this variability was substantially less, with almost no difference occurring between the three treatments (Figure 7.14b; Table 7.3).

286

Figure 7.13: The influence of burrow location on (a) the rate of total bank retreat, and (b) the size of the initial collapse, considering mean values +/- 1 standard error (SEM).

Figure 7.14: The influence of burrow location on the rate of bank retreat when (a) total retreat, and (b) diffuse erosion are considered, considering mean values +/- 1 standard error (SEM).

287

Figure 7.15: The rate of bank retreat over time in Friedkin experiments when burrows were (a) on the top row; (b) on the bottom row, and (c) clustered. Coloured lines show mean values, and grey lines show individual experimental runs. (d) A direct comparison of mean values of retreat when differing burrow locations were considered.

288

) )

-1 -1

8 Burrows 0.085 0.074 1.8 25.4 20.2 12.9 72.0 33.7 18.2 Randomly Distributed 7 Burrows 0.158 0.133 1.4 8.2 14.8 10.2 54.2 37.1 27.3 Bottom of Bank

Mean 7 Burrows 0.125 0.105 1.6 7.2 15.8 10.4 62.0 34.4 19.2 Top of Bank 7 Burrows 0.193 0.126 3.0 13.7 30.2 10.2 100.8 32.2 28.8 Clustered 8 Burrows 0.091 0.074 1.0 9.1 18.4 10.7 26.1 31.8 19.9 Randomly Distributed 7 Burrows 0.163 0.153 2.0 9.3 1.6 8.7 63.4 3.5 20.3 Bottom of Bank 7 Burrows

Median 0.100 0.099 1.0 5.7 1.2 9.8 19.5 3.9 21.5 Top of Bank 7 Burrows 0.139 0.096 3.0 14.3 2.9 10.1 89.8 2.6 26.0 Clustered

Table 7.3: Bank retreat metrics (n = 5) of different burrow locations from Friedkin flume experiments.

Whilst there was no difference in the width, length, or area of collapses that occurred between treatments, clustered burrows resulted more collapses occurring per run compared to when burrows were evenly distributed throughout the reach (median = top: 1.0; bottom: 2.0; clustered = 3.0; Figure 7.13d; Table 7.3), which resulted in a higher percentage of erosion occurring directly as a result of collapse (median = top: 5.7%; bottom = 9.3%; clustered = 14.3%; Figure 7.16; Table 7.3). Clustered burrows also resulted in the highest loss of material from initial collapse compared to when burrows were evenly distributed throughout the reach (Figure 7.17).

289

290

Figure 7.17: The effect of burrow locations on the rate of initial collapse, considering mean values +/- 1 standard error (SEM).

7.5 Discussion

Results observed in Chapter 6 indicated significant associations between the density of burrows and riverbank retreat, and the Friedkin flume experiments were designed to further investigate these field observations under controlled conditions. Parameters of the Friedkin flume were constructed to provide a repeatable, robust experimental setting, and were not measured against, and did not attempt to replicate, any natural system. Results from the experiments provide a semi-quantitative, deeper understanding of the processes occurring in burrowed river systems, but cannot yield quantitative predictions of burrow effects.

A criticism of the Friedkin methodology is the nature of the changing bank. As the bank retreats, the adjacent channel widens, increasing from a width of 50 mm to 90 mm by the end of an experimental run. This reduces the depth and velocity of the flow, and therefore the shear stress acting on the bank. This could affect erosion rates, but no notable decrease in erosion rates was observed through time (excluding the initial collapse phase). Moreover, the change in channel dimensions was consistent in all runs and all treatments, and so does not influence between-treatment differences.

Throughout the experiments, there were few statistically significant results. This may reflect the low number of replications (n = 5) undertaken, and the high variability observed between

291 runs. All previous studies using Friedkin Flume experiments have reported exclusively qualitative results, without statistical analysis (Friedkin 1945; Kleinhans et al. 2010; van de Lageweg et al. 2010; van Dijk et al. 2013a; van Dijk et al. 2013b; Kleinhans et al. 2014; Bodewes et al. 2017). It may be that due to the nature of the experiments undertaken, along with the relatively low number of replications, significance testing is not an appropriate method for assessing these results.

7.5.1 The effect of burrow density on bank retreat

7.5.1.1 Diffuse Erosion

It was hypothesised that increasing burrow density would lead to greater bank retreat, as was observed in the field (Chapter 6), but the opposite was observed in the Friedkin sand blocks. Erosion rates were the reverse of those hypothesised; experimental runs with no burrows eroded significantly faster than runs with burrows, and there was little difference between burrowed treatments (Figure 7.9; Figure 7.10). This was true for total retreat, diffuse erosion and retreat due to collapses. It is possible that this unexpected outcome is an artefact of the use of xanthan gum to provide bank cohesion. As a result, the simulated burrows made the banks more resistant to erosion by changing bank drainage properties and therefore the cohesion provided by the gum.

Mass wasting of cohesive materials occurs when erosional forces exceed resistive forces, and can be defined by the revised Mohr-Coulomb factor of safety failure criterion (Fredlund et al. 1978; Fredlund and Rahardjo 1993):

푆푟 = (휎 − 휇푤) tan(휗) + 푐 (Equation 7.1)

where Sr is the effective shear strength (kPa), σ is normal stress (kPa); µw is pore-water pressure (kPa), tan(φ) is the angle of internal friction (°); and c is cohesion (kPa).

Normal stress is defined by:

휎 = 푊 푐표푠훽 (Equation 7.2)

where W is the weight of the failure block (kN m-2), and β is the angle of the failure plane (°).

292

Banks fail when shear strength (Sr) is exceeded by the driving gravitational force (Sd; kPa):

푆푑 = 푊 푠𝑖푛훽 (Equation 7.3)

Failure can be calculated by the ratio between resisting and driving forces, and occurs when driving forces are dominant.

As water drains from a bank, it becomes more resistant to failure. Normal stress is reduced due to the reduction in mass, the angle of friction is increased due to reduced saturation and thus increased particle to particle bonding; cohesion is increased due to reduced pore water pressure and increased matric suction, and weight is reduced due to reduced saturation and thus lower bank mass (Rinaldi and Casagli 1999; Simon and Collison 2001; El-Dien et al. 2015; Fatahi et al. 2015). This change in resistance to failure by the bank is hypothesised to be affected by the presence of burrows, and is reviewed in Harvey et al. (2019).

The influences of bank drainage on bank resistance to failure were likely to have occurred during the setup for the experiment. As water was required to mix the sediment, and the banks were set in dry conditions, the bank became less saturated over time. The rate of drying would have been affected by burrows, which increased porosity and thus decreased the length of flow pathways. Flow paths were calculated as the mean distance of every particle in the bank to the nearest area of bank surface through an angle lower than 90o to vertically downwards. In the presence of 12 burrows, the mean drainage flow path distance of the study area of bank was reduced by 53.1% compared to baseline conditions (Table 7.4). The burrows also increased the surface area of the bank, with the presence of 12 burrows increasing the surface area exposed to the air by 61.3%, which would also contribute to an increased ability for the bank to drain.

The influences of bank drainage on bank resistance to failure is likely to have had substantial influence on the resistance of the bank to erosion, because the higher levels of saturation observed in the absence of burrows, and therefore more positive pore water pressure, would reduce the level of cohesion in the bank and increase the likelihood of mass failure and bank erosion (Simon et al. 2000; Simon and Collison 2001; Fox et al. 2007; Rinaldi and Darby 2007).

293

Surface area of Increase from Mean drainage Decrease from Relative drainage calculated block baseline flow path baseline index (mm2) conditions (%) distance (mm) conditions (%) No Burrows 16600 0 17.5 0.0 0.0 4 Burrows 19992 20.4 14.4 17.7 11.1 7 Burrows 22536 35.8 12.1 30.9 33.9 8 Burrows 23384 40.9 11.3 35.4 44.4

12 Burrows 26776 61.3 8.2 53.1 100.0

Table 7.4: Change in bank surface area and drainage flow path distance of the experimental bank in the presence of burrows.

The arguments above suggest that burrows may improve bank drainage at low flows, which could improve bank stability. However, in the field, bank failure scaled with burrow density suggesting that this effect, if present, was small relative to burrow-related drivers of instability (Chapter 6). This leaves open the question of why, in these experiments, erosion rate did not scale with burrow density, and it seems likely that the properties of the xanthan gum used as a cohesion agent are responsible.

Previous Friedkin experiments using xanthan gum have allowed the mixture to set for 15 minutes prior to the start of the experiment. Preliminary runs that adopted 15 minutes of drying revealed that the banks did not possess sufficient structural integrity to maintain a vertical profile or to maintain burrow cavities. Increased drying times produced harder sand blocks, so bank drying times (and thus cohesion) were increased until banks maintained a vertical profile and burrow subsidence did not occur. Oven-dried sediment blocks proved too brittle to be installed without breaking, and the rate of erosion of sediment blocks that were dried in situ overnight was too slow for individual experiments to be completed in a single day, which was a pragmatic requirement of the laboratory set up. After further trials, a drying time of 165 minutes was used, which was the shortest drying time that produced banks that could maintain a vertical profile while at the same time allowing for experimental runs to be completed in a single day.

The application of water was required to disperse and activate the xanthan gum, which set during dehydration (binding sand particles to sand particles). The efficiency of the xanthan gum used was therefore dependent on the ability of the substance to drain, with higher drainage resulting in a stronger setting of the xanthan gum and harder, less erodible test blocks.

294

Therefore, increasing the burrow density of the bank reflected three variables of two opposing effects, as opposed to simply destabilising effects as initially hypothesised:

i) Destabilising factors a. Burrow density ii) Stabilising factors a. Reduced bank saturation b. Increased xanthan gum cohesion

If the density of crayfish burrows had no impact on the resistance of the bank to erosion, as suggested by Figure 7.9 and Figure 7.10, then the rate of erosion would be expected to scale with the saturation of the banks and the strength of the xanthan gum, as observed in Bodewes et al. (2017; Figures 7.3 and 7.18 herein), where banks eroded more slowly with increased concentrations (and thus strength) of xanthan gum. In these experiments, a similar linear decrease in erosion rates could be expected if the strength of the xanthan gum was the only external influence on bank stability, and any deviation from this could be hypothesised to be a result of external factors, such as the influence of crayfish burrows. In the current experiments, a deviation from this trend was observed, with no trends or differences occurring between treatments with 4, 8, and 12 burrows. Therefore, it can be concluded that other external factors, other than the strength of the xanthan gum, were important in determining erosion rates, which can be attributed to the instability created by the burrows.

295

Figure 7.18: Conceptual model for expected erosion. (a) Considers the presence of stabilising impacts from drainage, and the absence of destabilising impacts of burrows. As the bank drains, the bank is stabilised by reduced pore water pressure and increased xanthan gum strength (blue), which reduces the expected erosion rate of the bank (black). (b) Considers the absence of stabilising impacts from drainage, and the presence of destabilising impacts of burrows. As burrow density increases (red), expected erosion increases (black). (c) Considers the presence of stabilising impacts from drainage, and the presence of destabilising impacts of burrows. The stabilising impacts are countered by the destabilising impacts of burrows, and so the expected erosion rate does not to change with increasing burrow density. Any deviation from a lack of association therefore represents an imbalance in effects, and thus provides evidence for the stabilising effects of xanthan gum and bank drainage, or evidence for the destabilising effects of burrows. Stabilising effects are expressed here as a power relationship as there are two contributing factors (pore water pressure and xanthan gum strength), and destabilising effects are expressed as a linear relationship as there is one contributing factor (burrow density).

296

Figure 7.19: (b) Observed erosion rate from Friedkin experiments, with reference to (a) conceptual model (see Figure 7.18).

It can therefore be hypothesised that the difference observed between the expected trend if burrows were to have no effect on bank retreat (Figure 7.19a) and the observed data (Figure 7.19b) is due to the influence of the burrows. However, it should be noted that the relative cohesive strength of xanthan gum and the destabilising effects of burrows are not quantified in this conceptual model, and so the slope of the expected erosion rate cannot be compared to the observed data. Nevertheless, the pattern observed suggests that the density of crayfish burrows is likely to have had a positive association with bank erosion in these experiments.

297

The absence of any trend between burrow density and erosion rates could also be interpreted as an absence of change in the resistance of the bank to erosion, and an absence of change in the resistance of the bank to changes in pore water pressure and xanthan gum strength. However, preliminary experiments without burrows demonstrated differences in bank retreat rates (Table 7.5) under changing:

i) Bank saturation: Banks with no burrows that were not allowed to dry eroded considerably more quickly than banks that were dried over a 2.75 hour and 24 hour period; ii) Bank saturation (initial conditions): Banks that had a higher initial water content (230 ml kg-1) eroded more quickly than banks with a lower initial water content (100 ml kg-1); iii) Xanthan gum concentration: Banks with a lower concentration of xanthan gum (0.5 g kg-1) were terminated due to immediate bank face failure to a shallow (≈45°) slope. Banks with a high (2 g kg-1) concentration of xanthan gum retreated more slowly than those with a medium (1 g kg-1) concentration.

These observations suggest that bank erosion rate was influenced by water saturation and xanthan gum strength. This suggests that the lack of any significant difference in erosion rate as density increases is indicative of a burrow density effect; it is likely that without the xanthan effect, erosion would scale with burrow density, consistent with observations from the field monitoring in Chapter 6.

Bank Retreat Bank Retreat Bank Retreat Preliminary Treatment (mm minute -1) Preliminary Treatment (mm minute -1) Preliminary Treatment (mm minute -1) Bank Drying Time Xanthan Gum Concentration Initial Water Concentration (1g kg -1 XG; 80 ml water) (2 hrs 45 mins; 80 ml water) (2hrs 45 mins; 1 g kg -1 XG) 0 hours 0.37 0.5 g kg -1 N/A 100 ml kg -1 0.25 2 hours 45 minutes 0.25 1.0 g kg -1 0.25 230 ml kg -1 1.00 24 hours < 0.09 2.0 g kg -1 < 0.09

Table 7.5: Rate of bank retreat under adjusted preliminary conditions.

298

7.5.1.2 Collapse Retreat

When only treatments with burrows were considered, the amount of bank material eroded during the initial collapse phase increased with burrow density (Figure 7.12). This suggests that any stabilisation caused by burrowing (because of greater drying of xanthan gum) was outweighed by reductions in stability caused by increasing numbers of burrows. The initial phase of collapse is a useful aspect of the Friedkin experiments to consider. Bank collapse is typically associated with high flow conditions (Julian and Torres 2006), and the initial phase of the experiments, representing the rising limb of a flood hydrograph (0 l hr-1 rising to 420 l hr-1) followed by persistent high flow (420 l hr-1) within the original channel dimensions, is possibly most representative of this field condition.

After the initial collapse had occurred, there were only small differences in erosion rates between treatments. In most experiments initial collapse recruited a large mass of material that was deposited at the base of the bank. Riverbank retreat is often mediated in mass failure driven systems through the deposition of submerged slump blocks of deposited material, which can promote erosion by directing flow toward the bank (Hackney et al. 2015) or protect the bank from high flow events by diverting flows away from the bank toe (Wood et al. 2001; Parker et al. 2011). The photographic method used in these experiments only recorded retreat from the top of the bank, and so processes such as undercutting and erosion of already collapsed sediment would not be accounted for here. These may explain the small differences in erosion rates that were recorded after the initial collapse phase between treatments, as erosion primarily occurred on the recruited material as opposed to the sand block.

Whilst there were no significant differences between the mass of sediment that eroded as a collapse between treatments (Table 7.1), differences were observed in the nature of collapse. Collapses that occurred in treatments with higher densities of burrows (8 and 12) were significantly wider, and although not significantly different, shorter in length than in treatments with lower densities of burrows (0 and 4; Figure 7.11). This suggests that at low burrow densities, collapses occurred as a whole bank process, meaning that collapses were occurring independently of the burrows, whereas burrows were responsible for driving collapse events at high burrow densities. When burrows were absent, collapse erosion typically occurred as a downstream process (Figure 7.20), whereas areas of the bank directly above burrows eroded locally, independently of the bank as a whole (Figure 7.21); increasingly modular bank retreat was observed with increasing local burrow density (Figure 7.22), where localised areas of

299 independent collapse occurred to the processes occurring along the full bank. This resulted in more localised, wider collapses that resulted in initially more pronounced variation in bank shape.

These observations support previous research that reported local scale collapses of P. clarkii burrows (Barbaresi et al. 2004), and the findings of Chapter 6, where an association between banks with a higher density of crayfish burrows and greater amounts of mass movement was observed in situ. This is supported by oblique imagery that demonstrates the direction of erosion occurring in the mass wasting of material under increasing burrow densities. When burrows were absent, erosional processes were first observed at the head of the bank, progressing downstream to the bank tail through time (Figure 7.21), whereas the direction of erosion observed when burrows were present was not parallel with the direction of flow (Figure 7.22). This therefore suggests that the nature of erosion changed when burrows were present, which drove localised collapse events.

Figure 7.20: Erosion through time in the treatment without burrows. Erosional processes were first observed at the head of the bank, progressing downstream to the bank tail through time. Flow is from right to left.

300

Figure 7.21: Oblique photographs of initial collapses in the presence of (a) four, (b) eight; and (c) twelve burrows. Arrows indicate the direction of erosion occurring at the moment of the photograph; the collapse observed in (a) was a rotational failure with sediment moving in the direction of the arrow; the collapse observed in (b) occurred from a fracture at the rear of the collapse, that resulted in the mass failure of sediment moving down the bank in the direction of the arrow, and the collapse in (c) started at the front of the bank, with collapsed sediment ‘pulling down’ sediment further into the bank, in the direction of the arrow. Erosion occurred locally, independently of whole bank processes, contrary to an absence of burrows (Figure 7.20), with collapses occurring directly above burrowed sections.

301

Figure 7.22: Processed images of banks after the initial collapse phase in the presence of (a) zero, (b) four; (c) eight, and (d) clustered burrows. Increasingly modular bank retreat is seen with increasing local burrow density.

7.5.2 The impact of vertical burrow position and clustering on bank retreat

In the burrow location experiments, seven burrows were created in all treatments, with the position, as opposed to the density of burrows, as the dependent variable. As all treatments were subject to the same burrow density, the methods employed are exempt from concerns about whether differences in burrow density led to differences in drying of xanthan and thence bank hardness (section 7.5.1.1).

Burrow distribution was important for determining bank retreat rate. Banks with burrows clustered in one location retreated considerably faster than banks with burrows arranged longitudinally (Figure 7.14; Figure 7.15). The retreat values reported in the current experiments are for the whole study area of the bank, and not for the localised area over which the clusters were positioned. Therefore, either (a) bank retreat was exceptionally high in the clustered area,

302 with no or little observed impact beyond the affected part, or (b) clustered burrows influenced the erosion rate of neighbouring sections of bank where no burrows were directly present. Consecutive aerial images show that the process driving this difference was accelerated collapse, with collapse metrics being approximately twice as high in runs with clustered burrows compared to other treatments (Table 7.3). Accelerated erosion was then recorded along the rest of the bank, particularly downstream of the collapse. This suggests that that even if burrows are dug in relatively limited areas, their detrimental geomorphological impacts may be observed beyond directly impacted reaches.

Differences in the same number of burrows located in different patterns suggest a compounding effect of burrows on erosional processes, which is important, because it suggests that previous numerical modelling considering the influence of one burrow on bank hydraulics and geotechnics (e.g. Camici et al. 2014; Orlandini et al. 2015; Taccari and van der Meij 2016a; Borgatti et al. 2017; Saghaee et al. 2017) may be underestimating the net effect of burrows as a whole on erosional processes. Clustered and randomly distributed burrows generating different amounts of erosion is also noteworthy because burrows are typically clustered in field settings. Stanton (2004) observed clustering of burrows in Gaddesby Brook, with the modal -1 -1 burrow density being 0 burrows m of riverbank, with some sites exceeding 9 burrows m . Burrows are typically clustered in field settings, with field surveys recording burrows occupying 0.2 - 23.5% (median 3.2%) and 0.4 – 27.1% (median 10.8%) of riverbank (Faller et al. (2016) and Chapter 3 respectively). In Chapter 3, two sites were surveyed where burrows exceeded a density of 1 burrow per meter of riverbank for the full survey (Gaddesby Brook (1.02 burrows m-1) and Broadmead Brook (1.05 burrows m-1)) but the mean distance between burrows was 0.27 m and 0.25 m respectively, which equates to burrows clustering in just 27.1% and 26.1% of available riverbank length. Whilst the Friedkin experiments are not intended to scale to real world outcomes, the patterns tested replicate those observed in nature (Figure 7.23), and so the patterns observed here are relevant for natural systems. It can therefore be hypothesised that rivers where burrows are heavily clustered are likely to experience greater rates of erosion than those where burrows are more evenly distributed. There is currently no research into why burrows are clustered together, and given the frequency of burrow clustering in the field, future research should aim to investigate the mechanism by which clustering promotes erosion.

303

Burrows arranged along the top and bottom row also had a faster retreat than seven randomly distributed burrows (Figure 7.13). This may be due to neighbouring burrows interacting to cause bank undercutting, which was observed in the longitudinally arranged treatments. Burrows arranged along the bottom row of the bank recruited more sediment during the initial collapse phase than burrows on the top row. This was expected. Considering the Mohr- Coulomb equation (see section 7.5.1.1), burrows at the bottom row were hypothesised to fail more quickly, due to the greater weight (83%) of material above the burrows than the top row burrows, meaning that failure would occur earlier under stress.

Figure 7.23: Crayfish burrows clustered on (a) the River Chess and (b) Gaddesby Brook, and (c) burrows arranged longitudinally at the bottom of the bank at Gaddesby Brook.

This pattern of burrows being arranged longitudinally along a bank face has also been observed in the field (Figure 7.23c). Burrows have been hypothesised to be a defence against high flows (Lindqvist et al. 1999), with crayfish invasion success (Mathers et al. 2020), activity (Johnson et al. 2014), and burrowing propensity (Chapter 4) being highest in low flow conditions.

304

Crayfish burrows are therefore most commonly constructed at the bottom of riverbanks, with 45% of burrows surveyed in Chapter 3 being underwater in low flow (Q78) conditions, and this suggests that burrows may therefore have a greater erosive impact than previously modelled, where modelling has considered measured mammal burrows towards the top of levee banks (Taccari and van der Meij 2016a; Borgatti et al. 2017), and hypothetical burrows in levees also towards the top of banks (Saghaee et al. 2017).

7.5.3 Discussion Summary

It was hypothesised that an increasing density of simulated burrows would increase the rate of bank erosion in the Friedkin experiments, but this was not observed, and there were no significant differences in erosion rate between density treatments except that erosion was higher without any burrows at all (Figure 7.9). It is hypothesised that the lack of association between burrow density and bank erosion is due to the use of xanthan gum as a cohesion agent. Drainage and drying time affect the hardness of the sand-xanthan mixture and therefore the erosivity of the test blocks. This is problematic because simulated burrows affect drainage and drying time, so that blocks with burrows probably resulted in more stable test blocks. The importance of this effect was explored relative to other drivers of stability and instability in a conceptual model (Figure 7.18). This suggests that the lack of any significant difference in erosion rate as density increases is indicative of a burrow density effect; it is likely that without the xanthan effect, erosion would scale with burrow density, consistent with observations from the field monitoring in Chapter 6.

The problem of the strength of the xanthan gum scaling with burrow density was not apparent when comparing runs with the same burrow numbers, as those tests were designed to evaluate the impact of differences in burrow locations. Results demonstrated that burrows arranged at a single level along the bottom of the bank and the clustering of burrows into one small area, both of which are common in field settings, resulted in higher rates of erosion (both diffuse erosion and collapse) than when burrows were located at the top of the bank or randomly distributed throughout the reach (Figure 7.13; Figure 7.17).

The mechanism by which erosion occurred also changed with changing burrow density and locations. When burrows were absent, erosional processes occurred parallel with the flow, with the bank retreating as a single unit (Figure 7.20). However, in the presence of burrows, modular retreat was observed, with erosion happening in directions not parallel with the flow (Figure

305

7.12; Figure 7.22). Compounding effects of burrows were observed, with burrows interacting to promote greater erosion. The mechanisms involved require further investigation, but the implication is that burrows may recruit more sediment to river systems than previously considered under single-burrow modelling scenarios. Overall, these results support observations made in the field that burrows play an important part in determining erosional rates and processes, demonstrate that burrows are key in driving local riverbank morphodynamics, and suggest that previous work, which has typically considered single burrows further up the bank, may have underestimated the impact of burrows on erosional processes.

306

Chapter 8

Summary, key themes, future research, and concluding remarks

307

8.1 Fulfilment of thesis objectives

At the start of this thesis, five objectives were identified (Chapter 1.2). This section considers how these objectives were addressed and fulfilled.

1. To quantify the mass of sediment excavated by signal crayfish burrowing through an extensive field study of 39 rivers across Great Britain.

The geomorphic importance of animals has been widely demonstrated (Crooks 2002; Moore 2006; Rice et al. 2012; Statzner 2012; Fei et al. 2014; Albertson and Allen 2015), especially in relation to invasive species (Fei et al. 2014; Emery-Butcher et al. 2020), and burrowing animals (Harvey et al. 2011; Haussmann 2017; Harvey et al. 2019). Despite this interest, only one study has attempted to quantify the mass of sediment excavated by burrowing signal crayfish at the catchment scale (Faller et al. 2016). Understanding the quantity of sediment directly excavated by crayfish is an important first step to understanding their catchment-wide effects.

This objective was thoroughly explored at the burrow scale, the bank scale and the reach scale in Chapter 3. A total of 38 sites on 29 crayfish invaded rivers were surveyed for burrows, which were recorded at 23 sites. Reach scale, local scale, and burrow scale dimensions, densities, and sediment budgets were calculated, with burrows recruiting up to 4.14 t km-1 of fine sediment (mean = 0.93 t km-1) at the reach scale. At the local scale, burrows recruited up to 14.3 kg m-1 (mean = 3.77 kg m-1), similar to the amounts reported in previous UK studies, but in this case the data were for a sample across the UK. This was also the first study to consider the individual sizes of all burrows recorded, rather than applying a standardised or averaged figure (previously 3 kg burrow-1, Faller et al. (2016); 1.15 kg burrow-1, mean calculated in this study). Variation in burrow size, density and sediment recruitment was observed at the bank and reach scales. Understanding this variability and what drives it was the focus of Objective 2.

2. To examine the biotic and abiotic drivers of signal crayfish burrowing behaviour through a field study of 39 rivers across Great Britain complemented by laboratory flume and mesocosm experiments using crayfish populations that are invasive (in the UK) and from their native geographical range (in the USA).

The presence, extent, and distribution of crayfish burrows differs between and within catchments (Guan and Wiles 1997; Stanton 2004; Faller et al. 2016; Rice et al. 2016), but these

308 differences have only been investigated by one study in sub-catchments of the River Thames (Faller et al. 2016). Chapter 3 examined the morphological, hydrological, chemical, and biological variations between sites with signal crayfish populations where burrows were present and absent, and considered associations between potential independent variables and reach scale burrow densities. Rivers with burrows were significantly narrower with a lower discharge, higher conductivity, and had finer-grained bed and bank materials than rivers without burrows. No difference in biological variables were observed, which was surprising. This was the first study to consider burrowing as a scaled, as opposed to a binary response, and modelled to account for variation in the size, as well as prevalence, of burrows. Previous studies have also only considered burrowing at a single point in time (Guan 1994; Guan and Wiles 1997; Stanton 2004; Faller et al. 2016) or assumed burrowing to be consistent through time (Rice et al. 2016). The results from this research suggest that with changing environmental conditions, crayfish burrowing may be a dynamic process, worthy of further temporal investigation.

Hypotheses about the controls on burrowing activity were subsequently examined in mesocosm studies in Chapters 4 and 5. Chapter 4 considered the significant associations observed between flow velocity and burrowing in the field through novel flume experiments. Flow has been considered important for understanding crayfish movement (Maude and Williams 1983; Clark et al. 2008; Salkonen et al. 2010; Johnson et al. 2014), colonisation success (Mathers et al. 2020), population density (Usio and Townsend 2000), and shelter use (Parvulescu et al. 2016; Ion et al. 2020), but this was the first study to consider flow in the context of burrowing. Significant responses were observed in the mass of sediment recruited by crayfish burrowing under different flow velocities, with the number and size of burrows decreasing as flow velocity increased.

Grain size was considered to be a primary factor driving shelter availability, and was scrutinised in detail in Chapter 5. Results supported the results from Chapter 3, with interacting associations between shelter availability and population density. The presence of a shelter reduced burrowing activity, and there was no association between population density and excavated sediment. However, the presence of a shelter mediated competitive interactions between crayfish, and a positive association between population density and excavated sediment was observed. This highlighted the importance of considering both biotic and abiotic factors together to understand zoogeomorphic processes.

309

Perhaps the most important result of Chapter 5 considers the population provenance of crayfish. Associations observed in Chapter 3 are dependent on consistent responses to external variables between independent crayfish populations. Animals have the capacity to change behaviour, especially when they are outside of their native range or are invasive (Wright et al. 2010; Sol and Weis 2019), and differences in the strength of a behaviour have been observed between native and invasive populations (Magurran et al. 1992; Holway and Suarez 1999; Sol and Lefebvre 2000; Jones and DiRienzo 2018). Burrowing by signal crayfish is a novel behaviour that has not been reported within its normal endemic range. Chapter 5 considered crayfish from two UK populations, an invasive USA population, and signal crayfish from their native USA range. The burrowing response of signal crayfish to shelter availability and crayfish density was consistent between populations, with no significant differences between the invasive UK populations and USA native populations. However, the USA invasive population burrowed significantly more, which facilitated the construction of a conceptual model of behavioural change for invasive populations. This was the first time that a novel behaviour has been empirically tested between native and invasive populations, and thus adds a novel context to examining the consideration of biotic and abiotic factors in modifying animal behaviour.

3. To examine the processes by which crayfish burrowing impacts the geomorphology of riverbank erosion, through field monitoring of riverbanks on two infested UK rivers complemented by laboratory physical modelling experiments in a purpose built Friedkin flume.

Evidence that burrowing can accelerate bank erosion comes from ex situ experiments (Onda and Itakura 1994; Vu et al., 2017), and physical (Viero et al., 2013; Saghaee et al., 2017) and numerical modelling (Camici et al., 2014; Orlandini et al., 2016; Borgatti et al., 2017) of burrows in levees. Harvey et al. (2019) hypothesise that crayfish burrowing can also impact riverbanks, which is supported by anecdotal and qualitative evidence (Barbaresi et al. 2004; West 2010; Arce and Dieguez-Uribeondo 2015; Faller et al. 2016), but there were no quantitative assessments prior to this thesis. Chapter 6 quantified the association between different burrow characteristics and the method of bank retreat through in situ monitoring on two rivers. Burrow volume was found to be significant for driving both diffuse erosion and bank collapse. Measurements were collected at discrete points in time, and so mechanisms could only be inferred. Additional work on how burrows may drive bank retreat was undertaken in idealised Friedkin flume experiments in Chapter 7. Clustered burrows promoted more

310 collapses than spatially distributed burrows, and facilitated faster diffuse erosion rates, supporting the results from Chapter 6. The ex situ nature of the experiments, which ran over a matter of hours, allowed careful observations to be made about the nature of the mechanisms involved. The experiments confirmed the supposition from field evidence that burrows promoted collapses through bank undercutting, with different collapse mechanisms evident between burrowed and non-burrowed banks. Together, these chapters provided supporting evidence for Harvey et al.’s (2019) hypotheses about how animal burrows can drive bank collapse and present the first quantitative data on the associations between animal burrows and the processes of sediment recruitment from riverbanks.

4. To quantify the effect of crayfish burrowing on volumes of riverbank erosion and investigate the relative roles of direct sediment input from crayfish burrows, accelerated bank erosion caused by crayfish burrows, and bank erosion in the absence of crayfish burrows in recruiting sediment to invaded river channels, using field monitoring of two UK rivers and laboratory physical modelling and flume experiments.

Chapter 6 quantified the mass of sediment recruited at the bank and reach scale to two lowland rivers through accelerated retreat and mass failure. These were the first estimates of accelerated sediment recruitment from riverbanks, rather than direct inputs via burrow excavation. At the bank scale, total accelerated retreat recruited an average of 48.7 kg m-1 a-1, ranging up to 189.1 kg m-1 a-1 at the River Bain, and an average of 18.2 kg m-1, ranging up to 44.6 kg m-1 a-1 of sediment at Gaddesby Brook. This research represents the first study to quantify a link between signal crayfish and the delivery of a substantial mass of sediment recruitment to rivers over time. Crayfish burrows significantly affected both retreat and mass failure in space and time. The temporal variability observed was linked with river flows. This was explored through ex situ mesocosm experiments in Chapter 4 where it was shown that different flow velocities led to significant differences in the mass of sediment excavated from crayfish burrows, the mass of sediment recruited through abiotic bank erosion, and the mass of sediment recruited as a result of crayfish burrowing activity. These results support the hypothesis of a relationship between erosion due to crayfish burrowing and river discharge. They were the first studies to isolate the relative importance of direct and indirect effects of burrowing, relative to purely abiotic processes, for sediment recruitment to rivers. It is clear that this type of budgeting is crucial for scaling and understanding zoogeomorphic impacts. As river flows are predicted to become more extreme (Kay and Jones 2011; Schneider et al. 2013), the importance of

311 biological activity such as crayfish burrowing in driving fluvial sediment loading will become more pronounced with climate change.

5. To construct models to predict the presence, extent, and geomorphic impacts of burrowing on UK rivers.

Signal crayfish are continuing to invade and expand their range within Great Britain (Almeida et al. 2013), at approximately 1.6% per annum (Chadwick 2019). Being able to model and predict their potential impacts is therefore of paramount importance given their demonstrated geomorphic impact from previous literature and throughout the research presented in this thesis. However, prior to the work done here, the only means of predicting how and where burrowing might spread was via a single presence or absence model, which could only explain 21% of the observed variation in burrow presence (Faller et al. 2016). This was substantially improved upon in Chapter 3, where morphological, hydrological, chemical, and biological variables were considered to build a logistic regression model that explained 68.3% of crayfish burrow presence. Further, this model was rearranged to allow estimation of the crayfish density required to initiate burrowing and so provide a tool that can be used to identify which rivers are most likely to experience burrowing impacts. This presence/absence model was complemented by multiple linear regression models designed to predict the amount of burrowing and recruited sediment at infested sites. These models significantly explained 62% and 72% of the variance in burrow density and the mass of sediment excavated, respectively.

Whilst the models created in Chapter 3 predict the mass of sediment excavated by burrowing, Chapter 6 indicated that this was a relatively minor sediment input compared with the indirect effects of burrowing on bank retreat. Regression modelling allowed for the associations between burrows and accelerated retreat to be quantified, and at the reach scale, accelerated erosion facilitated by burrows recruited 24.9 t km-1 a-1 of sediment into the River Bain. These models were then used to examine future invasion scenarios, with burrows recruiting over 100 t km-1 a-1 of fine sediment into the River Bain under the worst-case scenario. As these regression models considered biological variables, they provide the first tools to model the potential impacts of crayfish burrowing under current and future invasion scenarios. Employing these at a landscape scale is therefore a feasible and an important piece of future research.

312

8.2 Research themes

This thesis has investigated the biotic and abiotic controls on the burrowing behaviour of signal crayfish and quantified its geomorphic impacts on river systems. To achieve an overarching understanding, both in situ and ex situ approaches have been used over a range of spatial and temporal scales, and have considered aspects of geomorphology, ecology, and hydrology independently and in combination. This section discusses the wider themes arising through the chapters, and critically assesses the frameworks and methodologies employed.

8.2.1 Biological and geophysical energy should not be considered independently.

Whilst the geomorphic potential of biota is now well recognised within both ecosystem engineering and zoogeomorphology (Cooper 2002; Moore 2006; Statzner 2012; Rice et al. 2012; Fei et al. 2014; Emery-Butcher et al. 2020), the relative and combined contributions of biological and geophysical energy are poorly quantified in geomorphology (Philips 2009; Rice et al. 2016), and biotic energy is still largely disregarded from conceptual and numerical sediment transport models (Rice et al. 2019). The few conceptual models that do consider the importance of biota in fine sediment recruitment have considered biological and geophysical energy as independent (Figure 8.1). Moore (2006; Figure 8.1a) only considers the abiotic context as a modulator in determining the importance of work undertaken by organisms in comparison to their geophysical context; zoogeomorphic activity in high velocity flows may be deemed less important than in low flows due to the greater available geophysical energy. Mathers et al. (2019a; Figure 8.1b) built on this, and illustrated the cyclic importance of the physical environment in altering biotic interactions, and thus influencing zoogeomorphic forces, but conceptualised these zoogeomorphic effects as independent of external factors. Mathers (2017; Figure 8.1c) considered the geomorphic potential of a taxon, and suggested additive factors including physical settings and environmental controls.

313

Figure 8.1: Conceptual models of the interaction between biotic and abiotic forced in driving fluvial sediments from (a) Moore (2006), (b) Mathers et al. (2019a); (c) Mathers (2017), and (d) Rice et al. (2012).

Considering gravel beds, and the resultant recruitment of fine sediment, Rice et al. (2012; Figure 8.1d) considered biotic energy as priming sediment for transport by geophysical flows, but this is still expressed as an ‘in-series’ relationship, as opposed to ‘in parallel’. Through in situ monitoring, Chapter 6 quantified the mass of sediment recruited through accelerated retreat as a result of crayfish burrows, and thus was the first in situ sediment budget to calculate biological and geophysical sediment recruitment. Further, this research represents the first study to consider an interactive effect between biological and geophysical energy, and fully quantified its relative importance. These indirect biotic effects contributed a substantial mass of sediment to river systems, contributing an estimated 29.8% and 89.6% of sediment at the bank scale at the River Bain and Gaddesby Brook respectively, and 12.2% at the reach scale at the River Bain. This demonstrated the role that signal crayfish have in recruiting fine sediment to lowland river systems, but more generally showed the important role that biology has in

314 priming riverbanks for sediment erosion through altering their hydraulic properties and promoting geotechnical failure.

Further, results from Chapter 4 also illustrated the importance of the interactive effects of both biological and geophysical energy operating in parallel, together: flow velocity determined crayfish burrowing activity, which actively recruited sediment, which was transported at rates dependent on the initial flow. The compounding of these interactions was strongest in the medium flow velocity, with biological energy dominating low flow velocities, and geophysical energy dominating processes at high flow velocities. This led to the development of a numerically informed conceptual model to understand the dynamic nature of the interactions between geophysical and biological energy in sediment recruitment, but a more general framework is presented here (Figure 8.2), which for the first time conceptually considers the interactive processes of biotic and geophysical energy in determining sediment dynamics.

Figure 8.2: Conceptual framework for the importance of biotic energy in determining total geomorphic outputs, considering (a) the previously considered relative importance of biotic and geophysical energy, and (b) the proposed framework for the relative importance of biotic and geophysical energy.

315

8.2.2 The value of field data

This thesis has been based principally on primary data, with little use of remote sensing, meta- analysis, or big data. In a world of increased computing power, these techniques are an important means of understanding processes on a global scale in both ecology and geomorphology (Gomez et al. 2015; Knox et al. 2016; Brum et al. 2017; Piegay 2019). Research using secondary data sources is increasingly common (Carmel et al. 2013; Panuccio 2018), with an increase of 600% and 800% of modelling and big data studies, respectively, in the field of conservation since the 1980s (Rois-Saldana et al. 2018). However, this same period has also observed a 20% decrease in field-based studies, with a negative trend between journal rankings and the proportion of field studies published (Rios-Saldana et al. 2018). This is of particular concern given the necessity of field studies to inform modelling exercises (Church 2013; Price et al. 2018). As modelling becomes a more popular field of research, and the complexity of modelling increases, an increased quantity, depth and precision of fieldwork is required to complement that effort. Indeed, there has been a persistent call from modellers for an increase in global field data collection, with some speculating that the limit of modelling has been reached considering the currently available data (Rios-Saldana et al. 2018). Considering zoogeomorphology, Larson et al. (2018) suggest that meta barcoding has the potential to unveil the historical importance of biogeomorphic influences on landscape development. Zoogeomorphology is understudied (Philips 2009; Rice et al. 2016; Rice et al. 2019), and understanding modern day analogue systems is key in providing data to inform the large-scale modelling based research such as proposed by Larson et al. (2018).

Field observations and understanding behavioural processes and mechanisms are also important for modelling future landscapes. A highlighted future research avenue in this thesis is modelling sediment input to river systems by crayfish burrowing at the landscape scale. Crayfish burrowing has been documented in the UK (e.g. Guan 1994; Harvey et al. 2014; Faller et al. 2016), but has largely considered burrowing as a lowland river processes. Chapter 3 recorded burrows in upland catchments, and expanded the previously recognised range and distribution of crayfish burrowing behaviour. These upland environments may have been previously excluded in hypothetical modelling exercises considering currently existing data, and so this demonstrates the value of fieldwork of informing as well as validating modelling studies. In addition to these arguments that field data are needed to maximise the potential of modelling studies, field studies have value in the own right, uncovering important and unique

316 observations and mechanisms that would not be possible using desk-based approaches. For example, the remote sensing of riverbanks using accurate aerial imagery could be used to understand rates of riverbank retreat (cf. the processes and methods used in Chapter 7). However, these would not allow for the mechanism of burrows promoting undercutting to be observed, which was discovered through repeated field visits in Chapter 6.

8.2.3 The use of the mesocosm

Two of the five chapters presented here utilised mesocosm experiments to understand the processes observed in in situ. Whilst mesocosm studies are frequently used in the field of ecology, including for work with crustaceans (e.g. Childress and Herrnkind 1996; Rice et al. 2012; Harvey et al. 2014; Hughes et al 2014; Houghton et al. 2017; Toscano 2017; Ion et al. 2020; Wei et al. 2020), their use to study animal behaviour has been criticised on many grounds, in particular:

(i) mesocosms do not replicate natural settings (e.g. Underwood 1986; Skelly and Kiesecker 2001; Skelly 2002); (ii) animals do not behave ‘naturally’ in mesocosm settings (e.g. Skelly 2002; Kemp et al. 2009); and (iii) upscaling behaviours observed in mesocosms are not necessarily applicable to natural settings

Even though prominent critics have also recognised the value of mesocosm experiments (e.g. Skelly and Kiesecker 2001), these issues are worth discussing in the context of the work completed here.

8.2.3.1 Mesocosms do not replicate natural settings

Mesocosm studies have often been criticised for not replicating a natural setting (e.g. Underwood 1986; Skelly and Kiesecker 2001; Skelly 2002), but mesocosms do not attempt to replicate natural settings, and indeed this is one of the major advantages of ex situ experiments. Natural settings are complicated with a vast array of variables, which are inconsistent both spatially and temporally.

317

‘‘It is inconceivable that only the desired, experimentally manipulated variable of interest will actually differ between any two pieces of the world’

(Underwood 1986: 249)

Mesocosm experiments in an unnatural setting allow for variables to be purposefully selected and removed from complex environments so that direct associations can be understood in isolation without the influence of other factors. Variables can be controlled, meaning that a considerably larger number of treatment levels and replicates can be undertaken than would be possible in a field setting (Kareiva and Anderson 1988; Skelly and Kiesecker 2001).

In the context of this research, experiments were undertaken in mesocosms in order to understand the mechanisms behind the burrowing behaviour of signal crayfish. Chapter 3 identified relationships between crayfish burrowing with flow velocity, sediment size, and population density. However, a large proportion of variables in river systems, including those considered here, express collinearity, and considering them independently is therefore a necessary process to understand their relative and combined importance. Further, field settings do not allow for the processes to be understood. For example, Chapter 3 considered crayfish burrow densities that were present, but the exact processes could only be statistically inferred through modelling using observed data from one snapshot in time. Mesocosm experiments, such as in Chapters 4 and 5 allowed for a temporal resolution with variable manipulation that would be unfeasible in situ.

8.2.3.2 Animals do not behave ‘naturally’ in mesocosm settings

The use of mesocosms for behavioural experiments is dependent on fauna expressing behaviours to the same response cues in an ex situ setting in the same manner that they would in situ. Upscaling the ecological interactions observed in mesocosms between algae may be acceptable (Spivak et al 2011), but ecological interactions between larger fauna in mesocosms have been shown to be demonstrably different to patterns witnessed in the field (Hogan et al. 1988; Veasey et al. 1996; Skelly 2002; Mallapur et al. 2009; Valuska and Mench 2013; Breton and Barrot 2014). The influence of walls is a particular driver of non-natural behaviours in mesocosm experiments (Kemp et al. 2009), and indeed non-natural behaviours were observed in Chapter 4, where crayfish spent time climbing mesocosm walls, which was consistent with previous crayfish mesocosm observations (Rice et al. 2012).

318

‘Scale sensitive’ experimental designs are therefore called upon for mesocosm experiments relating to algae and micro-fauna (Petersen et al. 1999), but these principles are equally valid here. Experiments attempted to be scale-sensitive in terms of time, size, and population density. In Chapter 5, experiments were conducted over 96 hours, as this was the longest time taken for signal crayfish to completely construct a burrow in previous experiments (Stanton 2004), and population densities used also attempted to simulate densities found in wild populations, with the high density treatment corresponding to the highest recorded densities of adult signal crayfish in UK populations (Bubb et al. 2004). Mesocosm size was also considered, with the largest possible mesocosms being used that the laboratory space would allow. The mesocosm length (excluding the single bank) was 0.33 m, representing a stream 0.66 m wide (where two riverbanks are present); smaller stream widths were recorded in six of the 39 rivers sampled in the field survey (Chapter 3), suggesting that the ratio of substrate to riverbank did not promote burrowing behaviour beyond what is found under natural conditions. A limitation of the mesocosm design is river length; rivers would still allow for longitudinal movement that a mesocosm cannot. Due to scaling limitations, assessment of animal behaviour and study results with in situ observations is therefore necessary to validate observations in the mesocosm setting, which was undertaken to ratify the observations between in situ and ex situ studies (Chapters 3, 4, and 5).

In the mesocosm experiments undertaken here, a full spectrum of burrowing was witnessed in the experiments, from not burrowing at all, to multiple burrows of different sizes. The artificial environment therefore did not influence behaviour to a point where it masked the effects of the behaviour in relation to the treatments employed. Further, these experiments did not seek to measure burrows as absolute values, but provide a quantitative mechanism of measuring the response of the crayfish behaviour to external cues. Crayfish behaviour was not limited or exaggerated by the mesocosms. Constructed burrows were of a comparable size and architecture to those observed in the field and the results from mesocosm studies were consistent with field observations; for example, flow velocity was significantly associated with both crayfish burrow presence and size in both Chapters 3 (in situ) and 4 (ex situ).

8.2.3.3 Upscaling behaviours observed in mesocosms are not necessarily applicable to natural settings

There is also debate regarding the applicability of mesocosms and upscaling the results to represent system-wide processes (Carpenter 1996; Mowitt et al. 2006; Houde and Petersen

319

2009). Mesocosms are powerful tools for understanding organism responses to isolated variables. However, associations observed in ex situ settings may be absent, or at very least dampened, when applied to complex natural settings (Carpenter 1996). Extrapolating results from these experiments into rivers is therefore difficult due to the purposeful discrepancy of conditions between the mesocosm set up and the conditions found in rivers. Whilst the processes observed in the mesocosm experiments support the observations from in situ non- manipulated settings, the quantification of these processes requires further work.

Future work could include repeating the experimental design of Chapter 5 in enclosures in a river setting. This was decided against for this project due to logistical challenges (e.g. removing vegetation caught on cages), statistical challenges (e.g. undertaking enough replicates), replicability challenges (e.g. creating identical conditions in Montana where river banks are of different properties), and ethical and legal challenges (e.g. ‘housing’ invasive species in a natural setting). However, enclosure experiments could be a strong step in providing intermediary evidence between the experimental and observational work undertaken here.

8.2.3.4 Mesocosms and Invasive Species

The mesocosm setting is also a necessary setting for undertaking invasive species research. Globally important issues requiring foresight, such as climate change and invasive species, are not amenable to field study, and so replica settings need to be created to allow investigations (Benton et al. 2007). Mesocosm studies have therefore been successfully used to understand the consequences of phenomena not yet observed in the field, or phenomena that would have negative impacts on the ecology and functioning of a system, such as simulating drought (Ledger et al. 2011, Lancaster and Ledger 2015), fine sediment deposition (Wagenhoff et al. 2012) and metal contamination (Richardson and Kiffney 2009).

In the field of invasion, previous studies have often transplanted the invasive study organism, such as Anolis lizards (Anolis sagrei) in the Bahamas (Losos et al. 1997), the guppy (Poecilia reticulate) in Trinidad (Carvalho et al. 1996), and freshwater shrimp (Gammarus pulex) onto the Isle of Man (Dick et al. 1997). However, this is no longer ethically nor legally acceptable, and so mesocosms, if designed and implemented correctly, provide a suitable, biosecure arena for conducting experiments. Comparing native and invasive populations in situ may also not be possible due to the differing biotic and abiotic variables within the studied system.

320

Mesocosms provide a tool for standardising external cues to compare geographically distinct populations. Therefore, experiments investigating the behaviour of invasive species have been typically been undertaken in mesocosm or laboratory conditions, which ensures that the spread of the invasive organism in question is not unnecessarily supported (Stewart et al. 2013).

8.2.3.5 The Appropriateness of the Mesocosm

Whilst the applicability and difficulties of upscaling mesocosm experiments have been questioned, it has also been argued that a range of experimental approaches are necessary to understand the functioning of a system (Drenner and Mazumder 1999). Here, the combination of experimental mesocosms and field studies allowed for a much deeper understanding than either approach alone.

8.3 Future research directions

This thesis has made substantive contributions to the current understanding of crayfish burrowing and its geomorphological impacts on riverbank processes. It has also highlighted a number of areas where this work needs to be extended in the future, which are detailed in this section.

8.3.1 Quantifying burrows

The methods used here to calculate burrow size were very simple, and assumed burrows to be single chambered elliptical cylinders, as in Faller et al. (2016). Whilst the majority (>90%) of burrows are single chambered (Guan 1994; Stanton 2004), there is currently no research considering methods for estimating crayfish burrow volumes, and so whilst the methods employed here are precise, they may not be accurate. Quantifying burrow dimensions has previously been undertaken using structure-from-motion (Sofia et al. 2016), ground penetrating radar (Chlaib et al. 2014), and physical casting (Hasiotis et al. 1999; Talley et al. 2001; Rudnick et al. 2005; Wei et al. 2020). Burrow casting was attempted in Chapter 5 (Figure 8.3), but was not pursued due to the time pressures of experiment completion. However, preliminary trials proved successful, and crayfish burrow casting may be a useful way to pursue this for future research to better allow for accurate burrow measurements and thus improved sediment recruitment estimates.

321

Figure 8.3: A successful preliminary trial at casting burrows in mesocosms.

8.3.2 Temporal persistence of burrows

The principal aim of this thesis was to assess the mass of sediment that crayfish recruit to river systems. Whilst the research presented here assesses many aspects of this question, further work is required. Chapter 3 surveyed the mass of sediment excavated directly by burrowing across the UK, and Chapter 6 modelled the accelerated retreat of riverbanks and related sediment inputs. However, the models produced in the two chapters have not been used in conjunction to assess the impacts of accelerated retreat beyond the studied catchments because of the large variability in bank retreat rates observed between the two tested rivers in Chapter 6. Whilst Chapter 6 considered sediment recruitment over time, Chapter 3 only considered one snapshot in time, and gives no temporal resolution for this primary sediment input, and so understanding the temporal dynamics of crayfish burrows is a vital avenue for future research to contextualise the results presented here. Research is currently ongoing to answer this question. Over 1,500 individual crayfish burrows on five of the rivers considered in Chapter 3 have been monitored for four years to understand their temporal dynamics. ‘Burrow half-life’ is circa 500 days, and the annual mass of sediment recruited is, on average, 56% of the total observed mass from all burrows that are present at a moment in time (Table 8.1).

322

Mass of Sediment Recruited from Burrows per Year 2014-2017 2017-2018 Mean

M A %M A M A %M A M A %M A (t km -1 a -1 ) (% a -1 ) (t km -1 a -1 ) (% a -1 ) (t km -1 a -1 ) (% a -1 )

Eye Brook Upper 0.29 67.4 0.23 43.1 0.28 61.3 Eye Brook Lower 0.07 58.1 0.40 89.9 0.15 66.1 Gaddesby Brook Upper 1.08 69.7 0.58 52.4 0.96 65.4 Gaddesby Brook Lower 0.38 46.0 0.62 54.6 0.44 48.2 Gwash Uppermost 0.14 94.1 0.23 69.3 0.16 87.9 Gwash Upper 0.03 20.1 0.34 74.2 0.11 33.6 Gwash Lowest 0.40 63.4 0.22 55.0 0.36 61.3 Gwash Lower 0.28 49.1 0.21 36.3 0.26 45.9 Nene 0.07 40.4 0.60 68.2 0.20 47.4 Welland 0.24 44.3 0.07 19.9 0.20 38.2

Mean 0.30 55.3 0.35 56.3 0.31 55.5

Table 8.1: Sediment recruited directly through burrowing over time at the ten studied sites.

These preliminary figures of sediment excavated over time, combined with the models of accelerated bank retreat in the presence of burrows presented in Chapter 6 can be crudely applied to the nationwide burrow surveys undertaken in Chapter 3, to roughly estimate the mass of sediment annually recruited to river channels (Table 8.2). ‘Primary’ sediment recruitment rates can be calculated as a function of observed burrows surveyed in Chapter 3 and annual sediment recruitment rates calculated in this temporal study. ‘Secondary’ sediment recruitment rates can be modelled using the calculations presented in Chapter 6 in conjunction with the calculated ‘primary’ recruitment rates. The accuracy of this preliminary modelling cannot be determined without testing the models using independent data, but the independent variables used are likely to yield good predictions due to the strength of the informing data. Whilst untested, the models can be used to give a gross indication of how crayfish burrowing is affecting, and could affect, UK river systems.

323

Annual Sediment Excavated (t km-1 a-1) Primary Secondary (Burrows) (Accelerated Retreat) Bentley Brook 0.17 4.96

Bookill Gill Beck (Upper) 0.21 6.28

Bookill Gill Beck (Middle) 0.02 0.66

Bookill Gill Beck (Lower) 0.10 2.97 Broadmead Brook 0.85 25.11

Chess 1.89 55.84

Churn 0.17 4.96

Clyde (Allershaw Islands) 0.16 4.63

Clyde (Elvanfoot) 0.22 6.61

Allershaw Burn 0.06 1.65

Eye Brook 0.55 16.19

Gaddesby Brook (Gaddesby) 1.23 36.34

Gaddesby Brook (Twyford) 2.32 68.39

Greet 0.64 18.83

Nene 0.57 16.85

Ouzel (Upper) 0.56 16.52

Ouzel (Lower) 0.73 21.48

Potwell Dyke 0.35 10.24

Thames 0.07 1.98

Torkington (Upper) 0.06 1.65

Torkington (Lower) 0.01 0.17

Tove 0.82 24.12

Welland 0.21 6.28

Mean 0.52 15.33

Median 0.22 6.61

Standard Deviation 0.60 17.67

Minimum 0.01 0.17 Maximum 2.32 68.39

Table 8.2: Modelled mass of sediment excavated to all rivers considered in Chapter 3. ‘Primary’ rates are annual sediment recruitment rates calculated from observed burrow masses multiplied by 56%, in line with preliminary results from the described temporal study. ‘Secondary’ rates are annual sediment recruitment rates of accelerate bank retreat, calculated as primary rates multiplied by 29.5, the mean factor of increase from burrowed to accelerated retreat observed in Chapter 6.

324

8.3.3 Burrows and hydrology

Burrowing has previously been considered as a response to biotic (Guan 1994) and morphological (Faller et al. 2016) conditions, but the importance of hydrology in determining crayfish burrowing behaviour was demonstrated in Chapters 3 and 4. Flow velocity and river flows have been considered as influential considering crayfish shelter use (Parvulescu et al. 2016), but this is the first research considering flows and burrowing behaviour. Given that river discharge regimes are predicted to become more extreme (Kay and Jones 2011; Schneider et al. 2013), and the significant associations observed here between flow velocity and crayfish burrow presence (negative; Chapter 3) and sediment recruited through crayfish burrowing (positive; Chapter 4), understanding these associations and quantifying the conceptual models presented is a key area of future research to understand the combined roles of hydrology and biology in recruiting sediment to fluvial systems.

8.3.4 Modelling sediment

This thesis has presented numerical and conceptual models that consider the stage of invasion at which burrowing will be initiated in a given river (Chapter 3; Chapter 5), and models predicting the likelihood and extent of burrowing when burrowing occurs (Chapter 3). These can be combined to model the presence and extent of burrowing under current and future invasion scenarios. Large scale geospatial modelling of crayfish-excavated and facilitated sediment is therefore an important future avenue of research that can help to understand the role of crayfish in contributing to sediment dynamics at larger, landscape scales now and in the future, and to deploy appropriately future management techniques. This could be undertaken through GIS techniques using big data, such as River Habitat Survey data, to estimate the key independent variables used in the models presented here under difference invasion scenarios.

325

8.4 Concluding remarks

Animals are an important component in the processes of sediment transport. This thesis has investigated the novel burrowing behaviour of the invasive signal crayfish (Pacifastacus leniusculus), using a combination of in situ and ex situ approaches to understand burrowing at the bank, reach, and landscape scale. This thesis has identified, quantified, and modelled the specific drivers of crayfish burrowing, and proposed models for the prediction of sediment recruitment to rivers via direct and indirect effects. The relevance of the species invasive nature has been emphasised and considered in detail through mesocosm experiments that tested how burrowing behaviour varies between native and different invasive populations. The work reported here has utilised ideas and techniques from zoogeomorphology, ecosystem engineering, and hydroecology, and highlights the need to consider disciplines in parallel as opposed to independently to fully understand riverine processes.

However, this thesis has highlighted important questions that remain unanswered considering the geomorphic impacts of signal crayfish, and also invasive species and zoogeomorphology more widely. Scaling up the research presented in this thesis from laboratory or local scales to the landscape scale will allow for the true impact of animals, and biotic energy more generally, in driving earth system processes. Partitioning quantitative local sediment budgets into biotic, abiotic and interactive elements (as in Chapters 4 and 6) and conducting distributed work across multiple geographical locations (as in Chapters 3, 5 and 6) are important initial steps, and it is hoped and encouraged that this work will be built upon to increase the recognition of the role of animals in earth surface processes. It is hoped that future research will build upon the concepts, methodologies, and results presented here to continue to enhance our understanding of the role of animals, and in particular invasive species, in determining the ecology and geomorphology of the world’s threatened freshwater environments.

326

References

Aalto, R., Maurice-Bourgoin, L., Dunne, T., Montgomery, D.R., Nittrouer, C.A. and Guyot, J- L. (2003) Episodic sediment accumulation on Amazonian flood plains influenced by El Nino/Southern Oscillation. Nature, 425, 493-497.

Abernethy, B. and Rutherfurd, I.D. (2000) The effect of riparian tree roots on the mass-stability of riverbanks. Earth Surface Processes and Landforms, 25(9), 921-937.

Abrahams, M. (2012) This Is Improbable: Cheese String Theory, Magnetic Chickens, and other WTF Research. Oneworld: Croydon.

Acquistapace, P., Hazlett, B.A. and Gherardi, F. (2003) Unsuccessful predation and learning of predator cues by crayfish. Journal of Crustacean Biology, 23(2), 364-370.

Albertson, L.K. and Allen, D.C. (2015) Meta-analysis: abundance, behaviour, and hydraulic energy shape biotic effects on sediment transport in streams. Ecology, 96(5), 1329-1339.

Albertson, L.K. and Daniels, M.D. (2018) Crayfish ecosystem engineering effects on riverbed disturbance and topography are mediated by size and behaviour. Freshwater Science, 37(4), 836-844.

Albertson, L.K., Cardinale, B.J. and Sklar, L.S. (2014a) Non-additive increases in sediment stability are generated by macroinvertebrate species interactions in laboratory streams. PLoS ONE, 9(8), e103417.

Albertson, L.K., Sklar, L.S., Pontau, P., Dow, M. and Cardinale, B.J. (2014b) A mechanistic model linking insect (Hydropsychidae) silk nets to incipient sediment motion in gravel-bedded streams. Journal of Geophysical Research: Earth Processes, 119, 1833-1852.

Allen, D.C., Wynn-Thompson, T.M., Kopp, D.A. and Cardinale, B.J. (2018) Riparian plant biodiversity reduces stream channel migration rates in three rivers in Michigan, U.S.A. Ecohydrology, 11(4), e1972.

Almeida, D., Ellis, A., England, J. and Copp, G.H. (2014) Time-series analysis of native and non-native crayfish dynamics in the Thames River Bains (south-eastern England). Aquatic Conservation: Marine and Freshwater Ecosystems, 24(2), 192-202.

327

Angeler, D.G., Sanchez-Carrillo, S., Alvarez-Cobelas, M., Cirujano, S. and Medina, L. (2004) Exotic crayfish activity and its effects on water quality: preliminary implications for the alternative stable equilibria in Mediterranean wetlands. Journal of Mediterranean Ecology, 4(1), 13-21.

Anthony, K.R.N. and Fabricius, K.E. (2000) Shifting roles of heterotrophy and autotrophy in coral energetics under varying turbidity. Journal of Experimental Marine Biology and Ecology, 252(2), 221-253.

Arce, J.A. and Dieguez-Uribeando, J. (2015) Strucutral damage caused by the invasive crayfish Procambarus clarkii (Girard, 1852) in rice fields of the Iberian Peninsula: a study case. Fundamentals of Applied Limnology, 186(3), 259-269.

Argent, D.G. and Flebbe, P.A. (1999) Fine sediment effects on brook trout eggs in laboratory streams. Fisheries Research, 39(3), 253-262.

Arulanandan, K., Gillogley, E. and Tully, R. (1980) Development of a quantitative method to predict critical shear stress and rate of erosion of natural undisturbed cohesive soils. Technical Report, Waterways Experiment Station (U.S.), 80(5), 1-99.

Babcock, L.E., Miller, M.F., Isbell, J.L., Collinson, J.W. and Hasiotis, S.T. (1998) Paleozoic- Mesozoic crayfish from Antarctica: earliest evidence of freshwater decapod crustaceans. Geology, 26(6), 539-542.

Bai, S. and Lung, W.S. (2005) Modelling sediment impact on the transport of fecal bacteria. Water Research, 39(20), 5232-5240.

Barbaresi, S., Tricarico, E. and Gherardi, F. (2004) Factors inducing the intense burrowing activity by the red swamp crayfish, Procambarus clarkii, an invasive species. Naturwissenschaften, 91, 342-435.

Barber-Meyer, S.M., Mech, L.D. and White, P.J. (2008) Elk calf survival and mortality following wolf restoration to Yellowstone National Park. Wildlife Monographs, 169(1), 1-30.

Barko, J.W. and Smart, R.M. (1980) Mobilization of sediment phosphorus by submersed freshwater macrophytes. Freshwater Biology, 10(3), 229-238.

328

Barko, J.W., Gunnison, D. and Carpenter, S.R. (1991) Sediment interactions with submersed macrophyte growth and community dynamics. Aquatic Botany, 41(1-3), 41-65.

Barrett, S.C.H. and Richardson, B. (1986) Genetic attributes of invading species. In R. Groves and J. Burdon (eds.) Ecology of Biological Invasions: An Australian Perspective. Australian Academy of Sciences: Canberra. pp.21-33.

Basic, T., Britton, J.R., Jackson, M.C., Reading, P. and Grey, J. (2015) Angling baits and invasive crayfish as important trophic subsidies for a large cyprinid fish. Aquatic Sciences, 77(1), 153-160.

Batalla, R.J. (2003) Sediment deficit in rivers caused by dams and instream gravel mining. A review with examples from NE Spain. Cuaternario y Geomorfologia, 17, 79-91.

Bauer, R.T. (1998) Gill-cleaning mechanisms of the crayfish Procambarus clarkii (Astacidea: cambaridae): experimental testing of setobranch function. Invertebrate Biology, 117, 129-143.

Baynes, E.R.C., can de lageweg, W.I., McLellenad, S.J., Parsons, D.R., Aberle, J.A., Dijkstra, J., Henry, P-Y., Rice, S.P., Thom, M. and Moulin, F. (2018c) Beyond equilibrium: Re- evaulating physical modelling of fluvial systems to represent climate changes. Earth-Science Reviews, 181, 82-97.

Baynes, E.R.C., Lague, D. and Kermarrec, J.-J. (2018a) Supercritical river terraces generated by hydraulic and geomorphic interactions. Geology, 46(6), 499-502.

Baynes, E.R.C., Lague, D., Attal, M., Gangloff, A., Kirstein, L.A. and Dugmore, A.J. (2018b) River self-organisation inhibits discharge control on waterfall migration. Scientific Reports, 8, 2444.

BBC (2018) BBC News web-site. Available at: https://www.bbc.co.uk/news/uk-england- somerset-44314534. Access date: 12th April 2020.

Beirne, P., O’Donnell, I. and Janssen, J. (2018) Murdering Animals: Writings on Theriocide, Homicide, and Nonspeciesist Criminology. Palgrave: London.

329

Belchier, M., Edsman, L., Sheehy, M.R. and Shelton, P.M. (1998) Estimating age and growth in long‐ lived temperate freshwater crayfish using lipofuscin. Freshwater Biology, 39(3), 439- 446.

Bendoni, M., Georgiou, I.Y., Roelvink, D. and Oumeraci, H. (2019) Numerical modelling of the erosion of marsh boundaries due to wave impact. Coastal Engineering, 152, 103514.

Benfield, M. and Minello, T. (1996) Relative effects of turbidity and light intensity on reactive distance and feeding on an estuarine fish. Environmental Biology of Fishes, 46(2), 211-216.

Benton, T.G., Solan, M., Travis, J.M.R. and Sait, S.M. (2007) Microcosm experiments can inform global ecological problems. Trends in Ecology and Evolution, 22(10), 516-521.

Bergman, D.A. and Moore, P.A. (2003). Field observations of intraspecific agonistic behavior of two crayfish species, Orconectes rusticus and Orconectes virilis, in different habitats. The Biological Bulletin, 205(1), 26-35.

Berke, S.K. (2012) Functional groups of ecosystem engineers: A proposed classification with comments on current issues. Integrative and Comparative Biology, 50, 147-157.

Berrill, M. and Chenoweth, B. (1982) The burrowing ability of nonburrowing crayfish. The American Midland Naturalist Journal, 108, 199-201.

Bertoldi, W., Gurnell, A., Surian, N., Tockner, K., Zanomi, L., Ziliani, L. and Zolezzi, G. (2009) Understanding reference processes: linkages between river flows, sediment dynamics and vegetated landforms along the Tagliamento River, Italy. River Research and Applications, 25(5), 501-516.

Bertoldi, W., Welber, M., Gurnell, A.M., Mao, L., Comiti, F. and Tal, M. (2015) Physical modelling of the combined effect of vegetation and wood on river morphology. Geomorphology, 246(1), 178-187.

Beschta, R. and Ripple, W. (2006) River channel dynamics following extirpation of wolves in northwestern Yellowstone National Park, USA. Earth Surface Processes and Landforms, 31(12), 1525-1539.

330

Beschta, R.L. and Ripple, W.J. (2012) The role of large predators in maintaining riparian plant communities and river morphology. Geomorphology, 157-158, 88-98.

Besser, J.M., Brumbaugh, W.G., May, W.T., and Schmitt, C.J. (2007) Biomonitoring of Lead, Zinc and Cadmium in streams draining lead-mining and non-mining areas, Southeast Missouri, USA. Environmental Monitoring and Assessment, 129(1-3), 227-241.

Bhargava, D.S. and Mariam, D.W. (1991) Light penetration depth, turbidity and reflectance related relationship and models. ISPRS Journal of Photogrammetry and Remote Sensing, 46(4), 217-230.

Biggs, B.J.F., Goring, D.G. and Nikora, V.I. (1998) Subsidy and stress responses of stream periphyton to gradients in water velocity as a function of community growth form. Journal of Phycology, 34, 598-607.

Biggs, B.J.F., Duncan, M.J., Francoeur, S.N. and Meyer, W.D. (1997) Physical characterisation of micro-form bed cluster refugia in 12 New Zealand streams. New Zealand Journal of Marine and Freshwater Research, 31, 413–422.

Bilotta, G.S. and Brazier, R.E. (2008) Understanding the influence of suspended solids on water quality and aquatic biota. Water Research, 42, 2849-2861.

Bird, S., Hogan, D. and Schwab, J. (2010) Photogrammetric monitoring of small streams under a riparian forest canopy. Earth Surface Processes and Landforms, 35(8), 952-970.

Blake, M.A. and Hart, P.J.B. (1993) The behavioural responses of juvenile signal crayfish Pacifastacus leniusculus to stimuli from perch and eels. Freshwater Biology, 29, 89-97.

Blake, M.A. and Hart, P.J.B. (1995) The vulnerability of juvenile signal crayfish to perch and eel predation. Freshwater Biology, 33, 233-244.

Blashford-Snell, J. (1994) Something Lost Behind the Ranges. Harper Collins: London.

Blettler, M.C.M., Amsler, M.L., de Drago, I.E., Espinola, L.A., Eberle, E., Paira, A., Best, J.L., Parsons, D.R. and Drago, E.E. (2015) The impact of significant input of fine sediment on benthic fauna at the tributary junctions: a case study of the Bermejo-Paraguay River confluence, Argentina. Ecohydrology, 8, 340-352.

331

Blossey, B. (1999) Before, during, and after: The need for long-term monitoring in invasive plant species management. Biological Invasions, 1(2-3), 301-311.

Blott, S.J. and Pye, K. (2001) GRADISTAT: a grain size distribution and statistics package for the analysis of unconsolidated sediments. Earth Surface Processes and Landforms, 26, 1237- 1248.

Bodewes, B., van de Lageweg, W., McLelland, S. and Parsons, D. (2017) Modelling channel stability in experimental river systems. British Society for Geomorphology Annual Meeting 2017, Poster.

Boese, A. (2008) Elephants on Acid, and Other Bizarre Experiments. Pan Macmillan: London.

Borgatti, L., Forte, E., Mocnik, A., Zambrini, R., Cervi, F., Martinucci, D., Pellegrini, F., Pillon, S., Prizzon, A. and Zamariolo, A. (2017) Detection and characterization of animal burrows within river embankments by means of coupled remote sensing and geophysical techniques: Lessons from River Panaro (northern Italy). Engineering Geology, 226, 277-289.

Bovbjerg, R.V. (1953) Dominance order in the crayfish Orconectes virilis (Hagen). Physiological Zoology, 26(2), 173-178.

Box, J.B. and Mossa, J. (1999) Sediment, land use, and freshwater mussels: prospects and problems. Freshwater Science, 18(1), 99-117.

Boyle, J. (2001) Inorganic geochemical methods in Paleolimnology. In W. Last and J. Smol (eds.) Tracking environmental change using lake sediments. Vol. 2, Physical and geochemical methods. The Netherlands: Kluwer. Pp. 83-141.

Breinholt, J., Moler, P. and Crandall, K. (2007) Population structure of two crayfish with diverse physiological requirements. In C. Held, S. Koenemann and C. Schubart (Eds.) Phylogeography and Population Genetics in Crustacea. CRC Press: New York.

Breton, G. and Barrot, S. (2014) Influence of enclosure size on the distances covered and paced by captive tigers (Panthera tigris). Applied animal behaviour science, 154, 66-75.

332

Breukelaar, A.W., Lammens, E.H.R.R., Breteler, J.G.P. and Tatrai, I. (1994) Effects of benthivorous bream (Abramis brama) and carp (Cyprinus carpio) on sediment resuspension and concentrations of nutrients and chlorophyll a. Freshwater Biology, 32(1), 113-121.

Bridges, E.M. and Harding, D.M. (1971) Micro-erosion processes and factors affecting slope development in the lower Swansea Valley. Institute of British Geographers Special Publication, 3, 65–79.

Britannica (2020a) Encyclopaedia Britannica: Elk. Available at: https://www.britannica.com/animal/placental-mammal . Access date: 13th May 2020.

Britannica (2020b) Encyclopaedia Britannica: Gray Wolf. Available at: https://www.britannica.com/animal/gray-wolf . Access date: 13th May 2020.

British Geological Survey (2020) British Geological Survey web-site. Available at: www.bgs.ac.uk . Access date: 1st February 2020.

Britton, J.R., Berry, M., Sewell, S., Lees, C. and Reading, P. (2017) Importance of small fishes and invasive crayfish in otter Lutra lutra diet in an English chalk stream. Knowledge and Management of Aquatic Ecosystems, 418, 13.

Brookes, A. (1995) River channel restoration: theory and practice. In A.M. Gurnell and G.E. Petts (Eds.) Changing River Channels. John Wiley and Son: Chichester. pp. 369-388.

Brown, A.V. and Brussock, P.P. (1991) Comparisons of benthic invertebrates between riffles and pools. Hydrobiologia, 220, 99-108.

Brown, D.J. and Brewis, J.M. (1978) A critical look at trapping as a method of sampling a population of Austropotamobius pallipes (Lereboullet) in a mark and recapture study. Freshwater Crayfish, 4, 159-164.

Brum, F.T., Graham, C.H., Costa, G.C., Hedges, S.B., Penone,C., Radeloff, V.C., Rondinini, C., Loyola, R. and Davidson, A. (2017). Global priorities for conservation across multiple dimensions of mammalian diversity. Proceesings of the National Academy of Sciences of the United States of America, 114, 7641e7646.

333

Bubb, D.H, Lucas, M.C. and Thom, T.J. (2002) Winter movements and activity of signal crayfish Pacifastacus leniusculus in an upland river, determined by radio telemetry. In E.B. Thorstad, I.A. Fleming and T.F. Naesje (Eds.) Aquatic Telemetry. Netherlands: Springer. pp. 111-119.

Bubb, D.H., Thom, T.J. and Lucas, M.C. (2004) Movement and dispersal of the invasive signal crayfish Pacifastacus leniusculus in upland rivers. Freshwater Biology, 49(3), 357-368.

Bubb, D.H., Thom, T.J. and Lucas, M.C. (2008) Spatial ecology of the white‐ clawed crayfish in an upland stream and implications for the conservation of this endangered species. Aquatic Conservation, 18, 647-657.

Buffington, J.M. and Montgomery, D.R. (1997) A systematic analysis of eight decades of incipient motion studies, with special reference to gravel-bedded rivers. Water Resources Research, 33(8), 1993-2029.

Bujang, M.A., Sa’at, N., Sidik, T.M.I.T.A.B. and Joo, L.C. (2018) Sample size guidelines for logistic regression from observational studies with large population: emphasis on the accuracy between statistics and parameters based on real life clinical data. The Malaysian Journal of Medical Sciences, 25(4), 122-130.

Bursac, Z., Gauss, C.H., Williams, D.K. and Hosmer, D.W. (2008) Purposeful selection of variables in logistic regression. Source Code for Biology and Medicine, 3: 17. doi: 10.1186/1751-0473-3-17.

Burt, T.P., Howden, N.J.K., Worrall, F. and Whelan, M.J. (2008) Importance of long-term monitoring for detecting environmental change: lessons from a lowland river in south east England. Biogeosciences, 5, 1529–1535.

Butler, D.R. and Malanson, G.P. (1995) Sedimentation rates and patterns in beaver ponds in a mountain environment. Geomorphology, 13(1-4), 255-269.

Butler, D.R. and Malanson, G.P. (2005) The geomorphic influences of beaver dams and failures of beaver dams. Geomorphology, 71(1-2), 48-60.

Butler, D.R. (1995) Zoogeomorphology: Animals as Geomorphic Agents. Cambridge: Cambridge University Press.

334

Butler, D.R. (2006) Human-induced changes in animal populations and distributions, and the subsequent effects on fluvial systems. Geomorphology, 79(3-4), 448-459.

Butler, D.R. and Sawyer, C.F. (2012) Introduction to the special issue— zoogeomorphology and ecosystem engineering. Geomorphology, 157-158: 1-5.

Butler, D.R., Anzah, F., Goff, P.D. and Villa, J. (2018). Zoogeomorphology and resilience theory. Geomorphology, 305, 154-162.

Byrne, C.F., Lynch, J.M. and Bracken, J.J. (1999) A sampling strategy for stream populations of white-clawed crayfish, Austropotamobius pallipes (Lereboullet) (Crustacea, Astacidae). In Biology and environment: proceedings of the Royal Irish Academy, 99(2), 89-94.

Caine, N. (2004) Mechanical and chemical denudation in mountain systems. In P. Owens and O. Slaymaker (eds.) Mountain Geomorphology. Arnold: London. pp. 132-152.

Camici, S., Barbetta, S. and Moramarco, T. (2014) Levee body vulnerability to seepage: the case study of the levee failure along the Foenna stream on 1 January 2006 (central Italy). Journal of Flood Risk Management, 10(3), 314-325.

Carpenter, S.R. (1996) Microcosm experiments have limited relevance for community and ecosystem ecology. Ecology, 77, 667-680.

Carvalho, G.R., Shaw, P.W., Hauser, L., Seghers, B.H. and Magurran, A.E. (1996) Artificial introductions, evolutionary change and population differentiation in Trinidadian guppies (Poecilia reticulata: Poeciliidae). Biological Journal of the Linnean Society, 57(3), 219-234.

Chadwick, D.D.A. (2019) Invasion of the signal crayfish, Pacifastcus leniusculus, in England: implications for the conservation of the white-clawed crayfish, Austropotamobius pallipes. Published PhD Thesis, University College London.

Childress, M.J. and Herrnkind, W.F. (1996) The ontogeny of social behaviour among juvenile Caribbean spiny lobsters. Animal Behaviour, 51(3), 675-687.

Chlaib, H.K., Mahdi, H., Al-Shukri, H., Su, M.S., Catakli, A. and Abd, N. (2014) Using ground penetrating radar in levee assessment to detect small scale animal burrows. Journal of Applied Geophysics, 103, 121-131.

335

Choi, S-U., Yoon, B. and Woo, H. (2005) Effects of dam-induced flow regime change on downstream river morphology and vegetation cover in the Hwang River, Korea. River Research and Applications, 21(2-3), 315-325.

Church, M. (2000) Gravel-bed Rivers. In T. Burt and R. Allison (Eds.) Sediment Cascades: An Integrated Approach. John Wiley and Sons Ltd.: Chichester.

Church, M. (2013) Refocussing geomorphology: Field work in four acts. Geomorphology, 200, 184-192.

Clark, J.M., Kershner, M.W. and Holomuzki, J.R. (2008) Grain size and sorting effects on size- dependent responses by lotic crayfish to high flows. Hydrobiologia, 610, 55-66.

Clark, P.F. and Robbins, R.S. (2008) The Wey Navigation System, a fish and wildlife channel across Bulldogs Island and large burrowing invasive decapod species. A Report for the Environment Agency.

Clarke, S.J. (2002) Vegetation growth in rivers: influences upon sediment and nutrient dynamics. Progress in Physical Geography, 26(2), 159-172.

Collins, A.L. and Walling, D.E. (2007) Sources of fine sediment recovered from the channel bed of lowland groundwater-fed catchments in the UK. Geomorphology, 88(1-2), 120-138.

Collins, A.L., Anthony, S.G., Hawley, J. and Turner, T. (2009) The potential impact of projected change in farming by 2015 on the importance of the agricultural sector as a sediment source in England and Wales. Catena, 79(3), 243-250.

Coombes, M.A. (2016) Biogemorphology: diverse, integrative and useful. Earth Surface Processes and Landforms, 41, 2296-2300.

Cooper, R.J., Outram, F.N. and Hiscock, K.M. (2016) Diel turbidity cycles in a headwater stream: evidence of nocturnal bioturbation? Journal of Soils and Sediments, 16(6), 1815-1824.

Coops, H., Geilen, N., Verheij, H.J., Boeters, R. and van der Velde, G. (1996) interactions between waves, bank erosion and emergent vegetation: an experimental study in a wave tank. Aquatic Botany, 53, 187-198.

336

Corenblit, D., Tabacchi, E., Steiger, J. and Gurnell, A.M. (2007) Reciprocal interactions and adjustments between fluvial landforms and vegetation dynamics in river corridors: A review of complementary approaches. Earth-Science Reviews, 84(1-2), 56-86.

Cotterill, G.S., Cross, P.C., Middleton, A.D., Rogerson, J.D., Scurlock, B.M. and du Toit, J.T. (2018) Hidden cost of disease in a free-ranging ungulate: brucellosis reduces mid-winter pregnancy in elk. Ecology and Evolution, 8(22), 10733-10742.

Couper, P. (2003) Effects of silt-clay content on the susceptibility of river banks to subaerial erosion. Geomorphology, 56(1-2), 95-108.

Couper, P., Stott, T. and Maddock, I. (2002) Insights into river bank erosion processes derived from analysis of negative erosion-pin recordings: observations from three recent UK studies. Earth Surface Processes and Landforms, 27, 59-79.

Couper, P.R. and Maddock, A.P. (2001) Subaerial river bank erosion processes and their interaction with other bank erosion mechanisms on the River Arrow, Warwickshire, UK. Earth Surface Processes and Landforms, 26, 631-646.

Cover, M.R., May, C.L., Dietrich, W.E. and Resh, V.H. (2006) Quantitative linkages between sediment supply, streambed fine sediment, and benthic macroinvertebrates in the Klamath Mountaines, northern California. Proceedings of the Eighth Federal Interagency Sedimentation Conference (8thFISC), April 2-6, 2006, Reno, NV, USA.

Cowx, I.G. (1990) Growth and reproduction tactic of roach, Rutilus rutilus (L.), and dace, Leuciscus leuciscus (L.), populations in the Rivers Exe and Culm, England. Polskie Archiwum Hydrobiologii, 37, 195-210.

Craven, D., Thakuri, M.P., Cameron, E.K., Frelichs, L.E., Ejour, R., Blair, R.B., Blossey, B., Burti, J., Choi, A., Avalos, A., Fahey, T.J., Fisichelli, N.A., Gibson, K., Handa, I.T., Hopfensperger, K., Loss, S.R., Nuzzo, V., Maerz, J.C., Sackett, T., Scarenbroch, B.C., Smith, S.M., Vellend, M., Umek, L.G., and Eisenhauer, N. (2017) The unseen invaders: introduced earthworms as drivers of change in plant communities in North American forests (a meta- analysis). Global Change Biology, 23, 1065-1074.

337

Creed, R.P. and Reed, J.M. (2004) Ecosystem engineering by crayfish in a headwater stream community. Freshwater Science, 23(2), 224-236.

Crooks, J.A. (2002) Characterizing ecosystem-level consequences of biological invasions: the role of ecosystem engineers. Oikos, 97, 153-155.

Darwin, C. (1981) The Formation of Vegetable Mould, Through the Action of Worms. London: John Murray.

Davies, P.E. and Cook, L.S.J. (1993) Catastrophic macroinvertebrate drift and sublethal effects on brown trout, Salmo trutta, caused by cypermethrin spraying on a Tasmanian stream. Aquatic Toxicology, 27(3-4), 201-224.

Daws, A.G., Hock, K. and Huber, R. (2011) Spatial structure of hierarchical groups: testing for processes of aggregation, clustering, and spatial centrality in crayfish (Orconectes rusticus). Marine and Freshwater Behaviour and Physiology, 44, 209-222.

Dick, J.T.A., Nelson, N. and Bishop, J.D.D. (1997) Introduction experiments with Gammarus spp. (Crustacea: Amphipoda) in the Isle of Man (British Isles), 1949–1995. Zoology, 242(2), 209-216.

Diehl, S. (1992) Fish predation and benthic community structure: The role of omnivory and habitat complexity. Ecology, 72(5), 1646-1661.

Diez, J.R., Elosegi, A. and Pozo, J. (2001). Woody debris in north Iberian streams: influence of geomorphology, vegetation, and management. Environmental Management, 28(5), 687-698.

DiStefano, R.J., Decoske, J.J., Vangilder, T.M. and Barnes, L.S. (2003) Microhabitat partitioning among three crayfish species in two Missouri streams, U.S.A. Crustaceana, 76(3), 343-362.

Docker, B.B. and Hubble, T.C.T. (2008) Quantifying root-reinforcement of river bank coils by four Australian tree species. Geomorphology, 100 (3-4), 401-418.

Doeg, T.J. and Milledge, G.A. (1991) Effect of experimentally increasing concentrations of suspended sediment on macroinvertebrate drift. Australian Journal of Marine and Freshwater Research, 42(5), 519-526.

338

Dollar, E.S.J., James, C.S., Rogers, K.H. and Thoms, M.C. (2007) A framework for interdisciplinary understanding of rivers as ecosystems. Geomorphology, 89: 147–162.

Dolloff, C.A. (1987) Seasonal population characteristics and habitat use by juvenile coho salmon in a small southeast Alaska stream. Transactions of the American Fisheries Society, 116, 829-838.

Doretto, A., Piano, E., Bona, F. and Fenoglio, S. (2018) How to assess the impact of fine sediments on the macroinvertebrate communities of alpine streams? A selection of the best metrics. Ecological Indicators, 84, 60-69.

Douglas, I. (1996) The impact of land-use changes, especially logging, shifting cultivation, mining and urbanization on sediment yields in humid tropical Southeast Asia: a review with special reference to Borneo. Erosion and Sediment Yield: Global and Regional Perspectives (Proceesings of the Exeter Symposium, July 1996). IAHS no. 236, 463-471.

Drenner, R.W. and Mazumder, A. (1999) Microcosm experiments have limited relevance for community and ecosystem ecology: Comment. Ecology, 80(3), 1081-1085.

Droppo, I.G. (2001) Rethinking what constitutes suspended sediment. Hydrological Processes, 15, 1551-1564.

Dubiel, R.F., Blodgett, R.H. and Brown, T.M. (1987) Lungfish burrows in the upper Triassic Chinle and Dolores formations, Colorado Plateau. Journal of Sedimentary Petrouxy, 57(3), 512-521.

Duijsings, J. (1987) A sediment budget for a forested catchment in Luxembourg and its implications for channel development. Earth Surface Processes and Landforms, 12(2), 173- 184.

Dyer, J.J., Worthington, T.A. and Brewer, S.K. (2015) Response of crayfish to hyporheic water availability and excess sedimentation. Hydrobiologia, 747, 147-157.

Early, R., Bradley, B.A., Dukes, J.S., Lawler, J.J., Olden, J.D., Blumenthal, D.M., Gonzalez, P., Grosholz, E.D., Ibañez, I., Miller, L.P. and Sorte, C.J. (2016) Global threats from invasive

339 alien species in the twenty-first century and national response capacities. Nature Communications, 7.

El-Dien, A.A., Takebayashi, H. and Fujita, M. (2015) Erosion and collapse of riverbanks under different flood conditions. Annual Journal of Hydraulic Engineering, 59.

Emery‐ Butcher, H.E., Beatty, S.J. and Robson, B.J. (2020) The impacts of invasive ecosystem engineers in freshwaters: A review. Freshwater Biology, 65, 999-1015.

Engel, S. (1998), The role and interactions of submersed macrophytes in a shallow Wisconsin Lake USA. Journal of Freshwater Ecology, 4, 329-342.

Englund, G. and Krupa, J.J. (2008) Habitat use by crayfish in stream pools: influence of predators, depth and body size. Freshwater Biology, 43, 75-83.

Engstrom, D.R., Almendinger, J.E. and Wolin, J.A. (2009) Historical changes in sediment and phosphorus loading to the upper Mississippi River: mass-balance reconstructions from the sediments of Lake Pepin. Journal of Paleolimnology, 41(4), 563-588.

Environment Agency (2016) Freshwater Fish Counts for all Species, all Areas and all Years. Retrieved from https://data.gov.uk/dataset/f49b8e4b-8673-498e-bead- 98e6847831c6/freshwater-fish-counts-for-all-species-all-areas-and-all-years, August 22nd 2019.

Escapa, M., Minkoff, D.R., Perillo, G.M. and Iribarne, O. (2007) Direct and indirect effects of burrowing crab Chasmagnathus granulatus activities on erosion of southwest Atlantic Sarcocornia‐ dominated marshes. Limnology and Oceanography, 52(6), 2340-2349.

European Environment Agency (2012) Hydromorphological alterations and pressures in European rivers, lakes, transitional and coastal waters; thematic assessment for the EEA Water 2012. Available at: https://www.eionet.europa.eu/etcs/etc-icm/products/etc-icm- reports/hydromorphological-alterations-and-pressures-in-european-rivers-lakes-transitional- and-coastal-waters-etc-icm-technical-report-2-2012. Access date: 11th April 2020.

340

Evans, D.H., Piermarini, P.M. and Choe, K.P. (2005) The multifunctional fish gill: Dominant site of gas exchange, osmoregulation, acid-base regulation, and excretion of nitrogenous waste. Physiological Reviews, 85(1), 97-177.

Evans, E.P. (1906) The Criminal Prosecution and Capital Punishment of Animals. William Heinemann: London.

Everall, N.C., Johnson, M.F., Wood, P.J. and Mattingley, L. (2018) Sensitivity of the early life stages of a mayfly to fine sediment and orthophosphate levels. Environmental Pollution, 237, 792-802.

EWT (2016) Riversearch: How to survey for otters, water voles and water shrews. Available at: http://www.essexwt.org.uk/sites/default/files/how_to_survey.pdf . [Access date: 21st Dec 2016].

Faller, M., Harvey, G.L., Henshaw, A.J., Bertoldi, W., Bruno, M.C. and England, J. (2016) River bank burrowing by invasive crayfish: Spatial distribution, biophysical controls and biogemorphic significance. Science of the Total Environment, 569-570, 1190-1200.

Farnsworth , K.L. and Milliman, J.D. (2003) Effects of climatic and anthropogenic change on small mountainous rivers: the Salinas River example. Global and Planetary Change, 39(1-2), 53-64.

Fatahi, B., Pathirage, U., Indraratna, B., Pallewattha, M. and Khabbaz, H. (2015) The role of native vegetation in stabilizing formation soil for transport corridors: and Australian experience. In: B. Indraratna, J. Chu and C. Rujikiatkamjorn (Eds.) Ground Improvement Case Histories: Chemical, Electrokinetic, Thermal and Bioengineering. Elsevier: Oxford. pp. 591- 628.

Fei, S., Phillips, J. and Shous, M. (2014) Biogeomorphic impcats of invasive species. Annual Review of Ecology, Evolution and (Systemat), 45, 69-87.

Field-Dodgson, M.S. (1987) The effect of salmon redd excavation on stream substrate and benthic community of two salmon spawning streams in Canterbury, New Zealand. Hydrobiologia, 154(1), 3-11.

341

Financial Times (2020) Japan identifies first case of virus that originated in China. Financial Times. Available at: https://www.ft.com/content/a99a8cfa-3810-11ea-a6d3-9a26f8c3cba4 . Access date: 13th May 2020.

Flinders, C.A. and Magoulick, D.D. (2005) Distribution, habitat use and life history of stream- dwelling crayfish in the Spring River drainage of Arkansas and Missouri with a focus on the imperilled Mammoth Spring Crayfish (Orconectes marchandi). The American Midland Naturalist, 154(2), 358-374.

Fortin, D., Beyer, H.L., Boyce, M.S., Smith, D.W., Duchesne, T. and Mao, J.S. (2005) Wolves influence elk movement: Behaviour shapes a trophic cascade in Yellowstone National Park. Ecology, 86(5), 1320-1330.

Foucher, A., Salvador‐ Blanes, S., Vandromme, R., Cerdan, O. and Desmet, M. (2017) Quantification of bank erosion in a drained agricultural lowland catchment. Hydrological Processes. 31, 1424–1437.

Fox, G.A., Wilson, G.V., Simon, A., Langedoen, E.J., Akay, O. and Fuchs, J.W. (2007) Measuring streambank erosion due to ground water seepage: correlation to bank pore water pressure, precipitation and stream stage. Earth Surface Processes and Landforms, 32(10), 1558-1573.

Fox, M.G. and Keast, A. (1990) Effects of winterkill on population structure, body size, and prey consumption patterns of pumpkinseed in isolated beaver ponds. Canadian Journal of Zoology, 68, 2489-2498.

Frankel, K.L., Pazzaglia, F.J. and Vaughn, J.D. (2007 Knickpoint evolution in a vertically bedded substrate, upstream-dipping terraces, and Atlantic slope bedrock channels. Geological Society of America Bulletin, 119(3-4), 476–486.

Frazar, S., Gold, A.J., Addy, K., Moatar, F., Birgand, F., Schroth, A.W., Kellogg, D.Q. and Pradhanang, S.M. (2019) Contrasting behaviour of nitrate and phosphate flux from high flow events on small agricultural and urban watersheds. Biogeochemistry, 145, 141-160.

Fredlund, D., Morgenstern, N. and Widger, R. (1978) The shear strength of saturated soils. Canadian Geotechnical Journal, 15, 313-321.

342

Fredlund, D.G. and Rahardjo, H. (1993) Soil Mechanics of Unsaturated Soils. Wiley: New York, p.517.

Fredlund, D.G., Morgenstern, N.R. and Widger, R.A. (1978) The shear strength of saturated soils. Canadian Geotechnical Journal, 15, 313-321.

Freeman, M.A., Turnbull, J.F., Yeomans, W.E. and Bean, C.W. (2010) Prospects for management strategies of invasive crayfish populations with an emphasis on biological control. Aquatic Conservation: Marine and Freshwater Ecosystems, 20, 211-223.

Fremier, A.K., Yanites, B.J. and Yager, E.M. (2018). Sex that moves mountains: The influence of spawning fish on river profiles over geologic timescales. Geomorphology, 305, 163-172.

Friedkin, J.F. (1945) A laboratory study of the meandering of alluvial rivers. USACE Waterways Experiment Station, Vicksburg, Mississippi, USA.

Friel, J.P. (2008) Pseudobunocephalus, a new genus of banjo catfish with the description of a new species from the Orinoco River system of Colombia and Venezuela (Siluriformes: Aspredinidae). Neotropical Ichthyology, 6(3), 293-300.

Furness, A.I. (2015) The evolution of an annual life cycle in killifish: adaptation to ephemeral aquatic environments through embryonic diapause. Biological Reviews, 91, 796-812.

Gardiner, T. (1983) Some factors promoting channel bank erosion, River Lagan, County Down. Journal of Earth Sciences, 5(2), 231-239.

Gardner, M.B. (1981) Effects of turbidity on feeding rates and selectivity of bluegills. Transactions of the American Fisheries Society, 110, 446-450.

Garner, G., Malcolm, I.A., Sadler, J.P. and Hannah, D.M. (2017) The role of riparian vegetation density, channel orientation and water velocity in determining river temperature dynamics. Journal of Hydrology, 553, 471-485.

Gatto, L.W. (1995) Soil freeze-thaw effects on bank erodibility and stability. USACE, Cold Regions Research Engineering Laboratory, Special Report 95-24.

343

Gayraud, S. and Philippe, M. (2003) Influence of bed-sediment features on the interstitial habitat available for macroinvertebrates in 15 French streams. International Review of Hydrobiology, 88(1), 77-93.

Gherardi, F. (2013) Crayfish as global invaders: Distribution, impact on ecosystem services and management options. Freshwater Crayfish, 19(2), 177-187.

Gherardi, F. and Barbaresi, S. (2000) Invasive crayfish: activity patterns of Procambarus clarkii in the rice fields of the lower Guadalquivir (Spain). Archiv fur Hydrobiologie, 150(1), 153- 168.

Gherardi, F., Coignet, A., Souty-Grosst, C., Spigoli, D. and Aquiloni, L. (2013) Climate warming and the agonistic behaviour of invasive crayfishes in Europe. Freshwater Biology, 58, 1958-1967.

Gladman, Z.F., Yeomans, W.E., Adams, C.E., Bean, C.W., McColl, D., Olszewska, J.P., McGillivray, C.W. and McCluskey, R. (2010) Detecting North American signal crayfish (Pacifastacus leniusculus) in riffles. Aquatic Conservation: Marine and Freshwater Ecosystems, 20, 588-594.

Global Invasive Species Database (2020) 100 of the World's Worst Invasive Alien Species. Available at: http://www.iucngisd.org/gisd/100_worst.php. Access date: 14th April 2020.

Goessmann, C., Hemelrijk, C. and Huber, R. (2000) The formation and maintenance of crayfish hierarchies: behavioural and self-structuring properties. Behavioural Ecology and Sociobiology, 48, 418-428.

Gomez, C., Oguchi, T. and Evans, I.S. (2015) Spatial analysis in geomorphology (1): Present directions, from collection to processing. Geomorphology, doi: 10.1016/j.geomorph.2015.04.02 .

Gomi, T., Kobayashi, S., Negishi, J.N. and Imaizumi, F. (2010) Short-term responses of macroinvertebrate drift following experimental sediment flushing in a Japanses headwater channel. Landscape and Ecological Engineering, 6, 257-270.

344

Gottesfeld, A.S., Hassan, M.A., Tunnicliffe, J.F. and Poirier, R.W. (2004) Sediment dispersion in salmon spawning streams: the influence of floods and salmon redd construction. Journal of the American Water Resources Association, 40(4), 1071-1086.

Gradall, K.S. and Swenson, W.A. (1982) Responses of brook trout and creek chub to turbidity. Transactions of the American Fisheries Society, 110, 446-450.

Grauso, S., Pagano, A., Fattoruso, G., De Bonis, P., Onori, F., Regina, P. and Tebano, C. (2008). Relations between climatic–geomorphological parameters and sediment yield in a mediterranean semi-arid area (Sicily, southern Italy). Environmental Geology, 54(2), 219-234.

Gray, L.J. and Ward, J.V. (1982) Effects of sediment releases from a reservoir on stream macroinvertebrates. Hydrobiologia, 96, 177-184.

Green, N. (2009) Guidance on the use of Artificial Refuge Traps for White-clawed and nonnative crayfish species. Report for Green Ecology Ltd.

Green, S.J., Atkins, J.L. and Cote, I.M. (2011) Foraging behaviour and prey consumption in the Indo-Pacific lionfish on Bahamian coral reefs. Marine Ecology Progress Series, 433, 159- 167.

Greenwood, P. and Kuhn, N.J. (2014) Does the invasive plant, Impatiens glandulifera, promote soil erosion along the riparian zone? An investigation on a small watercourse in northwest Switzerland. Journal of Soils and Sediments, 14, 637-650.

Greenwood, P., Baumann, P., Pulley, S. and Kuhn, N.J. (2018) The invasive alien plant, Impatiens glandulifera (Himalayan Balsam),and increased soil erosion: causation or association? Case studies from a river system in Switzerland and the UK. Journal of Soils and Sediments, 18(12), 3463-3477.

Griffith, M.B. and Perry, S.A. (1993) Colonization and processing of leaf litter by macroinvertebrate shredders in streams of contrasting pH. Freshwater Biology, 30(1), 93-103.

Griffiths, S.W., Collen, P. and Armstrong, J.D. (2004) Competition for shelter among over‐ wintering signal crayfish and juvenile Atlantic salmon. Journal of Fish Biology, 65(2), 436- 447.

345

Grimaud, J-L., Paola, C. and Voller, V. (2016) Experimental migration of knickpoints: influence of style of base-level fall and bed lithology. Earth Surface Dynamics, 4, 11-23.

Guan, R.Z. (1994) Burrowing behaviour of signal crayfish, Pacifastacus leniusculus (Dana), in the River Great Ouse, England. Freshwater Forum, 4, 155-168.

Guan, R.Z. and Wiles, P.R. (1997) The home range of signal crayfish in a British lowland river. Freshwater Forum, 8, 45-54.

Guan, R.Z. and Wiles, P.R. (1999) Growth and reproduction of the introduced signal crayfish Pacifastacus leniusculus in a British lowland river. Fisheries Research, 42(3) 245-259.

Gurnell, A.M., Corenblit, D., Garcia de Jalon, D., Monzalel del Tanago, M., Grabowski, R.C., O’Hare, M.T. and Szewczyk, M. (2016) A conceptual model of vegetation- hydrogeomorphology interactions within river corridors. River Research and Applications, 32, 142-163.

Gurnell, A.M., Gregory, K.J., Hollis, S. and Hill, C.T. (1985) Detrended correspondence analysis of heathland vegetation: The identification of runoff contributing areas. Earth Surface Processes and Landforms, 10, 343-351.

Gurnell, A.M., van Oosterhout, M.P., de Vlieger, B. and Goodson, J.M. (2006) Reach-scale impacts of aquatic plant growth on physical habitat. River Research and Applications, 22, 667- 680.

Hackney, C., Best, J., Leyland, J., Darby, S., Parsons, D., Aalto, R. and Nicholas, A. (2015) Modulation of outer bank erosion by slump blocks: Disentangling the protective and destructive role of failed material on the three‐ dimensional flow structure. Geophysical Research Letters, 42(24), 10,663-10,670.

Hackney, C., Best, J., Leyland, J., Darby, S.E., Parsons, D., Aalto, R. and Nicholas, A. (2015) Modulation of outer bank erosion by slump blocks: Disentangling the protective and destructive role of failed material on the three‐ dimensional flow structure. Geophysical Research Letters, 42(24), 10,663-10,670.

346

Haggard, B.E., Stanley, E.H. and Hyler, R. (1999) Sediment-Phosphorus relationships in three northcentral Oklahoma streams. Transactions of the ASSAE, 42(6), 1709-1714.

Hannah, D.M., Wood, P.J. and Sadler, J.P. (2004) Ecohydrology and Hydroecology: A ‘new paradigm’? Hydrological Processes, 18, 3439-3445.

Hargeby, A., Andersson, G., Blindow, I. and Johansson, S. (1994) Trophic web structure in a shallow eutrophic lake during a dominance shift from phytoplankton to submerged macrophytes. Hydrobiologia, 279(1), 83-90.

Harris, H.E., Baxter, C.V., Davis, J.M. (2018) Wildfire and debris flows affect prey subsidies with implications for riparian and riverine predators. Aquatic Sciences, 80, 37.

Harris, R.R. and Young, H.J. (1996) Distribution, densities and population characteristics of signal crayfish, Pacifastacus leniusculus (Dana), in the Gaddesby Brook, Leicestershire. Report for the National Rivers Authority.

Harvey, G.L. and Bertoldi, W. (2015) Dynamics riverine landscapes: the role of ecosystem engineers. Earth Surface Processes and Landforms, 40, 1701-1704.

Harvey, G.L., Henshaw, A.J., Brasington, J. and England, J. (2019) Burrowing invasive species: And unquantified erosion risk at the aquatic-terrestrial interface. Reviews of Geophysics, 57(3), 1018-1036.

Harvey, G.L., Henshaw, A.J., Moorhouse, T.P., Clifford, N.J., Holah, H., Grey, J. and Macdonald, D.W. (2014) Invasive crayfish as drivers of fine sediment dynamics in rivers: field and laboratory evidence. Earth Surface Processes and Landforms, 39, 259-271.

Harvey, G.L., Moorhouse, T.P., Clifford, N.J., Henshaw, A.J., Johnson, M.F., Macdonald, D.W., Reid, I. and Rice, S.P. (2011) Evaluating the role of invasive aquatic species as drivers of fine sediment-related river management problems: The case of the signal crayfish (Pacifastacus leniusculus). Progress in Physical Geography, 35, 517-533.

Hasegawa, K. (1989) Universal Bank Erosion Coefficient for Meandering Rivers. Journal of Hydraulic Engineering, 115(6), 744-765.

347

Hasiotis, S.T. and Mitchell, C.E. (1993) A comparison of crayfish burrow morphologies: Triassic and Holocene fossil, paleo- and neo-ichnological evidence, and the identification of their burrowing signatures. Ichnos, 2, 291-314.

Hassan, M.A., Gottesfeld, A.S., Montgomery, D.R., Tunnicliffe, J.F., Clarke, G.K., Wynn, G., Jones-Cox, H., Poirier, R., MacIsaac, E., Herunter, H. and Macdonald, S.J. (2008) Salmon‐ driven bed load transport and bed morphology in mountain streams. Geophysical Research Letters, 35(4).

Hassard, F., Gwyther, C.L., Farkas, K., Andrews, A., Jones, V., Cox, B., Brett, H., Jones, D.L., McDonald, J.E. and Malham, S.K. (2016) Abundance and distribution of enteric bacteria and viruses in coastal and estuarine sediments – a review. Frontiers in Micobiology, 7, 1692.

Hastings, A., Byers, J.E., Crooks, J.A., Cuddington, K., Jones, C.G., Lambrinos, J.G., Talley, T.S., and Wilson, W.G. (2007) Ecosystem engineering in space and time. Ecology Letters, 10, 153-164.

Haussmann, N.S. (2017) Soil movement by burrowing mammals: A review comparing excavation size and rate to body mass of excavators. Progress in Physical Geography, 41(1), 29-45.

Hazlett, B.A., Acquitapace, P. and Gherardi, F. (2002) Differences in memory capabilities in invasive and native crayfish. Journal of Crustacean Biology, 22(2), 439-448.

Heaming, P.D. (2012) Ecosystem Engineers: organisms that create, modify and maintain habitats. ECOLOGY.INFO #12.

Hembree, D.I., Martin, L.D. and Hasiotis, S.T. (2004) Amphibian burrows and ephemeral ponds of the Lower Permian Speiser Shale, Kansas: evidence for seasonality in the midcontinent. Palaeogeography, Palaeoclimatology, Palaeoecology, 203, 127-152.

Henshaw, A.J., Thorne, C.R. and Clifford, N.J. (2013) Identifying causes and controls of river bank erosion in a British upland catchment. CATENA, 100, 107-119.

Hey, R.D. and Thorne, C.R. (1986) Stable channels with mobile gravel beds. Journal of Hydraulic Engineering, 112(8), 671-689.

348

Hickin, E.J. (1984) Vegetation and river channel dynamics. The Canadian Geographer, 28(2), 111-126.

Hirvonen, H., Holopainen, S., Lempiainen, N., Selin, M. and Tulonen, J. (2007) Sniffing the trade-off: Effects of eel odours on nocturnal foraging activity of native and introduced crayfish juveniles. Marine and Freshwater Behaviour and Physiology, 40(3), 213-218.

Hjulstrom, F. (1935) The Morphological Activity of Rivers as Illustrated by River Fyris. Uppsala: Almqvist and Wiksells.

Hoffman, D.F. and Gabet, E.J. (2007) Effects of sediment pulses on channel morphology in a gravel-bed river. GSA Bulletin, 119(1-2), 116-125.

Hogan, E.S., Houpt, K.A. and Sweeney, K. (1988) The effect of enclosure size on social interactions and daily activity patterns of the captive Asiatic wild horse (Equus przewalskii). Applied Animal Behaviour Science, 21(1-2), 147-168.

Holdich, D.M. and Reeve, I.D. (1991) Distribution of freshwater crayfish in the British Isles, with particular reference to crayfish plague, alien introductions and water quality. Aquatic Conservation: Marine and Freshwater Ecosystems, 1, 139-158.

Holdich, D. M., Reader, J. P. and Rogers, W. D. (1994) Crayfish Conservation. Interim Report for 1993 (Environment Agency): R&D Project 378.

Holway, D.A. and Suarez, A.V. (1999) Animal behavior: an essential component of invasion biology. Trends in Ecology and Evolution, 14(8), 328-330.

Holzenthal, R., Blahnik, R., Prather, A. and Kjer, K. (2010) Trichiptera: Tree of Life web project. Available at: http://tolweb.org/Trichoptera/8230/2010.01.12 [Access date: 15th Nov 2016].

Hondolero, D. and Edwards, M.S. (2017) Changes in ecosystem engineers: the effects of kelp forest type on current and benthic assemblages in Kachemak Bay, Alaska. Marine Biology, 164: 81.

Hooke, J.M. (1979) An analysis of the process of river bank erosion. Journal of Hydrology, 42, 49-62.

349

Houde, E.D. and Petersen, J.E. (2009) Physical and ecological complexity. In: J.E. Petersen, V.S. Kennedy, W.C. Dennison and W.M. Kemp (Eds) Enclosed Experimental Ecosystems and Scale: Tools for Understanding and Managing Coastal Ecosystems. Springer: New York.

Houghton, R.J., Wood, C. and Lambin, X. (2017) Size-mediated, density-dependent cannibalism in the signal crayfish Pacifastacus leniusculus (Dana, 1852) (Decapoda, Astacidea), an invasive crayfish in Britain. Crustaceana, 90(4), 417-435.

Hudina, S., Zganec, K. and Hock, K. (2015) Differences in aggressive behaviour along the expanding range of an invasive crayfish: an important component of invasion dynamics. Biological Invasions, 17(11), 3101-3112.

Huettmann, F. and Diamond, A.W. (2010) Using PCA scores to classify species communities: An example for pelagic seabird distribution. Journal of Applied Statistics, 28(7), 843-853.

Hughes, A.R., Mann, D.A. and Kimbro, D.L. (2014) Predatory fish sounds can alter crab foraging behaviour and influence bivalve abundance. Proceedings of the Royal Society B, 281, 20140715.

Hughes, R.G. and Paramor, O.A.L. (2004) On the loss of saltmarshes in south-east England and methods for their restoration. Journal of Applied Ecology, 41, 440-448.

Hughes, Z.J., FitzGerald, D.M., Wilson, C.A., Pennings, S.C., Więski, K. and Mahadevan, A. (2009) Rapid headward erosion of marsh creeks in response to relative sea level rise. Geophysical Research Letters, 36(3).

Hunter, B.A., Johnson, M.S. and Thompson, D.J. (1987) Ecotoxicology of copper and cadmium in a contaminated grassland ecosystem. II. Invertebrates. Journal of Applied Ecology, 24, 587-599.

Hynes, H.B.N. (1970) The Ecology of Running Waters. The Blackburn Press: Blackburn.

IBM (2015) IBM SPSS Statistics for Windows, Version 23.0. Armonk, NY: IBM Corp.

Independent (2013) Independent web-site. Available at: https://www.independent.co.uk/news/uk/politics/badgers-put-village-at-risk-after-digging- through-flood-defences-8456043.html. Access date: 12th April 2020.

350

Ion, M.C., Puha, A.E., Suciu, T. and Parvulescu, L. (2020) Get a grip: unusual disturbances drive crayfish to improvise. Behaviour, 157, 101-120.

Irish Waterway History (2020) Irish Waterways History web-site. Available at: https://irishwaterwayshistory.com/abandoned-or-little-used-irish-waterways/midlands-turf- waterways/the-finnery-river-navigation/ . Access date: 13th May 2020.

Issa, F.A., Adamson, D.J. and Edwards, D.H. (1999) Dominance hierarchy formation in juvenile crayfish Procambarus clarkii. Journal of Experimental Biology, 202, 3497-3506.

Jackson, T.R., Apte, S.V., Haggerty, R. and Budwig, R. (2015) Flow structure and mean residence times of lateral cavities in open channel flows: influence of bed roughness and shape. Environmental Fluid Mechanics, 15, 1069-1100.

Jaffe, R., McKnight, D., Maie, N., Cory, R., McDowell, W.H. and Campbell, J.L. (2008) Spatial and temporal variations in DOM composition in ecosystems: The importance of long‐ term monitoring of optical properties. Journal of Geophysical Research: Biogeosciences, 113, G4.

James, J., Thomas, J.R., Ellis, A., Young, K.A., England, J. and Cable, J. (2016) Over-invasion in a freshwater ecosystem: newly introduced virile crayfish (Orconectes rusticus) outcompete established invasive signal crayfish (Pacifastacus leniusculus). Marine and Freshwater Behaviour and Physiology, 49(1), 9-18.

Janzekovic, F. and Novak, T. (2012) PCA – A Powerful Method for Analyze Ecological Niches. In P. Sanguansat (Ed.) Principal Component Analysis – Multidisciplinary Applications. InTech: Rijeka.

Jensen, D.W., Steel, E.A., Fullerton, A.H. and Pess, G.R. (2009) Impact of fine sediment on egg-to-fry survival of Pacific Salmon: A meta-analysis of published studies. Reviews in Fisheries Science, 17(3), 348-359.

Jewson, D.H. and Taylor, J.A. (1978) The influence of tubidity on net phytoplankton photosynthesis in some Irish lakes. Freshwater Biology, 8(6), 573-584.

351

Jia, Y., Qu, Z., Wang, C. and Yang, Z. (2017) Effects of heavy metal accumulation in freshwater fishes: species, tissues, and sizes. Environmental Science and Pollution Research, 24, 9379-9386.

Johnsen, S.I. and Taugbol, T. (2010) NOBANIS – Invasive Alien Species Fact Sheet – Pacifastacus leniusculus. NOBANIS web-site. Available at: www.nobanis.org [Access date 27 Oct 2016].

Johnson, J.P. and Whipple, K.X. (2007) Feedbacks between erosion and sediment transport in experimental bedrock channels. Earth Surface Processes and Landforms, 32(7), 1048-1062.

Johnson, M.F. (2010) The disturbance of fluvial gravel substrates by signal crayfish (Pacifastacus leniusculus) and the implications for coarse sediment transport in gravel-bed rivers. PhD Thesis, Loughborough University.

Johnson, M.F., Reid, I., Rice, S.P. and Wood, P.J. (2009) Stabilization of fine gravels by net- spinning caddisfly larvae. Earth Surface Processes and Landforms, 34(3), 413-423.

Johnson, M.F., Rice, S.P. and Reid, I. (2010) Topographic disturbance of subaqueous gravel substrates by signal crayfish (Pacifastacus leniusculus). Geomorphology, 123, 269-278.

Johnson, M.F., Rice, S.P. and Reid, I. (2011) Increase in coarse sediment transport associated with disturbance of gravel river beds by signal crayfish (Pacifastacus leniusculus). Earth Surface Processes and Landforms, doi: 10.1002/esp/2192.

Johnson, M.F., Rice, S.P. and Reid, I. (2014) The activity of signal crayfish (Pacifastacus leniusculus) in relation to thermal and hydraulic dynamics of an alluvial stream, UK. Hydrobiologia, 724, 41-54.

Johnson, M.F., Thorne, C.R., Castro, J.M., Kondolf, G.M., Mazzacano, C.S., Rood, S.B. and Westbrook, C. (2019) Biomic river restoration: A new focus for river management. River Research and Applications, 1-10. doi: 10.1002/rra.3529 .

Jones, C. and DiRienzo, N. (2018) Behavioral variation post-invasion: Resemblance in some, but not all, behavioral patterns among invasive and native praying mantids. Behavioural Processes, 152, 92-99.

352

Jones, C.G., Lawton, J.H. and Shachak, M. (1994) Organisms as ecosystem engineers. In F. Samson and F. Knopf (Eds.) Ecosystem Management. New York: Springer. pp. 130-147.

Jones, J.I., Collins, A.L., Naden, P.S. and Sear, D.A. (2012a) The relationship between fine sediment and macrophytes in rivers. River Research and Applications, 28(7), 1006-1018.

Jones, J.I., Murphy, J.F., Collins, A.L., Sear, D.A., Naden, P.S. and Armitage, P.D. (2012b) The impact of fine sediment on macro-invertebrates. River Research and Applications, 28, 1055-1071.

Jugie, M., Gob, F., Virmoux, C., Brunstein, D., Tamisier, V., Le Coeur, C. and Grancher, D. (2018) Characterizing and quantifying the discontinuous bank erosion of a small low energy river using Structure-from-Motion Photogrammetry and erosion pins. Journal of Hydrology, 563, 418-434.

Julian, J.P. and Torres, R. (2006) Hydraulic erosion of cohesive riverbanks. Geomorphology, 76, 193-206.

Kaller, M.D. and Hartman, K.J. (2004) Evidence of a threshold level of fine sediment accumulation for altering benthic macroinvertebrate communities. Hydrobiologia, 518(1), 95- 104.

Karatayev, A.Y., Burlakova, L.E., Padilla, D.K., Mastitsky, S.E. and Olenin, S. (2009) Invaders are not a random selection of species. Biological Invasions, 11, 2009-2019.

Kareiva, P. and Andersen, M. (1988) Spatial Aspects of Species Interactions: the Wedding of Models and Experiments. In A. Hastings (ed.) Community Ecology. Lecture Notes in Biomathematics, vol 77. Springer: Berlin. pp. 35-50.

Kay, A.L. and Jones, D.A. (2011) Transient changes in flood frequency and timing in Britain under potential projections of climate change. International Journal of Climatology, 32(4), 489-502.

Kemp, P., Sear, D., Collins, A., Naden, P. and Jones, I. (2011) The impacts of fine sediment on riverine fish. Hydrological Processes, 25, 1800-1821.

353

Kemp, W., Petersen, J.E., Houde, E., Chen, C., Cornwell, J. and Porter, E. (2009) Spatial and temporal scaling. In J.E. Petersen, V.S. Kennedy, W.C. Dennison and M. Kemp (eds.) Enclosed Experimental Ecosystems and Scale. Springer: New York. pp. 49-62.

Kimiaghalam, N., Clark, S.P. and Ahmari, H. (2016) An experimental study on the effects of physical, mechanical, and electrochemical properties of natural cohesive soils on critical shear stress and erosion rate. International Journal of Sediment Research, 31(1), 1-15.

Kimmerer, W.J., Gartside, E. and Orsi, J.J. (1994). Predation by an introduced clam as the likely cause of substantial declines in zooplankton of San Francisco Bay. Marine Ecology Progress Series, 81-93.

Kleinhans, M., van Dijk, W., van de Lageweg, W., Hoendervoogt, R., Markies, H. and Schuurman, F. (2010) From nature to lab: scaling self-formed meandering and braided rivers. River Flow 2010, 1001-1009.

Kleinhans, M.G., van Dijk, W.M., van de Lageweg, W.I., Hoyal, D.C.J.D., Markies, H., van Maarseveen, M., Roosendaal, C., van Weesep, W., van Breemen, D., Hoendervoogt, R. and Cheshier, N. (2014) Quantifiable effectiveness of experimental scaling of river- and delta morphodynamics and stratigraphy. Earth Science Reviews, 133, 43-61.

Knighton, A.A. (1973) Riverbank erosion in relation to streamflow conditions, River Bollin- Dean, Cheshire. East Midlands Geographer, 6, 416-426.

Knox, J., Caccache, A., Hess, T. and Haro, D. (2016) Meta-analysis of climate impacts and uncertainty on crop yields in Europe. Environmental Research Letters, 11, 113004.

Kondolf, G.M. (1997) Hungry water: effects of dams and gravel mining on river channels. Environmental Management, 21, 533-551.

Kondolf, G.M., Loire, R., Piegay, H. and Malavoi, J-R. (2019) Dams and channel morphology. Environmental Flow Assessment: Methods and Applications, 143-161.

Korppoo, M., Huttenen, M., Huttenen, I., Piirainen, V. and Vehvilainen, B. (2017) Simulation of bioavailable phosphorus and nitrogen loading in an agricultural river basin in Finland using VEMALA v.3. Journal of Hydrology, 549, 363-373.

354

Kouba, A., Petrusek, A. and Kozak, P. (2014) Continental-wide distribution of crayfish species in Europe: update and maps. Knowledge and Management of Aquatic Ecosystems, 413, 5.

Kozak, P., Kajtman, J., Kouril, J. and Policar, T. (2002) The effect of crayfish density and shelter number on the daily activity of signal crayfish. Freshwater Crayfish, 13, 457-462.

Kravitz, E.A. and Huber, R. (2003) Aggression in invertebrates. Current Opinion in Neurobiology, 13(6), 736-743.

Kronvang, B., Andersen, H.E., Larsen, S.E. and Audet, J. (2013) Importance of bank erosion for sediment input, storage and export at the catchment scale. Journal of Soils and Sediments. 13(1), 230-241.

Krzeminska, D., Kerkhof, T., Skalsveen, K. and Stolte, J. (2019) Effect of riparian vegetation on stream bank stability in small agricultural catchments. CATENA, 172, 87-96.

Kundzewicz, Z.W. (2002) Ecohydrology – seeking consensus on interpretation of the notion. Hydrological Science Journal, 47, 799-804.

Ladle, M. and Griffiths, B.S. (1980) A study on the faeces of some chalk stream invertebrates. Hydrobiologia, 303, 195-206.

Lake, R.G. and Hinch, S.G. (1999) Acute effects of suspended sediment angularity on juvenile coho salmon (Oncorhynchus kisutch). Canadian Journal of Fisheries and Aquatic Science. 56, 862-867.

Larsen, A., Alvarez, N., Sperisen, C. and Lane, S. N. (2018) Biotic drivers of river and floodplain geomorphology – New molecular methods for assessing present‐day and past biota. Earth Surface Processes and Landforms, 43(1), 333-338.

Lamb, M.P., Finnegan, N.J., Scheingross, J.S. and Sklar, L.S. (2015). New insights into the mechanics of fluvial bedrock erosion through flume experiments and theory. Geomorphology, 244, 33–55.

Lancaster, J. and Ledger, M.E. (2015) Population‐ level responses of stream macroinvertebrates to drying can be density‐ independent or density‐ dependent. Freshwater Biology, 60(12), 2559-2570.

355

Land Trust (2019) New beaver dams at Camp Polk Meadow Preserve. Available at: https://www.deschuteslandtrust.org/news/news-items/2019-news-items/2019-cpm-beaver- update . Access date: 8th May 2020.

Lapointe, M. and Carson, M. (1986) Migration patterns of an asymmetric meandering river: The Rouge River, Quebec. Water Resources Research, 22(5) 731-743.

Larson, E.R. and Olden, J.D. (2011) The state of crayfish in the Pacific Northwest. Fisheries, 36(2), 60-73.

Larson, E.R., DiStefano, R.J., Magoulick, D.D. and Westhoff, J.T. (2008) Efficiency of a quadrat sampling technique for estimating riffle-dwelling crayfish density. North American Journal of Fisheries Management, 28(4), 1036-1043.

Laubel, A., Kronvang, B., Hald, A.B., and Jensen, C. (2003). Hydromorphological and biological factors influencing sediment and phosphorus loss via bank erosion in small lowland rural streams in Denmark. Hydrological Processes, 17(17), 3443-3463.

Lawler, D.M. (1978) The use of erosion pins in river banks. Swansea Geographer, 16, 9–18.

Lawler, D.M. (1992) Design and installation of a novel automatic erosion monitoring system. Earth Surface Processes and Landforms, 17(5), 455-463.

Lawler, D.M. (1993) The measurement of river bank erosion and lateral channel change: a review. Earth Surface Processes and Landforms, Technical Software Bulletin 18, 777–821.

Lawler, D.M., Grove, J.R., Couperthwaite, J.S., and Leeks, G.J.L. (1999). Downstream change in river bank erosion rates in the Swale–Ouse system, northern England. Hydrological Processes, 13(7), 977.

Ledger, M.E., Edwards, F.K., Brown, L.E., Milner, A.M. and Woodward, G. (2011) Impact of simulated drought on ecosystem biomass production: an experimental test in stream mesocosms. Global Change Biology, 17(7), 2288-2297.

Lemly, A.D. (1982) Modification of benthic invertebrate communities in polluted streams: Combined effects of sedimentation and nutrient enrichment. Hydrobiologia, 87, 229-245.

356

Levri, E.P., Luft, R. and Xiaosong, L. (2019) Predator detection and a possible dispersal behaviour of the invasive New Zealand mud snail, Potamopyrgus antipodarum (Gray, 1843). Aquatic Invasions, 14(3), 417-432.

Lewis, K. (1973) The effect of suspended coal particles on the life forms of the aquatic moss Eurhynchium riparioides (Hedw.). Freshwater Biology, 3, 251-257.

Light, T. (2003) Success and failure in a lotic crayfish invasion: The roles of hydrologic variability and habitat alteration. Freshwater Biology, 48(10), 1886-1897.

Lindqvist, O.V., Huner, J.V., Henttonen, P. and Kononen, H. (1999) A comparison of life history strategies and energy reserves of crayfishes occupying permanent and temporary water bodies. Freshwater Crayfish, 12, 449-461.

Lisle, T.E., and Church, M. (2002), Sediment transport-storage relations for degrading, gravel bed channels. Water Resources Research, 38(11), doi: 10.1029/2001WR001086 .

Lloyd, J. (Producer; 2003…2020) Quite Interesting [Television Series]. Television Centre: London.

Losos, J.B., Warheitt, K.I. and Schoener, T.W. (1997) Adaptive differentiation following experimental island colonization in Anolis lizards. Nature, 387, 70-73.

Loughman, Z.J., Welsh, S.A., Sadecky, N.M., Dillard, Z.W. and Scott, R.K. (2016) Evaluation of physiochemical and physical habitat associations for Cambarus callainus (Big Sandy crayfish), an imperilled crayfish endemic to the Central Appalachians. Aquatic Conservation: Marine and Freshwater Ecosystems, 27(4), 755-763.

Ludlam, J.P. and Magoulick, D.D. (2009) Spatial and temporal variation in the effects of fish and crayfish on benthic communities during stream drying. Journal of the North American Benthological Society, 28(2), 371-382.

Ma, T., Sheng, Y., Meng, Y. and Sun, J. (2019) Multistage remediation of heavy metal contaminated river sediments in a mining region based on particle size. Chemosphere, 225, 83- 92.

357

Magbanua, F.S., Townsend, C.R,, Hageman, K.J. and Matthaei, C.D. (2013) Individual and combined effects of fine sediment and the herbicide glyphosate on benthic macroinvertebrates and stream ecosystem function. Freshwater Biology, 58(8), 1729-1744.

Magilligan, F.J., Buraas, E.M. and Renshaw, C.E. (2015) The efficacy of stream power and flow duration on geomorphic responses to catastrophic flooding. Geomorphology, 228, 175- 188.

Magurran, A.E., Seghers, B.H., Carvalho, G.R. and Shaw, P.W. (1992) Behavioural consequences of an artificial introduction of guppies (Poecilia reticulata) in N. Trinidad: evidence for the evolution of anti-predator behaviour in the wild. Proceedings of the Royal Society B, 248(1322), 177-122.

Mahadler, B., Schwartz, J., Palomino, A.M. and Zirkle, J. (2016) Relationship between cohesive soil erosion behaviour and the physical geochemical properties of soil in Tennessee, USA. World Environmental and Water Resources Congress 2016, DOI: 10.1061/9780784479872.036 .

Major, J.J. (2004) Post-eruption suspended sediment transport at Mount St. Helens: Decadal- scale relationships with landscape adjustments and river discharges. Journal of Geophysical Research, 109, 1-22.

Mallapur, A., Miller, C., Christman, M.C. and Estevez, I. (2009) Short-term and long-term movement patterns in confined environments by domestic fowl: Influence of group size and enclosure size. Applied Animal Behaviour Science, 117(1-2), 28-34.

Mao, J.S., Boyce, M.S., Smith, D.W., Singer, F.J., Vales, D.J., Vore, J.M. and Merrill, E.H. (2005) Habitat selection by elk before and after wolf reintroduction in Yellowstone National Park. Journal of Wildlife Management, 69(4), 1691-1707.

Mare, C., Landman, M. and Kerley, G.I.H. (2019) Rocking the landscape: Chacma baboons (Papio ursinus) as zoogeomorphic agents. Geomorphology, 327, 504-510.

Marston, R.A., Girel, J., Pautou, G., Piegay, H., Bravard, J.P. and Arneson, C. (1995) Channel metamorphosis, floodplain disturbance and vegetation development, Ain River, France. Geomorphology, 13, 121–132.

358

Martin, A.J., Rich, T.H., Poore, G.C.B., Shultz, M.B., Austin, C.M., Kool, L. and Vickers- Rich, P. (2008) Fossil evidence in Australia for oldest known freshwater crayfish for Gondwana. Gondwana Research, 14, 287-296.

Martin, B.R. (2011) The Research Excellence Framework and the ‘impact agenda’: are we creating a Frankenstein monster? Research evaluation, 20(3), 247-254.

Martin, L.D. and Bennett, D.K. (1977) The burrows of the Mioscene beaver Palaeocastor, Western Nebraska, U.S.A. Palaeogeography, Palaeoclimatology, Palaeoecology, 22(3), 173- 193.

Mason, R.J., Rice, S.P., Wood, P.J. and Johnson, M. F. (2019). The zoogeomorphology of case‐ building caddisfly: Quantifying sediment use. Earth Surface Processes and Landforms, 44(12), 2510-2525.

Mather, M.E. and Stein, R.A. (1993) Using growth/mortality trade-offs to explore a crayfish species replacement in stream riffles and pools. Canadian Journal of Fisheries and Aquatic Sciences, 50, 88–96.

Mathers, K. (2017) The influence of signal crayfish on fine sediment dynamics and macroinvertebrate communities in lowland rivers. Published PhD Thesis, Loughborough University.

Mathers, K., White, J.C., Fornaldi, R. and Chadd, R. (2020) Flow regimes control the establishment of invasive crayfish and alter their effects on lotic macroinvertebrate communities. Journal of Applied Limnology, DOI: 10.1111/1365-2664.13584.

Mathers, K.L. and Wood, P.J. (2016) Fine sediment deposition and interstitial flow effects on macroinvertebrate community composition within riffle heads and tails. Hydrobiologia, 776(1), 147-160.

Mathers, K.L., Chadd, R.P., Dunbar, M.J., Extence, C.A., Reeds, J., Rice, S.P. and Wood, P.J. (2016) The long-term effects of invasive signal crayfish (Pacifastacus leniusculus) on instream macroinvertebrate communities. Science of the Total Environment, 556, 207-218.

359

Mathers, K.L., Rice, S.P. and Wood, P.J. (2017) Temporal effects of enhanced fine sediment loading on macroinvertebrate community structure and functional traits. Science of the Total Environment, 599-600, 513-522.

Mathers, K.L., Rice, S.P. and Wood, P.J. (2019a) Predator, prey, and substrate interactions: the role of faunal activity and substrate characteristics. Ecosphere, 10(1), 1-18.

Mathers, K.L., Rice, S.P. and Wood, P J. (2019b) Discharge and suspended sediment time series as controls on fine sediment ingress into gravel river beds. Catena, 173, 253-263.

Matsuzaki, S.M., Usio, N., Takamura, N. and Washitani, I. (2009) Contrasting impacts of invasive engineers on freshwater ecosystems: an experiment and meta-analysis. Oecologia, 158(4), 673-686.

Matthaei, C.D., Arbuckle, C.J. and Townsend, C.R. (2000) Stable surface stones as refugia for invertebrates during disturbance in a New Zealand stream. Journal of the North American Benthological Society, 19, 82–93.

Matthaei, C.D., Peacock, K.A. and Townsend, C.R. (1999) Patchy surface stone movement during disturbance in a New Zealand stream and its potential significance for the fauna. Limnology and Oceanography, 44, 1091–1102.

Maude, S.H. and Williams, D.D. (1983) Behaviour of crayfish in water currents: hydrodynamics of eight species with reference to their distribution patterns in southern Ontario. Canadian Journal of Fisheries and Aquatic Sciences, 40, 68-77.

McCarthy, T.S., Ellery, W.N. and Bloem, A. (1998) Some observations on the geomorphological impact of hippopotamus (Hoppopotamus amphibious L.) in the Okavango Delta, Botswana. African Journal of Ecology, 36(1), 44-56.

McCulloch, D.L. (1986) Benthic macroinvertebrate distributions in the riffle-pool communities of two east Texas streams. Hydrobiologia, 135, 61-70.

McKenzie, M., Mathers, K.L., Wood, P.J., England, J., Foster, I., Lawler, D. and Wilkes, M. (2020) Potential physical effects of suspended fine sediment on lotic macroinvertebrates. Hydrobiologia, 847, 697-711.

360

McShane, B.B., Gal, D., Gelman, A., Robert, C. and Tackett, J.L. (2019) Abandon statistical significance. The American Statistician, 73, 235-245.

Melero, Y., Palazon, S. and Lambin, X. (2014) Invasive crayfish reduce food limitation of alien American mink and increase their resilience to control. Oecologia, 174(2), 427-434.

Meysman, F.J.R., Middelburg, J.J. and Heip, C.H.R. (2006) Bioturbation: a fresh look at Darwin’s last idea. Trends in Ecology and Evolution, 21, 688-695.

Micheletti, M., Chandler, J.H. and Lane, S.N. (2014) Investigating the geomorphological potential of freely available and accessible structure-from-motion photogrammetry using a smartphone. Earth Surface Processes and Landforms, 40(4), 473-486.

Micheli, E.R. and Kirchner, J.W. (2002a) Effects of wet meadow riparian vegetation on streambank erosion. 1. Remote sensing measurements of streambank migration and erodibility. Earth Surface Processes and Landforms, 27, 627-639.

Micheli, E.R. and Kirchner, J.W. (2002b) Effects of wet meadow riparian vegetation on streambank erosion. 2. Measurements of vegetated bank strength and consequences for failure mechanics. Earth Surface Processes and Landforms, 27, 687-697.

Middleton, A.D., Kauffman, M.J., McWhirter, D.E., Cook, J.G., Cook, R.C., Nelson, A.A., Jiminez, M.D. and Klaver, R.W. (2013) Animal migration amid shifting patterns of phonology and predation: lessons from a Yellowstone elk herd. Ecology, 94(6), 1245-1256.

Midgley, T.L., Fox, G.A. and Heeren, D.M. (2012) Evaluation of the bank stability and toe erosion model (BSTEM) for predicting lateral retreat on composite streambanks. Geomorphology, 145-146, 107-114.

Miller, G.L. and Ash, S.R. (1988) The oldest freshwater decapod crustacean, from the Triassic of Arizona. Palaeontology, 31(2), 273-279.

Miller, O., Albayrak, I., Nikora, V. O’Hare, M. (2011) Biomechanical properties of aquatic plants and their effects on plant-flow interactions in streams and rivers. Aquatic Sciences, 74(1), 31-44.

361

Milliman, J.D. and Mead, R.H. (1983) World-wide delivery of river sediment to the oceans. The Journal of Geology, 91(1), 1-21.

Mills, D.N. (2019) Ecological impacts of a new invasive species in UK rivers the quagga mussel, Dreissena rostriformis bugensis (bivalve: dreissenidae; Andrusov 1987). Published PhD Thesis: King’s College London.

Molinas, A. and Wu, B. (2001) Transport of sediment in large sand-bed rivers. Journal of Hydraulic Research, 39(2), 135-146.

Monk, W.A., Wood, P.J., Hannah, D.M. and Wilson, D.A. (2006) Selection of river flow indices for the assessment of hydroecological change. River Research and Applications, 23(1), 113-122.

Montague, C.L. (1980) A natural history of temperate Western Atlantic fiddler crabs (genus Uca) with reference to their impact on the salt marsh. Contributions in Marine Science, 23, 25- 55.

Montana Field Guide (2019) Signal Crayfish — Pacifastacus leniusculus. Retrieved from http://fieldguide.mt.gov/, August 22nd 2019.

Montgomery, D.R. (2007) Soil erosion and agricultural sustainability. PNAS, 104(33), doi: 10.1073/pnas.0611508104 .

Montrenko, A., Strijov, V. and Weber, G-W. (2014) Sample size determination for logistic regression. Journal of Computational and Applied Mathematics, 255, 743-752.

Moore, J.W. (2006) Animal Ecosystem Engineers in Streams. BioScience, 56(3), 237-246.

Moorhouse, T.P. and Macdonald, D.W. (2011a) The effect of removal by trapping on body condition in populations of signal crayfish. Biological conservation, 144(6), 1826-1831.

Moorhouse, T.P. and MacDonald, D.W. (2011b) Immigration rates of signal crayfish (Pacifastacus leniusculus) in response to manual control measures. Freshwater Biology, 56, 993-1001.

Mortimer, K., Rowson, R., Mackie, A.S.Y, Clark, P.F., Maslen, C., Smith, A.S. and Harrower, C. (2012) Steep Holm Island, Bristol Channel, UK: evidence of Larus fuscus Linnaeus, 1758

362

(lesser black-backed gull) feeding on the invasive signal crayfish, Pacifastacus leniusculus Dana, 1852. BioInvasions Records, 1(3), 201-208.

Motrenko, A., Strijov, V. and Weber, G.W. (2014). Sample size determination for logistic regression. Journal of Computational and Applied Mathematics, 255, 743-752.

Mowitt, W.P., Houde, E.D., Hinkle, D.C. and Sanford, A. (2006) Growth of planktivorous bay anchovy Anchoa mitchilli, top-down control, and scale-dependence in estuarine mesocosms. Marine Ecology – Progress Series, 308, 255–269.

MSA (2014) Hanford Reach Fall Chinook Redd Monitoring Report for Calendar Year 2014. Report for the U.S. Department of Energy, Report Number HNF-58823. Available at: https://www.hanford.gov/files.cfm/HNF-58823_-_Rev_00.pdf . Access date: 8th May 2020.

Munroe, R. (2015) What If? John Murray: St Ives.

Murgatrody, A.L. and Ternan, J.L. (1983) The impact of afforestation on stream bank erosion and channel form. Earth Surface Processes and Landforms, 8(4), 357-369.

Naciri-Graven, Y. and Goudet, J. (2003) The additive genetic variance after bottlenecks is affected by the number of loci involved in epistatic interactions. Evolution: An International Journal of Organic Evolution, 57, 706-716.

Naden, P.S., Murphy, J.F., Old, G.H., Newman, J., Scarlett, P., Harman, M., Duerdoth, C.P., Hawczak, A., Pretty, J.L., Arnold, A., Laize, C., Hornby, D.D., Collines, A.L., Sear, D.A. and Jones, J.I. (2016). Understanding the controls on deposited fine sediment in the streams of agricultural catchments. Science of the Total Environment, 547, 366-381.

Naismith, J. (Producer; 2006…2020) The Unbelievable Truth [Radio Series]. Shaw Theatre: London.

National Biodiversity Network (2019) National Biodiversity Network web-site. Available at: www.nbn.org.uk . Access date: 3rd December 2019.

National River Flow Archive (2020) National River Flow Archive web-site. Available at: www.nrfa.ceh.ac.uk . Access date: 1st February 2020.

363

Nekrasov, A. (2019) School of fish, Green Humphead Parrotfish - Bolbometopon muricatum feeding on a coral reef, Oceania, Indonesia, Southeast Asia. Available at: https://www.istockphoto.com/gb/video/school-of-fish-green-humphead-parrotfish- bolbometopon-muricatum-feeding-on-a-coral-gm1134737903-301637655 . Access date: 8th May 2020.

Nelson, D. and Sommers, L. (1996) Total Carbon, Organic Carbon, and Organic Matter, in D. Sparks, A. Page, P. Helmke and R. Loeppert (eds.) Methods of Soil Analysis Part 3 – Chemical Methods. Madison: Soil Science Society of America. pp. 61-1010.

Newson, M.D. and Newson, C.L. (2000) Geomorphology, ecology and river channel habitat: mesoscale approaches to basin-scale challenges. Progress in Physical Geography, 24, 195- 217.

Noro, C.K. and Backup, L. (2010) The burrows of Parastacus defossus (Decapoda: Parastacidae), a fossorial freshwater crayfish from southern Brazil. Zoologia, 27, 341-346.

Nuara, M., Hornby, D.D., Collins, A.L., Sear, D.A., Hill, C., Jones, J.I. and Naden, P.S. (2016) Mapping the combined risk of agricultural fine sediment input and accumulation for riverine ecosystems across England and Wales. Ecological Indicators, 70, 209-221.

Nystrom, P., Stenroth, P., Holmqvist, N., Berglund, O., Larsson, P. and Graneli, W. (2006) Crayfish in lakes and streams: individual and population responses to predation, productivity and substratum availability. Freshwater Biology, 51, 2096-2113.

O’Callaghan, P., Jocque, M. and Kelly-Quinn, M. (2015) Nutrient- and sediment-induced macroinvertebrate drift in Honduran cloud forest streams. Hydrobiologia, 758, 75-86.

O’Hare, J.M., O’Hare, M.T., Gurnell, A.M., Dunbar, M.J., Scarlett, P.M. and Laize, C. (2011) Physical constraints on the distribution of macrophytes linked with flow and sediment dynamics in British rivers. River Research and Applications, 27, 671-683.

Olden, J.D. and Poff, N.L. (2003) Redundancy and the choice of hydrologic indices for characterizing streamflow regimes. River Research and Applications, 19(2), 101-121.

Onda, Y. and Itakura, N. (1997). An experimental study on the burrowing activity of river crabs on subsurface water movement and piping erosion. Geomorphology, 20, 279-288.

364

Orlandi, S., Moretti, G. and Albertson, J.D. (2015) Evidence of an emerging levee failure mechanism causing disastrous floods in Italy. Water Resource Research, 51, 7995-8011.

Owens, P.N., Walling, D.E. and Leeks, G.J.L. (2000) Tracing fluvial suspended sediment sources in the catchment of the River Tweed, Scotland, using composite fingerprints and a numerical mixing model. In I. Foster (ed.) Tracers In Geomorphology. Wiley: Chichester. pp. 291–308.

Oxford Mail (2017) Oxford Mail web-site. Available at: https://www.oxfordmail.co.uk/news/15255425.view-magdalen-bridge-oxford-restored- former-glory-just-time-may-morning/. Access date: 12th April 2020.

Ozalp, C., Pinarbasi, A. and Sahin, B. (2010). Experimental measurement of flow past cavities of different shapes. Experimental Thermal and Fluid Science, 34, 505-515.

Paaijmans, K.P., Takken, W., Githeko, A.K. and Jacobs, A.F.G. (2008) The effect of water turbidity on the near-surface water temperature of larval habitats of the malaria mosquito Anopheles gambiae. International Journal of Biometeorology, 52(8), 747-753.

Palmer, J.A., Schilling, K.E., Isenhart, T.M., Schultz, R.C. and Tomer, M.D. (2014) Streambank erosion rates and loads within a single watershed: Bridging the gap between temporal and spatial scales. Geomorphology, 209, 66-78.

Panuccio, M. (2018) Ecological publishing nowadays: the abandonment of binoculars and the raise of the Artificial Intelligence. Avocetta, 42, 55-57.

Parker, G.A. (1974) Assessment strategy and the evaluation of fighting behaviour. Journal of Theoretical Biology, 47, 223-243.

Parker, G., Shimizu, Y., Wilkerson, G.V., Eke, E.C., Abad, J.D., Lauer, J.W., Paola, C., Dietrich, W.E., and Voller, V.R. (2011) A new framework for modelling the migration of meandering rivers. Earth Surface Processes Landforms, 36(1), 70–86.

Parkyn, S.M. (2015) A Review of Current Techniques for Sampling Freshwater Crayfish. In T. Kawai, A. Faulks and G. Scholtz (Eds.) Freshwater Crayfish: A Global Review. CRC Press: Boca Raton. pp. 205-220.

365

Parkyn, S.M., Rabeni, C.F. and Collier, K.J. (1997) Effects of crayfish (Paranephrops planifrons: Parastacidae) on in‐ stream processes and benthic faunas: A density manipulation experiment. New Zealand journal of Marine and Freshwater Research, 31(5), 685-692.

Parr, T.W., Sier, A.R.J., Battarbee, R.W., Mackay, A. and Burgess, J. (2003) Detecting environmental change: science and society—perspectives on long-term re-search and monitoring in the 21st century. Science of the Total Environment, 310(1-3), 1-8.

Parvulescu, L., Zaharia, C., Groza, M-I., Csillik, O., Satmari, A. and Dragut, L. (2016) Flash- flood potential: a proxy for crayfish habitat stability. Ecohydrology, 9(8), 1507-1516.

Pavlou, M., Ambler, G., Seaman, S.R., Guttmann, O., Elliott, P., King, M. and Omar, R.Z. (2015) How to develop a more accurate risk prediction model when there are few events. BMJ, 351:h3868. doi: https://doi.org/10.1136/bmj.h3868.

Peakall, J., Ashworth, P.J. and Best, J.L. (2007) Meander-bend evolution, alluvial architecture, and the role of cohesion in sinuous river channels: a flume study. Journal of Sedimentary Research, 77, 197-212.

Peay, S. (2001) Eradication of alien crayfish populations. R&D Technical Report W1-037/TR1, Scott Wilson Resource Consultants.

Peay, S. and Rogers, D. (1999) The peristaltic spread of signal crayfish (Pacifastacus leniusculus) in the River Wharfe, Yorkshire, England. Freshwater Crayfish, 12, 665-676.

Peay, S., Dunn, A.M., Kunin, W.E., McKimm, R. and Harrod, C. (2015). A method test of the use of electric shock treatment to control invasive signal crayfish in streams. Aquatic Conservation: Marine and Freshwater Ecosystems, 25(6), 874-880.

Peay, S., Hiley, P.D., Collen, P. and Martin, I. (2006) Biocide treatment of ponds in Scotland to eradicate signal crayfish. Bulletin Français de la Pêche et de la Pisciculture, 380-381, 1363- 1379.

Pecorelli, J. (2018) ZSL’s Freshwater Mussel Survey in the Upper Tidal Thames. Available at: https://www.zsl.org/sites/default/files/media/2018- 06/INNS%20Survey%20Report_ZSL%202017_Final_07.02.18_0.pdf. Access date: 14th April 2020.

366

Peixoto, J.M.A., Nelson, B.W. and Wittmann, F. (2009) Spatial and temporal dynamics of river channel migration and vegetation in central Amazonian white-water floodplains by remote- sensing techniques. Remote Sensing of Environment, 113(10), 2258-2266.

Perillo, G.M., Minkoff, D.R. and Piccolo, M.C. (2005) Novel mechanism of stream formation in coastal wetlands by crab–fish–groundwater interaction. Geo-Marine Letters, 25(4), 214-220.

Perry, C.T., Kench, P.S., O’Leary, M.J., Morgan, K.M. and Januchowski-Hartley, F. (2015) Linking reed ecology to island building: Parrotfish identified as major producers of island- building sediment in the Maldives. Geology, 46(6), 503-506.

Petersen, J.E., Cornwell, J.C. and Kemp, W.M. (1999) Implicit scaling in the design of experimental aquatic ecosystems. Oikos, 85(1), 3-18.

Petrusek, A., Filipová, L., Kozubíková-Balcarová, E. and Grandjean, F. (2017) High genetic variation of invasive signal crayfish in Europe reflects multiple introductions and secondary translocations. Freshwater Science, 36(4), 838-850.

Philips, J.D. (2009) Biological energy in landscape evolution. American Journal of Science, 309, 271-290.

Piegay, H. (2019) Quantitative Geomorphology. International Encyclopedia of Geography: People, the Earth, Environment and Technology, doi: 10.1002/9781118786352.wbieg0417.pub2 .

Piegay, H. and Gurnell, A.M. (1997) Large woody debris and river geomorphological pattern: examples from S.E. France and S. England. Geomorphology, 19(1-2), 99-116.

Pimentel, D., McNair, S., Janecka, J., Wightman, J., Simmonds, C., O’Connell, C., Wong, E., Russel, L., Zern, J., Aquino, T. and Tsomondo, T. (2001) Economic and environmental threats of alien plant, animal, and microbe invasions. Agriculture, Ecosystems & Environment, 84(1), 1-20.

Pinto, C., Ing, R., Browning, B., Delboni, V., Wilson, H., Martyn, D. and Harvey, G.L. (2019) Hydromorphological, hydraulic and ecological effects of restored wood: findings and

367 reflections from an academic partnership approach. Water and Environment Journal, 33, 353- 365.

Pintor, L.M., Sih, A. and Bauer, M.L. (2008) Differences in aggression, activity and boldness between native and introduced populations of an invasive crayfish. Oikos, 117, 1629-1636.

Pledger, A.G., Rice, S.P. and Millet, J. (2014) Reduced bed material stability and increased bedload transport caused by foraging fish: a flume study with juvenile Barbel (Barbus barbus). Earth Surface Processes and Landforms, 39(11), 1500-1513.

Pledger, A.G., Rice, S.P. and Millett, J. (2017) Foraging fish as zoogeomorphic agents: an assessment of fish impacts at patch, barform, and reach scales. Journal of Geophysical Research: Earth Surface, 122, 2105-2123.

Pond, C.M. (1975) The role of the “walking legs” in aquatic and terrestrial locomotion of the crayfish Austropotamobius pallipes (Lereboullet). Journal of Experimental Biology, 62, 447- 454.

Porinchu, D.F. and MacDonald, G.M. (2003) The use and application of freshwater midges (Chironomidae: Insecta: Diptera) in geographical research. Progress in Physical Geography, 27(3), 378-422.

Porter, M.L., Perez-Losada, M. and Crandall, K.A. (2005) Model-based multi-locus estimation of decapod phylogeny and divergence times. Molecular Phylogenetics and Evolution, 37(2), 355-369.

Postma, G., Kleinhans, M.G., Meijer, P.T.H. and Eggenhuisen, J.T. (2008). Sediment transport in analogue flume models compared with real-world sedimentary systems: a new look at scaling evolution of sedimentary systems in a flume. Sedimentology, 55(6), 1541–1557.

Price, G.J., Louys, J., Faith, J.T., Lorenzen, E. and Westaway, M.C. (2018). Big data little help in megafauna mysteries. Nature, 558, 23-25.

Punzalan, D., Guiasu, R.C., Belchoir, D. and Dunham, D.W. (2001) Discrimination of conspecific-built chimneys from human-built ones by the burrowing crayfish, Fallicambarus fodiens (Decapoda, Cambaridae). Invertebrate Biology, 120(1), 58-66.

368

Puttock, A., Graham, H.A., Cunliffe, A.M., Elliott, M. and Brazier, R.E. (2017) Eurasian beaver activity increases water storage, attenuates flow and mitigates diffuse pollution from intensively-managed grasslands. Science of the Total Environment, 576, 430-443.

Rabeni, C.F. and Smale, M.A. (1995) Effects of siltation on stream fishes and the potential mitigating role of the buffering riparian zone. Hydrobiologia, 303, 211-219.

Ranta, E. and Lindstrom, K. (1993) Body size and shelter possession in mature signal crayfish, Pacifastacus leniusculus. Annales Zoologici Fennici, 30, 125-132.

Raposeiro, P.M., Ferreira, V., Gea, G. and Goncalves, V. (2018) Contribution of aquatic shredders to leaf litter decomposition in Atlantic island streams depends on shredder density and litter quality. Marine and Freshwater Research, 69(9), 1432-1439.

Ratliff, R. (1985) Meadows in the Sierra Nevada of California: State of Knowledge. General Technical Report PSW-84, United States Department of Agriculture. California: Pacific Southwest Forest and Range Experiment Station.

Redding, J.M., Schreck, C.B. and Everest, F.H. (1987) Physiological effects on coho salmon and steelhead of exposure to suspended solids. Transactions of the American Fisheries Society, 116, 737-744.

Reeve, I.D. (2004) The removal of the North American signal crayfish (Pacifastacus leniusculus) from the River Clyde. Scottish Natural Heritage Commissioned Report, No. 020 (ROAME No. F00LI12).

REFORM (2020) Restoring rivers FOR effective catchment Management web-site. Available at: https://www.reformrivers.eu/gallery/123 . Access date: 13th May 2020.

Reid, A.J., Carlson, A.K., Creed, I.F., Eliason, E.J., Gell, P.A., Johnson, P.T.J., Kidd, K.A., MacCormack, T.J., Olden, J.D., Ormerod, S.J., Smol, J.P., Taylor, W.W., Tockner, K., Vermaire, J.C., Dudgeon, D. and Cooke, S.J. (2019) Emerging threats and persistent conservation challenges for freshwater biodiversity. Biological Reviews, 94(3), 849-873.

369

Reinhardt, L., Jerolmack, D., Cardinale, B.J., Vanacker, V. and Wright, J. (2010) Dynamic interactions of life and its landscape: feedbacks at the interface of geomorphology and ecology. Earth Surface Processes and Landforms, 35, 78-101.

Reynolds, J., Souty-Grosset, C. and Richardson, A. (2013) Ecological Roles of Crayfish in Freshwater and Terrestrial Habitats. Freshwater Crayfish, 19(2), 197-218.

Reynolds, J.D. (2011) A review of ecological interactions between crayfish and fish, indigenous and introduced. Knowledge and Management of Aquatic Ecosystems, 401, 10.

Reznick, D.N. and Ghalambor, C.K. (2001) The population ecology of contemporary adaptations: what empirical studies reveal about the conditions that promote adaptive evolution. Genetica, 112, 183-198.

Ribbens J.C.H. and Graham J.L. (2004) Strategy for the containment and possible eradication of American signal crayfish (Pacifastacus leniusculus) in the River Dee catchment and Skyre Burn catchment, Dumfries and Galloway. Scottish Natural Heritage Commissioned Report No. 014. (ROAME No. F02LK05).

Rice, S.P. and Church, M. (1996) Sampling surficial fluvial gravels; the precision of size distribution percentile sediments. Journal of Sedimentary Research, 66(3), 654-665.

Rice, S.P., Johnson, M.F., Mathers, K., Reeds, J. and Extence, C. (2016) The importance of biotic entrainment for base flow fluvial sediment transport. Journal of Geophysical Research: Earth Surface, 21(5), 890-906.

Rice, S., Pledger, A., Toone, J. and Mathers, K. (2019) Zoogeomorphological behaviours in fish and the potential impact of benthic feeding on bed material mobility in fluvial landscapes. Earth Surface Processes and Landforms, 44(1), 54-66.

Rice, S.P., Johnson, M.F. and Reid, I. (2012) Animals and the Geomorphology of Gravel-bed Rivers. In: M. Church, P. Biron and A. Roy (eds.) Gravel Bed Rivers: Tools, Processes, Environments. Wiley-Blackwell. pp. 49-62.

370

Rice, S.P., Johnson, M.F., Extence, C., Reeds, J. and Longstaff, H. (2014) Diel patterns of suspended sediment flux and the zoogeomorphic agency of invasive crayfish. Cuadernos de Investigacion Geografica, 40(1), 7-27.

Rice, S.P., Lancaster, J.L., and Kemp, P.S. (2010) Experimentation at the interface of fluvial geomorphology, stream ecology and hydraulic engineering and the development of an effective, interdisciplinary river science. Earth Surface Processes and Landforms, 35: 64–77.

Rice, S.P. and Toone, J.A. (2011) Fluvial Audit of the Upper Dove Catchment. Report for Natural England, Bakewell Office.

Richards, C. and Bacon, K.L. (1994) Influence of fine sediment on macroinvertebrate colonization of surface and hyporheic stream substrates. Great Basin Naturalist, 54, 106-113.

Richardson, D.C., Kaplan, L.A., Newbold, J.D. and Aufdenkampe, A.K. (2009) Temporal dynamics of seston: A recurring nighttime peak and seasonal shifts in composition in a stream ecosystem. Limnology and Oceanography, 54(1), 344-354.

Richardson, J.S. and Kiffney, P.M. (2000) Responses of a macroinvertebrate community from a pristine, southern British columbia, Canada, stream to metals in experimental mesocosms. Environmental Toxicology and Chemistry, 19(3), 736-743.

Rinaldi, M. and Casagli, N. (1999) Stability of streambanks formed in partially saturated soils and effects of negative pore water pressures: the Sieve River (Italy). Geomorphology, 26(4), 253-277.

Rinaldi, M. and Darby, S.E. (2007) 9 Modelling river-bank-erosion processes and mass failure mechanisms: progress towards fully coupled simulations. Developments in Earth Surface Processes, 11, 213-239.

Ríos-Saldaña, C.A., Delibes-Mateos, M. and Ferreira, C.C. (2018) Are fieldwork studies being relegated to second place in conservation science?. Global Ecology and Conservation, 14, e00389.

Ripple, W.J. and Beschta, R.L. (2012) Trophic cascades in Yellowstone: The first 15 years after wolf reintroduction. Biological Conservation, 145(1), 205-213.

371

Risse, L.M., Nearing, M.A., Laflen, J.M. and Nicks, A.D. (1993) Error assessment in the universal soil loss equation. Soil Science Society of America Journal, 57(3), 825-833.

Robertson, M.P., Caithness, N. and Villet, M.H. (2009) A PCA-based modelling technique for predicting environmental suitability for organisms from presence records. Diversity and Distributions, 7(1-2), 15-27.

Rosell, F., Bozser, O., Collen, P. and Parker, H. (2005) Ecological impact of beavers Castor fiber and Castor canadensis and their ability to modify ecosystems. Mammal Review, 35, 3-4, 248-276.

Rosewarne, P.J., Svendsen, J.C., Mortimer, R.J.G. and Dunn, A.M. (2014) Muddied waters: suspended sediment impacts on gill structure and aerobic scope in an endangered native and an invasive freshwater crayfish. Hydrobiologia, 722, 61-74.

Rowe, D.K. and Dean, T.L. (1998) Effects of turbidity on the feeding ability of the juvenile migrant stage of six New Zealand freshwater fish species. New Zealand Journal of Marine and Freshwater Research, 32(1), 21-29.

RSPB (2016) Royal Society for the Protection of Birds web-site. Available at: http://www.rspb.org.uk/birds-and-wildlife/bird-and-wildlife-guides/bird-a- z/k/kingfisher/nesting.aspx . [Access date 21st Dec 2016].

Rudnick, D.A., Chan, V. and Resh, V.H. (2005) Morphology and Impacts of the Burrows of the Chinese Mitten Crab, Eriocheir sinensis H. Milne Edwards (Decapoda, Grapsoidea), in South San Francisco Bay, California, U.S.A.. Crustaceana, 78(7), 787-807.

Ryan, P.A. (1991) Environmental effects of sediment on New Zealand streams: a review. New Zealand Journal of Marine and Freshwater Research, 25, 207-221.

Saghaee, G., Mousa, A.A. and Meguid, M.A. (2017) Plausible failure mechanisms of wildlife- damaged earth levees: insights from centrifuge modelling and numerical analysis. Canadian Geotechnics, DOI: 10.1139/cgj-2016-0484.

372

Salehi, M.H., Beni, O.H., Marchegani, H.B., Borujeni, I.E. and Motaghian, H.R. (2011) Refining Soil Organic Matter Determination by Loss-on-Ignition. Pedosphere, 21(4), 473-482.

Salkonen, L., Pursiainen, M. and Tynkkynen, K. (2010) Response to simulated stream velocities by the noble crayfish Astacus astacus (Linnaeus) and the signal crayfish Pacifastacus leniusculus (Dana). Freshwater Crayfish, 17, 201-206.

Sanders, H. (2019) EnvironMental. Barley Twist: Nottingham Comedy Festival, Nottingham.

Sand-Jensen, K.A.J. (1998) Influence of submerged macrophytes on sediment composition and near-bed flow in lowland streams. Freshwater Biology, 39(4), 663-679.

Sandodden, R. and Johnsen, S.I. (2010) Eradication of introduced signal crayfish Pacifastacus leniusculus using the pharmaceutical BETAMAX VET. Aquatic Invasions, 5(1), 75-81.

Schlosser, I.J. and Kallenmeyn, L.W. (2000) Spatial variation in fish assemblages across a beaver-influenced successional landscape. Ecology, 81, 1371-1382.

Schmidt, J.C. and Graf, J.B. (1990) Aggradation and degradation of alluvial sand deposits, 1965 to 1986, Colorado River, Grand Canyon National Park, Arizona. United States Geological Survey, Professional Paper; (USA), 1493.

Schneider, C., Laize, C.L.R., Acreman, M.C. and Florke, M. (2013) How will climate change modify river flow regimes in Europe? Hydrological and Earth System Sciences, 17, 325-339.

Schumacher, B. (2002) Methods for the determination of total organic carbon (TOC) in soils and sediments. Ecological Risk Assessment Support Centre.

Scullion, J., Parish, C.A., Morgan, N. and Edwards, R.W. (1982) Comparisons of benthic macroinvertebrate fauna and substratum composition in riffles and pools in the impounded River Elan and unregulated River Wye, mid-Wales. Freshwater Biology, 12, 579-595.

Sear, D.A., Jones, J.I., Collines, A.L., Hulin, A., Burke, N., Bateman, S., Pattison, I. and Naden, P.S. (2016) Does fine sediment source as well as quantify affect salmonid embryo mortality and development? Science of the Total Environment, 541, 957-968.

373

Servizi, J.A. and Martens, D. (1987) Some effects of suspended Fraser River sediments on sockeye salmon (Oncorhynchus nerka). Canadian Special Publication Fisheries Aquatic Science, 96, 254-264.

Sherriff, S.C., Rowan, J.S., Fenton, O., Jordan, P., Melland, A.R., Mellander, P-E. and O hUallachain, D. (2016) Storm event suspended sediment-discharge hysteresis and controls in agricultural watersheds: Implications for watershed scale sediment management. Environmental Science and Technology, 50(4), 1769-1778.

Shimizu, S.J. and Goldman, C. (1983) Pacifastacus leniusculus (Dana) production in the Sacramento River. Freshwater Crayfish, 5, 210-228.

Sibley, P.J. (2003) Conservation management and legislation the UK experience. Knowledge and Management of Aquatic Ecosystems, 370-371, 209-217.

Sibley, P.J. (2000) Signal crayfish management in the River Wreake catchment. In: D. Rogers and J. Brickland (Eds.) Crayfish Conference Leeds. p.84-96.

Sidorchuk, A.Y. and Golosov, V.N. (2003) erosion and sedimentation on the Russian Plain, II: the history of erosion and sedimentation during the period of intensive agriculture. Hydrological Processes, 17, 3347-3358.

Simberloff, D., Martin, J.L., Genovesi, P., Maris, V., Wardle, D.A., Aronson, J., Courchamp, F., Galil, B., Garcia-Berthou, E., Pascal, M., Pysek, P., Sousa, R., Tabacchi, E. and Vila, M. (2013). Impacts of biological invasions: what's what and the way forward. Trends in Ecology & Evolution, 28(1), 58-66.

Simon, A. and Collison, A.J.C. (2001) Pore-water pressure effects on the detachment of cohesive streambeds: seepage forces and matric suction. Earth Surface Processes and Landforms, 26(13), 1421-1442.

Simon, A., Curini, A., Darby, S.E. and Langendoen, E.J. (2000) Bank and near-bank processes in an incised channel. Geomorphology, 35(3-4), 193-217.

Skelly, D.K. (2002) Experimental venue and estimation of interaction strength. Ecology, 83(8), 2097-2101.

374

Skelly, D.K. and Kiesecker, J.M. (2001) Venue and outcome in ecological experiments: manipulations of larval anurans. Oikos, 94, 1.

Smith, B.P.G., Naden, P.S., Leeks, G.J.L. and Wass, P.D. (2003) The influence of storm events on fine sediment transport, erosion and deposition within a reach of the River Swale, Yorkshire, UK. Science of the Total Environment, 314-316, 451-474.

Smith, G (2018) Step away from stepwise. Journal of Big Data, 5, 32.

Smith, R.M.H. (1987) Helical burrow casts of Therapsid origin from the Beaufort Group (Permian) of South Africa. Palaeogeography, Palaeoclimatology, Palaeoecology, 60, 155- 169.

Smith, S.M. and Green, C.W. (2017) Sediment suspension and elevation loss triggered by Altantic Mud Fiddler Crab (Ura pugnax) Bioturbation in Salt Marsh Dieback Areas of Southern New England. Journal of Coastal Research, 31(1), 88-94.

Soderback, B. (1994) Interactions among juveniles of two freshwater crayfish species and a predatory fish. Oecologia, 100(3), 229-235.

Sofia, G., Masin, R. and Tarolli, P. (2016) Prospects for crowdsourced information on the geomorphic ‘engineering’ by the invasive Coypu (Myocastor coypus). Earth Surface Processes and Landforms, 42(2), 367-377.

Sol, D. and Lefebvre, L. (2003) Behavioural flexibility predicts invasion success in birds introduced to New Zealand. Oikos, 90, 599-605.

Somerset Live (2018) Somerset Live web-site. Available at: https://www.somersetlive.co.uk/in-your-area/aggressive-crayfish-destroying-lakes-stunning- 1533159. Access date: 28th March 2020.

Soulsby, C., Youngson, A.F., Moir, H.J. and Malcolm, I.A. (2001) Fine sediment influence on salmonid spawning habitat in a lowland agricultural stream: a preliminary assessment. The Science of the Total Environment, 265(1-3), 295-307.

Spink, J. and Rowe, J. (2002). The current distribution of signal and native crayfish in the Broadmead Brook, Wiltshire. Freshwater Forum, 19, 3-10.

375

Spivak, A.C., Vanni, M.J. and Mette, E.M. (2011) Moving on up: can results from simple aquatic mesocosm experiments be applied across broad spatial scales? Freshwater Biology, 56, 279-291.

St. Laurent, K. (2016) Fiddler Crabs: From Burrows to Zoea. Wetland Animals. Available at: https://wmap.blogs.delaware.gov/2016/12/09/fiddler-crabs-from-burrows-to-zoea/ . Access date: 8th May 2020.

Stancliffe-Vaughan, A.E. (2015) Sampling UK Pacifastacus leniusculus (Dana, 1852): The effect of trapping on population structure. Published MPhil Thesis, Anglia Ruskin University.

Stanton, J.A. (2004) Burrowing behaviour and movement of the signal crayfish Pacifastacus leniusculus (Dana). Published PhD Thesis, Department of Biology, University of Leicester.

Statzner, B. (2012) Geomorphic implications of engineering bed sediments by lotic animals. Geomorphology, 157-158, 49-65.

Statzner, B. and Peltret, O. (2006) Assessing potential abiotic and biotic complications of crayfish-induced gravel transport in experimental streams. Geomorphology, 74(1-4), 245-256.

Statzner, B. and Sagnes, P. (2008) Crayfish and fish as bioturbators of streambed sediments: Assessing joint effects of species with different mechanistic abilities. Geomorphology, 93, 267- 287.

Statzner, B., Fuchs, U. and Higler, L.W.G. (1996) Sand erosion by mobile predaceous stream insects: Implications for ecology and hydrology. Water Resources Research, 32(7), 2279-2287.

Statzner, B., Gore, J.A. and Resh. V.H. (1988) Hydraulic stream ecology: observed patterns and potential applications. Journal of the North American Benthological Society, 7, 307–360.

Statzner, B., Peltret, O. and Tominova, S. (2003) Crayfish as geomorphic agents and ecosystem engineers: effects of a biomass gradient on baseflow and flood-induced transport of gravel and sand in experimental streams. Freshwater Biology, 48, 147-163.

Statzner, D., Fievet, E., Champagne, J.Y., Morel, R. and Herouin, E. (2000) Crayfish as geomorphic agents and ecosystem engineers: Biological behaviour affects sand and gravel erosion in experimental streams. Limnology and Oceanography, 45(5), 1030-1040.

376

Stebbing, P., Longshaw, M. and Scott, A. (2014) Review of methods for the management of non-indigenous crayfish, with particular reference to Great Britain. Ethology Ecology and Evolution, 26, 204-231.

Steele, C., Butler, L. and Kingsley, D. (2006) The publishing imperative: the pervasive influence of publication metrics. Learned Publishing, 19(4), 277-290.

Steneck, R.S., Graham, M.H., Bourque, B.J., Corbett, D., Erlandson, J.M., Estes, J.A. and Tegner, M.J. (2002) Kelp forest ecosystems: biodiversity, stability, resilience and future. Environmental Conservation, 29(4), 436-459.

Stephens, J.D., Allison, M.A., Di Leonardo, D.R., Weathers III, D.D., Ogston, A.S., McLachlan, R.L., Xing, F. and Meselhe, E.A. (2017) Sand dynamics in the Mekong River channel and export to the coastal ocean. Continental Shelf Research, 147, 38-50.

Stewart, R.I., Dossena, M., Bohan, D.A., Jeppesen, E., Kordas, R.L., Ledger, M.E., Meerhoff, M., Moss, B., Mulder, C., Shurin, J.B., Suttle, B., Thompson, R., Trimmer, M. and Woodward, G. (2013) Chapter Two - Mesocosm Experiments as a Tool for Ecological Climate-Change Research. Advances in Ecological Research, 48, 71-181.

Sutherland, A.B. and Meyer, J.L. (2007) Effects of increased suspended sediment on growth rate and gill condition of two southern Appalachian minnows. Environmental Biology of Fishes, 80, 389-403.

Suttle, K.B., Power, M.E., Levine, J.M. and McNeely, C. (2004). How fine sediment in riverbeds impairs growth and survival of juvenile salmonids. Ecological applications, 14(4), 969-974.

Sweka, J.A. and Hartman, K.J. (2001) Influence of turbidity on brook trout reactive distance and foraging success. Transactions of the American Fisheries Society, 130(1), 138-146.

Sweka, J.A. and Hartman, K.J. (2003) Reduction of reactive distance and foraging success in smallmouth bass, Micropterus dolomieu, exposed to elevated turbidity levels. Environmental Biology of Fishes, 67, 341-347.

377

Telegraph (2016) Invasive crayfish threated Oxford College. The Telegraph, available at: http://www.telegraph.co.uk/education/2016/08/11/invasive-crayfish-threaten-oxford-college/ . [Access date: 22nd Dec 2016].

Telegraph (2019) Telegraph web-site. Available at: https://www.telegraph.co.uk/news/2019/06/17/burrowing-badgers-blamed-catastrophic- flooding-forced-hundreds/. Access date: 12th April 2020.

Tabarestani, M.K. and Zarrati, A.R. (2015) Sediment transport during flood event: a review. International Journal of Environmental Science and Technology, 12(2), 775-788.

Taccari, M.L. and van der Meij, R. (2016a) Study of the effect of burrows of European Badgers (Meles meles) on the initiation of breaching in dikes. FLOODrisk 2016 3rd European Conference on Flood Risk Management. DOI: 10.1051/e3sconf/20160703001.

Taccari, M.L. and van der Meij, R. (2016b) Investigation of the influence of animal burrowing on the failure of the levee of San Matteo along the Secchia river. E3S Web of Conferences, 9, 19001.

Tal, M. and Paola, C. (2007) Dynamic single-thread channels maintained by the interaction of flow and vegetation. Geology, 35(4), 347-350.

Tal, M. and Paola, C. (2010) Effects of vegetation on channel morphodynamics: results and insights from laboratory experiments. Earth Surface Processes and Landforms, 35(9), 1014- 1028.

Talley, T.S., Crooks, J.A. and Levin, L.A. (2001) Habitat utilization and alteration by the invasive burrowing isopod, Sphaeroma quoyanum, in California salt marshes. Marine Biology, 138(3), 561-573.

Theurer, F.D., Lines, I. and Nelson, T. (1985) Interaction between riparian vegetation, water temperature, and salmonid habitat in the Tucannon River. Journal of the American Water Resources Association, 84140.

Thie, J. (2016) Exploring beaver habitat and distribution with Google Earth: The longest beaver dam in the world. Available at: http://www.geostrategis.com/p_beavers-longestdam.htm [Access date 14th Nov 2016].

378

Thoma, R.F. and Armitage, B.J. (2008) Burrowing Crayfish of Indiana. Indiana Department of Natural Resources.

Thompson, J.N. (1998) Rapid evolution as an ecological process. Trends in Ecological Evolution, 13, 329-332.

Thomson, J. R., Taylor, M. P. and Brierley, G. J. (2004). Are River Styles ecologically meaningful? A test of the ecological significance of a geomorphic river characterization scheme. Aquatic Conservation: Marine and Freshwater Ecosystems, 14(1), 25-48.

Thorne C.R. (1981) Field measurements of rates of bank erosion and bank material strength. In Erosion and Sediment Transport Measurement (Proceedings of the Florence Symposium, June 1981). IAHS Publication 133, International Association of Hydrological Sciences: Wallingford; 503–512.

Thorne, C. (1982) Processes and mechanisms of river bank erosion. In: R. Hey, C. Thorne and J. Bathurst (Eds.), Gravel-bed Rivers. Chichester: Wiley and Sons. pp. 227-259.

Thorne, C. (1990) Effects of vegetation on river-bank erosion and stability. In: J. Thornes (Ed.) Vegetation and Erosion. Chichester: Wiley and Sons. pp. 227-259.

Thorne, C.R., Zevenbergen, L.W., Pitlick, J.C., Rais, S., Bradley, J.B. and Julien, P.Y. (1985) Direct measurements of secondary currents in a meandering sand-bed river. Nature, 315, 746- 747.

Toscano, B.J. (2017) Prey behavioural reaction norms: response to threat predicts susceptibility to predation. Animal Behaviour, 132, 147-153.

Tree Works (2020) Tree Works web-site. Available at: http://treeworksguernsey.co.uk/tree- identification/weeping-willow/ . Access date: 13th May 2020.

Twardochleb, L.A. and Olden, J.D. (2013) A global meta-analysis of the ecological impacts of nonnative crayfish. Freshwater Science, 32(4), 1367-1382.

Underwood, T. (1986) The Analysis of Competition by Field Experiments. In J. Kikkawa and D. Anderson (eds.) Community Ecology: Patterns and Processes. Blackwell: Melbourne. p 240-267.

379

Usio, N. (2000) Effects of crayfish on leaf processing and invertebrate colonisation of leaves in a headwater stream: Decoupling of a trophic cascade. Oecologica, 124, 608-614.

Usio, N. and Townsend, C.R. (2004) Roles of crayfish: consequences of predation and bioturbation for stream invertebrates. Ecology, 85(3), 807-822.

Usio, N. and Townsend, C.R. (2000) Distribution of the New Zealand crayfish Paranephrops zealandicus in relation to stream physico‐ chemistry, predatory fish, and invertebrate prey. New Zealand Journal of Marine and Freshwater Research, 3, 557-567.

Usio, N. and Townsend, C.R. (2004) Roles of crayfish: consequences of predation and bioturbation for stream invertebrates. Ecology, 85(3), 807-822.

Vadher, A.N., Stubbington, R. and Wood, P.J. (2015) Fine sediment reduces vertical migrations of Gammarus pulex (Crustacea: Amphipoda) in response to surface water loss. Hydrobiologia, 753(1), 61-71.

Valuska, A.J. and Mench, J.A. (2013) Size does matter: The effect of enclosure size on aggression and affiliation between female New Zealand White rabbits during mixing. Applied animal behaviour science, 149(1-4), 72-76. van de Lageweg, W.I., van Dijk, W.M., Hoendervoogt, R. and Kleinhans, M.G. (2010) Effects of riparian vegetation on experimental channel dynamics. River Flow 2010, 1331-1338. van Dijk, W.M., Teske, R., van de Lageweg, W.I. and Kleihans, M.G. (2013b) Effects of vegetation distribution on experimental river channel dynamics. Water Resources Research, 49(11), 7558-7574. van Dijk, W.M., van de Lageweg, W.I. and Kleinhans, M.G. (2012) Experiemtnal meandering river with chute cutoffs. Journal of Sedimentary Research, 117, F03023. van Dijk, W.M., van de Lageweg, W.I. and Kleinhans, M.G. (2013a) Formation of a cohesive floodplain in a dynamic experimental meandering river. Earth Surface Processes and Landforms, 38, 1550-1565. van Smeden, M., Moons, K.G.M., de Groot, J.A.H., Collins, G.S., Altman, D.G., Eijkemans, J.C. and Reitsma, J.B. (2018) Sample size for binary logistic prediction models: Beyond events

380 per variable criteria. Statistical Methods in Medical Research, 28(8), doi: https://doi.org/10.1177/0962280218784726.

Varricchio, D.J., Martin, A.J. and Katsura, Y. (2007) First trace and body fossil evidence of a burrowing, denning dinosaur. Proceedings of the Royal Society B: Biological Sciences, 274(1616), 1361-1368.

Vaughan, I.P., Diamond, M., Gurnell, A.M., Hall, K.A., Jenkins, A., Milner, N.J., Naylor, L.A., Sear, D.A., Woodward, G. and Ormerod, S.J. (2009) Integrating ecology with hydromorphology: a priority for river science and management. Aquatic Conservation: Marine and Freshwater Ecosystems, 19: 113–125.

Veasey, J.S., Waran, N.K. and Young, R.J. (1996) On comparing the behaviour of zoo housed animals with wild conspecifics as a welfare indicator. Animal Welfare, 5, 13-24.

Vedia, I., Galicia, D., Baquero, E., Oscoz, J. and Miranda, R. (2017) Environmental factors influencing the distribution and abundance of the introduced signal crayfish in the north of Iberian Peninsula. Marine and Freshwater Research, 68(5), 900-908.

Veihe, A., Jensen, N.H., Schiotz, IG.. and Nielsen, S.L. (2010) Magnitude and process of bank erosion at a small stream in Denmark. Hydrological Processes, 25(10), 1597-1613.

Viero, D.P., D’Alpaos, A., Carniello, L. and Define, A. (2013) Mathematical modelling of flooding due to river bank failure. Advances in Water Resources, 59, 82-94.

Vila, M., Basnou, C., Pysek, P., Josefsson, M., Genovesi, P., Gollasch, S., Nentwig, W., Olenin, S., Roques, A., Roy, D., Hulme, P.E. and DAISIE partners (2010) How well do we understand the impacts of alien species on ecosystem services? A pan-European, cross-taxa assessment. Frontiers in Ecology and the Environment, 8(3), 135–144.

Viles, H.A. (1988) Biogeomorphology. Blackwell: Oxford.

Vinodhini, R. and Narayanan, M. (2008) Bioaccumulation of heavy metals in organs of fresh water fish Cyprinus carpio (Common carp). International Journal of Environmental Science and Technology, 5, 179-182.

381

Vorosmarty, C.J., Maybeck, M., Fekete, B., Sharma, K., Green, P. and Syvitski, J.P.M. (2003) Anthropogenic sediment retention: major global impacts from registered river impoundments. Global and Planetary Change, 39, 169-190.

Vu, H.D., Wieski, K., and Pennings, S.C. (2017) Ecosystem engineers drive creek formation in salt marshes. Ecology, 98(1), 162-174.

Wagenhoff, A., Townsend, C.R. and Matthaei, C.D. (2012) Macroinvertebrate responses along broad stressor gradients of deposited fine sediment and dissolved nutrients: a stream mesocosm experiment. Journal of Applied Ecology, 49(4), 892-902.

Walling, D.E. and Amos, C.M. (1999) Source, storage and mobilisation of fine sediment in a chalk stream system. Hydrological Processes, 13(3), 323-340.

Walling, D.E., Owens, P.N. and Leeks, G.J.L. (1999) Fingerprinting suspended sediment sources in the catchment of the River Ouse, Yorkshire, UK. Hydrological Processes, 13, 955- 975.

Walter, K. (2012) An evaluation of whether artificial refuge traps or baited traps are the most effective method for trapping White-clawed crayfish (Austropotamobius pallipes) in the Creedy Yeo River, Devon. The Plymouth Student Scientist, 5(2), 443-485.

Walters, D.M., Leigh, D.S., Freeman, M.C., Freeman, B.J. and Pringle, C.M. (2003). Geomorphology and fish assemblages in a Piedmont river basin, USA. Freshwater Biology, 48(11), 1950-1970.

Webb, P.W. (1979) Mechanics of escape responses in crayfish (Orconectes viridis). Journal of Experimental Biology, 79, 245-263.

Weber, M. (1946) Science as a Vocation. In: A.I. Tauber (Ed) Science and the Quest for Reality. Main Trends of the Modern World. Palgrave Macmillan: London.

Wei, G., Hossain, M.S., Kubec, J., Grabicova, K., Radnak, T., Burcic, M. and Kouba, A. (2020) Psychoactive compounds at environmental concentration alter burrowing behaviour in the freshwater crayfish. Science of the Total Environment, doi: 10.1016/j.scitotenv.2019.135138.

382

Weinländer, M. and Füreder, L. (2009). The continuing spread of Pacifastacus leniusculus in Carinthia (Austria). Knowledge and Management of Aquatic Ecosystems, (394-395), 17.

Weis, J.S. (2010) The role of behaviour in the success of invasive crustaceans. Marine and Freshwater Behaviour and Physiology, 43, 83-98.

Welch, S.M. and Eversole, A.G. (2006) The occurrence of primary burrowing crayfish in terrestrial habitat. Biological Conservation, 130(3), 458-464.

Welsh, S.A. and Loughman, Z.J. (2015) Physical habitat and water quality correlates of crayfish distributions in a mined watershed. Hydrobiologia, 745, 85-96.

West, R. (2010) Non-native crayfish: A new approach to trapping. A review of signal crayfish trapping on the River Lark at Barton Mills, Suffolk from 2001 to 2013. Lark Angling & Preservation Society.

Wheaton, J.M., Gibbins, C., Wainwright, J., Larsen, L. and McElroy, B. (2011). Preface: Multiscale feedbacks in ecogeomorphology. Geomorphology, 126(3-4), 265-268.

Whiles, M.R. and Dodds, W.K. (2001) Relationships between stream size, suspended particles, and filter-feeding macroinvertebrates in a great plains drainage network. Journal of Environmental Quality, 31(5), 1589-1600.

Whitledge, G.W. and Rabeni, C.F. (1996) Diel and seasonal variation in the food habits of crayfishes in a Missouri Ozark stream. Freshwater Crayfish, 11, 159-169.

Whitney, K.D. and Gabler, C.A. (2008) Rapid evolution in introduced species, ‘invasive traits’ and recipient communities: challenges for predicting invasive potential. Diversity and Distributions, 14(4), 569-580.

Whittingham, M.J., Stephens, P.A., Bradbury, R.B. and Freckleton, R.P. (2006) Why do we still use stepwise modelling in ecology and behaviour? Journal of Animal Ecology, 75(5), 1182-1189.

Wilby, R.L., Johnson, M.F. and Toone, J.A. (2012) The Loughborough University TEmperature Network (LUTEN): Rationale and analysis of stream temperature variations. In: R.J. Dawson, C.L. Walsh and C.G. Kilsby (eds.) Earth Systems Engineering 2012: A technical

383 symposium on systems engineering for sustainable adaptation to global change, Centre for Earth Systems Engineering Research, Newcastle University, U.K.

Wilcock, P., Pitlick, J. and Cui, Y. (2009) Sediment transport primer: estimating bed-material transport in gravel-bed rivers. Report for the US Department of Agriculture, Forest Service, Rocky Mountain Research Station. RMRS-GTR-226.

Wilkes, M.A., Gittins, J.R., Mathers, K.L., Mason, R., Casas‐ Mulet, R., Vanzo, D., Mckenzie, M., Murray-Bligh, J., England, J., Gurnell, A. and Jones, J.I. (2019) Physical and biological controls on fine sediment transport and storage in rivers. Wiley Interdisciplinary Reviews: Water, 6(2), e1331.

Wilkinson, B.H. (2005) Humans as geologic agents: A deep-time perspective. Geology, 33(3), 161-164.

Wilkinson, M.T., Richards, P.J. and Humphreys, G.S. (2009) Breaking ground: pedological, geological, and ecological implications of soil bioturbation. Earth Science Review, 97, 257- 272.

Williams, L.R., Taylor, C.M. and Warren, M.L. (2003) Influence of fish predation on assemblage structure of macroinvertebrates in an intermittent stream. Transactions of the American Fisheries Society, 132, 120-130.

Wilsdon, J. (2016). The Metric Tide: Independent Review Of The Role Of Metrics In Research Assessment And Management. Sage: Croydon.

Wilson, C.A., Hughes, Z.J. and FitzGerald, D.M. (2012) The effects of crab bioturbation on Mid-Atlantic saltmarsh tidal creek extension: geotechnical and geochemical changes. Estuarine, Coastal and Shelf Science, 106, 33-44.

Wilson, R.W., Bergman, H.L. and Wood, C.M. (1994) Metabolic costs and physiological consequences of acclimation to aluminium in juvenile rainbow trout (Oncorhynchus mykiss) 2: gill morphology, swimming performance, and aerobic scope. Canadian Journal of Fisheries and Aquatic Sciences, 51, 536-544.

Winterbottom, S.J. and Gilvear, D.J. (1997). Quantification of channel bed morphology in gravel‐ bed rivers using airborne multispectral imagery and aerial photography. Regulated

384

Rivers: Research & Management: An International Journal Devoted to River Research and Management, 13(6), 489-499.

Wolf, E.C., Cooper, D.J. and Hobbs, N.T. (2007) Hydrologic regime and herbivory stabilize an alternative state in Yellowstone National Park. Ecological Applications, 17(6), 1572-1587.

Wolman, M.G. (1954). A method of sampling coarse river‐ bed material. EOS, Transactions American Geophysical Union, 35(6), 951-956.

Wolman, M.G. (1959) Factors influencing erosion of a cohesive river bank. American Journal of Science, 257, 204-216.

Wood, A.L., Simon, A., Downs, P.W. and Thorne, C.R. (2001), Bank‐ toe processes in incised channels: The role of apparent cohesion in the entrainment of failed bank materials. Hydrological Processes, 15(1), 39–61.

Wood, K.A., Hayes, R.B., England, J. and Grey, J. (2017) Invasive crayfish impacts on native fish diet and growth vary with fish life stage. Aquatic Sciences, 79(1), 113-125.

Wood, P.J. and Armitage, P.D. (1997) Biological effects of fine sediment in the lotic environment. Environmental Management, 21(2), 203-217.

Wood, T.C., Kelley, R.E. and Moore, P.A. (2018) Feeding in fear: Indirect effects of predatory fish on macrophyte communities mediated by altered crayfish foraging behaviour. Freshwater Biology, 63, 1523-1533.

Woodruff, D.C. and Varricchio, D.J. (2011). Experimental modeling of a possible Oryctodromeus cubicularis (Dinosauria) burrow. Palaios, 26(3), 140-151.

Worrall, T.P., Dunbar, M.J., Extence, C.A., Laize, C.L.R., Monk, W.A. and Wood, P.J. (2013) The identification of hydrological indices for the characterization of macroinvertebrate community response to flow regime variability. Hydrological Sciences Journal, 59(3-4), 645- 659.

Wright, J.P. and Jones, C.G. (2006) The concept of organisms as ecosystem engineers ten years on: Progress, limitations and challenges. BioScience, 56(3), 203-209.

385

Wright, J.P., Jones, CG.. and Flecker, A.S. (2002) An ecosystem engineer, the beaver, increases species richness at the landscape scale. Ecosystems Ecology, 132, 96-101.

Xin, P., Guangqiu, J., Li, L., and Barry, D.A. (2009). Effects of crab burrows on pore water flows in salt marshes. Advances in Water Resources, 32, 439-449.

Xu, J. (2003) Sedimentation rates in the lower Yellow River over the past 2300 years as influenced by human activities and climate change. Hydrological Processes, 17, 3359-3371.

Yumoto, M., Ogata, T., Natsuoka, N. and Matsumoto, E. (2006) Riverbank freeze-thaw erosion along a small mountain stream, Nikko volcanic area, central Japan. Permafrost and Periglacial Processes, 17(4), 325-339.

Zamor, R.M. and Grossman, G.D. (2007) Turbidity affects foraging success of drift-feeding rosyside dace. Transactions of the American Fisheries Society, 136, 167-176.

Zanatell, B.A. and Peckarsky, B.L. (1996) Stonefield as ecological engineers – hungry predators reduce fine sediments in stream beds. Freshwater Biology, 36(3), 569-577.

Zelnik, I., Gregoric, N. and Tratnik, A. (2018) Diversity of macroinvertebrates positively correlates with diversity of macrophytes in karst ponds. Ecological Engineering, 117, 96-103.

Zhang, Y., Richardson, J.S. and Negishi, J.N. (2004) Detritus processing, ecosystem engineering and benthic diversity: A test of predator-omnivore interference. Journal of Animal Ecology, 73(4), 756-766.

Zuanon, J., Bockmann, F.A. and Sazima, I. (2006) A remarkable sand-dwelling fish assemblage from centra Amazonia, with comments on the evolution of psammophily in South American freshwater fishes. Neotropical Ichthyology, 4(1), 107-118.

386