Thermal and Hydraulic Characteristics of Landscapes And

There are landscapes and riverscapes, both are part of the waterscape.

Mullins Stream, a tributary of the Northwest Miramichi or Elmunokun in Mi’kmaq,

which means ‘a beaver hole’. A very lovely river (photo: June, 2017).

THERMAL AND HYDRAULIC CHARACTERISTICS OF LANDSCAPES AND

THEIR RIVERS

Antóin Mícheál O’Sullivan

Bachelor of Engineering (Honours) in Civil Engineering, 2014

Masters of Science in Engineering Geomorphology, Consultancy, and Practice, 2015

A Dissertation Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

in the Graduate Academic Unit of Forestry and Environmental Management (FOREM)

Supervisor: R. Allen Curry Ph.D., Biology/ FOREM

Examining Board: Tommi Linnansaari, Ph.D. FOREM/Biology Michelle Gray, Ph.D. FOREM David Lentz, Ph.D. Earth Sciences

External Examiner: Christian Torgersen Ph.D., U.S. Geological Survey, Forest and Rangeland Ecosystem Science Center, School of Environmental and Forest Sciences, College of the Environment, University of Washington, Seattle, USA

This dissertation is accepted by the Dean of Graduate Studies

THE UNIVERSITY OF NEW BRUNSWICK

March 2021

Landscapes and riverscapes are hydrologically interconnected through the four dimensions of space and time. While we have some understanding of local scale processes and global scale cycles, at the landscape and riverscape scales we have yet to truly solidify an understanding of interconnectedness. This dissertation examines the hydrological interconnectedness between landscapes and rivers using temperature as a tracer, and the spatial arrangement of aquatic habitats of rivers and wetlands. It was found that geology and specifically bedrock transmissivity exerts an apparent primary control on hydrological processes and river thermal regimes across the New Brunswick region. In areas with high conductance geology, topographic incisions (valleys) generated groundwater discharge which modified river flow and temperature. Inversely, topographic incisions in low conductance geology lacked groundwater discharge, and river flow and temperatures were more tightly coupled to climate. Upland wetlands on a high conductance geologic formation were associated with a lack of groundwater discharge and warm river temperatures, indicating potential recharge points. Winter satellite imagery can accurately delineate groundwater discharges in frozen rivers. Groundwater plume extent was observed to differentiate between summer and winter. High-resolution aerial imagery (30 cm) coupled with field calibration data was effective at modelling river bathymetry and hydraulics across a large catchment (> 1,000 km2). Applying these data to a machine learning algorithm revealed that low flow, aquatic habitats in the Little Southwest

Miramichi broadly follow pool-riffle-slack water sequences.

Earth’s abiotic and biotic systems are tightly interwoven. Changes within each reverberate through space-time moving towards a dynamic equilibrium. While some ii

changes occur at scales that override our ability to observe them, commonalities exist across Earth’s ecosystems, and these fuel adaptation and evolution. My thesis illustrates the interconnectedness of landscapes, wetlands that lie upon them and rivers that flow through them. I conclude by presenting a concept to guide future research direction, ‘the waterscape continuum concept’ (WCC). The continuum unifies surface water, soil and rock moisture, and groundwater. Applying the WCC, scientists, researchers, and practitioners are required to learn about climatic, geologic, and geomorphic processes - inherently leading to better experimental designs, and will reveal new findings of Earth system processes.

iii

DEDICATION

This work is inspired by the minds of William F. Ganong and Thomas C. Winter. Although my mind is light-years behind theirs, viewing through their minds’ lens provides a magnificent window to examine and question, the utter interconnectedness of Earth system processes.

To Emily, my rock.

ACKNOWLEDGEMENTS

The last five years of my life have unquestionably been the most rewarding. I found enormous enjoyment conducting the science that fills these pages. That I ‘work’ towards developing an understanding of the mechanisms that underlie nature is something I still giggle at, and an opportunity I am incredibly grateful for. I thank my supervisor, Prof. R.

Allen Curry. Allen gave me absolute freedom to develop my questions, which is a truly amazing gift. Allen is more than a supervisor/ mentor to me, he has become a 3rd parent.

I am indebted to him more than he may ever understand. It is funny to think a video call from my sister and brother-in-laws’ flat in London (and a promise of a bottle of Jameson for Allen) led me to New Brunswick. My supervisory committee, Dr. Wendy Monk, Prof.

Kerry MacQuarrie and Dr. Daniel Caissie have always given me an ear and advice when needed, for which I am extremely grateful. In particular, I thank Prof. Rick Cunjak, also a member of my supervisory committee, and Dr. Bob Newbury. Both have given me time and friendship, and for that, I am entirely thankful. I extend my thanks to Prof. Kevin

Devito at the University of Alberta, also a member of my supervisory committee. Kevin introduced me to non-linear hydrological systems, challenged and taught me to frame scientific questions, and has become a true friend. I also would like to make special mention of Prof. Tommi Linnansaari and Prof. Kurt Samways. My understanding of ecology, limited as it is, has mostly been syphoned from email exchanges with Tommi,

Kurt, Allen and Rick.

I could not have completed this work without the help of our team. While the team members change each year, there are constants that have helped my work. In particular, I v

thank Chris MacIntyre and Jani Helminen for both their friendship and support. I extend this to Liam MacNeil, Brian Connelly, Troy Robichaud, Ben Andrews, Colin DeCoste,

Craig Wall, Carson White, Allison Pugh, Matt Warner, Abby Culberson, Tyler Lynn,

Michael Duffy, Michael Arsenault, Connor O’Donovan, Jaime Leavitt, Mouhamed Ndong,

Dennis Connor, Thomas Coulson, and Nebai Mesanza, and anyone I’ve left out. Also to financial supporters: NSERC (184846-2011-RGPI), the Collaboration for Atlantic Salmon

Tomorrow (CAST), the Miramichi Salmon Association (MSA) Jack T. H Fenety

Conservation Scholarship, Forestry and Environmental Management – UNB, Bud and

Peggy Bird, Mr. and Mrs. Art Van Slyke, and the CRI Noel Hynes award.

Perhaps the greatest component of the last five years has been the friendships.

There are too many to mention individually, but special thanks to Mark Gautreau, and his family (including Coop), and Meghann Bruce (even including ‘that cat’) who have been my bedrock since 2015. Thank you to Marni Turnbull, Michelle Gray and Faith Sharpe for great chats. And the Linnansaari’s, the Samway’s (including Blitz), Tom Coulson,

Nebai Mesanza, René Malenfant, Tyler Lynn, Bernhard Wegscheider, Jan Niederwestberg,

Michelle Charest (and Mara), Toon Pronk, and Jeff Gaunce, among others for many great memories.

I thank my Mam (Betty), Dad (Tony) and brother (Fionn) as well as the rest of my family back in Ireland. Thanks for always supporting my pursuit on the path less taken, you already know how much you mean to me; and yes, I am now finished with my thesis!

To my sister (Fionnghuala), brother-in-law (Jonny), niece (Saoirse), and nephews vi

(Donnacha and Fiachra) in Cambridge, U.K., thank you for also being there for me. I especially would like to thank the Corey’s. Jane and Lee et al., thank you for inviting me into your home, feeding me, allowing me to finish the writing of this thesis in the incredible

Belleisle Bay, and offering me so much kindness - it seems like you have an endless reserve of it. You have become my 2nd family.

Lastly, and most importantly, I thank my partner in crime, Emily Corey, and our dogs Lux and Nova. To Lux and Nova, thank you for the check ins while I wrote this dissertation (which were entirely self-serving as you were both looking for pats or someone to throw a ball/ stick/ frisbee), for the entertainment you provide through moon walking on certain floors, chasing shadows, for runs in the woods, and being my best mates. To Em, I cannot really quantify how grateful I am that you were by my side throughout this process.

So, I won’t . . .  Thank you for highlighting to me ‘how water moves through landscapes’, for supporting me, and being present in my life. You’ll likely never truly understand how happy you make me. Without you, I could not have done this. I still question why you are with me; but when life gives you a gift, shut your mouth.

To anyone that reads this dissertation, I am sorry 

The keys to the past, present and future are in the ground. So, get diggin’!

vii

Table of Contents

ABSTRACT ...... ii

DEDICATION...... iv

ACKNOWLEDGEMENTS ...... v

List of Tables ...... xiii

List of Figures ...... xvi

Chapter 1: Introduction ...... 1 1.1 General Introduction ...... 1

1.2 Objectives ...... 4

1.3 Dissertation overview ...... 5

1.4 References ...... 7

Chapter 2: The influence of landscape characteristics on the spatial variability of river temperatures ...... 11 Abstract ...... 11

2.1 Introduction ...... 12

2.2 Methods...... 16

2.2.1 Study Area ...... 16

2.2.2 Geospatial Data ...... 22

2.2.3 Statistical analysis ...... 28

2.3 Results ...... 31

2.3.1 Burnthill Brook ...... 31

viii

2.3.2 Clearwater Brook ...... 35

2.3.3 Cains River...... 40

2.4 Discussion ...... 43

2.4.1 Miramichi Highland Watersheds ...... 44

2.4.2 Maritime Lowland Watersheds ...... 47

2.4.3 Interactions of Physiography and Catchment Characteristics ...... 49

2.5 References ...... 50

Chapter 3: Effects of topographic resolution and geologic setting on spatial statistical river temperature models...... 62 Abstract ...... 62

3.1 Introduction ...... 63

3.2 Methods...... 68

3.2.1 Study sites ...... 68

3.2.2 River temperature data – airborne thermal infrared (TIR) ...... 75

3.2.3 Model workflow...... 80

3.2.4 Spatial Statistical Stream Network Model ...... 89

3.3 Results ...... 92

3.3.1 Spatial variability of river temperature ...... 92

3.3.2 VIF analyses...... 95

3.3.3 River temperature model performance ...... 95

3.4 Discussion ...... 106

3.4.1 Spatial statistical river temperature models in complex geologic settings .... 106

3.4.2 Effects of topographic resolution on river temperature models ...... 108

3.4.3 The influence of wetlands on river temperature ...... 110

3.4.4 Predicting fine-scale groundwater thermal refugia ...... 111

3.4.5 Limitations ...... 111

3.5 References ...... 113

Chapter 4: Ice Cover Exists (ICE): A quick method to delineate groundwater inputs in running waters for cold and temperate regions ...... 131 Abstract ...... 131

4.1 Introduction ...... 132

4.2 Methods...... 135

4.2.1 Study Area ...... 135

4.2.2 Thermal Infrared Imagery ...... 138

4.2.3 Satellite imagery ...... 140

4.2.4 Maximum likelihood classification (MLC) ...... 142

4.3 Results ...... 145

4.3.1 Supervised maximum likelihood classification (sMLC) ...... 145

4.3.2 TIR and ICE input detections ...... 145

4.4 Discussion ...... 149

4.4.1 TIR and ICE groundwater detections ...... 149

4.4.2 Limitations ...... 156

4.5 References ...... 158

Chapter 5: Catchment-scale, high-resolution, hydraulic models and habitat maps – a salmonid’s perspective ...... 169 Abstract ...... 169

5.1 Introduction ...... 170

5.2 Methods...... 174

5.2.1 Area of study ...... 174

5.2.2 Data ...... 176

5.3 Modelling Framework ...... 179

5.4 Results ...... 189

5.4.1 Image-derived bathymetry ...... 189

5.4.2 Hydraulic characteristics ...... 191

5.4.3 Hydraulic habitats ...... 192

5.5 Discussion ...... 195

5.5.1 Image-derived depth maps ...... 195

5.5.2 Hydraulic characteristics ...... 196

5.5.3 Habitat classification ...... 197

5.5.4 Limitations ...... 199

5.6 References ...... 200

Chapter 6: Synthesis ...... 214 6.1 Significance of dissertation ...... 214

6.2 Synthisis and future directions ...... 217

6.2.1 Waterscape continuum concept ...... 218

6.3 References ...... 221

Appendix: Effects of topographic resolution and geologic setting on spatial statistical river temperature models...... 224 References ...... 242

Curriculum Vitae

xii

List of Tables

Table 2-1 Physical characteristics and total precipitation and mean air temperature prior

to the TIR measurements for the Cains River and Burnthill Brook watersheds

(March to July 2008), and for the Clearwater Brook watershed (March to July

2009). A single meteorological station borders Burnthill and Clearwater Brook

watersheds (Energy and Resource Development, 2016). 1Harvest occurred

within the 7 years prior to TIR acquisition...... 20

Table 2-2 Landscape characteristics of the tributary watersheds representing the dependent

variables in the tributary river temperature prediction models. For bedrock

geology, surficial geology, and soil classes, n total is the total number of classes

– accumulated across all study sites. Lentic Waterbodies represent lakes, small

ponds, and beaver ponds...... 27

Table 3-1 Landscape-catchment scale parameters used in the river temperature models.

Where $ indicates forest harvest was only considered during the 7 years prior to

thermal infrared (TIR acquisition), as per findings of groundwater thermal

fluxes in this region (see Alexander, 2006)...... 83

Table 3-2 The Cains River temperature models. The aspatial model is a multiple linear

regression model (MLR) and the spatial stream network models (SSN) are the

Euclidean Exponential (E.E) and Spherical (S), TU = tail-up, TD = tail-down, n

= number of river temperature observations; Coeff. = coefficient; Std. Error =

standard error; p = p-values; AIC = Akaike information criterion; CV R2 = leave

one out, cross-validation; and RMSE = root mean squared error. 1the mean value

of the metrics upriver contribution. Selected model is highlighted grey...... 96 xiii

Table 3-3 The North Pole Stream temperature models. The aspatial model is a multiple

linear regression model (MLR) and the spatial stream network models (SSN) are

Euclidean Exponential (E.E), Spherical (S) and Spherical Exponential (S.E.),

TU = tail-up, TD = tail-down, n = number of river temperature observations;

Coeff. = coefficient; Std. Error = standard error; p = p-values; AIC = Akaike

information criterion; CV R2 = leave one out, cross-validation; and RMSE = root

mean squared error. 1the mean value of the metrics upriver contribution, 2mean

air temperature between 15th July – 31 August, 2017. Selected model is

highlighted grey. DWC = deep water clastic bedrock; Precip. = precipitation.

...... 99

Table 4-1 River, date, and resolution of acquired image, along with resampled satellite

image resolution (see text)...... 141

Table 4-2 Summary of TIR and ICE inputs detected and the estimated, total areal extent

of groundwater input zones among study sites (see text)...... 146

Table 5-1 Channel Aspect Ratio (CAR) parameters developed for this study (see Equations

1-7), where mean channel width (n = 12,400 cross-sections) and X mean and

max represent the accumulated mean and maximum X values for n = 12,400

cross-sections. The river’s aspect ratio (A) is taken from Newbury (2015), 풅 =

average reach depth, and do is shallowest depth set to 0 following Legleiter

(2015)...... 184

Table 5-2 Results for CAR and field-calibrated (FC) depth models. The calibration form

and associated parameters are shown, along with statistics for the calculated

depth maps (n = 762)...... 189 xiv

Table 5-3 Summary statistics for arising hydraulic characteristics...... 192

Table 5-4 Summary statistics for the four hydraulic habitat classes and totals for our study

area...... 194

Table A-1 Summary of catchment characteristics ...... 225

Table A-2 Summary of surficial geologic units of the Cains River catchment and the North

Pole Stream catchment...... 226

Table A-3 Two-sample t-tests comparing elevations measured by a LiDAR bare earth

DEM, ‘LiDAR DEM’, and an SRTM DEM ‘SRTM’; and statistical outputs of

a DEM difference calculated by subtracting the LiDAR DEM from the SRTM

denoted by the head ‘DEM Diff.’ ...... 231

Table A-4 Variance inflation factor (VIF) analysis of multicollinearity for final Cains River

model variables ...... 234

Table A-5 Variance inflation factor (VIF) analysis of multicollinearity for final North Pole

Stream model variables ...... 235

List of Figures

Figure 1-1 An illustration of the peak-to-peak amplitude (or ‘range’) of thermal signals

(defined by range of temperature over time) of air temperature (red), surface-

water (black) and a groundwater dominant river (blue). The figure illustrates a

dampened signal of surface-water and groundwater in contrast to air

temperature. Note: This figure does not consider phase shifts which will modify

the time at which temperature changes occur in time for each signal...... 3

Figure 2-1 Topography and physiographic units of New Brunswick (a), and (b) topography

of the Miramichi watershed, where the Cains River watershed is contained

within the Maritime Plains physiographic unit, and the Clearwater Brook and

Burnthill Brook watersheds lie within the Miramichi Highlands physiographic

unit...... 18

Figure 2-2 Delineation of tributary watershed for (a) Clearwater Brook (n = 13); (b)

Burnthill Brook (n = 8); and (c) Cains (n = 14) watersheds, along with tributary

temperature samples sites...... 24

Figure 2-3 (a) Variables of Importance (VIPs) and (b) standardized coefficients (±95%

C.I.) for a Partial Least Squares (PLS) model with two latent vectors predicting

tributary median surface water temperature within the Burnthill Brook

watershed. ‘Solar Radiation (mean)’ is the mean calculated solar radiation

value (Wh/m2) for each tributary, ‘WS Slope (max)’ is the maximum watershed

slope calculated for each tributary watershed, Lentic refers to lentic water

xvi

bodies (lakes and ponds), and, ‘Geological Contacts’ are contact zones between

differing bedrock lithologies...... 32

Figure 2-4 (a) Burnthill Brook Longitudinal Temperature Profile (LTP) where km 0

identifies the headwater reaches, and moving downstream and three warming

reaches are identified between km 0.2 – 3.3. The black line denotes the

mainstem temperature profile, the blue dots indicate inflows, and the orange

dashed line is the inflows trend line. (b) The spatial extent of TIR data and

Burnthill stream network and lentic bodies (in blue) imposed on the watershed

slope of the Burnthill Brook...... 34

Figure 2-5 (a) VIPs and (b) standardized coefficients (±95% C.I.) for a PLS model with

two latent vectors predicting the tributary watershed median surface water

temperature within the Clearwater Brook watershed. Where ‘Solar radiation

(max)’ is the maximum calculated solar radiation value (Wh/m2) for each

tributary; ‘Surficial Geo (GP3)’ is a glaciofluvial plain deposit composed of

gravels, sands and silts, and generally > 1.5 m thick; ‘WS slope (max)’ is the

maximum tributary watershed slope, ‘Current harvest’ is harvested area in the

7 years prior to TIR acquisition, and ‘Sinuosity’ is river sinuosity...... 36

Figure 2-6 Spatial position of glaciofluvial plain (GP3) deposit, generally >1.5 m and

composed of gravel, sand, and minor silt, within the Clearwater Brook

watershed, the accumulated solar radiation across the watershed, and the spatial

position of lentic bodies (BK = brook)...... 37

Figure 2-7 (a) Clearwater Brook Longitudinal Temperature Profile (LTP) where km 0

identifies the base of the headwater basin. The black line denotes the mainstem xvii

temperature profile, the blue dots indicates inflows, and the orange dashed line

is the inflows trend line. (b) Lentic bodies – blue polygon, stream networks,

and extent of TIR imposed on the slope within the Clearwater Brook’s

watershed...... 38

Figure 2-8 Clearwater Brook mainstem temperature gradient, where (a) is a reach with

inputs from tributaries with high solar radiation exposures and lentic bodies

warming in a downstream direction and, (b) is a warming reach where the river

crosses a geological bedrock contact (Middle Ordovician upstream – Middle

Devonian downstream)...... 40

Figure 2-9 (a) VIPs; and (b) standardized coefficients (±95% C.I.) for a PLS model with 1

latent vector predicting the tributary median surface water temperature within

the Cains River watershed. Where ‘WS Elev (max)’ is the maximum watershed

elevation in the sampled tributary; ‘Wetlands’ are area of wetlands upstream of

the observed temperature point; ‘% thru WL’ is the length of river running

through a wetland complex; and ‘Stream Slope (mean)’ is the mean river slope

in the sampled tributary...... 41

Figure 2-10 (a) Cains River LTP where km 0 identifies the headwater reaches, and moving

downstream. The black line denotes the mainstem temperature profile, the blue

dots indicates inflows, and the orange trend line is the linear regression (LR)

for inflows. (b) displays TIR extent and wetland distribution imposed on the

Cains River watershed’s elevation profile...... 42

Figure 3-1 Elevation map of New Brunswick (NB) superimposed with physiographic

regions (a) and bedrock geology also illustrating physiographic regions (b), xviii

where the blue polygon is the Cains River catchment and the red polygon is the

North Pole Stream catchment. Longitudinal profile of the Cains River mainstem

Longitudinal profile of the North Pole Stream mainstem (d) showing elevation

and underlying bedrock (Chain of Rocks/Granite)...... 71

Figure 3-2 (a) The Cains River catchment, (b) a bare earth digital elevation model (DEM),

Figure 3-3 The North Pole Stream catchment (a) boundary with wetlands, (b) a bare earth

digital elevation model (DEM) where the black star denotes a steeply incised

channel, (c) lithology where the black diamond overlays a headwater lake, and

(d) surficial geology...... 75

Figure 3-4 Longitudinal river temperature profiles derived from thermal infrared (TIR)

surveys (see text) for the Cains River – July 2008 (a) and North Pole Stream –

July 2009 (b). The inset figures compare average daily river temperatures

recorded by three and two in-stream temperature loggers (Cains and North Pole

Strem, respectively - i to iii) between July 15 – August 31, 2017 (see text) and

the TIR survey river temperature measurements. Solid black line is the in-

stream, daily temperature (average); dashed line is the overall in-stream,

average daily temperature; red solid line is the measured TIR temperature at the

site. An example illustrating the differences between hydrographic networks

(HN) resolutions (1 km2 and 3 ha) and their corresponding random sampling

points (circles) in the Cains River (c). A fine-scale illustration of discrete inputs

xix

identified by the 3 ha HN, where (i) is a groundwater seep and (ii) is an

unmapped tributary (d)...... 77

Figure 4-1 Little Southwest Miramichi River, NB and its tributary Otter Brook, A – B

(source: GeoNB). A - The reach upstream of the outlet, March 9, 2017. B –

Outlet of Otter Brook discharging into the Little Southwest Miramichi River,

March 9, 2017. C – The spatial extent of the likely Otter Brook plume along

the south-eastern bank of the main river, March 9, 2017...... 137

Figure 4-2 The Miramichi River watershed, NB, Canada, and rivers studied, where: 1 =

North Branch of the Southwest Miramichi River; 2= McKiel Lake; 3 = Burnthill

Brook; 4 = Clearwater Brook; 5 = Little Southwest Miramichi River; 6 =

Southwest Miramichi River (and meteorological station), and 7 = Cains River.

...... 138

Figure 4-3 Three examples of detected groundwater contributing areas showing the aerial,

red-green-blue (RGB) image, corresponding thermal infrared image (TIR), and

winter satellite image (Satellite) with polygons delimiting the contributing

areas. A is a tributary input, B is a backwater channel, or alcove, and C is a

seepage zone...... 140

Figure 4-4 Illustration of the work flow for processing TIR (left) and satellite (right)

images. TIR (1) is airborne-derived TIR image, TIR (2) is the delineated

groundwater inputs based on a boundary condition of <0.5 oC from the

mainstem, and TIR (3) displays the groundwater inputs (red) overlaid on the

aerial image. Satellite (1) is the satellite image where ICE is met. Satellite (2)

is ICE with shadows clipped. Satellite (3) is the delineation of the training data xx

(red) for a supervised, maximum likelihood classification (sMLC) of ice cover.

Satellite (4) is the sMLC-derived classification. Satellite (5) displays the

predicted, groundwater inputs (blue) on the aerial image, where TIR (3) and

ICE (5) are the same image...... 145

Figure 4-5 An example of the groundwater contributing area as delineated by TIR and ICE

in the main Cains River (alcove-derived input). (A) ICE image showing the

extent of a thermal plume during ice cover (3rd April 2014). (B) TIR image

showing the thermal plume during summer (23rd July 2008). (C) The

comparison of the TIR (red) and ICE (blue) estimates of the groundwater input

zones...... 148

Figure 4-6 (A) the spatial extent of a winter thermal plume (as shown in Figure 4-5). (B)

a sample section to demonstrate thermo-mechanical processes occurring at the

groundwater input/ river interface. (C) shows the velocity distribution with

depth considering friction at both the substrate and ice interface and a

theoretical maximum velocity occurring at the mid-point between ice and

substrate. (D) shows advective heat dispersion occurring only in the free

flowing plume. (E) shows solar radiation reflectance (green arrow) and

absorption (red arrow), where more absorption occurs in the open plume (O.P.)

leading to heat gains. (F) shows shortwave absorption (green arrow) in the

plume (but minimized on ice) and longwave emittance (black arrow) from the

plume into the river ice and (G) shows heat gain from the groundwater input to

the frozen river with the magnitude of heat gain illustrated by the size of the

yellow arrow, where the length of the arrow denotes a greater heat transfer, xxi

dampening in a downstream direction with air conductance and evaporation.

...... 153

Figure 4-7 (A) is the spatial extent of a summer thermal plume (as shown in Figure 4-6).

(B) is a sample section to demonstrate thermo-mechanical processes occurring

at the groundwater input/ river interface. (C) shows the velocity distribution

with depth considering friction at the substrate and a theoretical maximum

velocity occurring at just below the water surface. (D) shows advective heat

dispersion occurring across the entire river. (E) shows solar radiation

reflectance (green arrow) and absorption (red arrow) leading to heat gains in

both the free flowing river (F.F.) and O.P. (F) shows shortwave absorption

(green arrow) and longwave emittance (black arrow) across the entire river and

(G) shows heat gain from the groundwater input to the F.F. river with the

magnitude of heat gain illustrated by the size of the yellow arrow, where the

length of the arrow denotes a greater heat transfer, dampening in a downstream

direction with air conductance and evaporation...... 155

Figure 4-8 (A) an aerial image of a steep confined bedrock channel (gold arrows) in the

North Pole Stream – located on the Miramichi Highlands, (B) the slope map of

the associated reach, (C) associated TIR imagery (29 July 2009) draped over a

high-resolution hillshade model with a noted lack of groundwater inputs, and

(D) ICE image (2 February 2017) showing unfrozen sections (red arrows)

which co-occur in the steep confined bedrock channel section...... 157

Figure 5-1 (a) Eastern Canada, where New Brunswick is defined by the black polygon. (b)

The Miramichi River Catchment (black polygon) and the Little Southwest xxii

Miramichi (LSW – red polygon). (c) The LSW (red polygon) and the North

Pole Stream (green polygon) showing 124 km of the mapped, mainstem river

(blue line), sites of the RTK dGPS (black bar), ADCP site (gold bar), the river

gauge (black arrow Environment Canada Hydrograph Station - 01BP001).175

Figure 5-2 A histogram of field collected depth measurements (n = 1,282) via RTK dGPS

and ADCP, n = 298 and n = 984, respectively; where the x-axis = observed

depths, and the y-axis = number of observations, prior to screening for shadows

in the images (a). Histogram for the same field observations post removing sites

that co-occur with shadows (n = 762; n = 280 - RTK GPS, n = 482-ADCP) (b).

...... 178

Figure 5-3 Example of the image processing steps in a selected reach. (a) Original,

unprocessed image. (b) Normalized difference water index (NDWI)

highlighting the river at the blue end of the colour gradient. (c) River polygon

with wetted river extracted – gravels and boulders removed. (d) The final

filtered image with shadows, gravel bars and emergent boulders removed. 180

Figure 5-4 The process of establishing a relationship between the X image quantity and

channel morphology from a filtered image (see text for explanation). (a)

Filtered image, where flow direction is defined by the blue arrow. (b) Band 1

(red) illustrating a dark, pool-like feature with low digital number (DN). (c)

The X image derived from band 1, with a cross-section through a pool-like

feature (red line). (d) The X-values for the cross-section (0 = left bank viewing

upriver) demonstrating the indirect relationship (b1 < 0) for the final depth

calculations...... 185 xxiii

Figure 5-5 (a) CAR model of depths based on n = 762 field observations. The black line

is the model with confidence intervals (C.I., 95 %) for means (dashed lines) and

confidence intervals (C.I., 95 %) for observations (red lines). An example reach

illustrating the CAR depth map (b). Field-calibrated exponential model (n =

762) (c). The black line is the model with confidence intervals (C.I., 95 %) for

means (dashed lines) and confidence intervals (C.I., 95 %) for observations (red

lines). An example reach illustrating the field-calibrated, image-derived depth

map (d)...... 190

Figure 5-6 (a) An example reach with riffles and pools showing the assessed cross-sections

at 10 m centres used as a parameter in the development of the velocity map (b).

...... 192

Figure 5-7 Kernel density plot depicting the relationship between depth (x-axis) and

velocity (y-axis), and habitat classes as defined velocity and depth (n = 21,000

points) (a). An example of a reach (b) and the classified habits (c)...... 194

Figure 6-1 An illustration of the Earth’s structure (redrawn from USGS, 1999),

highlighting the hydrosphere, a region extending from the 640 km in the mantle

to Earth’s atmosphere (a). A simple representation of the Earth’s water cycle

comprising atmospheric/ meteoric water, surface water, rock/soil moisture and

groundwater, evaportranspiration, and water use by humans (c). An illustration

of a waterscape continuum, connecting landscape, flora, and water body (lentic

or lotic) hydrological processes, and its associated flow paths and age of water

as a function of hydraulic conductance and gradient...... 219

xxiv

Figure A-1 Water table ratio values for North America, and the United Kingdom indicating

regions denoting a region that has groundwater regimes that are both connected

(WTR >1) and disconnected (WTR<1) from topography. Regions with strong

connection of topography to groundwater represent areas where spatial

statistical models show strong fits and low RMSPE’s as observed for spatial

statistical river temperature models in the Pacific Northwestern U.S. (Isaak et

al., 2017) and Scotland (Jackson et al., 2017). In comparison, lower spatial

statistical river temperature model fits and higher RMSPE’s are observed where

more heterogeneous WTR exist as observed in north eastern US (Detenbeck et

al., 2016). Note, New Brunswick has heterogeneous geology and large regions

predicted to have groundwater regimes that are both connected (WTR >1) and

large areas that are disconnected from topography. WTR data from Cuthbert et

al., (2019), accessed at https://figshare.com/authors/Mark_Cuthbert/3847054

...... 224

Figure A-2 An overview of riparian zones in the Cains River catchment. (a) an example

of the headwaters of the Cains River, where short wetland vegetation is

dominant via inspection of a high density LiDAR point cloud, and (b) an

example of the lower river section of the Cains, where a riparian zone flanks

the river and an alluvial terrace is present in this lower river section, adjacent to

tall (26 – 30 m) coniferous and deciduous trees...... 228

Figure A-3 An overview of riparian zones in the North Pole Stream catchment. An

example of the headwaters of the North Pole Stream, illustrating streams with

relatively tall vegetation via inspection of a high density LiDAR point cloud xxv

(a). An example of a riparian zone defined by vegetation less than 5 m in height,

which occur in the headwaters and is representative of the wetland vegetation

in the middle reaches (b). An example of LiDAR point cloud detailing tall

vegetation and a confined river section in the lower reaches of the catchment

(c)...... 229

Figure A-4 The Cains River (a) and the North Pole Stream (b) DEM difference where the

LiDAR DEM was subtracted from the SRTM DEM. Areas of disagreement

between the SRTM and LiDAR, defined by the black lines in (a and b), in the

Cains (c) and the North Pole Stream (d) rather than bare earth, where the red

and blue polygons highlight SRTM errors associated false measurements. . 232

Figure A-5 The relationship between measured thermal infrared (TIR) river temperature

on 29th July 2009 in the North Pole Stream and stream distance, where the red

line denotes the mean temperature (a). The relationship between modelled solar

radiation (for a clear sky) at the same locations for the time of flight and stream

distance, where the red line defines the mean solar radiation (b). The

relationship between log normal TIR temperature and log normal solar radiation

where a weak relationship is shown between the 2 variables (c). An example

section of a solar radiation model for the Cains River over 1 day (23rd July 2008

– clear sky conditions) (d)...... 233

Figure A-6 Model plots for observed and predicted temperatures for the Cains River 1 km2

STRM aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN

2: Spherical Euclidean; spherical tail-down (c) models. The associated model

xxvi

predicted standard errors (SE) are presented adjacent to the corresponding

models...... 236

Figure A-7 Model plots for observed and predicted temperatures for the Cains River 1 km2

LiDAR aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN

2: Spherical tail-down (c) models. The associated model predicted standard

errors (SE) are presented adjacent to the corresponding models...... 237

Figure A-8 Model plots for observed and predicted temperatures for the Cains River 3 ha

aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN 2:

Spherical tail-down (c) models. The associated model predicted standard errors

(SE) are presented adjacent to the corresponding models...... 238

Figure A-9 Model plots for observed and predicted temperatures for the North Pole Stream

SRTM 1 km2 aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial

–SSN 2: Spherical tail-down (c) models. The associated model predicted

standard errors (SE) are presented adjacent to the corresponding models. .. 239

Figure A-10 Model plots for observed and predicted temperatures for the North Pole

Stream 1 km2 LiDAR aspatial (a), spatial –SSN 1: Euclidean exponential (b)

and spatial –SSN 2: Spherical tail-down (c) models. The associated model

predicted standard errors (SE) are presented adjacent to the corresponding

models...... 240

Figure A-11 Model plots for observed and predicted temperatures for the North Pole

Stream 3 ha aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial

–SSN 2: Spherical tail-down (c) models. The associated model predicted

standard errors (SE) are presented adjacent to the corresponding models. .. 241 xxvii

Chapter 1: Introduction

1.1 General Introduction

Landscape heterogeneity is characterized by geologic structures (the evolution of the landscape), geomorphic forms (the shape of the landscape), hydrographic features, such as rivers, streams, wetlands, lakes, soil / rock moisture, and groundwater, and the biota that have evolved to these physio-chemical features. Geologic and geomorphic features provide the scaffolding for soil/ rock moisture, groundwater and surface water, and therefore, the foundation for flora and fauna, as water is the life of the biosphere (Ripl,

2003; Cuthbert and Ashley, 2014). Grouped together, these features define ecosystems.

The pathways for the movement of water across and through Earth, or flow paths, are mostly defined by topography, and permeability of the geomaterial (Devito et al., 2005;

McGuire et al., 2005). Heterogeneities of the Earth’s geologic and geomorphic composition, coupled with evaporation, govern flow paths that follow surface, shallow subsurface, or deep subsurface path ways (Dunne, 1980; Winter et al., 1998). These flow paths are also influenced by the Earth’s biota, mostly through transpiration (Condon et al.,

2019; Kirchner et al., 2020). Forests are mass consumers of water (Núñez et al., 2006;

D'Orangeville et al., 2018), while humans also consume enormous amounts of water, whether for drinking, agriculture energy, or industry (Gleeson et al., 2020). Flow paths also determine the age of water, or the resident time, and hydrologic mixing processes across and in the landscape (Soulsby et al., 2000; McGuire et al., 2005). While much knowledge has been gained about the hydrological processes that operate on Earth, much still remains unknown (e.g., Kirchner et al., 2010). Understanding the movements and

behaviour of fresh water has implication for the thermal, chemical and nutrient regimes inherent in surface water, soil/ rock moisture and groundwater bodies, and this ultimately affects biota and landscape evolution (Hynes, 1970; Phillips, 2009).

Temperature exerts a critical control on ecosystem processes. Landscape thermal regimes affect microbial community assemblages (Erny et al., 2015), soil respiration (the production of carbon dioxide by soil biota - Carey et al., 2016), state changes of water

(Arenson et al., 2002), geomorphic processes, such as gelifluction, landslides, and pingo development (Mackay, 1983), and geochemical processes (Sposito et al., 1999).

Riverscape thermal regimes similarly alter physical processes, which affect the spatial and temporal distribution of aquatic denizens (Curry and Devito, 1996; Linnansaari and

Cunjak, 2010). Atlantic Salmon (Salmo salar) for instance, a cold-water ectotherm will seek cold-water source to offset temperature induced physiological stress during periods of elevated river temperatures (see Wilbur et al., 2020; Corey et al., 2019). While the same processes drive thermal regimes of both landscapes and riverscape, i.e., radiation, convection, conduction, advection, the amplitude and period of temperature regimes differs between both. This has implications for groundwater (attributed to landscapes) and surface-water thermal regimes (Kurylyk et al., 2013; Monk et al., 2013). Groundwater, depending on the depth it flows from, will display a dampened signal in contrast to air temperature. Conversely, surface water temperature, dependent on contribution from groundwater and influence by solar radiation, will display a thermal signal that is more aligned with air temperature - this is illustrated in Figure 1.1. Here, a groundwater dominant river, sourced from a deep groundwater flow path, is shown to have a dampened

thermal signal over time in contrast to the air temperature, while the surface water thermal signal is shown to correspond with air temperature (note: No phase shift is displayed here as the purpose is only to illustrate the peak-to-peak amplitude - or ‘range’ - of thermal regimes only). Atlantic Salmon seeking cold-water sources during extreme heat events underlies the importance of understanding the linkages of landscape and riverscape hydrological processes to protect and conserve these critical habitats.

Figure 1-1 An illustration of the peak-to-peak amplitude (or ‘range’) of thermal signals (defined by range of temperature over time) of air temperature (red), surface- water (black) and a groundwater dominant river (blue). The figure illustrates a dampened signal of surface-water and groundwater in contrast to air temperature.

Note: This figure does not consider phase shifts which will modify the time at which temperature changes occur in time for each signal.

Rivers, streams, wetlands and lakes delimit the Earth’s surface water bodies. These water bodies provide habitat for all of Earth’s freshwater biota, from bacteria to apex fishes, 3

mammals, and reptiles (Fausch et al., 2002; Verhougstraete et al., 2015). These waterbodies are a function of past geologic, climatic and biological evolution (Šamonil et al., 2015; Pawlik et al., 2016), and are under a constant state of flux as time denudates and redistributes geologic and biotic material across the riverscape and its flood plain (Leopold et al., 1964; Junk, 1989). Fluvial geomorphic structures within these water bodies define the habitat templates which govern aquatic species spatial distribution, persistence, extinction, adaptation and/or evolution (Cunjak and Green, 1983; Fausch, 1984). As flow regimes flux, the spatial arrangement of fluvial geomorphic features, such as pools, riffles and runs, also flux (Gowan and Fausch, 2002; Wegscheider et al., 2020). During high flows, for instance, a river’s depth and velocity regime may characterise it as a run; however, as flows recede the depth and velocity regime will also change. As the depth decreases higher near-bed velocity gradients will ensue, transposing the run to a riffle

(Newbury and Bates, 2017); hence, aquatic habitats are dynamic. These fluvial geomorphic fluxes influence like history strategies for aquatic species. Atlantic Salmon

(Salmo salar) have been observed to select spawning sites that have particular ratios of depth and velocity, demarcated by Froude numbers – a non-dimensional unit described by gravity, velocity and depth (Moir et al., 1998). Large sea-run Brook Trout (Salvelinus fontinalis) under thermal stress have been noted to select thermal refugia defined not just by temperature, but also by certain depth characteristics (Wilbur et al., 2020).

1.2 Objectives

The goal of my dissertation was to develop an understanding of the hydrological interconnectedness of landscapes and rivers, and the spatial arrangement of riverine 4

hydraulic habitats in a northern temperate landscape. To achieve this goal my objectives were: 1) investigate the influence of landscape features on river hydrological processes; 2) establish the applicability of spatial autocorrelation statistical models to predict river temperature in a post-glacially and tectonically complex landscape; 3) determine if groundwater discharge to river systems remains constant in summer and winter; 4) detail the spatial variability of hydraulic habitats. To address these objectives the following studies were conducted:

 landscape controls on the spatial variability of river temperature (Chapter 2)  effects of topographic resolution and geologic setting on spatial statistical river temperature models (Chapter 3)  thermal infrared (TIR) imagery and winter satellite imagery to map groundwater discharge along rivers (Chapter 4)  construction of high resolution quasi-2D hydraulic habitat maps at the catchment-scale (Chapter 5)

1.3 Dissertation overview

The primary focus of my dissertation is focused on increasing our understanding of hydrological interconnectedness of landscape and rivers including the habitats they create.

To address hydrological interconnectedness and examine the temporal variability of groundwater discharge along rivers, I utilised remote sensing and statistical analyses

(Chapters 2-4). I also looked at the spatial arrangement of hydraulic habitats using remote sensing and field data (Chapter 5).

In CHAPTER 2 (published in CATENA (2019): 10.1016/j.catena.2019.02.006), I examined landscape and solar radiation controls on the spatial variability river temperature

in two distinct physiographic settings. I used TIR, remotely sensed topographic derivatives, geologic, hydrographic and land use maps to predict and link river temperature to landscape structure and form.

In CHAPTER 3 (published in Water Research Resources (2020): doi.org/10.1029/2020WR028122), I focused on the effects of topographic resolution and geologic setting on spatial statistical river temperature models. Here, I used a high density of TIR derived river temperature measurements, coarse and high resolution topographic data, satellite and model derived land use and climate data, respectively, and hydrographic and geologic maps to predict river temperature using spatial statistical river network models. I also examined the utility of developing high resolution river networks (>3 ha) to predict the occurrence of fine-scale groundwater inputs. I hypothesized that spatial statistical river network models would outperform aspatial statistical models, and that the application of high resolution topographic data would increase model performance.

In CHAPTER 4 (published in Hydrological Processes (2019): 10.1002/hyp.13557),

I investigated if groundwater discharge could be delimited in river systems from satellite red-green-blue (RGB) imagery in systems that freeze (in winter). I compared groundwater inputs identified in summer TIR to potential groundwater inputs demarcated in winter imagery. I hypothesized that deeper groundwater sources whose thermal regime was relatively temporally stable, i.e., less influenced by intra and inter-seasonal climatic flux, would not freeze in the winter and thus are detectable in RGB imagery.

In CHAPTER 5 (published in Journal of Ecohydraulics (2020):

10.1080/24705357.2020.1768600), I tested the applicability of developing bathymetric maps without field data, and thereafter the construction of habitat classification maps at the catchment-scale using simple hydraulic equations and a machine learning algorithm. I used high resolution (0.3 m) 4-band imagery, and field collected depth measurements to test and developed models.

In CHAPTER 6, I provide a synthesis of my findings into a general conceptual model linking the interconnectedness between landscapes and rivers from CHAPTERS 2-5. I conclude by presenting future directions of research.

1.4 References

Arenson, L., Hoelzle, M., & Springman, S. (2002). Borehole deformation measurements and internal structure of some rock glaciers in Switzerland. Permafrost and Periglacial Processes, 13(2), 117–135. https://doi.org/10.1002/ppp.414 Carey, J. C., Tang, J., Templer, P. H., Kroeger, K. D., Crowther, T. W., Burton, A. J., … Tietema, A. (2016). Temperature response of soil respiration largely unaltered with experimental warming. Proceedings of the National Academy of Sciences of the United States of America, 113(48), 13797–13802. https://doi.org/10.1073/pnas.1605365113 Condon, L. E., Atchley, A. L., & Maxwell, R. M. (2020). Evapotranspiration depletes groundwater under warming over the contiguous United States. Nature Communications, 11(1), 1–8. https://doi.org/10.1038/s41467-020-14688-0 Corey, E., Linnansaari, T., Dugdale, S. J., Bergeron, N., Gendron, J., Lapointe, M., & Cunjak, R. A. (2019). Comparing the behavioural thermoregulation response to heat stress by Atlantic Salmon parr ( Salmo salar ) in two rivers. Ecology of Freshwater Fish, eff.12487. https://doi.org/10.1111/eff.12487 Cunjak, R. A., & Green, J. M. (1983). Habitat utilization by brook char ( Salvelinus fontinalis ) and rainbow trout ( Salmo gairdneri ) in Newfoundland streams. Canadian Journal of Zoology, 61(6), 1214–1219. https://doi.org/10.1139/z83-165

Curry, R. A., & Devito, K. J. (1996). Hydrogeology of brook trout ( Salvelinus fontinalis) spawning and incubation habitats: implications for forestry and land use development. Canadian Journal of Forest Research, 26(5), 767–772. https://doi.org/10.1139/x26- 086 Cuthbert, M. O., & Ashley, G. M. (2014). A Spring Forward for Hominin Evolution in East Africa. PLoS ONE, 9(9), e107358. https://doi.org/10.1371/journal.pone.0107358 D’Orangeville, L., Houle, D., Duchesne, L., Phillips, R. P., Bergeron, Y., & Kneeshaw, D. (2018). Beneficial effects of climate warming on boreal tree growth may be transitory. Nature Communications, 9(1), 1–10. https://doi.org/10.1038/s41467-018-05705-4 Devito, K., Creed, I., Gan, T., Mendoza, C., Petrone, R., Silins, U., & Smerdon, B. (2005). A framework for broad-scale classification of hydrologic response units on the Boreal Plain: is topography the last thing to consider? Hydrological Processes, 19(8), 1705– 1714. https://doi.org/10.1002/hyp.5881 Dunne, T. (1980). Formation and controls of channel networks. Progress in Physical Geography, 4(2), 211–239. https://doi.org/10.1177/030913338000400204 Erny, D., De Angelis, A. L. H., Jaitin, D., Wieghofer, P., Staszewski, O., David, E., … Prinz, M. (2015). Host microbiota constantly control maturation and function of microglia in the CNS. Nature Neuroscience, 18(7), 965–977. https://doi.org/10.1038/nn.4030 Fausch, K. D. (1984). Profitable stream positions for salmonids: relating specific growth rate to net energy gain. Canadian Journal of Zoology, 62(3), 441–451. https://doi.org/10.1139/z84-067 Fausch, K. D., Torgersen, C. E., Baxter, C. V., & Li, H. W. (2002). Landscapes to Riverscapes: Bridging the Gap between Research and Conservation of Stream Fishes. BioScience, 52(6), 483. https://doi.org/10.1641/0006- 3568(2002)052[0483:ltrbtg]2.0.co;2 Gleeson, T., Cuthbert, M., Ferguson, G., & Perrone, D. (2020, May 1). Global Groundwater Sustainability, Resources, and Systems in the Anthropocene. Annual Review of Earth and Planetary Sciences. Annual Reviews Inc. https://doi.org/10.1146/annurev-earth-071719-055251 Gowan, C., & Fausch, K. D. (2002). Why do foraging stream salmonids move during summer? In Environmental Biology of Fishes (Vol. 64, pp. 139–153). Springer. https://doi.org/10.1023/A:1016010723609 Junk, W.J., Bayley, P.B. and Sparks, R.E., 1989. The flood pulse concept in river- floodplain systems. Canadian special publication of fisheries and aquatic sciences, 106(1), pp.110-127. Kirchner, J. W., Berghuijs, W. R., Allen, S. T., Hrachowitz, M., Hut, R., & Rizzo, D. M. (2020, February 13). Streamflow response to forest management. Nature. Nature Research. https://doi.org/10.1038/s41586-020-1940-6

Kirchner, J. W., Tetzlaff, D., & Soulsby, C. (2010). Comparing chloride and water isotopes as hydrological tracers in two Scottish catchments. Hydrological Processes, 24(12), 1631–1645. https://doi.org/10.1002/hyp.7676 Kurylyk, B. L., Bourque, C. P.-A., & MacQuarrie, K. T. B. (2013). Potential surface temperature and shallow groundwater temperature response to climate change: an example from a small forested catchment in east-central New Brunswick (Canada). Hydrology and Earth System Sciences, 17(7), 2701–2716. https://doi.org/10.5194/hess-17-2701-2013 Leopold, L. B. (Luna B., Wolman, M. G., & Miller, J. P. (John P. (1964). Fluvial processes in geomorphology. Dover Publications. Retrieved from https://pubs.er.usgs.gov/publication/70185663 Linnansaari, T., & Cunjak, R. A. (2010). Patterns in apparent survival of Atlantic Salmon (Salmo salar) parr in relation to variable ice conditions throughout winter. Canadian Journal of Fisheries and Aquatic Sciences, 67(11), 1744–1754. https://doi.org/10.1139/F10-093 Mackay, J. R. (1983). Downward water movement into frozen ground, western arctic coast, Canada. Canadian Journal of Earth Sciences, 20(1), 120–134. https://doi.org/10.1139/e83-012 McGuire, K. J., McDonnell, J. J., Weiler, M., Kendall, C., McGlynn, B. L., Welker, J. M., & Seibert, J. (2005). The role of topography on catchment-scale water residence time. Water Resources Research, 41(5). https://doi.org/10.1029/2004WR003657 Moir, H. J., Soulsby, C., & Youngson, A. (1998). Hydraulic and sedimentary characteristics of habitat utilized by Atlantic Salmon for spawning in the Girnock Burn, Scotland. Fisheries Management and Ecology, 5(3), 241–254. https://doi.org/10.1046/j.1365-2400.1998.00105.x Monk, W. a., Wilbur, N. M., Allen Curry, R., Gagnon, R., & Faux, R. N. (2013). Linking landscape variables to cold water refugia in rivers. Journal of Environmental Management, 118, 170–176. https://doi.org/10.1016/j.jenvman.2012.12.024 Newbury, R. W., & Bates, D. J. (2017). Dynamics of Flowing Water. In Methods in Stream Ecology: Third Edition (Vol. 1, pp. 71–87). Elsevier Inc. https://doi.org/10.1016/B978-0-12-416558-8.00004-4 Núñez, D., Nahuelhual, L., & Oyarzún, C. (2006). Forests and water: The value of native temperate forests in supplying water for human consumption. Ecological Economics, 58(3), 606–616. https://doi.org/10.1016/j.ecolecon.2005.08.010 Pawlik, Ł., Phillips, J. D., & Šamonil, P. (2016, August 1). Roots, rock, and regolith: Biomechanical and biochemical weathering by trees and its impact on hillslopes-A critical literature review. Earth-Science Reviews. Elsevier B.V. https://doi.org/10.1016/j.earscirev.2016.06.002 Phillips, J. D. (2009). Biological energy in landscape evolution. American Journal of Science, 309(4), 271–289. https://doi.org/10.2475/04.2009.01 9

Ripl, W. (2003). Water: the bloodstream of the biosphere. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 358(1440), 1921–1934. https://doi.org/10.1098/rstb.2003.1378 Šamonil, P., Phillips, J., Daněk, P., Beneš, V., & Pawlik, L. (2020). Soil, regolith, and weathered rock: Theoretical concepts and evolution in old-growth temperate forests, Central Europe. Geoderma, 368, 114261. https://doi.org/10.1016/j.geoderma.2020.114261 Soulsby, C., Malcolm, R., Helliwell, R., Ferrier, R. C., & Jenkins, A. (2000). Isotope hydrology of the Allt a’ Mharcaidh catchment, Cairngorms, Scotland: implications for hydrological pathways and residence times. Hydrological Processes, 14(4), 747– 762. https://doi.org/10.1002/(SICI)1099-1085(200003)14:4<747::AID- HYP970>3.0.CO;2-0 Sposito, G., Skipper, N. T., Sutton, R., Park, S. H., Soper, A. K., & Greathouse, J. A. (1999). Surface geochemistry of the clay minerals. Proceedings of the National Academy of Sciences of the United States of America, 96(7), 3358–3364. https://doi.org/10.1073/pnas.96.7.3358 Verhougstraete, M. P., Martin, S. L., Kendall, A. D., Hyndman, D. W., & Rose, J. B. (2015). Linking fecal bacteria in rivers to landscape, geochemical, and hydrologic factors and sources at the basin scale. Proceedings of the National Academy of Sciences of the United States of America, 112(33), 10419–10424. https://doi.org/10.1073/pnas.1415836112 Wegscheider, B., Linnansaari, T., & Curry, R. A. (2020). Mesohabitat modelling in fish ecology: A global synthesis. Fish and Fisheries. https://doi.org/10.1111/faf.12477 Wilbur, N. M., O’Sullivan, A. M., MacQuarrie, K. T. B., Linnansaari, T., & Curry, R. A. (2020). Characterizing physical habitat preferences and thermal refuge occupancy of brook trout ( Salvelinus fontinalis ) and Atlantic Salmon ( Salmo salar ) at high river temperatures. River Research and Applications, rra.3570. https://doi.org/10.1002/rra.3570 Winter, T. C., Harvey, J. W., Franke, O. L., & Alley, W. M. (1998). Groundwater and surface water: A single resource. USGS Publications, 79. https://doi.org/10.3389/fpsyg.2012.00044

Chapter 2: The influence of landscape characteristics on the spatial variability of river temperatures1

Abstract

River temperature is well established as an important variable for aquatic denizens.

Given its importance, researchers are attempting to disentangle complex interactions between landscapes and river temperatures. We investigated how landscape variables of geomorphology, geology, and vegetation can explain a large portion of the spatial variations among tributary temperatures in the large, Miramichi watershed, New

Brunswick. We utilize thermal infrared imagery (TIR) to characterize river temperature and remote sensed data to delineate landscape variables. Partial Least Squares regression

(PLS) models indicated that variability in river temperature was associated with landscape attributes, but these differed with physiography. For the Clearwater Brook and Burnthill

Brook watersheds located on the Miramichi Highlands physiographic unit, solar radiation exposure, surficial geology deposit (comprising gravels, sands, and minor silt), geological contact zones, and maximum watershed slope were strongly and positively correlated with river temperature. We stipulate that the interaction of slope steepness and geologic contact

1 O'Sullivan, A.M., Devito, K.J. and Curry, R.A., 2019. The influence of landscape characteristics on the spatial variability of river temperatures. CATENA, 177, pp.70-83. https://doi.org/10.1016/j.catena.2019.02.006

zones can produce complex local and larger groundwater interactions influencing river temperatures in this physiographic region. In the lower elevation, and lower relief, Cains

River watershed, located on the Maritime Plains physiographic unit, watershed elevation and wetlands were significant and positively correlated with river temperatures. Here the mainstem was cooler in the downstream reaches, likely due to the semi-confined channel interaction with groundwater discharge originating from the surficial deposits and fractured underlying sandstone of the lowlands, and possibly hyporheic processes. These findings illustrate how physiography and geomorphology influence thermal processes in flowing waters at both the landscape and local scales, and the resultant implications for management and conservation efforts both in terms of land use and climate change.

Keywords: Geologic contacts; geomorphology/ physiography; river temperature; surface- groundwater interactions; thermal infrared imagery; wetlands

2.1 Introduction

Water temperature is one of the driving factors for biological processes in river ecosystems (McCullough et al., 2009; Isaak et al., 2017) and often defines critical habitats for biota (Curry and Noakes, 1995; Malcom et al., 2004). Aquatic habitats are facing threats from human-altered landscapes including impacts on river thermal regimes (Steel et al., 2017), and the increasing air temperatures and changes in precipitation patterns with a warming climate (Koirala et al., 2014). For coldwater fishes in shallower rivers, warm summer temperatures increase physiological stress (Breau et al., 2011; Chadwick et al.

2015) and in response, individuals will relocate to colder river sections if such are

accessible (Ebersole et al. 2003; Dugdale et al. 2016). These summer refugia can also provide winter refugia when warmer groundwater discharges to rivers (Cunjak, 1996;

French, 2014). These findings highlight the ecological importance of thermal refugia

(Fullerton et al. 2018), especially for the survival of coldwater species during elevated temperature events in summer at least, and thus the need to develop a better understanding of watershed scale processes governing river temperature (Kurylyk et al., 2013; Isaak et al., 2015).

River flow and temperature regimes are a function myriad processes mostly governed by solar radiation, precipitation, and air temperature (Constantz, 1998; Soulsby et al., 2006), but modified by hydrogeology (Winter et al., 1998), geomorphology

(Dugdale et al., 2015), and landscape features and changes (Tague and Grant, 2004; Detty and McGuire, 2010). Natural landscape scale processes can also affect river temperatures, e.g., wildfires (Isaak et al., 2009). Anthropogenic activities such as forest removal can also impact and alter surface and subsurface flows and effect river temperatures (Curry et al.,

2002; Moore et al., 2005). The magnitude and scale of how these processes influence river temperature is derived from the geomorphology, geology, and climate of a region (Tague et al., 2007; Caissie et al., 2007). A river’s tributaries are typically smaller and often cooler than mainstems (Goniea et al., 2006; Dugdale et al., 2015) because of smaller scale effects that can increase groundwater contributions to baseflow and where the interception of solar radiation by riparian canopy can be important (Caissie, 2006). Smaller river systems are also responsive to thermal fluxes due to hyporheic exchange (Hatch et al., 2006; Westhoff et al., 2011). The presence of lentic water bodies, such as ponds and wetlands, can increase

river temperature via direct solar radiation on open, standing water (Majerova et al., 2015) or decrease temperatures via groundwater exchange (Weber et al., 2017).

Air temperature is traditionally used as an analogue for modelling of river temperature (e.g., Caissie et al., 2007), but it does not adequately predict the heterogeneity of river temperatures (e.g., Isaak et al. 2009). This is most problematic at finer reach scales where the patchiness of temperature patterns can be pronounced (Kurylyk et al., 2013;

Wawrzyniak et al., 2016). Additionally, air temperature cannot account for subsurface advection processes that alter and modify the thermal properties of groundwater, which affects river temperature (Briggs et al., 2017; Johnson et al., 2017). Groundwater flow paths are complicated by the heterogeneous composition of surficial geology and bedrock leading to variability in recharge and discharge through space and time (Soulsby et al.,

2006; Vidon and Smith, 2007). These geologic structures and associated processes can influence river flow and thermal regimes (Constanz, 1998; Larocque et al., 2010), which in turn affect biological processes (McCullough et al., 2009; Fuhrman et al., 2018).

However, researchers are just beginning to address these complexities across broader scales (Torgersen et al., 1999), such as the influence of valley morphology and river sinuosity (Dugdale et al., 2015), and geology (Briggs and Hare, 2018) on river temperature regimes. Landscape frameworks acknowledging surficial geology and bedrock geology

(Winter 2001; Devito et al., 2005) can be used to develop a more robust, conceptual understanding of hydrologic regimes at intra- and inter-watershed scales. These become especially important for understanding the impact of warming climate scenarios where high

frequency, high magnitude weather events push flows and temperatures to extreme highs and lows in running waters (Warren et al., 2011; van Vliet et al., 2013).

Until recently, analyzing the increasingly complex data sets describing spatial and temporal thermal structure of flowing waters has been beyond computing and programming abilities. Contemporaneous advancements in technologies such as

Geographical Information Systems (GIS), Light detection and Ranging (LiDAR), and thermal infrared imagery (TIR), and the development and application of robust multivariate statistical methods such as Partial Least Squared Regression (PLS; Wold et al., 1984) and

Spatial Statistical River Network models (Ver Hoef and Peterson, 2010) are allowing researchers to disentangle these complexities across scales. The result is an improvement of our understanding of a river’s thermal regime and its control of the biological components of the ecosystem (e.g., Petty et al., 2014; Steel et al., 2017), and the consolidation of associations amongst landscape variables across fine to coarse scales (e.g.,

Hrachowitz et al., 2010; Monk et al., 2013).

Studies are emerging that identify summer coldwater patches and the link to geomorphic features (Dugdale et al., 2015; Wawrzyniak et al., 2016). A next step is the investigation of the effects of physiography and geology at a range of scales on river temperature (Curry and Devito, 1996, Briggs et al., 2017). Hydrogeologic studies have established the effect of geological structure on groundwater discharge regimes (James et al., 2000; Briggs et al., 2018). Acknowledging the interconnectedness between the riverscape and the landscape (Hynes, 1975), we undertake an interdisciplinary study across hydrology, hydrogeology, forestry, geology, and geomorphology to examine the potential influence of landscape-to-reach scale variables on the spatial variability of surface water 15

temperatures in a river network. In this study, we examine watershed scale, surface water temperatures in the Miramichi River, New Brunswick, Canada which is a large system spanning multiple geomorphologies (the surface features), physiographic zones (geology and its evolution), and vegetation cover, and which supports intensive forest harvesting.

We extend upon studies by Monk et al. (2013) by comparing most probable landscape variables influencing river temperature in three watersheds in two physiographic units. By continuing to incorporate more components at the scale of the hydrological landscape, e.g., adding physiography and geology to define fine scale hydrological units (HUs) that overlay broader, hydrological response areas (HRA; Devito et al., 2017) of the complete hydraulic system (Winter, 2001), our goal is to improve models that identify the regulating components of a river’s temperature and simultaneously improve decision-making tools for managing land use and adapt to climate change to protect river ecosystems (Steel et al.,

2017; Johnson et al., 2017).

2.2 Methods

2.2.1 Study Area

New Brunswick lies in the Atlantic Maritimes Ecozone. It includes the Canadian

Appalachians underlain by three tectonostratigraphic zones (Williams, 1979). The region’s geological constituents are Pennsylvanian, Silurian, Ordovician, and

Mississippian, and it experienced Quaternary glaciation, isostatic depression and rebound, and submergence and emergence (Rampton et al., 1984). The surficial geology is

comprised of ablation and lodgment till, and associated sand and gravel deposited by Late

Wisconsinian ice or with minor reworking by water (Rampton et al., 1984).

The Miramichi River Watershed drains an area ≈ 14,000 km2 and encompasses two major physiographic units in New Brunswick - the Miramichi Highlands and the New

Brunswick Lowlands (Maritime Plains) (Figure 2-1). During the late Wisconsin glaciation, the entire watershed was covered with glacial ice (e.g., Curry, 2007), which advanced from the west and north creating hills from the Appalachian Mountains (Cunjak and Newbury,

2005). Following isostatic rebound and glacial retreat, meltwaters incised large meandering valleys into the landscape creating terraces and alluvium deposits commonly underlain by glaciofluvial, lacustrine and marine deposits (Rampton et al., 1884; Cunjak and Newbury, 2005). Groundwater discharge to streams and rivers is common (Monk et al. 2013; Kurylyk et al., 2014). Ninety percent of the watershed is covered by forest comprised mainly of softwoods with mixed hardwoods (State of the Environment Report for the Miramichi Watershed, 2007). The dominant land use within the watershed is forestry and that includes a high density of roads.

Figure 2-1 Topography and physiographic units of New Brunswick, and (b) topography of the Miramichi watershed, where the Cains River watershed is contained within the Maritime Plains physiographic unit, and the Clearwater Brook and Burnthill Brook watersheds lie within the Miramichi Highlands physiographic unit.

TIR data were collected for Burnthill Brook and the Cains River on the 23rd of July

2008, whilst the Clearwater Brook TIR dataset was acquired during 29th of July 2009. The accumulative precipitation (rain) in the Clearwater Brook and Burnthill Brook watershed’s

(http://www1.gnb.ca/0079/FireWeather/FireWeatherHourly-e.asp?stn=Clearwater), between March to July 2008 was 359 mm, and between March to July 2009 was 579 mm, whilst the Cains River (Bantalor Meteorological Station,

(http://www1.gnb.ca/0079/FireWeather/FireWeatherHourly-e.asp?stn=Bantalor)

experienced 389 mm rainfall between March to July 2008 (Table 2-1). The climate characteristics of the area are presented in Table 2-2 (Energy and Resource Development,

New Brunswick, 2016).

Clearwater Brook and Burnthill Brook are 5th order rivers draining an area about

325 km2 and 290 km2, respectively, and headwaters in the Miramichi Highlands (Figure 2-

1b). Both watersheds have similar bedrock geology and surficial geology, but differ in spatial distribution (Table 2-1). The watersheds are underlain by deep water clastic, mafic and felsic rock containing contact zones and faults throughout from tectontic movement and orogenic processes (Rampton et al., 1984). The age of lithologies in this region ranges from Cambrian to Late Devonian (540 – 359 million years ago (Mya)), where older rocks are more fractured and weathered than younger rocks (Rampton et al., 1984). The surficial geology comprises Late Wisconsinan (50,000 years before present (YBP)) loamy lodgment till, minor ablation till, hummocky, ribbed and rolling ablation moraines along with

Wisconsinan sand, gravels and minor silts (Rampton et al., 1984). The Cains River is a

5th order river with a watershed about 1,400 km2. It is the only major tributary of the

Miramichi River located entirely within the Maritime Plains of the New Brunswick lowlands physiographic unit (Figure 2-1b; Table 2-1). The Cains is underlain by fluviatile conglomerate and fractured sandstone sequences ranging from the Late Devonian to

Carboniferous (372.2 – 299 Mya) (Rampton et al., 1984). The watershed is blanketed with

Late Wisconsinan sandy till in the upper watershed, and Wisconsinan sand, gravels and silts in the lower watershed. There are bogs, fens, and swamps throughout, with the majority occurring in uplands areas (e.g. Rampton et al., 1984).

Table 2-1 Physical characteristics and total precipitation and mean air temperature prior to the TIR measurements for the Cains River and Burnthill Brook watersheds

(March to July 2008), and for the Clearwater Brook watershed (March to July 2009).

A single meteorological station borders Burnthill and Clearwater Brook watersheds

(Energy and Resource Development, 2016). 1Harvest occurred within the 7 years prior to TIR acquisition.

2.2.2 Geospatial Data

TIR data (wavelength band: 8-9.5 μm) was collected by Quantum Spatial using a thermal camera (FLIR Systems SC3000 LWIR) mounted to a gyro-stabilized unit on the underside of a helicopter (Monk et al., 2013). Approximately 30 km of Burnthill Brook were surveyed, and ≈ 52 km of the Cains River were surveyed in 2008, whilst ≈ 45 km of

Clearwater Brook were surveyed in 2009. Flight paths were flown longitudinal to the river channel, with the camera pointing towards nadir. The flight altitude was chosen to optimise resolution while providing a ground footprint that encompassed both the active channel and immediate floodplain. In-river thermographs (Onset, Hobo UA-002-64, pendant temperature loggers – accuracy 0.5 oC, housed in a white uPVC casing to shield solar radiation) were deployed throughout study areas, collecting data at 10-minute intervals, for calibration and verification of TIR accuracy (n=3 for the Cains – located at 52, 83, and 103 km upstream of the mouth, n=2 for Burnthill Brook – located at 3.5 and 12 km upstream of the mouth, and n=2 for Clearwater Brook – located at 10km, and 25km upstream of the mouth;). The Cains TIR computed radiant temperatures showed agreed with in-river thermographs (error of ±0.1 oC). However, the furthest downstream thermograph displayed a discrepancy of 1.9 oC. Upon inspection of the thermograph and TIR imagery, no source of error could be found. Quantum Spatial (unpublished report) suggest variable weather conditions during the Cains River acquisition, and especially near the downstream reaches, may have contributed to the observed difference. It should be noted that the in- river thermographs were not calibrated prior to deployment, e.g., immersion in an ice bath.

Therefore, it is possible the noted discrepancy between the computed radiant temperatures and the in-river thermograph, may be due to inherent inaccuracies in the thermograph itself, 22

or a combination of both weather conditions, and thermographs errors. There was a discrepancy of ±0.1 oC between TIR and river-based loggers for the Burnthill Brook. There was a divergence of -0.2 to 0.1 oC between TIR and river loggers for Clearwater Brook.

Images were collected every second, resulting in an image overlap of 40 – 70 % (all flights).

Individual frames were then manually geo-rectified by finding a minimum of 6 common ground control points (GCPs) between the image frames and the 1-m Softcopy Ortho- photomap Data Base (SODB) imagery downloaded from GeoNB (Quantum Spatial, unpublished data). The resultant TIR images had a pixel resolution of 60 cm.

To model temperature, we selected a subset of tributary watersheds within each main watershed, which were determined by the ability to detect water temperature at the mouth, e.g., limitations based on the TIR acquired data. These tributary watersheds ranged in size from 2 to 96 km2. At the confluence with the main river, but still within the tributary itself, water temperatures were quantified by querying pixel values (temperature) from the centre of the tributary channel and averaging the value of a 9-point sample to a GIS database (Quantum Spatial, unpublished data). Temperatures were only sampled along what appeared to be, via inspection of 1 m resolution aerial imagery and 60 cm resolution

TIR imagery, surface water flow. Upstream sample distance was dependent on where the major temperature differential was noted between the tributary and mainstem. Wilbur

(2012) demonstrated that the studied reaches were well-mixed and the likelihood of thermal stratification was low, i.e., the TIR extracted temperatures represent the complete water column in these systems. The dependent variable for the temperature models was the tributary’s median temperature and the final number of tributary watersheds sampled 23

were: n=8 for Burnthill Brook; n=14 for Clearwater Brook; and n=13 for Cains River

(Figure 2-2).

Figure 2-2 Delineation of tributary watershed for (a) Clearwater Brook (n=13); (b)

Burnthill Brook (n=8); and (c) Cains (n=14) watersheds, along with tributary temperature samples sites.

To investigate the spatial variability of each rivers’ longitudinal temperature profile

(LTP), the median temperature for each sample image was determined. Similar to the tributary analysis, temperatures were only sampled along what appeared to be, via aerial image and TIR inspection, the main flow channel in the river. This resulted in n=431 median river temperature measurements for Clearwater Brook, n =392 for Burnthill Brook, and n=631 for the Cains River.

Remotely sensed data were utilised to delineate model descriptors (the independent variables), e.g., fluvial geomorphologic features, terrestrial geomorphologic features, and forest inventory data. Watershed drainages and river networks were delineated from a 10 m digital elevation model (DEM), resampled from a 30 m DEM, and derived using ESRI’s

Arc Hydro toolset. To define artificial river networks, the 10 m DEM was reconditioned with a suite of National Hydro Network (NHN) data (the Canadian inland surface waters

GeoBase https://open.canada.ca/data/dataset/a4b190fe-e090-4e6d-881e-b87956c07977).

Reconditioning involves adjusting the DEM by ‘burning’ the vector layer into the raster

DEM. This reconditioning helps mitigate any errors which may be inherent in the DEM by imposing NHN river networks on the DEM (Li, 2014). Once reconditioned, a 1 km2 flow accumulation area was applied to initiate a river (ESRI, 2018). Watershed and river slopes were derived from the 10 m DEM using the Slope function in ArcMap (ESRI, 2018)

– where a slope value for each 10 m pixel was calculated. Geomorphic characteristics were computed using ArcMap (ESRI) zonal statistics and derived from the 10 m DEM, where minimum, maximum, and average of raster data, such as elevation or slope, are computed for an area of interest (Table 2-2). For instance, we characterise watershed slope by minimum, maximum, range, and average slope within the watershed of interest. We use the same method to calculate watershed elevation, aspect, solar radiation (see below), and river slope metrics.

A solar radiation exposure model for each watershed was built using ESRI’s

ArcGIS Area Solar Radiation tool, again using the 10 m re-sampled DEM (Table 2-2). The

10 m DEM was masked with the river network to produce a 10 m pixel size raster. This new river network raster was then utilized to develop the solar radiation model. The solar radiation analysis tool calculates insolation across a landscape based on methods from the hemispherical viewshed algorithm developed by Fu and Rich (2002). The tool was set to

collect total solar insolation on the day of TIR acquisition from 08:00 – 17:00, replicating solar azimuth and altitude on the day of TIR acquisition (Wh/m2).

Wetlands (measured in km2) were delineated by visual inspection of aerial imagery and sourced from GeoNB (http://www.snb.ca/geonb1/e/index-E.asp). River length migrating through wetlands was calculated by clipping the length of river network overlaying the wetland polygon (measured as km). Lentic bodies (km2) were also extracted from the GeoNB data sets. Sinuosity and gradient were derived from the DEM using Dilts

(2015) Stream Gradient and Sinuosity Toolbox (SGT) for ArcGIS

(https://www.arcgis.com/home/item.html?id=c8eb4ce1384e45258ccba1b33cd4e3cb).

The method involves dividing the river network into 200 m river segments based on the spatial distribution of meanders, and then gradient and sinuosity for each segment was calculated both for the artificially derived river network and the DEM it was built from.

Watershed total river gradient was measured in degrees (o) and total watercourse sinuosity was calculated as the ratio of river-line length to straight line length, where 1 is a straight river section and values >1 indicate greater sinuosity (dimensionless). Zonal statistics were calculated for both gradient and sinuosity to obtain average river gradient and sinuosity upstream of the sampled TIR point.

We used provincial datasets (http://www.snb.ca/geonb1/e/DC/catalogue-E.asp ) to examine and define the area’s geologic composition. Amongst the three watersheds, there were seven bedrock geologies units (n=7), nine surficial geology units (n=9), and twenty- three soil units (n=23; Table 2-2). Surficial geology and soil type were delineated as the

total area (km2) of the watershed. Bedrock geology was measured in km2. Faults and geological contacts were quantified by their length (km) passing through the watershed.

The spread of geologic data, and therefore the range of geologic variables, differed across our 3 study sites due to spatially dissimilar depositional modes.

Forest harvest data was obtained from the provincial dataset

(http://www.snb.ca/geonb1/e/DC/catalogue-E.asp). The state of forestry activity in a watershed was assessed based on the forestry applications in the seven (7) years prior to

TIR acquisition in 2008/9 (measured in km2 harvested with no differentiation of harvest type; Table 2-3). We used this time period because Alexander (2006) demonstrated that groundwater temperatures within a clear-cut with a buffer zone >15 m returned to pre-cut conditions after 7-8 years in the Little Southwest Miramichi watershed.

Table 2-2 Landscape characteristics of the tributary watersheds representing the dependent variables in the tributary river temperature prediction models. For bedrock geology, surficial geology, and soil classes, n total is the total number of classes – accumulated across all study sites. Lentic Waterbodies represent lakes, small ponds, and beaver ponds.

Landscape Landscape Metric Data Layer Data Source Characteristic Descriptor Non- Sinuosity dimensional Gradient Degrees (o)` Digital elevation NASA Shuttle Elevation Meters (m) model Radar Topography Fluvial Geomorphology 2 (resampled to 10 m, Mission Solar radiation Wh/m from 30 m) (SRTM) Aspect Degrees (o) River order Strahler

Watershed slope Degrees (o)

Watershed elevation Meters (m) Digital elevation NASA Shuttle Terrestrial model Radar Topography Geomorphology Aspect Degrees (o) (resampled to 10 m, Mission from 30 m) (SRTM) Watershed drainage km2 area Wetlands km2 New Brunswick Hydrography Lentic Waterbodies km2 Hydrographic GeoNB Network Peatlands km2

2 Bedrock (ntotal = 7) km

2 Surficial (ntotal = 9) km

2 Geology Soil (ntotal = 23) km Geological map GeoNB

Contact zones km

Fault lines km Department of Forest Harvest km2 Forest inventory Natural Resources (NB)

2.2.3 Statistical analysis

Partial Least Square Regression (PLS) was selected for modelling because of its ability to handle challenges of smaller data sets, i.e., overfitting models, with highly correlated predictor parameters (e.g., Carrascal et al., 2009, Wold, 2001). PLS combines features from Principal Component Analysis (PCA) and multiple linear regression (MLR)

to predict Y (an observation) from X (a descriptor) (Abdi, 2010), i.e., we predict the median river temperature of a tributary using landscape parameters from its watershed (see Table

- 2). PLS searches for a set of components (or latent vectors) that performs a simultaneous decomposition of X and Y applying the constraint that these latent vectors explain as much of the covariance between X and Y as possible (after Abdi, 2010; Eriksson et al., 1995).

Monk et al. (2013) highlight that multicollinearity issues that are common in environmental data are overcome by reducing descriptors to latent vectors. Resampling was performed to reduce the bias of estimators and evaluate the estimator’s variance. A jackknife technique was chosen over bootstrapping, again owing to the size of the sample size (Chin,

1998). Jackknifing involves iteratively subsampling according to an ascribed deletion number, where n samples are deleted. Pseudo-jackknife values (Ji) are calculated for each n subsample (see Chin, 1998). Then, the mean of these pseudovalues is calculated along with standard deviation and standard error. Finally, a t-statistic with n – 1 degrees of freedom is used to test if the original samples differs from the subsample.

An excellent overview of both PLS and jackknifing techniques is given in Chin

(1998). In our analysis ≈ 21 – 25% of the observations in each tributary watershed model were randomly selected and the jackknife technique was applied (i.e., n=2 (25% of total observations) for Burnthill Brook; n=3 (21% of total observations for Clearwater Brook; and n=3 (23% of total observations) for the Cains River)).

2 2 2 PLS performance is explained by the metrics Qy , Rx and Ry , referring to a measure of global contribution of the components to the predictive quality of the model, a

measure of the explanatory power of the components for the dependent variables of the model, and a measure of the explanatory power of the components for the explanatory variables of the model, respectively. Sharma and Kim (2012) used a Monte Carlo simulation to compare common PLS indices (coefficient of determination (R2) and

2 goodness of fit (GoF), Qy ), in selecting the best model and under differing conditions, e.g., sample size, data distribution. Their results showed that utilisation of R2 and GoF criteria may lead to the selection of highly fitting models, but ones which are unnecessarily complex. They concluded by recommending strongly the use of - which emerged as an equally good selection criterion as Akaike information criterion, (AICu), and Bayesian information criterion (BIC). We qualify model selection using both the more commonly

2 2 found Ry and . is set equal to (1 - 0.95 ) = 0.0975, where a latent vector is kept if

2 its Qy  0.0975 , corresponding to p < 0.05, (e.g. Eriksson et al., 1995; Wold et al., 2001, and Abdi, 2010).

PLS also utilizes a Variables Importance for the Projection (VIP) that measures the importance of an explanatory variable for the building of components. Eriksson et al.

(1995) proposed that highly influential variables are VIP > 1 and moderately influential variables have a VIP > 0.7. Our initial set of potential explanatory variables (where n is dependent on the watershed; see Table 2-2) was reduced by iteratively (minimum 10 iterations) removing variables with p > 0.05, and then re-running until only significant variables are remaining. We created a PLS model with 1 or 2 latent vectors for each of the

watersheds. Data preparation was conducted in Microsoft Excel, and statistical analyses were run in XLStat – Base (XLStat, 2016).

2.3 Results

2.3.1 Burnthill Brook

River Temperature Model The development and extraction of landscape variables for the Burnthill Brook tributary temperature model resulted in a total of n=49 descriptor variables for analysis.

The penultimate tributary median temperature model contained one latent vector and

2 2 2 produced a Qy = 0.635, Rx = 0.59, and Ry = 0.812. The addition of a second orthogonal latent vector increased to 0.703, to 0.869, and to 0.86. The final model’s root- mean-square-error (RMSE) was 0.301 oC, and the mean-square-error (MSE) was 0.091 oC.

Four influential parameters for the tributary’s median river temperature were identified and all parameters were positively correlated with river temperature (Figure 2-3). The mean river solar radiation and watershed maximum slope were predicted to be highly influential

(VIP>1.0), while lentic water bodies and geological contact zones were significant (VIP >

0.7).

Figure 2-3 (a) Variables of Importance (VIPs) and (b) standardized coefficients

(±95% C.I.) for a Partial Least Squares (PLS) model with two latent vectors predicting tributary median surface water temperature within the Burnthill Brook watershed. ‘Solar Radiation (mean)’ is the mean calculated solar radiation value

(Wh/m2) for each tributary, ‘WS Slope (max)’ is the maximum watershed slope calculated for each tributary watershed, Lentic refers to lentic water bodies (lakes and ponds), and, ‘Geological Contacts’ are contact zones between differing bedrock lithologies.

Longitudinal Temperature Profile

The mainstem, Longitudinal Temperature Profile (LTP) for Burnthill Brook was moderately spatially heterogeneous, with surface temperatures ranging from 15.4 – 17.8 oC over ≈ 30km (Figure 2-4). However, the variability in surface temperatures in the upper reach was as great as, or greater than variability over the entire mainstem of the river. A total of 57 inflow seeps, springs, side channels, and unnamed tributaries and the majority of these were not included in the tributary PLS analyses due to DEM resolution - were sampled along the mainstem of Burnthill Brook (Figure 2-4). Assuming a linear relationship between temperature and distance downstream for simplicity, the inflows displayed a general positive trend in temperature in the downstream direction (r = 0.03; p

< 0.05) (Figure 2-4a). The majority of these inflows (41 of 57) occurred within 15 km of the headwaters. The coolest inflow was 10.5 oC and occurred ≈ 1.25 km downstream of km 0 (headwaters), whilst the warmest inflow was 18.6 oC and occurred ≈ 4.2 km downstream from the headwater (Figure 2-4a). The inflow of the South Branch of Burnthill

Brook dropped the surface temperature of the mainstem river from 16.2 to 15.4 oC (km

14.3). Three warming reaches were observed at km 0.25, 1.1, and 2.5 and co-occurr with upland lentic bodies – we expand upon this in the discussion section.

Figure 2-4 (a) Burnthill Brook Longitudinal Temperature Profile (LTP) where km 0 identifies the headwater reaches, and moving downstream and three warming reaches are identified between km 0.2 – 3.3. The black line denotes the mainstem temperature profile, the blue dots indicate inflows, and the orange dashed line is the inflows trend line. (b) The spatial extent of TIR data and Burnthill stream network and lentic bodies (in blue) imposed on the watershed slope of the Burnthill Brook.

2.3.2 Clearwater Brook

River Temperature Model The extraction of landscape descriptors for the Clearwater Brook model predicting median tributary river temperatures resulted in n=51 variables for the PLS analysis. The final model for Clearwater Brook median tributary temperatures identified six influential variables (Figure 2-5), all of which were positively correlated with river temperature. The

2 2 2 model contained two latent vectors and produced a Qy = 0.843, Rx = 0.718, and Ry =

0.955. The model RMSE = 0.286 oC and MSE was 0.082 oC. The maximum river solar radiation, the occurrence of a glaciofluvial deposit (GP3) generally >1.5 m thick

(comprising gravels, sands, and silts), and the watershed’s maximum slope were highly influential VIPs. The spatial arrangement of these highly influential variables is displayed in Figures 2-6 and 2-7. Interestingly, named tributaries with the largest areas of the GP3 deposit, also have the highest levels of solar radiation amongst examined tributaries (see

Figure 2-7). Lentic bodies, geological contacts, and current harvest were moderately influential VIPs. The least significant variable was sinuosity, and had a negative correlation with median tributary temperature.

Figure 2-5 (a) VIPs and (b) standardized coefficients (±95% C.I.) for a PLS model with two latent vectors predicting the tributary watershed median surface water temperature within the Clearwater Brook watershed. Where ‘Solar radiation (max)’ is the maximum calculated solar radiation value (Wh/m2) for each tributary;

‘Surficial Geo (GP3)’ is a glaciofluvial plain deposit composed of gravels, sands and silts, and generally > 1.5 m thick; ‘WS slope (max)’ is the maximum tributary watershed slope, ‘Current harvest’ is harvested area in the 7 years prior to TIR acquisition, and ‘Sinuosity’ is river sinuosity.

Figure 2-6 Spatial position of glaciofluvial plain (GP3) deposit, generally >1.5 m and composed of gravel, sand, and minor silt, within the Clearwater Brook watershed, the accumulated solar radiation across the watershed, and the spatial position of lentic bodies (BK = brook).

Figure 2-7 (a) Clearwater Brook Longitudinal Temperature Profile (LTP) where km

0 identifies the base of the headwater basin. The black line denotes the mainstem temperature profile, the blue dots indicates inflows, and the orange dashed line is the inflows trend line. (b) Lentic bodies – blue polygon, stream networks, and extent of

TIR imposed on the slope within the Clearwater Brook’s watershed.

Longitudinal Temperature Profile

The LTP for the mainstem of Clearwater Brook displayed spatial heterogeneity and a warming trend with downstream distance (Figure 2-7). The coolest reach of the mainstem occurred in the headwaters (18.8 oC), whilst the warmest reach occurred ≈ 35 km downstream (21.1 oC). In addition, 3 warming reaches were observed at km 4.9-7.5, 11.5

– 13 and 27.5-34.5 (Figure 2-7a). Sixty-nine inflows to the mainstem were sampled

(including seeps, springs, side channels, and unnamed tributaries) and ranged from 13.2 –

20.8 oC (Figure 2-7). Similar to the mainstem, the temperatures of the smaller inflows displayed a general positive trend in the downstream direction (r = 0.02; p < 0.05). Further, a trend of increasing watershed slope – in both the mainstem and tributaries – for the lower section of the watershed co-occurs with an increase in water temperatures (see Figure 2-

7b). Additionally, an increase in river temperature (≈ 2.5 oC over a ≈ 4.8 km reach between

≈ 28.5 – 35.3 km) occurs in the lower section of the watershed and overlays a bedrock geological contact (Figure 2-8).

Figure 2-8 Clearwater Brook mainstem temperature gradient, where (a) is a reach with inputs from tributaries with high solar radiation exposures and lentic bodies warming in a downstream direction and, (b) is a warming reach where the river crosses a geological bedrock contact (Middle Ordovician upstream – Middle

Devonian downstream).

2.3.3 Cains River

River Temperature Model The extraction of landscape data for the Cains watershed tributary median temperature model lead to n = 43 landscape descriptors for the analysis. The final model for the Cains Watershed had one latent vector, and identified four influential variables for

2 2 the tributary’s river temperature (Figure 2-9). The models produced a Qy = 0.787, Rx =

2 o o 0.674, and Ry = 0.830, with a model RMSE of 0.752 C and MSE of 0.565 C. All VIP’s were positively correlated with river temperature, except for mean river slope, which was negatively correlated. VIPs > 1 predicted to be highly influential were ‘Maximum Elev’ = the maximum watershed elevation in the sampled tributary; ‘% thru WL’ = the length of river running through a wetland complex, and ‘Wetlands’ = area of wetlands upstream of the observed temperature point, whilst ‘Stream Slope’ = the mean river slope in the sampled tributary, had a 1 < VIP > 0.7.

Figure 2-9 (a) VIPs; and (b) standardized coefficients (±95% C.I.) for a PLS model with 1 latent vector predicting the tributary median surface water temperature within the Cains River watershed. Where ‘WS Elev (max)’ is the maximum watershed elevation in the sampled tributary; ‘Wetlands’ are area of wetlands upstream of the observed temperature point; ‘% thru WL’ is the length of river running through a wetland complex; and ‘Stream Slope (mean)’ is the mean river slope in the sampled tributary.

Longitudinal Temperature Profile

The LTP for the Cains River displayed spatial heterogeneity with temperature ranging from 18.6 – 20.8 oC over ≈ 52 km. In contrast to the Highland watersheds, the coolest reach of the mainstem occurred farthest downstream at 52 km from the headwaters

(Figure 2-10). The warmer surface temperatures observed in the upper reaches of the mainstem were associated with larger percent coverage of wetland and wetlands adjacent river. Similar to the mainstem temperatures, the 99 inflows (ranging from 9.6 – 20.6 oC) 41

entering the mainstem displayed a general negative trend in temperatures moving downstream (r = 0.134; p < 0.05).

Figure 2-10 (a) Cains River LTP where km 0 identifies the headwater reaches, and moving downstream. The black line denotes the mainstem temperature profile, the blue dots indicates inflows, and the orange trend line is the linear regression (LR) for inflows. (b) displays TIR extent and wetland distribution imposed on the Cains

River watershed’s elevation profile.

2.4 Discussion

We assessed probable landscape variables for their potential as predictors of surface water temperatures for headwater tributaries in the Miramichi River, NB, Canada. The variables, or characteristics, identified as significant for river temperatures were solar radiation, lentic and wetland water bodies, relief (slope), with suggestions of influence by overburden coarseness, bedrock permeability, and channel configuration in the landscape. These characteristics show similarities to land surface form and geologic frameworks used to conceive generalized hydrologic landscapes in other geographic regions and climates (Winter, 2001; Cowood et al., 2017). Further, these characteristics can be utilised to define finer scale hydrological units (HUs), such as areas of unique slope, soil, and vegetation, that overlie larger, hydrological response areas (HRA), defined by depth and texture of the geologic substrate to describe variations in surface and subsurface storage, transmissivity, and scale of flow path, for this region - similar to that proposed by

Devito et al., (2017). In our study the Highland and Lowland physiographic units define regions that vary in the proportion and configuration of landscape characteristics (HUs and

HRAs). The use of such HUs and HRAs within the context of larger regional physiographic units and climate can help in understanding regional differences in water movement and aid in better management of water resources.

Our analysis indicated 45 characteristics were not significant for Burnthill Brook median tributary temperatures; 44 characteristics were not significant for Clearwater Brook tributaries, and 39 characteristics were not significant in the Cains River tributaries. The

role of these characteristics influence on river temperature in general cannot be fully assessed without increased synoptic surveys of more tributary catchments that expand the range in landscape characteristics and potential HUs and HRAs to investigate the effects represented by each characteristic through both space and time. Eleven of the characteristics generated significant VIPs for the models predicting river temperatures and likely indicates the dominance of certain surface landforms and geology and their roles of influencing river temperatures in general within these two regional physiographic settings.

These are discussed in the context of potential HUs and HRAs in these two physiographic regions.

2.4.1 Miramichi Highland Watersheds

In both watersheds, solar radiation was a significant positive variable (VIP > 1) in the model projections of river temperatures. This response is predicted because, as Caissie

(2006) describes, heat flux at the air-surface interface is a result of energy exchange mainly through solar radiation or net short-wave radiation, net long-wave radiation, evaporative heat flux (evaporation), and convective heat transfer. The solar energy control of a river’s surface water temperatures is fundamental, but there are additional landscape-scale characteristics in the matrix of factors that control river temperatures (e.g., Johnson 2004;

Garner et al., 2017; and Wawrzyniak et al., 2017)

An additional landscape characteristic of significance for predicting temperatures in the highland tributaries was watershed slope, which was positively correlated with river temperature. The steepest sections, and tributaries, of both studied watersheds occur in their lower reaches (see Figures 2-4 and 2-7). In these sections, it is probable that shallow 44

surficial deposits exist, owing to their geomechanical composition and the reposing angle of the slope (Rampton et al., 1984). These shallow surficial deposits have less storage potential (Winter et al., 1998) and potentially creates near surface flow paths (Dunne,

1978). Shallow flow paths inherently lead to water which is closer to the surface, and will be more susceptible to fluxes in surface air temperature – whether warming in the summer, or cooling in the winter (Kurylyk et al., 2014; Briggs et al., 2017). This can be exasperated during drought conditions when shallow groundwater can disconnect from the river and surface water sources (Devito et al., 1996), thus, leading to river temperature increases.

Coarse textured, glaciofluvial outwash material was also a positive, significant variable predicting temperature in the Clearwater Brook. These deposits are permeable which is usually associated with groundwater discharge or hyporheic processes in rivers

(e.g., Wondzell and Gooseff, 2013; Wawrzyniak et al., 2016) that typically moderate temperature effects in flowing waters (e.g., Curry and Devito, 1996, Kurylyk et al., 2013).

The deposit occurs in tributary watersheds that receive the highest levels of solar radiation exposure and inputs from upland lentic waterbodies within Clearwater Brook (Figure 2-6).

The permeability and valley position of these deposits also suggests an area of recharge concurrent with higher solar radiation and thus possible introduction of warmer water into shallow or local groundwater (Kurylyk et al., 2013) that appears to influence the river.

Lentic waterbodies (lakes, ponds, and beaver ponds) were moderately significant and positively correlated with river temperatures in these highlands (1 > VIP > 0.7). The waterbodies are likely shallow as they sit in the low order portions of these watersheds,

thus they absorb solar radiation and heat in summer and discharge warm water to the river

(e.g., Collen and Gibson 2000). However, beaver dams and ponds, can alter groundwater flow paths and stream hyporheic exchange creating complex effects in receiving rivers

(e.g., Briggs et al., 2013; Weber et al., 2017). A better understanding of how landscape position of lentic body influences their role on river thermal regimes is required (Winter et al., 1998).

Interestingly, geological contact zones were projected as moderately significant and positively correlated with river temperature in both Highland region watersheds. The

Miramichi Highlands are part of the Appalachian orogeny and its basement is heterogeneous with various discontinuities typical of mountainous regions (Caine and

Tomusiak, 2003), comprising older and younger volcanic and sedimentary derived bedrock lithologies (Rampton et al., 1984). At contact zones contrasting bedrock types and permeability can influence groundwater transit time and flow path depth and modify groundwater and river base flow (Hale and McDonnell, 2016; Pfister et al., 2017). Older bedrock can be more fractured and more permeable than younger bedrock (Jefferson et al., 2010) and thus the contact zones in this study may alter gaining and losing characteristics for groundwater and thus affect the river’s thermal regime (e.g., Winter et al., 1998; Saar, 2011). This process may explain the increase in river temperature noted along the mainstem of Clearwater Brook. Here, the transition from a Middle Ordovician

(≈ 465 million years ago (Mya)) granite deposit to a Middle Devonian (≈ 387.5 Mya) granite deposit underlies a 2.5 oC increase in river temperature along ≈ 5 km reach (Figure

2-8).

Two characteristics of minor significance that influence river temperatures, and only in Clearwater Brook tributary watersheds, were percent coverage of recent harvesting and stream sinuosity. Harvesting may be significant in the Clearwater Brook, because this was the watershed with the most harvesting and potentially experienced reductions in interception and transpiration. Increased temperatures could be attributed to harvest activities in areas with shallow surface deposits that promote shallow water table depths and increases in temperatures of surface and subsurface water on route to the river

(Alexander, 2006; Moore et al., 2005; Smerdon et al., 2009). Also noteworthy is the lack of influence of river sinuosity (meanders) in this study as moderation of river temperatures

- decrease in summer and increase in winter- have been associated with the frequency of meanders that protrude into groundwater flow fields or promote hyporheic exchange flows

(HEF) in other studies (Dugdale et al., 2015; Wondzell and Gooseff, 2013). Rivers in the two study physiographic units may vary little in sinuosity due partially to the depth and composition of both the bedrock and surficial geological materials that the river moves through (Langbein and Leopold, 1966). Further exploration of the extent and processes explaining the relationship of river temperature and these metrics is the topic of future research.

2.4.2 Maritime Lowland Watersheds

In contrast to the Highlands watershed, for the physiographic setting of the Cains

River maximum watershed elevation (‘Maximum Elev’), percent of river moving through wetland (‘% thru wetlands), and total area of wetlands (‘Wetlands’) were significant, positive variables, whilst river slope (‘Stream slope’) in the watershed was a negative 47

variable in projecting tributary river temperatures. The warmer tributary temperatures at the headwaters of the Cains watershed is likely related to the extensive wetlands and associated advective processes in this low relief upland region with shallow surficial deposits. Wetlands have high solar radiation exposure owing to the lack of solar interception, thus heat up in the summer – unless mitigated by groundwater (Winter et al.,

1998). As these wetlands heat, they also increase the temperature of the receiving river.

Additionally, this is a physiographic setting where groundwater development and flow are restricted, potentially resulting in upland and hillslope hydrological disconnection from rivers or wetlands source areas during the summer or drought periods (Devito and Hill,

1999), and thereby, leading to increases in the receiving rivers’ temperature.

River slope was the only variable negatively correlated with river temperature.

Typically, steep slopes are indicative of shallow surficial deposits, with shallow groundwater flow paths, and are more regulated by air temperatures, and climatic variation

(Johnson et al., 2017, Briggs et al., 2017); however steep river slopes also typically experience low solar radiation exposure (Johnson, 2004). Nevertheless, the Cains Rivers’ underlying geology is permeable, highly fractured sandstone and fluviatile conglomerates

(Rampton et al., 1984). In this physiographic setting with permeable bedrock, we suggest that steep slopes in the mainstem and its tributaries intersect slower and deeper groundwater with cooler temperature discharging from the under-lying lithology (e.g.,

Winter et al., 1998; Hale and McDonnell, 2016). In this setting, the potential for HEF also increases due to large hydraulic gradients and permeable substrate (Clark et al., 1999;

Wondzell and Gooseff, 2013) and coupled with less solar radiation inputs cooler river

temperatures occur in steeper reaches of the Cain’s mainstem and its tributaries (Figure 2-

10).

2.4.3 Interactions of Physiography and Catchment Characteristics

Comparison of the Highlands and Lowland LTP illustrates how the difference in physiography, the depth, structure and configuration of geologic materials (i.e. HRA), influence how smaller scale variables (HU) function to influence river temperatures across different physiographic (i.e management) units. For example, the substantial increase in temperature with migration through wetlands, and lentic water bodies was observed on all three rivers, but the magnitude of the influence differed with location along the river. In both Highland rivers, cool temperatures in the mainstem and inflows were observed in the upstream semi-confined channel section, excluding wetland-derived inputs, and a warming trend was observed in the lower confined channel sections. The semi-confined channel sections on the headwaters are underlain by relatively deep coarse surficial deposits

(Rampton et al., 1984) that promote groundwater flow and HEF which maintain cooler river temperatures. Inversely, steep slopes and the underlying geology on the river and adjacent hill sides result in channel confinement at the lower sections of the rivers. The confinement is associated with shallow surficial deposits and reduced groundwater storage, and HEF potential (Wondzell and Gooseff, 2013) and river temperatures are likely moderated by solar radiation and air temperature. In contrast, the Maritime Lowland has a physiographic setting with lower relief and is underlain by homogeneous fractured sandstone and fluviatile conglomerates, with deeper surficial deposition in the lower reaches. The decrease in mainstem and inflow temperatures in a downstream direction of 49

the Cains River may be explained by: 1) the absence of wetlands and advective warming in the lower reaches; 2) deep groundwater inputs from the fractured bedrock geology, 3) greater surficial deposit depths in the downstream reaches leading to cooler groundwater inflow streams and seeps, and 4) river gradients increase in the lower sections leading to an increase in potential HEF processes (Wondzell and Gooseff, 2013).

The temperature of the small inflows and seeps along the LTP follow a similar trend to the mainstem in both physiographic settings. The ephemeral nature of inflows and seeps of the Highland rivers is apparent by the variability in number and spatial distribution in

Burnthill Brook (Figure 2-4) and Clearwater Brook (Figure 2-7). The greater number of inflows observed in the Clearwater may be a function of inter-annual variability in precipitation, as the Clearwater was sampled during a wetter year (see Table 2-2). The large number of small inflows and seeps on the Cains reflects deeper surficial deposits and permeable bedrock, and thus the temperatures are also more temporally stable than in the

Highland watersheds. The ephemeral character of small inflows and seeps in steep, confined river valley sections, their role in mitigating increases in river temperature and persistence with climate change requires further research.

2.5 References

Abdi, H. 2010. Partial least squares regression and projection on latent structure regression (PLS Regression). Wiley Interdiscip. Rev. Comput. Stat. 2(1): 97–106. John Wiley & Sons, Inc. doi:10.1002/wics.51. Alexander, M. D. (2006). The thermal regime of shallow groundwater in a clearcut and

forested streamside buffer, PhD Thesis, University of New Brunswick (April 2006),

1–436.

Breau, C., Cunjak, R.A., and Peake, S.J. 2011. Behaviour during elevated water temperatures: can physiology explain movement of juvenile Atlantic Salmon to cool water? J. Anim. Ecol. 80(4): 844–853. Blackwell Publishing Ltd. doi:10.1111/j.1365-2656.2011.01828.x. Briggs, M.A., and Hare, D.K. 2018. Explicit consideration of preferential groundwater discharges as surface water ecosystem control points. Hydrol. Process. 32(15): 2435–2440. Wiley-Blackwell. doi:10.1002/hyp.13178. Briggs, M.A., Johnson, Z.C., Snyder, C.D., Hitt, N.P., Kurylyk, B.L., Lautz, L., Irvine, D.J., Hurley, S.T., and Lane, J.W. 2018. Inferring watershed hydraulics and coldwater habitat persistence using multi-year air and stream temperature signals. Sci. Total Environ. 636: 1117–1127. Elsevier. doi:10.1016/J.SCITOTENV.2018.04.344. Briggs, M.A., Lane, J.W., Snyder, C.D., White, E.A., Johnson, Z.C., Nelms, D.L., and Hitt, N.P. 2017. Shallow bedrock limits groundwater seepage-based headwater climate refugia. Limnol. - Ecol. Manag. Inl. Waters. Urban & Fischer. doi:10.1016/J.LIMNO.2017.02.005. Briggs, M.A., Lautz, L.K., Hare, D.K., and González-Pinzón, R. 2013. Relating hyporheic fluxes, residence times, and redox-sensitive biogeochemical processes upstream of beaver dams. Freshw. Sci. 32(2): 622–641. The University of Chicago Press. doi:10.1899/12-110.1. Caine, J.S., and Tomusiak, S.R.A. 2003. Brittle structures and their role in controlling porosity and permeability in a complex Precambrian crystalline-rock aquifer system in the Colorado Rocky Mountain Front Range. Geol. Soc. Am. Bull. 115(11): 1410. GeoScienceWorld. doi:10.1130/B25088.1. Caissie, D. 2006. The thermal regime of rivers: a review. Freshw. Biol. 51(8): 1389–1406. doi:10.1111/j.1365-2427.2006.01597.x.

Caissie, D., Satish, M.G., and El-Jabi, N. 2007. Predicting water temperatures using a deterministic model: Application on Miramichi River catchments (New Brunswick, Canada). J. Hydrol. 336(3–4): 303–315. doi:10.1016/j.jhydrol.2007.01.008. Carrascal, L.M., Galván, I., and Gordo, O. 2009. Partial Least Squares Regression as an Alternative to Current Regression Methods Used in Ecology. WileyNordic Society Oikos. doi:10.2307/40235374. Chadwick, J.G., Nislow, K.H., McCormick, S.D., and McCormick, S.D. 2015. Thermal onset of cellular and endocrine stress responses correspond to ecological limits in brook trout, an iconic cold-water fish. Conserv. Physiol. 3(1): cov017. Oxford University Press. doi:10.1093/conphys/cov017. Chin, W.W. 1998. The partial least squares approach to structural equation modeling. In Modern methods for business research. Edited byG.A. Marcoulides. Lawrence Erlbaum Associates. Available from https://books.google.com/books?hl=en&lr=&id=EDZ5AgAAQBAJ&oi=fnd&pg= PA295&dq=Chin+1998+PLS&ots=48sJ6up5fp&sig=g2Hie3Yy1qK1G8VnO8O VS5kR5TE [accessed 17 October 2018]. Collen, P., and Gibson, R.J. 2000. The general ecology of beavers (Castor spp.), as related to their influence on stream ecosystems and riparian habitats, and the subsequent effects on fish – a review. Rev. Fish Biol. Fish. 10(4): 439–461. Kluwer Academic Publishers. doi:10.1023/A:1012262217012. Constantz, J. 1998. Interaction between stream temperature, streamflow, and groundwater exchanges in alpine streams. Water Resour. Res. 34(7): 1609–1615. doi:10.1029/98WR00998. Cowood, A.L., Moore, C.L., Cracknell, M.J., Young, J., Muller, R., Nicholson, A.T., Wooldridge, A.C., Jenkins, B.R., and Cook, W. 2017. Expansion of landscape characterisation methods within the Hydrogeological Landscape Framework: application in the Australian Capital Territory. Aust. J. Earth Sci. 64(8): 1073– 1084. Taylor & Francis. doi:10.1080/08120099.2017.1255656. Cunjak, R.A. 1996. Winter habitat of selected stream fishes and potential impacts from land-use activity. Can. J. Fish. Aquat. Sci. 53(S1): 267–282. NRC Research Press Ottawa, Canada . doi:10.1139/f95-275. 52

Cunjak, R.A., and Newbury, R.W. 2005. ATLANTIC COAST RIVERS OF CANADA. In Rivers of North America. Elsevier. pp. 938–980. doi:10.1016/B978-012088253- 3/50024-9. Curry, R.A. 2007. Late glacial impacts on dispersal and colonization of Atlantic Canada and Maine by freshwater fishes. Quat. Res. 67(2): 225–233. No longer published by Elsevier. doi:10.1016/J.YQRES.2006.11.002. Curry, R.A., and Devito, K.J. 1996. Hydrogeology of brook trout (Salvelinus fontinalis) spawning and incubation habitats: implications for forestry and land use development. Can. J. For. Res. 26(5): 767–772. NRC Research Press Ottawa, Canada. doi:10.1139/x26-086. Curry, R.A., and Noakes, D.L.G. 1995. Groundwater and the selection of spawning sites by brook trout ( Salvelinus fontinalis ). Can. J. Fish. Aquat. Sci. 52(8): 1733–1740. NRC Research Press Ottawa, Canada . doi:10.1139/f95-765. Curry, R.A., Scruton, D. a, and Clarke, K.D. 2002. The thermal regimes of brook trout incubation habitats and evidence of changes during forestry operations. Can. J. For. Res. 32(7): 1200–1207. doi:10.1139/x02-046. Detty, J.M., and McGuire, K.J. 2010. Threshold changes in storm runoff generation at a till-mantled headwater catchment. Water Resour. Res. 46(7): n/a-n/a. doi:10.1029/2009WR008102. Devito, K., Creed, I., Gan, T., Mendoza, C., Petrone, R., Silins, U., and Smerdon, B. 2005. A framework for broad-scale classification of hydrologic response units on the Boreal Plain: is topography the last thing to consider? Hydrol. Process. 19(8): 1705–1714. John Wiley & Sons, Ltd. doi:10.1002/hyp.5881. Devito, K.J., and Hill, A.R. 1999. Sulphate Mobilization and Pore Water Chemistry in Relation to Groundwater Hydrology and Summer Drought in two Conifer Swamps on the Canadian Shield. Water. Air. Soil Pollut. 113(1/4): 97–114. Kluwer Academic Publishers. doi:10.1023/A:1005081505086. Devito, K.J., Hill, A.R., and Roulet, N. 1996. Groundwater-surface water interactions in headwater forested wetlands of the Canadian Shield. J. Hydrol. 181(1–4): 127–147. Elsevier. doi:10.1016/0022-1694(95)02912-5.

Devito, K.J., Hokanson, K.J., Moore, P.A., Kettridge, N., Anderson, A.E., Chasmer, L., Hopkinson, C., Lukenbach, M.C., Mendoza, C.A., Morissette, J., Peters, D.L., Petrone, R.M., Silins, U., Smerdon, B., and Waddington, J.M. 2017. Landscape controls on long-term runoff in subhumid heterogeneous Boreal Plains catchments. Hydrol. Process. 31(15): 2737–2751. doi:10.1002/hyp.11213. Dilts, T. 2015. Stream Gradient & Sinuosity Toolbox for ArcGISv10. Available from https://www.arcgis.com/home/item.html?id=c8eb4ce1384e45258ccba1b33cd4e3c b [accessed 29 January 2018]. Dugdale, S.J., Bergeron, N.E., and St-Hilaire, A. 2015. Spatial distribution of thermal refuges analysed in relation to riverscape hydromorphology using airborne thermal infrared imagery. Remote Sens. Environ. 160: 43–55. Elsevier. doi:10.1016/J.RSE.2014.12.021. Dugdale, S.J., Franssen, J., Corey, E., Bergeron, N.E., Lapointe, M., and Cunjak, R.A. 2016. Mainstem movement of Atlantic Salmon parr in response to high river temperature. Ecol. Freshw. Fish 25(3): 429–445. doi:10.1111/eff.12224. Ebersole, J.L., Liss, W.J., and Frissell, C.A. 2003. Thermal heterogeneity, stream channel morphology, and salmonid abundance in northeastern Oregon streams. Can. J. Fish. Aquat. Sci. 60(10): 1266–1280. NRC Research Press Ottawa, Canada . doi:10.1139/f03-107. Eriksson, L., Hermens, J.L.M., Johansson, E., Verhaar, H.J.M., and Wold, S. 1995. Multivariate analysis of aquatic toxicity data with PLS. Aquat. Sci. 57(3): 217–241. Birkhäuser-Verlag. doi:10.1007/BF00877428. Esri: GIS Mapping Software, Spatial Data Analytics & Location Platform. Available from https://www.esri.com/en-us/home [accessed 29 May 2018]. French, W. 2014. Protected from the Elements: Winter Ecology of Brown Trout in Groundwater Buffered Streams. Available from https://conservancy.umn.edu/handle/11299/174220 [accessed 29 January 2018]. Fuhrman, A.E., Larsen, D.A., Steel, E.A., Young, G., and Beckman, B.R. 2018. Chinook salmon emergence phenotypes: Describing the relationships between temperature, emergence timing and condition factor in a reaction norm framework. Ecol. Freshw. Fish 27(1): 350–362. doi:10.1111/eff.12351. 54

Fullerton, A.H., Torgersen, C.E., Lawler, J.J., Steel, E.A., Ebersole, J.L., and Lee, S.Y. 2018. Longitudinal thermal heterogeneity in rivers and refugia for coldwater species: effects of scale and climate change. Aquat. Sci. 80(1): 3. Springer International Publishing. doi:10.1007/s00027-017-0557-9. 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. J. Hydrol. 553: 471–485. Elsevier. doi:10.1016/J.JHYDROL.2017.03.024. Hale, V.C., and McDonnell, J.J. 2016. Effect of bedrock permeability on stream base flow mean transit time scaling relations: 1. A multiscale catchment intercomparison. Water Resour. Res. 52(2): 1358–1374. Wiley-Blackwell. doi:10.1002/2014WR016124. Hatch, C.E., Fisher, A.T., Revenaugh, J.S., Constantz, J., and Ruehl, C. 2006. Quantifying surface water-groundwater interactions using time series analysis of streambed thermal records: Method development. Water Resour. Res. 42(10). doi:10.1029/2005WR004787. Hester, E.T., Young, K.I., and Widdowson, M.A. 2013. Mixing of surface and groundwater induced by riverbed dunes: Implications for hyporheic zone definitions and pollutant reactions. Water Resour. Res. 49(9): 5221–5237. doi:10.1002/wrcr.20399. Ver Hoef, J.M., and Peterson, E.E. 2010. A Moving Average Approach for Spatial Statistical Models of Stream Networks. J. Am. Stat. Assoc. 105(489): 6–18. doi:10.1198/jasa.2009.ap08248. Hrachowitz, M., Soulsby, C., Imholt, C., Malcolm, I. a., and Tetzlaff, D. 2010. Thermal regimes in a large upland salmon river: A simple model to identify the influence of landscape controls and climate change on maximum temperatures. Hydrol. Process. 24(July): 3374–3391. doi:10.1002/hyp.7756. Hynes, H.B.N. 1975. The stream and its valley. SIL Proceedings, 1922-2010 19(1): 1–15. Taylor & Francis. doi:10.1080/03680770.1974.11896033. Isaak, D., Luce, C., Rieman, B., Nagel, D., Peterson, E., Horan, D., Parkes, S., and Chandler, G. 2009. Effects of Climate Change and Recent Wildfires on Stream 55

Temperature and Thermal Habitat for Two Salmonids in a Mountain River Network. Ecol. Appl. 20(5): 100319061507001. doi:10.1890/09-0822. Isaak, D.J., Wenger, S.J., and Young, M.K. 2017. Big biology meets microclimatology: defining thermal niches of ectotherms at landscape scales for conservation planning. Ecol. Appl. 27(3): 977–990. doi:10.1002/eap.1501. Isaak, D.J., Young, M.K., Nagel, D.E., Horan, D.L., and Groce, M.C. 2015. The cold-water climate shield: delineating refugia for preserving salmonid fishes through the 21st century. Glob. Chang. Biol. 21(7): 2540–2553. doi:10.1111/gcb.12879. Jefferson, A., Grant, G.E., Lewis, S.L., and Lancaster, S.T. 2010. Coevolution of hydrology and topography on a basalt landscape in the Oregon Cascade Range, USA. Earth Surf. Process. Landforms: n/a-n/a. doi:10.1002/esp.1976. Johnson, S.L. 2004. Factors influencing stream temperatures in small streams: substrate effects and a shading experiment. Can. J. Fish. Aquat. Sci. 61(6): 913–923. doi:10.1139/F04-040. Koirala, S., Hirabayashi, Y., Mahendran, R., and Kanae, S. 2014. Global assessment of agreement among streamflow projections using CMIP5 model outputs. Environ. Res. Lett. 9(6): 64017. IOP Publishing. doi:10.1088/1748-9326/9/6/064017. Kurylyk, B.L., Bourque, C.P.-A., and MacQuarrie, K.T.B. 2013. Potential surface temperature and shallow groundwater temperature response to climate change: an example from a small forested catchment in east-central New Brunswick (Canada). Hydrol. Earth Syst. Sci. 17(7): 2701–2716. doi:10.5194/hess-17-2701-2013. Kurylyk, B.L., MacQuarrie, K.T.B., Caissie, D., and McKenzie, J.M. 2014. Shallow groundwater thermal sensitivity to climate change and land cover disturbances : derivation of analytical expressions and implications for stream temperature projections. Hydrol. Earth Syst. Sci. Discuss. 11: 12573–12626. doi:10.5194/hessd- 11-12573-2014. Langbein, W.B., and Leopold, L.B. 1966. River meanders - Theory of minimum variance. In Professional Paper. doi:10.3133/PP422H. Larocque, M., Fortin, V., Pharand, M.C., and Rivard, C. 2010. Groundwater contribution to river flows – using hydrograph separation, hydrological and hydrogeological

models in a southern Quebec aquifer. Hydrol. Earth Syst. Sci. Discuss. 7(5): 7809– 7838. doi:10.5194/hessd-7-7809-2010. Legleiter, C.J., Roberts, D.A., and Lawrence, R.L. 2009. Spectrally based remote sensing of river bathymetry. Earth Surf. Process. Landforms 34(8): 1039–1059. John Wiley & Sons, Ltd. doi:10.1002/esp.1787. Linnansaari, T., and Cunjak, R.A. 2010. Patterns in apparent survival of Atlantic Salmon (Salmo salar) parr in relation to variable ice conditions throughout winter. Can. J. Fish. Aquat. Sci. 67(11): 1744–1754. doi:10.1139/F10-093. Majerova, M., Neilson, B.T., Schmadel, N.M., Wheaton, J.M., and Snow, C.J. 2015. Impacts of beaver dams on hydrologic and temperature regimes in a mountain stream. Hydrol. Earth Syst. Sci. 19(8): 3541–3556. doi:10.5194/hess-19-3541- 2015. Malcolm, I.A., Soulsby, C., Youngson, A.F., Hannah, D.M., McLaren, I.S., and Thorne, A. 2004. Hydrological influences on hyporheic water quality: implications for salmon egg survival. Hydrol. Process. 18(9): 1543–1560. John Wiley & Sons, Ltd. doi:10.1002/hyp.1405. McCullough, D.A., Bartholow, J.M., Jager, H.I., Beschta, R.L., Cheslak, E.F., Deas, M.L., Ebersole, J.L., Foott, J.S., Johnson, S.L., Marine, K.R., Mesa, M.G., Petersen, J.H., Souchon, Y., Tiffan, K.F., and Wurtsbaugh, W.A. 2009. Research in Thermal Biology: Burning Questions for Coldwater Stream Fishes. Rev. Fish. Sci. 17(1): 90–115. doi:10.1080/10641260802590152. Monk, W. a., Wilbur, N.M., Allen Curry, R., Gagnon, R., and Faux, R.N. 2013. Linking landscape variables to cold water refugia in rivers. J. Environ. Manage. 118: 170– 176. Elsevier Ltd. doi:10.1016/j.jenvman.2012.12.024. Moore, R.D., Spittlehouse, D.L., and Story, A. 2005. Riparian microclimate and stream temperature response to forest harvesting: A Review. J. Am. Water Resour. Assoc. 41(4): 813–834. doi:10.1111/j.1752-1688.2005.tb03772.x. Nielsen, J.L., Lisle, T.E., and Ozaki, V. 1994. Thermally Stratified Pools and Their Use by Steelhead in Northern California Streams. Trans. Am. Fish. Soc. 123(4): 613–626. Taylor & Francis Group . doi:10.1577/1548- 8659(1994)123<0613:TSPATU>2.3.CO;2. 57

Petty, J.T., Thorne, D., Huntsman, B.M., and Mazik, P.M. 2014. The temperature– productivity squeeze: constraints on brook trout growth along an Appalachian river continuum. Hydrobiologia 727(1): 151–166. Springer International Publishing. doi:10.1007/s10750-013-1794-0. Pfister, L., Martínez-Carreras, N., Hissler, C., Klaus, J., Carrer, G.E., Stewart, M.K., and McDonnell, J.J. 2017. Bedrock geology controls on catchment storage, mixing, and release: A comparative analysis of 16 nested catchments. Hydrol. Process. 31(10): 1828–1845. doi:10.1002/hyp.11134. Rampton, V.N., Gauthier, R.C., Thibault, J., and Seaman, A.A. 1984. Quaternary geology of New Brunswick. In Memoir 416. Saar, M.O. 2011. Review: Geothermal heat as a tracer of large-scale groundwater flow and as a means to determine permeability fields. Hydrogeol. J. 19(1): 31–52. Springer- Verlag. doi:10.1007/s10040-010-0657-2. Sharma, P., and Kim, K. 2012. Model Selection in Information Systems Research Using Partial Least Squares Based Structural Equation Modeling. ICIS 2012 Proc. Available from https://aisel.aisnet.org/icis2012/proceedings/ResearchMethods/6 [accessed 29 January 2018]. Smerdon, B.D., Redding, T.E., and Beckers, J. 2009. An overview of the effects of forest management on groundwater hydrology. BC J. Ecosyst. Manag. 10(1): 22–44. Available from http://jem-online.org/index.php/jem/article/view/409. Soulsby, C., Tetzlaff, D., Dunn, S.M., and Waldron, S. 2006. Scaling up and out in runoff process understanding: insights from nested experimental catchment studies. Hydrol. Process. 20(11): 2461–2465. John Wiley & Sons, Ltd. doi:10.1002/hyp.6338. Steel, E.A., Beechie, T.J., Torgersen, C.E., and Fullerton, A.H. 2017. Envisioning, Quantifying, and Managing Thermal Regimes on River Networks. Bioscience 67(6): 506–522. Oxford University Press. doi:10.1093/biosci/bix047. Tague, C., Farrell, M., Grant, G., Lewis, S., and Rey, S. 2007. Hydrogeologic controls on summer stream temperatures in the McKenzie River basin, Oregon. Hydrol. Process. 21(24): 3288–3300. John Wiley & Sons, Ltd. doi:10.1002/hyp.6538.

Tague, C., Grant, G., Farrell, M., Choate, J., and Jefferson, A. 2008. Deep groundwater mediates streamflow response to climate warming in the Oregon Cascades. Clim. Change 86(1–2): 189–210. doi:10.1007/s10584-007-9294-8. Tague, C., and Grant, G.E. 2004. A geological framework for interpreting the low-flow regimes of Cascade streams, Willamette River Basin, Oregon. Water Resour. Res. 40(4). doi:10.1029/2003WR002629. Torgersen, C.E., Price, D.M., Li, H.W., and McIntosh, B.A. 1999. Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecol. Appl. 9(1): 301–319. Wiley-Blackwell. doi:10.1890/1051- 0761(1999)009[0301:MTRASH]2.0.CO;2. Tóth, J. 1962. A theory of groundwater motion in small drainage basins in central Alberta, Canada. J. Geophys. Res. 67(11): 4375–4388. doi:10.1029/JZ067i011p04375. Tóth, J. 1963. A theoretical analysis of groundwater flow in small drainage basins. J. Geophys. Res. 68(16): 4795–4812. doi:10.1029/JZ068i016p04795. Vidon, P., and Smith, A.P. 2007. Upland Controls on the Hydrological Functioning of Riparian Zones in Glacial Till Valleys of the Midwest1. JAWRA J. Am. Water Resour. Assoc. 43(6): 1524–1539. Blackwell Publishing Ltd. doi:10.1111/j.1752- 1688.2007.00125.x. Vidon, P.G.F., and Hill, A.R. 2004. Landscape controls on the hydrology of stream riparian zones. J. Hydrol. 292(1–4): 210–228. Elsevier. doi:10.1016/J.JHYDROL.2004.01.005. van Vliet, M.T.H., Franssen, W.H.P., Yearsley, J.R., Ludwig, F., Haddeland, I., Lettenmaier, D.P., and Kabat, P. 2013. Global river discharge and water temperature under climate change. Glob. Environ. Chang. 23(2): 450–464. Pergamon. doi:10.1016/J.GLOENVCHA.2012.11.002. Warren, R., Price, J., Fischlin, A., de la Nava Santos, S., and Midgley, G. 2011. Increasing impacts of climate change upon ecosystems with increasing global mean temperature rise. Clim. Change 106(2): 141–177. Springer Netherlands. doi:10.1007/s10584-010-9923-5. Wawrzyniak, V., Allemand, P., Bailly, S., Lejot, J., and Piégay, H. 2017. Coupling LiDAR and thermal imagery to model the effects of riparian vegetation shade and 59

groundwater inputs on summer river temperature. Sci. Total Environ. 592: 616– 626. doi:10.1016/j.scitotenv.2017.03.019. Wawrzyniak, V., Piégay, H., Allemand, P., Vaudor, L., Goma, R., and Grandjean, P. 2016. Effects of geomorphology and groundwater level on the spatio-temporal variability of riverine cold water patches assessed using thermal infrared (TIR) remote sensing. Remote Sens. Environ. 175: 337–348. Elsevier. doi:10.1016/J.RSE.2015.12.050. Weber, N., Bouwes, N., Pollock, M.M., Volk, C., Wheaton, J.M., Wathen, G., Wirtz, J., and Jordan, C.E. 2017. Alteration of stream temperature by natural and artificial beaver dams. PLoS One 12(5): e0176313. Public Library of Science. doi:10.1371/journal.pone.0176313. Westhoff, M.C., Gooseff, M.N., Bogaard, T.A., and Savenije, H.H.G. 2011. Quantifying hyporheic exchange at high spatial resolution using natural temperature variations along a first-order stream. Water Resour. Res. 47(10). doi:10.1029/2010WR009767. Wilbur, N. 2012. Characterizing thermal refugia for brook trout (Salvelinus fontinalis) and Atlantic Salmon (Salmo salar) In the Cains River, New Brunswick, Canada. University of New Brunwick. Williams, H. 1979. Appalachian Orogen in Canada. Can. J. Earth Sci. 16(3): 792–807. NRC Research Press Ottawa, Canada . doi:10.1139/e79-070. Winter, T.C. 2001. The concept of hydrologic landscapes. J. Am. Water Resour. Assoc. 37(2): 335–349. Wiley/Blackwell (10.1111). doi:10.1111/j.1752- 1688.2001.tb00973.x. Winter, T.C., Harvey, J.W., Franke, O.L., and Alley, W.M. 1998. Groundwater and surface water: A single resource. USGS Publ.: 79. doi:10.3389/fpsyg.2012.00044. Winter, T.C., Rosenberry, D.O., and LaBaugh, J.W. 2003. Where Does the Groundwater in Small Watersheds Come From? Groundwater 41(7): 989–1000. Blackwell Publishing Ltd. doi:10.1111/j.1745-6584.2003.tb02440.x. Wold, S., Ruhe, A., Wold, H., and Dunn, III, W.J. 1984. The Collinearity Problem in Linear Regression. The Partial Least Squares (PLS) Approach to Generalized Inverses.

SIAM J. Sci. Stat. Comput. 5(3): 735–743. Society for Industrial and Applied Mathematics. doi:10.1137/0905052. Wold, S., Sjöström, M., and Eriksson, L. 2001. PLS-regression: a basic tool of chemometrics. Chemom. Intell. Lab. Syst. 58(2): 109–130. Elsevier. doi:10.1016/S0169-7439(01)00155-1. Wondzell, S.M., and Gooseff, M.N. 2013. 9.13 Geomorphic Controls on Hyporheic Exchange Across Scales: Watersheds to Particles. In Treatise on Geomorphology. Elsevier. pp. 203–218. doi:10.1016/B978-0-12-374739-6.00238-4.

Chapter 3: Effects of topographic resolution and geologic setting on spatial statistical river temperature models2

Abstract

River temperature exerts a critical control on habitat for aquatic biota. As the climate warms in Eastern Canada, threats to habitats of cold-water species will increase, underpinning the necessity to develop an understanding of landscape-scale, thermal regimes of flowing waters. We assessed the performance of spatial statistical network

(SSN) models of river temperature using high resolution thermal infrared imagery (0.6 m) and LiDAR (1 m) compared to NASA’s Shuttle Radar Topography Mission (SRTM - 30 m) topographic data, and interrogate LiDAR derived fine-scale models (3 ha) to describe groundwater connectivity to surface waters in catchments with shallow overburden and varied bedrock geology. LiDAR improved model performance in a catchment underlain by a homogeneous, high hydraulic conductance bedrock (Cains River), but did not improve model performance in a catchment with heterogeneous bedrock and variable hydraulic conductance (North Pole Stream). We hypothesize that differences in bedrock hydraulic conductance modified the control of topography on subsurface flows and discharge patterns to the rivers, and thus produced the mixed performance of the SSN models. At

2 O’Sullivan, A.M., Devito, K.J., Ogilvie, J., Linnansaari, T., Pronk, T., Allard, S., and Curry, R.A. 2020. Effects of geological setting on spatial statistical network predictions of river temperatures. Water Resources Research. https://doi.org/10.1029/2020WR028122

finer scales, river reaches in steep valleys incising high hydraulic conductance bedrock produced groundwater discharge which was absent in incised valleys with low hydraulic conductance bedrock. These findings indicate that while topography exerts an important control on landscape-scale hydrological processes, geologic setting is a similarly important influence on hydrological processes. We suggest the inclusion of a 3rd dimension of spatial autocorrelation, representative of the vertical plane, would broaden the geographic applicability of spatial statistical models for river temperature studies.

Keywords: bedrock geology, groundwater, LiDAR, spatial statistics, river temperatures, wetlands.

3.1 Introduction

Models of river temperature regimes based on spatial statistical networks have been applied with success in the northwestern United States, leading to the development of regional-scale temperature maps (e.g., Isaak et al., 2009, 2015). These models exploit spatial autocorrelation to describe spatial variability in rivers, i.e., the similarity of one river segment to another in relation to space (e.g., Peterson and Ver Hoef, 2010; Isaak et al.,

2014). Similar spatial models have been applied with some success in Scotland (Jackson et al., 2017; Jackson et al., 2018). Cuthbert et al., (2019) illustrated that both northwestern

United States and Scotland are largely defined by water table ratios (WTR) that are typically < 1, i.e., the relationship between climate and groundwater systems is uni- directional, they are recharge controlled and the groundwater table is more disconnected from topography, e.g., Haitjema and Mitchell-Bruker (2005; Figure A-1) (but see Tague 63

and Grant, 2004; Jefferson et al., 2010). While localised groundwater inflows can occur in these systems (Leach and Moore, 2011; MacDonald et al., 2014), there is evidence that the regional interactions between climate, groundwater and topography are uni-directional and the mode of interactions is mostly recharge controlled in both regions (Gleeson et al.,

2011; Cuthbert et al., 2019). Both the northwestern United States and Scotland are also hydrologically defined by montane high elevation sub-catchments that are likely to be groundwater exporters (Schaller and Fan, 2009; Fan, 2019). It can be hypothesized that regional river flows and thus thermal regimes in these regions are mostly driven by surface water processes and tightly coupled to climate (see Isaak et al., 2018; 2020). Eastern

Canada is a region with a complex tectonic and glacial past that has created heterogeneous bedrock geology, shallow surficial geology, and variable relief across its’ ancient highlands and expansive lowlands (Rampton et al., 1984) resulting in highly variable WTRs, both uni and bi-directional modes of interactions (Gleeson et al., 2011) – see Figure A-1. Bi- directional systems indicate a topographic control on groundwater table conditions and climatic interactions that can recharge and/or extract, via evapotranspiration, moisture from the groundwater table, if it is shallow (see Cuthbert et al., 2019). In these systems, local, intermediate, and regional groundwater flows influence river thermal regimes (e.g.,

Kurylyk et al., 2014; O’Sullivan et al., 2019a) and spatial statistical network models of river temperatures have yet to be tested.

The thermal regime of a river is controlled by heat energy transfers at the air-water and river-river bed interfaces (Jackman and Yotsukura, 1977; Caissie, 2006a). At the surface, air-water interface fluxes in solar radiation (both short and longwave) and sensible and latent heat influence river temperatures (Webb et al. 2008; Dugdale et al., 2017).

These processes are spatio-temporally dynamic and can be modified by fluvial geomorphology (e.g., river width - Caissie et al., 2020) and vegetation (e.g., solar radiation exposure - Johnson, 2004). At the river-river bed interface, advective and conductive heat transfers provide the primary path for heat energy exchanges (Kurylyk et al., 2016; Glose et al., 2017). These processes also vary through space and time, and are modified by landscape features (Moore, 2005; Webb et al., 2008). In addition, a river’s thermal regime can be influenced by groundwater discharge (Kurylyk, 2014; Brunner et al., 2017), hyporheic flow (Hannah et al., 2009; Cardenas, 2015), wetlands, lakes, reservoirs, and dams (Webb et al., 2008; O’Sullivan et al., 2019a), and inflows from sub-catchments

(Jones and Schmidt, 2016; Johnson et al., 2017). While there have been many studies that examine site-specific temperature variation in rivers (e.g., Kelleher et al., 2012) and others that use models to predict regional-scale temperature regimes (e.g., Isaak et al., 2009), understanding the complexity of temperature variability in rivers is far from resolved (e.g.,

Webb et al., 2008; Hrachowitz et al., 2010; Fullerton et al., 2015, Steel et al., 2017).

Advances in remote sensing have facilitated the effective depiction of river networks by exploiting topographic data generated from various sensors, e.g., spaceborne radar and photogrammetric systems, such as NASA’s Shuttle Radar Topography Mission

(SRTM), airborne light detection and ranging (LiDAR), and photogrammetry. By combining these techniques, we are beginning to understand the complex spatial arrangement of Earth's water-ways at scales relevant to river networks (e.g., Allen and

Pavelsky 2018). Surface water flow paths can be built from topography-driven, flow accumulation models using digital elevation models (DEMs). Such models have proven useful in hydrologic (e.g., O’Loughlin et al., 2016) and aquatic ecology studies (e.g., Isaak

et al., 2015). Whilst these models and others, e.g., O’Loughlin et al., (2013) and Pavelsky et al.,(2014), are improving our understanding of the hydrologic structures and processes of river networks, limitations of the resolution persist and thus interpretations regarding river networks is limited (Murphy et al., 2008; Allen and Pavelsky, 2015). Further, hydrologic networks derived from surface topography can fail to capture subsurface flow networks (Devito et al., 2005; Gleeson and Manning, 2008).

Salmonids, such as Atlantic Salmon (Salmo salar) and Brook Trout (Salvelinus fontinalis), are coldwater species with narrow thermal tolerances (Hokanson et al., 1973;

Elliot and Elliot, 2010). During high temperature events, salmonids will seek out thermal refugia, i.e., colder water, to mitigate physiological stress (Wilbur et al., 2020; Corey et al., 2017). These refugia are most often groundwater features (Curry and Devito 1996) and can vary from fine-scale, localised bank seeps (Wilbur, 2012) and hyporheic up- wellings (Krause et al., 2011), to broader-scale, tributary inflows (Corey et al., 2019). At the southern extent of Atlantic Salmon and Brook Trout ranges, increasing river temperature associated with climate change are beginning to have negative effects on both species (Dugdale et al., 2016; Corey et al., 2019). Fisheries managers are quickly trying to understand the hydrological processes underpinning these mostly groundwater-derived, thermal refugia to best conserve river habitats and species (Dzara et al., 2019; Kirk et al.,

2020).

Over the past few decades there has been an exponential growth in our ability to accurately map topography and river networks at high spatial resolutions (e.g., Murphy et al., 2008). This is matched by gains in our capacity to accurately map surface water

temperatures at high-spatial resolutions (e.g., Torgersen et al., 1999; O’Sullivan et al.,

2019a) and regional-scales (Fullerton et al., 2015; Sass et al., 2014). Coupled with an increase in computing capacity, these advances are facilitating the development of integrated hydrological modelling efforts (e.g., Gilbert and Maxwell, 2017; Condon et al.,

2020). Even so, these models require high performance computers, and underpin the favour of statistical methods for regional and national-scale river temperature modelling studies (Jackson et al., 2017; Isaak et al., 2020). Concomitantly, the advent of cutting edge spatial statistical models are providing unprecedented details on the spatial configuration of river temperatures at catchment to national-scales (Isaak et al., 2018; Jackson et al.,

2018). Several issues with integrating these technological and statistical advances have emerged: 1) does river temperature model performance improve by increasing topographic resolution?; 2) how applicable are spatial statistical river networks models where the relationship between topography and groundwater varies at catchment and regional scales?, and 3) does high-resolution topographic data facilitate modelling the spatial, thermal variability of fine-scale inflows such as brooks and seeps that may represent thermal refugia for biota?

In this study, we hypothesize that spatial statistical river network models (SSNs - see Ver Hoef et al., 2006) and fine-scale LiDAR DEMs will improve river temperature model performance regardless of geologic setting, and both tools can facilitate high- resolution temperature models of fine-scale inflows. To test these hypotheses, we compare river temperature models constructed from coarse (SRTM) and high-resolution (LiDAR) topographic data in SSNs. We also investigate the potential of LiDAR derived, high-

resolution river networks to model the spatial thermal variability of fine-scale inflows associated with subsurface (groundwater) discharge, i.e., potential thermal refugia. Our study area is the Miramichi River catchment, New Brunswick, Eastern Canada; an area characterised by complex local, intermediate and regional groundwater flow paths, which can be connected and disconnected from topography (Kurylyk, 2014; O’Sullivan et al.,

2019a). Our datasets include high-resolution thermal infrared imagery (TIR – pixel = 0.6 m), SRTM (coarse resolution) and LiDAR (high resolution) topographic data, satellite, climate, and geological data.

3.2 Methods

3.2.1 Study sites

New Brunswick is located in eastern Canada and is characterized by six major physiographic regions (Figure 3-1a and b). The regions WTR is defined by areas which exhibit both groundwater connection and disconnection to/ from topography (Figure 3-1c).

We focus on the Maritimes Plain - Cains River, and Miramichi Highlands - North Pole

Stream (Figure 3-1b). These areas are chosen as Cains River and the North Pole Stream have landscape characteristics to test hypotheses 3 and 4 respectively. The climate of the

Cains River is defined by average annual temperatures between 4.5 – 5.5 °C, and average annual precipitation = 950 - 1000 mm (Fortin and Dubreuil, 2020), and the North Pole

Stream observes average annual temperature between 1.5 - 3 °C, and average annual precipitation = 1050 - 1150 mm (between 1981-2010, see Table A-1). Both catchments are mostly Acadian forest, characterized by softwood and hardwood tree species, including red spruce, balsam fir, black spruce, yellow birch, eastern white pine, eastern white-cedar,

eastern hemlock and sugar maple and dominated by forestry activities with few other disturbances (Linke et al., 2017)

Figure 3-1 Elevation map of New Brunswick (NB) superimposed with physiographic regions (a) and bedrock geology also illustrating physiographic regions (b), where the blue polygon is the Cains River catchment and the red polygon is the North Pole Stream catchment. Longitudinal profile of the Cains

River mainstem (c) showing elevation and the underlying bedrock (Minto

Formation). Longitudinal profile of the North Pole Stream mainstem (d) showing elevation and underlying bedrock (Chain of Rocks/Granite).

Cains River catchment

The Cains River catchment (Cains) is a 5th order tributary of the Miramichi River catchment (Figure 3-2a), and is located on the Maritime Plains physiographic unit, draining an area ≈ 1,400 km2 (Monk et al., 2013). Elevation ranges from 11 – 188 m.a.s.l. (see

Figure 3-1d and 3-2b) and it is characterized by gentle relief, with a high density of upland wetlands including peat bogs in the headwater reaches (Figure 3-2a), and steeply incised valleys in the mid and lower (downriver) sections (mean slope = 18 °; maximum slope =

69 °). The catchment is underlain by homogeneous, Late Devonian to Carboniferous sandstone and fluviatile conglomerate deposits (372.2 - 299 Mya; Rast and Stringer, 1974) which constitute the Minto Formation that has a relatively high hydraulic conductance

(Stapinsky et al., 2002). The catchment is overlain by predominantly shallow (0-5 m), sandy lodgment tills (50%), along with Late Wisconsinan (21 – 25 Ka) minor ablation and sandy / silty blanket deposits (Figure 3-2d; Rampton et al. 1984). Riparian zones are characterised by wetland vegetation in the headwater sections, transitioning to conifers and deciduous vegetation in the mid to lower sections which co-occur with incised valleys and alluvial terraces (Figure A-2).

Figure 3-2 (a) The Cains River catchment, (b) a bare earth digital elevation model

(DEM), (c) lithology, and (d) surficial geology.

North Pole Stream catchment

The North Pole Stream catchment (North Pole Stream) is a 4th order tributary of the Miramichi River and has a surface drainage of ≈ 214 km2 (Figure 3-3a). The North

Pole Stream is located entirely on the Miramichi Highlands with elevation ranging from

206.5 – 730.5 m.a.s.l. (Figure 3-3b and see Figure 3-1). The catchment is characterized by steeply incised channels in the headwaters most probably originating from glacial melt waters (mean slope = 11 °; maximum slope = 58 °), and transitions to hummocky relief in

the mid to lower reaches (Ganong, 1906; Rampton et al., 1984) (Figure 3-3b). The underlying bedrock is high hydraulic conductance in the northwest headwaters and southeast, downriver sections - Cambrian-Ordovician deep water clastic (DWC) deposit

(488 Mya) comprised of fine to medium grained lithic turbidites. There is a low hydraulic conductance, Middle Ordovician, felsic intrusion (granite - 464 Mya) located in the headwaters of the northeast section of the catchment. The remainder of catchment is underlain by an early Devonian, felsic intrusion (granite - 419 Mya) of low hydraulic conductance (Figure 3-3c, see also Figure 3-1d; Gleeson et al., 2011). As a result, the mainstem of the river flows from a zone of high to low hydraulic conductance bedrock

(Figure 3-3c). The catchment is overlain by shallow surficial deposits (0-3 m) comprised of moraine and ablation tills and colluvium deposited heterogeneously of Late Wisconsin age (Figure 3-3d; Rampton et al., 1984). Riparian zones across the catchment are characterised by a mixture of coniferous and deciduous tree species with wetlands in the upper (upriver) headwater sections which co-occur with incised valleys, transitioning to wetland vegetation in mid-catchment, and conifer and deciduous species in the lower- catchment with incised valleys (Figure A-3).

Figure 3-3 The North Pole Stream catchment (a) boundary with wetlands, (b) a bare earth digital elevation model (DEM) where the black star denotes a steeply incised channel, (c) lithology where the black diamond overlays a headwater lake, and (d) surficial geology.

3.2.2 River temperature data – airborne thermal infrared (TIR)

Surface water temperatures for both study areas were measured with airborne thermal infrared imagery (TIR). For both TIR missions, a FLIR Systems SC3000 LWIR sensor (8-9.5µ) was mounted to the underside of a Bell Long Ranger helicopter (Monk et

al., 2013). During 23 July 2008, ≈ 52 km of the mainstem of the Cains was flown (see

O’Sullivan et al., 2019a for details; Figure 3-4a). The total acquisition time was 51 minutes, thus limiting possible effects of diurnal temperature fluxes. The TIR image had a pixel resolution of 0.6 m2. On the 29 July 2009, ≈ 40 km of the mainstem of the North

Pole Stream was surveyed (Figure 3-4b). The total time for imagery acquisition was 22 minutes, thus limiting the effects of diurnal variability. Air temperatures on the day of acquisition ranged from 26.6 – 31.7 °C (measured at Miramichi, New Brunswick -

Environment Canada, 2019). Kinetic temperatures, measured by in-river thermographs housed in a white uPVC solar radiation shield (n=3), were utilized to calibrate radiant temperatures as measured by TIR (Torgersen et al., 1999). The measured difference between kinetic and radiant temperatures ranged from -0.4 to 0.4 °C, (Quantum Spatial, unpublished report). The TIR image also had a pixel resolution of 0.6 m2.

Figure 3-4 Longitudinal river temperature profiles derived from thermal infrared (TIR) surveys (see text) for the Cains River – July

2008 (a) and North Pole Stream – July 2009 (b). The inset figures

compare average daily river temperatures recorded by three and two in-stream temperature loggers (Cains and North Pole Strem, respectively - i to iii) between July 15 – August 31 2017 (see text) and the TIR survey river temperature measurements. Solid black line is the in-stream, daily temperature (average); dashed line is the overall in-stream, average daily temperature; red solid line is the measured

TIR temperature at the site. An example illustrating the differences between hydrographic networks (HN) resolutions (1 km2 and 3 ha) and their corresponding random sampling points (circles) in the Cains

River (c). A fine-scale illustration of discrete inputs identified by the 3 ha HN, where (i) is a groundwater seep and (ii) is an unmapped tributary (d).

The Cains River is well-mixed, where temperature differences throughout the water column are < 1 °C (Wilbur, 2012), and thus the TIR imagery is representative of temperature throughout the water column (see also Wilbur et al., 2020). In the North Pole

Stream, catchment scale high-resolution (0.5 m pixel) bathymetric and hydraulic mapping during summer 2019 established Reynolds numbers (Re) averaging 92,280 ± 86,606 (mean

± SD ) during low flows (O’Sullivan et al., 2020). Laminar flows occur when Re < 2,300 and can be suggestive of thermal stratification limiting the ability of TIR to represent temperature throughout the water column (Torgersen et al., 2001). Re values in the North

Pole Stream were > 2,300, hence it is most probable the North Pole Stream is thermally well-mixed.

TIR data represent a point-in-time measurement. We assessed the longer, in-season and the interannual temperatures in the modelled reaches of both rivers by deploying n = 5 temperature loggers in the Cains (n = 3) and North Pole Stream (n = 2) TIR reaches between

July 15 – August 31, 2017 (Figure 3-4 a and b, insets). There was no statistical difference between TIR temperatures and the average instream temperature (t 0.05, 8 = 0.48; p = 0.65; instream temperature = 17.9 °C ± 1.6 °C (mean ± SD); TIR temperature = 18.4 °C ± 1.8

°C). We continue with the assumption our TIR data is representative of mean river temperature between July 15 and August 31.

River temperature was extracted by overlaying the TIR image on the generated river networks (1 km2 and 3 ha). Points were randomly selected within the spatial extent

of the TIR using the ‘create random points’ function in ArcMap (ESRI, 2019). We then cross-examined the TIR imagery against 1 m2 resolution aerial imagery and when necessary, we adjusted the random point to lie in the mid-point of the mainstem or an inflow (when the latter was selected). Surface water temperature at a point was quantified by querying and averaging pixel values in a 9-point sample (9 pixels) around the randomly selected temperature point as in O’Sullivan et al. (2019a). Overlaying the 1 km2 river network on the Cains TIR imagery, we generated n = 102 random points and n = 229 for the 3 ha network (Figure 3-4). In the North Pole Stream, the same procedure produced n

= 103 points for the 1 km2 network and n = 231 points for the 3 ha network (Figure 3-4).

3.2.3 Model workflow

The general model predicts river temperature at points based on the characteristics of the catchment (area defined by topography) and river network upriver of the temperature observation (Figure 3-5). The catchment and network characteristics are derived from GIS analyses (ArcMap 10.5, ESRI, 2019) and are described in model parameters below.

Steps 1 - 5: Catchment and river network data - Airborne thermal infrared (TIR), climatic, geological, wetlands, and land-use data were assembled into consistent scales, i.e., regardless of topographic resolution these characteristics had the same resolution for every model (Figure 3-5, Table 3-1).

Step 6: Physical geography - We generated two DEMs from SRTM data (30 m2 resolution,

Step 6.1) and LiDAR data (1 m2 resolution, Step 6.2; Figure 3-5, and A-). We synchronized a national hydrographic network (NHN - Natural Resources Canada) to both DEM’s, where

a stream was initiated when flow accumulation = 1 km2 (Step 6.1a and 6.2a; Figure 3-5).

To examine fine-scale inflows and their thermal regimes, we built 3 ha catchments for both study sites, where a flow accumulation of 3 ha initiates a stream (Step 6.2b; Figure 3-5).

The 3 ha scale was chosen because it captured the occurrence of actual seepage locations along the river (A.M. O’Sullivan and R.A. Curry, unpublished data).

Step 7 - River network temperature network – Sample points were randomly selected across the TIR data suite for the 1 km2 networks and these were used in both SRTM and LiDAR derived models. We then built the SSN and aspatial models and compared outputs (Steps

7.1 and 7.2; Figure 3-5). The same procedure was followed for the 3 ha catchment models

(Step 7.3 - Figure 3-5).

Figure 3-5 An overview of the workflow undertaken to test models (see text for further explanation), where 7.3 denotes the test of predicting fine scale (3 ha) potential thermal refugia.

Table 3-1 Landscape-catchment scale parameters used in the river temperature models. Where $ indicates forest harvest was only consider during the 7 years prior to thermal infrared (TIR acquisition), as per findings of groundwater thermal fluxes in this region (see Alexander, 2006).

Descriptor Metric Inclusion rationale Data Source

We investigate if increasing the resolution of stream networks will identify differing LiDAR (NB-ERD) (1 m2 stream network densities due to Stream resolution)/ km underlying geologic structure (Jefferson et Network al., 2010), and also identify fine-scale streams/ inflows, which can be critical SRTM (NASA) (30 m2 thermal refugia. resolution)/ NHN NRCAN Wetlands can cool and heat rivers, dependent on surface area, depth, 2 Wetlands km groundwater regime and solar radiation exposure (Winter et al.,1998; Monk et al.,2013) In this study region, drainage area has been found to be a good analogue for river

Hydrography discharge (Caissie, 2015). The GeoNB Drainage area km2 catchments in this study are remote and (drainage areas in this study: 1 lack stream gauge data, therefore we use km2 and 3 ha) drainage area as a proxy for discharge. Similar to wetlands, lentic bodies can modify river temperatures. The rate of warming or cooling is dependent on Lentic bodies km2 surface area, depth, and groundwater regimes (Winter and LaBaugh, 2003; Weber et al., 2017)

Air Degree Celcius Climate is well established as a driver of DAYMET Temperature (°C) river temperatures (Caissie, 2006) (1 km resolution) Precipitation mm

LiDAR (NB-ERD) (1.5 m2 Climate Watt hours per Solar radiation is well established as a resolution)/ Solar radiation primary component of stream metre squared temperature (Johnson, 2004) SRTM (NASA) (30 m2 (Wh/m2) resolution) Terrestrially, slope can modify solar radiation exposure (Johnson, 2004), and regulate groundwater potential (MCGuire

Slope Degree (°) and McDonnell, 2010). In-river slope, when occurring with particular substrate conditions, can induce hyporheic flows LiDAR (NB-ERD (1 m2 (Wondzell and Gooseff, 2013) resolution) / SRTM (NASA) (30 2 Higher elevation rivers have been shown m resolution) to be mostly controlled by air

Physical GeographyPhysical temperatures via adiabatic cooling (Isaak Elevation m et al., 2018). Schaller and Fan (2009) explain high elevation catchments tend to be groundwater exporters

Deeper valleys can promote incision into Valley depth/ groundwater flow paths, further deep valley wall m/ degrees (°) valleys can reduce air temperature and slope solar radiation effects (Johnson et al.,2017) More sinuous rivers can lead to intrusion Sinuosity Dimensionless into groundwater flow paths (Dunne,

1980; Dugdale et al.,2015)

Bedrock geology can modify river thermal Bedrock km2 regimes dependent on geological structure Geology (Tague et al., 2008; Jefferson et al., 2010)

Faults can create changes in flow and

thermal regimes in rivers (Constantz, Faults and 1998). Contact zone can establish Canadian Geological Survey/NB - km contact zones transitional points from gaining to loosing DERD (1:50,000) Geology reaches (Winter, 1998; O'Sullivan et al. 2019). The depth, and composition, of surficial geology has been shown to be critical in Surficial km2 determining thermal resilience of shallow Geology groundwater aquifers to climate change (Kurylyk et al.,2014; Briggs et al.,2017)

Forest harvesting can modify groundwater and surface water regimes which can Global Forest Change Forest Harvest$ km2 impact river temperatures (Alexander, (30 m resolution)

Land Use Land 2006; Smerdon et al.,2009)

Model parameters – Hydrography

The coarse resolution (1 km2) river network used the SRTM v3.0 Global 1-arc second DEM, or ‘SRTMGL1’ which provides 80 % global coverage at ≈ 30 m resolution and the National Hydrographic Network (NHN - Natural Resources Canada, 2019). The

DEM was downloaded from https://search.earthdata.nasa.gov. We synchronized the NHN with the SRTM DEM by utilising a deterministic eight (D8) flow model in the ArcHydro toolbox (ESRI, 2019). Prior to flow modelling, we hydroburned the SRTM DEM with the

NHN feature using the ‘recondition’ tool - in ArcHydro (Li, 2014). Following hydroburning, we filled all sinks in the DEM using the ‘fill sinks’ module and ran both

‘flow direction’ and ‘flow accumulation’ modules in the same toolbox to create rasters for flow direction and accumulation. We then set a river definition boundary based on a flow

accumulation threshold of 1 km2 (Uuemaa et al., 2018), i.e., the number of raster cells from the flow accumulation raster required to define a river. Next, the ‘stream segmentation’ module was run to develop a suite of unique identification numbers for each river reach.

A catchment raster was then constructed using the stream segmentation and flow direction rasters, where each raster cell in the catchment raster has a value establishing which catchment the cell belongs to (ESRI, 2019). This catchment raster was later transformed into a feature class. Finally, a river network feature class was built using the stream segmentation and flow direction rasters as inputs. The final output was a river network representative of the NHN conditioned to the SRTM DEM. We followed the same method to synchronize the NHN with the high resolution LiDAR DEM to facilitate comparison between models.

A fine resolution river network was derived from a bare earth DEM developed from

LiDAR data acquired for both study sites between 2015 and 2016 by the Province of New

Brunswick (Department of Energy and Resource Development, NB-DERD) using a Riegl

Q780i sensor, and was collected with a point density of 6 points metre-1 (vertical accuracy

= 0.28 m in 2015, 0.12 m in 2016). LiDAR was processed as described in Murphy et al.

(2008). To overcome computational constraints (processing time) inherent in hydrologically modelling 1 m resolution raster data at the catchment scale, we resampled the LiDAR bare earth DEM from 1 m to 2.5 m. To exploit the vertical resolution inherent in the LiDAR derived DEM, we did not run the reconditioning tool. Similar to the 1 km2 river network delineation, we initially filled sinks in the bare earth DEM. Post sink filling, we ran the flow direction and accumulation modules. Field work and a desktop study during summer 2018 established that 3 ha drainage areas can detail fine-scale inflows

(Figure 3-4 c and d). In order to examine fine scale hydrologic processes, we applied a 3 ha (0.03 km2) river definition boundary condition. We mirrored the workflow listed above to demarcate both the 3 ha catchment feature and the 3 ha HN feature class. Wetlands and lentic bodies were delineated using the NB-DERD wetland and lentic bodies maps that are derived from aerial imagery (http://www.snb.ca/geonb1/e/index-E.asp).

Model Parameters - Climate

Climate data for both catchments was downloaded from the DAYMET website

(https://daymet.ornl.gov/). DAYMET provides predicted daily measurements of minimum and maximum temperatures, and precipitation at a spatial resolution of 1 km2 (Thorton et al., 2016). Following our assumption that our TIR data is representative a mean river temperature between July 15 – August 31, we collated maximum and minimum daily temperatures between July 15 and August 31, 2008 for Cains River models, and July 15 –

August 31, 2009 for the North Pole Stream models. Average air temperature was calculated as the average of the maximum and minimum air temperatures using the raster calculator within ESRI’s ArcMap. River discharge is an important parameter for river thermal regimes (Baker et al., 2018). In the remote setting of this study river gauges are absent. In an effort to account for discharge we use precipitation as a proxy. We also examine if inter and intra-seasonal accumulative precipitation affects river temperatures across our study area. The time periods considered for accumulative precipitation (total mm) were: (1) October 1, 2007/08 to August 31, 2008/09; (2) January 1 to August 31,

2008/09; (3) March 1 to August 31, 2008/09; (4) June 1 to August 31, 2008/09; and (5)

July 15 to August 31, 2008/09. The 2007/08 time period is used for the Cains River models, 86

and the 2008/09 period is used the North Pole Stream models. All precipitation data was downloaded from the DAYMET website ((https://daymet.ornl.gov/).

Solar radiation was calculated using the ‘Potential Incoming Solar Radiation’ module in SAGA GIS (System for Automated Geoscientific Analyses) and is measured in

Watt hours per metre2 (Wh/m2; see Böhner and Antonić, 2009; Conard et al., 2015). For the SRTM DEM, we calculated the accumulated direct solar radiation between July 15 to

August 31, 2008/09; our modelled time-period. We made an assumption of clear sky conditions as this period tends to be mostly dry in NB (Fortin and Dubreuil, 2020). We then calculated the average daily solar radiation for the n=47 days. Scale can impact solar radiation modelling efforts, but high-resolution LiDAR can improve the accuracy of regional-scale river temperature models (Loicq et al., 2018). For our resolution models, a digital surface model (DSM) developed from first LiDAR returns, is used to generate solar radiation models. An example of the point cloud density used for the LiDAR derived solar radiation models is given in Figure A-2 and A-3. We resampled the DSM from 1 m to 1.5 m pixel resolution using the ‘resampling’ tool in ArcMap (ESRI, 2019) to reduce computational constraints. We applied the same model conditions as defined by the SRTM

DEM solar radiation module above. The resulting high resolution, solar radiation model is shown in Figure A-5.

Model Parameters - Physical Geography Catchment averages for slope and elevation for both the SRTM and LiDAR DEMs were obtained using the ‘Zonal statistics’ function for means in ArcMap (see O’Sullivan et al.,

2019a). Catchment slope was calculated by first running the ‘Slope’ function in ArcMap 87

for the DEM. Then the ‘Zonal statistics’ function is applied where the catchment slope raster is the input for the upriver, areal extent defined by the temperature point, and average was extracted. This was repeated for average catchment elevation. River sinuosity

(dimensionless) and gradient (degrees) were derived as per O’Sullivan et al., (2019a) and based on 30 m river sections for the river network, i.e., average sinuosity/gradient is based on the 30 m section measures upstream of a temperature point. Valley depth, defined as the vertical distance to the river channel from the surrounding ridges, was calculated using the ‘Valley depth’ module in the terrain analyses suite of the SAGA (Conrad et al., 2015).

To delineate valley wall slope, we used the valley depth raster as a filter using the ‘extract mask’ function in ArcMap, where the catchment slope raster was the input and the valley depth was the areal extent of the extraction. We then ran zonal statistics for average valley depth and valley wall slope, where the statistics were calculated for the raster of interest, i.e., valley depth or valley wall slope, in the same manner as average catchment slope and elevation outlined above.

Model parameters - Geology

Bedrock and surficial geology maps were acquired from NB-DERD

(http://www.snb.ca/geonb1/e/index-E.asp) and the Geological Survey of Canada

(http://www.nrcan.gc.ca/earth-sciences/geography/topographic-information/free-data- geogratis/11042), respectively (Table 3-1). Bedrock and surficial geology were partitioned based on the topographically defined catchment upriver of a temperature point. Metrics for faults and contact zones (km) were extracted from the NB-DERD bedrock geological data

set and defined as the total linear distance in the topographic catchment upriver of the river temperature point (see O’Sullivan et al., 2019a).

Land use

Forest harvest data was extracted from a suite of satellite-derived forest inventory maps, or

‘Hansen Maps’, with a 30 m pixel resolution (Hansen et al., 2013). Hansen Maps present forest loss and gain for the entire planet at 30m resolution for a time period between 2000-

2016 (https://earthenginepartners.appspot.com/science-2013-global-forest). Linke et al.,

(2017) found these satellite data can accurately define harvest year and area in the

Miramichi, with accuracies of 85 and 88 % respectively. We used the forest harvest data suite to describe land use disturbances in both catchments, where area disturbed is measured in km2. Alexander (2006) found in a catchment lying on a transition zone between the Miramichi Highlands and the Maritime plains, groundwater and surface water conditions returned to those of pre-harvest seven years after a harvest event. As such, we defined harvested forest as areas where harvest occurred in the seven years prior to the acquisition of TIR data.

3.2.4 Spatial Statistical Stream Network Model

We created spatial statistical stream network models (SSN) developed by Ver Hoef et al.,

(2006) which uses moving average constructions to build covariance structures. The SSN model accounts for flow connectivity using ‘tail-up’ and ‘tail-down’ autocovariance models (e.g., Peterson et al., 2013). Tail-up models restrict spatial autocorrelation to flow- connected sections, i.e., sections where water flows from the upriver segment to the

downriver segment (Peterson and Ver Hoef, 2010). Tail-down models permit spatial autocorrelation between flow connected and flow unconnected sections, i.e., sections which share a river network (see Isaak et al., 2014 for review). Peterson et al., (2013) acknowledge the need to account for Euclidean patterns which may be inherent in rivers, e.g., intra and inter-sub-catchment movement of groundwater. Attending to this, Ver Hoef and Peterson (2010) combine river-network models and autocorrelation models designed for Euclidean distances to produce a suite of mixed models based on variance components.

The spatial linear model takes the form:

(1) where y is a vector of measured river temperature, X is a matrix of covariates, β is a parameter vector (parameters in Table 1), zTU, zTD and zE are vectors of zero-mean random variables with an autocorrelation structure based on tail-up (TU), tail-down (TD), and

Euclidean (E) covariance functions, and ε is a vector of independent random errors

(Peterson et al., 2013; Ver Hoef and Peterson, 2010; Isaak et al., 2014). To measure dissimilarity between observations through space, the SSN model uses a ‘Torgegram’, which is a semi-variogram for a river network (Peterson and Ver Hoef, 2010). Following examination of each Torgegram, we selected the SSN module was most appropriate for our data and model development (Ver Hoef and Peterson, 2010).

All GIS, SSN processing was completed in ESRI’s ArcMap (version 10.5), using the STARS, Spatial Tools for the Analysis of River Systems, (version 2.0.6; Peterson and

Ver Hoef, 2014); 90

https://www.fs.fed.us/rm/boise/AWAE/projects/SSN_STARS/software_data.html). The

SSN model was completed in RStudio (version 3.0.2 – R version 3.3), using the ‘SSN’ package (Ver Hoef, 2014).

We compared the SSN models to a simple, aspatial, multiple linear regression model (MLR) in the form:

푌̂ = 푏0 + 푏1푋1 + 푏2푋2+ . . . . +푏푝푋푝 (2)

where 푌̂= predicted or expected value of the dependent variable, X1 through Xp are p distinct independent or predictor variables (parameters in Table 3-1), b0 is the value of 푌̂ when all independent variables equal zero, and b1 though bp are the estimated regression coefficients.

To mitigate against issues associated with multicollinearity, we ran a variance inflation factor (VIF). Parameters were iteratively removed, where VIF values > 10 were removed (as per Hair et al., 1995). We kept certain parameters to test the hypotheses outlined earlier, such as bedrock geology, so long as no multi-collinearity existed. The remaining parameters became the final set of descriptors for the models. All VIF analyses were conducted in XLSTAT. Descriptor reduction in the temperature models was derived from successive modelling iterations where statistically insignificant parameters (p>0.05) were removed until only significant descriptors remained. Diagnostics of model performance followed a leave-one-out cross-validation, coefficient of determination (CV

R2- Isaak et al., 2017), Akaike information criterion (AIC- Akaike, 1974) and Root Mean

Square Error (RMSE - Isaak et al., 2017).

3.3 Results

3.3.1 Spatial variability of river temperature

The river temperature observed in the Cains River TIR ranged from 10.5 to 20.8

°C. The Cains River 1 km2 temperature models used n=102 river points and these temperatures ranged from 14.6 – 20.8 °C (average = 19.4 ± 1.0 SD °C; Figure 3-6a). The

Torgegram for the 1 km2 model illustrates a homogeneous, flow-connected temperature profile with little variance (Figure 3-6b). Dissimilarity of flow-unconnected branches increased with distance. In general, the farthest reaches upriver in the mainstem were warmer than the downstream reaches (Figure 3-8 a and b), which mirrors the findings of semi-variance in Figure 3-6b. The Cains River 3 ha temperature models used n=229 river points and these temperatures ranged from 13.1– 20.8 °C (average = 18.6 ±1.87 °C; Figure

3-6a). The Torgegram for the 3 ha river network model illustrates homogeneous flow- connected sites with little variance, and flow unconnected sites display dissimilarly as distance increases (Figure 3-6c). Similarly, this finding mirrors the spatial arrangement of temperatures (Figure 3-8a and b). This suggests that positive spatial auto-correlation is inherent in the Cains River inflows, while the mainstem is relatively homogeneous and does not reveal a detectable trend.

Figure 3-6 (a) Frequency of observations made across a range of temperatures for the

3 ha network (blue line), and the 1 km2 network (black line) across the Cains River.

(b) Torgegram for the 1 km2 network across 35 km, where the blue dots are flow- connected (mainstem) and green dots are flow-unconnected (inflows). (c) Torgegram

for the 3 ha network across the same 35 km. (d) Frequency of observations made across a range of temperatures for the 3 ha network, and the 1 km2 network across the North Pole Stream. (e) Torgegram for the 1 km2 network across 30 km, where the blue dots are flow-connected and green dots are flow-unconnected. (f) Torgegram for the 3 ha network across the same 30 km, where the blue dots are flow-connected and green dots are flow-unconnected. Across each panel, the size of the circles relates to the number of pairs of sites in each bin.

The river temperatures observed in the North Pole Stream TIR ranged from 8.7 to

18.8 °C. The North Pole Stream 1 km2 temperature model used n=103 points and these temperatures ranged from 12.9 – 18.8 °C (average = 15.9 ±1.3 °C; Figure 3-6d). The

Torgegram for the 1 km2 model in North Pole Stream reveals an increase of river temperature dissimilarity with distance for the mainstem and inflow temperature profiles, with both displaying a warming trend downriver (Figure 3-6e). The variance structure is indicative of an exponential type curve (Figure 3-6e). The North Pole Stream 3 ha temperature models used n=231 points and these temperatures ranged from 11.4 – 19 °C

(average = 15.8 ±1.5 °C; Figure 3-6d). The Torgegram for the 3 ha model sites illustrates semi-variance, and followed an exponential trend (Figure 3-6f). Inflows followed the same trend as those identified in the 1 km2 river network, but at the catchment-scale followed the same trend of growing more dissimilar with distance (Figure 3-6f). This highlights that spatial auto-correlation is inherent in the both inflows and the mainstem of the North Pole

Stream.

3.3.2 VIF analyses

Cains River

A total of 18 descriptors were extracted for all models in the Cains River catchment.

The suite of descriptors for the the Cains River 1 km2 SRTM model was reduced to 6 by the VIF analysis (see Table A-4). The suite of descriptors for the Cains River 1 km2 LiDAR models was reduced to 5 by the VIF analysis (see Table A-4). The suite of descriptors for the Cains River 3 ha LiDAR models was reduced to 5 descriptors by the VIF analysis (see

Table A-4).

North Pole Stream

A total of 27 descriptors were extracted for all models in the North Pole Stream

River catchment. The suite of descriptors for the North Pole Stream 1 km2 SRTM models was reduced to 6 by a VIF analysis (see Table A-5). The suite of descriptors for the North

Pole Stream 1 km2 LiDAR models explained was reduced to 5 descriptors (see Table A-

5). The suite of descriptors for the North Pole Stream 3 ha LiDAR models was reduced to

8 descriptors (see Table A-5).

3.3.3 River temperature model performance

Cains River models

Analysing the Torgegrams of the Cains River led to the selection of a Euclidean

SSN model, based on the shortest distance between temperature data, and a Spherical SSN model to capture the temperature effects of flow-unconnected sites (e.g., Ver Hoef at al., 95

2006; Ver Hoef and Peterson, 2010). For the 1 km2 SRTM model the best temperature predicting model was an Exponential Euclidean SSN model (AIC = 152.9; CV R2 = 0.84;

RMSE = 0.46 °C; Table 3-2). The SSN models also had lower residuals when compared with the aspatial model (Figure A-6). A Spherical (tail down) SSN model was selected as the best model for the 1 km2 LiDAR suite of models (AIC 157.3; CV R2 = 0.83; RMSE =

0.48). Examining the residuals in both SRTM and LiDAR models, the LiDAR model was also a better predictor of river temperature when compared to its SRTM counter-model

(Figures A6 and A7). There was little difference between the aspatial and spatial models for the 3 ha river temperature models. An aspatial model was selected for the 3 ha model of fine-scale inflows based on AIC, as this metric illustrated the only discernable difference

(AIC = 749.84; CV R2 = 0.6; RMSE = 1.21 °C – Table 3-2; Figure A-8). The 3 ha model was less accurate in contrast to the other 2 model resolutions.

Table 3-2 The Cains River temperature models. The aspatial model is a multiple linear regression model (MLR) and the spatial stream network models (SSN) are the

Euclidean Exponential (E.E) and Spherical (S), TU = tail-up, TD = tail-down, n = number of river temperature observations; Coeff. = coefficient; Std. Error = standard error; p = p-values; AIC = Akaike information criterion; CV R2 = leave one out, cross- validation; and RMSE = root mean squared error. 1the mean value of the metrics upriver contribution. Selected model is highlighted grey.

Model Descriptor Coeff. Std. p-value AIC Δ CV RMSE Error AIC R 2

(Intercept) 20.24 0.15 <0.001 Aspatial Wetlands 0.05 0.01 <0.001 200.6 47.7 0.76 0.7 (MLR) Valley wall -0.91 0.06 slope1 <0.001

(Intercept) 18.05 1.30 <0.001

SRTM

SSN Wetlands 0.09 0.01 <0.001

2 Exponential Valley 152.9 2.4 0.84 0.46 (n=102) (Euclidean) wall -0.33 0.08 Cains Cains River 1 1 km 1 km slope <0.001 SSN - (Intercept) 18.71 0.54 <0.001 Spherical Wetlands 0.08 0.01 <0.001 (Euclidean), 161.0 0 0.81 0.51 Spherical Valley 1 -0.58 0.13 <0.001 (taildown) wall slope

(Intercept) 23.31 0.32 <0.001 Aspatial Wetlands 0.07 0.01 <0.001 190.1 32.8 0.77 0.6 (MLR) Catchment 1 -3.78 0.23 slope <0.001 (Intercept) 19.37 1.69 <0.001

SSN Wetlands 0.09 0.01 <0.001

LiDAR

158.5 1.2 0.83 0.48

Exponential Catchment 2 1 -1.46 0.50

(n=102) (Euclidean) slope <0.001

Cains Cains River

1 km 1 km (Intercept) 19.42 2.60 <0.001 SSN - Spherical Wetlands 0.10 0.01 <0.001 157.3 0 0.83 0.48 (tail down) Catchment -1.42 0.51 slope1 <0.001

(Intercept) 12.96 0.39 <0.001 Aspatial Wetlands 0.11 0.01 749.4 0.0 0.6 1.21 (MLR) <0.001

Elevation1 0.04 0.00

<0.001

LiDAR (Intercept) 12.97 0.45 <0.001

(n=228) SSN Wetlands 0.11 0.01 <0.001 752.7 3.3 0.58 1.21

Cains Cains River 3 ha Exponential 1 (Euclidean) Elevation 0.04 0.00 <0.001 SSN - (Intercept) 12.97 0.46 <0.001 Spherical Wetlands 0.11 0.01 <0.001 752.7 3.3 0.58 1.21 (tail down) Elevation1 0.04 0.00 <0.001

North Pole Stream catchment models

Inspection of the Torgegrams for the North Pole Stream led to the selection of a

Euclidean SSN model, based on the shortest distance between temperature data, and a

Spherical SSN model, to better capture the temperature effects of tributaries (e.g., Ver Hoef at al., 2006; Ver Hoef and Peterson, 2010); which is similar to the approach used for the

Cains River catchment. For the 1 km2 SRTM model suite an SSN model had the lowest

AIC value (Table 3-3). However, the model fit and associated residuals were higher and tighter, respectively, for the aspatial model (Figure A-9), leading to the selection of the aspatial model as the best model (AIC = 255.6; CV R2 = 0.69; RMSE = 0.8 °C; Table 3-3).

For the North Pole Stream 1 km2 LiDAR a Euclidean Exponential SSN was selected as the best model (AIC = 262.48; CV R2 = 0.17; RMSE = 0.82 °C; Table 3-3; Figure A-10).

Examining the RMSE, the SRTM model was a better predictor of river temperature when compared to its LiDAR counter-model (Table 3-3). For the 3 ha North Pole Stream suite of models both AIC and RMSE was similar across the board (Table 3-3) and examination of model fits and residuals provided little additional insights (Figure A-11). In such instances it is difficult to distinguish the best model by AIC alone (Burnham and Anderson,

2004); hence, we selected the model with the highest CV R2. An aspatial MLR model was selected for the 3 ha model of fine-scale inflows (AIC = 786.1.3; CV R2 = 0.35; RMSE =

1.27 °C – Table 3-3). Echoing the Cains River catchment 3 ha models, the North Pole

Stream 3 ha’s was less accurate in contrast to the other two model resolutions.

Table 3-3 The North Pole Stream temperature models. The aspatial model is a multiple linear regression model (MLR) and the spatial stream network models (SSN) are Euclidean Exponential (E.E), Spherical (S) and Spherical Exponential (S.E.), TU

= tail-up, TD = tail-down, n = number of river temperature observations; Coeff. = coefficient; Std. Error = standard error; p = p-values; AIC = Akaike information criterion; CV R2 = leave one out, cross-validation; and RMSE = root mean squared error. 1the mean value of the metrics upriver contribution, 2mean air temperature between 15th July – 31 August, 2017. Selected model is highlighted grey. DWC = deep water clastic bedrock; Precip. = precipitation.

Model Descriptor Coeff. Std. p- AIC Δ CV RMSE Error value AIC R 2 (Intercept) 11.90 1.07 <0.001 DWC -0.08 0.02 <0.001 Felsic Aspatial 0.02 0.00 (granite) <0.001 255.6 26.3 0.69 0.8 (MLR) Valley 0.25 0.04 depth1 <0.001 Valley wall -0.76 0.15 slope1 <0.001

(Intercept) 13.69 1.40 <0.001

Felsic

0.00 0.00 (granite) 0.03

SRTM SSN-

- Sinuosity -2.01 0.81 0.015

2 Euclidean 230.3 1 0.37 0.7 Valley (n=103) (Exponential) 0.24 0.04 depth1 <0.001

1 km 1 km Valley wall North Pole North Stream Pole -0.54 0.15 slope1 <0.001 (Intercept) 13.64 1.14 <0.001 Felsic 0.00 0.00 (granite) 0.019 SSN - Sinuosity -2.12 0.83 0.012 Spherical 229.3 0 0.37 0.69 Valley (tail down) 0.23 0.04 depth1 <0.001 Valley wall -0.53 0.15 slope1 <0.001 (Intercept) -4.73 5.56 0.4 Air Aspatial 2 1.24 0.34 0.001 269.3 temperature 6.86 0.62 0.86 (MLR) DWC -0.11 0.02 <0.001 4 Felsic 0.02 0.00 (granite) <0.001 (Intercept) 12.71 1.72 <0.001

DWC -0.09 0.03 <0.001 Felsic SSN- 0.02 0.00 (granite) <0.001 262.4

LiDAR Euclidean 0 0.17 0.83 8

- (Exponential) Valley

2 1 0.24 0.08 0.003

(n=103) depth Catchment

1 km 1 -0.26 0.13 0.047

North Pole North Stream Pole slope (Intercept) 12.51 1.54 <0.001 DWC -0.10 0.03 <0.001 Felsic SSN - 0.02 0.00 (granite) <0.001 263.3 Spherical 0.87 0.18 0.82 Valley 5 (tail down) 0.24 0.08 0.002 depth1 Catchment -0.27 0.13 0.04 slope1 100

(Intercept) 27.12 3.79 <0.001 DWC -0.09 0.02 <0.001 Felsic Aspatial 0.02 0.00 (granite) <0.001 786.1 1.4 0.35 1.27 (MLR) Precip. (Oct -0.05 0.02 0.003 - July) Stream slope -0.09 0.03 0.007 (Intercept) 27.62 5.54 <0.001

DWC -0.10 0.02 <0.001

Felsic SSN- 0.02 0.00

LiDAR

(granite) <0.001

- Euclidean 784.9 0.3 0.19 1.27

(n=231) Precip. (Oct (Exponential) -0.05 0.02 0.04

3 ha 3 ha - July)

North Pole North Stream Pole Stream slope -0.08 0.03 0.02 (Intercept) 27.86 5.46 <0.001 DWC -0.10 0.02 <0.001 SSN - Felsic 0.02 0.00 Spherical (granite) <0.001 784.6 0 0.19 1.25 (tail down) Precip. (Oct -0.05 0.02 0.03 - July) Stream slope -0.08 0.03 0.02

Model descriptors

In the Cains, several descriptors were observed to be statistically significant across all model types and spatial resolutions (Table 3-2). Wetlands were identified as a significant and positive descriptor for river temperature in all models (Table 3-2). An example of wetlands flanking a warm river section is shown in Figure 3-7a. Valley wall slope was a negative significant descriptor all SRTM 1 km2 river temperature models

(Table 3-2). Catchment slope was a negative descriptor in all LiDAR 1 km2 (Table 3-2).

Figure 3-7b shows the co-occurrence of steep slopes and groundwater inflows, and a decrease in inflow temperatures in sections with incised valleys is highlighted in Figures

3-8a and b. River elevation was a positive descriptor for river temperature in all 3 ha models (Table 3-2).

101

Figure 3-7 Examples of parameters of importance in the Cains River catchment models. An upland wetland with no groundwater inflows and mainstem temperatures >20 °C (a) and incised valleys (> 70 ° - black arrows) with associated discrete groundwater inflows (13 – 14 oC) to the mainstem (blue arrows) (b). An example of contrasting thermal regimes associated with underlying bedrock geology in the North Pole Stream catchment. River temperature decreases by 3.9 °C over 1.6 km (white arrows) due to diffuse groundwater inflows occurring in the high K sedimentary bedrock, with inflows of 8.7 °C (c). An incised channel in the low K granite bedrock located in the lower sections of the catchment lacking groundwater inflows (d). 102

103

Figure 3-8 An illustration of the mainstem, longitudinal elevations, river temperature profiles (inflows – blue dots; mainstem – black line; inflow trend

– yellow dashed line), and bedrock hydraulic conductance (high and low K) for the Cains River (a and b) and North Pole Stream (b and d). The Cains River profile shows few groundwater inflows in the headwater, along with wetlands and a decrease of mainstem and inflow temperature downriver where valley incisions occur (b). The North Pole Stream profile showing an ≈ 4 °C decrease in 1.6 km below a lake and increasing temperature of the mainstem and inflows downriver over the granite bedrock.

104

In the North Pole Stream, a deep water clastic (DWC) bedrock deposit was a negative significant descriptor in 1 of 3 SRTM 1 km 2, 3 of 3 LiDAR 1 km2 and 3 of 3 3 ha models (Table 3-3). Valley wall slope was a significant negative descriptor in all 3

SRTM 1 km2 models, catchment slope was a significant negative descriptor in 2 of 3

LiDAR 1 km2 models and stream slope was a significant negative descriptor in all 3 3 ha models (Table 3-3). The coolest sections of the North Pole Stream catchment occur in the highest elevation areas, which have the steepest sections of the catchment, and are underlain by DWC bedrock deposit. In this DWC section, TIR identified discrete groundwater discharge (8.7 – 10.9 °C) into the lake and diffuse groundwater discharge below the lake (Figure 3-7c). This diffusive discharge reduced temperatures from 17.5 °C

(lake-outflow) to 13.6 °C in ≈ 1.6 km (Figure 3-7c). This change is also apparent in the mainstem temperature longitudinal profile in Figure 3-8c and d. A felsic granite bedrock deposit was a positive significant descriptor in all models (Table 3-3). Valley depth was a positive significant descriptor in all 3 SRTM 1 km2 models, and 2 of 3 LiDAR 1 km2 models (Table 3-3). A river reach lacking groundwater inflows in a deep, incised valley in the catchment’s lower section underlain by granite bedrock is illustrated in Figure 3-7d.

Average air temperature (15th July – 31 August) was a significant positive descriptor in 1

LiDAR 1 km2 and Precipitation between October and July was a positive significant descriptor in all 3 3 ha models (Table 3-3).

105

3.4 Discussion

3.4.1 Spatial statistical river temperature models in complex geologic settings

Spatial statistical river network models are emerging as a cutting-edge tools for riverine analyses (e.g., Isaak et al., 2009; Jackson et al., 2017). These models are based on Tobler’s first law of geography “everything is related to everything else, but near things are more related than distant things.” (Tobler, 1970). They have proven to be powerful tools in river temperature studies in high relief, montane regions (Steel et al., 2017; Jackson et al., 2018), and preliminary efforts to model aquatic biota populations are encouraging (Isaak et al.,

2017; Hocking et al., 2018). In this study, we found mixed success and sometimes poor outcomes for the SSN models, and the reason appears to be a function of geology.

Emerging from our current and past studies (Monk et al., 2013; O’Sullivan et al., 2019a) and relatively poor SSN performance in regions with similarly variable geology and topography (e.g., Detenbeck et al., 2016 - New England, North Eastern U.S.A.) is a working hypothesis that spatial statistical river network models are effective in regions where the relationship between climate, groundwater and topography is regionally constant, i.e., uni-directional or bi-directional (Tóth, 1962; Gleeson and Manning, 2008).

New Brunswick is characterised by a complex geologic and climatic past and we hypothesized that SSN models would increase model performance in this geologically complex settings; however, SSN models were found to have mixed success. We hypothesize that the utility of current SSN models is defined by an area’s geology and 106

topography. One line of support for our hypothesis is the Cains River which is underlain by homogeneous high conductance bedrock, where groundwater regimes are suggested to be constant throughout and tightly linked to topography (Rivard et al., 2008; Cuthbert et al., 2019). We suggest that in areas where the relationship between groundwater and topography is relatively uniform, regardless of connectivity, SSN models can effectively predict river temperature. This is consistent with Isaak et al., (2017) and Jackson et al.,

(2018) who report strong SSN model fits and low RMSE in regions which have groundwater and topography relationships that are regionally consistent (Figure A-1). The

North Pole Stream is characterized by bedrock with high conductance in the headwater sections and low conductance bedrock in the middle to downriver sections (Figure 3-3).

The catchment also has greater variability of topography in the headwaters than the downstream reaches, and thus we predict contrasting groundwater and topography interactions within the catchment between the headwater and lower sections (Haitjema and

Mitchell-Bruker, 2005; Gleeson et al., 2011). In such geological settings, we suggest SSN models in their current form are confounded by opposing groundwater and topography interactions, i.e., the apparent diffuse groundwater inflows in the headwater reaches and a lack of groundwater inflows in the downstream reaches (see Figure 3-7). Schaller and Fan

(2009) demonstrated that at regional to catchment-scales, underlying geologic structure can disrupt, weaken, and even reverse smooth hydrological transitions. We hypothesis these processes can also disrupt spatial autocorrelation in the river networks (e.g., Grant et al.,

2017). Detenbeck et al., (2016) working in a geologic setting similar region to NB, also reported SSN models with weaker fits and higher RMSEs than those of Isaak et al., (2017) and Jackson et al., (2018) (see Figure A-1). The mixed successes of the SSN models for

107

river temperature networks appears to be associated with the degree and variability of connectivity between local/ intermediate/ regional groundwater and topography.

Research is increasingly improving our understanding of local, intermediate and regional subsurface processes and thus the need to incorporate such processes into hydrological modelling (e.g., Jefferson et al., 2010; Schaller and Fan, 2009; Gleeson et al.

2011; O’Sullivan et al., 2019a; Condon et al. 2020). The inclusion of metrics that capture subsurface processes in SSN models may overcome complexities that appear to be arising from variable groundwater and topography relationships. More broadly, we suggest geologic setting, coupled with climate and solar radiation, is a first order control of river thermal and flow regimes (e.g., Tague and Grant, 2004; Carlier et al., 2019) and thus a critical factor determining the utility and efficacy of spatial statistical river network models.

It may be possible to add a 3rd dimension of spatial autocorrelation in a vertical plane that would capture a more holistic model of hydrological processes underpinning river temperatures.

3.4.2 Effects of topographic resolution on river temperature models

LiDAR continues to provide unprecedented insights into geomorphic and hydrological processes (e.g., Dietrich and Perron, 2006; Legleiter and Fosness, 2019;

Murphy et al., 2008). In the Cains River we observed statistically significant differences between SRTM and LiDAR DEMs (Table A-3), and noted that SRTM DEMs measured forest canopy as opposed to bare earth (see Figure A-4). We found high resolution LiDAR increased the accuracy of river temperature modelling in a geological setting where surface 108

and subsurface hydrological processes are likely tightly coupled, i.e., homogeneous conductance and subsurface flow paths that are replicas of topography (Haitjema and

Mitchell-Bruker, 2005; Cuthbert et al., 2019). The importance of high-resolution LiDAR is apparent when viewing incised valleys of the sub-catchments along the Cains River where discrete groundwater inflows appear and are most probably deeper, bedrock-origin subsurface flows (see Figure 3-8b; Rivard et al., 2008; Kurylyk, 2014). We conclude that in regions where topographic controls on groundwater are spatially constant, SSN models will likely benefit from LiDAR’s robust quantification of topography.

In the North Pole Stream we unexpectedly found no statistical difference between

SRTM and LiDAR DEMs (Table A3). However, similar to the Cains Rivers, the North

Pole Stream SRTM DEM had a positive bias for vegetated areas, and a negative bias when measuring over waterbodies data (see Figure A-4), suggesting LiDAR is a more accurate representation of topographic structure, regardless. In this setting high-resolution LiDAR did not improve the performance of river temperature models (Table 3-3). The Torgegram

(variogram) of the North Pole Stream reveals dissimilarities along the river (Figure 3-6).

The upriver, headwaters have a conductive bedrock with steep topography which creates diffuse inflows that decrease river temperatures (by ≈ 4 °C over 1.6 km; Figure 3-7c).

Contrastingly, a deeply incised valley with low conductance bedrock produced no apparent groundwater inflows to the river, i.e., groundwater and topography disconnection (Figure

3-7d). In such steep and confined, low conductance, shallow bedrock valleys, river temperatures can be warmed from heat gains by solar radiation (Caissie 2006a; Figure A-

4), shallow groundwater heating (Briggs et al., 2017), and fluid friction (O’Sullivan et al.,

2019b). We suggest that LiDAR has limited ability to improve river temperature model

109

performance in areas defined by opposing groundwater and topography interactions

(Schaller and Fan, 2009; Gleeson et al., 2011), unless the models also include parameters that address subsurface processes (Larocque and Broda, 2016).

3.4.3 The influence of wetlands on river temperature

In the catchment of lower relief and homogeneous bedrock composition, Cains

River, wetlands and lentic bodies showed positive effects on river temperatures, regardless of topographic or river network resolution. Monk et al., (2013) and O’Sullivan et al.,

(2019a) also linked wetland complexes to increases in tributary temperatures in the Cains

River. In the Cains River, wetlands and lentic bodies are mostly located in upland areas, generally shallow, and thus most probably have little groundwater influence (Euliss et al.,

2004) and experience heat gains from solar radiation and air temperature during summer resulting in warm water discharge to the river network (Hondzo and Stefan, 1993; Mooij et al., 2007). In the Cains, we also hypothesize that wetlands and lentic bodies recharge the underlying sandstone aquifer (Rivard et al., 2008). However, the effects of wetlands and lentic bodies on river thermal regimes will be largely dictated by groundwater connectivity (Winter, 1988; Winter and LaBaugh, 2003), and these interactions can vary through space and time as a function of climate (LaBaugh et al., 1998; Devito et al., 2017).

Future work is required to establish the hydrologic function of wetlands for river thermal regimes.

110

3.4.4 Predicting fine-scale groundwater thermal refugia

The importance of thermal refugia for coldwater species in rivers will continue to increase in a warming climate (Dugdale et al., 2016; Corey et al., 2019). Modelling fine- scale inflows is complicated because surface water and groundwater thermal regimes are governed by different thermo-mechanical processes and flux at different amplitudes and magnitudes (Anderson, 2005; Glose et al., 2017). Some fine-scale inflows may also be ephemeral, further complicating modelling efforts (Leibowitz et al., 2008; Ebersole et al.,

2015). Even so, the fine-scale models for the Cains River outperformed those in the North

Pole Stream. We extend our hypothesis that in areas where groundwater flows are replicated by topography (see Haitjema and Mitchell-Bruker, 2005), fine-scale groundwater inflows can be accurately modelled using high-resolution LiDAR (see Table

3-2 – 3 ha model). In geologic settings where groundwater and topography relationships are variable, such as the North Pole Stream, efforts to model fine-scale inflows will most likely produce weaker results (see Table 3-3 - 3 ha model). This highlights the necessity of acquiring higher-resolution subsurface data.

3.4.5 Limitations

Adjacent to our findings, it is worth considering the limitations of our modelling approach. TIR presents a snapshot in time; while we investigated how representative these

TIR data are of temporal means, hydroclimatic variability across years will undoubtedly influence hydrological processes. Our remote sites lack discharge data, which is a critical component of river temperature energy budgets (Baker et al., 2018). Land use in our study areas is dominated by forestry, and changes to the forest landscape have occurred between

111

2008/09 and 2017 (Linke et al., 2017); these changes can modify river flow and thermal regimes (Moore, 2005; Smerdon et al., 2009). However, the true effect of harvest on river flows and temperature remains largely unresolved (Alexander, 2006; Kirchner et al., 2020).

The climate data used in these models (DayMet) are limited in areas lacking meteorological stations, such as our study areas. By developing solar radiation models from high- resolution LiDAR we have addressed concerns of spatial resolution (e.g., Loicq et al.,

2018). However, these solar radiation models are temporally limited as we assumed a constant clear sky view. The geologic data used in this study is 1:50,000 scale; at this scale, these data most likely only capture broad-scale hydrological processes, and do not account for important fine-scaler hydrological processes that modify river thermal regimes.

Finally, ephemeral flows, which may not be account for in our modelling efforts, likely exists across both sites.

112

3.5 References

Akaike, H. (1974). A New Look at the Statistical Model Identification. IEEE Transactions

on Automatic Control, 19(6), 716–723. https://doi.org/10.1109/TAC.1974.1100705

Alexander, M. D. (2006). The thermal regime of shallow groundwater in a clearcut and

forested streamside buffer, PhD Thesis, University of New Brunswick (April 2006),

1–436.

Allen, G. H., & Pavelsky, T. M. (2015). Patterns of river width and surface area revealed

by the satellite-derived North American River Width data set. Geophysical Research

Letters, 42(2), 395–402. https://doi.org/10.1002/2014GL062764

Allen, G. H., Pavelsky, T. M., Barefoot, E. A., Lamb, M. P., Butman, D., Tashie, A., &

Gleason, C. J. (2018). Similarity of stream width distributions across headwater

systems. Nature Communications, 9(1), 610. https://doi.org/10.1038/s41467-018-

02991-w

Baker, E. A., Lautz, L. K., Kelleher, C. A., & McKenzie, J. M. (2018). The importance of

incorporating diurnally fluctuating stream discharge in stream temperature energy

balance models. Hydrological Processes, 32(18), 2901–2914.

https://doi.org/10.1002/hyp.13226

Briggs, M. A., Lane, J. W., Snyder, C. D., White, E. A., Johnson, Z. C., Nelms, D. L., &

Hitt, N. P. (2017). Shallow bedrock limits groundwater seepage-based headwater

climate refugia. Limnologica - Ecology and Management of Inland Waters.

https://doi.org/10.1016/J.LIMNO.2017.02.005

113

Brunner, P., Therrien, R., Renard, P., Simmons, C. T., & Franssen, H.-J. H. (2017).

Advances in understanding river-groundwater interactions. Reviews of Geophysics,

55(3), 818–854. https://doi.org/10.1002/2017RG000556

Burnham, K. P., & Anderson, D. R. (2004). Multimodel Inference. Sociological Methods

& Research, 33(2), 261–304. https://doi.org/10.1177/0049124104268644

Caissie, D. (2006a). The thermal regime of rivers: a review. Freshwater Biology, 51(8),

1389–1406. https://doi.org/10.1111/j.1365-2427.2006.01597.x

Caissie, D. (2006b). River discharge and channel width relationships for New Brunswick

rivers. Canadian Technical Report of Fisheries and Aquatic Sciences 2637

Caissie, Daniel, Ashkar, F., & El‐Jabi, N. (2020). Analysis of air/river maximum daily

temperature characteristics using the peaks over threshold approach. Ecohydrology,

13(1). https://doi.org/10.1002/eco.2176

Cardenas, M. B. (2015). Hyporheic zone hydrologic science: A historical account of its

emergence and a prospectus. Water Resources Research, 51(5), 3601–3616.

https://doi.org/10.1002/2015WR017028

Carlier, C., Wirth, S. B., Cochand, F., Hunkeler, D., & Brunner, P. (2019). Exploring

Geological and Topographical Controls on Low Flows with Hydrogeological Models.

Groundwater, 57(1), 48–62. https://doi.org/10.1111/gwat.12845

Condon, L. E., Atchley, A. L., & Maxwell, R. M. (2020). Evapotranspiration depletes

groundwater under warming over the contiguous United States. Nature

Communications, 11(1), 1–8. https://doi.org/10.1038/s41467-020-14688-0

114

Conrad, O., Bechtel, B., Bock, M., Dietrich, H., Fischer, E., Gerlitz, L., Böhner, J. (2015).

System for Automated Geoscientific Analyses (SAGA) v. 2.1.4. Geoscientific Model

Development, 8(7), 1991–2007. https://doi.org/10.5194/gmd-8-1991-2015

Corey, E., Linnansaari, T., Cunjak, R. A., & Currie, S. (2017). Physiological effects of

environmentally relevant, multi-day thermal stress on wild juvenile Atlantic Salmon

(Salmo salar). Conservation Physiology, 5(1), cox014.

https://doi.org/10.1093/conphys/cox014

Corey, E., Linnansaari, T., Dugdale, S. J., Bergeron, N., Gendron, J., Lapointe, M., &

Cunjak, R. A. (2019). Comparing the behavioural thermoregulation response to heat

stress by Atlantic Salmon parr ( Salmo salar ) in two rivers. Ecology of Freshwater

Fish, eff.12487. https://doi.org/10.1111/eff.12487

Curry, R. A., & Devito, K. J. (1996). Hydrogeology of brook trout ( Salvelinus fontinalis)

spawning and incubation habitats: implications for forestry and land use development.

Canadian Journal of Forest Research, 26(5), 767–772. https://doi.org/10.1139/x26-

086

Cuthbert, M. O., Gleeson, T., Moosdorf, N., Befus, K. M., Schneider, A., Hartmann, J., &

Lehner, B. (2019, February 1). Global patterns and dynamics of climate–groundwater

interactions. Nature Climate Change. Nature Publishing Group.

https://doi.org/10.1038/s41558-018-0386-4

Detenbeck, N. E., Morrison, A. C., Abele, R. W., & Kopp, D. A. (2016). Spatial statistical

network models for stream and river temperature in New England, USA. Water

Resources Research, 52(8), 6018–6040. https://doi.org/10.1002/2015WR018349 115

Detty, J. M., & McGuire, K. J. (2010). Threshold changes in storm runoff generation at a

till-mantled headwater catchment. Water Resources Research, 46(7), n/a-n/a.

https://doi.org/10.1029/2009WR008102

Devito, K., Creed, I., Gan, T., Mendoza, C., Petrone, R., Silins, U., & Smerdon, B. (2005).

A framework for broad-scale classification of hydrologic response units on the Boreal

Plain: Is topography the last thing to consider? Hydrological Processes, 19(8), 1705–

1714. https://doi.org/10.1002/hyp.5881

Devito, K. J., Hokanson, K. J., Moore, P. A., Kettridge, N., Anderson, A. E., Chasmer, L.,

… Waddington, J. M. (2017). Landscape controls on long-term runoff in subhumid

heterogeneous Boreal Plains catchments. Hydrological Processes, 31(15), 2737–

2751. https://doi.org/10.1002/hyp.11213

Dietrich, W. E., & Perron, J. T. (2006, January 26). The search for a topographic signature

of life. Nature. Nature Publishing Group. https://doi.org/10.1038/nature04452

Dugdale, S. J., Bergeron, N. E., & St-Hilaire, A. (2015). Spatial distribution of thermal

refuges analysed in relation to riverscape hydromorphology using airborne thermal

infrared imagery. Remote Sensing of Environment, 160, 43–55.

https://doi.org/10.1016/J.RSE.2014.12.021

Dugdale, S. J., Hannah, D. M., & Malcolm, I. A. (2017). River temperature modelling: A

review of process-based approaches and future directions. Earth-Science Reviews.

Elsevier B.V. https://doi.org/10.1016/j.earscirev.2017.10.009

Dunne, T. (1980). Formation and controls of channel networks. Progress in Physical

Geography, 4(2), 211–239. https://doi.org/10.1177/030913338000400204 116

Dzara, J. R., Neilson, B. T., & Null, S. E. (2019). Quantifying thermal refugia connectivity

by combining temperature modeling, distributed temperature sensing, and thermal

infrared imaging. Hydrology and Earth System Sciences, 23(7), 2965–2982.

https://doi.org/10.5194/hess-23-2965-2019

Ebersole, J. L., Wigington, P. J., Leibowitz, S. G., Comeleo, R. L., & Van Sickle, J. (2015).

Predicting the occurrence of cold-water patches at intermittent and ephemeral

tributary con fl uences with warm rivers. Freshwater Science, 34(1), 111–124.

https://doi.org/10.1086/678127.

Elliott, J. M., & Elliott, J. A. (2010). Temperature requirements of Atlantic Salmon Salmo

salar, brown trout Salmo trutta and Arctic charr Salvelinus alpinus: Predicting the

effects of climate change. Journal of Fish Biology, 77(8), 1793–1817.

https://doi.org/10.1111/j.1095-8649.2010.02762.x

Esri: GIS Mapping Software, Spatial Data Analytics & Location Platform. (2019.).

Euliss, N. H., LaBaugh, J. W., Fredrickson, L. H., Mushet, D. M., Laubhan, M. K.,

Swanson, G. A., … Nelson, R. D. (2004). The wetland continuum: A conceptual

framework for interpreting biological studies. Wetlands, 24(2), 448–458.

https://doi.org/10.1672/0277-5212(2004)024[0448:twcacf]2.0.co;2

Fan, Y., Clark, M., Lawrence, D. M., Swenson, S., Band, L. E., Brantley, S. L., …

Yamazaki, D. (2019). Hillslope Hydrology in Global Change Research and Earth

System Modeling. Water Resources Research, 55(2), 1737–1772.

https://doi.org/10.1029/2018WR023903

Fan, Ying. (2019). Are catchments leaky? WIREs Water, 6(6). 117

https://doi.org/10.1002/wat2.1386

Fortin, G., & Dubreuil, V. (2020). A geostatistical approach to create a new climate types

map at regional scale: case study of New Brunswick, Canada. Theoretical and Applied

Climatology, 139(1–2), 323–334. https://doi.org/10.1007/s00704-019-02961-2

Fullerton, A. H., Torgersen, C. E., Lawler, J. J., Faux, R. N., Steel, E. A., Beechie, T. J.,

Ebersole, J.L, Leibowitz, S. G. (2015). Rethinking the longitudinal stream

temperature paradigm: region-wide comparison of thermal infrared imagery reveals

unexpected complexity of river temperatures. Hydrological Processes, 4737(June),

n/a-n/a. https://doi.org/10.1002/hyp.10506

Ganong, W. F. (1906). Notes on the natural history and physiography of New Brunswick.

(Barnes and Co., Ed.). Saint John, New Brunswick.

Gilbert, J. M., & Maxwell, R. M. (2017). Examining regional groundwater and surface

water dynamics using an integrated hydrologic model of the San Joaquin River basin.

Hydrology and Earth System Sciences, 21(2), 923–947. https://doi.org/10.5194/hess-

21-923-2017

Gleeson, T., & Manning, A. H. (2008). Regional groundwater flow in mountainous terrain:

Three-dimensional simulations of topographic and hydrogeologic controls. Water

Resources Research, 44(10). https://doi.org/10.1029/2008WR006848

Gleeson, T., Marklund, L., Smith, L., & Manning, A. H. (2011). Classifying the water table

at regional to continental scales. Geophysical Research Letters, 38(5),

https://doi.org/10.1029/2010GL046427

118

Glose, A. M., Lautz, L. K., & Baker, E. A. (2017). Stream heat budget modeling with

HFLUX: Model development, evaluation, and applications across contrasting sites

and seasons. Environmental Modelling and Software, 92, 213–228.

https://doi.org/10.1016/j.envsoft.2017.02.021

Grant, G. E., O’Connor, J., & Safran, E. (2017). Excursions in fluvial (dis)continuity.

Geomorphology, 277, 145–153. https://doi.org/10.1016/j.geomorph.2016.08.033

Haitjema, H. M., & Mitchell-Bruker, S. (2005). Are Water Tables a Subdued Replica of

the Topography? Groundwater, 0(0), 050824075421008.

https://doi.org/10.1111/j.1745-6584.2005.00090.x

Hannah, D. M., Malcolm, I. A., & Bradley, C. (2009). Seasonal hyporheic temperature

dynamics over riffle bedforms. Hydrological Processes, 23(15), 2178–2194.

https://doi.org/10.1002/hyp.7256

Hansen, M. C., Potapov, P. V., Moore, R., Hancher, M., Turubanova, S. A., Tyukavina,

Thau, A., D., Stehman, S. V., Goetz, S. J., Loveland, T. R., Kommareddy, A., Egorov,

A., Chini, L., Justice1, C. O., Townshend, J. R. G. (2013). High-resolution global

maps of 21st-century forest cover change. Science, 342(6160), 850–853.

https://doi.org/10.1126/science.1244693

Hocking, D. J., Thorson, J. T., O’Neil, K., & Letcher, B. H. (2018). A geostatistical state-

space model of animal densities for stream networks. Ecological Applications, 28(7),

1782–1796. https://doi.org/10.1002/eap.1767

Hokanson, K. E. F., McCormick, J. H., Jones, B. R., & Tucker, J. H. (1973). Thermal

Requirements for Maturation, Spawning, and Embryo Survival of the Brook Trout, 119

Salvelinus fontinalis. Journal of the Fisheries Research Board of Canada, 30(7), 975–

984. https://doi.org/10.1139/f73-158

Hondzo, M., & Stefan, H. G. (1993). Regional water temperature characteristics of lakes

subjected to climate change. Climatic Change, 24(3), 187–211.

https://doi.org/10.1007/BF01091829

Hrachowitz, M., Soulsby, C., Imholt, C., Malcolm, I. a., & Tetzlaff, D. (2010). Thermal

regimes in a large upland salmon river: A simple model to identify the influence of

landscape controls and climate change on maximum temperatures. Hydrological

Processes, 24(July), 3374–3391. https://doi.org/10.1002/hyp.7756

Isaak, D., Luce, C., Rieman, B., Nagel, D., Peterson, E., Horan, D., … Chandler, G. (2009).

Effects of Climate Change and Recent Wildfires on Stream Temperature and Thermal

Habitat for Two Salmonids in a Mountain River Network. Ecological Applications,

20(5), 100319061507001. https://doi.org/10.1890/09-0822

Isaak, D. J., Peterson, E. E., Ver Hoef, J. M., Wenger, S. J., Falke, J. A., Torgersen, C. E.,

Sowder, C., Steel, E. A., Fortin, M-J., Jordan, C.E., Ruesch, A.S., Som, N., Monestiez,

P. (2014). Applications of spatial statistical network models to stream data. Wiley

Interdisciplinary Reviews: Water, 1(3), 277–294. https://doi.org/10.1002/wat2.1023

Isaak, D. J., Young, M. K., Nagel, D. E., Horan, D. L., & Groce, M. C. (2015). The cold-

water climate shield: delineating refugia for preserving salmonid fishes through the

21st century. Global Change Biology, 21(7), 2540–2553.

https://doi.org/10.1111/gcb.12879

Isaak, D. J., Ver Hoef, J. M., Peterson, E. E., Horan, D. L., & Nagel, D. E. (2017). Scalable

120

population estimates using spatial-stream-network (SSN) models, fish density

surveys, and national geospatial database frameworks for streams. Canadian Journal

of Fisheries and Aquatic Sciences, 74(2), 147–156. https://doi.org/10.1139/cjfas-

2016-0247

Isaak, D. J., Luce, C. H., Chandler, G. L., Horan, D. L., & Wollrab, S. P. (2018). Principal

components of thermal regimes in mountain river networks. Hydrology and Earth

System Sciences, 22(12), 6225–6240. https://doi.org/10.5194/hess-22-6225-2018

Isaak, D. J., Luce, C.H., Horan, D.L., Chandler, G.L., Wollrab, S.P., Dubois, W.B., Nagel,

D.E., (in press). Thermal regimes of perennial rivers and streams in the western

United States. Journal of the American Water Resources Association. DOI:

10.1111/1752-1688.12864

Jackman, A. P., & Yotsukura, N. (1977). Thermal loading of natural streams. Professional

Paper. https://doi.org/10.3133/pp991

Jackson, F.L., Hannah, D. M., Fryer, R. J., Millar, C. P., & Malcolm, I. A. (2017).

Development of spatial regression models for predicting summer river temperatures

from landscape characteristics: Implications for land and fisheries management.

Hydrological Processes, 31(6), 1225–1238. https://doi.org/10.1002/hyp.11087

Jackson, Faye L., Fryer, R. J., Hannah, D. M., Millar, C. P., & Malcolm, I. A. (2018). A

spatio-temporal statistical model of maximum daily river temperatures to inform the

management of Scotland’s Atlantic Salmon rivers under climate change. Science of

the Total Environment, 612, 1543–1558.

https://doi.org/10.1016/j.scitotenv.2017.09.010

121

Jefferson, A., Grant, G. E., Lewis, S. L., & Lancaster, S. T. (2010). Coevolution of

hydrology and topography on a basalt landscape in the Oregon Cascade Range, USA.

Earth Surface Processes and Landforms. https://doi.org/10.1002/esp.1976

Johnson, S. L. (2004). Factors influencing stream temperatures in small streams: Substrate

effects and a shading experiment. Canadian Journal of Fisheries and Aquatic

Sciences, 61(6), 913–923. https://doi.org/10.1139/F04-040

Johnson, Z. C., Snyder, C. D., & Hitt, N. P. (2017). Landform features and seasonal

precipitation predict shallow groundwater influence on temperature in headwater

streams. Water Resources Research, 53(7), 5788–5812.

https://doi.org/10.1002/2017WR020455

Jones, N. E., & Schmidt, B. J. (2017). Tributary effects in rivers: Interactions of spatial

scale, network structure, and landscape characteristics. Canadian Journal of Fisheries

and Aquatic Sciences, 74(4), 503–510. https://doi.org/10.1139/cjfas-2015-0493

Kelleher, C., Wagener, T., Gooseff, M., McGlynn, B., McGuire, K., & Marshall, L. (2012).

Investigating controls on the thermal sensitivity of Pennsylvania streams.

Hydrological Processes, 26(5), 771–785. https://doi.org/10.1002/hyp.8186

Kirchner, J. W., Berghuijs, W. R., Allen, S. T., Hrachowitz, M., Hut, R., & Rizzo, D. M.

(2020, February 13). Streamflow response to forest management. Nature. Nature

Research. https://doi.org/10.1038/s41586-020-1940-6

Kirk, R. W. Van, Contor, B. A., Morrisett, C. N., Null, S. E., & Loibman, A. S. (2020).

Potential for Managed Aquifer Recharge to Enhance Fish Habitat in a Regulated

River. Water, 12(3), 673. https://doi.org/10.3390/w12030673 122

Krause, S., Hannah, D. M., Fleckenstein, J. H., Heppell, C. M., Kaeser, D., Pickup, R.,

Pinay, G., Robertson, A.L., Wood, P. J. (2011). Inter-disciplinary perspectives on

processes in the hyporheic zone. Ecohydrology, 4(4), 481–499.

https://doi.org/10.1002/eco.176

Kurylyk, B. L., MacQuarrie, K. T. B., Caissie, D., & McKenzie, J. M. (2014). Shallow

groundwater thermal sensitivity to climate change and land cover disturbances :

derivation of analytical expressions and implications for stream temperature

projections. Hydrology and Earth System Sciences Discussions, 11, 12573–12626.

https://doi.org/10.5194/hessd-11-12573-2014

Kurylyk, B. L., Moore, R. D., & MacQuarrie, K. T. B. (2016). Scientific briefing:

quantifying streambed heat advection associated with groundwater-surface water

interactions. Hydrological Processes, 30(6), 987–992.

https://doi.org/10.1002/hyp.10709

LaBaugh, J. W., Winter, T. C., & Rosenberry, D. O. (1998). Hydrologic functions of prairie

wetlands. Great Plains Research, 8, 17–37. https://doi.org/10.2307/24156332

Larocque, M., & Broda, S. (2016). Groundwater–surface water interactions in Canada.

Canadian Water Resources Journal / Revue Canadienne Des Ressources Hydriques,

41(4), 451–454. https://doi.org/10.1080/07011784.2016.1176537

Leach, J. A., & Moore, R. D. (2011). Stream temperature dynamics in two

hydrogeomorphically distinct reaches. Hydrological Processes, 25(5), 679–690.

https://doi.org/10.1002/hyp.7854

Legleiter, C., & Fosness, R. (2019). Defining the Limits of Spectrally Based Bathymetric 123

Mapping on a Large River. Remote Sensing, 11(6), 665.

https://doi.org/10.3390/rs11060665

Leibowitz, S. G., Wigington, P. J., Rains, M. C., & Downing, D. M. (2008). Non-navigable

streams and adjacent wetlands: Addressing science needs following the Supreme

Court’s Rapanos decision. Frontiers in Ecology and the Environment. John Wiley &

Sons, Ltd. https://doi.org/10.1890/070068

Li, Z. (2014). Watershed modeling using arc hydro based on DEMs: a case study in

Jackpine watershed. Environmental Systems Research, 3(1), 11.

https://doi.org/10.1186/2193-2697-3-11

Linke, J., Fortin, M.-J., Courtenay, S., & Cormier, R. (2017). High-resolution global maps

of 21st-century annual forest loss: Independent accuracy assessment and application

in a temperate forest region of Atlantic Canada. Remote Sensing of Environment, 188,

164–176. https://doi.org/10.1016/J.RSE.2016.10.040

Loicq, P., Moatar, F., Jullian, Y., Dugdale, S. J., & Hannah, D. M. (2018). Improving

representation of riparian vegetation shading in a regional stream temperature model

using LiDAR data. Science of the Total Environment, 624, 480–490.

https://doi.org/10.1016/j.scitotenv.2017.12.129

MacDonald, R. J., Boon, S., Byrne, J. M., & Silins, U. (2014). A comparison of surface

and subsurface controls on summer temperature in a headwater stream. Hydrological

Processes, 28(4), 2338–2347. https://doi.org/10.1002/hyp.9756

Monk, W. A., Wilbur, N. M., Curry, R.A., Gagnon, R., & Faux, R. N. (2013). Linking

landscape variables to cold water refugia in rivers. Journal of Environmental 124

Management, 118, 170–176. https://doi.org/10.1016/j.jenvman.2012.12.024

Mooij, W. M., Janse, J. H., De Senerpont Domis, L. N., Hülsmann, S., & Ibelings, B. W.

(2007). Predicting the effect of climate change on temperate shallow lakes with the

ecosystem model PCLake. In Shallow Lakes in a Changing World (pp. 443–454).

Springer Netherlands. https://doi.org/10.1007/978-1-4020-6399-2_40

Moore, R. D., Spittlehouse, D. L., & Story, A. (2005). Riparian microlimate and stream

temperature response to forest harvesting: A review. Journal of the American Water

Resources Association, 41(4), 813–834. https://doi.org/10.1111/j.1752-

1688.2005.tb03772.x

Murphy, P. N. C., Ogilvie, J., Meng, F. R., & Arp, P. (2008). Stream network modelling

using lidar and photogrammetric digital elevation models: A comparison and field

verification. Hydrological Processes, 22(12), 1747–1754.

https://doi.org/10.1002/hyp.6770

O’Loughlin, F. E., Neal, J., Yamazaki, D., & Bates, P. D. (2016). ICESat-derived inland

water surface spot heights. Water Resources Research, 52(4), 3276–3284.

https://doi.org/10.1002/2015WR018237

O’Loughlin, F., Trigg, M. A., Schumann, G. J.-P., & Bates, P. D. (2013). Hydraulic

characterization of the middle reach of the Congo River. Water Resources Research,

49(8), 5059–5070. https://doi.org/10.1002/wrcr.20398

O’Sullivan, A.M., Devito, K. J., & Curry, R. A. (2019a). The influence of landscape

characteristics on the spatial variability of river temperatures. Catena, 177.

https://doi.org/10.1016/j.catena.2019.02.006 125

O’Sullivan, A. M., Linnansaari, T., & Curry, R. A. (2019b). Ice Cover Exists (ICE): A

quick method to delineate groundwater inputs in running waters for cold and

temperate regions. Hydrological Processes, hyp.13557.

https://doi.org/10.1002/hyp.13557

O'Sullivan, A.M., Wegcheider, B., Helminen, J., Cormier, J.G., Linnansaari, T., Wilson,

D.A., & Curry, R.A. (in press). Catchment-scale, high-resolution, hydraulic models

and habitat maps - a salmonid's perspective. Journal of Ecohydraulics:

DOI:10.1080/24705357.2020.1768600

Pavelsky, T. M., Durand, M. T., Andreadis, K. M., Beighley, R. E., Paiva, R. C. D., Allen,

G. H., & Miller, Z. F. (2014). Assessing the potential global extent of SWOT river

discharge observations. Journal of Hydrology, 519, 1516–1525.

https://doi.org/10.1016/J.JHYDROL.2014.08.044

Peterson, E. E., & Ver Hoef, J. M. (2010). A mixed-model moving-average approach to

geostatistical modeling in stream networks. Ecology, 91(3), 644–651. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/20426324

Peterson, E. E., Ver Hoef, J. M., Isaak, D. J., Falke, J. A., Fortin, M.-J., Jordan, C. E., …

Wenger, S. J. (2013). Modelling dendritic ecological networks in space: an integrated

network perspective. Ecology Letters, 16(5), 707–719.

https://doi.org/10.1111/ele.12084

Rampton, V. N., Gauthier, R. C., Thibault, J., & Seaman, A. A. (1984). Quaternary geology

of New Brunswick. Memoir 416. Geological Survey of Canada, Memoir 416.

https://doi.org/10.4095/119730

126

Rivard, C., Michaud, Y., Lefebvre, R., Deblonde, C., & Rivera, A. (2008). Characterization

of a Regional Aquifer System in the Maritimes Basin, Eastern Canada. Water

Resources Management, 22(11), 1649–1675. https://doi.org/10.1007/s11269-008-

9247-7

Sass, G. Z., Creed, I. F., Riddell, J., & Bayley, S. E. (2014). Regional-scale mapping of

groundwater discharge zones using thermal satellite imagery. Hydrological

Processes, 28(23), 5662–5673. https://doi.org/10.1002/hyp.10068

Schaller, M. F., & Fan, Y. (2009). River basins as groundwater exporters and importers:

Implications for water cycle and climate modeling. Journal of Geophysical Research,

114(D4), D04103. https://doi.org/10.1029/2008JD010636

Smerdon, B. D., Redding, T. E., & Beckers, J. (2009). An overview of the effects of forest

management on groundwater hydrology. BC Journal of Ecosystems and Management,

10(1), 22–44. Retrieved from http://jem-online.org/index.php/jem/article/view/409

Stapinsky, M., Michaud, Y., Morin, R.H., Butler, K.E., Deblonde, C., Chi, G., Thériault,

T., Castonguay, S., Boisvert, V., Julien, H.P. and Conohan, D., 2002. Groundwater

resources assessment in the Carboniferous Maritimes Basin: preliminary results of the

hydrogeological characterization, New Brunswick, Nova Scotia, and Prince Edward

Island. Natural Resources Canada, Geological Survey of Canada.

https://doi.org/10.4095/213247

Steel, E. A., Beechie, T. J., Torgersen, C. E., & Fullerton, A. H. (2017). Envisioning,

Quantifying, and Managing Thermal Regimes on River Networks. BioScience, 67(6),

506–522. https://doi.org/10.1093/biosci/bix047 127

Tague, C., & Grant, G. E. (2004). A geological framework for interpreting the low-flow

regimes of Cascade streams, Willamette River Basin, Oregon. Water Resources

Research, 40(4). https://doi.org/10.1029/2003WR002629

Tague, C., Grant, G., Farrell, M., Choate, J., & Jefferson, A. (2008). Deep groundwater

mediates streamflow response to climate warming in the Oregon Cascades. Climatic

Change, 86(1–2), 189–210. https://doi.org/10.1007/s10584-007-9294-8

Thornton, P.E., M.M. Thornton, B.W. Mayer, Y. Wei, R. Devarakonda, R.S. Vose, and

R.B. Cook. 2016. Daymet: Daily Surface Weather Data on a 1-km Grid for North

America, Version 3. ORNL DAAC, Oak Ridge, Tennessee, USA.

https://doi.org/10.3334/ORNLDAAC/1328

Tobler, W. R. (1970). A Computer Movie Simulating Urban Growth in the Detroit Region.

Economic Geography, 46, 234. https://doi.org/10.2307/143141

Torgersen, C. E., Price, D. M., Li, H. W., & McIntosh, B. A. (1999). Multiscale thermal

refugia and stream habitat associations of chinook salmon in northeastern Oregon.

Ecological Applications, 9(1), 301–319. https://doi.org/10.1890/1051-

0761(1999)009[0301:MTRASH]2.0.CO;2

Tóth, J. (1962). A theory of groundwater motion in small drainage basins in central Alberta,

Canada. Journal of Geophysical Research, 67(11), 4375–4388.

https://doi.org/10.1029/JZ067i011p04375

Uuemaa, E., Hughes, A., Tanner, C., Uuemaa, E., Hughes, A. O., & Tanner, C. C. (2018).

Identifying Feasible Locations for Wetland Creation or Restoration in Catchments by

Suitability Modelling Using Light Detection and Ranging (LiDAR) Digital Elevation 128

Model (DEM). Water, 10(4), 464. https://doi.org/10.3390/w10040464

Ver Hoef, J. M., & Peterson, E. E. (2010). A Moving Average Approach for Spatial

Statistical Models of Stream Networks. Journal of the American Statistical

Association, 105(489), 6–18. https://doi.org/10.1198/jasa.2009.ap08248

Ver Hoef, J. M., Peterson, E., & Theobald, D. (2006). Spatial statistical models that use

flow and stream distance. Environmental and Ecological Statistics, 13(4), 449–464.

https://doi.org/10.1007/s10651-006-0022-8

Webb, B. W., Hannah, D. M., Moore, R. D., Brown, L. E., & Nobilis, F. (2008). Recent

advances in stream and river temperature research. Hydrological Processes, 22(7),

902–918. https://doi.org/10.1002/hyp.6994

Weber, N., Bouwes, N., Pollock, M.M., Volk, C., Wheaton, J.M., Wathen, G., Wirtz, J.

and Jordan, C.E., 2017. Alteration of stream temperature by natural and artificial

beaver dams. PloS one, 12(5), p.e0176313.

https://doi.org/10.1371/journal.pone.0176313

Wilbur, N. (2012). Characterizing thermal refugia for brook trout (Salvelinus fontinalis)

and Atlantic Salmon (Salmo salar) In the Cains River, New Brunswick, Canada.

University of New Brunwick.

Wilbur, N. M., O’Sullivan, A. M., MacQuarrie, K. T. B., Linnansaari, T., & Curry, R. A.

(2020). Characterizing physical habitat preferences and thermal refuge occupancy of

brook trout (Salvelinus fontinalis) and Atlantic Salmon (Salmo salar) at high river

temperatures. River Research and Applications, rra.3570.

https://doi.org/10.1002/rra.3570 129

Winter, T. C., Harvey, J. W., Franke, O. L., & Alley, W. M. (1998). Groundwater and

surface water; a single resource. Circular. https://doi.org/10.3133/CIR1139

Winter, T. C., & LaBaugh, J. W. (2003). Hydrologic considerations in defining isolated

wetlands. Wetlands, 23(3), 532. https://doi.org/10.1672/0277-

5212(2003)023[0532:hcidiw]2.0.co;2

Wondzell, S. M., & Gooseff, M. N. (2013). 9.13 Geomorphic Controls on Hyporheic

Exchange Across Scales: Watersheds to Particles. In Treatise on Geomorphology (pp.

203–218). Elsevier. https://doi.org/10.1016/B978-0-12-374739-6.00238-4

130

Chapter 4: Ice Cover Exists (ICE): A quick method to delineate groundwater inputs in running waters for cold and temperate regions3

Abstract

Groundwater can be important in regulating stream thermal regimes in cold, temperate regions and as such, it can be a significant factor for aquatic biota habits and habitats. Groundwater typically remains at a relatively constant temperature through time, i.e., it is warmer than surface water in the winter and cooler in the summer. Further, small tributaries are often dominated by groundwater during low flows of winter and summer.

We exploit these thermal patterns to identify and delineate tributary/groundwater inputs along a frozen river (ice-on) using publicly available satellite data, and we tested the findings against airborne, thermal infrared (TIR) data. We utilise a supervised maximum likelihood classification (sMLC) to identify possible groundwater inputs while the river is in a frozen state (kappa coefficient 96.77 when compared to visually delineated possible groundwater inputs). We then compare sMLC identified possible groundwater inputs to

TIR classified groundwater inputs which confirmed there was no statistical difference (2

=0.78), i.e., confirming groundwater inputs can be delineated in north temperate river systems using available satellite imagery of the system’s frozen state. Our results also

3 O'Sullivan, A.M., Linnansaari, T. and Curry, R.A., 2019. Ice cover exists: A quick method to delineate groundwater inputs in running waters for cold and temperate regions. Hydrological Processes, 33(26), pp.3297-3309. https://doi.org/10.1002/hyp.13557

131

established the spatial extent and influence of possible groundwater inputs in two seasons.

The thermal plumes were longer and narrower in winter; this is likely related to seasonal differences in dispersion regimes. We hypothesize that differences between summer and winter is related to either (1) tributaries which are modulated by shading in the summer, or

(2) aquifer disconnection from the river in the winter owing to frozen ground conditions and lack of aquifer recharge. This method of establishing tributary/groundwater inputs and contributions to surface water thermal regimes is relatively simple, and can be useful for science and management as long as ‘ice cover exists’ (ICE), that is, the system can achieve a frozen state.

Keywords: Groundwater; ice; satellite imagery; thermal infrared; summer/ winter; maximum likelihood classification; thermal dispersion; thermal refugia

4.1 Introduction

Groundwater contributions to streams vary depending on development, storage capacity, residence times, and flow paths (see among many examples, Larocque and Broda,

2016; Brunner et al., 2017), and these are functions of bedrock geology (O’Sullivan et al.,

2019), surficial geology (Vidon and Smith, 2007), topography (Johnson et al., 2017), vegetation (Groeneveld and Griepentrog, 1985), and climate (Kurylyk et al., 2014).

Groundwater interactions with surface water occur at various scales through space and time

(Winter, 1995; Callahan et al., 2015). At broader spatial scales, groundwater contributions to surface waters can be sourced from groundwater-dominant tributaries (e.g., Winter

2007), while at finer, reach scales, inputs can be localized, e.g., seepage along banks (e.g., 132

Kurylyk et al., 2013). The discharge regime of groundwater is typically relatively constant through time (Constantz, 1998), but can flux depending the aquifer storage capacity (Tague et al., 2007) and the climate regime (Briggs et al., 2017).

The thermal regime of groundwater is driven by thermal and hydraulic conductance characteristics (Arenson et al., 2002) and associated thermo-dynamic processes (Anderson,

2005), and can be modified by natural (Isaak et al., 2009) and anthropogenic landscape disturbances (Moore et al., 2005) as well as climate (Briggs et al., 2018). Owing to the relatively constant rate of discharge and temperature (e.g., Alexander and Caissie, 2003), groundwater contributions to flowing waters in north temperate regions are typically persistent and cooler than surface water in the summer (e.g., Kurylyk et al., 2014) and warmer in the winter (e.g., Linnansaari and Cunjak, 2010).

The thermal and discharge characteristics of groundwater-based features in flowing waters can control fish habitats and therefore their behaviour (Curry and Devito, 1996).

Fish use groundwater discharge directly for reproduction (Curry and Noakes, 1995;

Malcolm et al., 2004) and indirectly, capitalizing on groundwater-dominated tributary discharge to larger rivers that provide thermal refugia in summer (Breau et al., 2011;

Dugdale et al., 2016) and winter (Cunjak, 1996; Cunjak et al., 1998). These habitats and their role as refugia are becoming increasingly critical for the management and conservation of coldwater species as our climate changes (Isaak et al., 2015).

133

Thermal infrared imagery (TIR) is being used to delineate river temperature at fine scales over large areas (e.g., Torgensen et al., 1999; Handcock et al., 2012), examine the physical processes which regulate stream temperature regimes (e.g., Monk et al., 2013), and investigate behavioural and physiological responses of coldwater species under extreme heat events (e.g., Petty et al., 2012; Dugdale et al., 2016). While TIR has more than shown its novelty and efficacy in the fields of hydrology (e.g., Fullerton et al., 2015) and aquatic ecology (e.g., Wilbur 2012), it remains an expensive tool that presents a single, snap shot in time of a river’s temperature. Satellite imagery can also detect and delineate the thermal properties of landscapes and possible groundwater discharge zones, e.g., Sass et al. (2014).

However, the resolution of satellite-derived thermal images is coarse and limited for applications in stream and river ecosystem management and conservation (Briggs and

Hare, 2018), specifically, Landsat Enhanced Thematic Mapper Plus (ETM+) Band 6

(thermal band) has a ground resolution = 60 m resampled to 30 m

(https://landsat.usgs.gov).

There is however, the potential that temperature variances created by groundwater can be detected in flowing waters using satellite imagery by considering the thermal dynamics of discharging groundwater in running waters either directly, e.g., seepage zones or via the discharge zone of smaller groundwater-dominated tributaries in rivers. When flowing waters become ice covered in winter, the zones of groundwater discharge and tributary confluences can be void of ice depending on their aquifer geometry and composition

(Paznekas and Hayashi, 2016). We examined an extensive network of TIR delineated summer temperatures in a river (see Monk et al. 2013; O’Sullivan et al., 2019) to identify

134

thermal anomalies and compared these with freely available, satellite imagery from the same locations collected in winter. Our goal was to establish if such satellite imagery could be used as an effective method of demarcating potential groundwater sources along the river network. The ice coverage images also allowed us an opportunity to consider the processes creating the geometry of groundwater and tributary plumes in summer and winter towards future quantification of the spatiotemporal contributions of groundwater in a river network.

4.2 Methods

4.2.1 Study Area

Our study area extends across the Miramichi River catchment in central New

Brunswick, Canada (known in Mi'kmaq as the "Lustagoocheehk" which translates to

"goodly little river") (Figure 4-2). The watershed is ≈ 14,000 km2 and encompasses two physiographic regions – the Maritime Plains and the Miramichi Highlands. The Miramichi

Highlands are a component of the Acadian Orogeny – the northern most unit of the

Appalachian mountain range, while the Maritime Plains are a lowland area that has witnessed various geomorphic processes from oceanic to glacial (Rampton et al., 1984).

The region has a typical north, temperate climate with cold winters (mean temperatures

December to February ≈ -7 oC) and warm summers (mean temperatures in July ≈18.8 oC with highs to 30 oC; Cunjak and Newbury, 2005). Rivers and streams are ice covered by late December and remain covered until mid-late April (Cunjak and Caissie, 1993).

However, smaller streams that are groundwater dominated can remain ice free to varying

135

degrees (Figure 4-1, and Cunjak, 1998). In summer, river temperatures can exceed 30 oC in the Miramichi (see Corey et al., 2019), while in the winter, river temperatures are typically 0 oC (Cunjak and Caissie, 1993). Groundwater thermal regimes in the Miramichi vary with depth, and display a lagging behaviour in relation to seasonal air temperature fluxes. Kurylyk et al., (2013), for instance, found lag times in the order 1.3 months between seasonal air temperatures and groundwater temperatures at a depth of 2 m (average monthly summer ≈ 11.2 oC, average monthly winter ≈ 3 oC), while groundwater temperatures at 8.75 m observed a lag time of ≈ 5.7 months with an average monthly summer ≈ 6 oC, and average monthly winter ≈ 7 oC, in Catamaran Brook (Alexander,

2006). Therefore, in the summer groundwater temperatures can be between 2-5 times cooler than ambient river temperatures, but in the winter groundwater temperatures can be up to 7 oC warmer than ambient river temperatures (depth and geologic material dependent).

136

Figure 4-1 Little Southwest Miramichi River, NB and its tributary Otter Brook, A –

B (source: GeoNB). A - The reach upstream of the outlet, March 9, 2017. B – Outlet of Otter Brook discharging into the Little Southwest Miramichi River, March 9, 2017.

C – The spatial extent of the likely Otter Brook plume along the south-eastern bank of the main river, March 9, 2017.

The Miramichi River is noted for Atlantic Salmon (Salmo salar) and Brook Charr or Trout (Salvenius fontinalis) (Cunjak and Newbury, 2005). These coldwater salmonids are highly prized and managed, but the threat of a changing climate has already impacted neighbouring populations to the south and is likely affecting the Miramichi River populations as well (e.g., Monk and Curry, 2009; Daigle et al. 2015). Studies in the

Miramichi River watershed are similarly demonstrating that salmon and charr have a dependence on summer thermal refugia created by zones of groundwater discharge or groundwater dominated tributaries, e.g., Breau et al., (2011), and Corey et al., (2019).

Hence, the importance to managers of identifying groundwater inputs across broad scales to protect the thermal refugia necessary for these coldwater ecotherms (Kurylyk et al.,

2014). Our study examines 208 km of river across 7 tributaries of the Miramichi River watershed (Figure 4-2).

137

Figure 4-2 The Miramichi River watershed, NB, Canada, and rivers studied, where:

1 = North Branch of the Southwest Miramichi River; 2= McKiel Lake; 3 = Burnthill

Brook; 4 = Clearwater Brook; 5 = Little Southwest Miramichi River; 6 = Southwest

Miramichi River (and meteorological station), and 7 = Cains River.

4.2.2 Thermal Infrared Imagery

The TIR (8-9.5 µm) data were collected by a FLIR Systems SC3000 LWIR, mounted to a gyro-stablized unit attached to a helicopter (Bell Long Ranger). One image was collected every second along with a geotag, resulting in an image over-lapped of 40-

70% (R. Faux, Quantum Spatial, unpublished data). In-stream thermographs were 138

deployed to calibrate TIR data, where radiant temperatures are calibrated based on the ground truthed kinetic temperatures. Post calibration, images were geo-referenced and orthorectified with a pixel resolution of 0.6m (see Table 1 - Monk et al., 2013). The final

TIR data are classified at 0.1 oC intervals. TIR data was collected between 22 – 24 July

2008 for Burnthill Brook (Bb), Cains River (Cr), Little Southwest Miramichi River - lower reach (LSw-L), McKiel Lake (McKiel), and the Southwest Miramichi River (Sw), while

TIR for Clearwater Brook (Cb), Little Southwest Miramichi River – upper (LSw-U), and

North Branch of the Southwest Miramichi River (NBSw) was acquired on the 29 July 2009

(Figure 4-3).

To delineate the spatial extent of thermal plumes in reaches examined, we used the raster math ‘condition’ function in ArcMap to identify pixels within the TIR data set with a > 0.5 oC (Torgersen et al., 2001) temperature differential compared to the main bulk temperature as determined by neighbouring pixels with the TIR data suite. We visually classified the groundwater source as ‘tributary’ – inputs from a stream (Figure 4-3 A),

‘alcoves’ - inputs from backwater channels (Figure 4-3 B), and ‘seeps’ – inputs coming directly from the river bank (Figure 4-3 C).

139

Figure 4-3 Three examples of detected groundwater contributing areas showing the aerial, red-green-blue (RGB) image, corresponding thermal infrared image (TIR), and winter satellite image (Satellite) with polygons delimiting the contributing areas.

A is a tributary input, B is a backwater channel, or alcove, and C is a seepage zone.

4.2.3 Satellite imagery

Over the past few decades, the resolution (defined by pixel size) of commercial satellite imagery has increased exponentially (e.g., Legleiter and Overstreet, 2012). The evolution of these high-resolution data is mostly confined to ≈ 0.4 – 2.3 µm range, i.e., not inclusive of the thermal range (≈ 8-14 µm), e.g., DigitalGlobe WorldView 3. Pixel 140

resolution for panchromatic imagery is now available at <1 m scales, e.g., DigitalGlobe,

WorldView – 4 captures images at ≈ 30 cm resolution and such are available for purchase.

Further, Google Earth Pro hosts a suite of high-resolution imagery including historical images that are publicly available. We utilised these publically available Google Earth Pro images for our analysis. We selected images with the condition that ice conditions exist

(ICE). Herein we suggest ICE occurs during the ice-cover period, i.e., mid or late winter before the initiation of spring ice melt, and later we discuss the processes that lead to the ice conditions we capitalize upon. We searched Google Earth Pro for ICE images that overlapped with our TIR images. We identified ICE conditions on 3 April 2014 across all sites excluding the LSw–L which only had an image from 2 February 2017. We georeferenced the satellite images in ArcMap with a resulting pixel resolution from 0.3 –

1 m (Table 4-1). In an effort to equalize the cell sizes for both TIR and satellite images, we resampled the satellite images at 0.6 m using the ArcMap ‘resampling’ tool, applying a ‘nearest’ resampling technique (ESRI, 2018). Following resampling, we manually inspected the images for excessive shadowing and clipped the shadows from the satellite image in order to prevent misclassification in later steps.

Table 4-1 River, date, and resolution of acquired image, along with resampled satellite image resolution (see text).

141

Thermal infrared (TIR) Satellite (ICE) ICE resampled River Date/ Resolution (m2) Date/ Resolution (m2) Resolution (m2)

Bb 23 July 2008 / 0.6 3 April 2014 / 0.45-1.00 0.6 Cr 23 July 2008 / 0.6 3 April 2014 / 0.30-0.85 0.6 Cb 29 July 2009 / 0.6 3 April 2014 / 0.38-1.00 0.6 LSw-L 23 July 2008 / 0.6 2 Feb 2017 / 0.42-0.90 0.6 LSw-U 29 July 2009 / 0.6 3 April 2014 / 0.42-0.83 0.6 McKiel 24 July 2008 / 0.6 3 April 2014 / 0.95 0.6 NBSw 28 July 2009 / 0.6 3 April 2014 / 0.37 0.6 Sw 23 July 2008 / 0.6 3 April 2014 / 0.42 0.6

4.2.4 Maximum likelihood classification (MLC)

To classify possible groundwater inputs and ice across the suite of the ICE

(satellite) images, we executed a supervised maximum likelihood classification algorithm

(sMLC) in ArcMap. sMLC is a method utilised in land use classification (Townshend et al., 1987), forest classification (Brosofske et al., 2014), and water body classification

(Verpoorter et al., 2012), and relies on Bayes theory (ESRI, 2018). Sun et al., (2013) explain that a normal distribution is assumed for training samples. During classification, unclassified pixels are subscribed a classification based on the likelihood of falling into the probability density function of a training category. Our training dataset included 1,542 pixels for groundwater and 137,094 for ice, where each pixel is 0.6 m and distributed randomly throughout our satellite data in an effort to capture all variability.

We assessed the accuracy of the sMLC by calculating the model’s Kappa coefficient (k). The k is utilised in image classification and follows the form:

푝 −푝 푘 = 표 푒 (1) 1−푝푒 142

Where po = observed proportional agreement, and pe = the expected agreement by change and,

1 푝 = ∑푔 푓 (2) 표 푛 푖=1 푖푖 while,

1 푝 = ∑푔 푓 푓 (3) 푒 푛2 푖=1 푖+ +푖

where fi+ is the total for the ith row, and f+i is the total for the ith column.

Jensen (1996) explains a k value > 0.8 is suggestive of strong agreement between the classified image and the reference data, such as ground control point or a manual inspected point. A k value between 0.4 – 0.8 signifies moderate agreement between the classified image and the reference data, and a k value < 0.4 denotes a poor agreement between the classified image and the reference data.

Considering no satellite image ground control points were available, we visually examined the ICE image for ice discontinuities, i.e., possible groundwater inputs, and ice cover to develop a validation dataset for the kappa coefficient analysis. To execute this, we constructed points in ArcMap and manually assigned them a class dependent on the hydrological state they represent in the ICE image (free flowing or frozen – possible groundwater or ice) - a total of n=62 points: where n=30 were classified as possible groundwater, and n=32 were classified as ice-cover. Using these points, we extracted classes from our sMLC classification. This was completed by converting the sMLC layer from a polygon to a raster, and using the ‘extract values’ tool in ArcMap, to extract values 143

(or classifications) from the sMLC raster dataset to our validation points. TIR inputs were visually inspected and possible groundwater inputs were classified as alcove, seep, or tributary. To compare observations of groundwater inputs in both TIR (0.5 oC boundary condition derived) and ICE (sMLC derived), we conducted a Chi square test (χ 2) with an

α = 0.05. All statistical tests were performed in Microsoft Excel (2018). A workflow of classification techniques for both TIR and ICE images in illustrated in Figure 4-4.

144

Figure 4-4 Illustration of the work flow for processing TIR (left) and Satellite (right) images. TIR (1) is airborne-derived TIR image, TIR (2) is the delineated groundwater inputs based on a boundary condition of <0.5 oC from the mainstem, and TIR (3) displays the groundwater inputs (red) overlaid on the aerial image. Satellite (1) is the satellite image where ICE is met. Satellite (2) is ICE with shadows clipped. Satellite

(3) is the delineation of the training data (red) for a supervised, maximum likelihood classification (sMLC) of ice cover. Satellite (4) is the sMLC-derived classification.

Satellite (5) displays the predicted, groundwater inputs (blue) on the aerial image, where TIR (3) and ICE (5) are the same image.

4.3 Results

4.3.1 Supervised maximum likelihood classification (sMLC)

An excellent relationship was observed between the sMLC derived classification for ice- cover and possible groundwater inputs and the manually derived classifications in the ICE imagery – k =96.8. Moreover, possible groundwater inputs were classified with 100% accuracy in the sMLC model – when compared to the manually determined classification; while the sMLC model classified ice-cover with 96.88 % accuracy.

4.3.2 TIR and ICE input detections

The TIR boundary condition (> 0.5 oC difference from mainstem) resulted in the identification of 45 assumed, groundwater inputs across the seven rivers (Figure 4-2), i.e.,

145

tributaries (‘tribs’) n = 19, bank seepage zones (‘seep’) n = 17, and back-water zones

(‘alcove’) n = 9 (Table 4-2). The ICE sMLC analysis identified 48 assumed, groundwater- derived sites across all the eight sampled areas: n = 17 tributaries (89% of TIR delineated sites), n = 21 bank seepage zones (100% plus 3 additional sites), and n = 10 alcoves (100% plus 1 additional site) (Table 4-2). The χ2 analysis indicated that ICE detections of groundwater inputs matched the TIR detections (χ2 = 0.489, df = 2, p = 0.78). At most locations, the area of the thermal plume was greater in the ICE image (Table 4-2). As an example, Figure 4-5 shows a site on the Cains River where groundwater input from a valley bottom, alcove extends along a much longer zone in winter.

Table 4-2 Summary of TIR and ICE inputs detected and the estimated, total areal extent of groundwater input zones among study sites (see text).

146

TIR ICE River Input Inputs Estimated area Inputs Estimated area detected (m2) detected (m2) Bb Trib 2 208 2 930 Seep 0 0 1 424

Alcove 0 0 1 484

Cr Trib 10 2,042 10 4,235 Seep 7 207 8 1,014

Alcove 1 4,088 1 3,984 Cb Trib 3 1,484 3 826 Seep 3 143 4 562

Alcove 6 1,196 6 5,906 LSw-L Trib 0 ------Seep 4 246 4 260 Alcove 1 603 1 931

LSw-U Trib 2 667 2 1,129 Seep 2 58 3 68 Alcove 0 ------

McKiel Trib 1 256 0 0 Seep 0 ------Alcove 0 ------

NBSw Trib 1 454 0 0 Seep 0 ------Alcove 1 336 1 445

Sw Trib 0 ------Seep 1 6 1 14 Alcove 0 ------

TOTAL 45 11,994 48 21,212

147

Figure 4-5 An example of the groundwater contributing area as delineated by TIR and ICE in the main Cains River (alcove-derived input). (A) ICE image showing the extent of a thermal plume during ice cover (3rd April, 2014). (B) TIR image showing the thermal plume during summer (23rd July, 2008). (C) The comparison of the TIR

(red) and ICE (blue) estimates of the groundwater input zones.

148

4.4 Discussion

4.4.1 TIR and ICE groundwater detections

It is evident that satellite images in winter, when ice conditions exist, ICE can accurately detect thermal anomalies that are consistent with groundwater inputs detected using mid-summer TIR data. The sMLC model results give confidence that landscape scale classifications of possible groundwater inputs can be achieved using a relatively simple and time efficient, machine-learning technique, hence removing the laborious and time-consuming task of manually delimiting possible groundwater and ice-cover from ICE images. That the ICE satellite images detected 89-100% of tributary, bank seepage, and alcove (backwater) sourced thermal anomalies, indicates that resource managers have an accessible tool to locate potential thermal habitats (e.g., Curry and Noakes, 1995) or refugia

(Wilbur, 2012). Another potential application of ICE images is aufeis, or naled, mapping with fine resolution satellite imagery. Here, rather than identifying ‘ice free conditions’ in the winter, ICE images could be utilised to delimit ice ‘build-up’ along the river channel due to aufeis formations – formed by the freezing of groundwater overflow (Wanty et al.,

2007). Work by Morse and Wolfe, (2015); Pavelsky and Zarnetske, (2017) detailed how coarse resolution (10-30 m pixel size) satellite imagery can be used to identify aufeis formations. However, with fine resolution satellite imagery (e.g., WV-3, 30 cm pixels), it is likely possible to identify fine-scale localised aufeis build up, and thereby, discrete groundwater inputs.

149

The tool can be used simply to identify sites, but also has some potential to spatially quantify thermal habitats. We demonstrated spatial differences of mapped inputs between the TIR and ICE methods. The root of this differential is likely multifaceted. There are temporal factors to consider: the TIR data was acquired in 2008/09, while the ICE images were collected in 2014 and 2017. Inter-annual variability in aquifer conditions such as antecedent wetness are likely (Schneider, 1961), and Dugdale et al., (2013) highlighted the spatiotemporal variability in possible thermal refugia due to such annual variances.

Further, it is well established that groundwater thermal regimes display lagging behaviour with seasonal air temperature variations; (see Alexander and Caissie, 2003; Kurylyk et al.,

2013). Solar radiation shading is known to reduce temperature in small tributaries

(Johnson, 2004). Therefore, it is also possible that the differences are related to tributaries cooled by solar radiation interception rather than groundwater inputs in the summer; thus, freeze in the winter (e.g., McKiel and NBSw – Table 4-2). However, in-season ice conditions at a site are likely the more significant factor and these are a function of both base flow and surface air conditions. Long, cold periods will grow surface ice as well as ice bridges over running waters, which can be further expanded by snowfall, or drifting, and thus minimizing the usefulness of satellite images in the temperate regions of eastern

Canada where the Miramichi River lies, the best ICE conditions are predicted from late

February to April. Note however that in-season conditions vary amongst regions and even sites, as we demonstrated with the February satellite image for one site (LSw-U).

We quantified the spatial variability of the thermal anomalies in summer and winter and the satellite images typically generated larger areas of thermal impacts at a site (Table

150

4-2). The difference has important consequences for fish, as they behave differently in response to thermal habitats in winter and summer (e.g., Linnansaari and Cunjak, 2013;

French et al., 2014). In addition, we know very little about water that flows under ice in smaller rivers, and therefore the value of the winter thermal anomalies that are apparent in the ICE images as discussed in Cunjak (1996). We are beginning to consider the hydrological processes leading to the occurrence and persistence of these winter anomalies to best inform both research and management of fish and other species during winter in rivers.

These ice cover conditions can also tell us about the thermo-mechanical processes occurring in rivers. It is well established that density differentials between cold groundwater inputs and ambient river temperatures in the summer can lead to thermal stratification, where warmer water sits on top of colder, denser water (Torgersen et al.,

1999). This is a particular limitation of TIR, which only measures the top 1 cm of the water column (Handcock et al., 2012). In some instances, this thermal stratification could explain the difference between thermal plume lengths in summer and winter (Table 4-2). This also suggests that ICE defined thermal plumes may be a more valid representation of depth- averaged temperature for groundwater inputs. However, extensive field work by Wilbur

(2012) established that Cains River TIR dataset represents a river which is vertically well mixed and that little, to no, thermal stratification existed (differences < 1 oC through the water column). Therefore, we suggest this points towards a different mechanism underpinning thermal dispersion – at least in the Cains River - for discrete groundwater inputs. We use Site 7 (Cains River reach) – Figure 4-5 - to demonstrate the potential

151

mechanism of how thermal conditions develop and persist (Figure 4-6, and 4-6B). The state change of a river from free flowing to frozen leads to a change in its discharge regime, and its frictional regime – (see Newbury, 1968 for an overview). As surface ice develops, an increase in friction occurs at the river/surface ice interface (Hoyt, 1913); coupling friction at the river/substrate interface, and a reduction of discharge to base flow levels, leads to a velocity decrease (Figure 4-6C). The frozen state modifies the water’s viscosity

(inverse relationship to temperature) and thus affects the turbulence regime of the river.

The relationship between turbulence (or mixing) and velocity and viscosity can be simply explained by the Reynolds number (Re). Here, low Re values relate to less turbulent flow regimes, and higher Re values identify more turbulent flow regimes (see Equation 4).

푢퐿 푅 = (4) 푒 휇

Where Re is the Reynolds number, u = velocity of the fluid (m/s), L is a linear dimension

(m), and µ is the fluids viscosity (m2∙s-1).

Additionally, Kurylyk et al., (2016) provided an excellent overview on the relationship between streambed heat advection, discharge and temperature, and therefore depth, cross-sectional width and velocity (in particular see equations 2 and 3 in Kurylyk et al., 2016). Considering this relationship, it can be shown during base flows regimes, and under ice conditions, a fluids velocity (u) will decrease, while a reduction in a fluids temperature leads to an increase in a fluids viscosity (µ), thereby reducing Re, hence a less turbulent flow regime, and less dispersion. In Figure 4-5 and 4-6, the longer thermal plume is likely driven by a less turbulent flow regime, thus, reducing dispersion, and resulting in

152

a longer zone of influence from the groundwater inputs in the winter (Figure 4-6D). Further modifying the spatial extent of the thermal plume is heat gained by solar radiation absorption in the open plume (no ice cover) and the emissivity of open water. The heat gained by solar radiation will mitigate, to some degree, heat loss via conduction and the cooler air preventing ice formation on the plume (Figure 4-6E). The plume will also emit longwave radiation (Figure 4-6F), which also limits ice formation at the plume. Ice formation develops on the plume when the magnitude of heat transfer by air conduction and evaporation overcomes heat gains, i.e., the length of the yellow arrow and dampening in a downstream direction in Figure 4-6G.

Figure 4-6 (A) the spatial extent of a winter thermal plume (as shown in Figure 4-5).

(B) a sample section to demonstrate thermo-mechanical processes occurring at the groundwater input/ river interface. (C) shows the velocity distribution with depth considering friction at both the substrate and ice interface and a theoretical maximum velocity occurring at the mid-point between ice and substrate. (D) shows advective 153

heat dispersion occurring only in the free flowing plume. (E) shows solar radiation reflectance (green arrow) and absorption (red arrow), where more absorption occurs in the open plume (O.P.) leading to heat gains. (F) shows shortwave absorption (green arrow) in the plume (but minimized on ice) and longwave emittance (black arrow) from the plume into the river ice and (G) shows heat gain from the groundwater input to the frozen river with the magnitude of heat gain illustrated by the size of the yellow arrow, where the length of the arrow denotes a greater heat transfer, dampening in a downstream direction with air conductance and evaporation.

In the summer, the frictional regime of the river reverts to that of a free flowing system, i.e., friction at the substrate/ river interface and minimal friction at the water’s surface (Newbury, 1968), while discharge increases across the river (excluding drought conditions) (see Figure 4-7C). Coupling the reduction of friction and an increase in discharge during summer, leads to an increase in the river’s velocity (Figure 4-7C), and the increase in air temperature leads to an increase in river temperature and a decrease in the rivers viscosity (Newbury 1968). Once again examining equation (4) it can be shown that an increase in u with a decrease in µ leads to an increase in Re and is indicative of a more turbulent flow regime. This increase in turbulence leads to a more dispersive environment, thus a higher rate of dispersion between the warm, free flowing water and the thermal plume (Kurylyk et al., 2016) -Figure 4-7D. Inversely, heat gained by solar radiation in the summer limits the spatial extent of the thermal plume (Figure 4-7E) with further impedance by heat gains from longwave emittance from the free flowing river (Figure 4-7F). Due to high turbulence regimes which persist in the free flowing river, the heat exchange between

154

the groundwater input and the mainstem leads to quicker dispersion of the thermal plume

– Figure 4-7G. As a result, a shorter thermal plume seems most probable in the summer.

Figure 4-7 (A) is the spatial extent of a summer thermal plume (as shown in Figure 4-

6). (B) is a sample section to demonstrate thermo-mechanical processes occurring at the groundwater input/ river interface. (C) shows the velocity distribution with depth considering friction at the substrate and a theoretical maximum velocity occurring at just below the water surface. (D) shows advective heat dispersion occurring across the entire river. (E) shows solar radiation reflectance (green arrow) and absorption

(red arrow) leading to heat gains in both the free flowing river (F.F.) and O.P. (F) shows shortwave absorption (green arrow) and longwave emittance (black arrow) across the entire river and (G) shows heat gain from the groundwater input to the

F.F. river with the magnitude of heat gain illustrated by the size of the yellow arrow,

155

where the length of the arrow denotes a greater heat transfer, dampening in a downstream direction with air conductance and evaporation.

4.4.2 Limitations

ICE performs favourably at delimiting groundwater inputs, but turbulent flow reaches can also create ice-free zones in the satellite images. Reaches with turbulent flow and thus frictional heating and heat loss (see for example Newbury, 1968), e.g., areas of steep banks and exposed bedrock or shallow unconsolidated overburden, flow conditions can limit ice cover (Woo, 2012). As an example, in the North Pole Stream (Figure 4-8) no groundwater inputs are detected in the summer TIR image (Figure 4-8C) and the winter satellite image shows unfrozen sections along the mid-channel, which co-occurs with a confining, bedrock section (Figure 4-8D). This is also highlighted in Figure 4-4, where the sMLC classified an open section as a potential groundwater input; however, this ice free section is almost certainly a function of fluid friction (personal communication Bob

Newbury). This situation reminds us that while remote sensing can be a useful tool of water resources managers, a good understanding of a region’s physiography and hydrological processes are required to ensure correct interpretation of image-based analyses (Pietroniro and Prowse, 2002). Lastly, ICE also requires cloud free images.

156

Figure 4-8 (A) an aerial image of a steep confined bedrock channel (gold arrows) in the North Pole Stream – located on the Miramichi Highlands, (B) the slope map of the associated reach, (C) associated TIR imagery (29 July, 2009) draped over a high- resolution hillshade model with a noted lack of groundwater inputs, and (D) ICE image (2 February, 2017) showing unfrozen sections (red arrows) which co-occur in the steep confined bedrock channel section.

157

It is worth noting, both ICE, and TIR methods, can be limited in their ability to identify groundwater inputs, as both methods represent a snapshot in time. This limitation can be underpinned by aquifer dynamics, and may be especially important for shallow, localised aquifers (see Vidon and Smith, 2007; Dugdale et al., 2013), and that are responsive to contemporary air temperature fluxes (see Kurylyk et al., 2014; Brigg et al.,

2017). However, for groundwater inputs originating from deeper flow paths, this limitation may be moot, as deeper flows are less responsive to contemporary climatic fluxes (e.g.,

Kurylyk et al., 2013).

4.5 References

Alexander, M.D., and Caissie, D. 2003. Variability and comparison of hyporheic water

temperatures and seepage fluxes in a small Atlantic Salmon stream. doi:DOI:

10.1111/j.1745-6584.2003.tb02570.x.

Anderson, M.P. 2005. Heat as a Groundwater Tracer. Groundwater 43(6): 951–968.

Wiley/Blackwell (10.1111). doi:10.1111/j.1745-6584.2005.00052.x.

Arenson, L., Hoelzle, M., and Springman, S. 2002. Borehole deformation measurements

and internal structure of some rock glaciers in Switzerland. Permafr. Periglac.

Process. 13(2): 117–135. Wiley-Blackwell. doi:10.1002/ppp.414.

Breau, C., Cunjak, R.A., and Peake, S.J. 2011. Behaviour during elevated water

temperatures: can physiology explain movement of juvenile Atlantic Salmon to

cool water? J. Anim. Ecol. 80(4): 844–853. John Wiley & Sons, Ltd (10.1111).

doi:10.1111/j.1365-2656.2011.01828.x. 158

Briggs, M.A., and Hare, D.K. 2018. Explicit consideration of preferential groundwater

discharges as surface water ecosystem control points. Hydrol. Process. 32(15):

2435–2440. Wiley-Blackwell. doi:10.1002/hyp.13178.

Briggs, M.A., Johnson, Z.C., Snyder, C.D., Hitt, N.P., Kurylyk, B.L., Lautz, L., Irvine,

D.J., Hurley, S.T., and Lane, J.W. 2018. Inferring watershed hydraulics and cold-

water habitat persistence using multi-year air and stream temperature signals. Sci.

Total Environ. 636: 1117–1127. Elsevier.

doi:10.1016/J.SCITOTENV.2018.04.344.

Briggs, M.A., Lane, J.W., Snyder, C.D., White, E.A., Johnson, Z.C., Nelms, D.L., and Hitt,

N.P. 2017. Shallow bedrock limits groundwater seepage-based headwater climate

refugia. Limnol. - Ecol. Manag. Inl. Waters. Urban & Fischer.

doi:10.1016/J.LIMNO.2017.02.005.

Brosofske, K.D., Froese, R.E., Falkowski, M.J., and Banskota, A. 2014. A Review of

Methods for Mapping and Prediction of Inventory Attributes for Operational Forest

Management. For. Sci. 60(4): 733–756. Oxford University Press.

doi:10.5849/forsci.12-134.

Brunner, P., Therrien, R., Renard, P., Simmons, C.T., and Franssen, H.-J.H. 2017.

Advances in understanding river-groundwater interactions. Rev. Geophys. 55(3):

818–854. John Wiley & Sons, Ltd. doi:10.1002/2017RG000556.

Callahan, M.K., Rains, M.C., Bellino, J.C., Walker, C.M., Baird, S.J., Whigham, D.F., and

King, R.S. 2015. Controls on Temperature in Salmonid-Bearing Headwater

Streams in Two Common Hydrogeologic Settings, Kenai Peninsula, Alaska.

159

JAWRA J. Am. Water Resour. Assoc. 51(1): 84–98. John Wiley & Sons, Ltd

(10.1111). doi:10.1111/jawr.12235.

Constantz, J. 1998. Interaction between stream temperature, streamflow, and groundwater

exchanges in alpine streams. Water Resour. Res. 34(7): 1609–1615.

doi:10.1029/98WR00998.

Corey, E., Linnansaari, T., Dugdale, S. J., Bergeron, N., Gendron, J., Lapointe, M., &

Cunjak, R. A. (2019). Comparing the behavioural thermoregulation response to

heat stress by Atlantic Salmon parr ( Salmo salar ) in two rivers. Ecology of

Freshwater Fish, eff.12487. https://doi.org/10.1111/eff.12487

Cunjak, R.A. 1996. Winter habitat of selected stream fishes and potential impacts from

land-use activity. Can. J. Fish. Aquat. Sci. 53(S1): 267–282. NRC Research Press

Ottawa, Canada . doi:10.1139/f95-275.

Cunjak, R.A., and Caissie, D. 1993. Proceedings of the 7th Workshop on the Hydraulics

of Ice Covered Rivers (1993). In Frazil ice accumulation in a large salmon pool in

the Miramichi River, New Brunswick: Ecological implications for overwintering

fishes. Saskatoon, Saskatchewan. pp. 279–295.

Cunjak, R.A., Prowse, T.D., and Parrish, D.L. 1998. Atlantic Salmon ( Salmo salar ) in

winter: "the season of parr discontent"? Can. J. Fish. Aquat. Sci.

55(S1): 161–180. NRC Research Press Ottawa, Canada . doi:10.1139/d98-008.

Cunjak, R. a., and Newbury, R.W. 2005. Atlantic Coast Rivers of Canada. Rivers North

Am.: 938–980. doi:10.1016/B978-012088253-3/50024-9.

Curry, R.A., and Devito, K.J. 1996. Hydrogeology of brook trout ( Salvelinus fontinalis)

spawning and incubation habitats: implications for forestry and land use

160

development. Can. J. For. Res. 26(5): 767–772. NRC Research Press Ottawa,

Canada. doi:10.1139/x26-086.

Curry, R.A., and Noakes, D.L.G. 1995. Groundwater and the selection of spawning sites

by brook trout (Salvelinus fontinalis). Can. J. Fish. Aquat. Sci. 52(hypothesis 1):

1733–1740. doi:10.1139/f95-765.

Daigle, A., Jeong, D. Il, and Lapointe, M.F. 2015. Climate change and resilience of

tributary thermal refugia for salmonids in eastern Canadian rivers. Hydrol. Sci. J.

60(6): 1044–1063. doi:10.1080/02626667.2014.898121.

Dugdale, S.J., Bergeron, N.E., and St-Hilaire, A. 2013. Temporal variability of thermal

refuges and water temperature patterns in an Atlantic Salmon river. Remote Sens.

Environ. 136: 358–373. Elsevier Inc. doi:10.1016/j.rse.2013.05.018.

Dugdale, S.J., Franssen, J., Corey, E., Bergeron, N.E., Lapointe, M., and Cunjak, R.A.

2016. Mainstem movement of Atlantic Salmon parr in response to high river

temperature. Ecol. Freshw. Fish 25(3): 429–445. Wiley/Blackwell (10.1111).

doi:10.1111/eff.12224.

Dugdale, S.J., Kelleher, C.A., Malcolm, I.A., Caldwell, S., and Hannah, D.M. 2019.

Assessing the potential of drone‐based thermal infrared imagery for quantifying

river temperature heterogeneity. Hydrol. Process. 33(7): 1152–1163. John Wiley &

Sons, Ltd. doi:10.1002/hyp.13395.

Esri: GIS Mapping Software, Spatial Data Analytics & Location Platform. (n.d.).

Available from https://www.esri.com/en-us/home [accessed 29 May 2018].

161

French, W. 2014. Protected from the Elements: Winter Ecology of Brown Trout in

Groundwater Buffered Streams. Available from

https://conservancy.umn.edu/handle/11299/174220 [accessed 29 January 2018].

French, W.E., Vondracek, B., Ferrington, L.C., Finlay, J.C., and Dieterman, D.J. 2014.

Winter feeding, growth and condition of brown trout Salmo trutta in a groundwater-

dominated stream. J. Freshw. Ecol. 29(2): 187–200. Taylor & Francis.

doi:10.1080/02705060.2013.847868.

Fullerton, A.H., Torgersen, C.E., Lawler, J.J., Faux, R.N., Steel, E.A., Beechie, T.J.,

Ebersole, J.L., and Leibowitz, S.G. 2015. Rethinking the longitudinal stream

temperature paradigm: region-wide comparison of thermal infrared imagery reveals

unexpected complexity of river temperatures. Hydrol. Process. 4737(June).

doi:10.1002/hyp.10506.

Groeneveld, D., Technical, T.G.-U.F.S.G., and 1985, undefined. (n.d.). Interdependence

of groundwater, riparian vegetation, and streambank stability: a case study.

fs.fed.us. Available from

https://www.fs.fed.us/rm/pubs_rm/rm_gtr120/rm_gtr120_044_048.pdf [accessed

16 May 2018].

Handcock, R.N., Torgersen, C.E., Cherkauer, K. a., Gillespie, A.R., Tockner, K., Faux,

R.N., and Tan, J. 2012. Thermal Infrared Remote Sensing of Water Temperature in

Riverine Landscapes. Fluv. Remote Sens. Sci. Manag.: 85–113.

doi:10.1002/9781119940791.ch5.

Hare, D.K., Briggs, M.A., Rosenberry, D.O., Boutt, D.F., and Lane, J.W. 2015. A

comparison of thermal infrared to fiber-optic distributed temperature sensing for

162

evaluation of groundwater discharge to surface water. J. Hydrol. 530: 153–166.

Elsevier B.V. doi:10.1016/j.jhydrol.2015.09.059.

Harwood, L.A., and Babaluk, J.A. 2014. Spawning, Overwintering and Summer Feeding

Habitats Used by Anadromous Arctic Char (Salvelinus alpinus) of the Hornaday

River, Northwest Territories, Canada. ARCTIC 67(4): 449.

doi:10.14430/arctic4422.

Hoyt WG. (1913) The effects of ice on stream flow. US Government Printing Office.

IPCC — Intergovernmental Panel on Climate Change. (n.d.). Available from

https://www.ipcc.ch/ [accessed 7 May 2019].

Isaak, D., Luce, C., Rieman, B., Nagel, D., Peterson, E., Horan, D., Parkes, S., and

Chandler, G. 2009. Effects of Climate Change and Recent Wildfires on Stream

Temperature and Thermal Habitat for Two Salmonids in a Mountain River

Network. Ecol. Appl. 20(5): 100319061507001. doi:10.1890/09-0822.

Isaak, D.J., Young, M.K., Nagel, D.E., Horan, D.L., and Groce, M.C. 2015. The cold-water

climate shield: delineating refugia for preserving salmonid fishes through the 21st

century. Glob. Chang. Biol. 21(7): 2540–2553. doi:10.1111/gcb.12879.

Jensen, J.R., and R., J. 1996. Introductory digital image processing : a remote sensing

perspective. Prentice Hall. Available from

https://dl.acm.org/citation.cfm?id=526335 [accessed 28 January 2019].

Johnson, S.L. 2004. Factors influencing stream temperatures in small streams: substrate

effects and a shading experiment. Can. J. Fish. Aquat. Sci. 61(6): 913–923. NRC

Research Press Ottawa, Canada . doi:10.1139/f04-040.

163

Johnson, Z.C., Snyder, C.D., and Hitt, N.P. 2017. Landform features and seasonal

precipitation predict shallow groundwater influence on temperature in headwater

streams. Water Resour. Res. 53(7): 5788–5812. Wiley-Blackwell.

doi:10.1002/2017WR020455.

Kurylyk, B.L., Bourque, C.P.-A., and MacQuarrie, K.T.B. 2013. Potential surface

temperature and shallow groundwater temperature response to climate change: an

example from a small forested catchment in east-central New Brunswick (Canada).

Hydrol. Earth Syst. Sci. 17(7): 2701–2716. doi:10.5194/hess-17-2701-2013.

Kurylyk, B.L., MacQuarrie, K.T.B., Caissie, D., and McKenzie, J.M. 2014. Shallow

groundwater thermal sensitivity to climate change and land cover disturbances :

derivation of analytical expressions and implications for stream temperature

projections. Hydrol. Earth Syst. Sci. Discuss. 11: 12573–12626. doi:10.5194/hessd-

11-12573-2014.

Kurylyk, B.L., Moore, R.D., and MacQuarrie, K.T.B. 2016. Scientific briefing: quantifying

streambed heat advection associated with groundwater-surface water interactions.

Hydrol. Process. 30(6): 987–992. John Wiley & Sons, Ltd. doi:10.1002/hyp.10709.

Larocque, M., and Broda, S. 2016. Groundwater–surface water interactions in Canada.

Can. Water Resour. J. / Rev. Can. des ressources hydriques 41(4): 451–454. Taylor

& Francis. doi:10.1080/07011784.2016.1176537.

Legleiter, C.J., and Overstreet, B.T. 2012. Mapping gravel bed river bathymetry from

space. J. Geophys. Res. Earth Surf. 117(F4). John Wiley & Sons, Ltd.

doi:10.1029/2012JF002539.

164

Linnansaari, T., and Cunjak, R.A. 2010. Patterns in apparent survival of Atlantic Salmon

(Salmo salar) parr in relation to variable ice conditions throughout winter. Can. J.

Fish. Aquat. Sci. 67(11): 1744–1754. doi:10.1139/F10-093.

Malcolm, I.A., Soulsby, C., Youngson, A.F., Hannah, D.M., McLaren, I.S., and Thorne,

A. 2004. Hydrological influences on hyporheic water quality: implications for

salmon egg survival. Hydrol. Process. 18(9): 1543–1560. John Wiley & Sons, Ltd.

doi:10.1002/hyp.1405.

Monk, W. a., Wilbur, N.M., Allen Curry, R., Gagnon, R., and Faux, R.N. 2013. Linking

landscape variables to cold water refugia in rivers. J. Environ. Manage. 118: 170–

176. Elsevier Ltd. doi:10.1016/j.jenvman.2012.12.024.

Monk, Wendy A., and R. Allen Curry. "Models of past, present, and future stream

temperatures for selected Atlantic Salmon rivers in Northeastern North America."

American Fisheries Society Symposium. Vol. 69. 2009.

Moore, R.D., Spittlehouse, D.L., and Story, A. 2005. Riparian microclimate and stream

temperature response to forest harvesting: A review. J. Am. Water Resour. Assoc.

41(4): 813–834. Blackwell Publishing Ltd. doi:10.1111/j.1752-

1688.2005.tb03772.x.

Newbury, R. 1968. The Nelson River: a study of subarctic river processes. Johns Hopskins.

Available from https://www.worldcat.org/title/nelson-river-a-study-of-subarctic-

river-processes/oclc/30731228 [accessed 28 January 2019].

O’Sullivan, A.M., Devito, K.J., and Curry, R.A. 2019. The influence of landscape

characteristics on the spatial variability of river temperatures. CATENA 177: 70–

83. Elsevier. doi:10.1016/J.CATENA.2019.02.006.

165

Pavelsky, T.M., and Zarnetske, J.P. 2017. Rapid decline in river icings detected in Arctic

Alaska: Implications for a changing hydrologic cycle and river ecosystems.

Geophys. Res. Lett. 44(7): 3228–3235. John Wiley & Sons, Ltd.

doi:10.1002/2016GL072397.

Paznekas, A., and Hayashi, M. 2016. Groundwater contribution to winter streamflow in

the Canadian Rockies. Can. Water Resour. J. / Rev. Can. des ressources hydriques

41(4): 484–499. Taylor & Francis. doi:10.1080/07011784.2015.1060870.

Petty, J.T., Hansbarger, J.L., Huntsman, B.M., and Mazik, P.M. 2012. Brook Trout

Movement in Response to Temperature, Flow, and Thermal Refugia within a

Complex Appalachian Riverscape. Trans. Am. Fish. Soc. 141(4): 1060–1073. John

Wiley & Sons, Ltd. doi:10.1080/00028487.2012.681102.

Pietroniro, A., and Prowse, T.D. 2002. Applications of remote sensing in hydrology.

Hydrol. Process. 16(8): 1537–1541. John Wiley & Sons, Ltd.

doi:10.1002/hyp.1018.

Sass, G.Z., Creed, I.F., Riddell, J., and Bayley, S.E. 2014. Regional-scale mapping of

groundwater discharge zones using thermal satellite imagery. Hydrol. Process.

28(23): 5662–5673. John Wiley & Sons, Ltd. doi:10.1002/hyp.10068.

Sawaya, K.E., Olmanson, L.G., Heinert, N.J., Brezonik, P.L., and Bauer, M.E. 2003.

Extending satellite remote sensing to local scales: land and water resource

monitoring using high-resolution imagery. Remote Sens. Environ. 88(1–2): 144–

156. Elsevier. doi:10.1016/J.RSE.2003.04.006.

Schneider, R. 1961. Correlation of Ground- Water Levels and Air Temperatures in the

Winter and Spring in Minnesota. Washington.

166

Sun, J., Yang, J., Zhang, C., Yun, W., and Qu, J. 2013. Automatic remotely sensed image

classification in a grid environment based on the maximum likelihood method.

Math. Comput. Model. 58(3–4): 573–581. Pergamon.

doi:10.1016/J.MCM.2011.10.063.

Tague, C., Farrell, M., Grant, G., Lewis, S., and Rey, S. 2007. Hydrogeologic controls on

summer stream temperatures in the McKenzie River basin, Oregon. Hydrol.

Process. 21(24): 3288–3300. John Wiley & Sons, Ltd. doi:10.1002/hyp.6538.

Torgersen, C.E., Faux, R.N., McIntosh, B.A., Poage, N.J., and Norton, D.J. 2001. Airborne

thermal remote sensing for water temperature assessment in rivers and streams.

Remote Sens. Environ. 76(3): 386–398. doi:10.1016/S0034-4257(01)00186-9.

Torgersen, C.E., Price, D.M., Li, H.W., and McIntosh, B.A. 1999. Multiscale thermal

refugia and stream habitat associations of chinook salmon in northeastern Oregon.

Ecol. Appl. 9(1): 301–319. Wiley-Blackwell. doi:10.1890/1051-

0761(1999)009[0301:MTRASH]2.0.CO;2.

Townshend, J.R.G., Justice, C.O., and Kalb, V. 1987. Characterization and classification

of South American land cover types using satellite data. Int. J. Remote Sens. 8(8):

1189–1207. Taylor & Francis Group . doi:10.1080/01431168708954764.

Vidon, P., and Smith, A.P. 2007. Upland Controls on the Hydrological Functioning of

Riparian Zones in Glacial Till Valleys of the Midwest1. JAWRA J. Am. Water

Resour. Assoc. 43(6): 1524–1539. Blackwell Publishing Ltd. doi:10.1111/j.1752-

1688.2007.00125.x.

Wanty, R.B., Wang, B., Vohden, J., Day, W.C., and Gough, L.P. 2007. Aufeis

accumulations in stream bottoms in arctic and subarctic environments as a possible

167

indicator of geologic structure: Chapter F in Recent U.S. Geological Survey studies

in the Tintina Gold Province, Alaska, United States, and Yukon, Canada--results of

a 5-year project. In Scientific Investigations Report. doi:10.3133/SIR20075289F.

Wilbur, N. 2012. Characterizing thermal refugia for brook trout (Salvelinus fontinalis) and

Atlantic Salmon (Salmo salar) in the Cains River, New Brunswick, Canada.

University of New Brunwick.

Winter, T.C. 1995. Hydrological Processes and the Water Budget of Lakes. In Physics and

Chemistry of Lakes. Springer Berlin Heidelberg, Berlin, Heidelberg. pp. 37–62.

doi:10.1007/978-3-642-85132-2_2.

Winter, T.C. 2007. The Role of Groundwater in Generating Streamflow in Headwater

Areas and in Maintaining Base Flow1. JAWRA J. Am. Water Resour. Assoc. 43(1):

15–25. Wiley/Blackwell (10.1111). doi:10.1111/j.1752-1688.2007.00003.x.

Woo, M. 2012. Rivers in Cold Regions. In Permafrost Hydrology. Springer Berlin

Heidelberg, Berlin, Heidelberg. pp. 407–454. doi:10.1007/978-3-642-23462-0_9.

168

Chapter 5: Catchment-scale, high-resolution, hydraulic models and habitat maps – a salmonid’s perspective4

Abstract

The advent of remotely-sensed high-resolution imagery has led to the development of methods to map river bathymetry. In this study, we utilized high-resolution imagery to map river depth and quantify hydraulic habitats at the catchment-scale (> 1000 km2) during low flows. Using 0.3 m airborne multi-spectral imagery (resampled to 0.5 m), we mapped contiguous river depth (124 km) within a well-established Atlantic Salmon (Salmo salar) and Brook Trout (Salvelinus fontinalis) river – The Little Southwest Miramichi, New

Brunswick, Canada. We built image-derived depth maps with and without field data calibration. The model without field calibration data (flow resistance equation‐based imaging of river depths) accurately described river depths (R2 = 72.7; RMSE = 0.167 m; n

= 762); however, it overestimated shallow depths. The field-calibrated model removed shallow depth errors (R2 = 76.4; RMSE = 0.155 m; n = 762). We mapped velocity using a relationship between river geometry and discharge, and coalesced the field-calibrated depth and velocity maps to create Froude and Reynolds number maps. Finally, we performed an unsupervised classification model to delineate hydraulically relevant habitat units for

4 O’Sullivan, A.M., Wegscheider, B., Helminen, J., Cormier, J.G., Linnansaari, T., Wilson, D.A., Curry, R.A. (In press) Catchment-scale, high-resolution, hydraulic models and habitat maps – a salmonid’s perspective. Journal of Ecohydraulics. 10.1080/24705357.2020.1768600

169

salmonids. This approach provides an unprecedented view of catchment-scale hydraulic habitats that will advance both hydrological process research and river resources management.

Keywords: aerial imagery; Atlantic Salmon; Brook Trout; catchment-scale; FREEBIRD; habitat; remotely-sensed bathymetry

5.1 Introduction

The occurrence and persistence of aquatic biota is tightly coupled to their habitats

(Southwood 1977) and thus, the broad-scale aquatic habitats of rivers are often described by fluvial geomorphic structures (Magilligan et al., 2016; Katopodis and Kemp, 2018), flow regimes (Poff and Allan, 1997; Monk et al., 2011), and thermal regimes (O’Sullivan et al., 2019a; Fullerton et al., 2018). These habitats are, however, defined by their more localized hydraulic and thermal properties. For example, fish have flow and temperature preferences (Corey et al. 2019; Wilbur et al. 2020) and they select reproductive habitats that can be defined by hydraulic characteristics (Curry and Noakes 1995; Moir et al., 2006).

Strong associations with hydraulic parameters, defined by depth and velocity, also occur for benthic macroinvertebrates (Reid and Thomas, 2008; Shearer et al., 2015) and macrophytes (Riis and Biggs, 2003; Schoelynck et al., 2013). In spite of their importance, mapping these fine-scale processes at broad, catchment-scales has still largely eluded contemporary aquatic sciences (Hugue et al., 2016; Legleiter et al., 2009).

170

The development of high-resolution hydrodynamic models has facilitated flow simulations at reach-scales and across various flow conditions (Marzadri et al., 2013;

Fleischmann et al., 2019). The next challenge is modelling flow-habitat relationships across the larger riverscape or catchment scales, e.g., drainage areas > 1,000 km2 (Hauer et al., 2009; Harby et al., 2017). With the exponential growth in computing power and high-resolution, broad-scale mapping of river depth, it is only a matter of time before high- resolution, catchment-scale 3D hydrodynamic models are developed. Adjacent to hydrodynamic models, mesohabitat models offer a solution that is computationally less demanding and are based on physical and hydraulic attributes associated with flow such as velocity or discharge and water depth (Hauer et al., 2013; Dunbar et al., 2012). The mapping of meso-scale habitats can be broadly grouped into field-based, remote sensing, and hydrodynamic modelling approaches (Parasiewicz et al., 2013; Wilkes et al., 2015).

The mesohabitat approaches are promising, but still require significant field data which restricts their upscaling, particularly in remote, large rivers.

Image-derived depth maps have proven efficient in mapping both marine and freshwater systems (Weatherall et al., 2015; Marcus and Fonstad, 2008). The method relies on a relationship between spectral irradiance and water depth, where spectral signatures are correlated to water depth (Winterbottom and Gilvear, 1997; Legleiter et al., 2004).

Using these methods, depth along river reaches (m) to sections (km) have been mapped

(Kasvi et al., 2019; Hugue et al., 2016); however, these methods rely on the availability of field calibration data (Flener et al., 2012; Legleiter and Overstreet, 2012). Researchers have begun to test models which construct image-derived depth maps where field data are lacking or unavailable, e.g., hydraulically assisted bathymetry (HAB; Fonstad and Marcus 171

2005) and Flow REsistance Equation‐Based Imaging of River Depths (FREEBIRD;

Legleiter 2015). The advent of high-resolution airborne LiDAR also presents an opportunity to delineate aquatic habitats (see Hauer et al., 2009). Such methods may provide novel pathways to unlock high-resolution, image-based depth maps of rivers and the creation of habitat maps at unprecedented scales (e.g., Marcus and Fonstad, 2010;

Carbonneau et al., 2012).

Delineating flowing water habitats is typically based on subjective criteria such as

“deep” or “fast” water with uncertainty in repeatability and variation between different users (e.g., Eisner et al., 2005). Borsányi et al., (2004) defined mapping techniques and objective criteria or thresholds for slope, surface turbulence, velocity, and depth to delineate mesohabitats in Norwegian rivers. Supervised and unsupervised methods have been used to cluster areas with similar hydraulic properties (Legleiter and Goodchild, 2005;

Boruah et al., 2008). Hauer et al., (2009) used 2D hydrodynamic modelling and hydraulic thresholds for flow velocity, water depth, and bottom shear stress to simulate and map hydro-morphologic units such as riffles and pools. The definition of hydraulic thresholds involves an inherent degree of subjectivity, and thus unsupervised clustering algorithms such as k-means (Farò et al., 2018) and fuzzy clustering (Wallis et al., 2012) have also been used to delineate habitat heterogeneity in ecohydraulic studies.

Salmonids, such as Atlantic Salmon (Salmo salar) and Brook Trout (Salvelinus fontinalis), have evolved to optimize certain thermal and hydraulic habitats (Linnansaari and Cunjak, 2010; Curry and Devito, 1996). While many studies have focused on the effects of extreme temperatures on salmonids (Torgersen et al., 1995; Corey et al., 2017), 172

fewer studies have focused on hydraulic habitat use across age-classes through space and time during extreme low flows (Heggenes and Borgstrøm, 1991; Holm et al., 2001). Recent work by Wilbur et al., (2020) revealed that even during extreme temperature events, hydraulic parameters were important for characterising thermal refugia for both Atlantic

Salmon and Brook Trout, which varied between and within species. Similarly, population modelling efforts typically rely on coarse data to represent habitats across life stages, which seldom, if ever, account for hydraulic controls (see Amiro, 1983; DFO, 2018). This limitation identifies a critical pitfall for population dynamic modelling and stock assessment (Chaput et al., 2016), particularly for declining Atlantic Salmon populations

(DFO, 2018). An improved understanding of the composition of riverine thermal and hydraulic habitats is required, across all scales, to identify and preserve these critical habitats given the potential threat of a warming climate, and to aid management decisions.

While thermal infrared imagery has proven highly efficient for mapping catchment-scale river temperatures (e.g., Torgersen et al., 1999; Dugdale et al., 2015), catchment-scale high-resolution hydraulic habitat maps remain absent from river habitat research. To address this knowledge gap, we examine the utility of coupling high-resolution airborne imagery (0.3 m), passive depth-mapping techniques (with or without field calibration data), hydraulic models, and unsupervised classification methods to spatially characterise and classify hydraulic habitat heterogeneity that is relevant for Atlantic Salmon and Brook

Trout at the catchment-scale (> 1,000 km2).

173

5.2 Methods

5.2.1 Area of study

The Little Southwest Miramichi (LSW) River, known by local Mi’kmaq as

Tuadook (“a difficult, dangerous river”), is a tributary of the Miramichi River located in central New Brunswick (Figure 5-1a and b). The LSW drains an area ≈ 1,300 km2 and encompasses two physiographic regions, the Miramichi Highlands and the Maritime

Plains. The river is typical of a temperate, post glacial river, underlain by alluvium and shallow bedrock, and flanked by paleo-terraces (Ganong, 1906; Cunjak and Newbury,

2005). The headwaters of the LSW mainstem originate in a gentle relief upland located in the Miramichi Highlands, and are hydrologically characterised by lakes and wetlands

(mean river width ≈ 20 - 30 m). The North Pole Stream (or Kadunnatquegak – “Watching salmon in a pool”) is a headwater sub-catchment (Figure 5-1c), draining an area of 214 km

2 and is also hydrologically defined by lakes and wetlands (mean river width ≈ 8- 20 m).

The mid-sections of the LSW catchment are defined by a more confined channel (mean river width ≈ 60-80 m). Moving downriver, into the Maritime Plains section, the river becomes more sinuous, wide and shallow (mean width = 60-100 m; Corey et al., 2019).

Water clarity throughout the system during low flows is defined by little turbidity

(O’Sullivan, unpublished data); hence, qualifying the catchment as an ideal candidate for image derived depth mapping (Legleiter et al., 2004). The catchment has a northern temperate climate with annual precipitation ≈ 1,158 mm of which 300 mm falls as snow

(Cunjak and Caissie, 1993). The mean annual temperature is ≈ 4.1 oC (Cunjak and

Newbury, 2005). River discharge (Q) ranges from 6.81 m3s-1 (low flow) to 241 m3s-1 (2-

174

year flood) with a mean annual Q = 32.2 m3s-1 (Caissie, 2006). The catchment is well known for its Atlantic Salmon and sea-run Brook Trout (Cunjak and Newbury, 2005).

Here, our modelling efforts are focused along 124 km of river, which encompass the mainstem LSW and the North Pole Stream (Figure 5-1). A river gauge is located in the lower section of our study area and is the source of discharge data for our study (Figure 5-

1c; Environment Canada Hydrograph Station - 01BP001).

Figure 5-1 (a) Eastern Canada, where New Brunswick is defined by the black polygon.

(b) The Miramichi River Catchment (black polygon) and the Little Southwest

Miramichi (LSW – red polygon). (c) The LSW (red polygon) and the North Pole

Stream (green polygon) showing 124 km of the mapped, mainstem river (blue line), sites of the RTK dGPS (black bar), ADCP site (gold bar), the river gauge (black arrow

Environment Canada Hydrograph Station - 01BP001).

175

5.2.2 Data

High resolution imagery

Four-band red (R - 0.64-0.67 µm), green (G - 0.53-0.59 µm), blue (B - 0.45-0.51

µm), and near infrared (NIR - 0.76-0.9 µm), 0.3 m pixel resolution airborne-derived imagery was collected by the Province of New Brunswick Natural Resources and Energy

Development (NB-NRED) on 1 October 2017. These data were not atmospherically corrected. All four-band imagery was resampled to 0.5 m resolution as up-sampling high- resolution imagery has been found to reduce noise in the image due to inherent smoothing

(Jordan and Fonstad, 2005). Resampling was completed using a bilinear interpolation in

ArcMap (ESRI, 2018). Mean discharge on the day of image acquisition was Q = 9 m3s-1

(27.9 % of mean annual Q) at the river gauge (Figure 5-1c), and river stage was 1.18 m, which is a low flow period (Caissie, 2006).

Field data

Direct measures of water depth were collected during 19 August 2019, a day when daily discharge closely matched the discharge on day of image acquisition: Q = 7.3 m3s-1,

24.2 % of mean annual Q, and river stage = 1.14 m. Sample sites were selected to capture the range of depth variability across the catchment, and were based on prior knowledge of the area. These sites represented a range of depths from shallow (0.15 m) to deep (> 1 m)

(Figure 5-1c). Depth measurements were made with a Javad Triumph LS’s real time kinematic (RTK) dGPS at the upper, middle, and lower reaches of the study area in wadable sections (Figure 5-1c). Elevation was measured at the edge of the wetted river and used as a reference point. Transects were then taken across the channel at ≈ 1.5 m intervals, at ≈ 5

176

m centres along the river, and elevation was measured at each point. These elevation measurements were subtracted from the reference point to calculate depth for each point.

Additionally, a remote-controlled acoustic Doppler current profiler (ADCP) – ARCboat

(HR Wallingford) measured depth in deeper sections of the study area (9,464 m2) sampled at 3 MHz (Figure 5-1). In total, we collected n = 1,282 (n = 298 RTK GPS, n = 984 ADCP) unique depths in the study area, where the depths ranged from 0 to 2.69 m (Figure 5-2a).

Upon removing shadows (screening) from the image n = 762 spatially unique depths remained for further analysis throughout our study site (n = 280, RTK GPS, n = 482 ADCP)

(Figure 5-2b).

177

Figure 5-2 A histogram of field collected depth measurements (n = 1,282) via RTK dGPS and ADCP, n = 298 and n = 984, respectively; where the x-axis = observed depths, and the y-axis = number of observations, prior to screening for shadows in the images (a). Histogram for the same field observations post removing sites that co- occur with shadows (n = 762; n = 280 - RTK GPS, n = 482-ADCP) (b).

178

5.3 Modelling Framework

Our approach followed four steps: (1) Image filtering - we processed the images to extract the river from the image. We then removed shadows and exposed boulders/dry gravel bars from the image leaving only the wetted river and thus, river width. (2) Depth model – Initially, three spectral bands (R – band 1, G – band 2, B – band 3) were isolated and analyzed to select band(s) that best characterized the depth. Using FREEBIRD and validation field data, we then selected a best-fit model to derive depth. (3) Hydraulic characteristics – We used the image-derived depths, measured discharge, and river width to calculated velocity. We then calculated the hydraulic variables, Froude number and

Reynolds number. (4) To mitigate against model overfitting, we applied a Principal

Component Analysis (PCA) to reduce the number of hydraulic parameters. Using the reduced suite of hydraulic parameters, we ran an unsupervised classification model that generated a map of potential Atlantic Salmon and Brook Trout habitat.

(1) Image filtering

The presence of shadows, dry gravel areas, boulders, and vegetation can reduce the accuracy of image-derived depth mapping (Legleiter et al., 2004). We conducted a filtering analysis to remove these artifacts from the image. First, we exploited all four (R,

G, B and NIR) spectral bands and used a Normalized Difference Water Index (NDWI -

McFeeters, 1996) to extract the approximate river from the image. NDWI was conducted in ArcMap using the ‘band arithmetic’ function (Figures 5-3a and b). To remove gravels and boulders from the NDWI defined water body and establish the wetted river, we visually inspected R, G, and B bands against the RGB image to ascertain which band best

179

characterised dry gravels and boulders. Band 2 (green) was found to best define dry gravels and boulders. We then manually queried the dry gravel and boulder digital numbers (DN), established DN > 100 characterised these features, and used a conditional function in

‘raster calculator’ ArcMap 10.5 to remove gravels and boulders from the image (Figure 5-

3c). We used the same technique to remove shadows from the image, i.e., band 1 (red) best delimited shadows (DN < 12). The outputs from the two steps were merged to produce a final, ‘filtered’ image (Figure 5-3d).

Figure 5-3 Example of the image processing steps in a selected reach. (a) Original, unprocessed image. (b) Normalized difference water index (NDWI) highlighting the river at the blue end of the colour gradient. (c) River polygon with wetted river 180

extracted – gravels and boulders removed. (d) The final filtered image with shadows, gravel bars and emergent boulders removed.

(2) Depth Modelling

FREEBIRD – channel aspect ratio (CAR)

Legleiter’s (2015) FREEBIRD maps river depths from high-resolution imagery based on a linear relationship between the image and depth:

푑 = 푏0 + 푏1푋 (1) where d = depth, X = a image-derived quantity and is further developed below, and b0 – intercept and b1 = slope, with the coefficient vector b = [b0 b1]’.

Typically, b0 and b1 are regressed against in-situ field data. However, one FREEBIRD module relies on the principles of open channel flow, i.e., continuity and flow resistance, to estimate b0 and b1 based on the ‘Channel Aspect Ratio’ (CAR). We used CAR which develops a relationship between the image and water depth based on the principle of light attenuation within the water column (e.g., Marcus and Fonstad, 2005; Legleiter, 2012;

Kasvi et al., 2019). The first step in CAR is to develop an image-derived X quantity. We chose to use a transform as per Lyzenga, (1981):

푋 = ln (퐷푁 − min(퐷푁) + 1) (2)

where X is an image-derived quantity and DN = digital number of the band.

181

Visual inspection of our image revealed band 1 (red) gave the best representation of channel morphology (Legleiter, 2015) and was chosen for the image-derived quantity, X

(Figure 5-4a and b). The next step in CAR is to determine if the X image has a direct, or indirect relationship with depth. This is established by taking a cross-section (XS) across the X image (Figure 5-4c). A direct relationship exists if a XS of the X image follows the same geometry one would expect in a river channel: low X image values in shallow areas, high values in deeper areas (b1 > 0) and an indirect relationship denoted by an X image cross-section with high X image values in shallow areas and low values in deeper areas (b1

< 0). This relationship determines whether the model uses maximum or minimum X values to determine the b0 and b1 coefficients (Legleiter, 2015):

max (푋푗) , 푖푓 푏1 < 0 푋0,푗 = { (3) min(푋푗) , 푖푓 푏1 > 0

where j indexes pixel values along the ith XS. Combining equation (3) and equation (1) produces:

푑0 = 푏0 + 푏1푋0,푖 (4)

where d0 is the shallowest depth along a XS.

The relationship between X and depth in our image was indirect (Figure 3d), therefore, we use maximum Xj.

182

The mean depth (di) for a XS can then be related to mean pixel value 푋̅푖 along the XS, leading to:

푑̅푖 = 푏0 + 푏1푋̅푖 (5)

Equations (4) and (5) are rearranged to form terms for the intercept, b0) and slope b1 coefficients relative to the ith XS:

푑̅̅̅푖−푑0 푏0,푖 = 푑0 − 푋̅푖 (6) 푋̅̅푖̅̅−̅푋0,푖

푑̅̅̅푖−푑0 푏1,푖 = (7) 푋̅̅̅푖−푋0,푖

CAR requires an initial estimate of the channel aspect ratio (A), which is width (w)

/ average depth (푑̅). Applying this relationship across a set of XS, A = 푤̅̅̅푖 / 푑̅푖. Rearranging this relationship produces the average depth, 푑̅, where the lack of subscript i indicates the average reach depth versus the XS averaged depth, i.e., the pooled average of the scenes w and d as determined by a suite of XS. In the Miramichi River, work by Newbury (2015) estimated A = 44, and Legleiter (2015) found that CAR was not greatly influenced by misspecification of A. We used A = 44 and derived w for XS’s at 10 m centres along the study area (n = 12,400).

Width extraction from an image can be complicated by the persistence of shadows, where removal of shadows can lead to an under-estimate of actual width. This has implications for CAR and for hydraulic models that utilise width as a parameter, e.g., velocity. To account for these potential issues, we extracted river width on the image prior to filtering for shadows. Then, for each location, we calculated river width by collapsing

183

the river polygon to a centre line. We then used a smooth line function in ArcMap to remove sharp angles which are not representative of the river centre line. Lastly, we constructed perpendicular lines at 10 m centres along the smoothed centre line and extracted river width along these XS’s. Equations (6) and (7) are used to calculate 푏0,푖 and

푏1,푖 for each XS. We assumed d0 = 0 and 푏0,푖 and 푏1,푖 are pooled to determine a coefficient specific to the study area: 푏0 and 푏1. These coefficients are used as the slope and intercept parameters in equation (1) and produce a depth map of the area. All processing took place in ESRI’s ArcMap 10.5 (2019). A thorough review of these methods is detailed in

Legleiter (2015).

Table 5-1 Channel Aspect Ratio (CAR) parameters developed for this study (see

Equations 1-7), where mean channel width (n = 12,400 cross-sections) and X mean and max represent the accumulated mean and maximum X values for n = 12,400 cross-sections. The river’s aspect ratio (A) is taken from Newbury (2015), 풅̅ = average reach depth, and do is shallowest depth set to 0 following Legleiter (2015).

CAR Parameters Mean channel width (m) 42.8 X mean 3.5 X max 4.8 A 44 푑̅ 0.97

do 0

184

Figure 5-4 The process of establishing a relationship between the X image quantity and channel morphology from a filtered image (see text for explanation). (a) Filtered image, where flow direction is defined by the blue arrow. (b) Band 1 (red) illustrating a dark, pool-like feature with low digital number (DN). (c) The X image derived from band 1, with a cross-section through a pool-like feature (red line). (d) The X-values for the cross-section (0 = left bank viewing upriver) demonstrating the indirect relationship (b1 < 0) for the final depth calculations.

Field-calibrated image-derived depth model

A field-calibrated model requires the development of an X image. Similar to the CAR method, the X image takes the form of equation (2) and uses the data described above. We

185

also utilised band 1 to ensure consistency for comparing model outputs. The X image was regressed against the observed depths (n = 762) to construct a relationship, (e.g., linear, logarithmic, exponential, etc.), to predict the depth across the entire study area. Finally, model selection was based on comparing coefficients of determination (R2) and root mean square errors (RMSE) for both models (Legleiter and Fosness, 2019). All processing was completed in ArcMap 10.5 using the ‘raster calculator’.

(3) Hydraulic characteristics

A key hydraulic characteristic of river habitats is velocity (e.g., Riss and Biggs,

2003; Wilbur et al., 2020). Velocity can be calculated using the continuity relationship for a cross-sectional area (Hugue et al., 2016):

푄 = 푉퐴 (8) where Q = discharge (m3s-1), V = velocity (ms-1) and A = cross-sectional area (m2).

Rearranging equation (8) to calculate velocity produces:

푄 푉 = (9) 푤×푑 where w = width (m) of a cross-section and d = depth (m) at the cell of interest. In this approach, d varies across the XS as defined by the image-derived depth map. For each XS

(10 m centres described above) we created a raster using the ‘polyline to raster’ function in ArcMap at 15 m pixel resolution. Velocity was calculated for each cell by applying depth, width rasters, and Q in equation (9). We assumed a constant Q = 9 m3s-1 for the mainstem LSW which was Q on the day the image was acquired (river gauge, see Figure

5-1). For the North Pole Stream, we assumed a constant Q = 0.67 m3s-1 which is derived

186

from Caissie, (2015 and D. Caissie, DFO, pers. com.). All calculations were completed using ArcMap’s raster calculator.

Froude number (Fr) is a dimensionless hydraulic characteristic that has been applied to aquatic habitat modelling efforts with some success (e.g., Jowett et al., 1991;

Moir et al., 1998):

푣 퐹푟 = (10) √푔푑 where v = raster cell velocity (ms-1), g = gravitational acceleration constant (9.81 ms-2), and d = depth at the cell of interest (m). Depth and velocity rasters were used to construct a study-wide Fr map.

Reynolds number (Re ) is a dimensionless unit that describes the state of a fluid in motion, i.e., laminar (Re < 2,300), transitional (2,300 < Re < 4,000), or turbulent (Re >

4,000). Recent work has emphasized the influence of turbulence on behaviour of fishes

(e.g., Hockley et al., 2014):

푣푑 푅푒 = (11) 휇 where v = raster cell velocity (ms-1), d = depth at the cell of interest (m), and µ = kinematic viscosity of a fluid (ms-2) for water at 20 °C, η=1x10-6 m2s-1. Depth and velocity rasters were used to construct study-wide Re map.

(4) Habitat classification

Acknowledging that the suite of hydraulic parameters may be correlated we applied a Principal Component Analysis (PCA) to reduce parameters and protect against overfitting 187

(e.g., Wold et al., 1987). We randomly created 21,000 sample points across our study area in ArcMap using the ‘create random points’ tool, where the bounding area was the filtered river polygon. We then extracted depth, velocity, Froude and Reynolds values at each individual point. We ran a PCA and from the first two significant components, selected two variables (see results for a description) for the classification model. Statistical analyses were performed in Microsoft Excel (2016) using the XLSTAT extension (2019), and data were standardized prior to analysis.

In habitat classification models, supervised approaches frequently rely on ‘expert opinion’ to elucidate boundary conditions that define habitat types, e.g., velocity thresholds

(Borsányi et al., 2004); however, such results are undoubtedly biased by the experts’ opinions and beliefs (Hauer et al., 2009). We applied a well-established, unsupervised, machine learning technique – ISODATA clustering (Jensen, 1996) to mitigate potential expert biases. ISODATA clustering is similar to k-means clustering where an arbitrary cluster vector is assigned initially. Next, each pixel is classified to a nearby cluster and then the two steps are iterated until the difference between each iteration does not change or is negligible (Mather, 2004). The analysis was performed in ArcMap’s Iso Cluster

Unsupervised Classification toolbox (Calvert et al., 2015). We selected four (4) classes to be created in the analysis, and this was based on established, generalized fish habitat types (e.g., Rosenfeld, 2003; Wyrick et al., 2014). For the Iso Cluster Unsupervised classification run, we defined the minimum number of cells in a valid class = 20, to be sampled at an interval = 10.

188

5.4 Results

5.4.1 Image-derived bathymetry

The total area for river extraction across the 124 km study area was 6.17 km2 (the wetted area) of which 0.87 km2 (14 %) was removed due to shadowing. The final areal extent of imagery analysed was 5.3 km2. The CAR model (Table 5-2) had an R2 = 0.73 (RMSE =

0.167 m) when regressed against the n = 762 field-measured depths (Figure 5-5a). The model produced a maximum, minimum, and mean depth of 3.27, -1.73, and 0.06 m respectively (Table 5-2). An example section of the CAR derived depth map is presented in Figure 5-5b. The field-calibration data was fitted with an exponential model (Table 5-

2): R2 = 0.76, RMSE = 0.155 m, n = 762, see Figure 5-5c). The predicted maximum, minimum, and mean depth were 4.09, 0.05, and 0.41 m, respectively (Table 5-2). An example section of the field calibrated model depth map is presented in Figure 5-5d. The field-calibrated model performed best statistically and importantly, generated no negative depths; therefore, our habitat analyses were based on the field calibrated model.

Table 5-2 Results for CAR and field-calibrated (FC) depth models. The calibration form and associated parameters are shown, along with statistics for the calculated depth maps (n = 762).

189

CAR1 FC1 Linear: Exponential: Form b1X Calibration d = b0 +b1X d = b0e

Calculated parameters b0 = 3.67; b1 = -0.98 b0 =13.48; b1 = -0.966 Minimum -1.73 0.05 Maximum 3.27 4.09 Depth (m) Mean 0.06 0.41 Std. Dev 0.50 0.30 RMSE 0.17 0.16

Figure 5-5 (a) CAR model of depths based on n = 762 field observations. The black line is the model with confidence intervals (C.I., 95 %) for means (dashed lines) and

190

confidence intervals (C.I., 95 %) for observations (red lines). An example reach illustrating the CAR depth map (b). Field-calibrated exponential model (n = 762) (c).

The black line is the model with confidence intervals (C.I., 95 %) for means (dashed lines) and confidence intervals (C.I., 95 %) for observations (red lines). An example reach illustrating the field-calibrated, image-derived depth map (d).

5.4.2 Hydraulic characteristics

The derived velocities (equation 9) had a mean = 0.39 ms-1 and a maximum = 2.49 ms-1 (Table 5-3). An example reach illustrating an aerial image of a riffle-pool sequence and XS at 10 m centres is shown in Figure 6a, along with the resulting velocity map (Figure

5-6b). The mean Fr = 0.25, which is indicative of sub-critical flow state (Fr <1), and the maximum Fr = 1.19, which is representative of super-critical flow (Fr >1; see Table 5-3).

Mean Re ≈ 135,000 and the minimum Re ≈ 15,990, suggestive of a turbulent flow state across the catchment (Re > 4,000; see Table 5-3).

191

Figure 5-6 (a) An example reach with riffles and pools showing the assessed cross- sections at 10 m centres used as a parameter in the development of the velocity map

(b).

Table 5-3 Summary statistics for arising hydraulic characteristics.

Hydraulic characteristic Statistics

Minimum 0 Velocity Maximum 2.49 (ms-1) Mean 0.39 Std. Dev. 0.29 Minimum 0 Froude number Maximum 1.19 (non-dimensional) Mean 0.25 Std. Dev. 0.19 Minimum 15,990 Reynolds number Maximum 1,115,195 (non-dimensional) Mean 135,151 Std. Dev. 66,743

5.4.3 Hydraulic habitats

The PCA explained 84.5 % of the variance with 65.3 and 19.2 % in the first and second components, respectively. Depth was the only negative element in the first component.

All variables were positive in the second component. Depth and velocity had the highest factor loadings for both PCA components (factor loading for depth and velocity in the first component = -0.58 and 0.968, respectively; factor loading for depth and velocity in the second component = 0.769 and 0.152, respectively). Therefore, we selected depth and velocity for the unsupervised classification to create four habitat classes as both are well- 192

established key habitat features for Atlantic Salmon and Brook Trout (Figure 5-7a). We defined the four classes as shallow-rapid (mean depth = 0.18 m, mean velocity = 1.55 ms-

1), shallow-fast (mean depth = 0.28 m, mean velocity = 0.5 ms-1), shallow-slow (mean depth = 0.32 m, mean velocity = 0.15 ms-1), and deep-slow (mean depth = 1.12 m, mean velocity = 0.08 ms-1; Table 5-4) habitats. Most of the study reach (124 km) was shallow- fast (39%) which can also be described as ‘riffle-like’ and shallow-slow (34%) or ‘slack water-like’, with 17% deep-slow or ‘pool-like’ and 9% shallow-rapid or ‘riffle-like’ (Table

5-4; Figure 5-7).

193

Figure 5-7 Kernel density plot depicting the relationship between depth (x-axis) and velocity (y-axis), and habitat classes as defined velocity and depth (n=21,000 points)

(a). An example of a reach (b) and the classified habits (c).

Table 5-4 Summary statistics for the four hydraulic habitat classes and totals for our study area.

Shallow rapid Shallow fast Shallow slow Deep slow Mean velocity 1.55 0.5 0.15 0.08 (ms-1)

Mean velocity, 0.4 0.1 0.09 0.1 1 Stand. Dev. (ms-1)

Mean depth (m) 0.18 0.28 0.32 1.12

Mean depth, 0.07 0.08 0.13 0.59 1 Stand. Dev. (m) Maximum velocity 2.49 0.75 0.46 0.69 (ms-1) Minimum velocity 0.64 0.33 0.01 0 (ms-1) Maximum depth (m) 0.87 0.68 1.24 4.09 Minimum depth (m) 0.05 0.05 0.05 0.23 Total area classified 0.5 2.1 1.8 0.9 (km2)

Total area classified 9.2 39.4 34.3 16.8 (% of total area)

194

5.5 Discussion

5.5.1 Image-derived depth maps

High-resolution remotely sensed data are globally ubiquitous (e.g., Hugue et al.,

2016; O’Sullivan et al., 2019b). This is creating global scale databases that are facilitating high-resolution analyses of hydrologic, geomorphic, and ecologic processes (e.g., Chen et al., 2013; Marcus and Fonstad, 2008; Legleiter and Goodchild, 2005). In this study, we tested the efficacy of high-resolution, airborne, 4-band imagery to map river depth in a wide, shallow river using a method which does not require calibration data (CAR;

Legleiter, 2015) and a second method that utilises field data for calibration. We observed an R2 ≈ 0.73 for the CAR method when regressed against field data, but the CAR method produced negative depths in shallow areas. While CAR shows promise under certain conditions (see Legleiter, 2015), in its current form it has limited utility for depth mapping for aquatic habitat assessment. Similar to others, our field-calibrated image-derived depth model (actual depth data) was best explained with an exponential model (Hugue et al.,

2016; Legleiter and Fosness, 2019), and produced a strong regression, R2 ≈ 0.76. This is likely underpinned by the exponential attenuation of light in water (Legleiter et al., 2004).

Our conclusion is that high-resolution images can generate meaningful depth maps across riverscape-scales. However, challenges remain due to shadow obstructions (Kasvi et al.,

2019), turbidity (Marcus and Fonstad, 2008), accuracy at depths (Legleiter and Overstreet,

2012), chlorophyll (Legleiter et al., 2004), and logistics of acquiring field data to replicate conditions when the image was collected (Hugue et al., 2016). Regardless, work continues

195

to overcome these limitations. For example, Legleiter and Fosness (2019) recently found hyperspectral imagery accurately mapped river depths > 9 m.

5.5.2 Hydraulic characteristics

The LSW is a post glacial river, underlain by alluvial deposits and shallow bedrock, and has been hydraulically classified as having pool-riffle-run type habitats (Chadwick,

1995; Cunjak and Newbury, 2005). Not much is understood about the catchments’ hydraulic habitat composition during low flows (Swansburg et al., 2004). Our analysis suggests few if any run-type habitats exist during low flows, and instead the river is defined by pools, riffles, and slow-slack waters. Rivers in the northern temperate zone experience seasonally dynamic flows (Monk and Curry, 2009; El-Jabi and Caissie, 2019), and their late summer low flows transpose deep to shallow-type habitats, i.e., runs to riffles.

Viewing the spatial, and suggested temporal, characteristics of the hydraulic habitats in the

LSW highlights potential fluxes in river connectivity that likely impact fishes. As deep water runs transition to shallow water riffles, movements by fish are potentially more restricted. In the LSW, summer low flow periods can also be defined by extreme water temperatures, sometimes > 30 °C (e.g., Corey et al., 2019). Coalescing the flux of river connectivity with the extreme water temperatures in summer (Daigle et al., 2015), and observations of thermal refugia use in the LSW (Breau et al., 2007), illustrate river hydraulics control habitats for coldwater fish like salmonids. While changes in river flow are predicted for the Miramichi River (Swanburg et al., 2004) it is interesting that the flux in hydraulic conditions in the LSW was reported >100 years ago when Ganong (1906) found paddling during late summer on the LSW was “a succession of boulder rips and 196

deadwater stretches, requiring wading, dragging and portaging of heavier equipment” –

Rees (2016).

5.5.3 Habitat classification

Our hydraulic approach produced very detailed habitat characteristics to support both science and management of fishes. For example, ~9 % of the 124 km study area is shallow- rapid flowing water averaging 0.18 m deep and a 1.55 ms-1 velocity. These velocity values are well outside of observed habitat selected by juvenile Atlantic Salmon and Brook Trout

(e.g., Höjesjö et al., 2016; Wilbur et al., 2020). The shallow-fast habitats had an average depth = 0.28 m and an average velocity = 0.5 ms-1 which are selected by Atlantic Salmon fry (0+ years) and parr (≥1+) (Gibson, 1978; Heggenes et al., 1999), but these velocities are outside the known preference for Brook Trout (Cunjak and Green, 1983; Wilbur et al.,

2020). Based on our improved habitat maps, 48 % of the study area is unsuitable for Brook

Trout during summer low flows, a period also defined by high temperature and thermoregulation behaviour (Corey et al., 2019, Wilbur et al., 2020). Another example is that 17 % of the area was denoted as deep-slow (average depth = 1.12 m and average velocity = 0.08 ms-1) habitats which are most probably representative of pools. Pools provide a niche for larger Atlantic Salmon parr (Heggenes et al., 1999; Linnansaari et al.

2010), and pools are holding habitats for adults during upstream migrations (Bardonnet and Baglinière, 2000; Enders et al., 2009), as well as adult, migratory Brook Trout (Curry et al., 2006; Curry et al., 2010; Wilbur et al., 2020). Low flows are characterised by periods of extreme temperature, whether summer or winter. As such, low flow hydraulics may

197

likely define to some degree, winter habitats for these salmonids (Cunjak, 1988;

Linnansaari and Cunjak, 2013).

Early pioneering work by Amiro (1983) sought to quantify Atlantic Salmon habitat across the Miramichi River catchment using topographic maps and ortho-imagery. River gradients were extracted from 1:50,000 (25 m resolution) topographic maps, and visually/qualitatively classified habitats using the 1:20,000 orthoimages (10 m resolution, collected 1974-77). Amiro classified habitat by establishing if the water surface was broken or flat, and then coupled with gradients obtained from the topographic maps to develop a map of habitats. These maps are still in use today by the Department of Fisheries and Oceans, Canada (e.g., DFO, 2018). However, the scale is coarse and the images were collected across 3-years and likely are impacted by geomorphic variability over the time period in this catchment underlain by alluvial sediments. They also do not account for spatio-temporal variation in fluvial geomorphology (45-42 years of geomorphic processes), and thereby changes in the spatial arrangement of contemporary habitats. In addition, the quantification of hydraulic spatial heterogeneity was lacking in the Amiro study. Therefore, these maps forego any consideration for hydraulic effects on habitat selection across life histories stages and strategies. Our classification approach demonstrated how a closer approximation to hydraulic conditions and their resultant habitats can be obtained using contemporary remote sensing techniques. The potential to advance both science and management with these new, detailed depth, hydraulic and habitat maps is immense.

198

5.5.4 Limitations

These depth and hydraulic maps are not without limitations. Images must meet the criteria of light attenuation (Marcus et al., 2003; Legleiter et al., 2004; Legleiter, 2015).

This places limits on images related to several factors including a single interpretation of water level (a snap shot), image resolution, water clarity, and water depth. Recent work by Legleiter and Fosness (2019) has illustrated that hyperspectral imagery may overcome these depth limitations. Mapping is limited where shadows exists. In this study shadows comprised ≈ 14 % of the river images, and similar issues are likely in deeply incised rivers.

Additionally, these habitat models depend on knowing discharge in mapped reaches and then an extrapolation based on discharge and depth based on uniform flow. Discharge data is rare and often unattainable in remote locations. Our assumption of continuous discharge is most likely an over simplification of discharge in this catchment, which has many sub- catchments and is known to have groundwater gaining and losing reaches. Rivers rarely if ever exhibit uniform flow; they are hydraulically complex and create myriad hydraulic habitats for fishes and biota in general through space and time. Although every effort was made to mitigate issues involving width extraction, errors may be inherent due to the presence of shadows. These width errors have implications for both depth mapping via

CAR and hydraulic modelling. Similarly, river thermal regimes are not spatio-temporally homogeneous (Fullerton et al., 2015; O’Sullivan et al., 2019a), and the assumption of a uniform kinematic viscosity potentially affects our hydraulic parameters. These factors highlight the importance of adding actual field data (depth, velocity, cross-sectional area, temperature, and discharge) to ensure accurate calibration of models, i.e., “all models are wrong, some are useful” (Box, 1976). While our unsupervised classification method

199

sought to mitigate expert biases, biases are still inherent, underpinned by sampling efforts, where more dominant habitat types are likely to be overly represented, along with biases driven by the number of classes we sought to define. Northern temperate rivers are dynamic, ice can and does cause broad changes in fluvial geomorphic structure each season. This is a particular limitation of calibrating depth-derived imagery with data collected outside of image acquisition. Finally, while we applied a random sampling technique to select variables for classification modelling, spatial autocorrelation issues may persist (but see Grant et al., (2017)). Even with these limitations, these hydraulic habitat maps provide an important next step for catchment-scale habitat mapping, and will have great interest for ecologists, biologists, river scientists, and land and river managers alike.

5.6 References

Amiro, P. G. (1983). Aerial Photographic Measurement of Atlantic Salmon Habitat of the Miramichi River, New Brunswick. DFO Canadian Science Advisory Secretariat Science Response 83/74.

Bardonnet, A., & Baglinière, J.-L. (2000). Freshwater habitat of Atlantic Salmon ( Salmo salar ). Canadian Journal of Fisheries and Aquatic Sciences, 57(2), 497–506. https://doi.org/10.1139/f99-226

Borsányi, P., Alfredsen, K., Harby, A., Ugedal, O., & Kraxner, C. (2004). A Meso-scale Habitat Classification Method for Production Modelling of Atlantic Salmon in Norway. Hydroécologie Appliquée, 14, 119–138. https://doi.org/10.1051/hydro:2004008

Boruah, S., Gilvear, D., Hunter, P., & Sharma, N. (2008). Quantifying channel planform and physical habitat dynamics on a large braided river using satellite data—the Brahmaputra, India. River Research and Applications, 24(5), 650–660. https://doi.org/10.1002/rra.1132

200

Box, G. E. P. (1976). Science and statistics. Journal of the American Statistical Association, 71(356), 791–799. https://doi.org/10.1080/01621459.1976.10480949

Breau, C., Cunjak, R. A., & Bremset, G. (2007). Age-specific aggregation of wild juvenile Atlantic Salmon Salmo salar at cool water sources during high temperature events. Journal of Fish Biology, 71(4), 1179–1191. https://doi.org/10.1111/j.1095- 8649.2007.01591.x

Caissie, D. 2006. River discharge and channel width relationships for New Brunswick

rivers. Can. Tech. Rep. Fish. Aquat. Sci. 2637: 26p.

Caissie, J., Caissie, D., & El-Jabi, N. (2015). Hydrologically Based Environmental Flow Methods Applied to Rivers in the Maritime Provinces (Canada). River Research and Applications, 31(6), 651–662. https://doi.org/10.1002/rra.2772

Calvert, J., Strong, J. A., Service, M., McGonigle, C., & Quinn, R. (2015). An evaluation of supervised and unsupervised classification techniques for marine benthic habitat mapping using multibeam echosounder data. ICES Journal of Marine Science, 72(5), 1498–1513. https://doi.org/10.1093/icesjms/fsu223

Carbonneau, P., Fonstad, M. A., Marcus, W. A., & Dugdale, S. J. (2012). Making riverscapes real. Geomorphology, 137(1), 74–86. https://doi.org/10.1016/j.geomorph.2010.09.030

Chadwick, E.M.P. (1995) Water, science, and the public: the Miramichi ecosystem. Can. Spec. Pub. Fish. Aquatic. Sci. 123: 300 p

Chaput, G., Douglas, S. G., & Hayward, J. (2016). Biological Characteristics and Population Dynamics of Atlantic Salmon ( Salmo salar ) from the Miramichi River , New Brunswick , Canada. Canadian Science Advisory Secretariat (CSAS) Research Document. Moncton, NB.

Chen, F., Lin, H., Zhou, W., Hong, T., & Wang, G. (2013). Surface deformation detected by ALOS PALSAR small baseline SAR interferometry over permafrost environment of Beiluhe section, Tibet Plateau, China. Remote Sensing of Environment, 138, 10–

201

18. https://doi.org/10.1016/j.rse.2013.07.006

Corey, E., Linnansaari, T., Cunjak, R. A., & Currie, S. (2017). Physiological effects of environmentally relevant, multi-day thermal stress on wild juvenile Atlantic Salmon (Salmo salar). Conservation Physiology, 5(1), cox014. https://doi.org/10.1093/conphys/cox014

Corey, E., Linnansaari, T., Dugdale, S. J., Bergeron, N., Gendron, J., Lapointe, M., & Cunjak, R. A. (2019). Comparing the behavioural thermoregulation response to heat stress by Atlantic Salmon parr ( Salmo salar ) in two rivers. Ecology of Freshwater Fish, eff.12487. https://doi.org/10.1111/eff.12487

Cunjak, R. A. (1988). Behaviour and Microhabitat of Young Atlantic Salmon ( Salmo salar ) during Winter. Canadian Journal of Fisheries and Aquatic Sciences, 45(12), 2156– 2160. https://doi.org/10.1139/f88-250

Cunjak, R. A., & Caissie, D. (1993). Proceedings of the 7th Workshop on the Hydraulics of Ice Covered Rivers (1993). In Frazil ice accumulation in a large salmon pool in the Miramichi River, New Brunswick: Ecological implications for overwintering fishes (pp. 279–295). Saskatoon, Saskatchewan.

Cunjak, R. A., & Green, J. M. (1983). Habitat utilization by brook char ( Salvelinus fontinalis ) and rainbow trout ( Salmo gairdneri ) in Newfoundland streams. Canadian Journal of Zoology, 61(6), 1214–1219. https://doi.org/10.1139/z83-165

Cunjak, R. A., & Newbury, R. W. (2005). Atlantic coast rivers of CANADA. In Rivers of North America (pp. 938–980). Elsevier. https://doi.org/10.1016/B978-012088253- 3/50024-9

Curry, R. A., & Noakes, D. L. G. (1995). Groundwater and the selection of spawning sites by brook trout (Salvelinus fontinalis). Canadian Journal of Fisheries and Aquatic Sciences, 52, 1733–1740. https://doi.org/10.1139/f95-765

086

Curry, R. A., Van De Sande, J., & Whoriskey, F. G. (2006). Temporal and spatial habitats of anadromous brook charr in the Laval River and its estuary. Environmental Biology of Fishes, 76(2–4), 361–370. https://doi.org/10.1007/s10641-006-9041-4

Curry, R. A., Bernatchez, L., Whoriskey, F., & Audet, C. (2010). The origins and persistence of anadromy in brook charr. Reviews in Fish Biology and Fisheries, 20(4), 557–570. https://doi.org/10.1007/s11160-010-9160-z

Daigle, A., Jeong, D. Il, & Lapointe, M. F. (2015). Climate change and resilience of tributary thermal refugia for salmonids in eastern Canadian rivers. Hydrological Sciences Journal, 60(6), 1044–1063. https://doi.org/10.1080/02626667.2014.898121

DFO. (2018). Update of indicators to 2017 of adult Altantic salmon for the Miramichi River (NB), Salmon Fishing Area 16, DFO Gulf Region. DFO Canadian Science Advisory Secretariat Science Response 2017/043.

Dugdale, S. J., Bergeron, N. E., & St-Hilaire, A. (2015). Spatial distribution of thermal refuges analysed in relation to riverscape hydromorphology using airborne thermal infrared imagery. Remote Sensing of Environment, 160, 43–55. https://doi.org/10.1016/J.RSE.2014.12.021

Dugdale, S. J., Franssen, J., Corey, E., Bergeron, N. E., Lapointe, M., & Cunjak, R. A. (2016). Mainstem movement of Atlantic Salmon parr in response to high river temperature. Ecology of Freshwater Fish, 25(3), 429–445. https://doi.org/10.1111/eff.12224

Dunbar, M. J., Alfredsen, K., & Harby, A. (2012). Hydraulic-habitat modelling for setting environmental river flow needs for salmonids. Fisheries Management and Ecology, 19(6), 500–517. https://doi.org/10.1111/j.1365-2400.2011.00825.x

Eisner, A., C. Young, M. Schneider, & I. Kopecki, 2005. MesoCASiMiR – new mapping method and comparison with other current approaches. COST 626 European Aquatic Modelling Network- Proceedings from the final meeting in Silkebork, Denmark., 397.

203

El‐Jabi, N., & Caissie, D. (2019). Characterization of natural and environmental flows in New Brunswick, Canada. River Research and Applications, 35(1), 14–24. https://doi.org/10.1002/rra.3387

Enders, E. C., Scruton, D. A., & Clarke, K. D. (2009). The “natural flow paradigm” and Atlantic Salmon-moving from concept to practice. River Research and Applications, 25(1), 2–15. https://doi.org/10.1002/rra.1214

Fausch, K. D., Torgersen, C. E., Baxter, C. V., & Li, H. W. (2002). Landscapes to Riverscapes: Bridging the Gap between Research and Conservation of Stream Fishes. BioScience, 52(6), 483. https://doi.org/10.1641/0006- 3568(2002)052[0483:ltrbtg]2.0.co;2

Farò, D., G. Zolezzi, P. Vezza, K. Baumgartner, R. Klar, & A. Andreaoli, 2018. Testing 2D hydraulic modeling of mesoscale fish habitats in the Mareit / Mareta River (Italy ). 12th International Symposium on Ecohydraulics. Tokyo, Japan.

Fleischmann, A., Paiva, R., & Collischonn, W. (2019). Can regional to continental river hydrodynamic models be locally relevant? A cross-scale comparison. Journal of Hydrology X, 3. https://doi.org/10.1016/j.hydroa.2019.100027

Flener, C., Lotsari, E., Alho, P., & Käyhkö, J. (2012). Comparison of empirical and theoretical remote sensing based bathymetry models in river environments. River Research and Applications, 28(1), 118–133. https://doi.org/10.1002/rra.1441

Fonstad, M. A., & Marcus, W. A. (2005). Remote sensing of stream depths with hydraulically assisted bathymetry (HAB) models. Geomorphology, 72(1–4), 320– 339. https://doi.org/10.1016/j.geomorph.2005.06.005

Fullerton, A.H., Torgersen, C.E., Lawler, J.J., Faux, R.N., Steel, E.A., Beechie, T.J., Ebersole, J.L., and Leibowitz, S.G. 2015. Rethinking the longitudinal stream temperature paradigm: region-wide comparison of thermal infrared imagery reveals unexpected complexity of river temperatures. Hydrol. Process. 4737 (June). doi:10.1002/hyp.10506.

Fullerton, A. H., Torgersen, C. E., Lawler, J. J., Steel, E. A., Ebersole, J. L., & Lee, S. Y. 204

(2018). Longitudinal thermal heterogeneity in rivers and refugia for coldwater species: effects of scale and climate change. Aquatic Sciences, 80(1), 3. https://doi.org/10.1007/s00027-017-0557-9

Ganong, W. F. (1906). Notes on the natural history and physiography of New Brunswick. (B. and Co., Ed.). Saint John, New Brunswick.

Gibson, R. J. (1978). The Behavior of Juvenile Atlantic Salmon (Salmo salar) and Brook Trout (Salvelinus fontinalis) with Regard to Temperature and to Water Velocity. Transactions of the American Fisheries Society, 107(5), 703–712. https://doi.org/10.1577/1548-8659(1978)107<703:TBOJAS>2.0.CO;2

Grant, G. E., O’Connor, J., & Safran, E. (2017). Excursions in fluvial (dis)continuity. Geomorphology, 277, 145–153. https://doi.org/10.1016/j.geomorph.2016.08.033

Harby, A., Martinez- Capel, F., Lamouroux, N., 2017. From Microhabitat Ecohydraulics to an Improved Management of River Catchments: Bridging the gap Between Scales. River Research and Applications, 33, 189–191. https://doi.org/10.1002/rra.3114

Harrison, L. R., Bray, E., Overstreet, B., Legleiter, C. J., Brown, R. A., Merz, J. E., … Dunne, T. (2019). Physical Controls on Salmon Redd Site Selection in Restored Reaches of a Regulated, Gravel-Bed River. Water Resources Research. https://doi.org/10.1029/2018WR024428

Hauer, C., Mandlburger, G., & Habersack, H. (2009). Hydraulically related hydro- morphological units: description based on a new conceptual mesohabitat evaluation model (MEM) using LiDAR data as geometric input. River Research and Applications, 25(1), 29–47. https://doi.org/10.1002/rra.1083

Hauer, C., Unfer, G., Holzmann, H., Schmutz, S., & Habersack, H. (2013). The impact of discharge change on physical instream habitats and its response to river morphology. Climatic Change, 116(3–4), 827–850. https://doi.org/10.1007/s10584-012-0507-4

Heggenes, J., Bagliniere, J. L., & Cunjak, R. A. (1999). Spatial niche variability for young Atlantic Salmon (Salmo salar) and brown trout (S. trutta) in heterogeneous streams. Ecology of Freshwater Fish, 8(1), 1–21. https://doi.org/10.1111/j.1600- 205

0633.1999.tb00048.x

Heggenes, J., & Borgstrom, R. (1991). Effect of habitat types on survival, spatial distribution and production of an allopatric cohort of Atlantic Salmon, Salmo salar L., under conditions of low competition. Journal of Fish Biology, 38(2), 267–280. https://doi.org/10.1111/j.1095-8649.1991.tb03113.x

Heggenes, J., & Saltveit, S. J. (1990). Seasonal and spatial microhabitat selection and segregation in young Atlantic Salmon, Salmo salar L., and brown trout, Salmo trutta L., in a Norwegian river. Journal of Fish Biology, 36(5), 707–720. https://doi.org/10.1111/j.1095-8649.1990.tb04325.x

Hockley, F. A., Wilson, C. A. M. E., Brew, A., & Cable, J. (2014). Fish responses to flow velocity and turbulence in relation to size, sex and parasite load. Journal of the Royal Society Interface, 11(91). https://doi.org/10.1098/rsif.2013.0814

Höjesjö, J., Kaspersson, R., & Armstrong, J. D. (2016). Size-related habitat use in juvenile Atlantic Salmon: the importance of intercohort competition. Canadian Journal of Fisheries and Aquatic Sciences, 73(8), 1182–1189. https://doi.org/10.1139/cjfas- 2015-0446

Holm, C. F., Armstrong, J. D., & Gilvear, D. J. (2001). Investigating a major assumption of predictive instream habitat models: is water velocity preference of juvenile Atlantic Salmon independent of discharge? Journal of Fish Biology, 59(6), 1653–1666. https://doi.org/10.1111/j.1095-8649.2001.tb00228.x

Hugue, F., Lapointe, M., Eaton, B. C., & Lepoutre, A. (2016). Satellite-based remote sensing of running water habitats at large riverscape scales: Tools to analyze habitat heterogeneity for river ecosystem management. Geomorphology, 253, 353–369. https://doi.org/10.1016/j.geomorph.2015.10.025

Jensen, J. R., & R., J. (1996). Introductory digital image processing : a remote sensing perspective. Upper Saddle River, N.J.: Prentice Hall.

Lyzenga, D. R. (1981). Remote sensing of bottom reflectance and water attenuation parameters in shallow water using aircraft and landsat data. International Journal of 206

Remote Sensing, 2(1), 71–82. https://doi.org/10.1080/01431168108948342

Mather, P. M. (2004). Computer Processing of Remotely Sensed Data: An Introduction, 3rd edition. In Computer Processing of Remotely Sensed Data: An Introduction (3rd ed., p. 442). Chichester, UK: John Wiley and Sons.

Jordan, D.C., Fonstad, M.A., 2005. Two dimensional mapping of River Bathymetry and power using aerial photography and GIS on the Brazos river, Texas. Geocarto Int. 20, 13–20. https://doi.org/10.1080/10106040508542351

Jowett, I. G. (1993). A method for objectively identifying pool, run, and riffle habitats from physical measurements. New Zealand Journal of Marine and Freshwater Research, 27(2), 241–248. https://doi.org/10.1080/00288330.1993.9516563

Kasvi, E., Salmela, J., Lotsari, E., Kumpula, T., & Lane, S. N. (2019). Comparison of remote sensing based approaches for mapping bathymetry of shallow, clear water rivers. Geomorphology, 333, 180–197. https://doi.org/10.1016/j.geomorph.2019.02.017

Katopodis, C., & Kemp, P. S. (2018). Challenges of integrating habitat for aquatic life and morphodynamics offer a plethora of opportunities for advances in Ecohydraulics. Journal of Ecohydraulics, 3(1), 1–3. https://doi.org/10.1080/24705357.2018.1484331

Kinzel, P. J., Legleiter, C. J., & Nelson, J. M. (2013). Mapping River Bathymetry With a Small Footprint Green LiDAR: Applications and Challenges 1. JAWRA Journal of the American Water Resources Association, 49(1), 183–204. https://doi.org/10.1111/jawr.12008

Legleiter, C. J., Roberts, D. A., Marcus, W. A., & Fonstad, M. A. (2004). Passive optical remote sensing of river channel morphology and in-stream habitat: Physical basis and feasibility. Remote Sensing of Environment, 93(4), 493–510. https://doi.org/10.1016/j.rse.2004.07.019

Legleiter, C. J., & Goodchild, M. F. (2005). Alternative representations of in‐stream habitat: classification using remote sensing, hydraulic modeling, and fuzzy logic. International Journal of Geographical Information Science, 19(1), 29–50. 207

https://doi.org/10.1080/13658810412331280220

Legleiter, C. J., Roberts, D. A., & Lawrence, R. L. (2009). Spectrally based remote sensing of river bathymetry. Earth Surface Processes and Landforms, 34(8), 1039–1059. https://doi.org/10.1002/esp.1787

Legleiter, C. J., & Overstreet, B. T. (2012). Mapping gravel bed river bathymetry from space. Journal of Geophysical Research: Earth Surface, 117(F4). https://doi.org/10.1029/2012JF002539

Legleiter, C. J. (2015). Calibrating remotely sensed river bathymetry in the absence of field measurements: Flow REsistance Equation‐Based Imaging of River Depths (FREEBIRD). Water Resources Research, 51(4), 2865–2884. https://doi.org/10.1002/2014WR016624

Legleiter, C., & Fosness, R. (2019). Defining the Limits of Spectrally Based Bathymetric Mapping on a Large River. Remote Sensing, 11(6), 665. https://doi.org/10.3390/rs11060665

Legleiter, C. J. (2012). Remote measurement of river morphology via fusion of LiDAR topography and spectrally based bathymetry. Earth Surface Processes and Landforms, 37(5), 499–518. https://doi.org/10.1002/esp.2262

Linnansaari, T., & Cunjak, R. A. (2010). Patterns in apparent survival of Atlantic Salmon (Salmo salar) parr in relation to variable ice conditions throughout winter. Canadian Journal of Fisheries and Aquatic Sciences, 67(11), 1744–1754. https://doi.org/10.1139/F10-093

Linnansaari, T., Keskinen, A., Romakkaniemi, A., Erkinaro, J., & Orell, P. (2010). Deep habitats are important for juvenile Atlantic Salmon Salmo salar L. in large rivers. Ecology of Freshwater Fish, 19(4), 618–626. https://doi.org/10.1111/j.1600- 0633.2010.00443.x

Linnansaari, T., & Cunjak, R. A. (2013). Effects of ice on behaviour of juvenile Atlantic Salmon ( Salmo salar ). Canadian Journal of Fisheries and Aquatic Sciences, 70(10), 1488–1497. https://doi.org/10.1139/cjfas-2012-0236 208

Lyzenga, D. R. (1981). Remote sensing of bottom reflectance and water attenuation parameters in shallow water using aircraft and landsat data. International Journal of Remote Sensing, 2(1), 71–82. https://doi.org/10.1080/01431168108948342

Magilligan, F. J., Nislow, K. H., Kynard, B. E., & Hackman, A. M. (2016). Immediate changes in stream channel geomorphology, aquatic habitat, and fish assemblages following dam removal in a small upland catchment. Geomorphology, 252, 158–170. https://doi.org/10.1016/j.geomorph.2015.07.027

Marcus, W. A. (2002). Mapping of stream microhabitats with high spatial resolution hyperspectral imagery. Journal of Geographical Systems, 4(1), 113–126. https://doi.org/10.1007/s101090100078

Marcus, W. A., Legleiter, C. J., Aspinall, R. J., Boardman, J. W., & Crabtree, R. L. (2003). High spatial resolution hyperspectral mapping of in-stream habitats, depths, and woody debris in mountain streams. Geomorphology, 55(1–4), 363–380. https://doi.org/10.1016/S0169-555X(03)00150-8

Marcus, W. A., & Fonstad, M. A. (2008). Optical remote mapping of rivers at sub-meter resolutions and watershed extents. Earth Surface Processes and Landforms, 33(1), 4– 24. https://doi.org/10.1002/esp.1637

Marcus, W. A., & Fonstad, M. A. (2010). Remote sensing of rivers: the emergence of a subdiscipline in the river sciences. Earth Surface Processes and Landforms, 35(15), 1867–1872. https://doi.org/10.1002/esp.2094

Marzadri, A., Tonina, D., & Bellin, A. (2013). Quantifying the importance of daily stream water temperature fluctuations on the hyporheic thermal regime: Implication for dissolved oxygen dynamics. Journal of Hydrology, 507, 241–248. https://doi.org/10.1016/j.jhydrol.2013.10.030

McFeeters, S. K. (1996). The use of the Normalized Difference Water Index (NDWI) in 209

the delineation of open water features. International Journal of Remote Sensing, 17(7), 1425–1432. https://doi.org/10.1080/01431169608948714

Merz, J. E., Setka, J. D., Pasternack, G. B., & Wheaton, J. M. (2004). Predicting benefits of spawning-habitat rehabilitation to salmonid (Oncorhynchus spp.) fry production in a regulated California river. Canadian Journal of Fisheries and Aquatic Sciences, 61(8), 1433–1446. https://doi.org/10.1139/F04-077

Moir, H. J., Soulsby, C., & Youngson, A. (1998). Hydraulic and sedimentary characteristics of habitat utilized by Atlantic Salmon for spawning in the Girnock Burn, Scotland. Fisheries Management and Ecology, 5(3), 241–254. https://doi.org/10.1046/j.1365-2400.1998.00105.x

Moir, H. J., Gibbins, C. N., Soulsby, C., & Webb, J. H. (2006). Discharge and hydraulic interactions in contrasting channel morphologies and their influence on site utilization by spawning Atlantic Salmon (Salmo salar). Canadian Journal of Fisheries and Aquatic Sciences, 63(11), 2567–2585. https://doi.org/10.1139/F06-137

Monk, W.A. and Curry, R.A., 2009. Models of past, present and future stream temperatures for selected Atlantic Salmon rivers in Northeastern North America. American Fisheries Society Symposium, 69, 215–230.

Monk, W. A., Peters, D. L., Curry, R.A., & Baird, D. J. (2011). Quantifying trends in indicator hydroecological variables for regime-based groups of Canadian rivers. Hydrological Processes. https://doi.org/10.1002/hyp.8137

Newbury, B. 2015. Stream Restoration Hydraulics, Project Casebook. UBR Services, Winfield, BC, Canada.

O’Sullivan, A. M., Devito, K. J., & Curry, R. A. (2019a). The influence of landscape characteristics on the spatial variability of river temperatures. Catena, 177, 70–83. https://doi.org/10.1016/J.Catena.2019.02.006

O’Sullivan, A. M., Linnansaari, T., & Curry, R. A. (2019b). Ice Cover Exists (ICE): A quick method to delineate groundwater inputs in running waters for cold and temperate regions. Hydrological Processes. https://doi.org/10.1002/hyp.13557 210

Parasiewicz P, Rogers JN, Vezza P, Gortázar J, Seager T, Pegg M, Wiśniewolski W, Comoglio C. Applications of the MesoHABSIM simulation model. Ecohydraulics: an integrated approach. 2013 Jan 1:109-24. https://doi.org/10.1002/9781118526576.ch6

Pasternack, G. B. (2020). River Restoration: Disappointing, Nascent, Yet Desperately Needed. In Reference Module in Earth Systems and Environmental Sciences. Elsevier. https://doi.org/10.1016/b978-0-12-409548-9.12449-2

Poff, N. L., Allan, J. D., Bain, M. B., Karr, J. R., Prestegaard, K. L., Richter, B. D., … Stromberg, J. C. (1997). The Natural Flow Regime. BioScience, 47(11), 769–784. https://doi.org/10.2307/1313099

Rees, R. (2016). New Brunswick was his Country: The Life of William Francis Ganong. (Nimbus Publishing). Halifax, NS. https://doi.org/10.1111/gere.12286

Reid, M. A., & Thoms, M. C. (2008). Surface flow types, near-bed hydraulics and the distribution of stream macroinvertebrates. Biogeosciences, 5(4), 1043–1055. https://doi.org/10.5194/bg-5-1043-2008

Riis, T., & Biggs, B. J. F. (2003). Hydrologic and hydraulic control of macrophyte establishment and performance in streams. Limnology and Oceanography, 48(4), 1488–1497. https://doi.org/10.4319/lo.2003.48.4.1488

Rosenfeld, J. (2003). Assessing the Habitat Requirements of Stream Fishes: An Overview and Evaluation of Different Approaches. Transactions of the American Fisheries Society, 132(5), 953–968. https://doi.org/10.1577/t01-126

Schoelynck, J., Meire, D., Bal, K., Buis, K., Troch, P., Bouma, T., Meire, P. and Temmerman, S., 2013. Submerged macrophytes avoiding a negative feedback in reaction to hydrodynamic stress. Limnologica, 43(5), pp.371-380. https://doi.org/10.1016/j.limno.2013.05.003

Shearer, K., Hayes, J., Jowett, I., & Olsen, D. (2015). Habitat suitability curves for benthic macroinvertebrates from a small New Zealand river. New Zealand Journal of Marine and Freshwater Research, 49(2), 178–191. 211

https://doi.org/10.1080/00288330.2014.988632

Southwood, T. R. E. (1977). Habitat, the Templet for Ecological Strategies? The Journal of Animal Ecology, 46(2), 336. https://doi.org/10.2307/3817

Swansburg, E., El-Jabi, N., Caissie, D., & Chaput, G. (2004). Hydrometeorological Trends in the Miramichi River, Canada: Implications for Atlantic Salmon Growth. North American Journal of Fisheries Management, 24(2), 561–576. https://doi.org/10.1577/m02-181.1

Torgersen, C.E. D.M. Price, H.W. Li, and B.A. McIntosh. (1995). Thermal refugia and chinook salmon habitat in Oregon: Applications of airborne thermal videography. In: Proceedings of the 15th Biennial Workshop on Color Photography and Videography (P. Mausel, ed.), pp. 167-171, American Society for Photogrammetry and Remote Sensing, Terre Haute, IN.

Torgersen, C. E., Price, D. M., Li, H. W., & McIntosh, B. A. (1999). Multiscale thermal refugia and stream habitat associations of chinook salmon in northeastern Oregon. Ecological Applications, 9(1), 301–319. https://doi.org/10.1890/1051- 0761(1999)009[0301:MTRASH]2.0.CO;2

Wallis, C., Maddock, I., Visser, F., & Acreman, M. (2012). A framework for evaluating the spatial configuration and temporal dynamics of hydraulic patches. River Research and Applications, 28(5), 585–593. https://doi.org/10.1002/rra.1468

Weatherall, P., Marks, K.M., Jakobsson, M., Schmitt, T., Tani, S., Arndt, J.E., Rovere, M., Chayes, D., Ferrini, V. and Wigley, R., 2015. A new digital bathymetric model of the world's oceans. Earth and Space Science, 2(8), pp.331-345. . https://doi.org/10.1002/2015EA000107

Wilbur, N. M., O’Sullivan, A. M., MacQuarrie, K. T. B., Linnansaari, T., & Curry, R. A. (2020). Characterizing physical habitat preferences and thermal refuge occupancy of brook trout (Salvelinus fontinalis) and Atlantic Salmon (Salmo salar) at high river temperatures. River Research and Applications, rra.3570. https://doi.org/10.1002/rra.3570

212

Wilkes, M. A., Maddock, I., Link, O., & Habit, E. (2016). A Community-Level, Mesoscale Analysis of Fish Assemblage Structure in Shoreline Habitats of a Large River Using Multivariate Regression Trees. River Research and Applications, 32(4), 652–665. https://doi.org/10.1002/rra.2879

Winterbottom, S. J., & Gilvear, D. J. (1997). Quantification of channel bed morphology in gravel-bed rivers using airborne multispectral imagery and aerial photography. Regulated Rivers: Research & Management, 13(6), 489–499. https://doi.org/10.1002/(SICI)1099-1646(199711/12)13:6<489::AID- RRR471>3.0.CO;2-X

Wohl, E., Lane, S. N., & Wilcox, A. C. (2015). The science and practice of river restoration. Water Resources Research, 51(8), 5974–5997. https://doi.org/10.1002/2014WR016874

Wold, S., Esbensen, K., & Geladi, P. (1987). A Pedagogy of Charity : Donald Davidson and the. Chemometrics and Intelligent Laboratory Systems, 37–52. https://doi.org/10.1016/0169-7439(87)80084-9

Wyrick, J. R., & Pasternack, G. B. (2014). Geospatial organization of fluvial landforms in a gravel-cobble river: Beyond the riffle-pool couplet. Geomorphology, 213, 48–65. https://doi.org/10.1016/j.geomorph.2013.12.040

213

Chapter 6: Synthesis

“By evaluating landscapes from a common conceptual framework, hydrologic processes common to some or all landscapes can be distinguished from hydrologic processes unique to specific landscapes.” – Thomas C. Winter (2001)

6.1 Significance of dissertation

The interconnectedness of hydrological processes across landscapes and riverscapes and the resonant effects on ecosystems – both terrestrial and freshwater – have been described eloquently (see Hynes, 1973; Winter et al., 1998); however, these interconnected structures and processes are abundantly complex. While we somewhat understand that the composition of water in rivers, streams, lakes, wetlands, and groundwater systems comprises water sourced across myriad spatio-temporal scales (see Soulsby et al., 2000;

McGuire and McDonnell, 2005; Kirchner et al., 2010; Harman, 2015), we are still far from resolving how landscapes modulate these intertwined processes (see for example Devito et al., 2005; Hokanson et al., 2019). The landscape’s governance of these physical processes, along with biota, ultimately underlies the spatial and temporal arrangement of rivers, streams, lakes, and wetlands, their thermal, nutrient, and hydraulic regimes, and therefore influence the occurrence and persistence of biota from bacteria to animals to plants.

My dissertation increased our understanding of the connections and interactions between landscapes and riverscapes. The riverscape itself represents a cacophony of complexities associated with hydraulic, thermal, light, chemical, and nutrient regimes, all

214

of which influence dispersion and diffusion from molecules to organisms through an unending thread of feedbacks between the abiotic environment and its biotic denizens. My dissertation offers evidence of (a) landscape controls on rivers as expressed in their thermal characteristics, (b) a quick method to delineate groundwater discharge in rivers, and (c) a high-resolution viewpoint of the longitudinal spatial variability of a riverscape’s hydraulic characteristics and the habitats they create. In answer to the objectives I presented in

Chapter 1:

1) Physiographic setting exerts a primary control on the spatial variability of river

temperatures via groundwater connectivity.

2) Landscape control of river temperatures includes the vertical dimension of

geology and as a consequence, surface spatial statistical models are confounded

in areas with complex hydrogeological settings.

3) Groundwater discharge along rivers can be delimited from winter imagery in

regions that freeze.

4) Aerial imagery can be utilised to develop high resolution, bathymetrical maps at

the catchment-scale. These maps can be interpreted in concert with discharge

data to develop detailed hydraulic and habitat maps.

Objective 1: I established that the spatial variability of river temperature can be associated with its physiographic setting. Geologic contact zones can demarcate points of hydrologic interfaces and in regions with shallow overburden, these interfaces can introduce groundwater to the river and modify the river’s thermal regimes. Upland wetlands can act

215

as heat sources for rivers that migrate through them because they likely lack a groundwater influence and their heat budgets are largely driven by solar radiation and air temperature, accompanied by precipitation.

Objective 2: Spatial statistical models have grown in popularity in recent decades, especially in hydrology studies. In regions where topography exerts a constant control on groundwater, whether connected or disconnected, surface spatial statistical river network models can offer an improvement over aspatial models for river temperature modelling. In these settings, high resolution topographic data can further improve models. However, in areas with complex geology where groundwater tables are both connected and disconnected from topography, surface spatial statistical models offer little to no improvement. At finer hydrographic network resolution, e.g., 3 ha drainage networks, these issues are more prominent. These findings established that geologic setting can confound spatial statistical autocorrelation and thus limit the utility of such models. I also demonstrated that geology and slope steepness or valley incision play a role in river thermal regimes related to the hydraulic conductance of the underlying geology. For example, deep valley incisions in hydraulically conductive bedrock, such as fractured sandstone, can create diffuse or discrete groundwater from bedrock discharges to the river. Conversely, a deeply incised valley in uniform, low conductance bedrock limits groundwater emerging at its surface and thus not influence river temperatures, although discrete discharges can occur with fractures.

216

Objective 3: In regions where winters temperatures lead to rivers freezing, groundwater discharge can be identified in wintertime via high resolution satellite imagery. Summer and winter temperature regimes are not only defined by seasonal temperatures, but also by changes in groundwater plume extent in rivers. In the winter, groundwater plumes are longer than summer plumes in some river reaches. I suggest this is driven by thermodynamic and hydraulic processes that are initiated by changes in temperature that induce changes in turbulence regimes along with the physical control of flow paths by ice.

These processes differ in summer creating higher mixing rates, driven again by temperature fluxes and associated state changes that result in shorter extents of groundwater plumes.

Objective 4: Aerial imagery can be utilised, in tandem with field data, to develop robust and accurate, sub-metre resolution, bathymetric maps at the catchment-scale. These maps provide the framework for the development of high-resolution hydraulic maps, and thereby, an unprecedented insight into the spatial complexity of freshwater aquatic habitats.

Such bathymetrical data may also facilitate the development of hydrodynamic models.

These methods, when coupled with biotic data, such as fish or benthic macro-invertebrate distributions have a high probability of uncovering many new levels of abiotic and biotic interconnectedness.

6.2 Synthisis and future directions

My dissertation sheds some new light on the fact that landscapes and riverscapes are inherently and utterly interconnected in an unbounded continuum. I believe Earth system research, whether abiotic or biotic, benefits when we acknowledge and incorporate 217

this interconnectedness into our thinking. In an effort to start conversations and guide the future of interdisciplinary work, I present a concept named the “waterscape continuum concept”.

6.2.1 Waterscape continuum concept

The term hydrosphere defines a planets combined water mass, including water under, on and above the planet’s surface. The Earth’s hydrosphere extends from the lithosphere (at a depth of 410-660 km) to the atmosphere (Figure 6-1a). Generally, the processes which are most influential for Earth’s ecosystems operate between a few hundred metres below the Earth’s surface (the soil/ rock moisture and groundwater component), on the Earth’s surface (the surface water component) and in the atmosphere (the atmospheric water component), and are expressed in the water cycle (Figure 6-1b). On and within the

Earth’s surface, a continuum of water exists and I term this a “waterscape continuum”. The waterscape continuum unifies surface water, soil/ rock moisture, and groundwater, where surface water is simply the component that intercepts topography and the soil/ rock moisture and groundwater are the subsurface components. The waterscape continuum extends across fine (sub-metre) to global scales (kilometres, e.g., Cuthbert et al., 2019), surfaces (Winter et al., 1998); depths of metres to kilometres (e.g., Fan, 2019), and time- scales of seconds to millennia (e.g., Maxwell et al., 2016; Figure 6-1c). The continuum is dynamic, and its spatial configuration varies temporally, expanding and contracting.

Fluxes in the continuum vary with (1) climate, (2) geologic composition, (3) topography,

(4) biota, and (5) the special case for biota, humans. The waterscape continuum’s structure, that is its geology’s hydraulic conductance, landscape topography, and human activities will determine its flow paths including the age of water throughout the waterscape (e.g., 218

Tóth, 1962; Sprenger et al., 2019; Figure 6-1c). The continuum is characterised by feedback loops, self-propagating throughout the waterscape.

Figure 6-1 An illustration of the Earth’s structure (redrawn from USGS, 1999), highlighting the hydrosphere, a region extending from the 640 km in the mantle to

Earth’s atmosphere (a). A simple representation of the Earth’s water cycle comprising atmospheric/ meteoric water, surface water, rock/soil moisture and groundwater, evaportranspiration, and water use by humans (c). An illustration of a waterscape continuum, connecting landscape, flora, and water body (lentic or lotic) hydrological processes, and its associated flow paths and age of water as a function of hydraulic conductance and gradient.

219

My ultimate goal is to expand the waterscape continuum into a Framework. This framework would be used to guide research towards maximizing successful outcomes in hydrology, ecology, and the breadth of interdiciplinarity inbetween. The framework will direct researchers to take their questions through the waterscape continum’s principles and evolve their questions to improve relevance and thus, revealing mechanisms that will support meaningful conservation and preservation tools.

EPILOGUE

While the complexities are enormous, the future for interdisciplinary research, and

the findings it may reveal, are truly incredible.

220

6.3 References

Cuthbert, M. O., Gleeson, T., Moosdorf, N., Befus, K. M., Schneider, A., Hartmann, J., &

Lehner, B. (2019, February 1). Global patterns and dynamics of climate–

groundwater interactions. Nature Climate Change. Nature Publishing Group.

https://doi.org/10.1038/s41558-018-0386-4

Devito, K., Creed, I., Gan, T., Mendoza, C., Petrone, R., Silins, U., & Smerdon, B.

(2005). A framework for broad-scale classification of hydrologic response units on

the Boreal Plain: is topography the last thing to consider? Hydrological Processes,

19(8), 1705–1714. https://doi.org/10.1002/hyp.5881

Harman, C. J. (2015). Time-variable transit time distributions and transport: Theory and

application to storage-dependent transport of chloride in a watershed. Water

Resources Research, 51(1), 1–30. https://doi.org/10.1002/2014WR015707

Hokanson, K. J., Mendoza, C. A., & Devito, K. J. (2019). Interactions Between Regional

Climate, Surficial Geology, and Topography: Characterizing Shallow Groundwater

Systems in Subhumid, Low‐Relief Landscapes. Water Resources Research, 55(1),

284–297. https://doi.org/10.1029/2018WR023934

Hynes, H. B. N. (1975). The stream and its valley. SIL Proceedings, 1922-2010, 19(1),

1–15. https://doi.org/10.1080/03680770.1974.11896033

Hynes, H. B. N. (1983). Groundwater and stream ecology. Hydrobiologia, 100(1), 93–99.

https://doi.org/10.1007/BF00027424

Kirchner, J. W., Tetzlaff, D., & Soulsby, C. (2010). Comparing chloride and water

221

isotopes as hydrological tracers in two Scottish catchments. Hydrological Processes,

24(12), 1631–1645. https://doi.org/10.1002/hyp.7676

Maxwell, R. M., Condon, L. E., Kollet, S. J., Maher, K., Haggerty, R., & Forrester, M.

M. (2016). The imprint of climate and geology on the residence times of

groundwater. Geophysical Research Letters, 43(2), 701–708.

https://doi.org/10.1002/2015GL066916

McGuire, K. J., McDonnell, J. J., Weiler, M., Kendall, C., McGlynn, B. L., Welker, J.

M., & Seibert, J. (2005). The role of topography on catchment-scale water residence

time. Water Resources Research, 41(5). https://doi.org/10.1029/2004WR003657

Soulsby, C., Malcolm, R., Helliwell, R., Ferrier, R. C., & Jenkins, A. (2000). Isotope

hydrology of the Allt a’ Mharcaidh catchment, Cairngorms, Scotland: implications

for hydrological pathways and residence times. Hydrological Processes, 14(4), 747–

762. https://doi.org/10.1002/(SICI)1099-1085(200003)14:4<747::AID-

HYP970>3.0.CO;2-0

Sprenger, M., Stumpp, C., Weiler, M., Aeschbach, W., Allen, S. T., Benettin, P., …

Werner, C. (2019, September 1). The Demographics of Water: A Review of Water

Ages in the Critical Zone. Reviews of Geophysics. Blackwell Publishing Ltd.

https://doi.org/10.1029/2018RG000633

Tóth, J. (1962). A theory of groundwater motion in small drainage basins in central

Alberta, Canada. Journal of Geophysical Research, 67(11), 4375–4388.

https://doi.org/10.1029/JZ067i011p04375

Winter, T. C., Harvey, J. W., Franke, O. L., & Alley, W. M. (1998). Groundwater and 222

surface water: A single resource. USGS Publications, 79. https://doi.org/10.3389/fpsyg.2012.00044

223

Appendix: Effects of topographic resolution and geologic setting on spatial statistical river temperature models

Introduction This supporting information provides additional details to those Chapter 3: Effects of topographic resolution and geologic setting on spatial statistical river temperature models.

Figure A-1 Water table ratio values for North America, and the United Kingdom indicating regions denoting a region that has groundwater regimes that are both connected (WTR >1) and disconnected (WTR<1) from topography. Regions with strong connection of topography to groundwater represent areas where spatial statistical models show strong fits and low

RMSPE’s as observed for spatial statistical river temperature models in the Pacific

Northwestern U.S. (Isaak et al., 2017) and Scotland (Jackson et al., 2017). In comparison, lower spatial statistical river temperature model fits and higher RMSPE’s are observed where more heterogeneous WTR exist as observed in north eastern US (Detenbeck et al., 224

2016). Note, New Brunswick has heterogeneous geology and large regions predicted to have groundwater regimes that are both connected (WTR >1) and large areas that are disconnected from topography. WTR data from Cuthbert et al., (2019), accessed at https://figshare.com/authors/Mark_Cuthbert/3847054

Catchment characteristics of the Cains River and North Pole Stream are presented in Table A-1.

A detailed overview of catchment surficial geology for both catchments is provided in Table A-

Table A-1 Summary of catchment characteristics

Cains River North Pole Stream Catchment area 1,400 214 (km2) Elevation (m.a.s.l.) 188; 11; 101 730; 265; 437 (max; min; mean) Catchment slope (degrees, °) 60.8; 0; 2.5 57.5; 0; 6.3 (max; min; mean) sandy lodgement till Moraine and ablation Surficial geology ablation and sandy/ tills silty blanket colluvium Sandstone Deep water clastic Bedrock geology Fluviatile Granite (464 Mya) conglomerate Granite (419 Mya) Air temperature 10 to 11; 7 to 8.5; (°C) -1 to -0.5; -3.5 to -2.5; (max; min; mean) 4.5 to 5.5 1.5 to 3 Precipitation 950 to 1000 1050 to 1150 (mm)

225

Table A-2 Summary of surficial geologic units of the Cains River catchment and the North

Pole Stream catchment.

Deposit Description Blanket and veneer: loamy lodgment till, minor ablation till, silt, sand, gravel, rubble generally 0.5 aMB2 to 3 m thick aMB3 Blanket loamy lodgement till, minor ablation till, silt, sand, gravel. Mainly stony till (more than 35% of clasts pebble-sized and larger) Morainal and colluvial sediments: loamy till and colluvium, regolith and weathered bedrock, and isolated boulder fields, undifferentiated; mixture of deposits formed directly from ice aMCB of unknown age and materials produced by weathering processes; generally greater than 1 m thick. Mainly stony deposits (more than 35% of clasts pebble-sized and larger) Hummocky, ribbed, and rolling ablation moraines: loamy ablation till, some lodgment till, minor aMM3 silt, sand, gravel, and boulders; generally greater than 1.5 m thick Mainly stony till (more than 35% of clasts pebble-sized and larger); Veener loamy lodgement till, minor ablation till, silt, sand, gravel. aMV3 Mainly stony till (more than 35% of clasts pebble-sized and larger) Blanket loamy lodgement till, minor ablation till, silt, sand, gravel. bMB3 Mainly bouldery till (more than 25% of clasts boulder-sized) Veener loamy lodgement till, minor ablation till, silt, sand, gravel. bMV3 Mainly bouldery till (more than 25% of clasts boulder-sized) Morainal sediments: lodgment till, ablation till, and associated sand and gravel deposited directly GX3 by Late Wisconsinan ice or with minor reworking by water. generally more than 2 m thick Morainal and colluvial sediments: loamy till and colluvium, regolith and weathered bedrock, and MCB isolated boulder fields, undifferentiated; mixture of deposits formed directly from ice of unknown age and materials produced by weathering processes; generally greater than 1 m thick.

bMM3 Hummocky, ribbed, and rolling ablation moraines: loamy ablation till, some lodgment till, minor silt, sand, gravel, and boulders; generally greater than 1.5 m thick ORGANIC SEDIMENTS: bogs, fens, swamps: peat, muck, minor silt and fine sand; generally 1 to OP 5 m thick; deposited in shallow basins and on poorly drained surfaces Blanket loamy lodgement till, minor ablation till, silt, sand, gravel. sMB3 Mainly sandy till (sand content > 50%) WB Blankets and plains: sand, silt, some gravel and clay; generally 0.5 to 3 m thick

A LiDAR point cloud was used to construct detailed solar radiation models for the

LiDAR models in Chapter 3. Figure A-2 and 3 illustrate examples of the riparian zone variability throughout the Cains River and North Pole Stream, respectively.

226

227

Figure A-2 An overview of riparian zones in the Cains River catchment. (a) an example of the headwaters of the Cains River, where short wetland vegetation is dominant via inspection of a high density LiDAR point cloud, and (b) an example of the lower river section of the

Cains, where a riparian zone flanks the river and an alluvial terrace is present in this lower river section, adjacent to tall (26 – 30 m) coniferous and deciduous trees.

228

Figure A-3 An overview of riparian zones in the North Pole Stream catchment. An example of the headwaters of the North Pole Stream, illustrating streams with relatively tall vegetation via inspection of a high density LiDAR point cloud (a). An example of a riparian zone defined by vegetation less than 5 m in height, which

229

occur in the headwaters and is representative of the wetland vegetation in the middle reaches (b). An example of LiDAR point cloud detailing tall vegetation and a confined river section in the lower reaches of the catchment (c).

In Chapter 3 I examined the effects of topographic resolution on spatial statistical river temperature models. I used two digital elevation models (DEMs) to examine these effects: a coarse 30 m resolution satellite derived DEM, NASA’s Shuttle Radar

Topography Mission (SRTM), and a high resolution (1 m) airborne LiDAR derived DEM.

To highlight differences between each dataset I computed a DEM difference. First, I applied a fishnet at a point density of 1 point/ 500 m across the study river catchments using the ‘Fishnet’ tool in ArcMap (ESRI, 2019). We then extracted values for the SRTM

DEM and LiDAR bare earth DEM to the fishnet points, using the ‘extract values to points’ tool in ArcMap. Upon extraction, we conducted t-tests in XLSTAT (www.xlstat.com) to investigate is any significant difference exists between the elevations measured in each

DEM. Additionally, we also subtracted the LiDAR bare earth DEM from the SRTM DEM to produce a DEM of elevation differences. Here, our objective was to compare observed elevation differences between the two DEMs on a cell by cell bases.

Applying a 1 point 500 m-1 fishnet across the Cains River produced 5,279 points.

The resulting two sample t-test (1 m bare earth LiDAR DEM; SRTM DEM) computed a p-value = 0.029 (α = 0.05), highlighting a statistically significant difference between elevations measured by SRTM and the LiDAR DEM (see Table A-3). Subtracting the bare earth DEM from the SRTM DEM produced a DEM difference across the catchment, with

230

a mean = 2.56 m, a minimum = -15.18 m, a maximum = 33.23 m, a range = 48.41 m, and a STD = 2.7 m (see Table A-3).

Applying the same point density to the North Pole Stream produced 846 points.

The two sample t-test (1 m bare earth LiDAR DEM; SRTM DEM) computed a p-value =

0.942 (α = 0.05), highlighting no statistically significant difference between elevations measured by SRTM and the LiDAR DEM (see Table A-3). Subtracting the bare earth

DEM from the SRTM DEM produced a DEM difference across the catchment, with a mean

= 0.4 m, a minimum = -19.61 m, a maximum = 23.78 m, a range = 43.39 m, and a STD =

3.28 m (see Table A-3).

Table A-3 Two-sample t-tests comparing elevations measured by a LiDAR bare earth DEM, ‘LiDAR DEM’, and an SRTM DEM ‘SRTM’; and statistical outputs of a DEM difference calculated by subtracting the LiDAR DEM from the SRTM denoted by the head ‘DEM Diff.’

Cains River North Pole Stream Metric LiDAR DEM SRTM DEM Diff. LiDAR DEM SRTM DEM Diff. n 5287 5287 Raster 846 846 Raster Min (m) 14.1 18.0 -15.2 270.9 271.0 -19.6 Max (m) 184.7 190.0 33.2 726.8 724.0 23.8 Mean (m) 107.8 110.3 2.6 464.2 464.6 0.4 Std. Dev. (m) 34.2 34.4 2.7 101.4 100.5 3.3 p-value (two- tail) 0.0001 NA 0.942 NA

231

The DEM difference raster reveals the spatial variability of errors associated with the SRTM across both the Cains River and North Pole Stream, Figures A-4a and b, respectively. A potential tree-stand captured by the SRTM is denoted by red polygons in the Cains River and North Pole Stream, Figures A-4c and d, respectively, highlighting positive biases in the SRTM DEM. While negative biases associated with a lake surface are found in the North Pole Stream (Figure A-4d).

Figure A-4 The Cains River (a) and the North Pole Stream (b) DEM difference where the LiDAR DEM was subtracted from the SRTM DEM. Areas of disagreement between the SRTM and LiDAR, defined by the black lines in (a and b), in the Cains

(c) and the North Pole Stream (d) rather than bare earth, where the red and blue polygons highlight SRTM errors associated false measurements. 232

Solar radiation is well established as a driver of river temperature. To examine the relationship between the longitudinal temperature profiles in the TIR data and solar radiation models I created a point comparison across the North Pole Stream. The longitudinal temperature profile for the North Pole Stream is shown in Figure A-5a. The

Solar radiation temperature profile is shown in Figure A-5b, and the relationship between log (temperature) and log (solar radiation) is given in Figure A-5c. A fine-scale example of a solar radiation modelled section is displayed in Figure A-5d.

Figure A-5 The relationship between measured thermal infrared (TIR) river temperature on 29th July, 2009 in the North Pole Stream and stream distance, where

233

the red line denotes the mean temperature (a). The relationship between modelled solar radiation (for a clear sky) at the same locations for the time of flight and stream distance, where the red line defines the mean solar radiation (b). The relationship between log normal TIR temperature and log normal solar radiation where a weak relationship is shown between the 2 variables (c). An example section of a solar radiation model for the Cains River over 1 day (23rd July, 2008 – clear sky conditions)

(d).

To prevent multicollinearity within the models developed in Chapter 3, I ran a variance inflation factor analysis (VIF). The final model variables for both the Cains River and North Pole Stream are presented in Table A-4 and 5.

Table A-4 Variance inflation factor (VIF) analysis of multicollinearity for final

Cains River model variables

Cains River 1 km SRTM 1 km LiDAR 3 ha R² VIF R² VIF R² VIF 0.4 1.8 0.3 1.5 0.5 2.2 Wetlands Wetlands Wetlands 7 8 7 8 5 0 Valley wall 0.6 2.9 Valley wall 0.8 6.7 Stream 0.3 1.5 slope1 6 1 slope1 5 7 slope 3 0 0.4 1.9 Solar 0.2 1.3 Solar 0.2 1.3 Valley depth1 9 5 radiation 3 0 radiation 6 5 0.3 1.4 0.2 1.3 0.7 4.3 Stream slope Sinuosity Elevation1 2 6 7 8 7 3 Precipitation 0.2 1.3 Catchment 0.8 7.5 Catchment 0.6 2.5

(Oct-July) 5 4 slope1 7 0 slope1 0 1

234

Catchment 0.6 2.6 slope1 3 8 1 denotes average

Table A-5 Variance inflation factor (VIF) analysis of multicollinearity for final

North Pole Stream model variables

North Pole Stream 1 km SRTM 1 km LiDAR 3 ha R² VIF R² VIF R² VIF Valley wall Valley wall 0.41 1.71 Valley depth1 0.42 1.74 0.46 1.87 slope1 slope1 Valley Felsic 0.36 1.56 0.77 4.38 Stream slope 0.19 1.24 depth1 (granite) Solar Deep water 0.04 1.04 0.76 4.23 Solar radiation 0.15 1.18 radiation clastic Catchment sinuosity 0.10 1.11 0.23 1.29 sinuosity 0.15 1.18 slope1 Felsic Air Precipitation 0.77 4.29 0.50 1.98 0.47 1.88 (granite) temperature1 (Oct-July) Deep water Felsic 0.76 4.21 0.87 7.42 clastic (granite) Deep water 0.82 5.42 clastic Catchment 0.48 1.91 slope1 1 denotes average

Prediction model outputs and Standard Errord (SE) for the models built in

Chapter 3 are presented in Figures A6-11.

235

Figure A-6 Model plots for observed and predicted temperatures for the Cains River

1 km2 STRM aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN

2: Spherical Euclidean; spherical tail-down (c) models. The associated model predicted standard errors (SE) are presented adjacent to the corresponding models.

236

Figure A-7 Model plots for observed and predicted temperatures for the Cains River

1 km2 LiDAR aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN

2: Spherical tail-down (c) models. The associated model predicted standard errors

(SE) are presented adjacent to the corresponding models.

237

Figure A-8 Model plots for observed and predicted temperatures for the Cains River

3 ha aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN 2:

Spherical tail-down (c) models. The associated model predicted standard errors (SE) are presented adjacent to the corresponding models. 238

Figure A-9 Model plots for observed and predicted temperatures for the North Pole

Stream SRTM 1 km2 aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN 2: Spherical tail-down (c) models. The associated model predicted standard errors (SE) are presented adjacent to the corresponding models.

239

Figure A-10 Model plots for observed and predicted temperatures for the North Pole

Stream 1 km2 LiDAR aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN 2: Spherical tail-down (c) models. The associated model predicted standard errors (SE) are presented adjacent to the corresponding models.

240

Figure A-11 Model plots for observed and predicted temperatures for the North Pole

Stream 3 ha aspatial (a), spatial –SSN 1: Euclidean exponential (b) and spatial –SSN

2: Spherical tail-down (c) models. The associated model predicted standard errors

(SE) are presented adjacent to the corresponding models.

241

References

Cuthbert, M. O., Gleeson, T., Moosdorf, N., Befus, K. M., Schneider, A., Hartmann, J., & Lehner, B. (2019). Global patterns and dynamics of climate–groundwater interactions. Nature Climate Change. Nature Publishing Group. https://doi.org/10.1038/s41558-018-0386-4 Detenbeck, N. E., Morrison, A. C., Abele, R. W., & Kopp, D. A. (2016). Spatial statistical network models for stream and river temperature in New England, USA. Water Resources Research, 52(8), 6018–6040. https://doi.org/10.1002/2015WR018349 Isaak, D. J., Wenger, S. J., Peterson, E. E., Ver Hoef, J. M., Nagel, D. E., Luce, C. H., … Parkes-Payne, S. (2017). The NorWeST Summer Stream Temperature Model and Scenarios for the Western U.S.: A Crowd-Sourced Database and New Geospatial Tools Foster a User Community and Predict Broad Climate Warming of Rivers and Streams. Water Resources Research, 53(11), 9181–9205. https://doi.org/10.1002/2017WR020969 Jackson, F. L., Hannah, D. M., Fryer, R. J., Millar, C. P., & Malcolm, I. A. (2017). Development of spatial regression models for predicting summer river temperatures from landscape characteristics: Implications for land and fisheries management. Hydrological Processes, 31(6), 1225–1238. https://doi.org/10.1002/hyp.11087

242

Curriculum Vitae

Candidate’s Full Name: Antóin Mícheál O’Sullivan

Universities Attended:

University of Sussex, September, 2015 Brighton, UK Master of Science in Engineering Geomorphology, Consultancy and Practice

Edinburgh Napier University, May, 2014 Edinburgh, UK Bachelor of Engineering (Honours) in Civil Engineering

Limerick Institute of Technology, May, 2011 Limerick, Ireland Bachelor of Engineering (Ordinary) in Civil Engineering

Publications:

Refereed Publications in Periodicals (Submitted or Published)

Authorship in bold

1. O’Sullivan, A.M., Corey, E., Linnansaari, T., Cunjak, R., and Curry, R.A. (in review). Salmonid thermal habitat contraction during an extreme heat event in a hydrogeologically complex setting. Science of the Total Environment 2. Sutton, A.O., Studd, E., Fernandes, T., Bates, A., Bramburger, Cooke, S.J., Hayden, B., Henry, H.A.L., Humphries, M., Martin, R., McMeans, B., Moise, E., O’Sullivan, A.M., Sharma, S., and Templer, P.H. (in review). Frozen out: unanswered questions about winter biology. Environmental Reviews 3. Furze, S., O’Sullivan, A.M., Pronk, T., Allard, S., and Curry, R.A. (in review). Landscape-scale, high resolution, depth-to-bedrock mapping in a post-glacial landscape with shallow bedrock. Journal of Advances in Modeling Earth Systems 4. Helminen, J., O’Sullivan, A.M., and Linnansaari, T. (in review). Measuring tailbeat frequencies of three fish species from Adaptive Resolution Imaging Sonar (ARIS) data. Transactions of the American Fisheries Society

5. Studd, E.K., Bates, A.E., Bramburger, A.J., Fernandes, T., Hayden, B., Henry, H.A.L., Humphries, M.M., Martin, R., McMeans, B.C., Moise, E.R.D., O’Sullivan, A.M., Sharma, S., Sinclair, B., Sutton, A.O., Templer, P.H., and Cooke, S.J. (in review). Twelve Maxims for Winter. BioScience 6. Arsenault, M., O’Sullivan, A.M., Ogilvie, J., Gillis, C-A., Linnansaari, T., and Curry, R.A. (in review). High-resolution LiDAR and GIS model reveals extent of stream fragmentation across a forested landscape. Hydrobiolgia 7. Williams, M., Hernandez, C., O'Sullivan, A.M., April, J., Regan, F., Bernatchez, L., Parle-McDermott, A. (2020). Comparing CRISPR-Cas and qPCR eDNA assays for the detection of Atlantic Salmon (Salmo salar L.). Environmental DNA 8. O’Sullivan, A.M., Devito, K.J., Ogilvie, J., Linnansaari, T., Pronk, T., Allard, S., and Curry, R.A. (2020). Effects of topographic resolution and geologic setting on spatial statistical river temperature models. Water Resources Research 9. O’Sullivan, A.M., Wegscheider, B., Helminen, J., Cormier, J.G., Linnansaari, T., Wilson, D.A. and Curry, R.A. (2020). Catchment-scale, high-resolution, hydraulic models and habitat maps–a salmonid's perspective. Journal of Ecohydraulics, pp.116. 10. O’Sullivan, A.M., Samways, K.M., Perreault, A., Hernandez, C., Gautreau, M.D., Curry, R.A., and Bernatchez, L. (2020). Space invaders: searching for invasive Smallmouth Bass (Micropterus dolomieu) in a renowned Atlantic Salmon (Salmo salar) river. Ecology and Evolution 11. Wilbur, N.M., O’Sullivan, A.M., MacQuarrie, K.T.B., Linnansaari, T., and Curry, R.A. (2020). Characterizing physical habitat preferences and thermal refuge occupancy of Brook Trout (Salvelinus fontinalis) and Atlantic Salmon (Salmo salar) at high river temperatures. River Research and Applications 12. Andrews, S.N., O’Sullivan, A.M., Helminen, J., Arluison, D.F., Samways, K.M., Linnansaari, T. and Curry, R.A. (2020). Development of Active Numerating Side Scan for a High-density Overwintering Location for Endemic Shortnose Sturgeon (Acipenser brevirostrum) in the Saint John River, New Brunswick. Diversity, 12(1), p.23.

13. O'Sullivan, A.M., Linnansaari, T. and Curry, R.A. (2019). Ice cover exists: A quick method to delineate groundwater inputs in running waters for cold and temperate regions. Hydrological Processes, 33(26), pp.3297-3309. 14. O'Sullivan, A.M., Devito, K.J. and Curry, R.A. (2019). The influence of landscape characteristics on the spatial variability of river temperatures. CATENA, 177, pp.70-83. 15. Arenson, L.U., Kääb, A. and O'Sullivan, A.M. (2016). Detection and analysis of ground deformation in permafrost environments. Permafrost and Periglacial Processes, 27(4), pp.339-351.

Conference Co-Convener:

 Permafrost Engineering in Mountainous Terrain Session - 11th International Conference on Permafrost: Potsdam, Germany (June, 2016)

Conference Presentations:

Invited guest speaker

1. O’Sullivan, A. M. (2020). Bottoms up: how landscape structure and composition influences river temperature. Geological Society of America. 26 – 30th October, 2020. Regular presentations

1. O’Sullivan, A. M., Linnansaari, T., and Curry, R. A. (2019). Necessary ice conditions exist (NiCE): A quick and free method to delineate groundwater inputs in running waters. Atlantic Salmon Ecosystem Forum, Québec, 12-13th March, 2019 2. O’Sullivan, A. M., Devito, K. J., Linnansaari, T., Ogilvie, J., Pronk, T., Allard, S. and Curry, R. A (2019). Cryptic streams: contrasting roles for stream temperature resilience is defined by geologic setting. New Brunswick Department of Natural Resources, Fredericton, NB. 19th April, 2019 3. O’Sullivan, A. M., Devito, K. J., Linnansaari, T., Ogilvie, J., Pronk, T., Allard, S. and Curry, R. A (2019). Cryptic streams: contrasting roles for stream temperature resilience is defined by geologic setting. 2nd CAST Science Colloquium, Miramichi, NB. 21st May, 2019

4. O’Sullivan, A. M., Devito, K. J., Linnansaari, T., and Curry, R. A (2019). What lies beneath: A bottom-up view of forest hydrology. Canaan-Washademoak watershed science meeting. Codys Womens Institute Hall. NB. 17th June, 2019 5. O’Sullivan, A. M., Devito, K. J., Linnansaari, T., and Curry, R. A (2019). Interactions between forests and streams . . . and the influence of geology. Fornebu Lumber, Miramichi, NB. 30th July, 2019 6. O’Sullivan, A. M., Linnansaari, T., and Curry, R. A. Ice conditions exist (ICE) (2019): A quick and free method to delineate groundwater inputs in running waters. CSEE, ESC, & AES 2019 Joint Meeting. Fredericton, NB. August 18th – 21st, 2019 7. O’Sullivan, A. M., Wegscheider, B., Linnansaari, T., Wilson, D., and Curry, R. A (2019). Remotely sensed, high resolution, catchment scale habitat mapping for rivers. International Society for River Science, Vienne, Austria. September 8th – 13th, 2019

Grants:

1. O’Sullivan (2020). What is an optimal thermal refugia for wild, adult Atlantic Salmon - $7,990. NBWTF 2. Linnansaari and O’Sullivan (2020). Forest Hydrology Project: Sustainable Co- Management of New Brunswick’s Forests, Water and Atlantic Salmon - $90,000. NBIF 3. O’Sullivan and Curry (2019). How physiography and climate change influences of forest harvesting effects on Atlantic Salmon habitats - $75,000. ASCF 4. O’Sullivan and Samways (2017). Fishing with eDNA: mapping the spatial distribution of Striped Bass, and searching for Smallmouth Bass, in the Miramichi Watershed, New Brunswick - $8,500. NBWTF 5. O’Sullivan and Samways (2017). Fishing with eDNA: mapping the spatial distribution of Striped Bass, and searching for Smallmouth Bass, in the Miramichi Watershed, New Brunswick – $19,750. ASCF 6. O’Sullivan and Curry (2016). Predicting Temperature Changes in the Miramichi Watershed - $25,000. ETF

7. O’Sullivan and Curry (2015). Mapping stream temperature - $4,570. NBWTF

Total funding from grants: $230,810

Scholarships:

 Jack T.H. Fenety Scholarship, 2020  Hynes Scholarship, 2019  Mr. and Mrs. Art Van Slyke, 2017 and 2018  Bud and Peggy Bird Wild Atlantic Salmon Conservation Scholarship 2017  Jack T.H. Fenety Scholarship, 2016  Canadian Rivers Institute WATER programme scholarship, 2015  CH2M HILL Scholarship (Engineering Geomorphology), 2014  University of Sussex Chancellor’s Masters Scholarship, 2014

Media:  Radio 1. Information Morning Fredericton. Low water levels in N.B. rivers. August, 19, 2020. https://www.cbc.ca/listen/live-radio/1-25-information-morning-fredericton-from- cbc-radio-new-brunswick-highlights/clip/15793387-low-water-levels-n.b.-rivers 2. Maritime Connection. The latest on the floods along the Saint John River. April, 21, 2019. https://www.cbc.ca/player/play/1501749315942 3. Information Morning Saint John. Rivers Institute sees low water levels in New Brunswick since dry summer. October 11, 2017. http://www.cbc.ca/player/play/1069691459820 4. Information Morning Fredericton. River Levels. October 10, 2017 http://www.cbc.ca/player/play/1068092995637

 Articles

1. Telegraph: Water temperature could shape salmon protection – expert. Nathan Delong. Telegraph, Miramichi Leader. April 11th 2018. https://www.telegraphjournal.com/miramichi- leader/story/100565983/collaboration-for-atlantic-salmon-tomorrow-cast-thermal- update?source=story-related

2. CBC: Dry summer is causing river levels to drop and stressing out fish

Jordan Gill · CBC News · Oct 10, 2017 https://www.cbc.ca/news/canada/new-brunswick/rivers-low-hurt-fish-1.4347224