THE PALAEOBIOLOGY OF AVALONIAN (EDIACARAN) RANGEOMORPHS

Charlotte Guenevere Kenchington

Selwyn College

University of Cambridge

December 2015

This dissertation is submitted for the degree of Doctor of Philosophy

SUMMARY: THE PALAEOBIOLOGY OF AVALONIAN (EDIACARAN) RANGEOMORPHS

CHARLOTTE G. KENCHINGTON

The Earth has supported life for most of its 4.5 billion year history, but the first macroscopic organisms only appeared some 600 million years ago, in the Ediacaran. Their world was fundamentally different from the one we know today, and many aspects of their biology and ecology remain a mystery. The late Ediacaran fossil assemblages of Avalonia represent some of the oldest evidence for complex macroscopic life, and are dominated by rangeomorphs, a group characterised by their self-similar branching architecture. In this thesis, I investigate several aspects of the preservation, classification and ecology of these enigmatic deep marine organisms.

The biotas of Charnwood Forest host several taxa which are new to science. Five of these are described here, and include two new genera, Orthiokaterna fordi gen. et sp. nov., and Undosyrus nemoralis gen. et sp. nov., and three new species: Primocandelabrum anatonos sp. nov., P. boytoni sp. nov., and P. katatonos sp. nov.. The Primocandelabrum species in particular encompass a great deal of variation in both branching characters and overall morphology. By using a novel multivariate statistical approach to analyse multiple characters in tandem, individual taxa can be discriminated from one another. Much of the observed variation is interpreted as intra-specific. This level of variation within a single taxon has not previously been recognised in rangeomorphs, and is likely attributable to (eco?)phenotypic rather than ontogenetic variability. Orthiokaterna displays eccentric branches, interpreted as a growth response to mechanical damage, reflecting a greater degree of growth plasticity than that recognised in other rangeomorphs, while Undosyrus had an external sheath, interpreted as modified rangeomorph elements serving a protective role.

Even without knowing the phylogenetic relationships of rangeomorphs, it is possible to resolve key aspects of their palaeoecology. The response(s) of communities in Charnwood Forest and Newfoundland to both ambient disturbance and to more substantial events is investigated by combining detailed petrographic analysis of the host sediments with multivariate statistical techniques. I demonstrate that higher taxonomic diversity is correlated with low–intermediate physical disturbance; that upright taxa (e.g. Charnia) dominate surfaces which experienced small-scale, sub-lethal sedimentation events and comparably high background sediment input; and that flat-lying forms (e.g. Fractofusus) preferentially occur on surfaces with low sediment input. The population demographics of several taxa also show evidence of multimodality: in some (including Charnia and Primocandelabrum), bimodality was

I induced by culling of part of an incumbent population by a substantial disturbance event, followed by re-colonisation; in others (e.g. Fractofusus), overlapping cohorts reflect non-continuous or pulsed reproduction. Disturbance (ambient and discrete events) demonstrably influenced community succession, with early-colonising taxa dominating horizons with low overall levels of disturbance, and those able to survive disturbance events dominating recovery populations and horizons with higher levels of disturbance. Based on their inferred life history traits and their environmental preferences, I propose a model of ecological succession for rangeomorph communities.

II DECLARATION

Declaration of Originality

This dissertation is the result of my own work and includes nothing which is the outcome of work done in collaboration except as declared below and specified in the text. It is not substantially the same as any that I have submitted, nor is it being concurrently submitted for a degree or diploma or other qualification at the University of Cambridge or any other University or similar institution.

Published work arising from the dissertation

Parts of Chapter 3 are based on material presented in Kenchington & Wilby (2014), paper entitled Of time and taphonomy: preservation in the Ediacaran and published in The Paleontological Society Papers, volume 20. Parts of the interpretation in Chapter 9 (and Fig. 9.13) are based on work I published in collaboration with P.R. Wilby and R.L. Wilby (Wilby et al. 2015), paper entitled: Role of low intensity environmental disturbance in structuring the earliest (Ediacaran) macrobenthic tiered communities, Palaeogeography, Palaeoclimatology, Palaeoecology, volume 434. Figure 10.1 was published in Mitchell et al. 2015, paper entitled: Reconstructing the reproductive mode of an Ediacaran macro-organism, published in Nature, volume 524. Figure 10.3 was published in Liu et al. 2015a, paper entitled: Remarkable insights into the paleoecology of the Avalonian Ediacaran macrobiota, Gondwana Research, volume 27.

Statement of Length

This dissertation is within the prescribed length as defined by the Degree Committee for Earth Sciences and Geography.

III IV CONTENTS

SUMMARY: THE PALAEOBIOLOGY OF AVALONIAN (EDIACARAN) RANGEOMORPHS ...... I

DECLARATION ...... III

FIGURES ...... XI

TABLES ...... XIII

ACKNOWLEDGEMENTS ...... XIV

1. INTRODUCTION: THE EDIACARAN AND AVALONIA ...... 1 1.1 Global context for life Ediacaran ...... 1 1.1.1 Glaciations and oxygenation ...... 3 1.2 Life in the Ediacaran ...... 3 1.2.1 Rangeomorphs ...... 5 1.2.2 Arboreomorphs...... 10 1.2.3 Discoidal forms ...... 10 1.2.4 Other taxa ...... 10 1.3 Rangeomorph growth and functional morphology ...... 11 1.4 The Ecology of Avalonian Rangeomorphs ...... 12

2. GEOLOGICAL CONTEXT FOR THE EDIACARAN OF AVALONIA ...... 14 2.1 Ediacaran Stratigraphy of Avalonia ...... 14 2.1.1 The Geological Setting of Charnwood Forest ...... 14 2.1.1.1 Age constraints for Charnwood Forest ...... 16 2.1.1.2 Fossil occurrences in Charnwood Forest ...... 17 2.1.1.3 Detailed lithological analysis of the Bradgate Formation ...... 19 2.1.2 The Geological Setting of the Ediacaran of Newfoundland ...... 20 2.1.2.1 Age constraints for Newfoundland ...... 24 2.1.2.2 Deep-water fossil occurrences in Newfoundland ...... 25 2.2 Comparing the Newfoundland and Charnwood Forest successions...... 25

3. THE TAPHONOMY OF AVALONIAN RANGEOMORPHS ...... 27 3.1. Introduction ...... 27 3.2 Metamorphism ...... 28 3.3. Mouldic preservation: mechanisms and mineralisation ...... 28 3.3.1 The “death mask” hypothesis ...... 29 3.3.2 The timing of pyritisation: new evidence from Charnwood Forest and Newfoundland ...... 30 3.3.3 Phosphatisation ...... 33 3.3.4 Discussion ...... 36 3.3.4.1 The location of pyritisation and variability of preservation ...... 36 3.3.4.2 The timing of pyritisation ...... 37

V 3.3.4.3 The importance of different mineral phases in Avalonian preservation ..... 38 3.4. Biostratinomy and the gradation of forms ...... 38 3.4.1 Alteration of the preserved morphology of multifoliate rangeomorphs ...... 39 3.4.2 Variation in preservation across a single specimen ...... 41 3.4.2.1 Variation across a single rangeomorph branch ...... 41 3.4.3 Iveshediomorphs ...... 43 3.5 Conclusions and future work ...... 44

4. A PLETHORA OF PRIMOCANDELABRUM: MUTLIVARIATE STATISTICAL ANALYSIS AND THE CLASSIFICATION OF RANGEOMORPHS ...... 46 4.1 Introduction ...... 46 4.1.1 Charnwood Forest – a plethora of multifoliate taxa ...... 47 4.2 Statistical techniques – an objective approach to taxonomy? ...... 50 4.2.1 Material ...... 50 4.2.2 Methods ...... 50 4.3 Results ...... 52 4.3.1 Hierarchical Clustering ...... 54 4.3.1.1 Scaled vs non-scaled ...... 55 4.3.1.2 All specimens vs min filt dataset ...... 55 4.3.1.3 Discrimination of Orthiokaterna ...... 55 4.3.1.4 Assignment of “dusters” ...... 56 4.3.1.5 Removing outsize specimens ...... 57 4.3.1.6 Assignment of Primocandelabrum individuals ...... 57 4.3.1.7 Removing “dusters” ...... 59 4.3.1.8 All characters versus morphologic characters versus branching characters only ...... 59 4.3.1.9 Reduced character matrix ...... 61 4.3.2 Assessment of cluster assignment ...... 61 4.3.3 Summary of results ...... 61 4.4 Discussion...... 63 4.4.1 The potential functional significance of rangeomorph branches ...... 69 4.4.2 Influence of taphonomy ...... 71 4.4.3 Validity of the statistical approach ...... 72 4.4.3.1 Selection of the most appropriate data sub(set) ...... 72 4.4.3.2 Categorical, continuous characters, or both? ...... 72 4.4.4 Placement of “dusters” ...... 73 4.4.5 Division of Primocandelabrum ...... 74 4.5 Systematic Palaeontology ...... 76 4.6 Conclusions ...... 77

5. ORTHIOKATERNA FORDI AND THE GROWTH OF RANGEOMORPHS ...... 79 5.1 Introduction ...... 79

VI 5.2 Systematic Palaeontology ...... 80 5.3 Comparison to existing taxa ...... 83 5.3.1 Orthiokaterna and Bradgatia ...... 84 5.3.2 Orthiokaterna and Primocandelabrum ...... 88 5.4 Discussion: taphonomy, anatomical reconstruction and functional morphology ...... 88 5.4.1 The disc ...... 88 5.4.1.1 Functional morphology of the disc ...... 89 5.4.2 The stem ...... 90 5.4.2.1 Functional morphology of the stem ...... 91 5.4.3 The frond ...... 91 5.4.3.1 Functional morphology of the frond ...... 93 5.4.4 Eccentric branches ...... 94 5.4.4.1 Potential mechanisms and consequences for rangeomorph biology ...... 95 5.4.4.2 Possible causes and functions of eccentric branches ...... 96 5.4.4.3 Restriction of eccentric branching to multifoliate rangeomorphs ...... 98 5.5 Conclusions ...... 98

6. UNDOSYRUS NEMORALIS: A STRATEGY AGAINST SEDIMENT OCCLUSION ...... 100 6.1 Introduction ...... 100 6.2 Systematic Palaeontology ...... 100 6.3 Discussion...... 104 6.3.1 Interpretation of specific features unique to the brush structure ...... 104 6.3.2 Comparison to other multifoliate taxa: Primocandelabrum and Orthiokaterna .. 106 6.3.2.1 Methods ...... 107 6.3.2.2 Results ...... 110 6.3.2.3 Discussion of results ...... 111 6.3.3 Hypothesis 1: the brush structure was not part of the rangeomorph ...... 111 6.3.3.1 Hypothesis 1a: the brush structure is a growth of another organism on a rangeomorph ...... 112 6.3.3.2 Hypothesis 1b: the brush structure is another taxon superimposed on a rangeomorph ...... 113 6.3.4 Hypothesis 2: the brush structure was part of the rangeomorph ...... 113 6.3.5 Was the rangeomorph Primocandelabrum, or another taxon? ...... 114 6.3.6 The purpose of the brush structure, and its bearings on rangeomorph biology ... 116 6.7 Conclusions ...... 119

7. TIERING IN AVALONIAN RANGEOMORPH-DOMINATED COMMUNITIES ...... 120 7.1 Introduction ...... 120 7.2 Material ...... 121 7.3 Methods ...... 122 7.4 Results ...... 124 7.4.1 Bed B...... 124

VII 7.4.2 Lower Mistaken Point (LMP) ...... 126 7.4.3 Mistaken Point D surface (MPD)...... 127 7.4.4 Mistaken Point E surface (MPE) ...... 127 7.4.5 Mistaken Point G surface (MPG)...... 130 7.4.6 Shingle Head (SH) ...... 132 7.4.7 Bristy Cove (BC)...... 132 7.4.8 Pigeon Cove (PC)...... 133 7.4.9 Summary of the results ...... 134 7.5 Discussion...... 135 7.5.1 Tiering and the rangeomorph mode of life ...... 137 7.6 Conclusions ...... 139

8. THE ROLE OF AMBIENT DISTURBANCE ON COMMUNITY COMPOSITION ...... 141 8.1 Introduction ...... 141 8.1.1 Disturbance and community structure ...... 141 8.2 Material and Methodology ...... 143 8.2.1 Methods ...... 143 8.2.2 Defining the explanatory variables: characterisation of the level of physical disturbance ...... 143 8.2.3 Defining the dependent variables: aspects of community composition ...... 144 8.2.4 Analyses of the correlation(s) between explanatory and dependent variables. ....145 8.2.4.1. Linear regression analysis...... 145 8.2.4.2 Principal components analysis and hierarchical clustering analysis ...... 146 8.3 Results ...... 147 8.3.1 Regression analyses ...... 147 8.3.2 Principal components analysis and hierarchical clustering analysis...... 149 8.3.2.1 Sedimentological variables only active...... 149 8.3.2.2 Biological variables only active...... 149 8.3.2.3 Morphotype variables only active ...... 151 8.3.2.4 All variables active ...... 152 8.3.3 Summary of analytical results ...... 153 8.4 Discussion...... 154 8.4.1 Interpretation of sedimentary fabrics and the coarse grain component ...... 154 8.4.1.1 Silt lenses and laminae ...... 154 8.4.1.2 Silt aggregates and floating grains ...... 155 8.4.2.1 Morphotype and sediment influx ...... 156 8.4.2.2 Branching architecture and ambient disturbance ...... 157 8.4.2.3 Taxonomic diversity and disturbance ...... 158 8.5.1 The persistence of Fractofusus ...... 159 8.6 Conclusions and future work ...... 160

VIII 9. MULTIMODALITY: DISTURBANCE AND REPRODUCTION ...... 161 9.1 Introduction ...... 161 9.2 Material ...... 162 9.3 Methods ...... 162 9.4 Results ...... 163 9.4.1 Bed B ...... 163 9.4.2 Lower Mistaken Point ...... 164 9.4.3 Mistaken Point D ...... 164 9.4.4 Mistaken Point E ...... 165 9.4.5 Mistaken Point G ...... 166 9.4.6 Pigeon Cove ...... 166 9.4.7 Bristy Cove ...... 167 9.4.8 Shingle Head...... 167 9.4.9 H14 surface ...... 168 9.4.10 Summary of results for each taxon ...... 168 9.5 Discussion: The myriad causes of multimodality...... 169 9.5.1 To log or not to log ...... 169 9.5.2 The normality of the distributions ...... 184 9.5.3 Multiple, discrete populations ...... 186 9.5.3.1 Bimodality on Bed B ...... 187 9.5.4 Multiple, overlapping populations ...... 189 9.5.5 Surfaces with multiple types of multimodality ...... 190 9.5.6 Unimodal distributions ...... 191 9.5.7 The influence of disturbance on community composition ...... 191 9.6 Conclusions ...... 192

10. THE PALAEOECOLOGY OF AVALONIAN RANGEOMORPH COMMUNITIES ...... 193 10.1 The ecological strategies of rangeomorphs and the succession of their communities ...193 10.1.1 Low-disturbance communities and adaptation of rangeomorphs to competition ...... 194 10.1.2 Higher disturbance environments and resilience to disturbance ...... 195 10.2 The evolution of a stem: response to competition or reproduction? ...... 197 10.3 The influence of rangeomorphs on their environment ...... 198

REFERENCES ...... 200

APPENDIX 1: MULTIVARIATE STATISTICAL TECHNIQUES ...... 219 1.1 Data treatment ...... 219 1.1.1 Iterations of the data ...... 219 1.1.2 Scaling data ...... 220 1.1.3 Dealing with missing data ...... 220 1.2 Multivariate analysis and reduction of dimensionality ...... 221

IX 1.2.1 Principal Components Analysis (PCA) ...... 221 1.2.3 Factor Analysis of Mixed Data (FAMD) ...... 222 1.3 Clustering ...... 222 1.3.1 Hierarchical clustering ...... 223 1.3.2 Model-based clustering and determination of the number of mixing components ...... 224 1.4 References unique to this section ...... 225

APPENDIX 2: DATA TABLES ...... 227 A.2.1 Specimens from Charnwood Forest ...... 227 A.2.2 Thin Section Data ...... 229 A.2.3 Framboid maximum dimension counts ...... 231 A.2.4 Raw measurements of Primocandelabrum and Orthiokaterna ...... 233 A.2.5 Proportions of Primocandelabrum and Orthiokaterna (divided by total height) 234 A.2.6 Branching architecture of Primocandelabrum and Orthiokaterna ...... 235 A.2.7 Taxonomic assignments of Primocandelabrum specimens ...... 237 A.2.8 Regression calculations for size data for Orthiokaterna ...... 238 A.2.9 Raw measurements and branching architecture of Undosyrus ...... 240 A.2.10 Peak heights and overlap distributions ...... 242

X FIGURES

Figure 1.1. Placing the Ediacaran in a global chronostratigraphic, geochemical, tectonic and evolutionary context...... 2 Figure 1.2 Avalonia...... 4 Figure 1.3. Rangeomorphs (a—d, f), arboreomorphs (e) and Aspidella (g)...... 6 Figure 1.4 Rangeomorph branching architecture...... 7 Figure 2.1. A) Geological map and B) stratigraphic column of the Ediacaran—Cambrian succession of Charnwood Forest, modified after Carney (1999)...... 16 Figure 2.2 Detailed log of the section of the Bradgate Formation around Bed B ...... 18 Figure 2.3. Geological maps and lithostratigraphy of Newfoundland...... 21 Figure 2.4. Epiclastic and volcaniclastic event beds in thin section...... 23 Figure 3.1. Distribution and occurrence of pyrite ...... 31 Figure 3.2. Pyrite textures ...... 32 Figure 3.3. Framboid density distribution profiles ...... 33 Figure 3.4. Pyrite textural relationships ...... 34 Figure 3.5. Pyrite and phosphate textural relationships ...... 35 Figure 3.6. Taphonomy of fronds...... 40 Figure 3.7. Taphonomic mechanism to explain the differential preservation across a rangeomorph element ...... 43 Figure 4.1. Stalked, multifoliate rangeomorphs from Bed B, Charnwood Forest...... 48 Figure 4.2. Continuous and categorical characters in rangeomorphs ...... 53 Figure 4.3. Profile plots ...... 56 Figure 4.4. Results of the cluster analysis ...... 58 Figure 4.5. Results of the cluster analysis ...... 60 Figure 4.6. Results of the cluster analysis ...... 62 Figure 4.7. Profile plots and scatter plots of Orthiokaterna and Primocandelabrum ...... 70 Figure 5.1. Orthiokaterna fordi...... 81 Figure 5.2. Detailed branching architecture of Orthiokaterna...... 84 Figure 5.3. Eccentric branching in Orthiokaterna ...... 85 Figure 5.4. Bradgatia...... 87 Figure 5.5. Artist’s reconstruction of Orthiokaterna fordi ...... 95 Figure 6.1. Undosyrus, specimen 3Bx (GSM 105872) ...... 101 Figure 6.2. Undosyrus specimen 5E4 (GSM 105962) ...... 103 Figure 6.3. Undosyrus specimen 3A1 (GSM 106009) ...... 105 Figure 6.4. Results of the hierarchical cluster analysis following principal component analyses of Primocandelabrum, Orthiokaterna, Undosyrus and “dusters”...... 108 Figure 6.5. Artist’s reconstruction of Undosyrus nemoralis ...... 115 Figure 7.1. Rangeomorph morphotypes...... 121 Figure 7.2. Distribution of frondose parts in the water column ...... 125

XI Figure 7.3. Density distribution plots of total heights of taxa and morphotypes in the water column ...... 128-129 Figure 7.4. Notched box plots of taxon and morphotype distributions of total heights in the water column ...... 131 Figure 7.5. Median distributions across taxa and surfaces ...... 133 Figure 8.1. Sedimentological fabrics...... 148 Figure 8.2. Results of the cluster analysis ...... 150 Figure 9.1. The population structures on Bed B, Charnwood Forest for raw (non-logged) values .170 Figure 9.2. The population structures on Bed B, Charnwood Forest for logged values ...... 171 Figure 9.3. The population structure on Lower Mistaken Point for raw (non-logged) values ...... 172 Figure 9.4. The population structure on Lower Mistaken Point for logged values ...... 173 Figure 9.5. The population structures on Mistaken Point D surface for raw (non-logged) values ..174 Figure 9.6. The population structures on Mistaken Point D surface for logged values ...... 175-176 Figure 9.7. The population structures on Mistaken Point E surface for raw (non-logged) values ...... 177-178 Figure 9.8. The population structures on Mistaken Point E surface for logged values ...... 179 Figure 9.9. The population structures on Mistaken Point G (MPG) and Pigeon Cove (PC) for raw (non-logged) values ...... 180 Figure 9.9. The population structures on Mistaken Point G (MPG) and Pigeon Cove (PC) for logged values ...... 181 Figure 9.11. The population structure on Bristy Cove (BC), Shingle Head (SH) and H14 for raw (non-logged) values ...... 182 Figure 9.12. The population structures on Bristy Cove (BC), Shingle Head (SH), and H14 for logged values ...... 183 Figure 9.13. Sequence of events leading to discrete bimodal population structures (e.g. on Bed B and LMP), shown for Charnia masoni ...... 188 Figure 10.1. Artist’s reconstruction of colonies of Fractofusus, with three generations depicted. . 194 Figure 10.2. Ecological succession for rangeomorph communities ...... 196 Figure 10.3. Middle- to late-stage communities in low to intermediate disturbance environments, characterised by a diverse community in which early stage species persist ...... 198

XII TABLES Table 1.1. Terminology for rangeomorph branching architecture ...... 9 Table 4.1. Quantitative (morphological) and qualitative (branching) variables categorising the clusters as determined by heirarchical clustering analysis ...... 65 Table 4.2. Quantitative (morphological) and qualitative (branching) variables categorising the clusters as determined by heirarchical clustering analysis ...... 67 Table 4.3. Comparing percentage of individuals placed in the same group (percentage group match) between pairs of iterations ...... 68 Table 4.4. Comparing percentage of individuals placed in the same group (percentage group match) between pairs of iterations excluding Orthiokaterna...... 69 Table 6.1. Results of the heirachical cluster analysis following principal components analysis. ....109 Table 8.1. Descriptions, abbreviations and ranges of the explanatory (sedimentary) variables...... 144 Table 8.2. Descriptions, abbreviations and ranges of the descriptor (biological) variables...... 145 Table 8.3. The ten most highly correlated pairs of variables from the linear regression analysis of single pairs of variables ...... 152 Table 8.4. Results from the stepwise deletion regression analyses...... 152 Table 9.1. Number of populations supported through model-based clsutering algorithms and normality of constituent populations for raw values...... 185

XIII ACKNOWLEDGEMENTS

My sincerest thanks go to my two supervisors for their support and for the opportunity to study these fascinating fossils. To Nick Butterfield, my thanks for his support throughout this project, for his ability to play devil’s advocate, and for always pushing me to challenge myself. To Phil Wilby, my thanks for guiding me through the early parts of this project, and for making me question everything Ediacaran. This project was funded by NERC (grant number NE/I005927/1), the British Geological Survey, and the Cambridge Home and European Student Scholarships.

I have been fortunate to engage with many people in discussing all things Ediacaran. Foremost among these is Alex Liu, who really helped me get to grips with rangeomorphs. His support throughout this project has been invaluable. Secondly, to Emily Mitchell, for discussing all things statistical and ecological. My thanks also to Marc Laflamme and Simon Conway Morris for their plentiful and informative discussions, and also to Jen Hoyal Cuthill and Sarah Tweedt.

Field work in the Mistaken Point Ecological Reserve and on the Bonavista Peninsula in Newfoundland was carried out under Scientific Research Permits granted by the Parks and Natural Area Division, Government of Newfoundland and Labrador. My thanks to Alex Liu and Jack Matthews for their assistance in the field, and for prooving to be excellent companions. Field work in Charnwood Forest was carried out with the permission and support of Natural England, the Bradgate Park Trust and with Charnwood Forest Golf Club. My thanks also to Matt Clapham for making his dataset of specimen heights available to use, to Emily Mitchell for providing the dataset for the H14 surface, and to Alex Liu for providing is taxonomic abundance dataset. This project required the making of hundreds of jesmonite resin casts, and so my sincerest thanks to Sue Martin for teaching me the technique and for making me feel welcome at the British Geological Survey. My thanks also to Louise Neep, Scott Renshaw and Paul Shepherd for arranging my visits their and for organising the loan of the casts, and to Paul Witney and Simon Harris for their guidance in fossil photography.

My heartfelt thanks also to various members of the Department of Earth Sciences for their support and friendship over the past four years, most especially to my officemate, Nick Wiggan, and to the incomparable Sarah Humbert.

My thanks to my parents, Richard and Sue, and to my brother Laurence for their support over the duration of this project, and especially during these last few months. Lastly, my deepest thanks to my partner Michael, without whose unwavering love and support this thesis would never have been written.

XIV C. Kenchington, 2014; published in Liu et al., 2015

“It’s life, Jim, but not as we know it”

- Spock, in Rory Kehoe’s Star Trekkin’

XV XVI 1. INTRODUCTION: THE EDIACARAN AND AVALONIA

The Ediacaran Period spans a critical interval in the history of life: the transition from the microbial world of the Proterozoic to the modern-style ecosystems of the Phanerozoic. Ediacaran macrofossils are the oldest record of complex life currently known from the rock record, and are inherently problematic: they have no known extant or fossil relatives, bearing only a passing resemblance to anything alive today or throughout geological history. After 3 billion years of exclusively microbial life, the appearance of animals heralded a drastic shift in ecological structure, nutrient cycling and oceanic chemistry (Butterfield, 2011). The enigmatic organisms that dominated the oceans during the Ediacaran shape our ideas of early metazoan evolution and the development of Phanerozoic ecosystem structure, and provide a means of calibrating molecular clocks and the origination of metazoan clades (Erwin et al. 2011). Even without knowing where these enigmatic organisms fit in the tree of life, it is possible to gain some insight into their basic biology – how they functioned, and how they interacted with their environment.

Ediacaran macrofossils have been the subject of intensive study for several decades, but many aspects of their biology have remained a mystery. In this thesis, I focus on some of oldest and most puzzling of the Ediacaran communities, those of the demonstrably deep-water successions of Newfoundland in Canada and Charnwood Forest in the UK. Three broad areas of study have the potential to substantially illuminate our understanding of these ancient organisms: their mode of preservation, their functional morphology, and their ecology. Any interpretation of an organism’s biology rests on understanding the taphonomic processes that influenced it during its preservation, and the environment in which it lived and was buried. By placing the fossils in their broad palaeoenvironmental (Chapter 2) and taphonomic context, original morphology can be discriminated from taphonomic artifice (Chapter 3). Given the lack of extant correlatives or close modern analogues for Ediacaran organisms, detailed aspects of their morphology can reveal how an organism and its constituent parts functioned in life (Chapters 4—6) – of critical importance for understanding how the organism lived. The ecology of an organism is a fundamental aspect of its biology. Its interaction with other organisms in the community, and the ways in which individuals and communities respond to environmental parameters (such as physical disturbance), reveal the underlying complexity of the organisms, the communities and their ecological networks.

1.1 Global context for life Ediacaran

The appearance of complex macro-organisms in the Ediacaran coincided with major global tectonic, climatic and geochemical upheaval, all of which have been linked to the biotic revolution of this time. By the latest Ediacaran, the low latitutude supercontinent of Rodinia/Pannotia had broken up, producing the main Palaeozoic continents of Baltica, Laurentia and Siberia (Li et al. 2008, Scotese 2009). The southern hemisphere landmass of Gondwana had almost assembled by this time (Fig. 1.1). A volcanic island

1 arc developed adjacent to a subduction zone a little north of this landmass, and once it had rifted from Gondwana it formed the microcontinent of Avalonia, the remnants of which comprise Newfoundland and Charnwood Forest (Fig. 1.2; Chapter 2). The Ediacaran was a period of prolonged and pronounced continental rifting and assembly (Li et al. 2008, Scotese 2009), and this dynamic period of tectonic evolution doubtless had significant influence on global climate and ocean chemistry.

The chemistry of Ediacaran oceans was different to that of Phanerozoic ones. It has been suggested on the basis of carbon isotopic data that late Ediacaran dissolved organic carbon (DOC) concentrations were 2 to 3 orders of magnitude higher than in modern oceans (Rothman et al. 2003). However, as has been noted by Bristow and Kennedy (2008), there is no independent geological evidence for such high DOC concentrations aside from the carbon isotopic data. This is of critical import for constraining the mode of life of Ediacaran organisms (Liu et al. 2015a). Ediacaran oceans were arguably lower in sulphate concentration (Halverson & Hurtgen 2007), and lower in oxygen than modern ones – although the exact timing, synchronicity and mechanisms behind oxygenation of the deep oceans is still a matter of some debate.

Chrono- 1 Geochemistry2, 3 Major evolutionary 5 Assemblage and 6 stratigraphy Glaciations and tectonic events 4 developments Taxonomic Range 2 2+ 2

δC13 (‰)

Ma Period 10.0 5.0 0.0 -5.0 White Sea Deep water O Deep water Fe Surface water O Metazoan-based Ecosystems Metazoan Biomineralisation Metazoan motility Animals Fungi Green algae Red algae Acanthomorphic Acritacrhs Nama Megascopic size Aspidella Rangeomorphs Charniodiscus Thectardis Avalon

525 Gondwanaland 541 Cambrian assembled 550

575 Gaskiers Early assembly of Gondwanaland 600 (S. latitudes) Ediacaran

625 635 Marinoan 650 Break-up of Rodinia near completion 5 675 Burrowing Through gut5 700 Sturtian Megascopic predation 7 Break-up of 5 Filter feeding? 725 Rodinia (low- Cryogenian latitude) begins Protistan predation8 Figure 1.1. Placing the Ediacaran in a global chronostratigraphic, geochemical, tectonic and evolutionary context. Figure modified after the following: 1) Gradstein et al. (2012); 2) Carbon isotope data after Shields-Zhou and Och (2011); 3) oxygenation data after Canfield et al. 2007; 4) Hoffman and Li 2009; 5) Lenton et al., 2014, and references therein; 6) Laflamme et al. (2013) and references therein; 7) Hua et al. (2003); 8) Cohen et al. 2011.

2 1.1.1 Glaciations and oxygenation

During the Cryogenian (the Period immediately preceding the Ediacaran, Fig. 1.1), the supercontinent of Rodinia lay over the equator, and the climate was characterised by severe, long-lasting and wide-spread glaciations – the evocative Sturtian (~715 Ma) and Marinoan (~635 Ma) “Snowball Earth” events (Fig. 1.1; Hoffman et al. 1998). While debate rages as to the exact extent of these glaciations – whether they were truly global, or whether there was a narrow band of ice-free ocean around the tropics (Allen & Etienne 2008) – glacial events of such magnitude have not since been repeated, and are not known from the 2 billion years prior to the Cryogenian. These “Snowball” events were associated with pronounced negative carbon isotope excursions, reflecting major perturbations to the global carbon cycle.

A third glacial event, the ~580 Ma Gaskiers glaciation (Fig. 1.1), was originally included in discussion of the “Snowball” events, and is similarly associated with a large negative carbon isotope excursion. However, this glaciation was restricted to the southern hemisphere, and spanned only 1 Ma – it was, therefore, considerably shorter than the earlier “Snowball” events, and likely represents just one of a series of regional and diachronous glaciations (e.g. Hoffman & Li 2009, Hebert et al. 2010). In Newfoundland, the change from glaciomarine- to turbidite-dominated sedimentology also coincides with redox geochemical signals taken to indicate a shift from ferruginous (anoxic) to oxic oceans (Canfield et al. 2008). It also coincides with a pronounced negative isotope excursion, correlated to similar excursions from sites across the globe (Halverson et al. 2005). This excursion has been taken to reflect oxidation of a large pool of dissolved organic carbon, itself reflecting oxygenation of the deep ocean (Halverson et al. 2005, Och & Shields-Zhou 2012). However, others have noted that the oxygenation of the deep oceans was diachronous (Johnston et al. 2012, Johnston et al. 2013), and therefore that the signal recorded at Newfoundland is likely only local.

The Gaskiers event in Newfoundland is just 3 Ma older than the appearance of the first macrofossils in that locality (Van Kranendonk et al. 2008). This temporal coincidence led to the glaciation and the associated geochemical transition to oxygenated water being held directly responsible for the evolution of large and complex multicellular life (e.g. Narbonne & Gehling 2003, Canfield et al. 2007). However, convincing arguments have been made for case that, rather than the rise in oxygen leading to the evolution of animals, the evolution of filter-feeding and burrowing animals capable of engineering their ecosystems on an unprecedented scale was responsible for oxygenation of the oceans (Lenton et al. 2014).

1.2 Life in the Ediacaran

Marine life diversified in the Neoproterozoic, with a diverse range of micro- and macroscopic groups appearing throughout the Era (Fig. 1.1). Prokaryotic and eukaryotic microbial life flourished through this time (e.g. Butterfield 2009b), and acritarchs developed increasingly complex morphologies (Grey et

3 A

Avalonia

B Laurentia Baltica

Avalonia

SE. Newfoundland Charnwood Forest, UK

Figure 1.2 Avalonia. A) Palaeogeographic reconstruction; map © Ron Blakey, Colorado Plateau Geosystems; approximate location of Avalonia after Scotese (2009) and Li et al., (2008); B) Current extent of the remnants of Avalonia, after Cocks et al. (1997) and Liu et al. (2015). al. 2011). The simple discoidal fossils for which the Ediacaran is remarkable started to appear from the lower to middle Ediacaran, a time that also records the predominantly macro-algal biotas of the Lantian of China (Yuan et al. 2011). It is not until the late Ediacaran that the complex soft-bodied biota for which the Period is perhaps best known make their first appearance.

The overwhelming majority of late Ediacaran organisms defy convincing phylogenetic placement. The most widely-studied and well-known macrofossil assemblages of the late Ediacaran are those of Newfoundland, Australia, Namibia, Russia, and the UK, and their historic treatment as one biota has in part hindered interpretation of their phylogenetic affinity (MacGabhann 2014). Fossils from the classic Ediacaran localities have been grouped into an extinct Kingdom, the Vendobionta (Seilacher 1984), and have been interpreted as fungal-grade organisms of uncertain phylogeny (Peterson et al. 2003). They have elsewhere been allied to various modern phyla, to everything from basal- to crown-group animals (Glaessner 1979, Clapham et al. 2003, Narbonne 2005, Sperling et al. 2011), giant protists (Seilacher et al. 2003), and even lichens (Retallack 1994). However, most workers now recognise several clades to be present across the various late Ediacaran assemblages, treating forms on a case-by-case basis.

4 The terminology used here is that of Erwin et al. (2011), but is approximately equivalent to that of Grazdhankin (2014) – although there are slight differences between the two schemes, they are closely comparable and differ in assignment of only a few groups.

Late Ediacaran macrofossils have been divided into three palaeogeographic regions, the Avalon (ca. 579—560 Ma, Noble et al. 2014, Van Kranendonk et al. 2008), White Sea (ca. 555 Ma, Martin et al. 2000) and Nama Assemblages (c. 545 Ma, Laflamme et al. 2013, Waggoner 2003). However, others have questioned whether this is a palaeogeographic, temporal or environmental signal (e.g. Gehling & Droser 2013, Grazhdankin 2014). Of all the currently recognised groups of complex Ediacaran macrofossils, only a handful are known from Avalonia, and only two occur in abundance in the deep-water successions of these regions that form the focus of this thesis: rangeomorphs, and arboreomorphs.

1.2.1 Rangeomorphs

The Rangeomorpha (Erwin et al. 2011, Grazhdankin 2014, Pflug 1972) are among the most recognisable of Ediacaran fossils (Fig. 1.3a—d, f). Rangeomorphs dominated the deep-water Ediacaran assemblages of Avalonia (Chapter 2), and are the most morphologically diverse group in those regions; they form the focus of this thesis. Twelve rangeomorph genera are known to date, of which eleven genera and a total of fourteen species are known from Avalonia (Liu et al. 2015a). Two new genera (Orthiokaterna fordi gen. et sp. nov., Chapter 5; and Undosyrus nemoralis gen. et sp. nov., Chapter 6), as well as three new species of a known genus (Primocandelabrum anatonos sp. nov., P. boyntoni sp. nov. and P. katatonos sp. nov., Chapter 4) are described in this thesis, all from Charnwood Forest. Rangeomorphs are frondose, and are characterized by pseudo-fractal, “modular” body plans, and gliding symmetry (Narbonne 2004, Brasier et al. 2012). This bauplan provides a relatively simple way of generating complex structures (Hoyal Cuthill & Conway Morris 2014), but is fundamentally different in its construction to that of any extant organism, making interpretation of their biology challenging.

The basic building blocks of rangeomorphs, their modules, are the rangeomorph branches or elements that give the group its name (Table 1.1). Each rangeomorph branch or element is itself comprised of multiple branches, giving rise to the “self-similar” branching architecture (Fig. 1.3, 1.4). These branches are organised into orders, with higher orders representing progressively finer subdivisions; up to five orders of subdivision are known from the best-preserved specimens (Liu et al. 2015b). The morphological appearance of the various orders of rangeomorph branching (Fig. 1.4) are used as a basis for their classification (Narbonne et al. 2009, Brasier et al. 2012). However, there are myriad inconsistencies in terms used between and even within working groups to describe the same structure, and this is addressed in Table 1.1 – terminology used throughout is modified from Brasier et al. (2012), except where otherwise specified.

5 Figure 1.3. Rangeomorphs (a—d, f), arboreomorphs (e) and Aspidella (g). A) Orthiokaterna, GSM105875, photo after Wilby et al. 2011; B) Charnia masoni and Bradgatia linfordensis, GSM105873, photo after Wilby et al. 2011; C) Beothukis plumosa (formerly Culmofrons) from the Bonavista Peninsula, Newfoundland (photo courtesy of A. Liu); D) Fractofusus from the Avalon Peninsula; E) an arboreomorph, Charniodiscus sp., GSM106069; F) Primocandelabrum, GSM105969b; G) several Aspidella from Ferryland Head on the Avalon Peninsula, Newfoundland, scale bar = 1 cm. The specimens in A, B, E and F are from Bed B of Charnwood Forest (photographs taken from plaster casts housed at the British Geological Survey), and all scale bars = 2 cm, unless otherwise specified.

6 The frond can be unipolar (Fig. 1.3c), bipolar (Fig. 1.3d) or multipolar (Fig. 1.3a, f), depending on the number of growth poles it possesses. In most rangeomorphs (e.g. Charnia, Fig. 1.3b), the frond consists of a single (unipolar) leaf-like structure (which is divided into first order branches), but some rangeomorphs possess multiple of these leaf-like structures (e.g. Bradgatia, Fig. 1.3b) and are thus multipolar. These main branches or leaf-like structures are here termed “folia” (pl.; “folium”, sing.), reconciling the unipolar and multipolar terms with “unifoliate” and “multifoliate” (Laflamme and Narbonne, 2008), respectively. In Chapter 5, I investigate whether each folium in a multifoliate rangeomorph (e.g. Orthiokaterna) is analogous to a primary branch (Brasier et al. 2012) in a unifoliate one, or to the entire unifoliate frond. If analogous to the frond, then multifoliate rangeomorphs record the repetition of major structural units. This is a fundamental aspect of a rangeomorph’s growth programme, and may reflect phylogenetic relationships within the group.

In addition to the frond, many rangeomorphs also possess a basal holdfast and a stem (Fig. 1.3a, c, f). Most taxa possess a basal holdfast, which varies from circular (discoidal) to tear-drop shaped (bulb-like) in preserved outline. The stem connects the frond to the disc, and its length is highly variable between species and genera. Possession of a stem is likely of significant ecological import (Chapter 7, 8), and its variation within a taxon may represent ecophenotypism (Chapter 4).

1st order: 1st order: 2nd order: 3rd order: Rotated, unfurled, Displayed, unfurled, Displayed, furled, Displayed, distal inflation, distal inflation, proximal inflation, furled, no inflation, radiating, radiating, radiating, subparallel, unconcealed unconcealed concealed concealed Figure 1.4 Rangeomorph branching architecture. Displayed/rotated: both rows/ one row of lower-order branches are visible. Furled/unfurled: branch edges are visible/tucked in, giving a scalloped/smooth outer margin. Inflation: shape of the branch increases distally/ proximally. Radiating/subparallel: branches emerge from central axis at increasing/ similar angles. Concealed/unconcealed: central axis is visible or concealed (dashed lines) by the branches arising from it. Full description of terms and alternate terminology presented in Table 1.1. 7 Term Definition Alternative terms Terms refering to the rangeomorph body Basal disc basal structure, interpreted as a holdfast; disc; holdfast disc holdfast, anchoring the organism on/in the sediment, often preserved with a circular outline Stem connects the holdfast to the stalk frond Frond that part of a rangeomorph which “petalodium” or “foliate leaf structure”2; “petalage”3; the comprises the rangeomorph term has been used in the literature to refer to either elements the frondose part of unifoliate rangeomorphs or to whole organisms (reviewed by 1, 2). Crown the “frond” or “petalage”3 of a This term is favoured over “frond” for multifoliate multifoliate rangeomorph rangeomorphs for clarity. While use of the term “frond” for the frondose part of unifoliate rangeomorphs is intuitive, given the shape connotations associated with the term some confusion may arise as to whether the term refers to the whole organism, the whole frondose part of the organism, or just a single folium. It is also unclear at this point in time whether the folium is homologous to the frond or primary branches of a unifoliate frond Rangeomorph modular unit, with each branch rangeomorph element4 branch acting as an axis for further branching; self-similar branching structures4 Folium the leaf-like structure or structures petals3 which comprise the frond Uni-/bi-/ Frond comprised of one/two/ likely corresponds to Uni-/bi-/multipolar1, i.e. possessing multifoliate multiple folia. one/two/multiple poles or directions of growth Terms relating to branch identification First order The largest visible order of Equivalent to primary branches in unifoliate branches branching; the branches rangeomorphs, and to folia in multifoliate emanating off the stem rangeomorphs. Differentiating between first order and primary branches permits use of first order branches when the relationship of the largest branches to the central axis (i.e. whether they are true primary branches or folia) cannot be determined. Second, third, Branches which emanate off a fourth (and lower order branch so on) order branches Primary Rangeomorph elements which branches branch of the central axis of a folium Secondary, Rangeomorph elements which tertiary, branch off the central axis off a quaternary (and lower branch rank (e.g. primary so on) branches off a secondary) Host branch The branch off which the rangeomorph element in question emanates Row One side of a rangeomorph petaloid2 element Eccentric Aberrant growth of branches to a branches larger size than their neighbouring branches of the same order, observed to display the branching pattern of the host branch rather than the neighbouring branches Subsidiary/ Branches arising from additional supplementary growth loci branches

8 Term Definition Alternative terms Terms refering to branching architecture Branching the presumably diagnostic characters relating Branching pattern architecture to several features of the branches, and their relationship to each other Rotated vs. displayed Rotated Branch has one row visible, but are presumed “Charnia-type” (single- to have two rows, with one is rotated out of the sided), although plane of view the interpretation is different: possession of one row5, as opposed to rotation of one row out of the plane of preservation1 Displayed Branch has two rows visible “Rangea-type” (double- sided) branches5 Furled vs. unfurled Refers to the distal tips of the high-order rangeomorph branches borne by the host branch. Furled Branch outline is smooth and relatively straight, as the tips of the branches are presumed to be tucked in Unfurled Branch outlines have an irregular or scalloped appearance. Overlapping of the branch margins may give unfurled branches a furled appearance locally. Subparallel vs. radiating Refers to the angle at which the branches emanate off the axis on which they are borne. Subparallel Branches emanate at the same angle all the way up their axis. Radiating Branches emanate at increasingly high angles, giving a fan-like appearance Constrained vs. unconstrained Constrained Branches conform tightly to their branching Fixed1 architecture Unconstrained Branches are more free to rotate about their Free1 emanation point, giving a looser or more untidy appearance to the branch; branching pattern is more susceptible to taphonomic disruption. Inflation Refers to the overall shape of the branch, contributed to by growth of the branch without addition of further higher-order branches Proximal The base of the rangeomorph element (i.e. ovate/lanceolate 2 closest to the axis of the host branch) is widest, giving a rounded tear-drop shape pointing away from the branch axis Median The central part of the rangeomorph element oval/elliptic2 is widest Distal The distal end of the rangeomorph element obovate2 (furthest away from the axis of the host branch) is widest, bar the youngest elements at the tip, giving a tear-drop shape pointing towards the branch axis Moderate Branches do not obviously inflate at any oblong/parallel 2 point along the growth axis, and so have a rectangular outline

Table 1.1. Terminology for rangeomorph branching architecture, terminology modified after 1) Brasier et al. 2012; 2) Laflamme and Narbonne 2008; 3) Fulde and Narbonne 2008; 4) Narbonne 2004; 5) Narbonne et al. 2009.

9 1.2.2 Arboreomorphs

Arboreomorphs are similar in gross appearance to unifoliate rangeomorphs (Fig. 1.3b, c), but are characterised by non-fractal branching patterns – a fundamentally different bauplan. The fronds of arboreomorphs are also symmetrical about their midline, in contrast to the gliding symmetry of rangeomorphs. The group is represented solely by the genus Charniodiscus (Fig. 1.3e). Arboreomorphs are comprised of a stalk, holdfast and frond, but known forms only have one leaf-like structure, and only two orders of branching are recorded (Laflamme et al. 2004). Their similarity of overall form to unifoliate rangeomorphs likely reflects convergence due to similar environmental constraints. Arboreomorphs are included in the frondomorphs of Grazhdankin (2014), but this group also includes a number of discoidal taxa, including Aspidella terranovica Billings (1872).

1.2.3 Discoidal forms

The Ediacaran is remarkable for its abundant and morphologically diverse discoidal fossils. While some, such as Cyclomedusa, are convincingly interpreted as microbial colonies (Grazhdankin & Gerdes 2007), many continue to defy interpretation of their ecology, mode of life and phylogenetic placement. These forms are included in the frondomorph group of Grazhdankin (2014). The genus Aspidella (Fig. 1.3g) was originally represented by a plethora of morphologically similar but distinctly named taxa, but has recently come to refer to the spectrum of discoidal fossils that are interpreted as the holdfasts of fronds – it is thus considered to be an organ taxon (Gehling et al. 2000a). Despite recent efforts (Burzynski & Narbonne 2015), the majority of Aspidella holdfast fossils cannot confidently be assigned to individual rangeomorph or arboreomorph taxa, and they conceivably include representatives from both groups. The genus as currently defined likely also includes non-holdfast forms, and/or the holdfasts of taxa neither rangeomorph nor arboreomorph (reviewed in Kenchington & Wilby 2014). Indeed, cnidaria- like behaviour (namely maintaining a constant level in the sediment through “shuffling” upwards) has recently been identified in certain small forms of Aspidella from the shallow-water successions of Newfoundland (Menon et al. 2013a), but not for the larger, concentric-ringed Aspidella morphs that are morphologically indistinguishable from (and indeed in some cases are still attached to) the holdfasts of arboreomorphs/rangeomorphs (Laflamme et al. 2011). Other small discoidal fossils from the Ediacaran of Shropshire (UK) have been re-interpreted as fluid escape structures (Menon et al. 2015), highlighting the importance of interpreting these morphologically simple fossils on a case-by-case basis.

1.2.4 Other taxa

There are several fossil groups in Avalonia that do not fit within any of the currently recognised taxonomic groupings (Erwin et al. 2011, Liu et al. 2015a). For example, there is the ladder-like body fossil Hadryniscala, the brush-like Parviscopa, and the string-like, holdfast-bearing Hadrynichorde, all described by Hofmann et al. (2008) from the Bonavista Peninsula of Newfoundland. The triangular

10 fossil, Thectardis, has provoked a great deal of discussion in the literature. Reconstructed as conical in life, it has been argued to represent a poriferan (Clapham et al. 2004, Sperling et al. 2011). Although the lack of any diagnostic features of crown-group sponges in Thectardis (Antcliffe et al. 2014) rule out assignment to Porifera, Thectardis could conceivably represent a stem-group lineage. The putative cnidarian Haootia quadriformis Liu et al. (2014), and potentially metazoan trace fossils (Liu et al. 2010), have been taken to record the first appearance of metazoans in the fossil record, and promise important insight into the ecology of these ancient communities (Liu et al. 2015a).

A large number of forms originally described as discrete taxa (e.g. the “lobate discs” of Mistaken Point, and Ivesia, Blackbrookia, and Shepshedia from Charnwood Forest) have recently been interpreted as the decayed remnants of rangeomorphs and arboreomorphs (i.e. taphomorphs of these taxa), and are together termed “iveshediomorphs” (Liu et al. 2011). Some forms currently incorporated into iveshediomorphs, such as the Blackbrookia fossils identified on the Bonavista Peninsula (Hofmann et al. 2008), have a discrete morphology and clear outlines that are quite different from typical iveshediomorphs, while other iveshediomorph forms may simply represent microbial colonies (Laflamme et al. 2012a).

The taphomorph interpretation has recently been extended to both Charniodiscus and Thectardis (Antcliffe et al. 2015), although with very little justification (Chapter 3). Discrimination of living from dead individuals on the surface has significant repercussions for interpretation of ecological interactions, standing biomass and even nutrient sources and modes of life (Liu et al. 2015a). The place of these two taxa in the community structure is discussed in Chapter 7, and their population structures are discussed in Chapter 9.

1.3 Rangeomorph growth and functional morphology

Many ecological studies rely upon consistently applied, accurate taxonomic classification. Especially for rangeomorphs, the taxonomic weight given to characters describing aspects of branching architecture and those referring to gross morphology is subjective, with different working groups placing greater emphasis on different characters. Chapter 4 focuses on the stalked, multifoliate rangeomorphs (Primocandelabrum, Orthiokaterna) for which the Charnwood Forest assemblage is remarkable. While many similar forms are known from Newfoundland, their numeric abundance and detailed preservation in Charnwood Forest allow detailed study of their morphology at a level which has not been possible on other sites. Multivariate statistical analyses are used to test the degree of co-variation between branching and gross morphological characters, to determine how successfully the various characters discriminate between taxa, and to quantify the extent of intra- vs. inter-specific variation.

11 Modular organisms are capable of great variation and adaptation in their morphology, the extent of which is in part dependent on the level of interconnectedness between modules (Chapman 1981, Dewel et al. 2001, Marfenin 1997).

The detailed architecture of rangeomorph branches (their modules) throughout ontogeny can reveal important insights into the complexity and plasticity of the rangeomorph growth programme. In Chapter 5, I focus on one stalked, multifoliate taxon from Charnwood Forest, Orthiokaterna fordi gen. et sp. nov.. This taxon has a distinct architecture at each level of branching, allowing unique aspects of its growth to be recognised. Given the lack of phylogenetic relatives, the next best avenue of research into rangeomorph biology is through identification of functional analogues, and the growth of Orthiokaterna is discussed in this context. A similar approach is taken in Chapter 6, where I describe Undosyrus nemoralis gen. et sp. nov. from Charnwood Forest. This form is associated with an intriguing lineated structure that has not been documented from any other Ediacaran taxon. Through study of the morphology of this structure and of the rangeomorph with which it is associated, I make certain inferences about the likely function of this structure and how it was generated, and subsequently discuss the plasticity of rangeomorph growth and the morphological adaptability of its modules.

1.4 The Ecology of Avalonian Rangeomorphs

The ecology of an organism – how it interacts with its community and its physical environment – is a fundamental component of its biology. The adaptability and environmental sensitivity of an organism reflects the complexity of its feedback systems. While some aspects of ecology can be ascertained through analysis of functional morphology (Chapter 6), others can only be identified through analysis of the structure and composition of the community as a whole. The rangeomorph-dominated Ediacaran communities of Avalonia represent the oldest known ecosystems comprised of diverse and complex macrofossils. The level of interaction between the organisms within a community, and between the organisms and their environment, provides a context within which to investigate the development of the more dynamic, modern-style ecosystem structures that originated in the Cambrian (Fig. 1.1).

The past few years have seen significant progress in understanding the ecology of rangeomorphs. Broad- scale environmental preferences of rangeomorphs and arboreomorphs (among other groups) have been indentified, highlighting the importance of taking the geological context into account when interpreting these deeply problematic organisms (Grazhdankin 2004, Gehling & Droser 2013). Through analysis of spatial patterns and size distributions in rangeomorph populations, both stolon-like, clonal reproduction and reproduction via (sexual?) waterborne propagules have been recognised in rangeomorphs (Darroch et al. 2013, Mitchell et al. 2015). Although there is still much debate regarding the mode of life of rangeomorphs, many have come to accept an osmotrophic feeding strategy for the group (absorbing

12 carbon directly out of the water column; Laflamme et al. 2009). While this explains the lack of preserved pores or zooids in specimens with 0.1 mm preservational resolution, there are serious doubts as to whether there would have been sufficient amounts of dissolved organic carbon in the ocean, and whether it would have been in a utilisable form (Liu et al. 2015a). Alternative modes of life include filter feeding via pinocytosis, or chemosymbiosis, or a combination of multiple modes of feeding (Chapter 5, 7).

In modern ecosystems, organisms are adapted to different niches to avoid inter-specific competition, and one aspect of this – tiering, or vertical partitioning of the ecosystem – has been argued to be present in Avalonian communities (Clapham & Narbonne 2002). Identification of this ecological structure has been used to support filter feeding and, alternatively, osmotrophic modes of life for the organisms in these communities, and even to infer a metazoan affinity for them (Clapham et al. 2003). The structure of several Avalonian communities is investigated in Chapter 6 to determine the presence and extent of tiering. The functional morphology and ecological importance of stems and of particular branching architectures is discussed in this context.

Although predation has played a critical role in structuring Phanerozoic ecosystems, it was minimal or absent at least until the terminal Ediacaran (Hua et al. 2003). In contrast, many of the physical processes that operated in Phanerozoic ecosystems would similarly have operated in Ediacaran ones (Wilby et al. 2015). Modern communities are affected by ambient disturbance and by discrete disturbance events, and the intensity of disturbance has a significant influence on their structure and composition (Dale et al. 2005). Organisms adopt one of two strategies: adaptation to competition, or tolerance to disturbance (White & Pickett 1985). A classic idea in ecology is the “intermediate disturbance hypothesis”, that is, the highest taxonomic diversity occurs at intermediate levels of disturbance as representatives of both the competitive and tolerant strategies coexist (White & Pickett 1985). In Chapter 8, I investigate the influence of physical disturbance as recorded by sediment influx on the taxonomic composition of rangeomorph/arboreomorph communities, identify which taxa are better adapted to low or high levels of ambient disturbance, and investigate the correlation between ambient disturbance and taxonomic diversity in these ancient communities. Discrete disturbance events influence not just the structure of the community, but also the structure of a population. In Chapter 9, I analyse the population demographics within several Avalonian communities, including those of Charnwood Forest and the classic Mistaken Point localities of Newfoundland (Chapter 2), and identify signals of reproductive events and the influence of disturbance events in these communities. By combining the results of these ecological studies, I infer the life history traits and ecological strategies of several rangeomorph taxa (Chapter 10), and consequently propose an environmentally sensitive model of succession in these enigmatic Ediacaran communities.

13 2. GEOLOGICAL CONTEXT FOR THE EDIACARAN OF AVALONIA

During the Ediacaran, a volcanic island arc developed off the northern margin of Gondwana, at around 60 oS (Li et al. 2008, 2013). This island arc formed the microcontinent of Avalonia, and complex macroscopic life thrived in the basins that developed in association with it. In the early Cambrian, Avalonia rifted from Gondwana, and drifted northwards until it collided with the northern landmasses of Baltica and Laurentia during the earliest phases of the Ordovocian–Silurian Caledonian orogenic cycle (Lambert & McKerrow 1976). The remnants of Avalonia today include the southern United Kingdom and south-east Newfoundland, and it is the fossils of these regions that form the focus of this thesis.

2.1 Ediacaran Stratigraphy of Avalonia

The Avalonian successions of the UK (Fig. 2.1, 2.2) and Newfoundland (Fig. 2.3) are best known for the diverse, demonstrably deep-water Ediacaran communities (e.g. Narbonne 2005). The biota is known from two regions in Newfoundland, namely the area around Mistaken Point on the Avalon Peninsula (2.3b), and the area around Catalina on the Bonavista Peninsula (2.3d); and from Charnwood Forest in the UK. The successions in all three regions are kilometres thick (Fig. 2.1b, 2.3c), but shallow-water indicators are absent (Charnwood Forest), or restricted to the uppermost parts of the succession and in very different facies to those that host the classic biotas (Newfoundland; Ichaso et al. 2007); an absence that provides the basis for interpreting a deep-water setting for these biotas. Approximately coeval, shallow-water deposits are known from elsewhere in both the UK (Shropshire and Carmarthenshire in the U.K. (e.g. Salter 1856, Cope 1977, McIlroy et al. 2005) and from the Ferryland area of Newfoundland (Gehling et al. 2000a; Fig. 2.3a), but the biotas in these areas are largely restricted to simple discs and microbial impressions. Confident correlation of these two shallow-water areas in terms of their age and depositional setting is incomplete. Outcrop quality is variable throughout the exposure, with differential ash cover, weathering and cleavage. As such, not all surfaces represent true expressions of the original substrate texture and, as a result, detailed sedimentological interpretations are necessarily couched with a reasonable degree of uncertainty.

2.1.1 The Geological Setting of Charnwood Forest

The Ediacaran succession of Charnwood Forest is encompassed by the c. 3800m thick Charnian Supergroup (Carney 1999, Moseley & Ford 1985). Two lithostratigraphic groups are recognized: the Ediacaran Blackbrook and Maplewell groups (Fig. 2.1a). The Cambrian Brand Group (Fig. 2.1a) was formerly included in the Charnian Supergroup (Moseley & Ford 1985), but has since been removed (Noble et al. 2014). Contacts between the groups are defined by prominent conglomerate or breccia units (Fig. 2.1b). The Blackbrook and Maplewell Groups comprise a sequence of largely fine-grained volcaniclastic sedimentary rocks, with minor pyroclastic volcanic breccias (Fig. 2.1b; Carney 1999), while the Brand

14 Group comprises a basal quartzite and then a thick mudstone succession. The predominantly high-silica andesites and porphyritic dacites of the Whitwick and Bardon Volcanic Complexes comprise the major magmatic centres in the area (Carney 1999, 2000), and likely represent the remnants of the volcanic sources from which the bulk of sediments were derived (Fig. 2.1a). The Charnwood Forest succession was deposited in a fore- or back-arc basin (Le Bas 1984, Pharaoh et al. 1987), in a setting comparable to the offshore of modern-day Montserrat (Carney 1999). The lack of any shallow-water indicators such as bi-directional cross-bedding or wave ripples, in even the deposits in comparatively close proximity to the volcanic centres, could suggest deposition well below storm wave base for the rest of the succession, including the fossiliferous intervals.

The lowest stratigraphic unit within the Charnian Supergroup is the Ives Head Formation of the Blackbrook Group, which is comprised largely of medium-grained, turbiditic tuffaceous sandstones (Fig. 2.1) with minor conglomerates and volcanic breccias (Carney 1999, 2000). Its top is defined as the South Quarry Breccia Member, a coarse-grained volcaniclastic unit (Moseley & Ford 1985). The overlying Blackbrook Reservoir Formation is comprised of tuffaceous pelites, coarse pelites and minor conglomerates (Moseley & Ford 1989, Carney 1999).

The succeeding Maplewell Group contains two formations (Moseley & Ford 1985; Fig. 2.1). The lower of these, the Beacon Hill Formation, is fine-grained and highly siliceous, probably reflecting a higher volcanic ash component within these sediments (Carney 1999). The Charnwood Lodge Volcanic Formation (Fig. 2.1a) is the proximal equivalent to the Beacon Hill Formation (Carney 2000). The Benscliffe Breccia Member, a thin volcanic breccia unit, defines the base of the Beacon Hill Formation (Fig. 2.1b; Moseley & Ford 1985). The Bradgate Formation is the youngest within the Charnian Supergroup. The base of the formation is defined by the Sliding Stones Breccia (Fig. 2.1b), a thick slump breccia with abundant contorted mudstone clasts, that reflects syn-sedimentary remobilisation of Beacon Hill sediments and acts as a marker unit across the region (Moseley & Ford 1989). This formation consists of predominantly fine-grained volcaniclastic material, with intermittent fine- to medium-grained tabular sandstone horizons (Fig. 2.2). Parallel and cross-lamination is sometimes present, and soft sediment deformation is common, most notably in the slump horizons, but also includes loaded beds, flame- structures, and internal homogenisation of beds. The Bradgate Formation is capped by the coarse pebble conglomerates of the Hanging Rocks Formation (Fig. 2.1b), which have been interpreted as channel-fill deposits (Moseley & Ford 1989). Stratigraphically above the Bradgate Formation is the Brand Group (Fig. 2.1b). Quartz arenites and greywackes of the Stable Pit Member of the Brand Hills Formation are overlain by the predominantly purple pelites and slates of the Swithland Formation.

15 Swithland A Formation B Brand Hills BRAND GROUP Formation HRF Bed B 556.6 ±6.4 Ma Bradgate Formation

SB 561.85 ±0.34 Ma

Beacon Hill

MAPLEWELL GROUP Formation

565.22 ±0.33 Ma

BB 569.08 ±0.45 Ma Blackbrook Reservoir Formation

SQ Ives Head 0 Formation Vertical BLACKBROOK GROUP scale of Bardon Hill Volcanic Complex column 300 m Fine Whitwick Volcanic Complex Medium Coarse, breccia, conglomerate SCD South Charnwood Diorites High pyroclastic Parallel bedding or lamination content CV Charnwood Lodge Volcanic Complex Convoluted lamination Volcanic breccia Load structures SP Stable Pit Quartzite Member Graded bedding ('slump') breccia Fossil bed Conglomerate Unconformity

Figure 2.1. A) Geological map and B) stratigraphic column of the Ediacaran—Cambrian succession of Charnwood Forest, modified after Carney (1999). Dates from Noble et al. (2014). HRF = Hanging Rocks Formation; SB =Sliding Stones Breccia Member; BB = Benscliffe Breccia Member; SQ = South Quarry Breccia Member.

2.1.1.1 Age constraints for Charnwood Forest The most recent and precise dating of the Ediacaran succession of Charnwood Forest was undertaken by Noble et al. (2014). Dating of this succession is complicated by the low zirconium content of the volcanic centres, resulting in a paucity of volcanogenic zircons that are demonstrably coeval to deposition. Zircons were dated from a volcaniclastic sandstone from the base of the Ives Head Formation, from a sandstone 1.5 m below the bedding plane that hosts the Lubcloud assemblage (Wilby et al. 2011), and from the South Quarry Breccia. Of indistinguishable morphology and giving ages within error of each other, the zircons were interpreted as inherited and provide a maximum age constraint of 611±1.1 Ma for the Ives Head Formation. The base of the Beacon Hill Formation is dated to 569±0.9 Ma, and its lower part to 565±0.9 Ma. The facetted morphology of the zircons combined with the presence of melt inclusions led to the interpretation of these dates as the ages of eruption/deposition. However, only an inherited population of zircons dated at ~613 Ma could be gained for the upper part of the formation. The base of the Bradgate Formation (the Park Breccia or Sliding Stones Slump Breccia), is dated to 561.9±0.9 Ma. The concordant dates of the younger zircons, their facetted morphology and melt inclusions suggest that this date is eruptive/depositional. The Hanging Rocks Formation was dated to 556±6.4 Ma. This date

16 was obtained from the four youngest (<10% discordant) grains, and their sharp facets suggested that the date was close to the age of deposition. As such, the date of this formation provides a minimum age constraint on the Bradgate Formation. The Swithland Formation contains specimens of the trace fossils Teichichnus and Arenicolites, confirming a Lower Cambrian age for the Brand Group (Bland 1994, Bland & Goldring 1995, McIlroy et al. 1998).

2.1.1.2 Fossil occurrences in Charnwood Forest Despite early (see Howe et al. 2012) and periodic interest by a few individuals (Boynton & Ford 1979, Boynton & Ford 1995, Brasier & Antcliffe 2009), the Charnwood Forest successions have long been considered the poor cousin of the Newfoundland ones. Its biggest claim to fame is that it was the first locality where these ancient fossils could be shown to be demonstrably Precambrian (Ford 1958), based on their position in the well-established regional stratigraphy (Lapworth 1888, Watts 1903). This was rectified when, in 2008, a programme of cleaning and silicon rubber moulding undertaken by the BGS revealed a biota on bedding surfaces in the Bradgate Formation (Fig. 2.1; Appendix A.2.1) whose diversity and preservational quality (Wilby et al. 2011) rivals that of all but the best surfaces in Newfoundland (Liu et al. 2015b).

The Ives Head Formation contains the holotypes of several taxa that were originally described by Boynton and Ford (1995, 1996) as discrete biological taxa. However, these fossils are now generally referred to as ivesheadiomorphs (sensu Liu et al. 2011), taphomorphs of uncertain affinity, but likely including rangeomorphs and arboreomorphs (Chapter 1, 3). Further refinement of the date of these fossils is crucial. A pre-600 Ma depositional age would make the Lubcloud fossils the oldest complex Ediacaran assemblages, on a par with the algal biotas of the Lantian. More importantly, a pre-582 Ma age would place them older than the Gaskiers glaciation, casting doubt on the importance of this event for evolution of the classic Ediacaran macrofossils. The overlying Blackbrook Formation is not currently known to be fossiliferous, but exposure of this formation is poor. Currently, only one discoidal fossil has been reported from the Beacon Hill Formation (Noble et al. 2014, Wilby et al. 2011). The Bradgate Formation contains the most iconic and the best preserved fossils within Charnwood Forest, including the holotype of Charnia masoni Ford (1958) and Charniodiscus concentricus Ford (1958). Of the two principal fossil-bearing localities in the Bradgate Formation, most of the fossils discussed in this thesis are from a single bedding plane, Bed B of Wilby et al. (2011), from the North Quarry locality (Fig. 2.1). The date of the Hanging Rocks Formation provides a minimum age constraint on the fossiliferous surfaces within the Bradgate Formation, and places them at most c. 3 Ma younger than the Mistaken Point surfaces (Noble et al. 2014).

17 m Average Grain Size m Average Grain Size m Average Grain Size CSiVFFMCVCGP CSiVFFMCVCGP CSiVFFMCVCGP 174 96 74

94 170 72

92 166 70

162 90 Bed B 68 150

88 66 148

86 52 146

84 50 144

98 48 142 CSiVF

96 FM 46 140

94 44 138

92 42 136

90 40 C 134

88 38 132

86 36 130

84 34 128

82 32 126

80 30 124

78 28 GC 122

76 CSi 26 120

Figure 2.2 Detailed log of the section of the Bradgate Formation around Bed B of North Quarry (Bed B) and GC; key to log overleaf.

18 2.1.1.3 Detailed lithological analysis of the Bradgate Formation The Bradgate Formation hosts the rangeomorph-dominated communities that the Ediacaran of Charnwood Forest is so well known for, and so its sedimentological context warrants a more detailed analysis. Five main lithologies can be identified.

The dominant lithology comprises greyish to blueish-green siltstone units of variable thickness (Fig. 2.2), which are variably interbedded with thin porcellaneous horizons and thin, tabular sandstones. Locally, the siltstone contains wispy laminae that weather pale in colour and which are less than a millimetre in thickness – they are most clearly identified in polished section. These laminae are of a finer grade than the siltstone, often pale green, yellow or white on polished surfaces. They could reflect the “wispy” lamination and mottled appearance documented from mud turbidites, but the lack of clear tractive structures precludes confident interpretation of the structures (cf. Rebesco et al., 2014).

In the field, the most striking beds in the succession are normally-graded event beds, which are a few centimetres to tens of metres thick, and are variably coarse gravel to fine sandstone grade at their base (Fig. 2.3a). The lower contacts of the beds are sharp, locally erosional and loaded, but lack evidence of scouring or fluting. The upper portion of the bed is typically siltstone grade, and laminated to very finely laminated. Lenses of coarser silt in these laminae can be identified in thin section, and are interpreted as deposition from mud turbidites (Chapter 7). Both epiclastic and volcaniclastic beds can be distinguished in the field and in thin section, following the conventions and definitions of Fisher & Schmincke (1984). Epiclastic gravity flows result from eroded volcanic rock, reworked and/or remobilised following an event such as an earthquake, but do not include remobilisation of unconsolidated pyroclastic fragments. Less than half of their thickness is graded, with more than half comprised of planar or cross laminated sediment. Clasts include lithic fragments, rounded—sub-angular quartz crystals and only a small component (less than a few percent volume) of feldspar crystals. Volcaniclastic gravity flows include remobilisation and redeposition of unlithified pyroclastic material, but also include pyroclastic flows that enter the water column immediately after eruption. Upon entering the water column, they typically separate into a gravity flow and a nepheloid plume of finer particles which settles out of the water column much more slowly. Clastic components include broken, euhedral or angular crystals, sub-rounded—

Key to lithologies and sedimentological features Lithologies Sedimentological features Contact type Siliceous, pale-weathering siltstone Lamina of sandstone (>50% crystals) Loaded base Siltstone interlaminated with very thinly laminated, Lamina of sanstone (<50% crystals) white/pink -weathering siliceous laminae (ash-fall tuffs?) Missing section, contact unseen Siltstone interlaminated with sandstone ribs (<3cm) Lamina of siliceous material (ash-fall tuff?) Erosional base Siltstone Soft-sediment deformation Sharp contact Contacts unseen; unit pieced together Sandstone with >50% euhedral/angular crystals Parallel lamination from adjacent outcrops Lithic sandstone (<50% crystals) Syn-sedimentary faulting Very thinly laminated, white/pink weathering siliceous Lithic granules beds (ash-fall tuffs?) Pyrite (euhedral)

19 angular quartz fragments, a relatively high proportion of feldspar crystals (more than a few percent volume), and a small proportion of lithic fragments which are typically of a similar composition to that of the bed as a whole (Fig. 2.4a). They typically have a high proportion of opaque minerals (Fig. 2.4a). For the pyroclastic flows, more than half of their thickness is graded, with less than half comprised of laminated sediment (Fig. 2.2).

One of the most visually distinctive lithologies in the Charnwood Forest successions are porcellaneous, pale-weathering beds. Typically no more than a few tens of millimetres thick, these beds are tabular with sharp bases and variably sharp or gradational upper contacts, pale white, green or pink in colour, and internally laminated. They are mudstone grade, weather proud of the surrounding sediment, and have flinty breakage. The porcellaneous beds are thought to represent periods of primary volcanic activity, evidenced by their relative abundance of pyroclastic material such as glass shards (Moseley and Ford 1989; Carney 1999), although these are rarely preserved. The features of these beds are consistent with interpretation as fine-grained ash-fall tuffs (cf. Norin 1958, Fisher & Schmincke 1984). Convolute lamination is common in these beds, and may be caused by seismic-induced dewatering, down-slope slumping, or, for ash-flow tuffs, from shear from the base surge above the bed (cf. Fisher & Schmincke 1984). The ash-fall tuffs likely include deposition from the nepheloid plume part of pyroclastic eruptions, but possibly also from discrete ash-generating events and subsequent settling out from the ashcloud.

While thick prominent slump beds define some of the formation contacts (Fig. 2.1), smaller-scale slump deposits are present throughout the succession, and vary from less than a centimetre to a couple of metres in thickness. All slump beds in the succession are typically matrix-supported, comprising a matrix of mud- to very fine sandstone grade around larger clasts that range from very fine sandstone grade to a few tens of centimetres in maximum diametre. The clasts themselves include quartz grains and lithic fragments, with the most common of these bring deformed, laminated mudstone. The edges of these clasts are often splayed, and some are fully enrolled, indicating that they were still soft upon transport, although partial lithification is indicated by the preservation of depositional lamination.

2.1.2 The Geological Setting of the Ediacaran of Newfoundland

The Ediacaran strata of Newfoundland comprises three lithostratigraphic groups: the Conception, St. John’s and Signal Hill groups (Fig. 2.3c), which record a progressively shallowing-up sequence (Williams & King 1979, O’Brien & King 2005, Hofmann et al. 2008). The Conception Group comprises thick- to medium-bedded siltstones and mudstones with frequent volcanic ash horizons, and are interpreted as abyssal plain/slope sediments deposited in a fore-arc basin (Wood et al. 2003, Ichaso et al. 2007). The overlying St. John’s Group comprises thin-bedded turbidites with a volumetrically minor volcaniclastic component, and is interpreted as basinal to delta-front deposits during the transition from fore-arc basin

20 -53.5 -53.2 -53.0 A B 2

SB St. John’s

BS FH 46.8 46.8 1

PC1 PC2 Gibbett Hill C CANT Fm. SH BC SP LMP Renews MPD, MPE, MPG Head Fm. 010km Cape Ballard Fm. 350m+ -53.5 -53.2 -53.0 Ferryland Head Fm. Fermeuse Fm. Gibbett -53.1 -53.0 Hill Fm. D CHF Renews

St. John’s GroupSt. John’s SHG Head Fm. 1km Fermeuse H14 Fm.

Trepassey Fm. GroupSt. John’s Signal Hill Group TPF MPF 565 48.5 ± 3 Ma Briscal Fm. 48.5

578.8 PU4 Mistaken Drook Fm. SPS Point Fm. ±0.5 Ma PU.V PU

Conception Group MC.1 PU.2 Gaskiers Fm. PU.1 Mall Bay Fm. STB base not exposed Conglomerate BC Sandstone Conception Group Drook Fm. Siltstone HS Shale 00.511.52km base not exposed Conglomerate Sandstone CHF = Cappahayden Fm. Siltstone TPF = Trepassey Fm. Shale MPF = Mistaken Point Fm. -53.1 -53.0

Figure 2.3. Geological maps and lithostratigraphy of Newfoundland. A) Outline of southern Newfoundland, Canada, with sites from the northern and east coasts of the Avalon Peninsula: BS = Brigus South; FH = Ferryland Head; SB = Spaniard’s Bay; B) Fossil localities of the southern Avalon Peninsula, after blue inset (1) of a). BC = Bristy Cove; CANT = C. antecedens surface; LMP = Lower Mistaken Point; MPD = Mistaken Point D; MPE = Mistaken Point E; MPG = Mistaken Point G; Pigeon Cove; SH = Shingle Head; SP = Sword Point; C) Stratigraphic columns correlating the lithostratigraphic sections of the Bonavista (left) and Avalon Peninsulas (right) of Newfoundland; D) Fossil localities of the Bonavista Peninsula, after Hofmann et al. (2008), purple inset (2) of a): BC = Back Cove; H14 = Hofmann 14; HS = Haootia type surface; MC = Murphy’s Cove; PU.number = Port Union site.number; PU = MUN Surface (Liu et al. 2015b); PU.V = Port Union Vinlandia type surface; SPS = St. Peter’s Surface; STB = South Thectardis Surface.

21 to a strike-slip pull-apart basin (Wood et al. 2003, Ichaso et al. 2007), while the Signal Hill Group records dominantly shallow-water marine to alluvial deposits (Hofmann et al. 2008).

In the study area (Fig. 2.3b, d), four of the five formations that comprise the Conception Group are exposed. The lowest of these is the Gaskiers Formation (Fig. 2.3d), a ~250 m thick sequence of diamictites interpreted as glaciomarine in origin (Williams & King 1979). The overlying Drook Formation is comprised of thick-bedded siltstones, predominantly the products of moderate to weak turbidite flows capped by hemipelagic fallout. It has volumetrically the highest proportion of volcanic ash horizons, which has been taken to suggest that that volcanic activity was greatest at this time (Wood et al. 2003, Ichaso et al. 2007). Thin (<1 cm) epicalstic and volcaniclastic beds are identified in thin section from within the hemipelagite layers. Epiclastic beds are recognised as thin, discrete, tabular laminae, sometimes normally graded, with sharp and often loaded bases (Fig. 2.4c—f). They are are comprised of lithic fragments, rounded—sub-angular quartz crystals and only a small component (less than a few percent volume) of feldspar crystals. The volcaniclastic event beds have sharp bases but diffuse upper contacts (through mixing with non-pyroclastic material), and lack internal current-induced sedimentary structures. They often have a high clay content (likely resulting from decay of more reactive minerals), and occasional double-graded beds (Fig. 2.4b), which can arise through the presence of large but low density glass shards or pumice fragments, are also observed. The Briscal Formation is only present in the southern Avalon Peninsula, and is thick-bedded and comparatively coarse-grained (Wood et al. 2003, Ichaso et al. 2007).

The Mistaken Point Formation is the youngest formation of the Conception Group (Fig. 2.3c). It is comprised of medium-bedded siltstones that record deposition by strong—moderate turbidite flows capped by hemipelagite fallout. Appreciable influence by contour currents in the hemipelagite is indicated by the perpendicular orientations of felled fronds and current ripples (Ichaso et al. 2007, Wood et al. 2003), indicating that the event beds are not contourite deposits. While the thickness of the mud- dominated turbidite units at Mistaken Point has been interpreted to reflect ponding of turbidites against a topographic high within the basin (Ichaso et al. 2007), it is possible that the thick muddy tops result from mud particles settling out of suspension following passing of the turbidite flow, as a “hemiturbidite” (cf. Stow & Wetzel 1990), or from settling of material from a detached turbidite (cf. Stanley 1983). Within the hemieplagites, deposition by the distal parts of mud turbidites are indicated by the presence of silt lenses (Chapter 8). The upper part of the Formation comprises thin-bedded turbidites and red-coloured horizons indicating slower rates of deposition (Ichaso et al. 2007). Especially noticeable in this Formation are coarse-grained crystal tuffs that overlay some of the classic fossil horizons (Narbonne 2005).

The St. John’s Group comprises two formations. The lower of these, the Trepassey Formation, comprises thin-bedded siltstones interpreted to record deposition by moderate to weak turbidite flows with a small 22 A B

mt

mt

C D

e

bl e

E F

e

dt

mt

lb

Figure 2.4. Epiclastic and volcaniclastic event beds in thin section. A) Two microtuffs (mt) with broken and euhedral plagioclase crystals, and angular quartz grains, Charnwood Forest GC. B) Double-graded microtuff with opaque mineral bands, PU.4a. C) Graded epiclastic event bed (e), comprised of coarse silt, PU2. D) Tabular, epiclastic event bed with a very coarse silt to very fine sand grade basal lag (bl), PU3. E) Tabular epiclastic event bed (e) and undulose microtuff (mt), PC1. F) Loaded base (lb) to a graded eipclastic event bed with diffuse top (dt), HF3. All scale bars 0.5 mm. 23 proportion of hemipelagite (Ichaso et al. 2007). Slump deposits start to become more abundant at this level, indicating instability of the slope (Narbonne et al. 2001, Wood et al. 2003). Especially on the Bonavista Peninsula, this Formation also includes a high proportion of porcellaneous horizons similar to those of Charnwood Forest. Thin (< 1 cm) epiclastic and volcaniclastic event beds are observed in thin section, as well as silt lenses (Appendix A.2.2) interpreted to record deposition by mud turbidites (Chapter 8). The overlying Fermeuse Formation is dominated by thin-bedded siltstones deposited from weak, slow turbidites, with thicker slump deposits than are documented lower in the succession (Ichaso et al. 2007). Symmetrically-undulating bedding has been interpreted as seismically induced, supporting a near-horizontal depositional slope, which has been used to suggest a basin-plain depositional setting for the lowest part of this formation (Mason et al. 2013). This formation marks the transition to a strike- slip pull-apart basin setting (Wood et al. 2003). The Renews Head Formation overlies the Fermeuse Formation, and is distinguished by its greater silt and sand content than the Fermeuse Formation, and by the presence of pyritic lenticular bedding, current ripples, pseudonodules, water-escape structures and sand dykes (Hofmann et al. 2008). Its general coarsening-up sequence records prograding deltaic sedimentation, and represents the transition from the marine deposition of the underlying formations to the dominantly terrestrial sedimentation of the overlying ones (Hofmann et al. 2008).

The succeeding Signal Hill Group comprises grey to reddish siltstones and sandstones, with wave ripples, desiccation cracks and cobble conglomerates occurring higher up the sequence (Williams & King 1979). Representing the final fill of the basins, these sediments contain the first evidence for sub-aerial exposure, and were deposited under shallow-marine to fluvial conditions (Hofmann et al. 2008). The transition to the Cambrian is not exposed in the study area, but is likely to be unconformable (O’Brien & King 2005).

2.1.2.1 Age constraints for Newfoundland The Newfoundland succession is remarkably poorly dated considering its importance in the evolutionary story of the oldest complex macrofossils: only two recent, high-precision dates are available for the whole of the succession. The Gaskiers glacial diamictite has been dated to 582.4±0.4 Ma, and a comparatively thick unit in the upper part of the Drook Formation (close to the Pigeon Cove locality, Fig. 2.4a) has been dated to 578.8±0.5 Ma (Van Kranendonk et al. 2008). A date of 565±3 Ma is recorded for the Mistaken Point Formation (Benus 1988), from an ash layer in close stratographic proximity to the classic sites of the area. This is the only other date recorded for the Newfoundland succession but, as no data is presented with it, there is no way to assess its reliability. No dates have yet been reported for the Bonavista succession.

24 2.1.2.2 Deep-water fossil occurrences in Newfoundland The Newfoundland successions have been the subject of sporadic study since the first discovery of discoidal fossils in the region (Billings 1872), and the first discovery of complex macrofossils from the famed Mistaken Point locality (Misra 1969). The study of Ediacaran macrofossils in general flourished following their placement in a spearate, extinct Kingdom (the Vendobionta; Seilacher 1992), which prompted more intesive study of the Newfoundland macrofossils.

The oldest definitive rangeomorphs (Chapter 1), including Trepassia wardae (Narbonne et al. 2009), are recorded from the Drook Formation on the Avalon Peninsula (Narbonne & Gehling 2003). Their age is constrained by the ca. 579 Ma date of the Pigeon Cove localities, and their close temporal coincidence with the glacial Gaskiers Formation has been the cause of the excitement surrounding this glacial event (Narbonne & Gehling 2003; section 1.1.1). No fossils are known from the equivalent formation on Bonavista Peninsula (Hofmann et al. 2008), but the Pigeon Cove localities of the Avalon Peninsula (Fig. 2.3b) lie in the upper part of the Drook Formation (Liu et al. 2012). The classic biotas of the Mistaken Point surfaces (Fig. 2.3b), as well as many of the most diverse communities on the Bonavista Peninsula (such as the Murphy’s Cove localities, Fig. 2.3d), lie in the Mistaken Point Formation (Fig. 2.3b). The Briscal Formation contains the Bristy Cove localities (Fig. 2.3b), which host the holotype of the iconic rangeomorph, Fractofusus andersoni (Gehling & Narbonne 2007), but this formation is absent on the Bonavista Peninsula.

The Trepassey Formation is fossiliferous on both the Avalon and Bonavista Peninsulas, and includes the Shingle Head and Sword Point localities on the Avalon Peninsula (Fig. 2.3b), and the Port Union surfaces as well as HF14 on the Bonavista Peninsula (Fig. 2.3d). This formation also hosts the exquisitely preserved fronds of the Spaniard’s Bay localities (Narbonne 2004, Narbonne et al. 2009, Brasier et al. 2013; Fig. 2.3a) and the MUN surface (PU, Fig. 2.3d) on the Bonavista Peninsula (Liu et al. 2015b). The Fermeuse Formation hosts quite different assemblages on the two peninsulas. On the Avalon Peninsula, fossils in this formation are restricted to discoidal holdfasts and Palaeopascichnus (Liu et al. 2015a), but on the Bonavista Peninsula, at least the lower parts of this formation host diverse rangeomorph communities (BC, Fig. 2.3d; Hofmann et al. 2008) as well as the body fossil of a putative cnidarian (Liu et al. 2014).

2.2 Comparing the Newfoundland and Charnwood Forest successions.

The broad-scale depositional environment of the Newfoundland and Charnwood Forest successions are very similar: both were deposited in fore- or back-arc basins, with an appreciable volcanic input. While the Bonavista and Avalon Peninsula successions were arguably deposited in different basins (O’Brien & King 2005), they are readily correlated based on their lithostratigraphy (O’Brien & King 2005). However, the relationship of these Newfoundland basins to the Charnwood Forest depositional system is uncertain. 25 Comparing the biotas of the regions within the context of their lithological variation and depositional environment has significant implications for the discussion of the controls on community assemblage, particularly for the importance of fossil endemism (Wilby et al. 2011) versus palaeoenvironment (cf. Grazhdankin 2004; Chapter 8).While lithostratigraphic correlation between the Charnwood Forest and Newfoundland succession is more difficult, their broader depositional setting can be compared through the relative proportions of the five main lithologies identified in the Charnwood Forest succession (section 2.1.1.3).

The Trepassey Formation of the St. John’s Group, especially the exposures on Bonavista Peninsula, is most similar lithologically to the Bradgate Formation of Charnwood Forest: both are characterised by thin turbidite beds and a high abundance of porcellaneous (ash-fall) beds. In contrast, the Mistaken Point Formation is characterised by much thicker event beds and correspondingly thicker hemipelagite layers; it also lacks abundant porcellaneous beds, and has a very low proportion of volcaniclastic event beds. Additionally, the rare ash beds in the Mistaken Point succession include medium- to coarse-grained horizons, which are rare to absent in the Bonavista and Charnwood Forest successions. As such, although there is a decrease in the abundance of crystal tuffs through the Conception, St. John’s and Signal Hill Groups (Ichaso et al. 2007), there is a higher proportion of pyroclastic gravity flow deposits in the Bonavista succession, and especially in the Charnwood Forest successions, than in the Mistaken Point succession.

Intriguingly, the fossil assemblages in the Trepassey Formation on the Bonavista Peninsula are most similar to those of Charnwood Forest. Both contain abundant and large Primocandelabrum, Charnia and Charniodiscus. Strikingly, Fractofusus, which is so common in the Mistaken Point Formation is, with a few notable exceptions, rare in Bonavista and absent in Charnwood Forest. Such broad-scale lithological comparisons are interesting, and have proved meaningful in shallow-water settings (Grazhdankin 2004, Gehling & Droser 2013). However, it is the local palaeoenvironmental conditions with which the organisms would have had to contend on a day-to-day basis that will have exerted the greatest influence on them. The influence of local and transient sedimentary parameters on the taxonomic composition and structure of the communities is discussed in Chapters 8 and 9.

26 3. THE TAPHONOMY OF AVALONIAN RANGEOMORPHS

Disclosure statement: Parts of this chapter are based on material presented in Kenchington & Wilby (2014).

3.1. Introduction

Ediacaran macrofossils are amongst the least well understood of any macrobiotic assemblage in terms of their biology, palaeoecology, and phylogenetic affinity. In addition to allowing genuine morphological features to be distinguished from taphonomic artefact, a thorough understanding of the processes involved in their preservation can reveal important insight into certain aspects of their biology, such as tissue composition; despite the presence of phosphatised microfossils (Xiao et al. 2014), there is no record of the phosphatised muscle tissue that is so common in the Phanerozoic (Briggs 2003). Mineralization in macro-organisms only evolved in the terminal Ediacaran (e.g. Grant 1990, Grotzinger et al. 2000, Penny et al. 2014), and the makers of known trace fossils largely remain elusive: without soft-part preservation, we would have little knowledge of macro-benthic life during most of this critical interval of Earth history.

In any time interval, the particular mode in which a fossil is preserved depends on a variety of factors, including the rate of burial, the composition of the sediment (Narbonne 2005, Grazhdankin et al. 2008, Wilson & Butterfield 2014), the presence and nature of the microbial mat community (Gehling 1999, Gehling et al. 2005), and the chemistry of the pore waters (Mapstone & McIlroy 2006, Callow & Brasier 2009). Soft tissue preservation in the Ediacaran is dominated by three major taphonomic modes: mouldic, replication by early diagenetic minerals, and carbonaceous compression (e.g. Gehling 1999, Grazhdankin et al. 2008, Zhu et al. 2008, Cai et al. 2012). While mineralisation and compression fossils are known from throughout the Phanerozoic, mouldic preservation of soft tissues is argued to be most typical of the Ediacaran, and reflective of unique palaeoenvironmental conditions (Narbonne 2005, Droser et al. 2006, Grazhdankin et al. 2008, Gehling & Droser 2013, Buatois et al. 2014). Scavenging and deep bioturbation were absent and, with the exceptions of rare surface trace makers (Liu et al. 2010) and mat grazers/ miners in the younger parts of the succession (Seilacher et al. 2005, Buatois et al. 2014), the organisms were largely sessile and immotile. Consequently, there was a general lack of biological disturbance and, it is assumed, an attendant lack of significant time-averaging (Seilacher 1992, Clapham et al. 2003; though see Liu et al. 2011; Wilby et al. 2015). The Ediacaran also potentially had a different sedimentary and oceanic chemistry to the Phanerozoic, with lower seawater sulphate concentrations (Canfield et al. 2008), a condensed sediment-water geochemical profile which would have favoured early diagenetic mineralization (Callow & Brasier 2009) and, putatively, higher concentrations of labile dissolved organic carbon in the deep oceans (Sperling et al. 2011). Microbial mat fabrics were conspicuous in the Ediacaran (Steiner & Reitner 2001, Noffke et al. 2002, Gehling et al. 2005, Grazhdankin & Gerdes 2007, Callow & Brasier 2009, Wilby et al. 2011, Lan & Chen 2012), and are often cited as key to preservation in this time interval in particular (e.g. Gehling 1999, Briggs 2003, Darroch et al. 2012, Buatois et al. 2014, Raff & Raff 2014). New insights from petrographic and morphologic studies undertaken on samples from both Charnwood Forest and Newfoundland help to refine understanding of preservation in the fine-grained, deep-water settings of the Ediacaran of Avalonia.

27 3.2 Metamorphism

Both the Charnwood Forest and Bonavista Peninsula Ediacaran sections are exposed as the cores of faulted anticlines (Carney 1999, Hofmann et al. 2008), and the Avalon Peninsula section has also been subject to folding and faulting (O’Brien & King 2005). Charnwood Forest has undergone metamorphism to lower epizonal greenschist facies, and is pervasively inflicted by two intersecting, penetrative cleavages. The Neoproterozoic units of the Avalon Peninsula were subjected to prehnite–pumpellyite- grade metamorphism in the Cambrian (Papezik 1974), corresponding to the anchizone metamorphic facies (200–300 °C) of (Merriman & Frey 1998), slightly lower than the epizonal facies reached by the rocks of Charnwood Forest during the Caledonian orogeny (Merriman & Frey 1998, Árkai et al. 2003, Carney et al. 2008). Laumontization of tuffs on the nearby Bonavista Peninsula has been suggested to indicate negligible metamorphism (Retallack 2014) but, as this mineral can form under a wide range of conditions (Coombs et al. 1959, Boles & Coombs 1977, Árkai et al. 2003), this is inconclusive.

The intersection of the tectonic cleavage with bedding planes throughout Avalonia locally destroys or at least distorts these low epirelief fossils (Wilby et al. 2011, Liu et al. 2015a, Mitchell et al. 2015), and produces thin lineations which can confuse their morphology. These overprint the morphology of all fossils in the region, and locally obscure fine morphological details. The lineations resulting from cleavage- bedding plane intersections occur at a similar spacing to the finer levels of branching architecture. The two structures, original and secondary, can be distinguished by comparing the orientations of linear structures to cleavage traces on the bedding surface, and determining whether the structure in question continues beyond the outline of the larger structures in which they occur and which are demonstrably biological (i.e. whether or not putative fourth order branch margins continue beyond the margin of the host third order branch, in which case they are treated as taphonomic artefacts).

3.3. Mouldic preservation: mechanisms and mineralisation

Mouldic preservation is the most abundant and typical preservational style of the Ediacaran (e.g. Gehling 1999, Steiner & Reitner 2001, Narbonne 2005, Grazhdankin et al. 2008, Cai et al. 2012, Meyer et al. 2014), and the “death mask” model of mouldic preservation (Gehling 1999) has become the iconic taphonomic mode of the period. Terms used to describe the nature of the moulds, and their relationship to the beds preserving the fossils, were introduced by Glaessner and Wade (1966). Features observed on the top surface of a bed are termed epirelief, and those seen on the base of a bed are hyporelief. Features which form hollows or depressions have negative relief and those which protrude above the surface have positive relief. It is generally assumed that negative epirelief preservation captures the surface in contact with the sediment, whereas positive epirelief preservation captures the surfaces of the organism in contact with the overlying bed (Narbonne 2005). The sense of relief with which a particular fossil is preserved with is thought to involve an interplay between the relative resistance of the soft parts to

28 collapse, and the timing of substrate lithification (Gehling 1999, Narbonne 2005): more robust (more resistant to decay and/or physical damage) or recalcitrant parts collapse or decay more slowly, and so are cast by still-soft material from the underlying bed being injected upwards (creating negative hyporelief/ positive epirelief impressions), whilst more fragile (more prone to decay and/or physical damage), fluid- filled or labile parts collapse or decay comparatively quickly, creating impressions which are filled and presumably cast by material from the overlying bed subsiding into the void (resulting in positive hyporelief/negative epirelief impressions). The latter process likely requires stabilization of the lower surface of the organism prior to complete decay in order to retain the observed level of morphological detail (Darroch et al. 2012). In the Avalon Assemblage sites of Newfoundland and Charnwood Forest, the fossils are only seen preserved as epirelief impressions; the counterparts (at least of the frondose parts) are as yet unknown.

3.3.1 The “death mask” hypothesis

Once formed, the impressions must be rapidly stabilized in order to be preserved. Groundbreaking work into understanding this mode of preservation, and particularly how biological structures could be preserved in their original positive relief, was made by Gehling (1999), based on observations from the Flinders Ranges. The elegant “death mask” hypothesis he proposed consists of four main stages:

1) Organisms living on a microbial mat were smothered by sediment; 2) Labile or fluid-filled organisms/tissues decayed rapidly, leaving impressions which were infilled by sediment from the overlying bed, while more robust organisms/tissues persisted; 3) Sulphur-reducing bacteria exploited the organic material of both carcasses and mat, releasing reduced sulphur compounds which combined with iron in the sediment porewaters, resulting in the formation of pyrite. This pyrite coated the lower surface of the now-collapsed labile organisms/ tissues, and/or the upper surface of the recalcitrant organisms/tissues, stabilizing the impressions and forming the so-called “death mask”; 4) Death masks which formed over more recalcitrant tissues were infilled from below by still- unlithified sediment. The pyrite thus formed is observed on the base of the burial event bed, comprising a sole veneer of sediment grains and interstitial pyrite; this layer is typically no more than a few sand grains thick (Gehling et al. 2005, Mapstone & McIlroy 2006).

Since its first proposal, the death mask model has been expanded upon and applied far beyond the deposits from which it was first described. Comparable pyrite sole veneers have been described from the Amadeus Basin of Australia (Mapstone & McIlroy 2006), and inferred from hematite partings (Gehling et al. 2005) and concentrations of Fe and S within preserved microbial mats (Laflamme et al. 2011) from Newfoundland. 29 3.3.2 The timing of pyritisation: new evidence from Charnwood Forest and Newfoundland

New thin sections from Avalonia, including three from Charnwood Forest and twenty-four from Newfoundland (Appendix A.2.2), provide an excellent opportunity to constrain the sequence and timing of mineralisation. In both regions, abundant pyrite is observed in the matrix of the sediment or ash overlying the fossil horizon (Fig. 3.1a, b). Pyrite is also occasionally found concentrated in discrete laminae (Fig. 3.1c), and in microtuff horizons not associated with fossils (Fig. 3.1d, e). There are two main types of pyrite growth in these sections: framboidal pyrite (Fig. 3.2a,b), and concentric-ringed, sub-spherical (colloform) growths (Fig. 3.2b—f), with both textures frequently occurring together and being of similar dimensions (Fig. 3.2). Framboids vary in measured section from ~5 to 20 μm, averaging between 6 and 9 μm in the sections that contain abundant framboids (Fig. 3.1—3.3; Appendix A.2.3). Some thin sections also contain massive pyrite (Fig. 3.2g, h), and others contain cubic pyrite, both as large, euhedral crystals (Fig. 3.2i) and as disseminated crystallites (Fig. 3.2j). Some individual framboids are mantled by eu-/subhedral outer layers, both as large, single cubic crystals and as a ring of smaller crystals with euhedral terminations (Fig. 3.2f). Others additionally display iron (hydr-)oxide weathering rinds (Fig. 3.2b). In some cases, clusters of several framboids or of colloform pyrite are mantled by contiguous outer layers (Fig. 3.2c—e).

The abundance of pyrite differs between thin sections (Figs. 3.1, 3.2, 3.4). Where it occurs in the matrix of the ash, pyrite is most commonly interstitial between crystals of quartz and feldspar (now replaced by quartz and clay minerals). In some thin sections, the pyrite growths are randomly distributed throughout the ash (Fig. 3.1c), whereas in others they are concentrated in regions, as polyframboidal (sensu Love 1971) and colloform masses (Fig. 3.2d, e). In one section, pyrite framboids are aligned in rows which appear to mimic original layering in the ash (Fig. 3.4c, f). In each case, pyrite forms within the ash, not exclusively (or even dominantly) along the interface between substrate and ash, as would be expected from the “death mask” model.

Both colloform and framboidal pyrite are observed to overprint diagenetic, iron-rich clay minerals that have replaced the matrix of the ash and the underlying sediment, truncating contacts between the clay and quartz intergrowths (Fig. 3.2a, b, f; Fig. 3.4, Fig. 3.5a,b). In no instance are the clay minerals observed to deflect around or to mantle the pyrite growths, as they would if the pyrite was present during clay mineral growth. Furthermore, it also appears to terminate against metamorphic titanite needles (Fig. 3.5a, b), although their spatial relationships are less certain. Cleavage is observed to offset and smear pyrite laminae (Fig. 3.1a), and so is interpreted to post-date pyrite formation. In Bed B of Charnwood Forest, sporadic spheroidal pyrite “suns” (Fig. 3.5c), together with a layer of colloform pyrite, coats clay minerals that are aligned parallel to cleavage (Fig. 3.5d—f). Only rare framboidal pyrite is observed

30 A alb fmm alb

alb

qtz fmm pyr

1 mm

B alb C ksp alb

fmm

pyr 20105.1b

fmm pyr 200 μm 200 μm

D E

pyr

1 mm 100 μm

Figure 3.1. Distribution and occurrence of pyrite, which appears white in backscatter images (a — d), black in transimtted light (f) . A) Pyrite in crystal tuff, predominantly in matrix, distributed in layers. Sample 10.9.12 (collected by A. Liu); B) Random distribution of pyrite throughout ash matrix. Sample 15105.2, SP.2; C) Discontinuous lenses of pyrite, no associated with ash or fossil-bearing horizon. Sample 20105.1b, H14.1b; D) Pyrite in microtuff with loaded base (close-up in inset), and in two overlying, discontinuous laminae, probably recording similar small-scale tuffaceous events with either a high iron content. Sample 22105.3, PU.3; E) Pyrite (opaque) distributed throughout a microtuff event not associated with a fossiliferous horizon, concentrated within and particularly at margins of inner lamina. Sample 19105.4, PU.4. Pyr = pyrite; qtz = quartz; alb = albite; ksp = K-feldspar; fmm = iron-magnesium-rich mica.

31 A B alb

pyr

pyr pyr alb fmm fmm

20 μm 10 μm

C qtz D E pyr

pyr pyr alb

20μm 10 μm

F

pyr fmm alb

pyr fmm

alb

10μm 20μm

G H I J pyr pyr

pyr pyr

100μm 10 μm 20 μm 20 μm

Figure 3.2. Pyrite textures, pyrite appears white in backscatter images. A) Framboids ranging from 5 to 20μm in diameter, distributed randomly through matrix of ash. Sample 10.9.12, BS; B) Concentric-ringed, framboidal and disseminated pyrite, and C) Concentric pyrite, outer lamina encasing both framboid and earlier laminae (black arrow). Sample 15105.2, SP.2; D, E) Colloform masses in matrix of ash; D) is inset of E). Sample 21105.1, SPS.1; F) Concentric-ringed and disseminated pyrite in matrix of ash, clearly overprinting growth of micaceous minerals in ash. Sample 15105.2, SP.2; G) Solid mass of pyrite encasing small crystals of ash matrix and H) inset. Sample 19105.2, MC.2; I) Late diagenetic cubic pyrite. Sample R003, Bed B. J) Disseminated pyrite growing with cubic form. Sample 21105.4, BC. Pyr = pyrite; qtz = quartz; alb = albite; ksp = K-feldspar; fmm = iron-magnesium-rich mica. 32 20105.1b 22105.3 15105.2 10.9.12 Density 0.00 0.05 0.10 0.15

0 5 10 15 20

diameter (μm)

Figure 3.3. Framboid density distribution profiles for the four thin sections in which more than 100 framboids occur in a single horizon (20105.1b, HF14.1b; 22105.3, PU.3; 15105.2, SP.2; 10.9.12, BS). Raw data in Appendix 2.3. in this section. The clays are interpreted to record mineralisation during bed-slip, and so the pyrite formation post-dates this late metamorphic fabric.

3.3.3 Phosphatisation

In one thin section from Charnwood Forest (Bed B; Fig. 2.1; Fig. 3.5c—f) and one from Newfoundland (St Peter’s Surface, Fig. 2.3d; Fig. 3.5 g—k), thin (~10 μm) fragments of microbial mat are preserved, identified by wavy laminae and clot structures. In the Bed B sample, this mat has been phosphatised, and occurs above an ash that underlies the fossiliferous surface (Wilby et al. 2015). In the St Peter’s section, the mat fragment appears to overlie the background (hemipelagite) sediment directly, but is patchy in its occurrence (Fig. 3.5g). Compared to the background sediment, the mat fragment has anomalously high concentrations of iron and phosphorous (Fig. 3.5j). This suggests that it was originally phosphatised but has been subsequently pyritised (Fig. 3.5k), both on a fine scale and as a single spherical growth (Fig. 3.5h). In other sections, large metamorphic apatite growths occur beneath the fossil surface (Fig 3.5 l, m).

33 A B fmm pyr fh fmm alb alb pyr alb pyr pyr fmm

fmm qtz alb qtz 100 μm 10 μm

C D ti

fmm alb alb pyr qtz his pyr fmm 20 μm

fmm E alb fmm alb alb pyr qtz

100 μm 20 μm

F 100 μm G ksp fmm pyr alb pyr fmm pyr

fmm qtz

20 μm pyr Figure 3.4. Pyrite textural relationships, pyrite appears white in backscatter images. A) Framboids with iron (hydr- )oxide weathering rind framboids overprinting clay and mica mineral growth in the ash matrix, and infilling embayments in quartz and albite crystals, and B) Framboids and disseminated pyrite overprinting clay and mica contacts. Sample 10.9.12, BS; C) Framboids overprinting diagenetic clays, quartz and feldpsars. Sample 20105.1b, H14.1b; D) and E) Section of C), to show D) contacts between crystals and E) framboidal texture; F) Framboids aligned in subparallel bands in the ash matrix, clearly overprinting the diagenetic, aligned mica crystals in the ash and overprinting breakdown products of recrystallised feldspars. Sample 10.9.12, BS. G) Colloform and framboidal pyrite infilling embayments in quartz crystal. Sample 21105.1, SPS.1. Pyr = pyrite; qtz = quartz; alb = albite; ksp = K-feldspar; fmm = iron-magnesium-rich mica; his = hole in slide.

34 A fmm B C ti fmm pyr qtz ti

pyr

qtz qtz 100 μm 10 μm 10 μm pyr

D E F pyr pyr

20 μm

fmm qtz fmm 20 μm 10 μm

G pho H I pyr pho fmm fmm

pyr

100 μm 10 μm 10 μm

J K L M qtz qtz alb

fmm pyr alb pho fmm 20 μm 20 μm 200 μm 100 μm fmm

Figure 3.5. Pyrite and phosphate textural relationships, pyrite appears white in backscatter images, phosphate mid- grey. A) Pyrite framboids in associated with titanite needles. The framboids appear to truncate the titanite, and B) inset. Sample 22105.3, PU.3. C—F) Sample R003, Bed B; C) Pyrite “sun”. D) Colloform pyrite layer coating diagenetic clay minerals interpreted to have formed during bed slip. The pyrite layer bisects the clay layer, and E, F) Insets of D) showing pyrite clearly overgrowing and mantling diagenetic clays; G—K) Sample 21105.1, SPS.1; G) Phosphastised and pyritised microbial mat in association with ash. Mat directly overlies background sediment, but its relationship with the ash is not conclusive; (left-hand) dashed box = area in J, K; (central) dotted box = H, (right-hand) solid box = I. H) Inset of G) showing mat texture and spherical pyrite growth. I) Portion of mat with clay minerals having the typical texture of those in the ash matrix. J) Phosphate abundance. K) Iron occurrence. L) Large cluster of apatite crystals intergrown with iron-rich micas. Sample 10.9.12, BS. M) Euhedral apatite crystals. Sample 19105.5, PU.5. Pyr = pyrite; qtz = quartz; alb = albite; ksp = K-feldspar; fmm = iron-magnesium-rich mica; his = hole in slide; pho = phosphate.

35 3.3.4 Discussion

Several factors influence the thickness and distribution of pyrite formation, in turn influencing the anatomical fidelity of the resulting impression – if only thin or patchy pyrite is formed, then incomplete fossils are formed, and conversely, if very thick layers of pyrite form, then some of the anatomical detail may be obscured (e.g. Darroch et al., 2012; Meyer et al., 2014). Most importantly, a balance exists between the quantity, quality and distribution of organic matter, and the availability of sulfate and iron ions in the system. Early diagenetic formation of pyrite requires pore-waters rich in dissolved iron (Raiswell 1993, Raiswell et al. 1993, Raiswell et al. 2008). If pore-waters are sulphide-rich, there must be a readily available and abundant source of iron from minerals with short half-lives, such as iron (hydr- )oxides (Pyzik & Sommer 1981, Canfield 1989). Microbes are also required in the system in order for the reaction to take place (Darroch et al. 2012).

Microbial mats are often implicated in preservation of soft tissues. This is especially true in the Ediacaran, where the buried microbial mat is argued to have provided both a diverse population of decay bacteria and a ready supply of organic matter at the horizon which hosts the fossils (Gehling 1999). Additionally, the re-establishment of the microbial mat on top of the event bed is argued to have sealed the sediment beneath, isolating the now anoxic/dysoxic pore waters from the oxic water column, controlling the availability of sulfate ions and maintaining anoxic pore waters in even porous sediments (Gehling et al. 2005, Callow & Brasier 2009). In “death mask” preservation, it is not the soft tissues themselves that are preserved, as this is unlikely to occur if there is abundant carbon in the system, or where diagenetic pyrite is abundant (Briggs et al. 1991, Raiswell et al. 1993, Briggs et al. 1996, Raiswell et al. 2008, Farrell et al. 2013, Farrell 2014), but rather their impressions left in the microbial mat.

3.3.4.1 The location of pyritisation and variability of preservation. The dominant occurrence of pyrite in ash rather than the background sediment may be explained by iron- rich phases (minerals, glass) in the ash matrix providing a local source of iron, and in turn suggesting that the pore-waters were more sulphidic than ferruginous. The original nature of such phases would have imparted an important control on the timing of pyrite formation: sulphide reacts with iron hydr(-oxides) to form pyrite within months, but with iron silicates on timescales >105 years (Canfield et al. 1992). Mobilisation of iron from silicates occurs on a similar timeline to pyritisation of recalcitrant organic tissues (Raiswell et al. 1993). The layered distribution of some framboids may reflect original compositional layering in the ash (cf. Fisher and Schmincke, 1984), perhaps recording different abundance of iron-rich phases in the ash.

The variability in pyritisation on different surfaces may be attributable to an original difference in the amount of organic material present at the site from where the thin section was taken or, on a larger scale, to different amount of iron and/or sulphide available for reaction. It is also possible that the 36 thickness of pyrite relates to the original thickness of microbial mat present – all other factors permitting, the mat would provide abundant organic matter to sustain sulphur-reducing bacteria, which would produce a thick layer of pyrite that has subsequently been remobilised to produce the observed textural relationships. Alternatively, thicker ashes or ones with a higher proportion of Fe-rich phases may have provided the iron for diagenetic (abiotic) pyrite production during decay of minerals/glass in the ash matrix. Coincidentally, the extra-cellular polysaccharides (EPS) produced by a thick (laterally extensive) mat may have served to stabilise the impressions (Darroch et al., 2012) and retain higher preservational fidelity than in a thin mat. Until fossils of different preservation qualities can be sectioned, this will remain a point of contention. The fact that the only fossils to have been sectioned are Aspidella (probable holdfasts) from the shale and siltstone sequence of the Fermeuse Formation of Newfoundland (Chapter 1), which lacks preserved fronds and likely had a very different taphonomic history (Laflamme et al. 2011), exacerbates our poor understanding of Ediacaran taphonomy.

3.3.4.2 The timing of pyritisation Framboidal pyrite can form in various environments, such as precipitation in (euxinic) water columns (e.g. Wilkin et al. 1996), microbially-induced or purely abiotic growth in sediment pore-waters (Wilkin & Barnes 1997, Ohfuji & Rickard 2005, Cavalazzi et al. 2014), during metamorphism even up to lower greenschist facies (Scott et al. 2009), and through hydrothermal activity (Wilson et al. 2003, Scott et al. 2009). While strongly negative sulphur isotopes in sections from the Fermeuse Formation in Newfoundland (Chapter 2) have been taken to indicate a microbial sulphur source (Wacey et al. 2015), this has not been extended to the deep-water Avalonian sections studied here (Chapter 2). In any case, the evidence of pyrite coating clusters of framboids observed here indicate numerous generations of pyrite formation (section 3.3.2).

Colloform pyrite, as seen here, has been recorded from a number of other deposits, but is formation is contentious and complex (e.g. Barrie et al. 2009): in some cases it is demonstrably ooidal, either original, where it is interpreted to record fluctuating redox conditions (Schieber & Riciputi 2005), or replacing chamosite (Schieber & Riciputi 2004). In others it forms masses which grew during hydrothermal mineralisation and/or diagenesis (e.g. Barrie et al. 2009): this growth texture alone cannot constrain timing, but its textural relationships to other minerals can.

The combination of framboidal and colloform pyrite in the studied sections (section 3.3.2) is remarkably similar to that observed in coal deposits from Brazil, but the timing of formation of the mineral in that deposit was not well constrained (Ribeiro Filho & Mussa 1977). Several lines of textural evidence can constrain timing of pyrite formation in the thin sections studied here: 1) the large size distributions and large maximum sizes of framboids here (Fig. 3.3) is consistent with their formation in the sediment rather than the water column (Wilkin et al. 1996, Wilkin et al. 1997); 2) pyrite layers mantling several growths

37 (framboids and concentric-ringed) and large, euhedral crystals supports at least some metamorphic pyrite growth; 3) the overprinting of metamorphic minerals by framboids and concentric-ringed pyrite indicates that mineral growth post-dates metamorphism; 4) smearing of pyrite into the plane of cleavage suggests that pyrite formation pre-dates cleavage formation, except in Charnwood Forest where pyritisation post- dates bed-slip minerlisation. Taking the sum of textural evidence, it is clear that pyritisation occurred too late during diagenesis and metamorphism to be responsible for primary fossilisation of the organisms present. This is consistent with the sequence of mineralisation observed in the Amadeus Basin (McIlroy and Mapstone, 2006), where pyrite (subsequently altered to haematite) forms only after authigenic quartz, illite, K-feldspar, chlorite, smectite and glauconite. Indeed, its occurrence in ash events associated with neither microbial mat nor hemipelagite supports the interpretation of an abiotic genesis for at least some of the observed pyrite.

3.3.4.3 The importance of different mineral phases in Avalonian preservation What, then, was responsible for preservation? Even if some pyrite formed early enough during diagenesis to form a “death mask”, then the level of morphological detail retained in fossils is difficult to reconcile with their collapse and/or decay prior to pyritization (which does not occur until step 3 of Gehling’s (1999) model. Recent experimental and fossil evidence suggests that either authigenic aluminosilicate templating or EPS from the microbial mat (Darroch et al. 2012) may have stabilized the impression during the earliest stages of its formation, until clay formation or mineralisation occurred. The observation of phosphatised mat fragments (section 3.3.3) suggests that phosphatisation was a more important process in preservation, at least in Avalonia, than previously thought. Its rarity compared to pyrite layers (or iron hydr-/oxide after pyrite; Fig. 3.4b) may be due in part to its removal from exposed surfaces via weathering and abrasion, or to dissolution and reprecipitation to form the metamorphic apatite growths observed in many thin sections. By extension, while pyrite may have served to stabilise the impression through metamorphism, it occurred too late to have been instrumental in the initial preservation of the fossil.

3.4. Biostratinomy and the gradation of forms

In addition to diagenetic and metamorphic processes, post-mortem, pre-burial processes can also influence an organism’s preserved morphology. Taxonomic classification frequently includes discussion of the morphological proportions of an organism (Chapter 4). Understanding the extent of post-mortem modification of morphology is crucial to prevent construction of taphospecies, and to acurately determine the functional morphology of an organism (Chapter 5, 6). Biostratinomy encompasses the effects of post-mortem compaction, contraction or expansion (whether by dehydration or bacterial decay), folding, and transport (Gehling et al. 2005), which necessarily influence the final morphology of the fossil. Compared to the Phanerozoic, fewer biostratinomic processes operated in Ediacaran times. Scavengers

38 consume or disarticulate carcasses, but are unknown from the period, and bioturbation was limited. This has been argued to have allowed the organisms to decay undisturbed on the sea-floor (Liu et al. 2011), representing a type of time-averaging – the remains of dead organisms persisted on the sea-floor and were preserved alongside living ones. However, biomineralisation only developed in the terminal Ediacaran, and the lack of hard parts in the majority of the organisms of this time greatly reduces opportunity for extensive time averaging and reworking. In contrast, microbial decay and abiotic (physical) disturbance would have endured throughout the Proterozoic and into the Phanerozoic.

Syn- or post-mortem distortion of the morphology of the organism by physical processes has been recorded from Ediacaran localities around the globe, and includes wrinkling (Gehling 1991), folding (Seilacher 1992, Evans et al. 2015) and ripping (Runnegar & Fedonkin 1992). If different parts of an organism had dissimilar rheologies, they will be affected differently by shared biostratinomic processes. For example, the crenellated skirt of Kimberella shows comparatively greater deformation or wrinkling than the rest of the organism, and is accordingly inferred to have been a broad, flattened “foot” which was less robust than a surrounding, unmineralized shell (Fedonkin & Waggoner 1997).

3.4.1 Alteration of the preserved morphology of multifoliate rangeomorphs

The strong alignment of individuals within Avalonian communities indicates that they were felled by a flow, likely as part of the depositional event that smothered them. This flow must have had some influence on the preserved morphology of the organisms, and must be accounted for when reconstructing the organisms and their communities. Indeed, the felling flow in Charnwood Forest was strong enough to twist and even enrol fronds (Fig. 3.6a—c). Fronds of increasing size may show variations in susceptibility to current-induced stacking of branches and compression of overall form. The observed change from an I-shape outline to V- to U- and to O-shape plan-view morphology in Bradgatia (Flude & Narbonne 2008) warrants discussion. Although originally proposed to be an ontogenetic series, with changing overall form through growth, it is here considered much more likely that it is a taphonomic interaction of growth and current: as additional branches are inserted and the existing folia expand outwards, the overall form becomes less easy to compress into a “leek-shape” via the current and post-felling compression, and so a progressively more chaotic and open structure results from increasing size (cf. Brasier et al. 2013).

Particularly if the branches are rigid, then their preserved lengths will reflect the length of the branch that impacted the sediment – an indeterminable amount may have remained aloft above the sediment and therefore is not be preserved. Similarly, the preserved length and width of the crown will necessarily reflect its compacted morphology – depending on how rigid the branches are, this could be quite different to its in-life proportions. For example, for entirely rigid branches, only a small section of the crown will be preserved. The number of folia preserved and identifiable is also a product of taphonomy, as only those folia that are in direct contact with the casting medium will be preserved in their entirety, 39 Figure 3.6. Taphonomy of fronds. A—C) The influence of current on morphology in Charnwood Forest. A) Small frond (white arrow) wrapped around the stem of a large Charniodiscus, GSM105968; B) an enrolled frond, GSM106097; C) twisting and disruption of branches of a large Charnia (white arrow), GSM105873; D) Differential preservation in a Beothukis specimen from Spaniard’s Bay (Chapter 2), scale bar = 5 mm; E) upper insert of d) showing typical, high- quality negative relief preservation in the area; and F) lower insert of d), showing poor-quality, positive preservation caused by deflation and infilling of the branch by sediment (Brasier et al. 2013); G) an “iveshediomorph” with typical reticulate texture, and H) a !pizza disc” iveshediomorph, Scale bars in a—c = 2 cm, and in g, h = 5 cm.

40 and any overlying ones will be preserved only as partial impressions and only if some part of them contacts the casting medium (discussed in more detail for Orthiokaterna, Chapter 5). This is expressed in Primocandelabrum by two main branches being preserved in most specimens, with the third indicated by a partial impression between the two clearly preserved stems. However, the extent of taphonomic influence on crown morphology and disc shape should be similar within a taxon, as the rheology of the organism, its overall proportions and susceptibility to taphonomic alteration should remain consistent through ontogeny and vary proportionally with total height.

3.4.2 Variation in preservation across a single specimen

In fossils from Newfoundland, the quality of preservation (as indicated by preservational resolution, clarity of outline and sharpness of relief) is seen to decrease along their length (Laflamme et al. 2007). On a finer scale, the branching pattern of rangeomorphs may also be affected during the burial event, for example the current-induced imbrication of primary branches recorded in specimens from Spaniard’s Bay, Newfoundland (Brasier et al. 2013). In that locality, differential positive and negative relief was interpreted to record deflation and infilling by sediment of some branches but not others (Fig. 3.6d—f). In Charnwood Forest, preservational quality varies across a specimen on two scales: across the whole organism, and across a single rangeomorph branch (at folium and first order levels).

In unifoliate rangeomorphs from Charnwood Forest, preservational quality decreases distally, explained as increasing amounts of ash falling between the frond and the preservational surface as it was felled (cf. Laflamme et al. 2007, Laflamme & Narbonne 2008). In contrast, arboreomorphs and mutifoliate rangeomorphs with long stems (e.g. Orthiokaterna, Primocandelabrum anatonos) typically have a clear and well-preserved holdfast and distal parts of the frond, but the stems and proximal parts of the frond are comparatively faint. This may be explained by these organisms having relatively inflexible stems (Chapter 5), capable of bending only in an arc and remaining aloft (and therefore above the plane of preservation) for at least part of its length. The poor preservation of proximal parts of the frond may also reflect this process, as the stiff stems would cause the frond to be held above the sediment (at least until ash separated the organism and the substrate). Alternatively, the branches may be stiffest at their base. The crown of multifoliate rangeomorphs also increases in preservational quality laterally. This is attributed to the decreased amount of branches compressed one on top of the other in these regions, compared to the proximal and central parts of the frond. The resulting compound impression would be more confused and distorted than parts recording only a single impression.

3.4.2.1 Variation across a single rangeomorph branch Preservational quality also varies across an individual rangeomorph branch at all resolvable levels (typically up to third order). Perpendicular to branching, the topographic profile slopes steeply downward, and then curves gently upwards (Fig. 3.7a). Quality of preservation is consistently best on the gently-

41 sloping surface, with poorer preservation on the steeply sloping surface. This profile and variation in preservation can help to identify the order of overlap between adjacent branches, which will in turn have influenced how the current interacted with the surface in life. For example, if there is no overlap and the elements are tucked one against the other, surface roughness will be lower than if there is a degree of overlap, affecting the thickness of the diffusive boundary layer over the surface (Chapter 7). Additionally, the direction of overlap (proximal over distal or distal over proximal) will alter the local path of the flow as it is channelled up the surface of the larger-scale structure.

Initially, it appears that the element whose outline is observed in its entirety is in contact with the sediment, and therefore that all obscured or partially preserved elements overlie this. However, this does not concur with all the features observed. In particular, to obtain the observed profile (sloping upwards distally), the overlying element would need to be preserved with a tick-shaped profile (Fig. 3.7c—d), with a steep curve downwards at the overlap point, rather than a simple sequence of stacked elements (Fig. 3.7e). Additionally, the overlying element would need to be pushed below the topographic level of the underlying element (Fig. 3.7b—d). Although this is potentially achievable through differential loading, it is here considered unlikely. Accordingly, the following taphonomic mechanism is proposed.

1) Upon felling of an individual, one element overlies another (Fig. 3.7e). 2) Following compaction (by the weight of the organism and/or the overlying sediment, or simply following taphonomic collapse of the element), a small amount of substrate (sediment or EPS from the microbial mat) is squeezed between the sediment-facing (underlying) element/branch and the overlying element/branch (Fig. 3.7f). This would raise the edge of the overlying element to give the gentle concave-up slope, and may cause the underlying branch to become depressed into the substrate, producing the steep slope downwards in profile. 3) The squeezed-up substrate takes the impression of the base of the overlying element, whilst the original substrate takes the impression of the underlying element (Fig. 3.7g). As the organism decays, both impressions are stabilized by EPS from the mat (cf. Darroch et al. 2012). The formation of clay minerals or phosphate further stabilises the sediment (Fig. 3.7g). 4) When the overlying sediment is weathered off to reveal the fossiliferous surface, it is the impression of the overlying element that is observed in its entirety, while the underlying element is visible only beyond the margin of the overlying element (Fig. 3.7h). This model explains not only the topographic profile observed, but also the variance in preservation across the profile in terms of resolution and sharpness. Similarly, displacement of substrate over parts of the organism may explain the variable preservation over the whole organism, as obscured parts would leave their impression on the substrate, but not on the upper sediment surface which is exposed at outcrop.

42 A B C D

E F G H

Event bed Sediment Rangeomorph element Outline of rangeomorph element

Good preservation Poor/no preservation Collapse of rangeomorph element Movement of sediment Phosphatisation and/or clay mineral precipitation Pyritisation Figure 3.7. Taphonomic mechanism to explain the differential preservation across a rangeomorph element, and the order of overlap between elements. A) Observed profile of rangeomorph element; B) rangeomorph is felled, branches overlap proximal over distal; C) proximal element becomes pushed down into sediment; D) profile is mineralised. Alternative mechanism: E) rangeomorph is felled, with elements overlying distal over proximal; F) ash is deposited, rangeomorph elements collapse/are compressed, and substrate ± microbial mat ± EPS are squeezed up between the elements; G) the underneath of the elements is stabilised by EPS and/or phosphatisation; H) pyritisation in the overlying ash further stabilises the impression (post-decay).

3.4.3 Iveshediomorphs

Forms currently referred to as “iveshediomorphs” (Fig. 3.6g, h) are contentious, and include a wide spectrum of morphologies. These were originally described from Charnwood Forest as discrete taxa, and include Ivesheadia, Blackbrookia, Pseudovendia and Shepshedia (Boynton & Ford 1979, Boynton & Ford 1995; similar forms in Newfoundland are referred to as “pizza discs” (Fig. 3.6h), “lobate discs” and “bubble discs” (Narbonne et al. 2001, Laflamme et al. 2012a). A full spectrum between such forms and fronds exhibiting fine detail has been documented from several bedding planes in Newfoundland, leading to the interpretation of “ivesheadiomorphs” as the remnants of dead organisms which were in the process of microbial decay at the time of burial (Liu et al. 2011). The irregular, unusually high relief and often network-like internal features of these forms (Fig. 3.6g) were suggested to represent a conflation of sediment trapped by EPS and gas derived from the decay process and associated decay microbes (Liu et al. 2011). However, other authors have suggested alternative explanations. Laflamme et al. (2012a) interpret these structures as purely microbial in origin, differing from the original interpretation only in the sense that the microbial colonies are growing on the substrate/microbial mat, rather than on a dead macro-organism. Alternatively, Wilby et al. (2011) propose that at least some of the forms may be created by differential loading on the fossil-bearing surface following collapse of organisms within the overlying bed.

The taphomorph interpretation has recently been extended to both Thectardis and Charniodiscus from Newfoundland (Antcliffe et al. 2015). Forms from Newfoundland and Australia that are ascribed

43 to Charniodiscus preserve only two orders of branching (Chapter 1), whereas C. concentricus from Charnwood Forest (the holotype of the genus) appears to record third-order branching. This is the main reason the Charniodiscus specimens in Newfoundland were assigned to taphomorph status. Thectardis has no internal branching. However, the proposed taphonomic index on which the interpretations of these fossils as taphomorphs are based relies on the organism originally having more than two orders of branching. Consequently, application of this index to any organism that originally had two or fewer orders of branching will inherently assign them as taphomorphs of varying degrees, and so the argument becomes circular. The consistent bilateral symmetry of Charniodiscus (unclear in C. concentricus due to the incomplete nature of its frond) is distinct from the gliding symmetry characteristic of rangeomorphs (Chapter 1), and taphonomic alteration of symmetry in all the described Charniodiscus species is implausible and lacks convincing explanation. Additionally, no rangeomorph is known that has the long stem and stem/frond angle of C. procerus, nor the distal spine of C. spinosus (Laflamme et al. 2004) – thus, special pleading is required for all rangeomorphs with those original morphologies to be preserved only as taphonomorphs. The clear outline of the both Charniodiscus and Thectardis, and their distinct spatial patterns (Mitchell et al. 2015), additionally count against a taphomorph interpretation for these taxa. It is, however, likely that Charniodiscus from Newfoundland and Australia belong to a different genus than C. concentricus, which may indeed be a rangeomorph.

3.5 Conclusions and future work

Taphonomic processes and biases impact all aspects of palaeobiology; understanding these is paramount if we are to further elucidate the nature of the original organisms and their communities. Artefacts can be induced at all stages throughout the taphonomy of an organism, during its demise and throughout its burial and the diagenetic to metamorphic history of the rocks in which it is found. Such artefacts must be identified and excluded from interpretations of its morphology. Great strides are being made, with elegant experimental work (McIlroy et al. 2009, Darroch et al. 2012) enhancing detailed petrographic and field-based studies (e.g. Gehling 1999, Xiao et al. 2005, Grazhdankin et al. 2008, Laflamme et al. 2011).

Detailed petrographic work has constrained the timing of pyritisation in volcaniclastic settings of Avalonia to late metamorphic, after growth of metamorphic clays and micas. The mechanisms triggering the growth of similarly-sized pyrite growths with distinct textures (framboidal and concentric-ringed) is unknown, but may be revealed through high-resolution mapping of trace elements and/or sulphur isotopes. Phosphatisation of microbial mats and/or formation of authigenic clay minerals were instead likely key to soft tissue preservation in these settings. Seventeen years on from the proposal of the death mask model (Gehling 1999), perhaps it is time to consider how the plethora of biotas featuring pyritization relate to this model, and to one another.

44 However, the extent of the influence exerted by palaeoenvironment, and of the disparate factors this includes, remains uncertain. Perhaps the most appropriate way to investigate these biases is through expansive experimental work. The effects of a range of physical and chemical parameters have been tested in various taphonomic experiments, although these crucially have been limited to crown group plants and animals (see e.g. Briggs 2003, Sansom 2014). Such studies have yet to be systematically extended to investigate systems which would be more applicable to the Ediacaran, such as those with microbial mats (with the exception of (Darroch et al. 2012), and more accurately reflecting Ediacaran geochemical conditions (Chapter 1). Trace metals such as molybdenum, which are limiting nutrients for life (e.g. Glass et al. 2012), are becoming widely used as tracers of productivity and ocean redox conditions in the Proterozoic (e.g. Scott et al. 2008), but nothing is known about their specific effects on decay-related microbial activity, and therefore on taphonomy. Given the differences in ocean chemistry between the Ediacaran and Phanerozoic (Chapter 1), perhaps factors such as trace metal concentrations may have contributed to the abundance mouldic soft tissue preservation that is so characteristic of the Ediacaran.

A major caveat to such taphonomic experimentation is the enduring uncertainty surrounding the original biological composition of Ediacaran organisms, which will likely only be resolved upon discovery of an assemblage preserving cellular-level detail. Currently, inferences regarding relative degrees of robustness and rigidity may be made based on biostratinomic grounds, but these too are susceptible to variations in other biogeochemical conditions. Taphonomic experiments sampling a wide variety of tissue and cell types from as many branches of the tree of life as possible may provide our best hope: by comparing the behaviours of different biological compositions to features seen in fossils, it may be possible to relate the two, and consequently to infer the original composition of the organism or its parts. Of course, this must be repeated for the many potential variables already discussed in order for any such inferences to be made with any degree of confidence, rendering the number of experiments required unfeasible. Paradigm shifts in understanding will probably depend not on the development of analytical techniques, but rather on the discovery of new, higher-resolution preservational windows.

45 4. A PLETHORA OF PRIMOCANDELABRUM: MUTLIVARIATE STATISTICAL ANALYSIS AND THE CLASSIFICATION OF RANGEOMORPHS

4.1 Introduction

Given the uncertainty regarding the biology of rangeomorphs, their taxonomy and phylogeny are fraught with difficulty (Narbonne 2005, Brasier et al. 2012, Liu et al. 2015a). There are two main sources of characters: branching architecture (Chapter 1), and overall morphology. Growth programmes (e.g. insertion of new branches vs inflation of existing ones) are also useful taxonomic characters (Liu et al. 2015b), but only a few taxa are preserved with sufficient resolution to determine the number of branches present throughout ontogeny. Shape metrics (length to width ratios) and overall morphology have been traditionally used in taxonomic diagnoses (e.g. Laflamme et al. 2004, Laflamme & Narbonne 2008), whereas recent diagnoses have focused on branching architecture (Brasier & Antcliffe 2009, Narbonne et al. 2009, Brasier et al. 2012, Laflamme et al. 2012b, Liu et al. 2015b). However, different schools of thought place greater emphasis on one or the other group of characters, and even adopt different terminology to describe the same characters (Chapter 1, Table 1.1).

In current taxonomic framework(s), the relative weight applied to the various characters is, essentially, arbitrary and subjective. On the one hand, the comparative resilience of branching characters to tectonic overprinting (shearing and distortion), and their possible application to fragmentary fossils, makes them appear more attractive for taxonomy than continuous characters (Brasier et al. 2012). It has recently been argued that categorical characters (presence/absence of a holdfast, and details of branching architecture) should be used for genus-level diagnoses, and continuous characters (aspects of morphology, such as proportional stem length) should be restricted to species level diagnoses (Liu et al. 2015b). This recommendation was based on the assumption that branching architecture reflects genetic control, but shape metrics are more susceptible to environmental influence and, therefore, convergence – an assumption that rests on the apparent consistency of branching across multiple specimens (Brasier & Antcliffe 2009, Brasier et al. 2012). Accordingly, forms that differ in even a single branching character (e.g. Vinlandia vs. Beothukis) are classified as different genera (Brasier et al., 2012), with possibilities that they reflect different species within one genus, or even variation within a species, being rejected. However, the vast majority of taxonomic diagnoses make no mention of variation in branching architecture within the taxa, but it is not clear whether this reflects its absence or its lack of recognition. The only documented variation within a taxon are local displaying of typically rotated branches (Narbonne et al. 2009), interpreted as a taphonomic overprint, and local unfurling in Bradgatia (Brasier et al. 2012), which has been attributed to ontogeny.

46 The abundance of stalked, multifoliate (Chapter 1) individuals on Bed B in Charnwood Forest (Wilby et al., 2011; Fig. 4.1) provide an excellent opportunity to test various aspects of taxonomy. In this chapter, I investigate if and how branching (i.e. categorical) characters correlate with continuous morphological characters within these individuals, and analyse the differences between the groups that fall out using the two groups of characters separately and in conjunction. Through analysis of all specimens of the stalked, multifoliate rangeomorphs on Bed B, I test the validity of using single differences in branching character states to discriminate genera.

4.1.1 Charnwood Forest – a plethora of multifoliate taxa

The Mercian Assemblage as typified by the Charnwood Forest North Quarry Bed B biota (Wilby et al. 2011) is distinguished by its abundance of stalked and multifoliate forms (Fig. 4.1; Appendix 2.1). These forms consist of a holdfast disc, a discrete stem and a multifoliate (multi-branched) crown (terminology reviewed in Table 1.1). The sixty-four specimens belonging to the “multifoliate, stalked” morphology on Bed B fall into two main forms based on their overall appearance. Nine conform to the “dumbbell- like taxon” of Wilby et al. (2011; henceforth referred to as Orthiokaterna, Fig. 4.1a; Ch. 5), and fifty- two are referable to Primocandelabrum (Fig. 4.1b—i), a taxon first described from Newfoundland (Hofmann et al. 2008). Wilby et al. (2011) noted that there appeared to be multiple forms within the Primocandelabrum specimens on Bed B.

Orthiokaterna has a crown that is circular in outline and comprises upwards of five first order branches, which are gracile and sinuous in outline, and are entirely concealed along their lengths; a proportionally longer stem and large holdfast disc (Fig. 4.1a; Ch. 5); and displayed first order branches.

As currently defined (based on the Newfoundland specimens), Primocandelabrum is principally identified by the triangular outline of its crown and its coarse branches (Hofmann et al. 2008), with P. hiemaloranum distinguished by its “Hiemalora-like” holdfast, that is, having rays which radiate out from the disc. Newfoundland specimens typically preserve two main branches, but the poor preservation of the type specimens of Primocandelbrum makes it impossible to determine the branching architecture for the taxon, and so these characters cannot be used to identify the genus or constituent species. The Newfoundland specimens of Primocandelabrum are often reconstructed as a two-dimensional fan or candelabrum in life (Hofmann et al. 2008), while those from NW Canada have a more brush-like appearance (Narbonne et al. 2014). In new specimens from Charnwood Forest which fit the very broad description of Primocandelabrum, branching characters can be identified, and it is clear in the best- preserved specimens that the crown comprises three first order branches which split from the main stem (Fig. 4.1b—l). All new Primocandelabrum specimens have rotated and unfurled first order branches. The preservation of only two main branches is likely due to failure of the third branch to be expressed in

47 Figure 4.1. Stalked, multifoliate rangeomorphs from Bed B, Charnwood Forest. A) Orthiokaterna, (GSM105875); B-I) Individuals referable to Primocandelabrum, showing the variation present in terms of crown size (width and height), disc size, stem length and stem width relative to total height: B) “Big Bertha” (GSM 105872), C) 6A2b (GSM105969b), D) 10C3c (GSM106051c), E) 10B8 (GSM106049), F) 19A4 (GSM106040), G) 3(1) (GSM105945), H) 6A2d (GSM 105969d), I) 20Ba (GSM 106039); J, K) “dusters” 8C3 (GSM106045) and 8C1 (GSM106043), respectively. Scale bars are all 2 cm. All photographs are of plaster casts housed at the British Geological Survey.

48 the fossil, as it is held above the plane of preservation (see section 3.4.1), or possibly due to taphonomic processes such as effacement (Liu et al. 2011) and modern erosion. However, none of the specimens from Bed B which are referable to Primocandelabrum possess a “Hiemalora-like” holdfast disc, although an isolated disc with radiating rays is known from Bradgate Park in Charnwood Forest (Wilby et al. 2011). While it is often assumed that the rays represent a persistent biological feature (Dzik 2003, Hofmann et al. 2008), it is possible that they appear at a particular life stage, or that they occur only at a particular horizon in the sediment profile which is not captured on Bed B.

Two stalked, multifoliate specimens from Bed B (Fig. 4.1j, k) are referable to published specimens of “dusters” from Mistaken Point, a poorly-defined, bucket group for small, brush-like fronds, which have numerous (more than three) first order branches emanating from a shared point at the top of the stem, and a small or bulb-shaped holdfast (Mason and Narbonne, 2016; Clapham and Narbonne, 2002; Fig. 4.1j, k). “Dusters” are often assumed to dominantly comprise Primocandelabrum (Liu et al. 2015a, Mitchell et al. 2015; Mason and Narbonne, 2016), but this has not yet been rigorously tested. Two “dusters” have been recently described as Plumeropriscum hofmanni (Mason and Narbonne, 2016). The gross morphology, number of branches and overall appearance of Plumeropriscum are superficially similar to that of the “dusters” on Bed B, but are distinct from all specimens of Orthiokaterna, even those of comparable size. However, the branching architecture of Plumeropriscum has not been described, and the published photographs are too poor quality to determine it - consequently, this taxon cannot be meaningfully compared to any of the forms studied here. Nine stalked, multifoliate taxa are assigned to Undosyrus, but were excluded from these analyses because branching characters could not confidently be determined for the majority of specimens, see Chapter 6.

There is a considerable degree of variation in overall morphology and in branching architecture amongst the specimens assigned to Primocandelabrum sp., leading to the suggestion by Wilby et al. (2011) that several taxa or forms were represented. Following existing guidelines (Brasier et al. 2012) results in construction of at least 15 new genera for these 42 specimens (Table A.2.6). This number of genera (let alone species) appears unrealistic given the close similarity of the individuals in their overall morphology and appearance, and is hardly useful on a broader scale. By extension, how different do branching patterns have to be before they should be considered to discriminate different genera – are single character state differences reliable for generic-level taxonomy (as recommended by (Brasier et al. 2012, Liu et al. 2015b)? Both inter- and intra-specific variation of continuous characters could manifest in discrete morphologically-defined populations and/or continuous spectra. For instance, the two discrete populations in the stem length and number of primary branches (Liu et al. 2015b) have been used to distinguish Beothukis mistakensis from B. plumosa (formerly Culmofrons plumosa, Laflamme et al. 2012b), but the two species rarely co-occur, and the relationship of this character to

49 environmental parameters (Chapter 8) or community structure (Chapter 7) has not been determined. If categorical characters correlate with each other, and with continuous ones, then the groups can be considered to be morphologically distinct and are therefore more likely to be taxonomically meaningful than ones where the characters vary essentially randomly. If categorical characters discriminate genera and continuous characters discriminate species (Liu et al. 2015b), then clear groups should be identified based on categorical characters only, which could subsequently be subdivided based on their morphologic characters.

4.2 Statistical techniques – an objective approach to taxonomy?

Phylogenetic analysis has long relied on statistical and computational approaches in order to deal with the vast quantity of data involved, and taxonomic work has sporadically made use of similar techniques (Jardine & Sibson 1971, Dillon & Goldstein 1984, Burge & Zhukovsky 2013, Lajus et al. 2015). Whilst parsimony and, more recently, use of Bayesian information criterion to compare different models are typically used in phylogenetic analysis (Jin & Nei 1990, Yang 1996, Ronquist & Huelsenbeck 2003, Nylander et al. 2004), taxonomic workers have used dimensionality reduction techniques such as principal components analysis and multiple correspondence analysis, and clustering algorithms (Bonder et al. 2012, Dunn & Everitt 2004, Jardine & Sibson 1971, Pagni et al. 2013). The approach adopted here is to combine various clustering methods with multivariate analytical techniques. Various iterations of the data are run in order to determine the extent of influence of outliers and of particular groups of characters on the groupings which result. All tests were run in the R statistical package, version 4.1.1 (2014). Orthiokaterna, which can be distinguished from Primocandelabrum based on traditional techniques (based on all branching character and gross morphology), is included in order to test the validity of the statistical approaches.

4.2.1 Material Jesmonite resin casts of 64 specimens of multifoliate, stalked taxa from Bed B on Charnwood Forest. Original moulds taken from the bedding surface in Charnwood Forest are held at the BGS (Appendix 2.1).

4.2.2 Methods

Tests were run on all specimens which fit into the broad “multifoliate, stalked” morphospace for which both categorical and continuous characters could be recorded. Casts were studied under controlled lighting conditions using a microscope equipped with a camera lucida. Continuous characters (Fig. 4.2a) were measured by hand to the nearest millimetre using a ruler and recorded for each specimen (Appendix 2.4), as this allowed the specimen to be viewed under different lighting conditions. Up to second order branching characters (Fig. 4.2b; Chapter 1) can be discerned in 48 specimens (40 Primocandelabrum sp., 6 Orthiokaterna and 2 “dusters”) and were diagnosed and recorded for each of these specimens (Appendix

50 4.6). Branches are termed first, second and third order for simplicity (corresponding to folia, primary branches and secondary branches). Third order characters can only be discerned in a few individuals. The architecture of the first order branches of the “dusters” is uncertain, and so these observations were excluded from the main analyses. Identification of branching characters likely represents the largest source of primary error, due to taphonomic overprint and complexities arising from the compound nature of the fossils (see section 3.4.1). To minimise these, only characters that were consistent across the specimen were included. Characters for which states could not confidently be determined were left blank to minimise errors introduced due to misidentification.

The data were not retrodeformed prior to analysis. The lack of independent strain markers on the surface renders the holdfasts of the organisms the only structure available for strain analysis. For all multifoliate specimens, the correlation between disc height and disc width is even across the surface (R2 = 0.9915). Together with the fact that the holdfasts exhibit little variation in ellipticity (σ = 2.6%, Wilby et al. 2015), this indicates that shear is even across the surface, operated parallel to the long axis, and affected all specimens equally. If the holdfasts were originally circular (the common assumption in retrodeformation of Ediacaran fossils; Wood et al. 2003), then all fossils have been compacted to 77% of their original height (i.e. compaction of 23%), parallel to their long axis. While this will have affected the absolute proportions, it will have done so equally to all fossils studied, and so the relative proportions may be robustly compared. Some of the continuous characters are strongly correlated to one another. For instance, the length of the left and right branches are the same length in most specimens (Appendix 2.4), and also correlate to the height and/or width of the crown (depending on whether it is taller than it is wide). Similarly, the width of the stem measured at the inflection point and at the base are correlated: those specimens with narrower stems tend to be narrower along their entire length. Accordingly, one iteration is run a reduced character matrix, using only those characters whose proportions do not inherently depend upon one another (Appendix 1.1.1). This avoids any bias due to double-correlation of the analyses towards characters that are themselves strongly correlated (Dillon & Goldstein 1984).

Several aspects of the data make it challenging to work with, requiring various data treatments (Appendix 1.1). The presence of outlying individuals can influence the results of each of the methods detailed below (cf. Dillon & Goldstein 1984), and so several iterations of the data were run to investigate the extent of the influence of these factors (Appendix 1.1.1). This includes organisms of a significantly larger size than the main population (i.e. lacking ontogenetic intermediates) and which may have had a different life history (Chapter 8), the influence of which on organism morphology cannot be constrained. The variance of each continuous character was investigated using the makeProfilePlot function made available by Coghlan (2014; Fig. 4.3). Total size of the individual is a source of great variance in the data, and necessarily affects the absolute values of all other characters. Accordingly, all continuous characters were divided

51 by total height (taken from top of crown to base of stem, Fig. 4.2a) to standardise them (Appendix A.2.5). Since categorical characters (here, branching architecture) are either binary or comprise a small number of discrete categories, they have a proportionately large variance when compared to continuous characters. While some methods scale the data as a part of their algorithm, others require the data to be scaled first – this scales all values to unit variance, so that all characters influence the results equally (Appendix 1.1.2). The majority of the fossils included in this study are incomplete, or include characters which cannot be determined with confidence and which are left blank. This necessarily results in a dataset which has a relatively large amount of missing values, which are problematic for any statistical test. The missing values were imputed following guidelines for best practice (Josse et al. 2012, Josse & Husson 2012), discussed in detail in Appendix 1.1.4.

Hierarchical clustering following principal components analysis (Husson et al. 2010b; HCPC) was performed on the data (Appendix 1.3). The principal components analysis required by the method is discussed in Appendix 1.2: principal components analysis (PCA) is run on continuous characters; multiple factorial analysis (MFA) is run on categorical characters; and factorial analysis of mixed data (FAMD) is run on mixed datasets, i.e. those with both categorical and continuous characters. The influence of the iteration on the results was examined by determining the percentage of individuals which were assigned to the same group for each pair of similar iterations. For the hierarchical clustering analysis following principal components analysis, the success of assignment to each cluster based on branching characters was assessed through the percentage of individuals displaying a particular character assigned to the cluster described by that character (100% indicates that all individuals which display that character are placed in the cluster), and the percentage of individuals in the cluster which displayed that character (100% indicates that all individuals in the cluster display that character). Successful discrimination of a cluster was additionally visually assessed on the principal components space, with a high degree of overlap between different clusters indicating poor discrimination at least on the axes of greatest variance. The success of assignment to a cluster based on morphological characters was assessed by comparing the average value for each character within a cluster with the average value across all clusters, through bivariate plots of pairs of characters, and through visual inspection of profile plots and principal components space.

4.3 Results

Between iterations, the continuous characters significantly contributing to construction of the dimensions are similar for each iteration, but the categorical characters significantly contributing to the dimensions are more variable between iterations. For all tests including Orthiokaterna, the first dimension accounted for approximately 45% of the total variance, and the second dimension for approximately 17%. When Orthiokaterna was excluded, the first dimension of the principal components space accounted for

52 A

8

9 12 10 11 5

4 6 7

3 2 2 cm 1

B

1st order: 1st order: 2nd order: 3rd order: Rotated, unfurled, Displayed, unfurled, Displayed, furled, Displayed, distal inflation, distal inflation, proximal inflation, furled, no inflation, radiating, radiating, radiating, subparallel, unconcealed unconcealed concealed concealed

Figure 4.2. Continuous and categorical characters in rangeomorphs. A) Morphological Characters. 1: Disc width; 2: Disc height; 3: Width of stem at base; 4: Width of stem at inflection point; 5: Width of stem at base branches; 6: Height of stem to inflection point; 7: Height of stem to base branches; 8: Crown width (widest point); 9: Crown height; 10: Length of left branch; 11: Length of right branch; 12: Total height (from centre of disc/base of stem). Dashed white line: outline of fossil; Blue line: 1st order branch outline; Purple line: 2nd order branch outline; Pink line: 3rd order branch outline. B) Branching characters: Displayed/rotated: both rows/ one row of lower-order branches are visible. Furled/unfurled: branch edges are visible/tucked in, giving a scalloped/smooth outer margin. Inflation: shape of the branch increases distally/ proximally. Radiating/subparallel: branches emerge from central axis at increasing/similar angles. Concealed/unconcealed: central axis is visible or concealed (dashed lines) by the branches arising from it.

53 around 30% of the total variance, and the second dimension for around 25% for all tests. For the reduced character matrix, the percentage of variance described by each dimension was comparable to that for the full character matrix, but slightly higher when morphological characters only were considered (55% and 27% compared to around 40% and 20%). The influence of iteration on these aspects is also discussed, particularly with regard to the assignment of individuals within the group assigned to Primocandelabrum. This is quantified for the hierarchical cluster analyses as the percentage of individuals which are assigned to the same group for each pair of iterations. Significance refers to 95% confidence (i.e. p<0.05) throughout.

4.3.1 Hierarchical Clustering

All iterations using Orthiokaterna show the greatest jump in inertia gain (see Appendix A.1.3.1) in moving from two to three clusters (Fig. 4.4a, c, e). For most iterations, the next greatest jump is in moving from three to four clusters (Fig. 4.4a, e), but for branching characters only, the next greatest jump was between six and seven clusters. The characteristics for each cluster are listed in Tables 4.1 and 4.2, and are summarised below. Some variability was observed between the different iterations, which is explored in more detail below.

For most analyses where Orthiokaterna was included, cluster 1 (Fig. 4.4b, d, f; Table 4.1) was typified by having a slightly short and narrow stem, small disc and tall crown; first order branches which were rotated, furled, concealed and with proximal inflation, and second order branches which were rotated, unfurled, concealed, radiating and with median inflation; cluster 2 (Fig. 4.4b, d, f; Table 4.1) was typified by having a wide and short stem, a large and wide crown with long left and right-hand branches (Fig. 4.2a), first order branches which were unconcealed and second order branches which were concealed and with median inflation; cluster 3 (corresponding to Orthiokaterna; Fig. 4.4b, d, f; Table 4.1) was characterised by having a long stem, large disc and small crown, first order branches which were displayed, furled, concealed and with distal inflation, and second order branches which were displayed, furled and with proximal inflation (where the variables were included and either category of these variables was significantly correlated to the cluster, or the mean of the quantitative variables was significantly different within the cluster than for the group as whole, see Table 4.1).

For most analyses where Orthiokaterna was excluded and the individuals were forced into two or three clusters, cluster 1 was typified by having a long and slightly narrow stem (at inflection), slightly large disc and small crown compared to the other clusters, first order branches which were concealed and showed median inflation, and second order branches which had median inflation (Table 4.2; Figs. 4.5, 4.6); cluster 2 was typified by a large crown, slightly small disc and short, slightly narrow stem (all approximating the average values for each variable), first order branches which were unconcealed and

54 with proximal inflation, and second order branches with median inflation (Table 4.2; Figs. 4.5, 4.6; cluster 3 (where selected) was typified by a large disc, wide crown with long left and right branches and a slightly wide stem, and unconcealed first and second order branches with either distal or proximal inflation, and furled second order branches (Table 4.2; Figs. 4.5, 4.6). For the reduced character matrix, the same morphological characters described each cluster. Similar branching character variables described clusters 1 and 3, but the only branching characters which significantly described cluster 2 were proximal inflation of first order branches (Table 4.2).

4.3.1.1 Scaled vs non-scaled Because the values in the disjunctive table represent the degree to which a given missing value belongs to the two dummy variables (Appendix 1.1.2), all variables when scaled are necessarily treated as continuous categories. Therefore, a PCA was run on the standardised disjunctive table. The HCPC conducted on this iteration showed the largest jump between two and three clusters (the same as for the non-standardised data), with the next largest jump between five and six clusters.

4.3.1.2 All specimens vs min filt dataset The HCPC run with all specimens supported three clusters, with the largest increase in inertia gain between two and three groups. There was also an 81% match of group assignment for individuals between the two tests (Table 4.3), but no clear correlation between total number of missing data and mismatch (APPENDIX). The variables most strongly correlated to each cluster were similar for each dataset (Table 4.1), but with additional qualitative variables significantly correlated to the clusters formed from the filtered dataset (Table 4.1).

4.3.1.3 Discrimination of Orthiokaterna For all tests including Orthiokaterna, individuals assigned to this taxon (specimen numbers 6, 5C6, 3E1(1), 19A1, 3D7 and 6B2b; Table A.2.1) plot away from the main group (Fig 4.4), and comprise the second group when specimens are forced into two clusters (Table ?). Iterations including only branching characters and only morphological characters each place one Primocandelabrum specimen (6A1 and 4A1b(right), respectively) into this cluster (Fig. 4.4). The unfiltered dataset places one Primocandelabrum (3F2) into this category. Two Orthiokaterna individuals (3D7 and 5A4) do not cluster with the others in the iterations including only branching characters, placing in cluster 1 when outsize specimens are included, and cluster 2 when outsize specimens are excluded. This is not due to the number of missing data, as other individuals with the same number of missing values consistently plot with the rest of the Orthiokaterna specimens, but is rather due to their proportionally tall crowns and short stems, which place them closer to Primocandelabrum proportions for these characters (Table A.4.1). Additionally, 5A4 has unfurled second order branches and is the smallest specimen, and 3D7 is missing the furled second order character (Table A.4.1) which is a significant variable in defining the clusters (Table

55 Crown.width Crown.width A Crown.height..i. B Crown.height..i. Length.left..i. Disc.width Length.right..i. Stem.height..c.i. Disc.width at.inflection Disc.height Stem.height..c.i. Stem.height..c.bb. Stem.width.at.base at.inflection at.top.flare vectori vectori 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5

0 102030405060 0 102030405060

Index Index

Crown.width Crown.width C Crown.height..i. D Crown.height..i. Disc.width Length.left..i. Stem.height..c.i. Length.right..i. at.inflection Disc.width Disc.height Stem.height..c.i. Stem.height..c.bb. Stem.width.at.base at.inflection at.top.flare vectori vectori -2-101234 0.0 0.5 1.0 1.5

0 102030405060 0 1020304050

Index Index Figure 4.3. Profile plots, a graphic representation of the variance within each morphologic character. A) All specimens, all morphological characters, data not scaled. The two outsize specimens are at the right edge of the plot. The greatest variance is within “crown width” and “disc width” , “length left” and “length right”. B) as a) but on the reduced morpholgic character matrix: the difference in variance between “crown width” and “disc width” and other characters is more pronounced. C) as b) but scaled to unit variance. D) All Primocandelabrum and “dusters”, including outsize specimen, Big Bertha (right-most data points).

4.2). The reduced character matrix places 4A1b (right), 13A4 and Big Bertha (when included) into the Orthiokaterna group for iterations using only morphological characters (Fig. 4.4e, f).

4.3.1.4 Assignment of “dusters” When Orthiokaterna is included and only morphology is examined, 8C1 clusters most frequently with Orthiokaterna, and 8C3 clusters most frequently with Primocandelabrum cluster 2. However, when both branching characters and outsize specimens are included, both 8C1 and 8C3 place in the same cluster (Primocandelabrum cluster 2).

When Orthiokaterna is excluded, both 8C1 and 8C3 are placed in the same cluster (cluster 1) for the majority of analyses. Exceptions to this are for the iterations using only branching characters and excluding Big Bertha, and on the reduced character matrix when only morphological characters are

56 included, where the two dusters place into different clusters. In each case, this is likely due to the fact that two of their second order branching characters and the proportional length of their stems differ (Table 4.1). Even when placed in different clusters, they are still in close proximity on the factor map and trees constructed, indicating their close similarity (Fig. 4.5). However, for the reduced character matrix (excluding Orthiokaterna), this is not so, and they place in quite different parts of the factor map on far from each other on the trees (Fig. 4.6). Including “displayed first order branching” observations in the data matrix places 8C1 with Orthiokaterna (when those individuals are included), but not 8C3. For these iterations, when Orthiokaterna is included, both 8C1 and 8C3 form a separate group together with Big Bertha when four clusters are selected, but 8C3 at least falls in with other Primocandelabrum specimens when two clusters are selected (strongest support).

4.3.1.5 Removing outsize specimens When Orthiokaterna was included, two or three clusters were supported for all iterations except branching characters only, for which including outsize specimens supported two or six clusters and excluding them two or five, and excluding concealed data but including outsize specimens, which supported two or four clusters. There was a 90—100% (Table 4.3) match in group assignment of specimens for all tests except those iterations including Orthiokaterna and including all characters (81%; Table 4.3) and branching characters only (62%; Table 4.3). The latter test also showed an appreciable difference between characters significantly correlated to cluster 1 in particular, with more variables significantly correlated to this cluster when outsize specimens are included (Table 4.3). When Orthiokaterna was excluded, the number of supported clusters was only influenced by removing Big Bertha for iterations including all characters, which supported three clusters including Big Bertha, and two or four clusters excluding Big Bertha.

4.3.1.6 Assignment of Primocandelabrum individuals Primocandelabrum specimens + “dusters” are consistently distinguished from Orthiokaterna in most iterations of the data, but determining the number and composition of clusters within Primocandelabrums is more variable. This variation is discussed below with respect to each data treatment, and the variation in the definition of each cluster is summarised in Tables 4.1 and 4.2. Comparing iterations including and excluding Orthiokaterna, when outsize (i.e. of significantly larger size than the bulk population) and strongly outlying specimens are included or only morphological characters are considered, there is a very low match (around 50%) in cluster assignment of Primocandelabrum specimens (Tables 4.3, 4.4). When all characters or only branching characters are considered and outsize specimens are excluded, this increases to 74% and around 95% respectively (but 81% for reduced data matrix excluding outsize specimens; Tables 4.3, 4.4). When the Primocandelabrum specimens are divided into three or more clusters, disc size becomes a significant contributing factor to cluster definition (Table 4.2); when the data is split into two clusters, disc size is only significant when only morphological characters are used, or the analyses are run on the reduced character matrix (Table 4.2). Three clusters were most strongly 57 A B morphological and branching characters morphological and branching characters cluster 1 cluster 2 8C2 cluster 3

01234 3(1) 10C3c 10C3a inertia gain Big Bertha 19B1cluster 2 19C4 3(2) 6 13A8 6A1 6B2b 13A4 19A1 6a2 l 10A23D8 3D7cluster 3 4A38B4 5C6 10C3b2B1 8B3b Dim 2 (18.36%) 10B85.00E+031B1 5F5 5A4 3E1(1) 5C57A5 10A75C1 19A37A4cluster 13C3 10C85.00E+02 10B3 8C3 8C1 5G34B319A4 4A1b10b10 (right) 5F3 -20246 20Ba 01234 6 5F3 5F5 3(1) 3(2) 4A3 8B4 7A5 2B1 4B3 1B1 5A4 7A4 6A1 5C6 5C1 3D7 3C3 8C2 8C3 3D8 8C1 5C5 5G3 6a2 l 10A2 10B8 13A4 13A8 19A4 19A1 6B2b 19A3 10B3 20Ba 10A7 19B1 8B3b 10C8 19C4 10b10 10C3c 10C3b 10C3a 3E1(1) 5.00E+03 5.00E+02

Big Bertha -4-202468 4A1b (right) Dim 1 (41.89%)

C D branching characters only branching characters only cluster 1 cluster 1 cluster 2 cluster 3 10C3a cluster 4 cluster 3 cluster 2 cluster 5 8C2 cluster 6 0.00 0.10 0.20 0.30

inertia gain 8B4 19C4 10C3bBig 7A510A2Bertha4A3 cluster13A8 3(2) 1 19B13(1)2B16a21B1 l 6B2b5C619A1 10C3c cluster6 6 3E1(1) 5G35C55C1 13A4 6A1 19A3 5.00E+025F5 3D75A4 10C83D810A77A4cluster8B3b 2 10B819A45.00E+033C34B3 Dim 2 (12.08%) 10B3 cluster 4 4A1b (right) 10b10 8C1

5F3 cluster8C320Ba 5 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 6 -1.0 -0.5 0.0 0.5 1.0 5F3 5F5 3(2) 3(1) 7A4 2B1 1B1 5A4 7A5 4A3 8B4 4B3 6A1 3D8 5C1 5C6 3C3 8C3 5C5 8C2 3D7 8C1 5G3 6a2 l 10B8 19B1 13A4 8B3b 10A7 19A1 10B3 19A4 13A8 10A2 6B2b 20Ba 19A3 19C4 10C8 10b10 10C3c 10C3a 10C3b 3E1(1) -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 5.00E+03 5.00E+02 Big Bertha 4A1b (right) Dim 1 (43.92%)

E morphological characters only F morphological characters only cluster 1 cluster 2 8C2 cluster 3 Big Bertha 3(1)19B1 10C3c

0.0 0.5 1.0 1.5 19C4 3D8 cluster 2 3(2) inertia gain 6 6B2b 8C1 10C3a 19A1 13A4 13A8 6A1 6a2 l cluster3D7 1 8B3b5.00E+0310B8 10A2 10A7 19A3 8C3 4A3 2B1 5A44A1b (right) 5F5 5C6 10b105C1 10C3b 5C5 Dim 2 (26.63%) cluster1B1 3 3C3 7A4 5.00E+02 8B4 3E1(1) 20Ba 10C8 19A45F3 10B3 7A55G3

4B3 -2 -1 0 1 2 3 0.0 0.5 1.0 1.5 6 5F3 5F5 3(1) 3(2) 8B4 2B1 4B3 7A5 5A4 4A3 6A1 7A4 1B1 5C5 8C2 8C3 5C6 3D7 5C1 3D8 8C1 3C3 5G3 6a2 l 19A1 19B1 19A3 10A2 6B2b 20Ba 19A4 13A4 10B8 13A8 10B3 10A7 8B3b 10C8 19C4 10b10 10C3c 10C3a 10C3b 3E1(1) 5.00E+03 5.00E+02 Big Bertha -4 -2 0 2 4A1b (right) Dim 1 (55.39%)

Figure 4.4. Results of the cluster analysis, HCPC on the data set including all individuals for which branching and morphological characters could be determined. All values were standardised to total height. A) Cluster dendrogram and B) factor map for analysis on morphological characters (reduced character matrix) combined with branching characters; C) Cluster dendrogram and D) factor map for branching characters only; E) Cluster dendrogram and F) factor map for morpholgical characters (reduced character matrix) only. In A) Inertia gain supports division into two or three clsuters; in C) it supports division into 2 or 6 clusters, and in E) it supports 2 or 3 clusters. Orthiokaterna plots separately in A, B, C and D, but with other specimens in E and F.

58 supported by inertia gain for most analyses (Figs. 4.5, 4.6). Additionally, when individuals were divided into three rather than two clusters, there was a higher percentage of individuals that displayed a given character used to describe the cluster contained within the cluster, and a higher percentage of individuals within a given cluster displayed the characters used to describe that cluster (Table 4.2). When divided into more than three clusters, there was a similar or greater percentage of cluster and character match, but only one or two characters significantly described each cluster.

4.3.1.7 Removing “dusters” When Primocandelabrum individuals are divided into two groups and all characters are considered, there is 74% match between datasets including and excluding Orthiokaterna (excluding outsize specimens; Table 4.3), but a 92% match when dusters are excluded from the Primocandelabrum individuals (Table 4.3). When only branching characters are considered, this increases to 95% when dusters are included but falls to 59% when these two individuals are excluded (Table 4.3). For only morphological characters, there is only a 47% match when dusters are included, and 67% when they are excluded (Table 4.3). For the reduced dataset, there is an 81% match when dusters are included, and an 85% match when excluded for all characters, but only a 46% and a 52% match when only morphological characters are considered (Table 4.3). When individuals are divided into three clusters, Big Bertha is excluded, and all characters are considered, there is a 97% match between the analyses excluding and including “dusters”. The cluster match drops to 90% between the iterations excluding and including “dusters” when only morphological characters are considered, and to 85% when only branching characters are considered (Table 4.4). For the reduced character matrix and all characters are considered, there is a 92% match between iterations including and excluding “dusters”, and 82% between iterations including and excluding “dusters” when only morphological characters are considered (Table 4.4).

4.3.1.8 All characters versus morphologic characters versus branching characters only There is typically a low (60—70%) agreement in group assignment between iterations including all characters and those including only branching characters, although this is generally slightly higher (around 75%) when outsize specimens are excluded (Table 4.3). The match in group assignment between iterations using only morphological characters and only branching characters is similarly low (around 60%; Table 4.3). The agreement between iterations using all characters and those using only morphologic characters is higher (85—95%). When the dataset excluding Orthiokaterna is divided into three clusters, the match between all characters and branching only is around 57%, between branching only and morphological characters only around 43%, and between all and morphological characters only 69% when Big Bertha is included, but 86% when this specimen is excluded (Table 4.4). The reduced character matrix produces similar results, with a match between all characters and branching only (including and excluding Big Bertha, and excluding dusters as well as Big Bertha, respectively) of 55%, 60% and 49%, between all

59 A B morphological and branching characters morphological and branching characters

cluster 1 cluster 2 cluster 3 8C2 cluster 4 13A4 0.00.40.81.2 inertia gain

4A1b (right) 3(1)

10b10 5C1 19C4 cluster 4 cluster 2 13A8 8B3b 6A1 10C3c 5.00E+03 19B1 20Ba 5F5 19A3

Dim 2 (24.84%) 3C3 10A7 5.00E+02 3D8 19A4 3(2) 10C3a cluster 1 1B1 10B8 10A2 cluster 3 4A3 6a2 l 10B3 7A4 2B1 5F3 10C8 5C5

-2 0 2 4 7A5 4B3 5G3 10C3b 8B4 0.0 0.5 1.0 1.5 5F3 5F5 3(1) 3(2) 1B1 6A1 7A4 7A5 8B4 4A3 2B1 4B3 3C3 8C2 5C1 5C5 3D8 5G3 6a2 l 13A8 19A4 10A7 13A4 10B8 10B3 20Ba 10A2 19A3 8B3b 19B1 10C8 19C4 10b10 10C3c 10C3b 10C3a -6 -4 -2 0 2 4 5.00E+03 5.00E+02 4A1b (right) Dim 1 (29.85%)

C D branching characters only branching characters only

cluster 1 cluster 1 cluster 2 cluster 3 cluster 2 5F3 cluster 4 0.00 0.10 cluster 5 cluster 3 cluster 4 inertia gain cluster 6 20Ba

8B4 7A5 Dim 2 (15.45%) 10A2 4A3 10C8 19B1 3D85.00E+02 10C3b 10A7 10C3a 10C3c3(1) 5G3 7A4 5F5 10B3 cluster13A8 1 2B1 10B8 4A1b (right) cluster 2 5C15C5cluster 6 19C4 3(2) 3C3 19A4 1B1 6a2 l 4B3 5.00E+03 6A1 8C2 19A3cluster 5 8B3b cluster 3 10b10 -0.5 0.0 0.5 1.0 1.5 13A4 0.00 0.05 0.10 0.15 5F5 5F3 3(2) 3(1) 8B4 1B1 7A4 7A5 4A3 6A1 4B3 2B1 8C2 5C5 5C1 3D8 3C3 5G3 6a2 l 10B8 13A8 10A2 13A4 10B3 19A4 20Ba 8B3b 10A7 19B1 19A3 19C4 10C8 10b10 10C3c 10C3a 10C3b 5.00E+02 5.00E+03

4A1b (right) -1 0 1 2 Dim 1 (38.13%)

E morphological characters only F morphological characters only cluster 1 cluster 2 8C2 cluster 3 3(1) 0.0 0.4 0.8 1.2 inertia gain 19B1

cluster 2 10C3c 3D8 3(2) 10C3a 19C4 13A8 10B8 6a2 l 13A4 5.00E+03 6A1 8B3b 4A1b (right) 10A2 2B1 19A3 5F5 10A7 4A3 10C3b

Dim 2 (35.80%) 5C1 cluster 3 cluster 1 5C5 10b10 7A4 8B4 5.00E+02 1B1 20Ba 3C3 19A4 10B3 10C8 5F3 7A5 5G3

4B3 0.0 0.2 0.4 0.6 0.8 1.0 1.2 5F3 5F5 3(2) 3(1) 4B3 4A3 1B1 8B4 7A5 6A1 2B1 7A4 5C5 3D8 8C2 5C1 3C3 5G3 6a2 l -3 -2 -1 0 1 2 3 4 8B3b 10B3 19B1 19A4 13A8 19A3 10A2 10A7 20Ba 13A4 10B8 19C4 10C8 10b10 10C3c 10C3a 10C3b 5.00E+02 5.00E+03

4A1b (right) -4 -2 0 2 Dim 1 (40.03%)

Figure 4.5. Results of the cluster analysis, HCPC on the data set including all individuals for which branching and morphological characters could be determined, excluding Orthiokaterna and “dusters” and using the reduced character matrix. All values were standardised to total height. A) Cluster dendrogram and B) factor map for analysis on morphological and branching characters; C) Cluster dendrogram and D) factor map for branching characters only; E) Cluster dendrogram and F) factor map for morpholgical characters only. Inertia gain plots support division into 4 clusters in A), 2, 3 or 6 clusters in C), and 2 or 3 clusters in E).

60 and morphological characters of 88%, 81% and 87%, and between morphological and branching only of 57%, 60% and 44% (Table 4.4).

4.3.1.9 Reduced character matrix When Orthiokaterna was included, the match in group assignment between the full character matrix and the reduced character matrix was around 80%, both for all characters and for morphological characters only (Table 4.3). When Orthiokaterna was excluded and the individuals were split into three groups, there was around a 90% match in assignment from iterations including branching characters, when Big Bertha and dusters were included and excluded (Table 4.4). When only morphological characters were considered, the match was only around 50%, except when both dusters and Big Bertha were excluded, which resulted in an 83% match (Table 4.4). When Orthiokaterna was excluded and individuals were divided into two groups, there was only a 50%, 62% and 73% match when all characters were considered (including and excluding Big Bertha, and excluding Big Bertha and dusters respectively), and around a 90% match when only morphological characters were considered (all iterations; Table 4.4).

4.3.2 Assessment of cluster assignment

For each cluster, three factors describe the success of the cluster discrimination: 1) the percentage of assignment of individuals displaying a particular character which are assigned to a cluster characterised by that character; 2) the percentage of individuals within that cluster which display that character; 3) the average value for a continuous character within a cluster as opposed to across clusters. These are listed in Tables 4.1, 4.2. When Orthiokaterna is included, the within-cluster averages for this group are less similar than the within-cluster averages for the two Primocandelabrum (± “duster”) clusters. There is also a greater discrimination of Orthiokaterna individuals based on categorical characters, evidenced by more characters with 100% inclusion or exclusion (i.e. 0% inclusion) of individuals into clusters than there is for the two Primocandelabrum clusters (Tables 4.1, 4.2). When individuals are sorted into the clusters that they are assigned by the majority of iterations, the total variance within clusters is easily visualised through profile plots (Fig. 4.7a, b). Variance within clusters determined through the hierarchical cluster analysis is lower than across all individuals (Fig. 4.3), evidenced by smoother lines. When bivariate character plots are constructed with individuals are coloured according to their cluster number (Fig. 4.7c—f), the clusters follow distinct trendlines for those characters which are shown to provide the strongest discrimination of each cluster (Tables 4.1, 4.2), but do not form clear trendlines for those characters that provide weak discrimination of clusters.

4.3.3 Summary of results

All tests discriminated Orthiokaterna, placing these individuals close together even when one or two additional specimens are included in the cluster, or when the smallest Orthiokaterna individuals are placed in a different cluster. Separation of Orthiokaterna from Primocandelabrum is further supported by 61 A morphological and branching characters B morphological and branching characters

inertia gain cluster 1 cluster 2 Big Bertha cluster 3 8C2 8C1 0.0 0.5 1.0 1.5

13A4 3(1) 4A1b (right) cluster 3 19C4 19B1 10C3c 10b10 5C1 13A8 20Ba cluster 1 19A3 10C3a 5F5 5.00E+03

Dim 2 (23.77%) 10A7 8B3b 6A1 3C3 8C3 10A2 3(2) 5.00E+02 1B1 3D8 10B8 4A3 5F3 19A4 2B1 6a2 l cluster 2 10C3b 7A5 5C5 5G3 7A4 8B4 -2 0 210B3 4 10C8 4B3 0.0 0.5 1.0 1.5 5F5 5F3 3(1) 3(2) 7A4 2B1 7A5 1B1 8B4 6A1 4A3 4B3 8C1 8C3 3D8 8C2 5C5 3C3 5C1 5G3 6a2 l 13A4 10B8 19B1 8B3b 10A2 10A7 19A4 13A8 19A3 10B3 20Ba 19C4 10C8 10b10 10C3c 10C3b 10C3a 5.00E+02 5.00E+03 Big Bertha 4A1b (right) -4 -2 0 2 4 Dim 1 (27.79%)

C branching characters only D branching characters only

cluster 1 cluster 2 cluster 3 8C1 cluster 4 cluster 5 0.00 0.10 cluster 6 inertia gain 5F3 cluster 5 8C3 20Ba cluster 3

4A1b (right) 19A3 19B1 Dim 2 (14.93%) 10C3b 7A510A2 10b10 Big Bertha 4A3 10C3c 3D8 10C8 13A83(1)2B1 5C5 5.00E+02 8B4cluster 2 10A710B3 19A4 5G35C16a2 l 10B8 10C3a cluster 4 7A45.00E+03 19C4 3(2) 1B1 5F5 3C3 4B38B3b 6A1 cluster 1 8C2 cluster 6 -0.5 0.0 0.5 1.0 1.5 13A4 0.00 0.05 0.10 0.15 5F3 5F5 3(1) 3(2) 2B1 1B1 7A4 8B4 7A5 4A3 4B3 6A1 5C1 3C3 8C1 8C3 8C2 3D8 5C5 5G3 6a2 l 19A3 10A7 19A4 19B1 20Ba 13A8 13A4 10B8 10B3 8B3b 10A2 10C8 19C4 10b10 10C3c 10C3a 10C3b 5.00E+02 5.00E+03 Big Bertha 4A1b (right) -1 0 1 2 Dim 1 (36.48%)

E morphological characters only F morphological characters only

cluster 1 cluster 2 cluster 3 3(1) 19B1 8C2 cluster 2

0.0 0.4 0.8 Big Bertha inertia gain 3(2) 3D8 10C3c 10C3a 6a2 l 10B8 13A8 6A1 19C4 2B1 5.00E+03 cluster 1 8B3b 10C3b 10A2 4A3 5F5 13A4 5C5 19A3 Dim 2 (33.61%) 8B4 8C3 8C1 7A4 cluster 3 10A7 4A1b (right) 10B3 1B1 5C1 10C8 5.00E+02 10b10 3C3 5G3 19A4 5F3 7A5 20Ba -2 -1 04B3 1 2 3 0.0 0.2 0.4 0.6 0.8 1.0 5F3 5F5 3(1) 3(2) 8B4 4B3 6A1 4A3 7A4 2B1 1B1 7A5 3C3 8C3 8C1 5C1 3D8 5C5 8C2 5G3 6a2 l 8B3b 10B8 10B3 19A4 10A7 19A3 13A4 13A8 10A2 19B1 20Ba 19C4 10C8 10b10 10C3c 10C3a 10C3b 5.00E+03 5.00E+02 Big Bertha 4A1b (right) -2 -1 0 1 2 3 4 Dim 1 (42.91%)

Figure 4.6. Results of the cluster analysis, HCPC on the data set including all individuals for which branching and morphological characters could be determined, excluding Orthiokaterna and using the reduced character matrix. All values were standardised to total height. A) Cluster dendrogram and B) factor map for analysis on morphological and branching characters; C) Cluster dendrogram and D) factor map for branching characters only; E) Cluster dendrogram and F) factor map for morpholgical characters only. Inertia gain plots support division into 2 or 3 clusters in A), 2, 3 or 6 clusters in C), and 2, 3 or 5 clusters in E).

62 profile plots of variance (Fig. 4.7a, b) and bivariate plots (Fig. 4.7c–f). The morphological and branching characters of the “dusters” place one (8C1, Fig. 4.1k) most frequently with Orthiokaterna, and one (8C3, Fig. 4.1j) within Primocandelabrum clusters. The outsize specimens are placed into an isolated cluster based on all characters, morphological characters only, and for branching characters only. They also plot as outliers to the bulk population in the principal components space, especially for morphological characters only (Fig. 4.6e, f).

The clustering consistently supports placement of Primocandelabrum individuals into two or three clusters (by analysis of inertia gain) that are distinct from Orthiokaterna. However, within Primocandelabrum, there is considerable variation in the composition of each cluster (both individuals placed into the cluster and the character descriptions of the cluster) depending on data treatment and iteration. The highest percentage match of group assignment between iterations, including for branching characters only and morphologic characters only, was achieved for analyses run on the reduced character matrix. The lowest percentage match between iterations resulted from analyses using only branching characters. Many characters define more than one cluster; that is, individuals in multiple clusters share characters. Only one or two clusters contain 100% of the individuals that display a particular character state, and <100% of individuals in the cluster share all the same characters (Table 4.2). Within Primocandelabrum, no iteration produces a set of clusters within which all individuals are identical in terms of their branching and distinct from other clusters (without dividing the individuals into an unwieldy number of taxa). Importantly for this study, morphological and branching characters do not correlate perfectly, as they do for Orthiokaterna (Table 4.1). Profile plots of the variance within the data show a smoother, more consistent correlation of morphological traits when sorted into three groups as determined by cluster analysis than when the individuals are unsorted (Fig. 4.7a, b).

4.4 Discussion

The argument that generic-level diagnoses should be based on categorical characters such as branching architecture, and that continuous characters be relegated to species-level diagnoses, rests on the assumption that branching is genetically controlled whereas morphological characters such as possession of a stem are more susceptible to environmental influence. If aspects of branching architecture can be influenced by environment, or if they provide some functional benefit, then this assumption breaks down – traits that provide an advantage through their functionality are more likely to be converged upon. The main advantage of the statistical techniques outlined here is that they allow characterisation of clusters of specimens based on all characters and on subsets of characters, and that they weight all characters equally. They also provide an efficient way to handle large numbers of specimens with subtly different character states, such as the Primocandelabrum individuals on Bed B, which become unmanageable using traditional techniques.

63 Length right 0.77 (0.64) 0.36 (0.64) 0.79 (0.68) 0.31 (0.68) 0.81 (0.68) 0.32 (0.68) 0.41 (0.68) 0.82 (0.68) 0.36 (0.68) 0.83 (0.68) Length left 0.82 (0.65) 0.36 (0.65) 0.83 (0.69) 0.30 (0.69) 0.89 (0.70) 0.32 (0.70) 0.39 (0.69) 0.89 (0.69) 0.36 (0.70) 0.89 (0.70) Crown height 0.72 (0.65) 0.44 (0.65) 0.73 (0.67) 0.40 (0.67) 0.76 (0.67) 0.41 (0.67) 0.45 (0.67) 0.72 (0.67) 0.78 (0.67) 0.43 (0.68) 0.72 (0.68) 0.78 (0.68) 0.72 (0.67) 0.40 (0.67) 0.76 (0.68) 0.40 (0.68) 0.45 (0.67) 0.73 (0.67) 0.44 (0.68) 0.73 (0.68) 0.75 (0.68) Crown width 0.45 (0.91) 1.23 (0.97) 0.39 (0.97) 1.28 (0.97) 0.40 (0.97) 0.56 (0.97) 1.31 (0.97) 0.46 (0.97) 1.31 (0.97) 1.35 (0.97) 0.39 (0.97) 1.28 (0.97) 0.37 (0.97) 0.59 (0.97) 1.38 (0.97) 0.52 (0.97) 1.37 (0.97) Stem width at inflection 0.13 (0.09) 0.07 (0.09) 0.14 (0.10) 0.14 (0.10) 0.07 (0.10) 0.08 (0.10) 0.15 (0.10) 0.08 (0.10) 0.15 (0.10) 0.15 (0.10) 0.08 (0.10) 0.14 (0.10) 0.07 (0.10) 0.08 (0.10) 0.15 (0.10) 0.08 (0.10) 0.15 (0.10) Stem width at top flare 0.27 (0.20) 0.04 (0.20) 0.32 (0.22) 0.27 (0.22) 0.07 (0.22) 0.36 (0.27) 0.37 (0.27) 0.23 (0.27) Stem width at base 0.14 (0.09) 0.08 (0.09) 0.04 (0.09) 0.08 (0.11) 0.18 (0.11) 0.03 (0.10) 0.16 (0.11) 0.08 (0.11) 0.06 (0.11) 0.08 (0.12) 0.19 (0.12) 0.08 (0.12) 0.19 (0.12) Stem height (c-bb) 0.37 (0.41) 0.40 (0.46) 0.65 (0.46) 0.38 (0.45) 0.67 (0.45) 0.65 (0.45) 0.41 (0.46) 0.67 (0.45) 0.41 (0.45) Stem height (c-i) 0.22 (0.31) 0.57 (0.31) 0.26 (0.32) 0.60 (0.32) 0.23 (0.31) 0.59 (0.31) 0.56 (0.32) 0.26 (0.32) 0.58 (0.32) 0.26 (0.31) 0.27 (0.32) 0.60 (0.32) 0.23 (0.31) 0.60 (0.31) 0.55 (0.32) 0.25 (0.32) 0.56 (0.32) 0.25 (0.32) Disc height 0.22 (0.27) 0.42 (0.27) 0.20 (0.27) 0.33 (0.27) 0.44 (0.27) 0.21 (0.26) 0.41 (0.26) 0.46 (0.28) 0.21 (0.28) 0.42 (0.27) 0.21 (0.27) Disc width 0.28 (0.34) 0.50 (0.34) 0.27 (0.35) 0.43 (0.35) 0.51 (0.35) 0.28 (0.33) 0.48 (0.33) 0.55 (0.36) 0.28 (0.36) 0.49 (0.34) 0.28 (0.34) 0.29 (0.35) 0.51 (0.35) 0.29 (0.33) 0.48 (0.33) 0.53 (0.36) 0.28 (0.36) 0.47 (0.34) 0.28 (0.34) fm brd(1) mf nbored mo mf mo (2) nbored fm brd(3) mf nbored mo all (1) all (2) all (3) mf (1) mf (2) mf (3) mfnbo (1) mfnbo (2) mfnbo (3) mf br (1) mf br (2) mf br (3) mf br nbb (1) mf br nbb (2) mf br nbb (3) mf mo (1) mf mo (2) mf mo (3) mf mo nbo (1) mf mo nbo (2) mf mo nbo (3) mf red (1) mf red (2) mf red (3) mf nbored (1) mf nbored (2) mf nbb red (3) mf mo red (1) mf mo red (2) mf mo red (3) Iteration (cluster number) Continued overleaf.

64 2nd (c) concealed or unconcealed (uc) c(35,57) re the average for all the average for re ter that are placed in the that are ter chical clustering analysis. taset (speicmens for which taset (speicmens for 2nd Inflation: proximal (p), median (m) or distal (d) m(78,44) m(76,48) p(60,43); m(0,0) m(76,57) m(41,100) m(0,0) p(60,50) m(24,19) m(76,61) m(6,8) m(94,53) p(60,43); m(0,0) m(12,12) m(88,63); p(0,0) 2nd radiatingor subparallel (s) (r) r(55,30) r(30,100) r(26,24) r(26,100) f(67,67); u(4,17) 2nd furled (f) or unfurled (u) f(63,45) f(14,4) f(71,71); u(4,14) f(67,57) f(14,3) f(86,100); u(0,0) f(0,0) f(83,100); u(0,0) u(75,70); f(0,0) f(71,71); u(4,14) f(0,0) 2nd displayed (d) or rotated (r) d(24, 100) r(100,57) r(0,0) d(24,100) d(22,100) 1st concealed or unconcealed uc(79, 71); c(7,10) uc(16, 9) c(32,82) c(16,27) uc(50,60); c(28,100); uc(0,0) uc(78, 82); c(8,12) c(67, 70); uc(17,13) c(60,42) c(0,0) uc(100,86); uc(0,0) c(83,95); c(0,0) uc(50,75); uc(0,0) c(28,100); c(4,6) uc(78,82); uc(22,17) c(71,71); c(25,100); uc(0,0) 1st Inflation: proximal(p), median (m) or distal(d) p(0,0) p(75, 89) p(0,0) p(59,83) p(0,0) m(100,57) p(91,81) p(0,0) p(0,0) p(78,83) p(0,0) 1st furled (f) 1st furled or unfurled (u) f(100, 27) u(8,43) f(100,43); u(59,91) f(100,43); u(8,43) u(82,86); f(0,0) f(100,50); u(5,33) f(100,60); u(3,20) f(100,43); u(8,43) f(100,50); u(6,33) 1st (d) Displayed or rotated (r) d(0,0) d(100,55), r(0,0) r(70,48); d(0,0) d(100,86); r(0,0) d(0,0) d(100,71); r(0,0) r(85,78); d(17,3) d(83,83); r(0,0) r(55,86); d(0,0) d(80,80); r(0,0) r(76,83); d(0,0) r(0,0) d(100,86); d(0,0) d(100,83); r(0,0) all (1) all (2) all (3) mf (1) mf (2) mf (3) mfnbo (1) mfnbo (2) mfnbo (3) mf br (1) mf br (2) mf br (3) mf br nbo (1) mf br nbo (2) mf br nbo (3) mf mo (1) mf mo (2) mf mo (3) mf mo nbo (1) mf mo nbo (2) mf mo nbo (3) mf red (1) mf red (2) mf red (3) mf nbo red (1) mf nbo red (2) mf nbo red (3) trto (cluster Iteration number) First values are the average for the cluster (number in bracket of row name); for quantitative characters, values in brackets a name); for in bracket of row (number the cluster the average for First values are of individuals with that charac (first value) the percentage qualitative characters, the numbers in brackets are specimens; for da mf = filtered that display the given character. of individuals within the cluster and (second value) the percentage cluster, Table 4.1. Quantitative (morphological) and qualitative (branching) variables categorising the clusters as determined by heirar Table

65 0.62 (0.74) 0.88 (0.74) 0.62 (0.73) 0.86 (0.73) 0.65 (0.75) 0.86 (0.75) 0.88 (0.74) 0.65 (0.74) 0.63 (0.73) 0.88 (0.73) 0.65 (0.75) 0.88 (0.75) 0.60 (0.75) 0.95 (0.75) 0.61 (.076) 0.94 (0.76) 0.63 (0.77) 0.94 (0.77) 0.95 (0.75) 0.58 (0.75) 0.60 (0.76) 0.97 (0.76) 0.63 (0.77) 0.97 (0.77) Length left Length right 0.64 (0.72) 0.78 (0.72) 0.64 (0.72) 0.78 (0.72) 0.63 (0.73) 0.77 (0.73) 0.77 (0.72) 0.76 (0.72) 0.59 (0.72) 0.63 (0.72) 0.77 (0.72) 0.64 (0.73) 0.80 (0.72) 0.63 (0.72) 0.79 (0.72) 0.79 (0.72) 0.62 (0.72) 0.64 (0.72) 0.79 (0.72) 0.64 (0.73) 0.80 (0.73) 0.79 (0.72) 0.62 (0.72) 0.63 (0.73) 0.80 (0.72) 0.8 (1.07) 1.52 (1.07) 0.80 (1.06) 1.38 (1.06) 0.84 (1.08) 1.38 (1.08) 1.40 (1.07) 0.88 (1.07) 0.81 (1.06) 1.40 (1.06) 0.85 (1.08) 1.40 (1.08) 0.81 (1.07) 1.41 (1.07) 1.62 (1.07) 0.84 (1.07) 0.80 (1.06) 1.34 (1.06) 0.90 (1.08) 1.39 (1.08) 1.70 (1.06) 0.84 (1.06) 0.84 (1.08) 1.38 (1.08) Crown width Crown height 0.08 (0.10) 0.07 (0.10) 0.16 (0.1) 0.06 (0.10) 0.16 (0.10) 0.15 (0.10) 0.08 (0.10) 0.08 (0.10) 0.15 (0.10) 0.08 (0.10) 0.16 (0.12) 0.07 (0.10) 0.15 (0.10) 0.06 (0.10) 0.16 (0.10) 0.07 (0.10) 0.16 (0.10) 0.07 (0.10) 0.16 (0.10) Stem width at inflection 0.22 (0.25) 0.35 (0.25) 0.21 (0.25) 0.33 (0.25) 0.33 (0.24) 0.22 (0.26) 0.21 (0.26) 0.21 (0.25) 0.30 (0.25) Stem width at top flare 0.18 (0.12) 0.17 (0.12) 0.08 (0.11) 0.17 (0.11) 0.18 (0.12) 0.09 (0.12) 0.09 (0.12) 0.19 (0.12) 0.09 (0.12) 0.19 (0.12) Stem width at base 0.53 (0.42) 0.34 (0.42) 0.54 (0.42) 0.33 (0.42) 0.52 (0.41) 0.34 (0.41) 0.36 (0.42) 0.56 (0.42) 0.54 (0.42) 0.34 (0.42) 0.51 (0.41) 0.33 (0.41) (c-bb) 0.36 (0.27) 0.21 (0.27) 0.36 (0.27) 0.21 (0.27) 0.37 (0.26) 0.21 (0.26) 0.23 (0.28) 0.39 (0.28) 0.37 (0.27) 0.22 (0.27) 0.35 (0.27) 0.20 (0.27) 0.37 (0.27) 0.21 (0.27) 0.21 (0.28) 0.38 (0.28) 0.37 (0.27) 0.21 (0.27) 0.36 (0.26) 0.19 (0.26) 0.21 (0.27) 0.38 (0.27) 0.37 (0.27) 0.20 (0.27) (c-i) 0.20 (0.25) 0.40 (0.25) 0.20 (0.24) 0.29 (0.24) 0.19 (0.23) 0.29 (0.23) 0.19 (0.26) 0.34 (0.26) 0.29 (0.25) 0.19 (0.25) 0.20 (0.24) Disc height Stem height 0.26 (0.33) 0.54 (0.33) 0.26 (0.31) 0.39 (0.31) 0.25 (0.31) 0.39 (0.31) 0.26 (0.33) 0.42 (0.33) 0.25 (0.32) 0.25 (0.31) 0.38 (0.31) 0.25 (0.33) 0.48 (0.33) 0.27 (0.33) 0.64 (0.33) 0.25 (0.31) 0.39 (0.31) 0.25 (0.31) 0.39 (0.31) 0.27 (0.32) 0.56 (0.32) 0.38 (0.31) 0.26 (0.31) Disc width Stem height all (1) all (2) all (3) nbb (1) nbb (2) nbb (3) nbb nd (1) nbb nd (2) nbb nd (3) mo (1) mo (2) mo (3) mo nbb (1) mo nbb (2) mo nbb (3) mo nbbnd (1) mo nbbnd (2) mo nbbnd (3) red (1) red (2) red (3) mo red (1) mo red (2) mo red (3) red nbb (1) red nbb (2) red nbb (3) red nd nbb (1) red nd nbb (2) red nd nbb (3) mo nbb red (1) mo nbb red (2) mo nbb red (3) mo nbb nd red (1) mo nbb nd red (2) mo nbb nd red (3) Iteration (cluster number) Continued overleaf.

66 uc (100,33) uc (100,22) uc (100,22) uc (100,25) c (14,40) uc (100,22) 2nd (c) concealed or unconcealed (uc) termined. Quantitative: chical clustering analysis. All chical clustering analysis. ens; Qualitative: first number ens; Qualitative: first number m(0,0) m(53,64) m(0,0) m(82,74) m(82,74) m(18,15) m (82, 74) m(13,13) m(73,58) m (53, 69) m(0,0) m(53,75) m(0,0) m(0,0) 2nd Inflation: proximal (p), median (m) or distal (d) rs” removed; mo = morphological rs” removed; he second number is percentage of is percentage he second number 2nd radiatingor subparallel (s) (r) f (100) f (100,100) 2nd furled (f) 2nd (f) furled or unfurled (u) d (60,71) 2nd displayed (d) or rotated (r) c (56,67); uc (17,20) c (56,71); uc (11,14) uc (39,78); c (6,11) c (53,82); uc (0,0) uc (41,78); c (6,11) uc (100,95); c (0,0) c (100,95); uc (6,5) uc (94,85); c (0,0) c (100,95); uc (6, 5) uc (94,100); c (0,0) c (88,79); uc (6,5) c (56,77); uc (6,8) uc (39, 88); c (0,0) c (56,83); uc (6,8) uc (44,80); c (0,0) c (65,85); uc (0,0) uc (47,89); c (0,0) 1st concealed or unconcealed m (100, 27) m (100,29) p (53,94) p (16,56); m (0,0) m (100,21) m (100,21) m (100,50); p (3,25) m (100,31) p (59, 90) m (100,33) p (56,95) p (16,56) 1st Inflation: proximal(p), median (m) or distal(d) 1st furled (f) 1st furled or unfurled (u) 1st (d) Displayed or rotated (r) all (2) all (3) nbb (1) nbb (2) nbb (3) nbbnd(1) nbbnd(2) nbbnd(3) br (1) br (2) br (3) br nbb (1) br nbb (2) br nbb (3) br nbbnd (1) br nbbnd (2) br nbbnd (3) red (1) red (2) red (3) red nbb (1) red nbb (2) red nbb (3) redndnbb(1) redndnbb(2) redndnbb(3) trto (cluster Iteration number) analyses excluding Orthiokaterna and only including specimens for which both morphological and branching characters could be de analyses excluding Orthiokaterna and only including specimens for all specim the average for name); values in brackets are in bracket of row (number the cluster the average for first values are (not included imputed, i.e. originally missing, values), and t with specified character of individuals in cluster is percentage nd nbb = outsize specimens and “duste nbb = outsize specimens removed; included in the cluster. individuals with that character matrix. character = reduced = branching characters only; red characters only analysed; br Table 4.2. Quantitative (morphological) and qualitative (branching) variables categorising the clusters as determined by heirar Table

67 90.48 46.51 po mo red nbb po red nbb 61.90 81.40 po mo red nbbnd 51.28 po red nbbnd 73.81 88.10 84.62 po mo red 90.48 43.18 nly individuals for which morphologial and nly individuals for into two groups two permit comparison with into two groups a excluded; nbb = outsize specimen Big Bertha ocandelabrum and “dusters”). Abbreviations: Abbreviations: ocandelabrum and “dusters”). po red 50.00 93.18 mo red nbo 83.33 51.28 46.51 red nbo 85.42 84.62 81.40 mo red 82.00 43.18 red 78.00 93.18 58.97 po br nbbnd po br nbb 95.35 75.61 68.29 97.56 59.09 69.05 69.05 97.56 po br po mo nbbnd 66.67 88.10 46.51 87.80 100.00 68.29 90.48 po mo nbb 45.45 95.24 100.00 69.05 90.48 po mo 92.31 73.81 po nbb nd 74.42 92.68 87.80 75.61 61.90 po nbb 47.73 92.68 95.24 69.05 50.00 po 85.11 59.57 100.00 46.51 66.67 83.33 mon bo 91.84 57.14 45.45 82.00 100.00 mo 74.47 61.70 59.57 95.35 58.97 brnbo br 63.27 61.70 57.14 59.09 80.85 74.47 85.11 74.42 92.31 85.42 nbo all 80.85 63.27 91.84 47.73 78.00 Iteration all nbo br brnbo mo monbo po po nbb po nbb nd po mo po mo nbb po mo nbbnd po br po br nbb po br nbbnd red mo red red nbo mo red nbo po red po mo red po red nbbnd po mo red nbbnd po red nbb po mo red nbb branching characters could be determined were included in each iteration. Analyses excluding Orthiokaterna (“po”) were divided Analyses excluding Orthiokaterna (“po”) were included in each iteration. branching characters could be determined were comprising Prim two groups with the other those analyses including Orthiokaterna (which typically separates into its own group, = branching characters only; po Orthiokatern mo = morphological characters only analysed; br nbo = outsize specimens removed; matrix. character = reduced excluded; nbbnd= Big Bertha and “dusters” red Table 4.3. Comparing percentage of individuals placed in the same group (percentage group match) between pairs of iterations. O group (percentage of individuals placed in the same group 4.3. Comparing percentage Table 68 Iteration po po po po mo po mo po mo po br po br po br po po mo po red po mo red po red po mo nbb nbb nd nbb nbbnd nbb nbbndred red nbbnd nbbnd nbb red nbb po 90.2487.18 69.05 57.14 88.10 po nbb 90.24 97.44 85.37 56.10 90.24 po nbb nd 87.1897.44 87.18 51.28 92.86 po mo 69.05 90.24 79.49 42.86 45.24 po mo nbb 85.37 90.24 89.74 46.34 53.66 po mo nbbnd 87.18 79.49 89.74 48.72 83.33 po br 57.14 42.86 100.00 54.7657.14 po br nbb 56.10 46.34 100.0084.62 84.62 59.52 po br nbbnd 51.28 48.72 84.62 84.62 48.72 43.59 po red 88.10 54.76 88.10 89.74 92.68 po mo red 45.24 57.14 88.10 82.05 100.00 po red nbb nd 92.86 48.72 89.7482.05 92.31 87.18 po mo red nbb nd 83.33 87.18 po red nbb 59.52 92.68 92.31 87.18 80.95 po mo red nbb 90.24 53.66 43.59 100.00 87.18 80.95 Table 4.4. Comparing percentage of individuals placed in the same group (percentage group match) between pairs of iterations excluding Orthiokaterna. Only individuals for which morphologial and branching characters could be determined were included in each iteration. All iterations were divided into three clusters, as this was supported by inertia gain and by cluster descriptions for most iterations. Abbreviations: po = Orthiokaterna excluded; nbb = outsize specimen Big Bertha excluded; nbbnd= Big Bertha and “dusters” excluded; mo = morphological characters only analysed; br = branching characters only; red = reduced character matrix. 4.4.1 The potential functional significance of rangeomorph branches

Furling of branches is observed in many rangeomorph taxa (Brasier et al. 2012) and is seemingly at odds with the presumed function of rangeomorph elements as exchange surfaces, as it decreases the surface area exposed to the water column (Laflamme et al. 2009, Liu et al. 2015a). However, there are several conceivable advantages that might have been bestowed upon a rangeomorph by possession of furled branches, particularly for higher order branches. First, it allows for tighter packing of branches, and minimizing overlap of rangeomorph elements. Secondly, it may have served to protect the tips (the likely site of growth, Antcliffe & Brasier 2007) from abrasion by neighbouring branches and from environmental damage (discussed further in Chapter 5, 8). Thirdly, the flow of water within a furl might be slowed, increasing its contact time with the rangeomorph surface – contrastingly, flow may be increased by the surface texture of the furl, as for the riblets on shark skin (Dean & Bhushan 2010). In contrast, possession of unfurled rangeomorph elements should serve to increase surface roughness of the rangeomorph exchange surface, reducing the thickness of the diffusive boundary layer over the surface and enhancing uptake of nutrients and/or removal of waste products (Chapter 6). Therefore, if these characters are functionally important, it might be expected that a trade-off between increased surface roughness and damage prevention would determine whether a rangeomorph adopted a furled or unfurled branching architecture.

The functional advantage of certain aspects of branching, such as proximal vs median vs distal inflation, is not readily apparent. It appears to be a function of the rate of insertion relative to inflation, and the inflation of individual branches relative to each other (R. Hoekzema, pers. comm.). Maintaining

69 A B Crown.width Crown.width Crown.height..i. Crown.height..i. Length.left..i. Length.left..i. Length.right..i. Length.right..i. Disc.width Disc.width Disc.height Disc.height Stem.height..c.i. Stem.height..c.i. Stem.height..c.bb. Stem.height..c.bb. Stem.width.at.base Stem.width.at.base at.inflection at.inflection Primocandelabrum at.top.flare at.top.flare cluster 3

Primocandelabrum O. fordi O. fordi Primocandelabrum cluster 2 cluster 1 vectori vectori 0 102030405060 0 102030405060 “dusters” “dusters”

0 1020304050 0 1020304050 Index Index

C D Orthiokaterna fordi Orthiokaterna fordi Primocandelabrum cluster 1 Primocandelabrum cluster 1 Primocandelabrum cluster 2 Primocandelabrum cluster 2 Primocandelabrum cluster 3 Primocandelabrum cluster 3 “dusters” “dusters” crown width total height 10 20 30 40 0 102030405060

0 10203040 10 20 30 40 disc width total height

E F Orthiokaterna fordi Orthiokaterna fordi Primocandelabrum cluster 1 Primocandelabrum cluster 1 Primocandelabrum cluster 2 Primocandelabrum cluster 2 Primocandelabrum cluster 3 Primocandelabrum cluster 3 “dusters” “dusters” stem width stem length 1234 0 5 10 15 20

10 20 30 40 0 102030405060 total height crown width

Figure 4.7. Profile plots and scatter plots of Orthiokaterna and Primocandelabrum. A) Profile plot of all individuals order by height and cluster; B) Profile plot with individuals ordered by group determined through cluster analyses and then by height. C—F) Simple biplots of all individuals within the analysed dataset (thopse for whom both branching and morphological characters could be determined. Primocandelabrum cluster 1, with proportionally long stems; Primocandelabrum cluster 2, with short stems (and unconcealed folia), and Primocandelabrum cluster 3, with wide crowns, and large discs. 70 a consistent mode of inflation may have served to reduce self-shading (cf. Enríquez & Pantoja-Reyes 2005), while the distal inflation in Bradgatia creates a hyperbolic surface that would have increased the surface area to volume ratio (R. Hoekzema, pers. comm.). Other aspects, such as rotated vs. displayed branches and concealed vs. unconcealed axes, may be implicated in the efficiency of branch packing. Given the current uncertainties regarding the functional constraints on branching architecture, and consequent possibilities of convergence, it is perhaps premature to place great phylogenetic relevance on such characters.

4.4.2 Influence of taphonomy

Any analysis of fossil morphology must take into account the biases that may be imposed by taphonomic processes Chapter 3. Taphonomic interference with branching architecture is described in Chapter 3, and includes local rotation of branches (Narbonne et al. 2009), deflation (Brasier et al. 2013), as well as displacement of branch order of overlap for unconstrained branches. That only branching characters that were consistent across the specimen were included in the analyses presented in this chapter should minimise the influence of taphonomy on branching characters in this study.

As discussed in Chapter 3, the overall morphology of a rangeomorph (i.e. continuous characters) is potentially susceptible to several sources of taphonomic overprint. For example, the preserved size of the holdfast and its number of concentric rings may be influenced by burial depth in the sediment, stem length may be increased through the stretching that generates torsion lines in the longest stems (section 5.5.2), and stem width may be obscured by sediment settling beneath the stem during felling. The shape of the crown, which was three-dimensional in life but is preserved as a compound, two-dimensional impression, was also likely to have been influenced during felling, depending on the stiffness and rheology of the branches (Chapter 3, 5).

One point of particular note is the possible introduction of taphospecies through inclusion of “iveshediomorphed” individuals. This term describes individuals where the organism has recently died and begun to decay (Liu et al. 2011; or alternatively where the frond is partially held above the plane of preservation; Wilby et al. 2011), but where the fossil is still identifiable as a described taxon – it is not yet a full “iveshediomorph” (Chapter 3, cf. Liu et al. 2011). The decay process has been demonstrated to influence the size and shape of the preserved remnants, through development of a globular mass of extracellular polysaccharides (Liu et al. 2011). Use of only individuals for which rangeomorph branching architecture can be determined minimises this risk, and so the preserved morphology included in these analyses can be assumed not to reflect taphonomically-induced proportions.

71 4.4.3 Validity of the statistical approach

The group of individuals assigned to Orthiokaterna is more different to the group containing Primocandelabrum than that group is within itself – the reason for placing them into a separate taxon using traditional taxonomic methods. Using the statistical techniques, Orthiokaterna is placed into an isolated cluster based on all characters, morphological characters only and branching characters only. That Orthiokaterna is discriminated for all of these iterations indicates that categorical character states correlate consistently with particular values for the continuous characters. Crucially, for the categorical characters that statistically significantly describe the Orthiokaterna cluster, a high proportion of the individuals within that cluster display a given character state, and a high proportion of the individuals that display that character state are assigned to the cluster (Table 4.3). Additionally, the mean values of the continuous characters for Orthiokaterna are statistically significantly different from the means of the continuous characters describing the Primocandelabrum clusters, and are more different to the means of the Primocandelabrum clusters than are the means within the Primocandelabrum clusters. This is strong support for these individuals representing a distinct and separate group, and indicates that the statistical approach described here is useful in discriminating taxonomic groups in mixed populations.

4.4.3.1 Selection of the most appropriate data sub(set) The degree to which group assignment matches between the iterations gives an indication of how susceptible the iterations are to small changes in the dataset: those iterations with the highest group match should be the most robust, capturing the underlying causes for the variation. The reduced character matrix (Appendix 1.1.1) has the advantage that the characters are not implicitly correlated (for example, branch length and crown width/height), and so cluster definition should reflect only natural correlations (e.g. one taxon having both a long stem and a wide crown). However, it removes some measures of variation, such as the variation in stem width along the length of the stem, and the shape of the crown. Analyses run on the reduced character matrix gave the best match between iterations, but this may reflect the lower number of continuous characters on which these analyses were based. The higher match for iterations using only categorical characters with those using only the reduced continuous character matrix may reflect a reduction in noise from the additional continuous characters used in the complete character matrix.

4.4.3.2 Categorical, continuous characters, or both? If the clusters identified by these analyses represent discrete taxa, which iteration should be used for taxonomic discrimination: those including all characters, only categorical characters or only continuous characters? If categorical characters do indeed represent generic-level diagnoses, then division of clusters should first be based on these characters, and only subsequently subdivided into species by variation in continuous characters. However, unless taxa with the same branching characters have converged on

72 forms that are indistinguishable based on the sum of their continuous characters, then the combination of continuous and categorical characters should provide the best discrimination of genera.

In the analyses of Primocandelabrum, the cluster assignment for iterations using only continuous characters most consistently agrees with the assignment from iterations using all characters, having a higher cluster assignment match than for categorical characters only with all characters. While variation in one continuous character (such as stem length) could reflect ecological drivers (see Chapter 6), variation in all continuous characters as documented here is much harder to explain away in this manner, especially as all specimens are found on the same bedding plane. Equally, the potential functional drivers for branching architecture as well as for continuous characters means that one set of characters is not inherently more robust than the others (see section 4.4.1). Accordingly, the clusters defined by all characters are interpreted here to most likely represent the original, biological clusters within the data. The greater mismatch in cluster assignment between iterations when only categorical characters are used, compared to when both categorical and continuous characters (or only continuous characters) are used, has two likely causes. Firstly, the clusters defined do not incorporate all of the individuals that share a character state. Secondly, the particular set of categorical characters that are used to describe the clusters varies widely between the different clusters in a given iteration and between iterations (Table 4.1, 4.2). In contrast, the same continuous characters are used to define the clusters both between clusters in a given iteration and between iterations (Table 4.1, 4.2).

Inclusion of outgroups is an important consideration. Their inclusion can help to root clusters by providing information on ancestral or shared characters, but they can also mask variation within the group under scrutiny, particularly if there is convergence of characters. When Orthiokaterna is included, the characteristics of individuals belonging to Orthiokaterna contribute to the definitions of the component space and of all the clusters, both masking variation within Primocandelabrum and influencing the way that group is divided (Table 4.1, 4.2). When Orthiokaterna is excluded, it is just the characters of Primocandelabrum that contribute to the component space – these iterations should most reliably and accurately identify natural clusters within Primocandelabrum. Inclusion/exclusion of the two “dusters” has little effect on the analyses, but also influences the clustering – particularly when 8C3 is included, as this individual is assigned to Orthiokaterna.

4.4.4 Placement of “dusters”

Separation of the two “dusters” is more problematic, as their morphological and branching characters place one (8C3, Fig. 4.1j) within Primocandelabrum clusters, and the other (8C1, Fig. 4.1k) most frequently with Orthiokaterna. There are three potential causes for the failure of “dusters” to fall consistently into the same cluster, and into a cluster separate to the Primocandelabrum specimens: 1) the limited size of the dataset; 2) they represent two different taxa; 3) the exclusion of a key taxonomic 73 character (the number of folia) from the analyses due to its uncertainty in the vast majority of specimens of both Primocandelabrum and “dusters” – as with Orthiokaterna, it is only possible to determine a minimum number of folia. Inclusion of number of folia may change the placement of “dusters” relative to Primocandelabrum, and would further distinguish Orthiokaterna from Primocandelabrum. Based on morphological characters, 8C1 is more similar to Orthiokaterna than it is to 8C3 or Primocandelabrum, and its branching structure (as far as can be determined) only differs to that of Orthiokaterna in the sense of inflation (Appendix 2.6), and possibly also in the rotated/displayed nature of its folia (Appendix 2.6). While its stem is thicker and its holdfast smaller than other specimens of Orthiokaterna even of similar size (Appendix 2.4, 2.5), its dominant placement in the Orthiokaterna cluster suggests that it is a different morph of Orthiokaterna, though whether a different species or simply a variant cannot be determined on the evidence from this solitary individual. All Primocandelanrum specimens appear to have three, rotated folia – because the number of folia and their rotated/displayed nature in 8C3 is uncertain, it is not clear whether this individual represents a separate taxon or a different expression of Primocandelabrum. This indicates that the number of folia and their branching architecture are key taxonomic characters. However, the consistent placement of 8C3 within the Primocandelabrum spectrum based on those characters that are available, and its close position to Primocandelabrum in the principal components space even when its folia are set as “displayed”, does not preclude its assignment to this genus.

4.4.5 Division of Primocandelabrum

The number of clusters within Primocandelabrum is difficult to assess, given the variability of cluster composition depending on data treatment, and the fact that many characters which define clusters do not do so exclusively (Table 4.2).

The clustering following principal components analyses consistently supports two or three clusters within Primocandelabrum, by analysis of inertia gain (Figs. 4.5, 4.6), with the best discrimination of clusters achieved when the specimens are divided into three clusters. Analyses based on only branching characters show additional increases in inertia gain at higher levels of cluster division (Figs. 4.5a, c, 4.6c), supporting more than three clusters, but even then best cluster discrimination is generally achieved when the group is divided into three rather than more than three clusters . When Primocandelabrum is divided into the three clusters determined by the majority of cluster analyses, profile plots of the data show a smoother, more consistent correlation of morphological traits than when the individuals are unsorted (Fig. 4.7a, b). Size trends appear to show a greater increase in all dimensions than those for stem width (which show a shallower increase with size; Fig. 4.7b). Division into three groups is also borne out by simple scatterplots (Fig. 4.7c, d), where the groups assigned by the principal component analyses (reduced and full character matrix) follow separate trends for many pairs of measurements. An

74 underlying ontogenetic cause for these differences can be ruled out by the fact that the large specimens do not cluster together or plot close to each other (Figs. 4.4, 4.7c—f).

So, there are three clusters that can be described within Primocandelabrum. The question then, is what taxonomic level these clusters represent: generic, species, or morph/variety. Orthiokaterna is consistently discriminated from Primocandelabrum based on categorical and continuous characters in isolation and in tandem, strongly supporting its separation into a different genus. However, such clear discrimination is not achieved for Primocandelabrum. With one or two exceptions (which are not even consistent across all iterations; Table 4.1, 4.2), there is not a 100% assignment for individuals which share the same categorical character state to the same cluster, nor do 100% of individuals within a cluster share the same character state (Table 4.1, 4.2). Furthermore, only two or three categorical characters statistically significantly contribute to cluster definition, with the other categorical characters showing no significant correlation to one cluster over another, even when only categorical characters are used. In contrast, Orthiokaterna is discriminated from Primocandelabrum based on almost all of its categorical characters.

The mean values of many continuous characters for each cluster within Primocandelabrum are statistically significant from the means of the characters for Primocandelabrum as a whole (Table 4.2). Additionally, morphological and branching characters do not correlate perfectly, as they do for Orthiokaterna: an individual could be assigned to a taxon based on its branching, or on its morphology, but the two are only likely to produce the same assignment around two-thirds of the time (Figs. 4.4, 4.5; Table 4.3, 4.4). This morphological variation does not appear to be the result of ontogenetic variation, as size ranges of specimens ascribable to each form overlap. Without dividing the Primocandelabrum population into at least 15 clusters (and ignoring variation in continuous characters), no iteration produces a set of clusters within which all individuals are identical in terms of their branching and, crucially, that are distinct from other clusters.

While the three groups can be distinghuished consistently based on the sum of their characters (continuous and categorical), the variation within the groups overlaps to a small extent: for example, the primary branches of both cluster 2 and cluster 3 may be subparallel or radiating, and their folia are unconcealed, but the two may be distinguished by the proportional widths of their holdfast discs, lengths of their stems and the width of their crowns, and by the concealed/unconcealed nature of their primary branches (Table 4.2). Based on this variation, it is possible to treat the three group as varieties of one species. While lumping all the individuals into one species, within which there is a large amount of intraspecific variation, would remove the problem of overlap of individual characters between species, the fact that three coherent groups can be defined supports there being three natural groups within Primocandelabrum. Accordingly, three species are defined here (Appendix 2.7). 75 4.5 Systematic Palaeontology

Material: The genus is described from fifty-two complete specimens, all from Bed B in Charnwood Forest (Chapter 1). Master moulds and casts are housed at the British Geological Survey, Keyworth.

Group: Rangeomorpha (Pflug 1972)

Genus: Primocandelabrum Hofmann, O’Brien and King (2008)

Type species: P. hiemaloranum Hofmann, O’Brien and King (2008)

Multifoliate rangeomorph comprising a stem, holdfast and crown. The crown is wider than it is tall, comprising around 70% (60—80%) of the total height of the organism taken from the base of the stem, and its width may be up to 1.5 times the total height of the organism. The stem comprises the remaining 30% (20—40%) of the total height, and its width is around 10% (7—15%) of the total height. The width of the holdfast is around 30% (20—40%) of the height of the organism. The crown is comprised of three main first order branches (folia), which are rotated, unfurled, and typically show proximal inflation. Second order (primary) branches are typically unfurled, displayed and show median inflation. Both orders may be unconcealed or concealed at their bases, typically appearing concealed towards the distal tips.Third order (secondary) branches are typically displayed, concealed, and unfurled, with either median or distal inflation.

P. anatonos sp. nov.

Holotype is designated as 19A4 (GSM106040; Fig. 4.1f), paratypes are 4A1b (GSM105953) and 19C4 (GSM106049).

This species has a proportionally long stem (average 35%, range 30—40% of the total height), and a crown that is both short (average 65%, range 50—75% of the total height) and narrow (average 80%, range 60—90% of the total height). It is characterised by folia and primary branches that are concealed. Folia show dominantly proximal (but sometimes median) inflation. Primary branches show dominantly median (but sometimes proximal) inflation, may be displayed or rotated and subparallel or radiating. Secondary branches are concealed but may be rotated or displayed, furled or unfurled, subparallel or radiating, and with no or distal inflation. Etymology: anatonos (Gr.), meaning stretching high or upwards, in reference to the proportionally long stem of this species.

P. boyntoni sp. nov.

Holotype is 6A2 l (GSM105969; Fig. 4.1c), paratypes are 10C3c (GSM106051c; Fig. 4.1d) and 10B8 (GSM106049; Fig. 4.1e).

76 This species is characterised by a proportionally small disc (~25%, 10—30% of the total height), a short stem (average 25%, range 10—30% of the total height), and a relatively wide crown (approximately equal to the total height, but 75—130%). The proportions of this taxon are close to the average for the genus Primocandelabrum. Its folia have proximal inflation and are typically unconcealed, and its primary branches are typically displayed, concealed and show proximal inflation, but which may be subparallel or radiating. Its secondary branches are concealed and are typically displayed and unfurled, but may be subparallel or radiating, and with median or distal inflation.

Etymology: Named for Helen Boynton, in recognition of her work on the biotas of Charnwood Forest.

P. katatonos sp. nov.

Holotype is designated as 3(1) (GSM105945; Fig. 4.1g); paratypes are “Big Bertha” (GSM105872; Fig. 4.1b) and 3(2) (GSM105948).

This species is characterised by a crown that is wider than the organism is tall (average 140%, range 110—175% of the total height), and a proportionally large disc (average 50%, range 30—90% of the total height). Its folia are unconcealed and inflate proximally, and its primary branches are displayed, concealed or unconcealed and may be furled or unfurled and may show distal inflation. Its secondary branches are concealed, and may be displayed or rotated, furled or unfurled, subparallel or radiating and with no, median or distal inflation.

Etymology: katatonos (Gr.), meaning wider than it is high, in reference to the proportionally wide crown and holdfast disc of this species.

4.6 Conclusions

Statistical techniques such as those described here provide a powerful way of analysing large datasets with missing values and, importantly, of analysing both categorical and continuous characters in tandem. The separation of Orthiokaterna from Primocandelabrum is supported by all cluster analyses using morphological characters only, branching characters only and both sets of characters combined (Fig. 4.6).

In the majority of iterations using FAMD, MCA and PCA analysis followed by HCPC, either two or three clusters are supported within Primocandelabrum (Figs. 4.5, 4.6). When three clusters are selected, the characters contributing to cluster definition better show better category division than those for two clusters, incorporating into the cluster a higher percentage of individuals fitting a particular category, and having a higher percentage of individuals within the cluster which fit that category (Table 4.2). Accordingly, individuals assigned to Primocandelabrum are divided into three species: P. anatonos, P.

77 boytnoni and P. katatonos. The inclusion of additional specimens into the dataset would doubtless help to confirm whether these are species or just varieties: if the variation is continuous and intra-specific, then the clustering would presumably break down, whereas if it is discrete, it would strengthen the clustering. The “dusters” are concluded to contain two separate taxa, one Orthiokaterna sp. (8C1), and one which may be Plumeropriscum (Mason and Narbonne, In Press), although the lack of information regarding the branching characters of this specimen makes this assignment tentative at best.

The variety of branching characters in specimens whose morphology is virtually indistinguishable and, likewise, the shared branching characteristics of individuals with at least superficially distinguishable morphology is surprising. This is especially so given the fact that few other published descriptions acknowledge variability in branching pattern within a taxon, unless attributed to processes such as ontogeny (Brasier et al. 2012). Certainly Charnia masoni seems to have a very consistent branching pattern across multiple specimens (Wilby et al. 2015), with the only variation observed clearly attributable to taphonomic processes. Charnwood Forest has very few other easily assignable unifoliate (Chapter 1) rangeomorphs, and so their variety in branching pattern is impossible to constrain. The only other multifoliate taxon known from more than a handful of individuals is Bradgatia, but branching pattern is not distinguishable across large parts of many of these specimens due to their complex three- dimensional shape and their preservation as a compound fossil (Chapter 3). Therefore, it is possible that intra-specific variability in branching pattern is a feature unique to multifoliate (or perhaps just stalked and mutlifoliate) taxa. This fits well with other observations about their growth which indicate a plastic component to their growth (Chapter 5). Newfoundland material would provide an excellent source of investigation into the variety in branching within unifoliate taxa such as Trepassia, Vinlandia and Beothukis to constrain this hypothesis.

78 5. ORTHIOKATERNA FORDI AND THE GROWTH OF RANGEOMORPHS

5.1 Introduction

Most communities in the Ediacaran successions of Avalonia are dominated by rangeomorphs (Narbonne 2005, Laflamme et al. 2013, Liu et al. 2015a). These enigmatic organisms are characterised by pseudo- fractal branching and a frond-like gross morphology, but their position on the tree of life remains largely a mystery. Whilst recent studies have advanced understanding of the mode of life (Laflamme et al. 2009), reproduction (Darroch et al. 2013, Mitchell et al. 2015) and growth (Wilby et al. 2015) of individual taxa, many aspects of rangeomorph biology are still relatively poorly understood (Liu et al. 2015a). Each new rangeomorph taxon can be used to scrutinise existing interpretations of their mode of life, to test models of growth and ecology, and to verify and/or modify taxonomic frameworks (Brasier et al. 2012, Liu et al. 2015b).

At present, rangeomorphs are classified on the basis of their gross morphology (Laflamme et al. 2004, Laflamme et al. 2012b) and details of their branching architecture (Narbonne 2004, Bamforth & Narbonne 2009, Brasier et al. 2012). The relative merits of morphological and branching characters in taxonomy are discussed in Chapter 4. There, it was concluded that combining morphological and branching characters provides the most reliable basis of taxon discrimination, and allows quantification of intraspecific variation in Primocandelabrum. It has the added benefit of permitting taxonomic identification in poorly preserved forms by form taxonomy, and in partial specimens by branching architecture.

Charnwood Forest is notable for its abundance of stalked and multifoliate (Chapter 1) rangeomorphs – this abundance permits a greater understanding of their functional morphology to be made by testing interpretations across numerous well-preserved individuals and across different taxa. The gross morphology of stalked, multifoliate fronds is strikingly similar to that of passive filter feeding invertebrates, which live in similar deep-water depositional environments to those interpreted for rangeomorphs. The physical constraints imposed on these modern organisms are therefore similar to those that rangeomorphs would have faced, and so they are particularly useful in studies of functional morphology. Here, I describe Orthiokaterna fordi, a multifoliate rangeomorph with a conspicuously large holdfast, a densely branched, goblet-shaped frond, and proportionally the longest stem recorded for any rangeomorph. Although its overall morphology is unique among rangeomorphs, it is superficially similar to crinoids and certain sea pens, and it represents the first claim for convergence upon this form which was so ubiquitous and successful in the Phanerozoic. The exquisite preservation of Orthiokaterna specimens combined with the fact that all resolvable branching orders possess distinct architecture permits identification of a novel growth structure, termed ‘eccentric branches’. These yield valuable insights into the growth of rangeomorphs, as well as their ability to recover from damage.

79 5.2 Systematic Palaeontology

Material: Seven specimens, all from a single bedding-surface (Bed B of Wilby et al., 2011) in the Bradgate Formation, Maplewell Group, Charnwood Forest, UK. Master moulds and casts are housed at the British Geological Survey, Keyworth, UK (GSM105875, GSM105957, GSM105958, GSM105969, GSM106012, GSM106040 and GSM106112; Fig. 5.1; originals are in situ. Two further fossils (GSM106034 and GSM106113) from the same surface are assigned to this taxon, based on their overall proportions, but are poorly preserved. One specimen, GSM106043, may represent a second species of Orthiokaterna based on its morphology, but the majority of its branching characters are more similar to those of P. boyntoni: its assignment to Orthiokaterna is dependent on the displayed nature of its folia, which is difficult to constrain with confidence in this individual (Chapter 4). Two isolated holdfast discs, one at Hofmann’s locality 5 on the Bonavista Peninsula, Newfoundland, and one in Charnwood Forest, bear the triangular structure that is characteristic of Orthiokaterna. However, neither has exposed/ preserved fronds, and so their assignment to this taxon is tentative.

Orthiokaterna fordi sp. nov.

2011 “dumbbell-like taxon”, “dumbbell-like frond” Wilby, Carney and Howe, p. 656, Fig. 2D; Fig. 5.

Group: Rangeomorpha (Pflug 1972)

Genus: Orthiokaterna gen. nov.

Type species: Orthiokaterna fordi

Holotype is designated as GSM106040, paratypes is GSM105875.

Diagnosis: Frondose rangeomorph comprising a disc and similarly-sized frond connected by a comparatively long, narrow stem. The disc is large, typically preserves several concentric rings, and frequently includes a triangular feature at its junction with the stem. The stem is long and preserved straight, comprising around 60% of the total height of the organism. The crown or frond has a sub-circular outline and is multipolar and multifoliate, comprising numerous independent branches emanating from a single point at the distal end of the stem. First order branches (folia) are displayed, furled or unfurled and unconstrained, and show distal inflation. Second order branches are displayed, furled, radiating and unconstrained, and show proximal inflation. Third order branches are displayed, furled, radiating and unconstrained, and show distal inflation. Fourth order branches are displayed, furled, constrained and shows slight radiation and slight distal inflation. Branch axes of all orders are concealed, and opposing ones are offset along the length of the host branch. First and second order branches commonly bear randomly distributed ‘eccentric branches’.

80 Figure 5.1. Orthiokaterna fordi. A) GSM105875, the largest known example; B) Interpretive line drawing (up to folium level detail) of GSM105875, dark blue area is the holdfast disc, with dark blue lines outlining its internal rings; medium blue are is stem, red lines within which are the “lineations” and “triangle”, bright blue lines outline the folia; C) GSM106040 (mould); D) GSM105959 ; E) GSM106112; F) GSM105957, the smallest well-preserved example; G) GSM 105958. All scale bars = 2 cm, all photographs are of casts held at the British Geological Survey unless otherwise stated.

Etymology: The branching part of this taxon is borne on a comparatively long, straight stem and thus was afforded a relatively high position above the sediment surface (Gr. orthio- meaning lofty, high) compared to its bed-fellows. The frond is profusely branched (Gr. katernes meaning with luxuriant branches). Named for Trevor Ford, in recognition of his seminal work on the fossils of Charnwood Forest and his contributions towards their understanding.

Description: Fossils of Orthiokaterna have a large disc at the proximal, and a similarly-sized sub-circular frondose portion (hereafter ‘crown’; Chapter 1) at the distal end, joined to the holdfast by a long, narrow stem (Fig. 1). The height of known specimens, from the base of the stem (i.e. centre of the disc) to the distal margin of the frond, range from 7.6 cm to 37.6 cm (Appendix 2.1, 2.4).

81 The diameter of the disc ranges from 2.7 cm to 27 cm, measures 49% of the total height of the organism on average (Appendix 2.4, 2.5). There is a statistically significant correlation between the proportion of the disc diameter to total height with increasing size (p=0.03; Appendix 2.8), but this is due to the influence of the largest individual – there is no correlation when the smaller individuals are considered (p=0.31; Appendix 2.8). The disc is comparatively low relief (<0.5 cm, approximately constant across the disc), has a smooth surface texture and a well-defined outer margin. It has variable numbers of concentric rings (both complete and partial; the total number is broadly proportional to the size of the disc. Rings vary inconsistently in width within and between specimens, from 1 mm to 32 mm wide, and are bounded on each side by a ridge of <1–2 mm width. The rings have a consistent right−left slant in profile: the ridges which define the rings consistently slope steeply on the right hand side (looking from the proximal to the distal end), and slope much more shallowly into the hollow of the ring on the left, a trend which continues right the way across the disc.

The stem is straight and of uniform width, except at its base where it expands into a triangular structure. It comprises between 58% and 69% of the total height of the organism, but this variation is not correlated to total height (Appendix 2.8). The stem of the largest specimen displays fine, closely-spaced, parallel, linear features along much of its length (Fig. 5.1a, b). A triangular feature forms the junction between the disc and stem, is typically approximately a third of the width of the disc (proportional to its size), and overlays the rings within the holdfast. The largest specimen (Fig. 5.1a, b) possesses two such structures, which overlap.

The frond of Orthiokaterna is broadly circular in outline and has a well-defined, scalloped distal margin (Fig. 5.1). In many specimens, there is a faint, incomplete duplicate of the outline a few millimetres beyond the margin (Fig. 5.1). The frond is slightly wider than it is high (Appendix 2.5, 2.8), and the width correlates to that of the disc on an almost 1:1 ratio (Appendix 2.5), but there is no significant variation in this ratio with total height. It is approximately symmetrical, with left and right most branches having equal length (Appendix 2.4). This shape is maintained throughout ontogeny – there is no appreciable variation in the ratio between crown proportions and total height with increasing size (Fig. 5.1; Appendix 2.8). The frond is multipolar and multifoliate, consisting of numerous folia (Chapter 1), which all emanate from the terminus of the stem (Fig. 5.1). Five folia are visible in the majority of specimens, but only four are clearly preserved in the smallest specimen (Fig. 5.1e). The largest specimen (Fig. 5.1a, b) has five clear folia, but additional (overlying) ones are suggested by the frond’s scalloped distal margin. The number of branching orders resolvable varies between specimens within this population, presumably as a function of their size and preservation. The pattern of overlap of folia is similar in all specimens, based on the proximal portions of the frond, where the original relationship is likely to be best conserved: typically, one folium is partially overlapped by those on either side, whereas all others are overlapped

82 only on one side. In the two largest specimens (Fig 5.1a, c), a single folium does not conform to this pattern in its medial and distal parts, but this is taken to record extrusion of the folium outside of the normal overlap pattern during felling (i.e. it is a taphonomic artefact). Where overlap of first and second order branches does occur, branches are typically overlapped by their distal neighbours.

At least four orders of branching are present in the best-preserved specimens (Fig 5.2, 5.3; a fifth is suggested in the largest specimen by linear features subdividing its fourth order branches. The folium is displayed, unconstrained and shows median-distal inflation (larger specimens show more pronounced distal inflation); furled/unfurled folia can be determined in six specimens – in three specimens, they are clearly unfurled at their distal tips but locally furled at their bases, but in the other three they appear furled along their length. First order branching is displayed, furled, radiating, unconstrained and shows moderate proximal-median inflation. Three first order branches over two specimens appear unfurled but, as this is a highly localised occurrence, it is presumed to be taphonomic. Second order branching is displayed, furled, radiating, unconstrained and inflates moderately distally. Third order branching is displayed, furled, constrained and shows moderate radiation and moderate distal inflation. The fourth and fifth orders appear furled, but no other detail can be confidently resolved. Branch axes of all orders are concealed. Three specimens have furled folia.

Individual branches on three out of seven well preserved specimens, including the two largest individuals, possess unusual features (Fig. 5.2, Fig. 5.3). These are also the three best-preserved specimens, and so it is possible that they occur in the other specimens but are not resolvable. They occur not just once, but several times in an individual (Fig. 5.2, 5.3). These branches are borne by first-, second- and third-order branches, and fit within the regular pattern of the host branch; that is, they appear to occupy the position of a normal branch of the same order as their neighbours. They are an unusually large size relative to their neighbours, and possess a branching pattern which mimics that of the host branch rather than that of their neighbours (Fig. 5.2, 5.3). Such branches are here termed “eccentric branches” (see section 5.5.4 for a discussion of their interpretation and a comparison to “subsidiary branches”). The number of adjacent eccentric branches decreases with branching order, from up to four adjacent braches emanating off a second order branch having eccentric branching, to single occurrences of eccentric branches being hosted off a folium.

5.3 Comparison to existing taxa

At least superficially, Orthiokaterna is similar to two other rangeomorph taxa: Bradgatia Boynton and Ford (1995) and Primocandelabrum Hofmann, O’Brien and King (2008). Both Bradgatia and Orthiokaterna have bushy, multifoliate fronds, and both Orthiokaterna and Primocandelabrum possess a discrete stem and holdfast. Additionally, the only specimens with holdfasts approaching the proportional

83 Figure 5.2. Detailed branching architecture of Orthiokaterna. A) GSM106040 B) close up of a) and C) interpretative line drawing of b). D) GSM106112; E) close-up of d) and F) interpretative line drawing of e). Key to colours on facing page. All scale bars = 2 cm, photographs are of casts housed at the British Geological Survey. Interpretative line drawings are digitsed camera lucida drawings. size of those of Orthiokaterna are the largest specimens of P. katatonos from Charnwood Forest (Chapter 4), and large specimens of Charniodiscus on the same surface. However, both existing taxa are readily distinguishable from Orthiokaterna.

5.3.1 Orthiokaterna and Bradgatia

Orthiokaterna and Bradgatia are most obviously distinguished by the presence of a large, circular holdfast and a long stem in Orthiokaterna, in contrast to the much smaller, bulb-shaped holdfast and lack of stem in Bradgatia (Fig 5.4a). No specimen of Bradgatia has been discovered which possesses even a small stem, and there are no known morphological intermediates between the Bradgatia and Orthiokaterna, either from Bed B or documented from other localities. This is quite different to the Beothukis mistakensis and Beothukis plumosa (formerly Culmofrons plumosa Laflamme et al. 2012b), which differ not in the presence/absence of a stem, but in the proportional length of the stem. Based on specimens from Charnwood Forest, both Bradgatia and Orthiokaterna are interpreted to have a multifoliate (multipolar) construction, although the possibility of a central axis which has been reduced to zero cannot be entirely excluded. By extension, each folium in these organisms is equivalent to the frond

84 Folium Figure 5.3. Eccentric branching in Orthiokaterna. Progressive First order branch close-ups of the area outlines by the white box in the preceeding Second order branch image, with interpretative line drawings (digitised from camera Third order branch lucida drawings). The final image in each series is of the same Fourth order branch view as the third image, but shows only the interpretative line Eccentric branch drawing. A) GSM106040; B) GSM106112; and C) GSM105875. All Eccentric branch (internal branching) scale bars = 2 cm, photographs are of casts housed at the British Cleavage crack Geological Survey. of a unifoliate (unipolar) rangeomorph and, if the number of folia is maintained throughout ontogeny (Flude & Narbonne 2008), this number may hold phylogenetic significance.

Based both on the revised diagnosis of B. linfordensis given by Brasier et al. (2012), and on observations of new specimens (Fig. 5.4a) and the holotype of this taxon, there are key differences in the branching architecture of the fronds of Orthiokaterna and Bradgatia. Some specimens of Bradgatia and of Orthiokaterna have folia which appear furled only at their base, considered to be ontogenetic (Brasier et al. 2012). However, the smallest specimen of Orthiokaterna possesses unfurled folia throughout (Fig.

85 5.1e), and the outermost branches of juvenile and adult specimens of Orthiokaterna (Fig. 5.1a, b, c) and Bradgatia (Fig. 5.4a) appear unfurled, with furling observed solely in the central parts of the fossils (Fig. 5.4b). Overlapping of folia is most prominent in the central parts, and so I consider the furled–unfurled progression in these specimens to be taphonomic, caused by the unfurled margins of the folia in contact with the preservation surface being obscured by those overlapping them, giving the straight outline of furled branches. The most prominent difference in branching architecture between the two taxa are in the structure of their primary branches – those of Orthiokaterna are furled and show proximal-median inflation, whereas those of Bradgatia are unfurled and inflate distally. Secondary and tertiary branches of Orthiokaterna are furled and have only moderate distal inflation, whereas those of Bradgatia are unfurled and show pronounced distal inflation. There is also less pronounced radiation in the tertiary branches of Orthiokaterna than in those of Bradgatia. Additionally, the frond of Orthiokaterna is of a consistent shape throughout ontogeny, whereas larger specimens of Bradgatia tend to be preserved with a more open, chaotic appearance – although this trend was interpreted as ontogenetic by Flude and Narbonne (2008), it likely reflects the influence of taphonomy on preserved morphology (Brasier et al. 2013). Folia of Orthiokaterna also appear straighter in preserved outline, in specimens of all sizes, than those of Bradgatia (Fig. 5.4a). This suggests an underlying difference in the resistance to flow deformation of their folia.

A recent interpretation of Bradgatia has the first order branches emanating off a central axis which has been reduced to a short length, and accordingly term them primary branches (Brasier et al. 2012, Brasier & Antcliffe 2009) – by definition, all primary braches in rangeomorphs (see Chapter 1) emanate sequentially from a central axis (Brasier et al. 2012). However, a major argument supporting this interpretation was that the branching of Beothukis then fit neatly between the branching of Charnia and of Bradgatia (Brasier & Antcliffe 2009) – this reconstruction of Bradgatia was at least partly dependent on suggested phylogenetic relationships and evolutionary progressions. In Orthiokaterna, it is clear that the first order branches (folia) share a single point of origination (the terminus of the stem) and, in a new specimen of Bradgatia from Charnwood Forest (Fig. 5.4), it is clear that the same is true - the first order branches emanate from a central point and are therefore termed folia. There are two possibilities for this arrangement. One is that the central axis off which true primary branches radiate has reduced to a single locus. In this case, each folium is equivalent to a primary branch and the term “folium” is redundant. This would be an extreme version of the Brasier et al. (2012) interpretation of Bradgatia. The other possibility is that each folium is analogous to the frond of a unipolar/unifoliate rangeomorph (i.e. excluding its holdfast), arranged either helically (cf. Hoyal Cuthill & Conway Morris 2014) or radially. Although this interpretation was rejected for Bradgatia on the grounds of parsimony (Brasier & Antcliffe 2009), such a construction was considered theoretically possible (“multipolar”; Brasier et al. 2012), and is consistent with the majority of interpretations for Bradgatia. 86 A B

Figure 5.4. Bradgatia, GSM105873. A) Whole specimen, showing bulbous holdfast and typical unfurled, displayed branching architecture; B) inset of A), showing the folia clearly emanating from a shared central point.

There are three possible alternatives for the similarity in branching pattern but difference in overall morphology between Bradgatia and Orthiokaterna. Firstly, the groups may represent ecophenotypes of a single taxon, as traits such as height in the water column (length of stalk) and size of holdfast would likely be susceptible to local environmental pressures (Chapter 7—9; Liu et al. 2015b) and attendant functional constraints. As examples of both taxa occur on the same bedding plane and in close proximity, such variability would presumably need to be highly localised and/or persistent in order to induce ecophentotypic variation. Secondly, the two groups could be part of an ontogenetic sequence. This possibility is considered unlikely on the basis that clear ontogenetic sequences are known for both forms, and that known specimens overlap in size range. Thirdly, whilst it is possible that the two groups represent different phases in a life cycle, or even different sexes, there is no recorded evidence for either of these in rangeomorphs. Finally, they represent two separate taxa. While the proportions of a stem alone may not be a taxonomically useful character (Liu et al. 2015b; though see Chapter 4), its presence or absence is likely to be important. Similarly, while both taxa possess holdfasts, they are of different shapes. The fact that the two groups possess distinct branching architecture which, crucially, does not overlap (Chapter 4), further precludes the two taxa from being different species of a single genus. Despite these differences, and provided that branching architecture is indeed a reliable indicator of phylogenetic relationships (Brasier et al. 2012) ,then Orthiokaterna and Bradgatia are considered more closely related to each other than to other currently-described rangeomorphs.

87 5.3.2 Orthiokaterna and Primocandelabrum

Like Orthiokaterna, Primocandelabrum possesses a disc, a straight stem, and a ‘bushy’ frond. However, unlike Primocandelabrum, Orthiokaterna is neither triangular in preserved outline, nor does it have “coarse” branches arranged in a form resembling a candelabrum (part of the generic-level diagnosis of Hofmann et al. (2008). The poor preservation of the type specimens of Primocandelbrum makes it impossible to determine the branching architecture for the taxon. However, new specimens from Charnwood Forest do conform to the Primocandelbrum diagnosis (Hofmann et al., 2008) and overall morphology, and so permit comparison of Primocandelabrum and Orthiokaterna at a more detailed level (Chapter 4). Although the two taxa fit into the same stalked, multifoliate rangeomorph morphospace, they can be confidently and consistently distinguished on the basis of both branching architecture and gross morphology (Chapter 4).

5.4 Discussion: taphonomy, anatomical reconstruction and functional morphology

In light of the lack of known phylogenetic relatives, functional morphology provides the most reliable means by which we may understand the biology and mode of life of these enigmatic organisms. The overall form of Orthiokaterna is superficially similar to that of modern passive suspension-feeding invertebrates (Gage & Tyler 1992), such as certain sea pens (e.g. Umbellula), crinoids (e.g. Metacrinus) and bryozoa (e.g. Kinetoskias). However, accurate reconstruction of organisms which are known exclusively from low-relief impressions must incorporate careful consideration of the effects which various taphonomic processes have had on the organism. The various preservational processes affecting all fossils in Charnwood Forest and Newfoundland, including current-induced modification of form (Brasier et al. 2013) and metamorphic effects are discussed in Chapter 3. Certain aspects of taphonomy (variation in preservation across a specimen, and determination of the sense of overlap between elements) apply to all taxa on the bedding plane are treated in Chapter 3. The most significant effect of taphonomy on the preserved morphology is likely in the exact proportions (section 4.4.2).

5.4.1 The disc

Rangeomorph holdfasts are interpreted as being either buried within the sediment (Gehling et al. 2000b, Narbonne & Gehling 2003, Laflamme & Narbonne 2008, Serezhnikova 2010, Tarhan et al. 2010, Brasier et al. 2013, Menon et al. 2013b) or within a mat (Seilacher 1999, Tarhan et al. 2010, Laflamme et al. 2011). In either case, their function is presumed to be as a holdfast, anchoring the frond to or within the substrate. These holdfasts are typically reconstructed as having been bulbous in life (Gehling et al. 2000b, Grazhdankin & Gerdes 2007, Laflamme & Narbonne 2008, Narbonne et al. 2009, Laflamme et al. 2011, Burzynski & Narbonne 2015), and rarely as flat or disc-like (Steiner & Reitner 2001, Burzynski & Narbonne 2015).

88 The holdfast of Orthiokaterna is conspicuously large relative to that of other rangeomorph taxa, and has a shallow expression proportional to its diameter (section 5.2). If the holdfast was originally spherical or bulb-shaped, as in Bradgatia (Fig. 5.4) and Charnia (Wilby et al. 2015), then it would have originally been up to 30 cm in circumference. Collapse of such a large spheroid would have likely have created a much deeper crater in the sediment and would be deeper in the middle, rather than of the near-constant depth observed in the disc of Orthiokaterna. If suggestions that the holdfast was contained entirely within a microbial mat (Laflamme et al. 2011) are true, then this would account for the high degree of compaction. However, in order to accommodate a spherical holdfast, the mat would necessarily have been originally very thick (i.e. thicker than the circumference). Current sedimentological and petrographical evidence suggests that the microbial mat present on the surface was thin (Chapter 3; Wilby et al. 2015), and certainly unlikely to have been thick enough to accommodate such a large holdfast of bulbous shape. Accordingly, the holdfast is interpreted to have been a strongly oblate spheroid or disc, quite different to the bulbous holdfasts of Bradgatia and Charnia (Wilby et al. 2015).

5.4.1.1 Functional morphology of the disc The proportionally large size of the holdfast in the largest specimen (Fig. 5.1a, b) likely reflects the increased anchorage required as the frond enters into regions of increasingly high flow and consequently high shear (Dade et al. 2001), supporting its interpretation as a holdfast. The fact that this increase is not directly proportional to size suggests that there is a critical threshold in the water column, above which proportionally large discs are required. The preservation of the holdfast as a collapse structure but the frond and stem as epirelief surficial impressions indicates that all specimens of Orthiokaterna (and indeed in rangeomorphs generally) collapsed at the junction between the stem and holdfast disc. Whilst the stem and junction show evidence of strain (Fig 5.2b), there is none in the holdfast or surrounding sediment (cf. “mops”; Tarhan et al. 2010). Therefore, whilst the large size of the Orthiokaterna holdfast would have served primarily to stabilize the organism, it seems that its anchoring capacity exceeded the strength of the rest of the organism, especially given the likelihood that holdfasts were subject to the same forces as the rest of the organism (see Singer et al., 2012).

If the sole function of a holdfast was to anchor the frond, then growth to a size larger than the biomechanical minimum would have been a waste of resources. In modern deep ocean systems, concentrations of dissolved organic carbon (DOC) can be up to an order of magnitude greater in the upper few centimetres of the sediment than in the immediately overlying water column (Hulth et al. 1997, Sierra et al. 2001, Papadimitriou et al. 2002), and is typically highest just below the sediment-water interface (Jorgensen & Boudreau 2001, Papadimitriou et al. 2002). This is likely to have been even more significant in deep- water Ediacaran systems, where surficial microbial mats (Seilacher 1999, Wilby et al. 2015), slow flows (Madsen et al. 2001, Ghisalberti et al. 2014), and rare to absent vertical bioturbation (Seilacher et al.

89 2005, Buatois et al. 2014) would have severely limited the exchange of substances between the sediment and water column through resuspension and/or diffusion (cf. Aller 2001, Boudreau 2001, Dade et al. 2001). Accordingly, geochemical gradients are thought to have been condensed in the topmost sediment during the Ediacaran (Callow & Brasier 2009) and, assuming this holds for the Charnwood Forest palaeoenvironment, the shallow position of the holdfast of Orthiokaterna would have maximised its potential for exploiting these gradients. Its large size could consequently represent an adaptation to allow it to participate in the acquisition of DOC from the substrate (cf. Dzik 2003). A similar interpretation has been postulated for the more morphologically complex discoidal structure Heimalora (Laflamme and Narbonne, 2008), which is characterised by ray-like protrusions and is thought to be an organ taxon of Primocandelabrum hiemaloranum (Hofmann et al2008). The ray-like protrusions of the P. hiemaloranum holdfast would have increased the surface area of the holdfast in contact with the substrate, but no such feature or other adaptation to increase surface area to volume ratios are observed in the holdfast of Orthiokaterna. However, provided that the uptake of substances being exchanged between frond and water column was the limiting step for the organism’s metabolism, then the low surface area of the holdfast (in comparison to that of the frond) would not have been deleterious.

5.4.2 The stem

The stem of Orthiokaterna is proportionally the longest recorded for any rangeomorph. It is consistently preserved straight, whereas the axes of other frondose organisms on the same surface, such as Charnia and Charniodiscus, are frequently sinusoidal, despite generally being shorter. This suggests that Orthiokaterna had a less flexible stem, which was likely capable of bending only in an open arc. Two specimens of Orthiokaterna (Fig 5.1a, g) possess irregular, linear features on the stem (Fig. 5.1a), which compare closely with those that have been attributed in other Ediacaran organisms to current-induced torsion (Tarhan et al. 2010). It is unlikely that these features are original, biological features because of their irregularity and occurrence on only two specimens (although it is worth noting that the stems of other specimens are comparatively poorly preserved). Stretching of the stem beyond the elastic limit of the tissues of which it is composed would irreparably damage it, leading to features similar to those observed here. Based on these arguments, the duplication of the distal margin of the frond (and of the triangular structure in the largest specimen, Fig. 5.1a, b) are interpreted to record relaxation of the organism following its (post-torsion) collapse onto the sediment.

The triangular feature at the junction between stem and holdfast is present in most specimens of Orthiokaterna, but has not (to date) been documented from any other taxon. The compound nature of the impression (overlaying the rings within the holdfast; Fig. 5.1a, b), combined with the interpretation of the fine lineations as by-products of torsion, suggests that the structure likely formed through the collapse of a biological structure rather than through sediment collapse (as interpreted for the rings

90 within the holdfast). Rather than an originally discrete, triangular structure, it is perhaps best interpreted as a flaring out of the stem where it joined the holdfast, which produced the observed triangular feature formed upon its collapse onto the sediment surface. Its preservation on the sediment surface (rather than as a taphonomic collapse crater) suggests that it was the only part of the holdfast to have breached the substrate surface.

5.4.2.1 Functional morphology of the stem It has been argued that the evolution of a distinct stem (i.e. one lacking frondose parts) enabled tiered communities to develop in the Ediacaran (Laflamme et al. 2012b) – if this theory holds, the proportionally long stem of Orthiokaterna would afford it a distinct advantage in a tiered community (Chapter 7). Generally speaking, stems (stalks) have three functions: 1) anchorage to the substrate; 2) elevation of the frond above the substrate and into a higher part of the water column; and 3) depending on the flexibility of the stem-frond junction, allowing the frond to be oriented into an optimum “feeding” position, as for crinoids (Baumiller and Ausich, 1996) and sunflowers (Vandenbrink et al. 2014). The mechanical properties of stems reflect a compromise between holding the frondose component(s) of an organism aloft in the optimum position, and resisting the deleterious effects of currents (Koehl 1977). While a flexible stem would permit adaptation to variable flow direction (Singer et al. 2012), it may also have resulted in the organism being dragged into lower, more densely occupied, parts of the water column, reducing the competitive advantage of possessing a stem. In modern stalked, sessile organisms with an upright posture, the stem is typically stiffer than the frond, and is stiffest at its base (Wainwright & Dillon 1969). The inflexible nature of the stem of Orthiokaterna was therefore likely optimized to maintain the frond in a relatively constant position within the water column, where there is the lowest density of fronds and, hypothetically, the lowest levels of competition (Chapter 7). The consistent failure of the stem at its junction with the holdfast (Fig. 5.1) indicates that this was the weakest part of the stem. Flaring of the stem at this point would have served to distribute the weight of the frond across a wider area of the holdfast distribute, and is here interpreted to have provided strength or support to the stem, thereby increasing its resistance to bending in the flow. In benthic communities, sessile organisms must either withstand sediment influx, or employ anti-fouling strategies (Chapter 6). The long stem of Orthiokaterna would have placed its frond, i.e. its exchange surface, out of the path of small-scale sedimentation events (Chapter 8). Although it may have evolved primarily in response to tiering pressure, its stem may have acted as a pre-adaptation enabling it to persist in environments with frequent sediment influx, where lower-tier organisms would need to find other ways of coping with sediment stress (Chapter 6).

5.4.3 The frond

Reconstruction of the frond of Orthiokaterna is important in understanding how it interacted with the current during life, but the degree to which biostratinomic processes such as currents modified the in

91 vivo shape of Orthiokaterna, in particular the morphology of the frond and the positions of its constituent branches, is debatable. However, all taxa on Bed B, including Orthiokaterna, are mutually aligned, having been felled by a single flow event (Wilby et al. 2011). Specimens of other taxa on the bed are enrolled, entwined, or have a partially disrupted branching pattern, suggesting that the felling flow influenced the preserved morphologies of those individuals (Chapter 3, Fig. 3.6a—c).By extension, it likely also had some influence on Orthiokaterna. Additionally, the fossils record a compound impression of an originally three-dimensional structure, resulting in the overlap and superimposition of branches and attendant complications to interpretations. Orthiokaterna exhibits comparatively clear preservation of morphological detail in the peripheral regions of the frond, a characteristic shared by other taxa on Bed B. This is attributed to largely single, rather than compound, branch impressions being recorded in these areas.

The frondose component of rangeomorphs varies greatly in both overall form and detailed branching architecture. The closest comparison to the frond of Orthiokaterna (at least at first appearances) is to be made with Bradgatia. This taxon has been interpreted as fan-shaped (Brasier et al. 2012), or as lettuce- or cabbage-shaped in life (Brasier & Antcliffe 2004, Flude & Narbonne 2008, Hoyal Cuthill & Conway Morris 2014). In Orthiokaterna, the apparent emanation of all folia from a central point and their (current-induced) alignment lends the preserved frond of Orthiokaterna a superficially fan-shaped appearance. The scalloped distal margin of the frond is attributable to the outlines of the individual folia which comprised it, but only those parts of the folia which came into contact with the sediment were preserved. While five folia are visible along their entire length (from their junction with the stem to their distal tips), the outlines of the distal tips of additional folia are preserved beyond them (Fig. 5.1a, b). Accordingly, the frond is interpreted to have had more than five folia in life. Unless the frond comprised two rows of folia arranged back-to-back, this number of branches is inconsistent with a fan shape. Assuming the folia were the same on both sides, a back-to-back arrangement would likely have rendered the inwards-facing sides of each folium redundant. If the frond were cabbage-like, with multiple folia nestled together, the large specimens especially would be expected to have a more chaotic appearance, as exemplified by the “O” morph of Bradgatia (Flude and Narbonne, 2008; Chapter 3). In contrast, the frond of Orthiokaterna is of a consistent shape in specimens of all sizes (Appendix 2.8), suggesting that it maintained a constant shape and resilience to current-compression throughout ontogeny. The consistent shape and slight curvature to the folia axes in Orthiokaterna suggests that these were comparatively inflexible, and consequently more capable of maintaining their shape and less prone to distortion and ruffling than Bradgatia upon felling. In contrast, the more sinuous shapes of first and higher-order branches suggest that these were more flexible than the folia – more similar to those of Bradgatia. Accordingly, I consider it most parsimonious that the frond of Orthiokaterna was goblet-

92 shaped (Fig. 5.5a), and that its relatively rigid folia allowed this shape to be conserved in the current and, to an extent, during felling.

The outer-most folium of Orthiokaterna (i.e. the one which is in full contact with the sediment and so which preserves both sides) is not always the central one – if it were, then the overlap would be explainable by the central folium being most pushed downstream by the current. The fact that it is not suggests that this is an original feature, with the folia having a controlled order of overlap in life. In some cases, the distal parts of some folia and first order branches are displaced from this otherwise regular pattern of overlap, suggesting that they pivoted about their bases (i.e. they are unconstrained). Neither twisting nor enrollment of folia or branches is observed in any specimen of Orthiokaterna. This suggests that the folia and first order branches, at least, were not capable of rotation about their axis to any significant degree. Additionally, although the majority of first and second order branches are furled along their entire length, some exhibit local unfurling at or near the tip. In other taxa, this has been attributed to an ontogenetic sequence (Brasier et al. 2012), but in others local disruptions to the branching architecture are interpreted as current-induced imbrication and displacement of branches (Brasier et al. 2013). It is not possible to rule out a similar biostratinomic cause for this feature in Orthiokaterna because of its limited and local occurrence.

5.4.3.1 Functional morphology of the frond In organisms that lack internal transport systems, surface area to volume ratios are a limiting factor to growth. The dense branching of rangeomorph fronds creates surfaces with high surface areas, and this has been taken to suggest that these primarily served as an exchange surface with the water column (Laflamme & Narbonne 2008, Laflamme et al. 2009, Brasier et al. 2012). While recent interpretations of rangeomorphs as osmotrophs state that the substance exchanged is dissolved organic carbon (DOC ; Laflamme et al. 2009), this interpretation has a number of significant problems (Liu et al 2015). The high SA/V of fronds may instead have served primarily to act as gas exchange surfaces, especially if the holdfast was involved in the uptake of DOC from the substrate.

Benthic flow in deep-water environments is typically complex, with temporal variations in the direction and strength of ambient flow occurring at any one location (Dade et al. 2001), a variability to which organisms must adapt in order to maximise the incidence on their surfaces of substances carried in the water column. The shapes of modern passive (invertebrate) suspension-feeders, for example, are adapted to maximize capture of particles under the flow-regimes of their specific environments (Gage & Tyler 1992). There is no reason to assume that benthic flow would have behaved differently in the Ediacaran and, even if it did not share a suspension-feeding lifestyle, Orthiokaterna would have faced similar constraints with regard to nutrients arriving from different directions. Some suspension-feeders actively orientate themselves, whilst others respond passively to flow, either short-term (by rotation) or

93 long-term (by growth). There is neither evidence to suggest that rangeomorphs were capable of active orientation, nor that the frond of Orthiokaterna could pivot or flex about its junction with the stem. The bowl-shaped frond of Orthiokaterna would have formed a near-continuous, high surface area interface with the water (Fig. 5.5a). Consequently, even without the ability to orient itself in the water column, Orthiokaterna would have been able to exploit currents from all directions equally – as for modern erect bryozoan colonies of an overall similar morphology (Ryaland & Warner 1986). It is also possible that this shape may have enhanced exchange efficiency by retarding the ambient flow as it was pulled into the bowl (cf. Vogel 1977), thereby increasing the frond’s contact time with the water. Additionally, any vortices produced behind the branches facing the flow (similar to those shed off crinoids; Baumiller 1997) would also have served to reduce the diffusive boundary layer along the frond surface within the bowl, and again on the down-flow (exterior) side. Further modification of the flow may have been induced by the distal-over-proximal sense of overlap of first and higher order branches (section 3.4.2.1). This arrangement may have served to funnel flow up the length of the frond and into the bowl, increasing the contact time between any given part of the water column and the surface of the frond and maximising its chance to uptake the nutrients/gases required for its metabolism (Chapter 7).

5.4.4 Eccentric branches

One of the most notable features of Orthiokaterna is its frequent production of eccentric branches (section 5.2; Figs. 5.2, 5.3), and their observation has permitted recognition of comparable structures in new specimens of Bradgatia and in Primocandelabrum. However, they are so far unrecognised from unifoliate rangeomorphs. The relationship of these structures to their host branch gives critical insight into the way they formed. Subsidiary branches or frondlets recognised in specimens of Bradgatia (Brasier et al. 2012) and Fractofusus (Gehling & Narbonne 2007) represent branches which have developed at additional growth loci on the host branch. In contrast, the observation that eccentric branches in Orthiokaterna (at least, those emanating from primary and secondary branches) occupy the position of a normal branch (Figs. 5.2, 5.3), rather than inserting between them, indicates that they have developed not from an additional locus, but rather from a normal branch which has “gone rogue”. This is illustrated in Fig. 5.5b.

The relationship between eccentric branches and folia is more difficult to constrain, as their bases are frequently obscured. In Bradgatia, eccentric branching can only be confidently distinguished arising from the folia (“primary branches” of Brasier et al., 2012). That eccentric branches can arise at any point along the length of their folia, and occur in even modest-sized specimens of both Bradgatia and Orthiokaterna (Fig. 5.3), suggests that they are not an ontogenetic feature (cf. Brasier et al. 2012). How might such branching have been achieved?

94 A B

Figure 5.5. Artist’s reconstruction of Orthiokaterna fordi. A) Entire organism; B) Eccentric branch emanating of a folium. The branching architecture of the eccentric branch matches that of the host branch (the folium) rather than its neighbouring primary branches.

5.4.4.1 Potential mechanisms and consequences for rangeomorph biology Identification of direct equivalents to eccentric branching is challenged by the lack of most organisms with distinct branching structures at different orders of branching. However, rangeomorphs are essentially modular in their construction (Narbonne 2004), and many modern modular organisms are capable of morphological plasticity (Sebens 1983, Marfenin 1997, Sánchez et al. 2003, Shaish et al. 2007), for example in response to environmental stimuli. At least in plants, phenotypic responses are thought to occur at the level of individual modules (De Kroon et al. 2005). The mechanisms and triggers underlying this morphological plasticity, however, are highly variable. Given that eccentric branches are larger than their neighbours, they likely reflect a response to anomalous concentrations of growth factor (or a substance that stimulates its local production). This is also consistent with the apparent reversion in an eccentric branch to the branching pattern diagnostic of one order lower (i.e. the order of the host branch): either the pattern of a given branch is controlled by concentration of growth factor when it starts growing (with lower concentrations inherently creating higher-order patterns as they are more distal to the source of growth factor), or the enhanced growth rate at the tip (the site of damage) creates disproportionately large distal branches within the subsidiary branch, causing apparent unfurling and distal (as opposed to proximal) inflation, creating an apparent reversion

95 which is simply a function of faster growth rather than a true influence on the genetically-controlled branching pattern.

Regardless of the trigger for eccentric branching, its occurrence permits certain inferences to be made regarding the biology of rangeomorphs. The anomalous concentration of growth factor and/or resources could have been delivered to the site of eccentric branching either as a targeted delivery, or via a diffusive gradient away from the point of production. The most likely method depends on whether such substances were produced/stored and released locally (i.e. at the site of damage), or centrally (i.e. one site in the whole organism). Diffusion of large organic molecules in aqueous solutions is slow (on the order of 6x10-6 cm2/s for glucose; Varga et al. 2009), whereas transport systems in organisms can transport substances much quicker. That fewer adjacent branches show eccentric branching at lower orders of branching (Figs. 5.2, 5.3) is important in distinguishing the mode of storage/transport. If the growth factor travelled from a central production/storage point through diffusion, then it would take a long time to reach the site of eccentric branching and would be of an extremely low concentration by the time it reached the site of branching. Additionally, the growth factor would be spread over a much wider area. A likely consequence of this would be a greater abundance of adjacent eccentric branches, perhaps in a continuum of size distributions from the origination point along the concentration gradient. Diffusion away from a local production/release point is much more likely: the travel distances from source to site of effect are much smaller (millimetres instead of centimetres or decimetres) and, by that token, the increasing number of adjacent eccentric branches with higher order is more easily explained. That the adjacent branches do not discernibly decrease in size (Figs. 5.1, 5.2) either side of a point of production, i.e. along a concentration (diffusion) gradient, suggests that either there were well constrained thresholds to growth expression, or that each branch which has developed eccentric growth was itself a point of a production/release. Based on the available specimens, there is no way to distinguish between the two: growth thresholds may be low enough to permit eccentric growth, but high enough to require a specific amount of growth factor, or all branches in an area may be more prone to whatever triggers the release of growth factor. If the substance is produced at a central location, then a targeted delivery system such as that observed in gorgonian octocorals (Sánchez & Lasker 2004) must be involved. This implies not only a transport system in rangeomorphs, but also a rudimentary communication system: consequently, rangeomorph modules would be a highly interconnected system rather than each existing in isolation (Chapman 1981). While local production and diffusion would seem most parsimonious on the balance of evidence, targeted delivery is most consistent with modern analogues.

5.4.4.2 Possible causes and functions of eccentric branches Regardless of the precise mechanism by which eccentric branches were induced, there are four possible causes for an anomalous concentration in growth factor and the consequent deviation in growth. Firstly, it may be caused through genetic mutation. Whilst random genetic mutations cannot be fully dismissed, 96 there is no a priori reason to believe that multifoliate rangeomorphs should be more susceptible to genetic mutation than their unifoliate or bifoliate compatriots. Their occurrence in very large specimens, multiple times in the same specimen (Fig. 5.2, 5.3), and in both Bradgatia and Orthiokaterna, all count against them being the result of a growth defect (i.e teratological) – such mutations are almost always deleterious to the survival of the afflicted individual, and their frequency of occurrence would require genetic mutations resulting in the same phenotypic expression to have occurred multiple times, in numerous individuals, and across two genera. Secondly, it could represent greater freedom of phenotypic expression in some rangeomorphs compared with others. While it is possible that eccentric branches represent a response to neighbour-shadowing (a form of competition), representing growth on the non- shaded side of the organism as it strives to obtain resources from an un-exploited area (Franco 1986), this would however require a high degree of sensitivity to external stimuli and a sophisticated feedback loop of responses similar to that of auxin in plants in response to light (Wang & Li 2008). Thirdly, it could represent a novel mode of rangeomorph reproduction, with eccentric branches representing daughter clones budding off from the parent, which would be released and settle elsewhere.

Finally, it may represent re-growth in response to branch loss or damage. In modern organisms as diverse as cnidarians and higher plants, physical damage can result in faster, “over-compensatory” growth (new growth exceeding normal rates). One example of particular interest is provided by gorgonian octocorals (Sánchez & Lasker 2004). When the apical growth tips of these organisms are removed, daughter branches near the site of injury develop into mother branches, which themselves branch at a higher rate than normal (Sánchez & Lasker 2004) – directly analagous to the pruning of plants that have apical dominance (Sánchez & Lasker 2004, Wang & Li 2008). Comparably, certain Silurian dendroids display abnormal bifurcation or unusually fast elongation of stipes in a portion of their rhabdosome (Bull 1996). This abnormal growth was interpreted to have served to fill in space created in the rhabdosome structure following damage, with normal growth demonstrably resuming once the space was filled. Such damage response-traits have been demonstrated to be beneficial for modern sessile organisms and, by extension, similar traits would likely have been advantagous for rangeomorphs. Firstly, they enable the organisms to better cope with environmental constraints and maintain their optimum form (Shaish et al. 2007, Shaish & Rinkevich 2009). Secondly, many modern modular organisms such as corals (Lirman 2000, Shaish et al. 2007) and succulents (Gorelick 2015) reproduce through fragmentation. Perhaps the fragment which was removed and induced growth of eccentric branches developed into a new individual. There are to date no known rangeomorphs whose branching pattern mimics that of a single branch of Orthiokaterna. In corals, however, the size of the fragment and the location of the break point in relation to a branch node affect the morphology of the daughter organism (Shaish et al. 2007) – perhaps the size or the branch order of a fragment released from Orthiokaterna influenced the branching architecture of the daughter organism(s). Additionally, given the patchy nature of preservation on Bed B, and its relatively small 97 lateral extent (compared to most modern benthic communities), it is possible that individuals matching our preconceived ideas of what an individual of Orthiokaterna which developed from a released fragment (or indeed from an eccentric branch) would look was simply not preserved. Our current knowledge of the mode of rangeomorph reproduction is still limited to a few individual taxa, and none of them multifoliate (Darroch et al. 2013, Liu et al. 2015a, Mitchell et al. 2015; it is perhaps presumptive to exclude this interpretation of eccentric branching out of hand.

5.4.4.3 Restriction of eccentric branching to multifoliate rangeomorphs While there is no way to test the possibility of certain rangeomorph taxa having greater innate phenotypic flexibility, there may be two reasons for the apparent dominance of eccentric branching in multifoliate (as opposed to unifoliate) rangeomorphs. Firstly, the flexible, comparatively highly overlapping branches of multifoliate rangeomorphs would have been particularly prone to abrasion by neighbouring ones (cf. Franco 1986). Secondly, it may have been selected against in unifoliate rangeomorphs because such branches would distort the outline of the frond and alter the currents around it (cf. Singer et al. 2012) – the behaviour of multifoliate rangeomorphs in flows has not yet been modelled. One mechanism by which this differential growth plasticity could have been achieved is by those rangeomorphs that did not develop eccentric growth (e.g. Charnia) having tighter thresholds on the effects of the growth factor – if a higher threshold level of growth factor was required before a branch could develop eccentric growth (either by altering branch pattern or purely by faster growth), then this growth would be less likely to result from relocation of resources and/or growth factor following damage (Sánchez & Lasker 2004). This suggests that different rangeomorph forms are not simply a case of restructuring branch patterns, but that there is a more nuanced and adaptive aspect to their growth and morphology.

5.5 Conclusions

Orthiokaterna fordi was an upright, stalked rangeomorph with a densely-branched, bowl-shaped frond. Together with Primocandelabrum and Bradgatia, Orthiokaterna occupied a multifoliate rangeomorph morphospace, whereas the majority of rangeomorphs and arboreomorphs adopt a unifoliate construction. The specialisation of a discrete stem in both Orthiokaterna and Primocandelabrum distinguishes them from many of their compatriots (such as Charnia) which had frondose parts along their entire lengths. The globular shape and multifoliate construction of the frond resulted in the presentation of an absorptive surface to the flow regardless of the flow direction, and may have enhanced nutrient uptake by providing prolonged contact of rangeomorph elements with the water within the bowl. Several aspects of the functional morphology of Orthiokaterna suggest adaptation to life in a depositional environment with a strong flow. Its large, shallow-buried holdfast would have served to anchor the organism, while its long stem raised the frond above its compatriots, exposing the frond to faster flow. While this would have reduced inter-specific competition (Chapter 6), the faster flows would have increased the shear on

98 the frond and the mechanical stress placed on the organism as a whole. The short, thick stems in most specimens of Primocandelabrum (Chapter 4) and the long, rigid stems in Orthiokaterna may represent different strategies to dealing with this stress. The disproportionately large holdfast of Orthiokaterna may have exploited nutrients in the substrate. Perhaps the high surface area of the frond was used primarily for gas exchange (rather than carbon capture), for metabolite release (cf. Singer et al., 2012), or for capturing larger particles (Chapter 6). Eccentric branches are interpreted as an over-compensatory growth response to damage through current or neighbour abrasion, and give important insight into the biology and growth mechanics of rangeomorphs, including differential phenotypic plasticity between fronds with multifoliate and unifoliate construction. The division into unifoliate and stem + mutlifoliate morphological categories may represent a compromise between increased uptake of nutrient from the water column, and expenditure of resources on construction of a mechanically resistant stem and holdfast. If rangeomorphs did indeed possess regenerative capabilities, these would have served to mitigate damage caused not only by self-harm, but also by sediment influx (Chapter 8). By the same token, this growth response would have mitigated damage caused by predation; eccentric branching may represent a pre-adaptation to coping with predation pressure.

99 6. UNDOSYRUS NEMORALIS: A STRATEGY AGAINST SEDIMENT OCCLUSION

6.1 Introduction

The Ediacaran is renowned for its abundance of enigmatic fossils. Among the most intriguing forms on Bed B of the Charnwood Forest biota (Chapter 1) are the colloquially-named “Basil Brushes”. These fossils are comprised of a stem and holdfast, and rangeomorph-type branching that is superimposed by an enigmatic lineated structure. There are ten known specimens of this form, all from the same bedding surface. Nothing directly comparable has previously been described, but there are many modern and fossil organisms, and even sedimentological features, which bear at least a superficial resemblance to the lineated structure. In this chapter, I describe the population of “Basil Brushes”, compare them to other fossils on Bed B, and critically assess possible interpretations for the lineated structure.

6.2 Systematic Palaeontology

Material: The genus is described from the nine complete specimens and one partial specimen (stalk and holdfast missing) that are known, all from Bed B in Charnwood Forest (Chapter 1). Original fossils remain in situ in the field; silicon rubber moulds taken from the bedding plane and counterpart jesmonite resin casts are housed at the BGS. Specimen numbers in A.2.1.

Group: Rangeomorpha (Pflug 1972)

Genus: Undosyrus gen. nov.

Diagnosis: Rangeomorph comprising a short stem, a small bulbous holdfast, and a proportionally large, multifoliate crown which is broadly triangular in outline. The crown possesses numerous primary branches or folia which emanate from a central point. This genus possesses a distinct, lineated “brush structure”, which is interpreted to record an external sheath. The brush structure bears a series of undulose, subparallel pairs of grooves separated by a higher-relief ridge and internally divided by a lower-relief ridge. First order branches are rotated, unfurled and concealed; second order branches are displayed, usually furled but locally unfurled, radiating, concealed and have median inflation; third order branches are rotated, furled, radiating, concealed and have distal inflation.

Etymology: Undosyrus Undo- from undosus, meaning billowy, full of waves, syrus (L) = broom.

Type species: Undosyrus nemoralis

Holotype is designated as GSM105872, paratypes are GSM105962 and GSM106009.

Diagnosis: - as for genus.

100 A B

Br Br

Ch Secondary branches Third order branches Fold in brush structure High-relief ridge Low-relief ridge Bifurcation of grooves Outer margin

C D cv

Figure 6.1. Undosyrus, specimen 3Bx (GSM 105872). A) Photograph of the whole specimen, scale bar = 2 cm. and B) interpretative line drawing. Br = underlying branches; Ch = small Charniodiscus specimen adjacent to the main specimen. C) Photograph of inset demarked in a), cleavage trace labelled cv and D) interpretative line drawing. Innterpretative line drawings are digitised camera lucida drawings.digitised camera lucida drawings.

Etymology: nemoralis = of the wood or forest (strictly woody, sylvan), a reference to Charnwood Forest, from where it is described.

Description: All fossils of Undosyrus comprise a basal holdfast, a discrete stem (i.e. one that lacks rangeomorph branches; Chapter 1), and a crown (Chapter 1) which bears the rangeomorph branches (Fig. 6.1a, b). They are aligned parallel to each other and to the other fossils on the surface, presumably

101 by the current which felled the organisms (Chapter 3; Wilby et al. 2011). The most distinctive part of the fossil is a lineated structure (hereafter referred to as the “brush structure” Fig. 6.1, 6.2, 6.3a) which appears to overlie the crown. No feature comparable to the brush structure is observed elsewhere on the surface, either in isolation or in association with an organism of different morphology or branching to that described here. The surface topography of the brush structure is quite variable, with centimetre- scale crests and troughs broadly parallel to the length of the organism. These are interpreted as the main or first order branches of the crown. Rangeomorph elements (sensu Narbonne 2004) are preserved predominantly at the distal tips of the specimens (e.g. Figs. 6.1, 6.2a, 6.3), but are also seen intermittently in proximal parts of some specimens (Fig. 6.2a—c). Only small parts of any one branch are visible (Figs. 6.1—6.3), making determination of detailed aspects of the rangeomorph branching architecture of the crown difficult.

While the rangeomorph branching and the brush structure are preserved in negative epirelief (see Chapter 1), the stem and holdfast are preserved in positive epirelief (unusual for this surface). The fossils were not retro-deformed, as deformation is demonstrably even across the surface (R2 = 0.99; Wilby et al. 2015) and is not more than 23% compression in the direction parallel to the length of the organism (assuming holdfasts were originally circular). Accordingly, the absolute proportions may vary between surfaces with different tectonic histories. The range, average and standard deviation of the proportions of the organisms in this study are presented in Appendix A.2.9.

The complete fossils range from 4 to >12 cm (mean = 7.8 cm, n=9) in total height (taken as the centre of the bulb of the holdfast to the distal edge of the crown, i.e. that furthest from the holdfast disc; Fig. 1). The stem flares directly into a bulbous holdfast (where the latter is preserved clearly), and comprises approximately 30% of the total height the organism. The stem width is approximately 10% of the total height of the organism, and the maximum width of the holdfast is 13% of the total height. The crown is broadly triangular in preserved outline, and comprises between 67 and 85% of the total height of the organism. Its widest point occurs between half and three-quarters of the way up the crown (Fig. 6.1), and is between 77 and 100% of the total height of the organism, and typically between 90 and 150% of the height of the crown.

The brush structure is comprised of a series of undulose, subparallel lineations. They have a consistently paired appearance, with pairs of 0.4—0.8 mm wide grooves internally separated by a lower-relief, 0.05— 0.14 mm wide ridge, and each pair separated by a higher-relief, 0.06—0.18 mm wide ridge. The width of the grooves (taken at their widest point) correlates with the width of the crown (R2= 0.45, p=0.035), and weakly with total height (R2= 0.38, p=0.079). The grooves lack internal divisions perpendicular to the lineations, and otherwise appear smooth in texture (Figs. 6.1d, 6.2c). The grooves are occasionally observed to bifurcate and truncate each other (Figs. 6.1b, c, d; 6.2c, d; 6.3). They also are truncated by 102 A B

D C

Primary branches Fold in brush structure Bifurcation of Secondary branches High-relief ridge grooves Third order branches Low-relief ridge Outer margin Figure 6.2. Undosyrus specimen 5E4 (GSM 105962). A) Photograph of the whole specimen, scale bar = 2 cm; B) dashed inset of a), showing rangeomorph elements exposed along the edge of the left-most visible branch and in the centre of the organism; the margins to the brush structure have the same weight as folds n the structure, and the brush lineations terminate sharply against them; C) digitised camera lucida of b), key to left; D) solid inset of a) showing the brush lineations draping around a disc from another individual, showing the flexibility of the structure. ridges of a similar scale to the higher-relief ridges but which do not form part of the paired groove-ridge set that describes the brush lineations. Where the grooves bifurcate, two lower-relief ridges are initiated, dividing both of the original grooves; the lower-relief internal division between the original pair then becomes a higher-relief division separating two pairs (Figs. 6.1c, d; 6.2c, d). The ridges and the grooves are symmetrical in profile. All lineations which truncate others are subparallel to the main lineations

103 (Figs. 6.1a, b; 6.3a), and are not observed to truncate more than five pairs of lineations (Fig. 6.3a). When truncated, there is no evidence of the truncated lineations continuing beyond the truncating lineation. The ridges appear to emanate from a shared point which is coincident with the region at which the underlying branching structure emerges from the stem (Fig. 6.1a, b). The junction between stem and brush structure is smooth (Fig. 6.1a, b). In addition to mantling the topography interpreted as the underlying branches of the crown, the brush structure is observed to drape over the holdfast disc of another fossil, whose frond is not preserved (Fig. 6.2a, b, d). The specimens have between 15 and 70 pairs of grooves, proportional to the reconstructed height of the individual (R2 = 0.72, p=0.004) and crown width (R2 = 0.58, p=0.01).

Between three and eight underlying branches are observed; this number is proportional to the reconstructed height of the organism (m = 0.451, R2 = 0.6352). The rangeomorph elements are preserved at a topographically lower level than the brush structure, and are always truncated by it (Figs. 6.1c, d; 6.2b, c; 6.3b, c). The width of a third order rangeomorph branch is typically 0.1 mm in width (apart from at the tip of the branch, where they are around half this width). This is comparable to the width of one pair of the brush lineations. Where they are observed, both in distal and proximal parts of the crown, the rangeomorph elements are truncated either side by the lineations of the brush structure, which appear to diverge at these points (Figs. 6.1c, d; 6.2a—c, 6.3b, c).The first order branching structure appears to be displayed and unfurled – neither the sense of inflation nor the concealment of branch axes can be determined. The second order appears displayed, furled or unfurled, concealed, arranged subparallel (i.e. not radiating) and with distal inflation (Figs. 6.1c, d, 6.2b, c, 6.3b, c). The third order branches appear displayed, unfurled, constrained, radiating or subparallel, and with median inflation or no inflation (Fig. 6.2b, c). It is important to note that for most specimens, only a small number of branching characters can be observed (Appendix A.2.9), due to their being obscured by the brush structure.

6.3 Discussion

In this section, features described from the brush are used to ascertain its relationship to the rangeomorph branches, and to determine its structural properties. This is followed by a thorough comparison of the rangeomorph part of Undosyrus to the fossils on the bedding surface to which it bears the closest resemblance, and a critical assessment of the relationship between the rangeomorph part and the brush structure. In the final section, insights into rangeomorph biology which can be gleaned from the presence and nature of the brush structure, as well as its function, are discussed.

6.3.1 Interpretation of specific features unique to the brush structure

The truncations of some brush structure lineations against others and against separate ridges are explainable as folding of the structure (Fig. 6.1). This folding, the sinuousity of its lineations and its mantling over the topography of the branches and over the holdfast of another organism all indicate

104 A B

C

Primary branches Fold in brush structure Secondary branches High-relief ridge Third order branches Low-relief ridge Figure 6.3. Undosyrus specimen 3A1 (GSM 106009). A) Photograph of the whole specimen 3A1, scale bar = 2 cm; stem and holdfast are missing due to the large crack which bisects the specimen, which has removed the fossil- bearing surface below it. B) inset of a), showing rangeomorph elements exposed in the distal portions, and C) digitised camera lucida (presumed second order rangeomorph branching outlined in purple, third order branching outline in pink). a level of flexibility to the structure. That the fold lines are subparallel to the other lineations and only truncate a small number of brush lineations suggest that the folds are quite small in scale, and therefore unlikely to be the cause of the underlying topography. Superimposition of coarser branches beneath/over the finer lineations of the brush structure: the brush structure appears to drape over or under these structures, which are interpreted as the axes of the primary branches or folia. There is no evidence of twisting of the structure, such as lineations truncated by one fold appearing the other side of the fold. The triangular shape of the brush structure, and the fact that rangeomorph elements are so frequently preserved at the distal end, suggests that it had an open top rather than a closed one, which would give a more circular outline. The lineations appear sutured together, given the way they fold, diverge and truncate each other (Fig. 6.1b, c, d). Their behaviour as a coherent, contiguous structure could be explained by a quilted structure (cf. Seilacher 1992), adhesion to each other, or their containment in a thin sheath as has been proposed for Rangea (Grazhdankin & Seilacher 2005).

105 If described using the terminology applied to rangeomorph branching, their paired nature is consistent with displayed rangeomorph branching, their straight edges with furled branching and their lack of internal divisions perpendicular to the brush lineations with undivided branching (Brasier et al. 2012).

The truncation of rangeomorph elements by the brush structure indicates that it was the brush structure that was in closest contact to the casting medium (the substrate surface; Chapter 3): rangeomorph elements are only preserved where they came into contact with the casting medium, in places where the brush structure is not preserved. While this strongly indicates that the brush structure was external to the rangeomorph elements, it is worth exploring the possibility that it records an internal structure. If so, it could arise from preservation of the internal and external structures on different planes – although its pervasive preservation over the specimen would require the entire space between the external and internal parts of the organism to be filled with sediment, and then the layer of sediment preserving the internal rather than the external structure to be expressed on the surface. As the current alignment of these individuals indicates that they were epifaunal, rather than infaunal, this sediment would have to have been injected at the moment of felling and in the opposite direction to the felling flow (assuming that the sediment entered in at the open top). Alternatively, it could be revealed by decay of the rangeomorph elements and outer branches, but this is hard to reconcile with the observation that the majority of the brush specimens are associated with rangeomorph elements for which sub-millimetric, third order branching detail is resolvable. Accordingly, it is concluded that the brush structure was external to the rangeomorph branches, and rangeomorph branches preserved at the distal end are interpreted to have protruded beyond the brush structure as the organism was felled.

If the lower surface of the organism is being preserved (as it appears, section 3.4.2.1), then the brush structure is draping over the branch axes, held aloft above the substrate surface; if the lower surface is being preserved, then it is possible that the branch axes prevented compaction, so that the portions of the brush structure between the branch axes is pushed down further into the sediment than the portions overlying the branches. The truncation by the brush structure of rangeomorph elements that are preserved in more proximal parts on either side, and the fact that the brush lineations appear to diverge at these points, suggests that at these locations, the elements have been extruded through the brush structure onto the substrate surface, perhaps through rips or tears.

6.3.2 Comparison to other multifoliate taxa: Primocandelabrum and Orthiokaterna

Based on the presence of the rangeomorph elements, at least part of the fossil described as Undosyrus must be a rangeomorph. The morphological proportions and branching architecture of the specimens assigned to Undosyrus are similar to those of the two other multifoliate genera on Bed B: Primocandelabrum (Chapter 4) and Orthiokaterna (Chapter 5). There is a wide range of morphology and branching within Primocandelabrum on Bed B, and so it would at first seem not unreasonable to assign an individual 106 with the branching and morphology of Undosyrus to this genus. However, Undosyrus appears to have more than four or five primary branches or folia, whereas Primocandelabrum appears to have two or three (Chapter 4). Moreover, Primocandelabrum and Undosyrus have different holdfast types. All known individuals of Primocandelabrum have a discoidal holdfast, that is, one where there is a sharp and clear junction between the stem and holdfast, and where the margin of the holdfast disc truncates against the margin of the stem (e.g. Fig. 4.1). In contrast, every individual of Undosyrus for which a holdfast is preserved possesses a bulbous holdfast more similar to that of Charnia (Fig. 1.3b), where there is a smooth gradation between stem and holdfast. That the stems and holdfasts of Undosyrus are preserved in positive epirelief as opposed to the negative epirelief preservation of the same structures in Primocandelabrum and Orthiokaterna (Chapters 4, 5) also suggests some other underlying difference between them. However, it is possible to argue a taphonomic explanation for the difference in the sense of relief (positive vs negative) of the holdfast – speculatively, covering of the structures by the brush structure providing greater resistance to compaction and/or decay. Given the uncertainty in determining the number of branches within Undosyrus in particular, the similarity of the overall morphology of Primocandelabrum and Undosyrus is worth further investigation. This is undertaken using the same multivariate statistical approach that was described in Chapter 4.

6.3.2.1 Methods In order to investigate the extent of the morphological similarity between Undosyrus, Primocandelabrum and Orthiokaterna, the morphological data for the three genera was analysed using principal components and hierarchical clustering methods. This technique allows for the three genera to compared using all morphological data in tandem. The FactoMineR package (Husson et al. 2010b) in R (R Core Team 2014) was used for these analyses, as it gave the most successful discrimination of taxa (Chapter 4).

Three iterations were undertaken. The first analyses the three multifoliate genera in order to test whether Undosyrus is more similar to Primocandelabrum or Orthiokaterna. The second includes Undosyrus and Primocandelabrum to determine the similarity between these two genera, as Orthiokaterna is shown to be more different from them than they are from each other. The third iteration includes a categorical variable for holdfast shape. The outsize specimen, Big Bertha, was excluded from all iterations – it places as an outlier to the main Primocandelabrum population and so would exert a strong bias on the analyses (see Appendix 1.1.1). Only partial diagnoses of branching character can be made for any one specimen of Undosyrus because they are obscured by the brush structure. While branching architecture is important in rangeomorph classification, it is possible to distinguish both genera and species based on the sum of morphological proportions alone (Chapter 4), and there is simply insufficient data to include the branching characters of Undosyrus in the analyses here. The smooth junction between the stem and holdfast of Undosyrus mean that holdfast height cannot be determined, and it is impossible to determine

107 A B Orthiokaterna included Orthiokaterna included cluster 1 cluster 2 8C2 3(1) cluster 3 19B1

0.0 1.0 10C3c 3D8 3(2) inertia gain 10C3a 19C4 6a2 l 13A8 10B86A1 6 8B3b5.00E+03 13A4 10A2 2B1 8C1 8C3cluster19A3 2 10C3b 6B2b 5F5 4A35C5 19A1 4A1b (right) 10A7 7A4 8B4 cluster3D7 1 5C1 1B1 5A4 10b10 5.00E+023C3 10C8 19A4 5C6 20Ba 5F3 7A5 5G310B3 Dim 2 (25.74%) 3D4 4B3 3E1(1) 20Ag 11D2 cluster 310A4 11D1 12A1 5.00E+04 5F6 3Bx -3 -2 -1 0 1 2 3 0.0 0.5 1.0 1.5 2.0 6 5E2 5F5 5F6 3Bx 5F3 5E3 5E4 3(1) 3(2) 5A4 6A1 2B1 4B3 7A5 8B4 7A4 1B1 4A3 5C5 8C1 3D7 3D4 3D8 5C6 3C3 8C2 8C3 5C1 5G3 6a2 l 20Ba 19A4 19B1 19A3 19A1 20Ag 10A4 6B2b 12A1 10B8 13A8 13A4 10A7 10B3 8B3b 10A2 11D1 11D2 19C4 10C8 10b10 10C3c 10C3a 10C3b 3E1(1) -6 -4 -2 0 2 4A1b (right) Dim 1 (46.70%)

C Primo. and Undosyrus only D Primo. and Undosyrus only cluster 1 cluster 2 cluster3(1) 3 8C2 cluster 3 19B1 cluster 4 0.0 0.4 0.8 1.2 cluster 5 inertia gain

3(2) 3D8 10C3c 10C3a 6a2 l 10B8 13A8 19C4 6A1 5E3cluster 4 cluster 1 8B3b 2B1 10A2 13A4 10C3b 19A3 3D45F5 4A1b (right) 8B4 5C5 4A3 7A4 5C1 Dim 2 (34.80%) 10A7 cluster 5 20Ag 11D21B1 5E2 10b10 10A410C8 3C3 20Ba 12A110B3cluster7A511D1 5.E42 19A4 5G3 5F3 3Bx 5F6 4B3 -2-101234 0.0 0.2 0.4 0.6 0.8 1.0 1.2 5F6 5F3 3Bx 5F5 5E3 5E2 5E4 3(1) 3(2) 8B4 7A4 7A5 1B1 4B3 2B1 4A3 6A1 3D4 3C3 5C5 3D8 8C2 5C1 5G3 6a2 l 19A3 13A4 10B8 13A8 20Ba 19A4 10A2 12A1 8B3b 10A7 10A4 19B1 10B3 20Ag 10C8 19C4 11D2 11D1 10b10 10C3c 10C3a 10C3b -4 -2 0 2 4 4A1b (right) Dim 1 (37.06%)

E F Primo. and Undosyrus only Primo. and Undosyrus only

cluster 1 cluster 2 6a2 l cluster 3 10C3a 8B4 3(2) 0.0 0.4 0.8 1.2 2B1 10C3b 3D8 19B1 inertia gain 5C5 3(1) cluster 2 0 . 7A4 6A110B8 1 10C8 4A3 10A4 10C3c 12A120Ag 10A2 19A3 4B3 10B35G3 8C2 7A5 5.00E+038B3b 3Bxcluster 1 11D2 1B1 13A8 11D1 5F65E4 19A4

5 19C4

. 5F3 Dim 2 (30.60%) 3D4 5.00E+0210A7cluster 3

0 5F5 3C3

20Ba10b105C1 13A4 4A1b (right) -3 -2 -1 0 1 2 3 0 . 0 3Bx 5F5 5F6 5F3 5E3 5E4 5E2 3(1) 3(2) 1B1 8B4 2B1 7A4 4A3 6A1 4B3 7A5 3D4 5C1 3D8 8C2 5C5 3C3 5G3 6a2 l 19A3 10B3 13A8 19B1 10A4 12A1 19A4 20Ba 10A7 10B8 8B3b 10A2 20Ag 13A4 11D2 10C8 11D1 19C4 10b10 10C3c 10C3a 10C3b -4 -2 0 2 4 4A1b (right) Dim 1 (34.71%)

Figure 6.4. Results of the hierarchical cluster analysis following principal component analyses of Primocandelabrum, Orthiokaterna, Undosyrus and “dusters”. A) Cluster dendrogram and inertia gain plot, and B) map when Orthiokaterna is included; C) cluster dendrogram with inertia gain plot and D) map when Orthiokaterna is excluded; E) cluster dendrogram with inertia gain plot and F) map when Orthiokaterna is excluded and holdfast shape (bulbous or disc) is included as a character. Green specimens numbers indicate Unodsyrus.

108 % assignment Orthiokaterna na na na na na na na na % assignment Undosyrus ber is percentage of individuals with that is percentage ber lues in brackets represent the average value lues in brackets represent qualitative values, first number is percentage is percentage qualitative values, first number % assignment Primo. 0, 0 43.59, 100 0, 0 56.41, 100 Discoidal holdfast 100 na na na na na na na na 100 Bulbous holdfast Crown width Crown height 1.18 (1.04) 0.80 (0.73) 0, 0 56.41, 0.15 (0.10) 1.33 (0.94) na 35.90, 93.33 Stem width at inflection (c-i) 0.20 (0.27) 1.21 (1.04)0.35 0.82 (0.27) (0.73) 0.15 (0.10) na 0.67 (0.73) na 35.90, 100 0, 0 0.21 (0.26) Disc width Stem height 0.24 (0.32) 0.25 (0.31) 0.08 (0.10) 0.74 (0.69) na 88.89, 25 0, 0 0.50 (0.32) 0.56 (0.31) 0.51 (0.94) na 5.13, 20 100, 80 0.17 (0.28)0.56 (0.28) 0.08 (0.10) 0.84 (1.04)0.37 (0.28) 0.37 (0.27) 0.07 0.15 (0.10) (0.10) 0.79 1.70 (1.04) (1.04) 0.63 (0.73) na na na 17.95, 46.67 88.89, 53.33 0.35 (0.27) 0.32 (0.26) 0.64 (0.73) 0, 0 43.59, 0.14 (0.27) 0.85 (1.04) 100,100 0, 0 0, 0 100, 100 all (1) all (2) all (3) excluding Orthiokaterna (1) excluding Orthiokaterna (2) excluding Orthiokaterna (3) excluding Orthiokaterna (4) excluding Orthiokaterna (5) exc. Orthiokaterna, inc.holdfast shape (1) exc. Orthiokaterna, inc.holdfast shape (2) exc. Orthiokaterna, inc.holdfast shape (3) Iteration (cluster number) for the entire dataset, and the value not in brackets is the average value for the character within the specified cluster. For For within the specified cluster. the character dataset, and the value not in brackets is average for the entire for (not including imputed, i.e. originally missing, values), and the second num with specified character of individuals in cluster included in the cluster. character Table 6.1. Results of the heirachical cluster analysis following principal components analysis. For quantitative values, the va analysis following principal components analysis. For 6.1. Results of the heirachical cluster Table

109 the extent to which the brush structure might have obscured the left and right branches. Accordingly, five independent morphometric measurements (disc width, stem height, stem width at inflection, crown width and crown height) were used in these analyses (see Appendix 1.1.1 for discussion).

The number of clusters and their descriptions were determined using the hierarchical clustering technique following principal components analysis described in Chapter 4. As before, the number of clusters was determined by the greatest gain in within-group inertia as depicted in the plots of inertia gain (Husson et al. 2010b). For the hierarchical clustering analysis following principal components analysis, the success of assignment to each cluster based on branching characters was assessed through the percentage of individuals displaying a particular character assigned to the cluster described by that character (100% indicates that all individuals which display that character are placed in the cluster), and the percentage of individuals in the cluster which displayed that character (100% indicates that all individuals in the cluster display that character).

6.3.2.2 Results Undosyrus plots in the same region of the principal components space as P. boyntoni, when Orthiokaterna is included (Fig. 6.4b) and excluded (Fig. 6.4d, f). When Orthiokaterna is included, the first dimension positively correlates to crown height and crown width, and negatively to disc width and stem height; the second dimension correlates positively to stem width, disc width and crown width; the third dimension correlates positively to disc width and negatively to stem inflection. When Orthiokaterna is excluded, the first dimension correlates positively to stem height and disc width, and negatively to crown height; the second dimension correlates positively to crown width, stem width and disc width; the third dimension correlates positively correlates to stem width and negatively to disc width.

When Orthiokaterna is included, Undosyrus plots within the same region of the principal components space as P. boyntoni (Fig. 6.4b). Three clusters are supported by inertia gain, and eight out of nine individuals assigned to Undosyrus plot in cluster 2 with 24 P. boyntoni specimens. This cluster is characterised by tall crowns, short and narrow stems, and small discs (Table 6.1). The remaining Undosyrus specimen (3D4) plots in cluster 3 with the remaining P. boyntoni specimens, bar two individuals (Fig. 6.4a, b). This cluster is characterised by wide crowns and wide stems (Table 6.1). Orthiokaterna plots into cluster 1 with the last two P. boyntoni specimens. When Orthiokaterna and “dusters” are excluded, three clusters are most strongly supported by inertia gain, although the division between five and six clusters is of a similar scale (Fig. 6.4c). When the individuals are placed into three clusters, eight out of nine specimens assigned to Undosyrus are placed into cluster 1 along with 23 individuals assigned to P. boyntoni (Fig. 6.4d). This cluster is characterised by slightly tall and narrow crowns, short and narrow stems, and small discs (Table 6.1). The remaining Undosyrus specimen (3D4) plots in cluster 3 with 11 P. boyntoni individuals. This cluster is characterised by tall stems and short and narrow crowns. When placed into

110 five clusters, eight out of nine Undosyrus specimens plot with 7 specimens of P. boyntoni in cluster 2. This cluster is characterised by narrow stems, narrow crowns and small discs. The remaining Undoysrus specimen (3D4) plots with 6 P. boyntoni specimens in cluster 4 (Fig. 6.4d). This cluster is characterised by long, wide stems, and short crowns (Table 6.1). 3D4, the specimen which consistently plots away from the other specimens of Undosyrus is distinguished by its proportionally long and wide stem (45% of the total height of the specimen, compared to an average value of 27%, 1 standard deviation = 9%). All other proportions fall within one standard deviation of the means. When holdfast shape is included in the analyses, Undosyrus is placed in a separate group both when Orthiokaterna is included and when it is excluded (Fig 6.4e, f). Despite this, the individuals of Undosyrus place in a similar region of the principal components space, both to each other and to Primocandelabrum specimens (Fig. 6.4f).

6.3.2.3 Discussion of results The morphology of individuals assigned to Undosyrus places them with P. boyntoni (Chapter 4), plotting in the same region of the principal components space. However, their branching structure (where possible to determine it) is most similar to P. anatonos (Chapter 4), with which one specimen of Undosyrus (3D4) clusters based on its morphology (Fig. 6.4d). The separation of Undosyrus into a separate category based on its holdfast shape is not surprising – all Undosyrus individuals plot towards the edge of the Primocandelabrum cluster (Fig. 6.4b, d, f), and inclusion of a category which discriminates the two groups 100% (which no other character does, Table 6.1) will inherently have a strong influence on the clustering.

In summary, when the brush structure is ignored, Undosyrus can only be distinguished from P. boytoni based on its holdfast shape. This raises two main possibilities. Firstly, the brush structure was not part of the rangeomorph with which it is associated, i.e., it is some other structure or organism superimposed on a few individuals of Primocandelabrum – therefore, the association is secondary or coincidental. Secondly, the brush structure was an integral part of the rangeomorph in life. In this case, either it was expressed by only some Primocandelabrum individuals; it was taphonomically lost from those that appear to lack it; or the rangeomorph associated with the brush structure was not P. boyntoni.

6.3.3 Hypothesis 1: the brush structure was not part of the rangeomorph

It is important to first establish that the brush structure is biogenic. Current lineations, ripple crests, and rill marks are not consistent with the symmetrical profile of the ridges and grooves, their close and regular spacing, and their consistent occurrence in pairs; tool marks are further rejected by the lack of a terminal depression. Erosion of cohesive, partially-lithified surficial mud has been reported to give rise to undulose, lineations a millimetre or so in depth and a few millimetres wide (cf. Schieber et al. 2010) – a scale similar to that of the brush lineations (allowing for compaction). However, the undulating

111 trace of the brush lineations, and the fact that they truncate each other (Figs. 6.1, 6.2) count against this interpretation.

The possibility of its representing a microbial mat fabric is intriguing – perhaps as a microbially-induced sediment structure (MISS; Noffke et al. 2001, Noffke 2009) onto which a rangeomorph was superimposed. The brush structure bears a resemblance to Arumberia Glaessner and Walter (1975), a surficial feature which consists of subparallel, undulose lines. Although originally considered to be a discrete biological taxon (Bland 1984, Glaessner & Walter 1975), this feature has since been interpreted to have formed from the action of currents on cohesive, perhaps microbially-bound, substrates (McIlroy & Walter 1997). The spacing of the lineations of Arumberia and Undosyrus overlap (0.05–0.7 mm for the brushes, and 0.3–7.0 mm for Arumberia; Glaessner & Walter 1975). However, the regular spacing and paired nature of the lineations, the sharp margin of the brush structure, and the fact that the width of the grooves between the lineations correlates with height of the structure (section 6.2), all serve to distinguish the lineations of Undosyrus from those of Arumberia. Less conclusively, there are several features expressed by the brush structure which are not entirely consistent with an Arumberia interpretation, although they are not noted from all Arumberia-type specimens: the brush lineations are subparallel rather than radiating, as they are in Arumberia (Glaessner & Walter 1975); no lineations are observed behind the apex of the brush structure; and the brush structures do not overlap (common in Arumberia; McIlroy and Walter, 1997). Regardless of the process(es) that form Arumberia-type structures, the cause for the exclusive association of the structure with such a small morphological subset of the rangeomorphs on Bed B would require explanation by an as-yet unknown causal link. No similar relationship between Arumberia and any organism has been documented elsewhere.

6.3.3.1 Hypothesis 1a: the brush structure is a growth of another organism on a rangeomorph The brush structure could record a symbiotic association between rangeomorph and a microbe such as a fungus or filamentous bacteria. The classic symbiotic association of organisms of a non-metazoan grade is, of course, that of the lichens – of which a few taxa are known from modern deep marine environments. Ediacaran organisms have been interpreted as terrestrial lichens (Retallack 1994) and, while the depositional environment of these organisms is unambiguously sub-tidal (Chapter 2), it is worth exploring the possibility that the organisms represent a marine fungal symbiosis similar to lichens. Although fungal structures are at least an order of magnitude smaller than the brush lineations, they can form a dense network similar in appearance to the brush structure (e.g. Visagie et al. 2009). Many marine fungi act as pathogens or parasites on other organisms, rather than as symbionts (Kohlmeyer & Kohlmeyer 1979, Porter 1986), and so perhaps the brush structure represents a (fungal) parasite on a rangeomorph. The siliclastic settings in which Undosyrus is preserved do not preserve the cellular level detail required to identify diagnostic features of fungi such as hyphae with perforated cell walls, or to

112 distinguish microbial from eukaryotic cells. While intriguing, it is not possible to conclusively reject or accept this hypothesis based in the available evidence.

In a similar vein, could the brush structure record a decay-related microbial growth? In this case, the brush structure would represent a different style of taphomorph to those currently recognised in the Ediacaran. Taphomorphs are usually expressed by a continuum of forms (here it would be from brush structure absent to brush structure present), but this is not observed in the fossils of Undosyrus, and the fine resolution of the rangeomorph elements (Figs. 6.1–6.3) is difficult to reconcile with a (long) period of decay prior to preservation (Liu et al. 2011).

6.3.3.2 Hypothesis 1b: the brush structure is another taxon superimposed on a rangeomorph The brush structure is also broadly comparable to the lineated pattern seen on Haootia quadriformis, an intriguing form recently described from the Bonavista Peninsula in Newfoundland that has been interpreted to record cnidarian (or cnidarian-grade) muscle tissue (Liu et al. 2014). Like the brush structures, Haootia is associated with a holdfast and a stem, and the lineations are undulose and truncate (overlap) each other. Unlike the brush structure, however, the lineations in Haootia do not occur in pairs, bundle together at their distal ends (relative to the holdfast) and, critically, are not associated with rangeomorph elements.

Two groups of Ediacaran taxa, erniettomorphs and dickinsoniomorphs consist of subparallel modular, tubular units which are similar in scale to the lineations of the brush structure (Laflamme et al. 2013). However, there is no evidence for either a central suture or rib in the brush structure, or for multiple vanes as there are in erniettomorphs and dickinsoniomorphs. This lack of a central midline also prevents direct comparison to the Cambrian taxon, Stromatoveris (Shu et al. 2006); additionally, despite sub- millimetric preservational resolution in the brush structures (Figs. 6.1, 6.2, 6.3), there is no evidence of the millimetre-scale zooids or polyps which are observed in Stromatoveris. The lack of these structures in rangeomorphs are the principal argument against this group being suspension feeders (Liu et al. 2015a).

6.3.4 Hypothesis 2: the brush structure was part of the rangeomorph

A different taphonomic explanation (rather than a decay-related one) could be that the brush lineations represent lines or stretch-marks induced in the rangeomorph through torsion, following tugging and/ or stretching of the organism prior to death. Such structures have been interpreted for holdfasts from Australian material (Tarhan et al. 2010), and in the stalks of two Orthiokaterna individuals (section 5.5.2). While in each of the cases the lineations are irregular in both spacing and relief, they taper to either end; this is seen in the brush structure lineations at their points of divergence, but not their points termination. If the brush structure lineations terminated in rangeomorph elements, it might be possible

113 that the lineations represent stretched or unravelled rangeomorph branches – however, the truncation of rangeomorph elements by the brush structure indicates that they are clearly separate (i.e. not attached to the rangeomorph elements themselves).

If the brush structure was instead an original part of the organism, one explanation might be that it records a teratology, where the brush structure represents perhaps rangeomorph elements which have suffered an altered growth structure. Undosyrus individuals attain heights of more than 12 cm (Appendix 2.9) – unusually, the teratology cannot have been severely deleterious. If the organism is Primocandelabrum, then while it is confined to a small subset of one variety of one taxon (as currently described from morphology and branching), it is very common in that variety. Also, not all rangeomorph elements are afflicted, not are those that are so located randomly or in one location – all lineations appear to emerge from apparently the same locus, in a continuous ring around the stem of the organism – remarkably controlled for a teratology.

Sheaths or membranes have been described from several Ediacaran organisms, including Rangea and Avalofractus, although they are typically featureless and smooth, and emanate from the basal holdfast (e.g. Grazhdankin & Seilacher 2005, Narbonne et al. 2009, although see Brasier et al. 2013). Rangea, the youngest rangeomorph known, and has undergone several reinterpretations (most recently in Vickers-Rich et al., 2013). In one of these, it has been reconstructed with an external sheath (Grazhdankin & Seilacher 2005), which bears “faint branch-like structures”. These are interpreted to record the places where the rangeomorph branches or quilts touched the membrane, perhaps in a similar manner to the underlying branch structures creating the topography over which the brush structure appears to drape (Fig. 6.1a, b). Intriguingly, small, curvilinear structures of a similar scale (0.75—1.54 mm ridges, separated by 0.25 mm wide grooves, as measured from published photographs Elliott et al. 2011, their Fig. 7) to the brush lineations have been reported from Farm Aar in Namibia, and interpreted as “membrane-like structures” (Elliott et al. 2011). Although there is no defined margin to the area in which these lineations in the “membrane-like structures” occur, and they lack the paired arrangement observed in the brush structure (Figs. 6.1, 6.2), they do at least indicate the possibility that textured or ornamented membranes could have existed in the Ediacaran. An external membrane or sheath is also consistent with the taphonomic evidence suggesting that the brush structure was external to the rangeomorph branches (section 6.3.1). It is therefore reconstructed as an external sheath (Fig. 6.5).

6.3.5 Was the rangeomorph Primocandelabrum, or another taxon?

If the brush structure was a part of a Primocandelabrum, then its absence from most individuals assigned to that taxon could be explained by disarticulation, where the structure was removed prior to smothering beneath the burial sediment. This disarticulation could have been active (i.e. abscission), or passive (taphonomic). In either case, the fact that no brush structure is preserved except in association with the 114 Figure 6.5. Artist’s reconstruction of Undosyrus nemoralis. The rangeomorph elements are protected by an external sheath. The base of the bulbous holdfast would have been buried to a shallow depth within the sediment substrate. rangeomorph would require all removed brush structures (from at least 53 otherwise well-preserved individuals, Chapter 4) to have been shed a long enough time before burial for them to have completely decayed (if active), or to have been carried out of the area of the bedding plane which is preserved (if active or passive). However, what process would have caused so select few individuals to have lost this structure?

Perhaps the brush structure was only expressed by a few P. boytoni individuals in life? It could represent an ecophenotypic variation but, as discussed in Chapter 5 for Orthiokaterna and Bradgatia, there is no evidence for the local environmental variability which would be needed to induce such variation. Having already dismissed a teratological explanation, the most plausible explanation is that the structure reflects a feature of one particular life stage. Individuals bearing the brush structure range from 4 to >12 cm (Appendix 2.9), a size range that overlaps with that of Primocandelabrum individuals. Interestingly, if size correlates with age (there are, admittedly, reasons why it might not, such as neighbour shading; Glynn 1973, Merz 1984, Okamura 1984, Huston & DeAngelis 1987, Barry & Tegner 1990; see Chapter 7, 10),

115 then individuals reached the point in their life cycle at which the brush structure is expressed at different sizes or ages. It is also interesting that no similar structure is observed on any of the equally numerous taxa on the surface, even those with otherwise broadly similar morphology (such as Orthiokaterna).

Rangeomorph taxonomy is fraught with difficulty, and remains under revision (Chapter 4): currently, genus-level discrimination is used for major differences in morphology and branching, and species- level discrimination is used for small variations in overall proportions or shape (Brasier et al. 2012). While the proportions of the holdfast, stalk and crown of Undosyrus, and its branching pattern, are all within the variation documented for Primocandelabrum (Chapter 4; Appendix 2.4, 2.6), possession of the brush structure, different number of folia and a different type of holdfast are enough by themselves to warrant to exclusion into a separate taxon (cf. Liu et al. 2015b). Each of the above explanations for the expression or retainment of the brush structure by only a few individuals of P. boytoni requires some special pleading, especially given the very similar morphology and branching pattern (where discernible) of the individuals which bear that structure. On the balance of evidence, it seems more parsimonious that individuals which bear the brush structure do belong to a separate genus.

6.3.6 The purpose of the brush structure, and its bearings on rangeomorph biology

It is concluded from the above points of discussion that the structure must have served some function and, importantly, may be able to tell us something about the biology of these organisms. Rangeomorphs are typically accepted to have gained their nutrition through osmotrophy (the absorption of dissolved organic carbon through an outer membrane), although it is not clear whether this is true osmotrophy (as for fungi), or whether it is phagotrophy or some manner of suspension feeding via small particles (Liu et al. 2015a). It seems initially counterintuitive for an osmotroph or a filter feeder to place a barrier between the feeding structures and the water column (the source of its food) – what benefit could such a structure have provided to the organism? The lack of large modern obligate osmotrophs makes direct comparison of such a structure with extant organisms impossible. However, useful comparisons can be made with suspension feeders, especially given that this is still a possible mode of feeding for rangeomorphs (Liu et al. 2015).

Structures bearing a similar appearance and scale to the brush structure have been described from the trilobed form, Ventogyrus, from the Ediacaran of the White Sea. These structures are argued to be internal, forming a branching channel or duct system that was impressed on the outer tissue and preserved on a different plane to it (Fedonkin & Ivantsov 2007) – the organism has been interpreted as a siphonophore on the basis of these structures. While the brush structure is arguably external to the rangeomorph elements (section 6.3.1), and lacks a central rib, suture or duct (Figs. 6.1, 6.3), it is possible that the brush structure performed some similar supportive or transportive function.

116 Alternatively, the brush structure could represent one way to maximise its feeding efficiency, similar to that seen in sponges, certain bryozoa and tunicates. While there is no evidence in rangeomorphs for the cilia or musculature required to actively generate a flow in the manner of these extant organisms, the brush structure may still have been capable of providing some of the benefits of internalising feeding structures. The topography the brush structure creates in the water column (a ridge followed by a depression) would have disrupted the flow over the structure, potentially inducing the formation of vortices even without an active pumping system (Sansón & van Heijst 2015). This could have slowed down the speed of water flow inside the sheath. A similar result could also have been achieved if the brush structure was porous, although these pores would have had to be smaller than the finest resolvable detail (<0.1 mm). Slowing the flow in this manner would have given the rangeomorph elements more time to absorb nutrients from the water, or to provide sufficient time for rangeomorph elements to secrete digestive enzymes, for these to degrade recalcitrant organic matter and then for the elements to absorb the digested products (Chapter 9).

In modern marine communities, sediment fouling represents a major source of stress to immotile, benthic organisms, and there is no reason to assume that this would have been different for organisms in the Ediacaran (Chapter 8, 9). Sediment particles can clog feeding structures, irritate tissues and damage the organism through physical abrasion. Whether the rangeomorphs gained their nutrition via osmotrophy or filter feeding, they would almost certainly have suffered from sediment stress either way: sediment influx has been demonstrated to have had a significant influence in structuring these ancient communities (Wilby et al. 2015; Chapter 9). Accordingly, it is not unlikely that rangeomorphs would have developed strategies for coping with sediment stress. Strategies for coping with this stress adopted by extant organisms include entrapment of grains in a mucous layer which is periodically sloughed; enclosing the feeding structures within a hard structure (e.g. the shell of brachiopods, the test of tulip worms); and retracting the entire body of the organism into the substrate during times of string flow (non-skeletal filter-feeding worms, sea pens). Although there is no evidence in rangeomorphs for the musculature required to retract the organism into the sediment, nor is there any evidence of biomineralisation until later in the Ediacaran (Grant 1990), rangeomorphs may still have been capable of protecting their feeding structures. In some rangeomorphs, this may be achieved by furling the elements so that the growing tips are protected (section 4.4.1; Chapter 8), but an alternative way would be to encase the entire feeding crown in a sheath composed of soft tissue, similar to the interpretation for the Cambrian filter feeder, Siphusauctum gregarium (O’Brien & Caron 2012). This strategy would additionally have enabled the individual to exploit the lower levels of the water column (which are typically underpopulated in environments facing comparatively high sediment stress; Chapter 8), and also potentially to avoid the strain imposed by the faster moving currents higher above the substrate surface. Accordingly, the primary function of the brush

117 structure is interpreted to have been protective, representing a novel adaptation to environmental stress in these unexpectedly dynamic environments (Chapter 8, 9).

Whatever its function, the brush structure has a number of implications for interpreting the biology of rangeomorphs. Differentiation of structures in this manner mean that the brush structure could be interpreted as a different tissue type – so little is known of the original construction of rangeomorphs. Plastic growth has been described from Charnia masoni (Wilby et al. 2015), and appears to be particularly prevalent in multifoliate rangeomorphs (Chapter 4, 5); the anomalous eccentric growth which is common in Orthiokaterna (Chapter 5), and to a lesser extent in Bradgatia and P. boyntoni, has not yet been recorded from unifoliate rangeomorphs. This aspect of their growth is interpreted as a response to damage that was quite likely caused by sediment-related abrasion (see Chapter 5). Undosyrus is a multifoliate rangeomorph and, given the similar appearance and scale of the brush structure lineations to that of the rangeomorph branching elements (Figs. 6.1, 6.2, 6.3), it is possible that the brush structure arose through modification of rangeomorph elements, and perhaps their suturing together to comprise a coherent structure. The brush structure therefore is interpreted to represent a more controlled aspect of the plastic and malleable architecture of rangeomorph branches. Although not strict tissue differentiation, the modifications of existing structures in this manner may give insight into the grade of organisation of rangeomorphs.

There is a remarkable similarity between the brush structure and the holdfast of an undescribed gorgonian from the Devonian of the Czech Republic (Mikulas & Pek 1995): the lineations and the specimen as a whole are of a similar scale. It is likely that if the brush structure were felled perpendicular rather than parallel to the substrate surface, it would appear to radiate out from a central boss or point, as do the lineations of the gorgonian holdfast (cf. Liu et al. 2014). While the observation that the brush structure emerges from a stalk which is itself attached to a holdfast, and the occurrence of rangeomorph branches at the distal end, precludes interpretation of the structure itself as a holdfast, the similarity in appearance of the two structures is intriguing. Parallels have been drawn between soft corals and rangeomorphs before, based on their overall appearance and habitat (Dzik 2002). While the direction of growth in the two groups is demonstrably opposite (Antcliffe & Brasier 2007), the similarity of the brush structure to that of gorgonian holdfasts may indicate a similar level of tissue/cell differentiation to the Octocorallia (the subclass that includes soft corals, sea pens and gorgonians).

Rangeomorphs have alternatively been argued to represent a similar grade of organisation to fungi (Peterson et al. 2003). The organism is unlikely to belong to crown-group fungi – their fruiting bodies are susceptible to abrasion from currents or sediment and, even the deep marine realm, they only reach a few mm in height (Kohlmeyer & Kohlmeyer 1979). Additionally, the few free-living marine fungi forms that are known from the deep marine realm are saprophytic on wood or chitin, requiring high rates of 118 dissolved oxygen (Kohlmeyer & Kohlmeyer 1979) – there are significant problems with a fungal-style osmotrophic mode of feeding for rangeomorphs (Chapter 9). However, the modification of parts of the rangeomorph into a external sheath and stalk could represent a similar level of differentiation to that observed in the fungal group, basidiomycetes (Kües & Liu 2000), or to the development of the cap in Acetabularia (Dassow et al. 2001, Henry et al. 2004), providing additional support for the rangeomorphs representing this level of organisation.

6.7 Conclusions

A new rangeomorph taxon is described from Charnwood Forest. It bears a novel feature, the brush structure. While it is possible that this structure represents a particular life-stage, a teratology, a taphomorph or even a microbial parasitic/symbiotic association, it is here interpreted to record an external sheath enclosing a rangeomorph of similar morphological proportions and branching architecture to P. boyntoni. Based on their similarity of scale, this structure is inferred to have developed through modification of rangeomorph branches: they would be classified as displayed, furled and undivided. This represents the first modification of existing tissues into a second tissue type (at least with a different function) in the Ediacaran, and further highlights the plastic nature of the rangeomorph growth architecture.

119 7. TIERING IN AVALONIAN RANGEOMORPH-DOMINATED COMMUNITIES

7.1 Introduction

Tiered community structures typify many Phanerozoic and modern communities, and develop as a response to competition for vertically distributed resources, on a variety of scales. In terrestrial forest communities and those of marine photoautotrophs, light is the limiting factor (Foster 1975, Cannell & Grace 1993), and tiering develops as photosynthetic organisms strive to out-grow their neighbours, or adapt to low-tier (low-light) niches (such as ferns). Light accessibility also drives morphologies towards those which minimise self-shading (Enríquez & Pantoja-Reyes 2005). In heterotrophic benthic marine communities, suspended particulate matter (delivered to the system principally via horizontal flows; Bottjer and Ausich, 1986) is the limiting nutrient (Brett & Liddell 1978, Watkins 1991). Consequently, benthic communities develop tiered structures as a means of partitioning these nutrients (Bottjer & Ausich 1986). Size distributions of rangeomorph populations on three Ediacaran bedding surfaces in Newfoundland (Mistaken Point D and E, and Lower Mistaken Point) have been taken as evidence of ecological tiering (Clapham & Narbonne 2002). In the Clapham and Narbonne study (2002), tier boundaries were found to be gradational, but the heights at which they occurred were used to imply a metazoan affinity for rangeomorphs (Clapham & Narbonne 2002, Narbonne 2005). The tiered structures interpreted for the communities have in turn been argued to support a filter feeding mode of life for rangeomorphs (Clapham & Narbonne 2002) or, more recently, an osmotrophic one (Ghisalberti et al. 2014).

In Phanerozoic communities, organisms adopt either a “colonial” tiering habit, in which they feed along their entire length, or a “solitary” tiering habit, in which they possess a specialised feeding structure which occupies a single tier, usually supported by a stem or stalk (Bottjer & Ausich 1986). Although the majority of organisms adopting a colonial tiering habit are colonial, the terms do not necessarily imply a colonial/solitary grade of organisation (Bottjer & Ausich 1986). In rangeomorphs, Charnia (Ford, 1958) and Trepassia (Brasier et al. 2012), which bear rangeomorph elements along their entire length, are examples of colonial tiering – the discussion as to whether or not rangeomorphs are colonial (comprised of numerous, separate individuals), or are merely modular remains uncertain and is a point of some contention (Dzik 2002, Xiao & Laflamme 2009). The oldest claim to the separate stalk/feeding structure specialisation is for Beothukis plumosa (Liu et al. 2015b), formerly Culmofrons (Laflamme et al. 2012b), and is exemplified by the abundant stalked taxa found on Bed B of Charnwood Forest (Wilby et al. 2011). It is likely that only the frondose part of the organism (with its high surface area; Chapter 5), rather than its stem, played a role in absorption and consequently exerted a tiering pressure. Therefore, while it is possible to use total height to determine the extent of tiering where all taxa which share a tiering habit (Bottjer & Ausich 1986), comparing the total heights of taxa with different tiering habits 120 Figure 7.1. Rangeomorph morphotypes. A–C) Stalked morphotype: A) Orthiokaterna, GSM105875; B) Primocandelabrum, GSM105969b; and C) Charniodiscus from Charnwood Forest, GSM106069; D) Upright, non- stalked morphotype: Charnia from Charnwood Forest, GSM105988; E–F) Flat-lying morphotype, E) Fractofusus and F) Pectinifrons, both from Newfoundland (photographs courtesy of A. Liu). All scale bars 2 cm, a—d are photographs of casts housed at the British Geological Survey.

(i.e. those with and without a naked stem) is misleading. This study attempts to improve on previous studies of tiering in Ediacaran organisms by taking its tiering habit into account, comparing only the range of the water column occupied by the frond.

7.2 Material

Specimens from Bed B from Charnwood Forest and seven bedding planes from Newfoundland were analysed for size data. Casts of fossils from Bed B and their specimen numbers are listed in Appendix 2.1. Bedding planes from Newfoundland (Chapter 2, Fig. 2.3) analysed for this study include the 121 Mistaken Point D, E and G (MPE, MPD, MPG) surfaces; the Pigeon Cove (PC), Bristy Cove (BC), Lower Mistaken Point (LMP) and Shingle Head (SH) surfaces in the Mistaken Point Ecological Reserve and environs (Clapham & Narbonne 2002, Darroch et al. 2013). Retro-deformed specimen heights for these surfaces were taken from the Clapham dataset (Clapham & Narbonne 2002, Clapham et al. 2003, Darroch et al. 2013). While retro-deformation removes the effects of shearing, there is no way to account for total compression; as such, only trends (rather than absolute values) can reliably be compared between surfaces. Measurements for Bed B were not retro-deformed before analysis, as strain has been shown to be have affected all individuals equally (Wilby et al. 2015).

7.3 Methods

In order to analyse the influence of tiering habit on community structure, taxa present were identified to genus level and grouped into three morphotypes.

1) Flat-lying (e.g. Fractofusus, Pectinifrons; Fig. 7.1e, f). This morphotype includes all rangeomorphs that are reconstructed as recumbent or prone against the substrate in life. This reconstruction is based on four lines of evidence (Gehling & Narbonne 2007, Laflamme & Narbonne 2008, Liu et al. 2015a): 1) lack of current alignment when other organisms are current-aligned; 2) lack of stem or holdfast in any known specimens; 3) bipolar growth; 4) even preservation across the frond – the comparatively poor preservation of lateral margins in other fronds has been taken to record ash that settled between the preserving surface and the frond during felling, and correspondingly clear lateral margins are taken to record organisms that were already against the preserving surface (Laflamme & Narbonne 2008). 2) Stalked (Fig. 7.1a—c). Stalked taxa are defined as those with a stem (Chapters 4, 5). This includes rangeomorphs (e.g. Orthiokaterna, Primocandelabrum, “dusters”) and the arboreomorphs (e.g. Charniodiscus procerus). Beothukis mistakensis and B. plumosa (formerly Culmofrons (Laflamme et al. 2012b, Liu et al. 2015b) have stems, although they are not always preserved in B. mistakensis and the proportional length of the stem is highly variable (Liu et al. 2015b), and so are included in this morphotype. 3) Upright but non-stalked (e.g. Charnia, Vinlandia, Trepassia; Fig. 7.1d). This category includes all organisms reconstructed as having been upright or gently reclined in life, but that lack a discrete stem. For the main part of this study, only the portion of the water column occupied by that part of the organism which bears the frondose branches was considered. Thus, for stalked organisms, the height in the water column was defined as between the top of the stem and the top of the frond. For the Bed B specimens, the top of the stem was taken as the point of inflection (Chapter 4), as it was not possible to determine the exact point at which the branches emanated from the stem in the majority of fronds.

122 These measurements were taken on all well-preserved specimens, that is, those for which both the distal margin and the centre of the holdfast could be identified, and for which a taxonomic assignment and likely in-life orientation could be determined (n = 146). Where stem height was included in the Clapham dataset it was incorporated into this study. Where it was not recorded for “dusters” an arbitrary stem height of 1/3 total height was used (comparable to recently published values for one taxon within the “dusters”; Mason and Narbonne, In Press). Fractofusus is variably reconstructed as having two (Liu et al. 2015a, Mitchell et al. 2015), or three (Gehling & Narbonne 2007, Narbonne et al. 2014) vanes/rows (Chapter 1). A generous estimate of the height above the seafloor of one half the width of the specimen was used for Fractofusus, allowing for the possibility of a third vane – the same proportion as that used by Clapham and Narbonne (2002). The maximum estimate for the height of Pectinifrons in the water column is taken as its width. One c. 60 cm long fossil was excluded from the Bed B community due to the poor preservation of its lower half and consequently the uncertainty of its morphotype. The single 1.8 m specimen of Frodondophyllas (Bamforth & Narbonne 2009) was excluded from the analysis of Lower Mistaken Point. Iveshediomorphs were not included in the study, as these are generally accepted to represent either microbially-induced sedimentary structures (Laflamme et al. 2012a), or the decaying remnants of dead organisms (Liu et al. 2015a, Liu et al. 2011). As such, while they may themselves provide a local source of nutrients, they are no longer in competition with the living organisms in the community (Liu et al. 2015a). The non-rangeomorph Thectardis is included in the whole community distribution but not in the upright, non-stalked morphotype.

To visualise the density of frondose parts in the water column at a particular height, the water column was divided into 1 cm bins, and the number of individuals whose frondose parts occurred in each 1 cm bin was plotted against height in the water column. Distribution curves were generated for individuals grouped by taxon, by morphotype (flat-lying, upright frond or stalked frond) and for the whole community (Fig. 7.2).

To investigate the influence of including tiering habit on the analyses, density distributions of the total heights of each classification grouping, morphotype and the whole population were plotted in R using the sm package (Bowman & Azzalini 2014). Notched boxplots of this data were constructed using ggplot2 (Wickham 2009), and allow determination of the overlap of the medians of the samples (Chambers et al. 1983), and of their interquartile ranges (IQR, that is, the range between the 25th and 75th percentiles, in which 50% of the individuals plot). For these analyses, visual inspection of density distributions was used to first assess the degree of overlap between taxa and morphotypes. The overlap of the medians was determined through visual inspection of notched box-plots, where failure of the notches to overlap provides strong evidence for the medians not overlapping within 95% confidence (Chambers et al. 1983). The overlap of IQRs was similarly determined through visual inspection.

123 7.4 Results

In order to analyse the degree of tiering for each taxon and each morphotype in individual communities, the results below are presented for each bedding surface (section 7.2). The distribution of frondose mass in the water column is presented in Fig. 7.2. The peaks in this distribution of frondose discussed below refer to that part of the water column in which the greatest numbers of individuals in a given taxon, morphotype or community have their frondose parts (Fig. 7.2): this part of the water column is the most densely-occupied by rangeomorph branches (the absorptive surfaces). The centimetre value of the peak corresponds to height above substrate (the left axis of the graph). The density distribution of the total heights of all taxa and morphotypes are presented in Fig. 7.3, and notched boxplots for the distributions in Fig. 7.4. Values for the median, inter-quartile range (IQR), and the range and peak of frondose part distributions are presented for each taxon and morphotype in Appendix A.2.10, together with a list of the other taxa/morphotypes with which these values overlap.

7.4.1 Bed B

When only the distributions of frondose parts in the water column are considered (i.e. to take account of tiering habit), the distribution for the whole community ranges from 0 to 46 cm, and peaks c. 5—6 cm above the substrate surface (Fig. 7.2 a). Flat-lying fronds are notably absent from this community. Stalked have frondose parts from 0 to 46 cm, with a peak in the distribution at 5—6 cm. Upright, non-stalked forms range from 0 to 35 cm, and their distributions peaks at 0—2 cm. While the two morphotypes overlap in their absolute range, their peaks are offset by 3 cm. The distribution for Orthiokaterna ranges from 5 to 38 cm, peaking at 7—8 cm, while the Primocandelabrum distribution ranges from 1 to 46 cm, with a peak at 5–6 cm. Undosyrus ranges from 0 to 17 cm and peaks at 5—6 cm; and Bradgatia ranges from 0 to 15 cm and has a broad peak from 0 to 5 cm. “Dusters” range from 4 to 9 cm, and peaks between 5 and 8 cm. The range and peak of Bradgatia and Undosyrus overlap with each other, with Primocandelabrum and with dusters; the peak of “dusters” also overlaps at its higher end with Orthiokaterna. Charnia ranges from 0 to 35 cm, and its peak is the lowest on the surface, at 0—2 cm. The Charniodiscus distribution ranges from 0 to 33 cm, and has a broad peak between 0 and 6 cm. The peaks of Bradgatia, Charniodiscus, Primocandelabrum and “dusters” overlap with each other and with the peak of the stalked morphotype, while the peak of Charnia overlaps with that of Bradgatia and Charniodiscus. The peak of Charniodiscus overlaps with that of Charnia at the lower end, and with Bradgatia, Charnia, Undosyrus and Primocandelabrum at the higher end. The upright, non-stalked morphotype peak overlaps with those of its three constituent taxa, but not with the peak of the stalked morphotype. The peak of Orthiokaterna overlaps with no other taxa.

When total heights are considered (Fig. 7.3a), the whole community has one main left-skewed peak at 7—8 cm, and a number of smaller sub-peaks or bumps at greater heights in the water column. The

124 ABCharnwood Forest Bed B B Lower Mistaken Point surface 50 50 Bradgatia “Charnia” Charnia 45 45 “Charnia 2” Charniodiscus Charniodiscus Duster 40 40 Primocandelabrum Duster Fractofusus 35 Orthiokaterna 35 Undosyrus Ostrich Feather 30 All stalked 30 Frond indet. All upright, non-stalked Unknown 25 All 25 All stalked All stalked, All upright, non-stalked 20 Charniodiscus removed 20 All Height above substrate Height above substrate 15 15

10 10

5 5

0 0 0204060801001200 50 100 150 200 250 300 Number of individuals Number of individuals

C Mistaken Point D surface CD Mistaken Point E surface 50 50 Bradgatia Bradgatia 45 “Charnia” 45 “Charnia” Duster Charniodiscus 40 Fractofusus 40 Duster Frond indet. Fractofusus 35 Hapsidophyllas 35 Thectardis Pectinifrons All stalked 30 30 Unknown All upright, non-stalked All upright, non-stalked 25 25 All All flat-liers 20 All 20 Height above substrate 15

Height above15 substrate

10 10

5 5

0 0 0 200 400 600 800 1000 1200 1400 1600 0 500 1000 1500 2000 2500 Number of individuals Number of individuals E F Mistaken Point G surface Shingle Head surface 50 50 Bradgatia Bradgatia “Charnia” 45 45 “C harnia” Charniodiscus Charniodiscus 40 Duster 40 Frond indet. Pectinifrons Unknown 35 Unknown 35 All stalked All upright, non-stalked 30 All upright, non-stalked 30 All All 25 25

20 20 Height above substrate Height above substrate 15 15

10 10

5 5

0 0 0 20 40 60 80 100 120 0 50 100 150 200 250 300 350 400 Number of individuals Number of individuals

G Bristy Cove surface H Pigeon Cove surface 50 50 “Charnia” “Charnia” Thectardis 45 Fractofusus 45 All upright, non-stalked Unknown 40 All 40 All upright, non-stalked All 35 35

30 30

25 25

20 20 Height above substrate Height above substrate 15 15

10 10

5 5

0 0 0 102030405060708090 0 20 40 60 80 100 120 140 160 180 Number of individuals Number of individuals

Figure 7.2. Distribution of frondose parts in the water column. A) Bed B, Charnwood Forest; B—H) Newfoundland: B) Lower Mistaken Point; C) Mistaken Point D; D) Mistaken Point E; E) Mistaken Point G; F) Shingle Head; G) Bristy Cove; H) Pigeon Cove.

125 morphotype and community distributions show a number of small, discrete peaks at greater heights in the water column, but all occur within the distributions of single taxa. The medians of Bradgatia, Charnia, Orthiokaterna, Primocandelabrum and Undosyrus overlap with all other taxa on the surface (Fig. 7.4a). The medians of Charniodiscus and Undosyrus do not overlap. The medians of both morphotypes overlap within 95% confidence. The interquartile range (IQR) of Bradgatia and Charnia overlap with those of all other taxa on the surface. The IQR of Charniodiscus and Orthiokaterna overlap with each other and with Bradgatia and Charnia, while the IQR of Primocandelabrum and Unodysrus overlap with each other and with Bradgatia and Charnia. The stalked morphotype IQR overlaps with that of the upright, non-stalked morphotype and with all taxa except Orthiokaterna (which is higher) and “dusters” (which are lower), while the upright, non-stalked morphotype overlaps with all taxa except Orthiokaterna and Charniodiscus.

7.4.2 Lower Mistaken Point (LMP)

The distribution of the frondose parts of the community on Lower Mistaken Point (Fig. 7.2b) ranges from 0 to 33 cm, and peaks at 2—3 cm. The stalked morphotype distribution ranges from 0 to 22 cm, and peaks at 2—3 cm. The constituent taxa of this morphotype all peak at this height, but have different ranges: Charniodiscus ranges from 2 to 22 cm, “dusters” from 0 to 12 cm, “ostrich feathers” from 0 to16 cm, and “Charnia 2” from 0 to 20 cm. Thus, the ranges and peaks for all taxa in the stalked morphotype overlap. Fractofusus is the only flat-lying taxon on the surface; its distribution ranges from 0 to 9 cm and peaks between 0 and 1 cm. The distribution for the non-stalked, upright morphotype ranges from 0 to 33 cm and peaks at 1—2 cm. “Charnia” and the frond indet. are included in that category; their distributions range from 0 to 33 cm and from 0 to 6 cm, and peak at 0—2 cm and 1—3 cm, respectively. The peaks of all taxa overlap, except for Fractofusus which only overlaps with “Charnia” and the frond indet.. The peaks of the stalked and flat-lying morphotypes do not overlap with each other, but both overlap with the upright, non-stalked morphotype (Fig. 7.2b; Appendix A.2.10).

The whole community has a left-skewed peak in the density distribution of total heights at around 4–5 cm, with a few small bumps or sub-peaks at greater heights in the water column (Fig. 7.3b). The peak for the flat-lying taxa is highly left-skewed and peaks at around 3 cm, while the distributions for all non-stalked, upright taxa (together and individually) are symmetrical, and narrow about their peak (at 4 cm). The stalked morphotype also peaks at around 4 cm and is left-skewed; “Charnia 2” shares this distribution. Charniodiscus peaks at 6 cm but has a large spread, “dusters” have a main peak at 4—5 cm and a second peak at 12 cm, and “ostrich feathers” has a strong peak at 4 cm and several small (higher) peaks. The medians for all taxa overlap, except for “Charnia” which does not overlap with frond indet. and “ostrich feathers”. The IQR of all taxa overlap, except for Charniodiscus which does not overlap with the “unnamed” frond and Fractofusus. Medians and IQR for all morphotypes overlap (Fig. 7.4b).

126 7.4.3 Mistaken Point D surface (MPD).

The community distribution of frondose parts ranges from 0 to 27 cm and peaks at 0—2 cm, co-incident with the peak of the flat-lying morphotype that is numerically dominant on this surface (A.2.10), and ranges from 0 to 23 cm (Fig. 7.2c). There are two flat-lying taxa on this surface, Pectinifrons, whose distribution ranges from 0 to 23 cm and peaks between 0 and 4 cm, and Fractofusus, whose distribution ranges from 0 to 6 cm and peaks between 0 and 2 cm. The upright, non-stalked morphotype distribution ranges from 0 to 27 cm has a broad peak between 0 and 6 cm; the range and peak overlap with those of the flat-lying taxa. Within this morphotype, “Charnia” ranges from 0 to 27 cm, and has a broad peak from 0 to 7 cm; and Bradgatia ranges from 0 to 19 cm and peaks between 0 and 6 cm. “Dusters” are the only stalked form known from this surface, and their distribution ranges from 0 to 7 cm, and peaks at 1—2 cm. On this surface, the ranges and peaks of all morphotypes and all individuals (within and between morphotypes) overlap.

The density distribution of total heights in the whole community is left-skewed and peaks at 6 cm – the flat-lying morphotype and Fractofusus have the same distribution (Fig. 7.3c). Pectinifrons has a peak at around 14—18 cm, but its distribution has a lot of spread. The upright, non-stalked morphotype distribution is approximately symmetrical about its peak at 12 cm. The Bradgatia distribution is broadly symmetrical about its peak at 13 cm, while “Charnia” has a main peak at 5 cm and a second, small peak at 27 cm and the “Unknown” group has one peak at 6 cm and another at 11 cm. “Dusters” are the only taxon of the stalked morphotype on this surface, and have a tight, approximately symmetrical distribution about their two overlapping peaks at 3 and 5 cm. On this surface, there is a limited degree of overlap between taxa, and the medians of the morphotypes do not overlap (Fig. 7.4c). The median of Bradgatia overlaps with that of “Charnia” and of the upright, non-stalked morphotype. The median of the frond indet. overlaps with that of the “dusters”, while the median of Pectinifrons overlaps with no other taxa. The IQR of “Charnia”, Bradgatia and Fractofusus overlap, while the IQR of Pectinifrons overlaps with that of Bradgatia and Charnia. The IQR of the “dusters” overlaps only with that of the frond indet., the IQR of which also overlaps with that of Fractofusus. The IQR of the morphotypes overlap with each other and with everything except “dusters” (which are of smaller sizes), and additionally with Pectinifrons for the flat-lying morphotype and the frond indet. for the upright, non-stalked morphotype.

7.4.4 Mistaken Point E surface (MPE)

The community distribution of frondose parts ranges from 0 to 22 cm and peaks between 0 and 1 cm, co- incident with the peak of Fractofusus, the only taxon of the flat-lying morphotype, which is numerically abundant on this surface and ranges from 0 to 11 cm (Fig. 7.2d). The upright, non-stalked morphotype ranges from 0 to 22 cm, and peaks between 0 and 3 cm. Its two constituent taxa on this surface Bradgatia, and “Charnia”, have distributions ranging 0 to 19 cm and 0 to 22 cm, respectively, and both have a broad

127 A Charnwood Forest Bed B B Lower Mistaken Point surface

Bradgatia “C harnia” Charnia “C harnia 2” Charniodiscus Charniodiscus Duster 0.3 Duster Primocandelabrum Fractofusus Orthiokaterna Undosyrus Ostrich Feather All stalked Frond indet. All upright, non-stalked Unknown All All stalked All stalked, All upright, non-stalked 0.10 Charniodiscus removed All

0.2 density density

0.05 0.1

0.00 0.0 0 10203040 0102030 Total height (cm) Total height (cm)

C Mistaken Point D surface D Mistaken Point E surface 0.3 Bradgatia Bradgatia “C harnia” “C harnia” Duster Charniodiscus 0.20 Fractofusus Duster Frond indet. Fractofusus Hapsidophyllas Thectardis Pectinifrons All stalked Unknown All upright, non-stalked All upright, non-stalked All All flat-liers All 0.15 0.2 density

0.10 density

0.1

0.05

0.00 0.0 0 10203040 0 102030 Total height (cm) Total height (cm)

Continued overleaf.

128 A B Mistaken Point G surface Shingle Head surface

Bradgatia Bradgatia “C harnia” “Charnia” Charniodiscus 0.3 Charniodiscus Duster Pectinifrons Frond indet. Unknown Unknown All stalked All upright, non-stalked All upright, non-stalked All All

0.2

0.2 density density

0.1 0.1

0.0 0.0 0 102030 0 10203040 Total height (cm) Total height (cm)

C Bristy Cove surface D Pigeon Cove surface 0.12 “C harnia” “Charnia” Fractofusus Thectardis Unknown All upright, non-stalked All upright, non-stalked All All

0.75 0.09 density density 0.50 0.06

0.25 0.03

0.00 0.00 02468 5 101520 Total height (cm) Total height (cm)

Figure 7.3. Density distribution plots of total heights of taxa and morphotypes in the water column. A) Bed B, Charnwood Forest; B—H) Newfoundland: B) Lower Mistaken Point; C) Mistaken Point D; D) Mistaken Point E; E) Mistaken Point G; F) Shingle Head; G) Bristy Cove; H) Pigeon Cove.

129 peak between 0 and 3 cm; the non-rangeomorph Thectardis has a very similar distribution, ranging from 0 to 17 cm and peaking between 0 and 5 cm. The stalked morphotype ranges from 0 to 22 cm and peaks at 1—2 cm. Its two component taxa on this surface, Charniodiscus and “dusters”, range from 0 to 21 cm and from 0 to 22 cm, and both peak at 1—2 cm. The ranges and peaks of the upright, non-stalked morphotype and its constituent taxa overlap with those of the stalked morphotype and its constituent taxa. The peaks of all individuals and morphotypes overlap, except for the peaks of Fractofusus and the upright, non-stalked morphotype which do not overlap with Charniodiscus and “dusters”; the Fractofusus peak additionally does not overlap with that of the stalked morphotype.

The community distribution of total heights is left-skewed and peaks at 5 cm (Fig. 7.3d). The distributions of the upright, non-stalked and flat-lying morphotypes, as well as Fractofusus and Bradgatia, peak at 6 cm, while the distributions of the stalked morphotype and its constituent taxa, Charniodiscus and “dusters”, peak at 3 cm. The distribution of “Charnia” has a peak at 7.5 cm, and two small peaks at 12.5 and 18 cm. The distribution of Thectardis is approximately symmetrical about a peak at 11 cm. The medians of “Charnia” and Fractofusus overlap, as do the medians of Charniodiscus and “dusters”, and those of “Charnia” and the “spoon frond” (Fig. 7.4d). Thectardis overlaps with the “spoon frond”, the stalked morphotype only with Charniodiscus and “dusters”, and the upright, non-stalked morphotype with Bradgatia. The IQR of Charniodiscus, Fractofusus, “Charnia” and Bradgatia overlap, as do those of Charniodiscus and “dusters”. The IQR of Fractofusus overlaps with those of all taxa except “dusters”, which are of smaller sizes. The IQR of the three morphotypes overlap but do not overlap with the IQR of Thectardis and additionally the “dusters” for the flat-lying and upright, non-stalked morphotype, and with “Charnia” for the stalked morphotype.

7.4.5 Mistaken Point G surface (MPG)

The community distribution of frondose parts ranges from 0 to 32 cm, and peaks at 2–3 cm, con-incident with the range and peak of the stalked morphotype (Fig. 7.2e). Three taxa on this surface fall into the stalked morphotype: Charniodiscus, which ranges from 0 to 32 cm and peaks between 5 and 11 cm; “dusters”, which range from 112 cm and peak at 4—5 cm; and frond indet., which ranges from 0 to 14 cm and peaks at 2—3 cm. The upright, non-stalked morphotype ranges from 0 to 13 cm and peaks between 0 and 3 cm, while its constituent taxa, Bradgatia and “Charnia”, range from 0 to 13 cm and 0 to 7 cm respectively, and both have a broad peak between 0 and 4 cm. While the peaks for the morphotypes overlap with each other and with the upright, non-stalked taxa, the peaks of Charniodiscus and “dusters” overlap neither with each other, nor with any other taxa (or even with the other taxa on this surface).

The distribution of total heights on the community is left-skewed and peaks at 4 cm, the same as the frondose morphotype and the “unknown” fronds (Fig. 7.3e). “Charnia” has a left-skewed distribution that peaks at 3.5 cm, while the left-skewed distribution of Bradgatia has a main peak at 4.5 cm and a 130 A Charnwood Forest Bed B B Lower Mistaken Point surface

Bradgatia Charnia

Charnia Charnia II

Charniodiscus Charniodiscus Duster Duster Fractofusus Orthiokaterna Frond indet. Primocandelabrum Ostrich feather Undosyrus Unknown ST ST

FR FR

All All

0 10 20 30 40 0 10 20 30 Total height (cm) Total height (cm)

C Mistaken Point D surface CD Mistaken Point E surface Bradgatia Bradgatia Charnia Charnia Charniodiscus Duster Duster Frond indet. Fractofusus Fractofusus Hapsidophyllas Hapsidophyllas Spoon Frond Pectinifrons Thectardis

Unknown Unknown

FL ST

FR FR

All All

0 10 20 30 40 0 10 20 30 Total height (cm) Total height (cm)

E F Mistaken Point G surface Shingle Head surface Bradgatia Bradgatia Charnia Charnia Charniodiscus Charniodiscus Duster

Frond indet. Pectinifrons

Unknown Unknown ST FR FR

All All

0 10 20 30 0 10 20 30 40 Total height (cm) Total height (cm)

G Bristy Cove surface H Pigeon Cove surface I 95% confidence of median Charnia 15th median Charnia percentile Thectardis Fractofusus outlier Unknown Unknown 25th mean FR percentile 75th percentile All All Inter-quartile range 95 percentile 0 2 4 6 8 0 5 10 15 20 Total height (cm) Total height (cm) ± 1 standard deviation

Figure 7.4. Notched box plots of taxon and morphotype distributions of total heights in the water column. A) Bed B, Charnwood Forest; B—H) Newfoundland: B) Lower Mistaken Point; C) Mistaken Point D; D) Mistaken Point E; E) Mistaken Point G; F) Shingle Head; G) Bristy Cove; H) Pigeon Cove. I) Illustration showing the various parts of the notched box plots and their meanings. second, small peak at 12 cm. The unnamed “frond” similarly has a left-skewed distribution, a main peak at 3.5 cm and two small peaks at 12 and 14 cm. The stalked morphotype distribution peaks at 10 cm: the Charniodiscus distribution peaks at 12 cm, while the “dusters” peak between 7 and 10 cm – there is a greater spread to the distributions of both the individual stalked taxa and the stalked morphotype

131 than for the upright, non-stalked morphotype and its constituent taxa. No flat-lying taxa are recorded from this surface. The medians of Bradgatia and the frond indet. do not overlap with each other, but do overlap with that of “Charnia” (Fig. 7.4e). The median of “Charnia” overlaps with all other taxa except Charniodiscus, and the median of “dusters” overlaps with all taxa except Charniodiscus. The median of the stalked morphotype overlaps only with its two constituent taxa on the surface (Charniodiscus and “dusters”), while the median of the non-stalked morphotype overlaps with Bradgatia, “Charnia” and “dusters”. The IQR of Bradgatia, “Charnia” and the frond indet. overlap, while that of Charniodiscus and “dusters” overlap with each other but no other taxa. The IQR of the stalked morphotype overlaps with the upright, non-stalked morphotype, with “duster” and Charniodiscus, with the IQR of the upright, non-stalked morphotype overlaps with all taxa except Charniodiscus.

7.4.6 Shingle Head (SH)

The distribution of the frondose parts of this community ranges from 0 to 47 cm and peaks at 1–2 cm. The only flat-lying taxon on this surface, Pectinifrons, is also the most numerically abundant taxon on the surface; its distribution ranges from 0 to 12 cm, and peaks between 0 and 2 cm. There are small populations of two non-stalked, upright taxa on the surface, Bradgatia and “Charnia”, which range from 0 to 10 cm and 0 to 47 cm respectively. Bradgatia peaks between 0 and 3 cm, while there is only one specimen of “Charnia”. The “Unknown” grouping lack stems, range from 0 to 46 cm and peak between 0 and 2 cm. Charniodiscus is the only stalked taxon on this surface, and is also known from one individual – its frond ranges from 1 to 5 cm. While the distribution peaks of Bradgatia, Pectinifrons and the “unknown frond” overlap, the Charniodiscus distribution only overlaps with these taxa for its lower 2 cm.

The distribution of the total heights of the whole community is left-skewed, and peaks at 6 cm – the same as the distribution of Pectinifrons. The distribution of the upright, non-stalked morphotype, and the “unknown frond” that dominates this morphotype, are left-skewed and have a main peak at 3 cm, and three smaller peaks at greater heights in the water column. The medians and IQR of Bradgatia and Pectinifrons do not overlap, but the median of the “unknown frond” overlaps with Bradgatia and its IQR overlaps with those of both Bradgatia and Pectinifrons. There is only one individual each of Charnia and Charniodiscus and so their medians cannot be meaningfully compared. The median of the upright, non-stalked morphotype overlaps only with that of Bradgatia, but its IQR overlaps with all taxa on the surface.

7.4.7 Bristy Cove (BC).

The community distribution for frondose parts ranges from 0 to 8 cm and peaks between 0 and 1 cm, coincident with the peak of Fractofusus, the numerically dominant taxon on this surface. Fractofusus is

132 A Medians of each taxon and morphotype B Median of medians

Bradgatia 0.12 “Charnia” Charniodiscus Duster 0.15 Fractofusus All stalked All upright, non-stalked All 0.09

0.10 density

0.06 density

0.05 0.03

0.00 0.00 0 10 20 30 40 0 10 20 30 40 medians median Figure 7.5. Median distributions across taxa and surfaces. A) Distributions of medians of total heights calculated from each surface, for all taxa and morphotypes. B) The distribution of the median of the total heights for all individuals across all surfaces. the only flat-lying taxon, and ranges from 0 to 6 cm. “Charnia”, the only stalked taxon on this surface, ranges from 0 to 8 cm, and peaks at 2—3 cm. The “unknown” group ranges from 0 to 8 cm and peaks between 0 and 2 cm, overlapping with the peak of the flat-lying taxon (Fractofusus). The peak of “Charnia” does not overlap with that of Fractofusus.

The community distribution for total heights is strongly left-skewed and peaks at 1.5 cm, the same as the distribution of the only flat-lying taxon on the surface, Fractofusus. The community has a number of small additional peaks at greater heights in the water column, and Fractofusus has two small additional peaks (at 3.5 and at 5.5 cm). The upright, non-stalked morphotype has a broad peak at around 4 cm. “Charnia” has a peak at 3 cm and 4.5 cm, and a small peak at 8.5 cm. The medians and IQR of Fractofusus and “Charnia” do not overlap.

7.4.8 Pigeon Cove (PC).

The community distribution of frondose parts ranges from 0 to 23 cm and peaks between 0 and 2 cm. The upright, non-stalked morphotype is the only one found on this low-diversity surface, and ranges from 0 to 23 cm and peaks between 0 and 5 cm. It contains two taxa: “Charnia” ranges from 0 to 23 cm, and has a broad peak between 0 and 6 cm, while the “unknown” non-stalked fronds (for whom no stalk length was recorded) range from 0—10 cm and peak between 0 and 5 cm. Thectardis, the numerically dominant

133 taxon on this surface, ranges from 0 to 16 cm and peaks between 0 and 2 cm. The peaks and distributions for all taxa on this surface overlap with each other and with the upright, non-stalked morphotype.

The distribution of the total heights of the whole community and of Thectardis is slightly left-skewed and has two overlapping peaks, between 6 and 10 cm. “Charnia” and the frondose morphotype have left-skewed distributions with a broad peak at around 12 cm and 10 cm respectively. The medians of “Charnia” and Thectardis do not overlap, but the median of the upright, non-stalked morphotype into which they are placed overlaps with both of them. The IQR of “Charnia” and Thectardis overlap with each other and that of the upright, non-stalked morphotype.

7.4.9 Summary of the results

When the total heights of the organisms are taken into account, the medians and IQR of all upright, non-stalked forms overlap with most or all other forms of this morphotype on each surface. On MPE, the median of Fractofusus overlaps with “Charnia”, and on LMP it overlaps with all other taxa, but on MPD and BC it overlaps with no other discrete taxa. Taking the IQR into account increases the overlap of Fractofusus with other taxa: it overlaps with the lower part of all taxa except Charniodiscus on LMP and with Pectinifrons on MPD, which are both higher in the water column. Fractofusus is higher in the water column (both median and IQR) than the “dusters” on MPE and MPD, and than the frond indet. on MPD; similarly, the median (but not the IQR) of Fractofusus is higher than that of Charniodiscus on MPE. The median and IQR of stalked taxa overlap with those of non-stalked ones on the majority of bedding planes, with only a few exceptions: on MPG, the median and IQR of Charniodiscus and “dusters” are higher in the water column than those of upright, non-stalked taxa; on LMP, the IQR of Charniodiscus is higher in the water column than that of Fractofusus, and its median is higher than those of “dusters” and the frond indet.; and on Bed B, the median and IQR Charniodiscus and the median of Orthiokaterna (but not the IQR) are higher in the water column than the “dusters” and Undosyrus. There is no correlation between the value of the median and taxon or morphotype (R2=0.196, p=0.95) and, when the density distribution of the medians on all bedding surfaces are plotted, the peak distributions and medians of all taxa overlap (Fig. 7.5). There is no correlation between morphotype abundance and density, but there is a statistically significant correlation between abundance of Fractofusus and density (R2=0.53, p=0.04). There is no evidence of the <8 cm, 822 cm and >35 cm tier divisions proposed by Clapham (2002) from any of the analyses undertaken here – the mean and median of the medians of all taxa and morphotypes individually fall between 5 and 10 cm (Fig. 7.5). When the medians of all taxa and morphotype are combined, there is a main peak at 5 cm, a second peak at 8 cm which overlaps with the 5 cm peak, as well as two small peaks at 18 cm and 47 cm (Fig. 7.5).

When tiering habit is taken into account (by analysing only the density of the frondose parts in the water column), the distribution of upright, non-stalked morphotype consistently overlaps with those of both 134 the flat-lying and stalked morphotypes. The distribution of the stalked morphotype overlaps with that of the flat-lying morphotype on all surfaces except LMP. The distributions of individual taxa overlap with others of the same and of different morphotypes, with only a few exceptions: on Bed B, the peak of Orthiokaterna is above those of all other taxa; on MPE, the peaks of Charniodiscus and “dusters” are above those of Fractofusus; and on MPG the peaks of Charniodiscus and “dusters” are above those of all non-stalked taxa on the surface.

7.5 Discussion

Tiering is a strategy for the vertical partition of the ecospace. In epifaunal benthos, vertical tiering reduces inter-specific competition for water-borne resources. These organisms reside in the benthic boundary layer (BBL), that part of the water column which is directly affected by the interface between sediment surface and the overlying water (Dade et al. 2001) – it is a site of both vertical and lateral transport of solutes (Boudreau 2001) and suspended particles (Hill & McCave 2001), and contains strong chemical and physical gradients (Jorgensen & Boudreau 2001). Rates of flow and eddy diffusivity increase with increasing height in the water column (Dade et al. 2001). Concentrations of solutes within the water (i.e. those not produced at the sediment surface) decrease towards the sediment surface (Boudreau 2001), while concentration of suspended particles decreases away from the sediment surface (Adams & Weatherly 1981). Growing higher into the water column exposes part or all of the feeding structures (for organisms with a colonial or solitary tiering habit, respectively) to parts of the flow with a higher overall concentration of solutes derived from the water column (Boudreau 2001), and a lower concentrations of suspended sediment (Hill & McCave 2001). The extent of ecological tiering in a community gives important information on the structure of the community, such as the likely severity of competition, and are important in analysing flow dynamics in the community – in sub-aqueous environments, hydrodynamic flow within the community and the incidence of flow on a particular individual is strongly affected by the location and size of individuals in the community with respect to one another (Carey 1983, Eckman et al. 1981, Ghisalberti & Nepf 2002, Johnson 1990).

The left-skewed shapes of the density distributions of total heights on each of the fossiliferous surfaces studied here indicate that small-sized individuals are numerically dominant on the surfaces. For all the communities studied here, the highest density of frondose mass is in the lower most few centimetres of the water column – it is this part of the water column where there is the greatest overlap between individuals (Fig. 7.2). The simplest interpretation of this trend is, similarly, that it reflects the greater numeric abundance of small individuals on the surface. However, the exact height of the peak distribution of both total height and frondose mass varies between communities, reflecting the peak of the most numerically abundant taxon/morphotype on the surface (Figs. 7.1, 7.2). For instance, the communities on Bed B, Lower Mistaken Point, Mistaken Point G, Shingle Head and Pigeon Cove have rare or no

135 flat-lying taxa, and so the highest density of frondose mass in the communities on these surfaces is situated a few centimetres above the substrate surface (Fig. 7.2). In contrast, Mistaken Point E, H14 and Mistaken Point D are dominated by flat-lying taxa, and so the peak of all taxa combined sits within the lowermost few centimetres of the water column (Fig. 7.2). The presence of small secondary peaks in the community distributions of total heights and frondose mass is interesting. However, as these secondary peaks are present within single taxa, they do not represent discrete tiers, and instead reflect other ecological processes – these are discussed in Chapter 9.

The evolution of discrete, naked stems has been argued to have been a direct response of, and subsequently a driver to, vertical tiering in rangeomorph-dominated communities (Laflamme et al. 2012b). When tiering habit is ignored, vertical partitioning is observed for a small number of taxa on three surfaces (MPD, MPE, MPG), but there is no consistency between the taxonomic composition of the tiers with, for example, “dusters” and Charniodiscus occupying low positions in the water column on MPE and MPD, but high positions in the water column on MPG. In contrast, in modern and Phanerozoic communities, taxa are adapted to specific vertical niches (e.g. Bottjer & Ausich 1986, Watkins 1991, Cannell & Grace 1993). When tiering habit is taken into account, giving a more accurate picture of vertical stratification in the water column (Bottjer & Ausich 1986), only taxa with proportionally long, discrete, naked stalks where the organisms attain relatively large heights (e.g. Orthiokaterna on Bed B, Charniodiscus and “dusters” on MPG and LMP), and flat-lying forms where other morphotypes gain large heights (e.g. BC, LMP compared to MPE, MPD) show any evidence of vertical partitioning. On most of the studied surfaces, forms with short stalks (e.g. Beothukis plumosa), forms that lacked stalks (e.g. Charnia), and especially forms that were flat-lying (e.g. Fractofusus), all have their frondose parts in the same densely- occupied part of the water column – there is little to no discernible vertical stratification for the majority of taxa and morphotypes. This lack of vertical tiering suggests that competition for water-borne resources was negligible in the majority of the studied communities.

These conclusions have direct implications for our understanding of the communities, and the ecological processes that operated on them. That the total heights of “dusters” and Charniodiscus are smaller than those reached by the flat-lying Fractofusus is interesting. While this has been interpreted as these taxa occupying the lowest tier on the community (Clapham & Narbonne 2002), this is demonstrably not the case (at least, when tiering habit is taken into account, see section 7.3). Of particular note, the small sizes of the majority of Charnia masoni individuals on Bed B has been argued to reflect intense competition preventing growth to optimal/maximum sizes (Wilby et al. 2015), based on comparison with modern benthic communities (Johnson 1990, Sebens 1983) and the neighbour –shading and down-stream effects that are common in dense benthic communities (Glynn 1973, Merz 1984, Okamura 1984). A similar argument could be made for the small sizes of stalked fronds on MPD and MPE, given the density of

136 individuals on MPE and MPD (39.7 and 23.5 individuals per m2, Clapham et al. 2003). As the stalked fronds occur in lower abundances on MPD and MPE compared to those of the non-stalked taxa (1187 vs. 1876 individuals), it is also possible that these taxa arrived in the community at comparatively late stages of the ecological succession (rather than the mid-stages suggested by (Clapham et al. 2003), and therefore that their apparent occupancy of a lower tier simply records the comparatively young ages of these individuals. Modern species characteristic of later stages in ecological succession typically have considerably slower rates of growth than those of earlier stages, potentially exaggerating the size signal.

At least in modern photoautotrophic communities, species occupying low tiers are adapted to maximise resource capture, or to avoid shading (e.g. shade-tolerant/avoidance species in the understory of rainforests, (Henry & Aarssen 1997, Valladares & Pearcy 2000). In these communities, Fractofusus occupies the lowest levels of the water column, at least in the greatest densities (Fig. 7.2). It has unfurled branching at all resolvable scales (Brasier et al. 2012; personal observations). Contrastingly, forms that grow taller in the water column and therefore enter regions of higher-velocity flow (e.g. Orthiokaterna and Charnia, see eection 7.2), have furled branches, particularly at higher orders. Surfaces of sediment and organisms alike are mantled by a diffusive boundary layer (DBL; Jorgensen 2001, Jumars et al. 2001), in which the rate of transport by molecular diffusion exceeds that by eddy diffusion. The thickness of this layer is affected by the rate of flow (Jorgensen & Des Marais 1990) and surface roughness (Jorgensen 2001) – placing the feeding structures into a region of faster flow speeds results in a thinner diffusive boundary layer over the surface of the feeding structures, and consequently a higher rate of flux of particles and solutes (Dade et al. 2001, Jorgensen 2001). Passive filter feeders (such as stalked crinoids) typically occupy high tiers, whereas active filter feeders (who are able to overcome lower nutrient concentrations by creating their own flow environment) occupy low tiers (Bottjer & Ausich 1986). However, given the lack of preserved cilia or other structures which could modify or create a current, it is assumed that all rangeomorphs were passive feeders (Clapham & Narbonne 2002). While faster flow speeds would have reduced the DBL over the surfaces of tall organisms, Fractofusus may have achieved this by adopting an unfurled architecture and thereby increasing its surface roughness (cf. Jorgensen 2001).

7.5.1 Tiering and the rangeomorph mode of life

The rangeomorph mode of life has remained elusive, despite considerable debate on the topic. Early descriptions of rangeomorphs as possible sea pens (Glaessner 1985) saw them widely interpreted as filter feeders, but chemoautotrophy, symbiosis and even photoautotrophy have also been proposed (Jenkins & Gehling 1978, Glaessner 1985, McMenamin 1986, McMenamin 1998). While photoautotrophy may be a viable option in shallow-water successions, the deep-water Avalonian communities would have been well below the photic zone (Chapter 2), particularly given the greater attenuation of light through the

137 turbid waters thought to have persisted at that time (Butterfield 2009a). Recently, however, it has become the trend to infer an osmotrophic mode of life for rangeomorphs – the absorption of dissolved organic carbon (DOC) through an outer membrane (Laflamme et al. 2009). The main arguments supporting an osmotrophic feeding strategy in rangeomorphs are their lack of visible, preserved feeding structures, and the high surface area to volume (SA/V) ratios of many rangeomorph taxa (Laflamme et al. 2009). However, there are numerous problems with this interpretation, based on functional morphological, physiological and geochemical grounds (discussed in Liu et al. 2015a). While the supposedly tiered structure of rangeomorph communities was initially taken to support a filter-feeding mode of life, this structure and its consequences for flow dynamics in the community has more recently been invoked to support an osmotrophic one (Ghisalberti et al. 2014).

The presence of epifauna on a sediment surface affects flow processes in the benthic boundary layer, principally by reducing flow velocity (Madsen et al. 2001), and increasing and rescaling flow turbulence (Jumars et al. 2001). This is particularly the case for dense communities, in which canopy flows develop (Ghisalberti & Nepf 2002, Raupach et al. 1996), increasing the velocity gradients and rates of vertical mixing normally observed in the boundary layer (Ghisalberti & Nepf 2009). Ghisalberti and colleagues used fluid dynamic modelling of rangeomorph communities to demonstrate that the interaction between bottom-currents and the presumed tiered frondose communities created a velocity profile in the water column, with taller organisms perceived to have access to flows of higher velocity, permitting faster rates of nutrient uptake (Ghisalberti et al. 2014). While such interaction between organism and flow would be expected to result in down-flow scours and up-flow sediment pile-up, the apparent absence of these structures on all beds (apart from Spaniard’s Bay, Brasier et al. 2013) may be a result of the background irregular surface topography or incompletely eroded overlying sediment obscuring such structures, or because sediment was microbially-bound and so resistant to such modifications. The flow profile is suggested to have been a major driver for the development of ecological tiering in rangeomorph communities (Ghisalberti et al. 2014, their fig. S3). However, in modern communities, the inclusion of abundant or dominant stalked taxa in the community have a significant influence on the structure of the canopy flow (Madsen et al. 2001), but this has not yet been applied to rangeomorph communities – the modelled communities have only small and/or rare stalked taxa (Ghisalberti et al. 2014). The exact density of the communities also impacts the structure of the flow, which has to date only been modelled for a fraction of the available bedding surfaces (see Ghisalberti et al, 2014 for a discussion). Crucially for interpretation of their mode of life, however, there is no requirement in the model for the substance taken up from the water column to be DOC. Tiered structures develop in response to light (Foster 1975, Cannell & Grace 1993), as well as to particulate matter in the water column (Bottjer & Ausich 1986, Watkins 1991) and in the sediment (Droser et al. 1994, McIlroy & Logan 1999); it reflects vertical partitioning of resource acquisition, but gives no indication whatever as to the composition of 138 those resources. Indeed, given the effect of currents on enzymatic DOC break-down, higher velocities may have been counter-productive for osmotrophs (discussed in Liu et al. 2015a). Therefore, even if the modelled canopy flows played a contributory role in structuring frondose communities and their interactions, they do not provide conclusive evidence for osmotrophy. Similarly, while it could be argued that the unfurled architecture of Fractofusus would have enhanced absorption of DOC, supporting an osmotrophic mode of life, it may simply have served to increase the efficiency of gas exchange between the organism and the water column (cf. Jorgensen & Des Marais 1990).

There is no reason why all Ediacaran organisms should have adopted the same mode of life – indeed, their likely disparate phylogenetic relationships (Grazhdankin 2014, Laflamme et al. 2013) makes this unlikely. Even within rangeomorphs, there may have been multiple modes of gaining their nutrition – unless nutrient sources were abundant, this would have ameliorated any inter-specific competition present. In modern systems, concentrations of dissolved and particulate organic matter are considerably higher in the sediment than in the water column (e.g. Hulth et al. 1997, Papadimitriou et al. 2002, Sierra et al. 2001). Extant organisms as disparate as sea urchin larvae (Shilling & Manahan 1990), dinoflagellates (Glibert & Legrand 2006) and algae (Gervais 1997, Vincent & Goldman 1980) have been shown to be capable of osmotrophy, but most of these organisms use it facultatively, to supplement photosynthesis (Glibert & Legrand 2006) or endocytosis (Stephens 1988; often to provide a source of nitrogen), or as a means of subsisting through unfavourable conditions (Vincent & Goldman 1980). A variety of options exist for the use of osmotrophy in combination with other feeding strategies. Low-lying forms such as Fractofusus may have exploited substances released from the microbial mat, or putatively from the clots of sediment and organic matter which are more abundant on surfaces where they occur than where they are absent (Chapter 8). In contrast, stalked organisms placed higher in the flow may have gained more of their nutrition through (particles) delivered by the lateral flow – for whom placing their fronds in a higher part of the water column would have been beneficial, both by reducing competition and increasing delivery of nutrients.

7.6 Conclusions

Weak tiering is identified on four communities when total heights are examined, but there is no consistency between the taxa occupying various levels in the water column, and no correlation between median total height and taxonomic or morphotype composition. When tiering habit is taken into account, stalked taxa with long stems and large total heights occupy slightly higher levels of the water column than non- stalked taxa and, where upright, non-stalked and stalked taxa attain large heights, Fractofusus occupies a lower level of the water column than the stalked taxa. However, the distributions of upright, non-stalked taxa overlap with those of stalked and flat-lying taxa – there is no vertical stratification of these forms, nor of those stalked forms with proportionately short stems (such as Beothukis). Accordingly, tiering is

139 interpreted to have been weak or absent on these communities, and cannot by itself explain the evolution of discrete, naked stems. If possession of a stalk provided any competitive advantage for nutrient acquisition, then these forms would be expected to dominate the most densely-occupied communities. However, there is no correlation between the proportion of stalked individuals and the organism density in a community, and only a few stalked individuals occupy a higher level of the water column than their non-stalked peers. Similarly, if the low-lying habit and unfurled branching of Fractofusus enabled it to exploit the densely-occupied lower levels of the water column, then why is it only present on four out of eight of the studied surfaces, and only abundant on three? Environmental processes that influenced the taxonomic composition of a community are explored in Chapter 8. The lack of prominent tiering in these rangeomorph-dominant communities has numerous implications for the ecology of rangeomorphs, most notably that competition for nutrients in the water column was far less severe than has been previously suggested.

140 8. THE ROLE OF AMBIENT DISTURBANCE ON COMMUNITY COMPOSITION

8.1 Introduction

The ecology of Ediacaran organisms has been the subject of much recent work. In the comparatively shallow-water Ediacaran localities of the White Sea, different fossil assemblages have been linked to broad-scale sedimentary facies. Among the first of these studies was undertaken on White Sea assemblages (Grazhdankin 2004), and was followed by work for the classic sites of Australia (Gehling & Droser 2013).

Within Avalonia (Chapter 1, 2), there is considerable variation in the taxonomic composition, diversity and size structure within communities preserved on different bedding surfaces. Previous studies of Avalonian sedimentology have broadly distinguished several facies, which have proved powerful in understanding basin evolution and the broader palaeoenvironment analysis (Wood et al. 2003, Ichaso et al. 2007). Despite this work, fossil assemblages have only been correlated to two broad depositional settings: turbidite-dominated, deep-sea slope where the classic Avalon macrofossil assemblages occur (Drook to Trepassey Formations, Chapter 2), and shelf settings, in which the Aspidella-dominated assemblages occur (upper Fermeuse Formation, Chapter 2). This broad-brush approach has led to the assumption that the deep-water Avalonian environment was relatively stable and quiescent, at least between depositional event beds (Chapter 2; Clapham et al. 2003, Narbonne 2005, Mason et al. 2013), and so the observed variation in community composition was attributed solely to ecological succession (Clapham et al. 2003). While the Clapham study represents a significant leap forward in our understanding of Avalonian ecology, it does not take into account the local, transient and spatially heterogenous conditions with which the organisms contended on a day-to-day basis, the importance of which was highlighted by recent work in the Mackenzie and Wernecke Mountain successions (Johnston et al. 2013). While we may not be able to determine factors such as the spatial variability in flow speeds and substrate composition in Ediacaran communities, it is possible to investigate the influence of physical disturbance (such as influx of sediment) that are captured in the rock record.

8.1.1 Disturbance and community structure

In modern systems, disturbance plays an important role in shaping the structure and composition of both marine and terrestrial communities (e.g. Connell, 1978; Huston, 1979). For example, marine communities may take decades to recover from dredging and storm events, if they do so at all, and frequently their compositions are altered towards those with life history traits that endow them with some resilience towards disturbance (Thrush & Dayton 2002, Browne et al. 2014). Similarly, plant communities recovering from volcanic eruptions and the lahars, fires and ash-fall that accompany them,

141 have drastically different compositions to those before the eruptions (Dale et al. 2005). While modern deepwater benthic communities are shaped by a variety of physical, temporal and biological parameters (Adams & Weatherly 1981, Etter & Grassle 1992, Gage & Tyler 1992, Gooday 2003, Carney 2005 , Jennings et al. 2013), some of these, such as predation, appear absent until the latest Ediacaran (Hua et al. 2003). In contrast, physical parameters, such as sedimentation rate and variation in current srength, are likely to have been an enduring factor in structuring deepwater communities.

Disturbance events fall on a continuum in terms of their severity, as characterised by their intensity, frequency, predictability, size, spatial extent and distribution (White & Pickett 1985, Turner et al. 1997, Dale et al. 2005). For instance, the Antarctic shelf is subjected to total denudation by ice scouring, but the intensity of this scouring depends on the depth to which the icebergs reach, and the path that they take across the shelf (Barnes & Conlan 2007). Similarly, the intensity and frequency of trawling control the degree to which the disturbed communities are affected (Thrush & Dayton 2002). Communities in intensely and frequently disturbed areas are very different to those in comparatively undisturbed areas.

There is a trade-off between adaptation for resource competition (e.g. for space, nutrients), and tolerance to disturbance. At low levels of disturbance (both in terms of intensity and time since last disturbance event), species that are competitively dominant for resources persist, while at high levels of disturbance, those better adapted to tolerate disturbance thrive – at intermediate levels of disturbance, there is a balance between species adopting both strategies (Connell 1978). Accordingly, communities experiencing intermediate levels of disturbance are often characterised by the highest taxonomic diversity – this relationship is termed the “intermediate-disturbance hypothesis” (Grime 1973, Huston 1979) and is seen in terrestrial and marine settings (Blackwood et al. 2010, Paterson et al. 2011, Portilla-Alonso & Martorell 2011). It most strongly affects temporal rather than spatial patterns of diversity, as communities take varying times to recover from disturbance events, depending on the magnitude of the event, and the mechanisms of temporal recovery are different from those of spatial recovery (Collins & Glenn 1997, Mackey & Currie 2001). The distal location of the fossil communities to the source of the disturbance, and their small areal extent, render spatial heterogeneity in the disturbance unimportant within a single community. The disturbance–diversity dynamic is particularly important for sessile communities, as their ability to colonise new areas is restricted to the motility of their propagules – they cannot themselves move to avoid disturbance, or to take advantage of recently denuded areas (Wootton 1998, Mackey & Currie 2001).

When determining the effect of such disturbance events on a community, they must be taken in the context of the ambient disturbance which is experienced. Communities in which the constituent organisms are more tolerant to ecological disturbance are typically less affected by comparatively major events, or better

142 able to recover from them, than ones in areas with relatively low levels of ambient disturbance (Dale et al. 2005). The ambient physical disturbance regime of an area incorporates factors such as the background sedimentation rates and grain size; high-frequency but comparatively low intensity disturbance events such as sedimentation events resulting from small storms or distal gravity-driven events. This chapter explores the extent and type of ambient physical disturbance that affected the fossiliferous sites (as expressed by sedimentary parameters), and how it affected the composition of the fossil communities. Chapter 9 discusses the effect of discrete, rare disturbance events on the communities.

8.2 Material and Methodology

In order to characterise the degree of physical disturbance, thin sections of sedimentological samples collected from 24 fossiliferous sites were examined, of which 23 were from Newfoundland (Appendix 2.2) and one from Charnwood Forest. Sedimentological samples from Newfoundland were collected during the 2012 field season under permit from the Parks and Natural Area Division, Government of Newfoundland and Labrador, supplemented by material collected by A.G. Liu during 2010-2012. Sedimentological samples from Bed B are housed at BGS. Detailed fossil occurrence data, providing information on the taxonomic composition of the community, was provided by A.G. Liu.

8.2.1 Methods

There are two main parts to this study. The first part involves characterisation of the level of physical disturbance, and documenting various aspects of the community composition for a given fossiliferous site. The second part involves analysis of the presence and types of correlation between the disturbance and composition parameters. These analyses determine the relationship between two sets of parameters: the explanatory or predictor variables, and the dependent or response variables. The dependent variables are the observation of interest – the effects or results as influenced by the explanatory variables. The analyses then determine if changes in the dependent variables correspond in a predictable way to changes in the explanatory variables. In this study, the explanatory variables are the sedimentological parameters that are taken to characterise the levels of physical disturbance. The dependent variables are the biological parameters, that is, aspects of the community composition.

8.2.2 Defining the explanatory variables: characterisation of the level of physical disturbance

Detailed petrographic analysis of thin sections was undertaken using conventional petrographic techniques combined with SEM-BSE equipped with EDX capabilities (for elemental mapping). The latter technique has the advantage that relict grain textures can be identified based on the concentration of different elements, providing a way of determining the grain size of the original sedimentary particles (including the matrix grain size) in these metamorphosed rocks. Using the light microscope, grain size was estimated using the built-in millimetre scale bar, and from images taken using the SEM-BSE.

143 Variable Abbreviation Description Range or levels Average matrix grain AMG This bed is defi ned as the portion of sediment above the size last event bed (Chapter 2) and capped by the fossiliferous horizon. Silt lenses in bed SL Comprised of a coarser grain size and typically much higher 0 (absent), 0.5 (present), 1 below proportion of quartz and opaque minerals than the matrix, (abundant) interpreted as the products of lateral fl ow, for example small current-induced ripples. Glebules G These are similar in grain size and composition to silt lenses, 0 (absent), 0.5 (present), 1 but are a much rounder shape, and are interpreted as sunken (abundant) grain rafts or clots Percentage of PFG The modal percentage of outsize or “fl oating” grains in the 0-5% fl oating grains bed below the fossiliferous surface Average grain size AFG The average grain size of outsize or “fl oating” grains in the of fl oating grains bed below the fossiliferous surface Maximum size of MFG The average grain size of outsize or “fl oating” grains in the bed fl oating grains below the fossiliferous surface; maximum size corresponds to the maximum dimension observed in thin section Thickness of the bed TBB The thickness of sediment between the fossiliferous surface below and the last event bed Table 8.1. Descriptions, abbreviations and ranges of the explanatory (sedimentary) variables. All variables relate to properties of the bed below the fossiliferous surface, defined as the portion of sediment above the last event bed and capped by the fossiliferous horizon. The various sedimentary fabrics observed are then placed into an interpreted depositional process/ event context. Seven sedimentological parameters were defined for the bed immediately beneath the fossiliferous surface (Table 8.1), quantifying the extent and type of sediment influx. The underlying (hemipelagite) bed is used as as it represents the background sedimentation between event beds, and thus records the detailed environmental conditions under which the community developed. These are the explanatory variables. The assumptions are that comparatively coarse matrix grain sizes, a high proportion of coarse grain components, and thin beds reflect more dynamic environments, with faster flows and a higher degree of disturbance through sediment small-scale influx. Contrastingly, sections with finer matrix grain sizes, thicker beds and a low proportion of coarse grain component were subjected to slower flows and lower levels of ambient disturbance through small-scale sediment influx.

8.2.3 Defining the dependent variables: aspects of community composition

Both rangeomorphs and arboreomorphs (Chapter 1) are divided into taxonomic (generic and species) groups, allowing for inclusion of their branching architecture and uni-/multifoliate habit to be incorporated (though see Chapter 4). The rangeomorph and arboreomorph taxa were characterised into three morphotypes, corresponding to the overall morphology and life habit of the organism: stalked (Fig. 7.1a—c); upright, non-stalked (Fig. 7.1d) and flat-lying (Fig. 7.1e, f). Definition of these morphotypes was discussed in Chapter 7. Form is related to function, and it might be expected that, if disturbance has an effect in shaping the composition of these ancient communities, then different morphotypes may correlate with the adaptation to competition for resources vs. tolerance to disturbance, and that is tested here. Eight dependent variables, corresponding to biological parameters (Table 8.2), were defined.

144 Variable Abbreviation Description Range or levels Total taxa present TT The total number of taxa identifi ed on the bedding surface: a crude measure of taxonomic diversity on the bedding plane Proportion of stalked PST The proportion of stalked taxa present on the bedding individuals surface: (number of stalked taxa/TT) Proportion of fl at-lying PFL The proportion of fl at-lying taxa present on the taxa bedding surface: (number of fl at-lying taxa/TT) Proportion of PFR The proportion of upright, non-stalked taxa present frondose taxa on the bedding surface: (number of upright, non- stalked taxa/TT) Abundance of FF The relative abundance of the fl at-lying taxon 0 (absent), 0.5 (present), 1 Fractofusus Fractofusus on the bedding surface (abundant), 1.5 (both species abundant) Multifoliate or U/M The relative dominance of taxa with multifoliate 0 (unifoliate dominant), 0.5 unifoliate construction versus unifoliate construction on the bedding surface (no dominance), 1 (multifoliate (rangeomorphs only) dominant) Displayed or rotated D/R The relative dominance of taxa with displayed versus 0 (rotated dominant), 0.5 (no branching rotated branching construction on the bedding dominance), 1 (displayed surface (rangeomorphs only) dominant) Furled or unfurled U/F The relative dominance of taxa with unfurled versus 0 (furled dominant), 0.5 (no branching furled branching construction on the bedding surface dominance), 1 (unfurled (rangeomorphs only) dominant) Table 8.2. Descriptions, abbreviations and ranges of the descriptor (biological) variables. 8.2.4 Analyses of the correlation(s) between explanatory and dependent variables.

Two methods were used to investigate the correlation: 1) linear regression analysis, and 2) principal components analysis. Both were undertaken in R (2014).

8.2.4.1. Linear regression analysis. Linear regression analysis examines the correlation between a dependent variable and explanatory variable(s). The first step in this method was to examine the correlation between single pairs of variables. Given the large number of possible permutations of variables, the most highly correlated variables were determined in R using the function made available by Coghlan (2014), which generates a list of all variables most highly correlated to each other. The p-values of the correlation were subsequently determined for all pairings between explanatory and dependent variables thus identified.

The second step in this method used stepwise backward regression analysis to evaluate the correlation between dependent variables and multiple explanatory variables together. The combination of explanatory variables for the modelled regression which best fit the data is interpreted as the best predictors for the dependent variable. The linear regression between a dependent variable (e.g. morphotype or taxon presence/abundance) and all possible explanatory variables (sedimentological parameters) is plotted. The steps are as follows:

1) The two-sided p-value calculated for the t-statistic, Pr(>|t|), is examined in order to identify the explanatory variable which has the lowest correlation to the dependent variable. 2) The linear regression is run again, but with the variable identified in (1) removed from the analysis.

145 3) Steps (1) and (2) are repeated until only one explanatory variable remains. For categorical variables, correlation statistics are calculated for each category, and frequently one category will be much more highly correlated with the descriptor variable than the other. In this case, two runs are made: one removes the categorical variable in step 2 when either category is the least highly correlated with the descriptor variable, and the second removes the categorical variable only when all the categories are least highly correlated variables. This has a small effect on the models which best describe the correlation. 4) The AIC (Akaike Information Criterion; Bozdogan 1987) is calculated for each regression model, whereby the model with the lowest AIC value represents that which best fits the data. Where two models have very similar AIC values, the R2 and p-value of each is examined, and the regression with the highest R2 and lowest p-value selected as the best-fit model.

8.2.4.2 Principal components analysis and hierarchical clustering analysis This method is described in detail in Chapter 4, and used the FactoMineR package as described there. It allows multiple explanatory and dependent variables to be analysed in tandem. It allows samples to be grouped into clusters that are more similar to each other, and describes the parameters that correspond to cluster definition. The principal components analysis (PCA) was used in preference to the factoral analysis of mixed data (FAMD), as the categorical variables are ordered, that is, they progress in a meaningful manner (e.g. absent = 0, present = 0.5, abundant = 1) and so do not impart the biases that result from arbitrary scoring of non-ordered categorical characters. The minimal amount of missing information in this dataset (Appendix 2.2) means that the analyses can be meaningfully run without first imputing missing values (Appendix 1.1.3). Consequently, it is possible to set variables as supplementary, that is, they do not contribute to construction of the principal components space or clustering analysis, but their correlation to the principal components axes can still be determined. This is extremely useful in determining the correlation between multiple descriptor and explanatory variables without biasing construction of the components space (Husson et al. 2010b, Husson et al. 2010a). All variables were scaled prior to analysis to remove potential bias induced from different measurement units (see discussion in Appendix 1.1.2).

Four iterations were run: 1) setting explanatory variables as supplementary ones, constructing the principal components space and clustering fossil sites on just the sedimentological variables; 2) setting descriptor variables as supplementary ones, constructing the principal components space and clustering fossil sites on just the biological variables; 3) setting all variables except morphotype as supplementary; and 4) using all descriptor and explanatory variables in constructing the principal components space and clustering the fossil sites. After PCA analysis, the bedding surfaces were grouped according to the variables used to construct the PCA space using the hierarchical clustering algorithm in FactoMineR (HCPC; Chapter 4; Appendix 1.3). 146 8.3 Results

The fossiliferous surfaces all lie above a hemipelagite layer of varying thickness (Appendix 2.2), which overlies alternatively epiclastic or volcaniclastic beds (Fig. 2.4; defined in Chapter 2). Within the hemipelagite facies, however, there is a great deal of lithological variability (Figs. 8.1; Appendix 2.2). The matrix is dominantly clay to very fine silt grade. Some beds have an entirely homogenous or mottled appearance over the few centimetres encompassed by the thin section, whereas others include variable proportions of coarser-grained material. The fine to coarse silt component is dominated by rounded to sub-angular quartz grains with rare eu-/subhedral quartz and plagioclase crystals, and frequently has a higher proportion of opaque crystals than the matrix (Fig. 8.1b). Three main occurrences of coarser-grained material are identified, in varying abundances in the studied sections (Table 8.1). The first comprises lenticular to discontinuous laminae which pinch out to either end (silt ripples), and are dispersed throughout the hemipelagite (Fig. 8.1). These are termed “silt lenses”. The second comprises regions of silt that are rounded to ellipsoidal in outline (both perpendicular and parallel to bedding), and that do not appear to be connected to each other. These often occur in bands, and are termed “silt aggregates”. The “silt aggregates” resemble the “intraclasts” described by (Brasier et al. 2013) from Spaniard’s Bay, although not all silt aggregates have the framboidal pyrite cement described by those authors (Fig. 8.1d—f), and their grain compositions vary in even adjacent instances. The “intraclast” term is further rejected as this term implies that the aggregates were remobilised (partially) lithified sediment, but there is no evidence for this. The final occurrence of silt is as isolated, “floating grains”. The average and maximum grain size of the coarse grain component in each of its three fabrics varies across the studied thin sections (Appendix 2.2).

8.3.1 Regression analyses

The results of the single variable regression analyses are presented in Table 8.3. The only two single variable correlations with p-values below 0.05 are the abundance of Fractofusus with the thickness of the bed below the fossiliferous surface (cor = 0.57, p=0.00297), and with the total number of taxa recorded from the bedding surface (cor = 0.49, p=0.01). When all variables were scaled, the only statistically significant correlation was between thickness of the bed below and the proportion of upright, non- stalked fronds (cor=-0.50, p=0.01).

The results of the stepwise backwards regression analysis are presented in Table 8.4. In these analyses, the total number of taxa present on a surface is higher with a lower percentage and smaller maximum grain size of floating grains, higher average size of floating grains and thicker beds (p=0.08364). There is a higher proportion of flat-lying fronds on the surface with smaller average grain size of the matrix (p=0.07661), Fractofusus is more abundant on a surface with a smaller percentage and smaller maximum grain size of floating grains, thicker beds and when silt aggregates are absent (p=0.00384). There is a

147 A fg B fg sl fg fg dl fg dl dfl sl sl

sl dfl

sl fg

C D e sl

e fg gb sl e gb fg fg gb

E F dfl gb fg dfl

gb e gb gb fg

G H

fg

fg

fg

148 higher proportion of upright, non-stalked taxa on a surface with larger average matrix grain size, smaller maximum grain size of floating grains, thinner beds, and when silt aggregates are present in the bed below (p=0.01152). There is a higher proportion of stalked fronds on a surface with a larger average but smaller maximum floating grain size, but this correlation is very weak (p=0.3344).

8.3.2 Principal components analysis and hierarchical clustering analysis.

The results are presented under their respective iteration.

8.3.2.1 Sedimentological variables only active. When bedding surfaces were grouped according to their sedimentary characteristics, three groups were supported by inertia gain (Appendix 1.3). Group 1 (Sword Point and Mistaken Point E), was characterised by thicker-than-average beds beneath the fossiliferous horizon, and a larger maximum but smaller-than-average grain size of floating grains. Their position on the PCA space (Fig. 8.2b) correlates with a higher proportion of flat-lying taxa, a higher abundance of Fractofusus, a dominance of displayed, multifoliate and unfurled taxa (Fig. 8.2a), and lower taxonomic diversity. Group 2 (including Shingle Head, H14 and Bristy Cove) was characterised by smaller-than-average maximum and average grain size of floating grains, and a lower-than-average percentage of floating grains. Their position in the PCA space (Fig. 8.2b) correlates with a higher proportion of flat-lying and upright fronds, present or abundant Fractofusus, and high taxonomic diversity (Fig. 8.2a). Group 3 (including Bed B, Pigeon Cove and Spaniard’s Bay) was characterised by larger-than-average grain sizes of matrix and floating grains, a higher-than-average percentage of floating grains, and a lower abundance of silt aggregates. Their position in the PCA space (Fig. 8.2b) correlates with a higher proportion of stalked and upright, non- stalked taxa, a lower abundance of Fractofusus, lower taxonomic diversity, and a dominance of rotated, unifoliate and furled taxa (Fig. 8.2a).

8.3.2.2 Biological variables only active. When the explanatory variables are set as supplementary, the fossiliferous horizons are clustered only by the dependent (biological) variables. When two groups were selected, Group 1 (including Sword Point, Bristy Cove and Spaniard’s Bay) was characterised by a higher-than-average proportion of upright, non-stalked taxa, a lower-than-average proportion of flat-lying taxa, a lower-than-average abundance of Fractofusus, and a dominance of unifoliate, rotated and furled taxa. Their position in the PCA space

Figure 8.1 (facing page). Sedimentological fabrics. A) Silt lenses (sl), discontinous laminae (dl) and floating grains (fg); PU17. B) Silt lenses (sl) and discontinuous laminae (dl) with a high proportion of opaque minerals, associated with diffuse lenses and laminae (dfl) and floating grains (fg); SP.1. C) Three tabular epicalstic event beds (e), with silt lenses (sl) and floating grains (fg) in the remainder; SPS1. D) “Silt aggregate” (gb) with typical rounded edges and variable amounts of quartz and opaque minerals, arranged in bands, and rust staining likely after pyrite; SB. E) “Silt aggregate” (gb) and floating grains (fg), MPE. F) Isolated “silt aggregate” (gb) with typical rounded edges, in a matrix with diffuse laminae (dfl) above an epiclastic event bed (e), HS. G, H) Homogenous mudstone with very few floating grains (fg) and diffuse mottling, HF14.1a and MC.1a (respectively). All scale bars 0.5 mm.

149 A B sedimentological characters sedimentological characters

cluster 1 cluster 2 MFG cluster 3 SP.1

SL TBB PFG cluster 1 AMG PC

G BS.1 NQ bed B MP D/R FF cluster 3 U/F TT UM AFG SB HS CA Dim 2 (20.94%) PST BC PU.V.3 PU.4a Dim 2 (20.94%) HF14.1a2 PFL PFR PU.V.4 MC.1a PU.2 BS.2 SPS.1 PU.1 SP.2 PU cluster 2 HF14.1a STB

GHSB -2 -1 0 1 2 3 4 SH -1.0 -0.5 0.0 0.5 1.0

-1.0 -0.5 0.0 0.5 1.0 -4 -2 0 2 Dim 1 (27.60%) Dim 1 (27.60%)

C D morphotype abundance characters morphotype abundance characters

cluster 1 cluster 2 SH cluster 3 PST PU.1 PFR

cluster 2 AFG PFL BS.2 SB BC SPS.1 U/M NQ bed B STB PU GHSB PC PFG CA SP.2 HS AMG D/R cluster 1 BS.1 U/F PU.V.3 PU.V.4 PU.4a HF14.1a TT MC.1a cluster 3 SL G PU.2 MP Dim 2 (24.99%) Dim 2 (24.99%) MFG FF TBB SP.1

HF14.1a2 -3 -2 -1 0 1 2 -1.0 -0.5 0.0 0.5 1.0

-1.0 -0.5 0.0 0.5 1.0 -2 0 2 4 Dim 1 (43.10%) Dim 1 (43.10%)

E F all characters all characters

cluster 1 cluster 2 SH cluster 3 NQ bed B PU.1 AFG HS U/M PU.4a PST STB cluster 2 PFR PFG U/F PU BS.1 HF14.1a AMG D/R CA PU.V.3 GHSB PFL MC.1aSB PC BC SPS.1 BS.2 cluster 1 PU.2 TT PU.V.4 SL SP.2 Dim 2 (15.38%)

Dim 2 (14.85%) MFG FF HF14.1a2 G SP.1 cluster 3 TBB MP -4 -2 0 2 -1.0 -0.5 0.0 0.5 1.0

-1.0 -0.5 0.0 0.5 1.0 -4 -2 0 2 4 Dim 1 (26.30%) Dim 1 (25.44%)

Figure 8.2. Results of the cluster analysis, HCPC on the data set including all biological and sedimentological characters. A) Cluster dendrogram and B) factor map for analysis with sedimentological characters only active; C) Cluster dendrogram and D) factor map for morphotype abundance only active; E) Cluster dendrogram and F) factor map for all characters active. Inertia gain plots support division into 3 clusters in each instance.

150 correlates with coarser grain size of matrix and floating grains, a higher percentage of floating grains, and a higher abundance of silt aggregates. Group 2, including Mistaken Point E surface, Bed B and Shingle Head, was characterised by a higher proportion of flat-lying taxa, a higher abundance of Fractofusus, a lower proportion of upright, non-stalked fronds, and a dominance of multifoliate, displayed and unfurled taxa. Their position in the PCA space correlates with finer average and maximum grain size of floating grains, finer matrix grain size, thicker beds and less abundance silt aggregates, but a higher abundance of silt lenses. When divided into five clusters, Group 1 (including Sword Point and Spaniard’s Bay) was characterised by a higher-than-average proportion of upright, non-stalked fronds, a lower-than- average proportion of stalked taxa, low taxonomic diversity, and a dominance of furled and rotated taxa. Their position correlates with coarser-than-average maximum and average floating grain size and average matrix grain size, thin beds, a higher percentage of floating grains and a higher abundance of silt aggregates. Group 2 (including the Port Union surfaces and Pigeon Cove) was characterised by a dominance of unifoliate and furled taxa, and their position correlated with intermediate grain size parameters, percentage of floating grains, bed thickness and abundance of silt aggregates. Group 3 (including Bed B and the “Wrinkly M” surface) was characterised a dominance of multifoliate taxa, and their position correlates with intermediate values for all sedimentary parameters. Group 4 (including Mistaken Point E and H14) was characterised by a high abundance of Fractofusus, and a dominance of displayed and unfurled taxa. Their position correlates with thick beds, finer-than-average maximum and average grain size of floating grains and matrix, a lower abundance of silt aggregates, and higher taxonomic diversity. Group 5 (including Shingle Head) was characterised by a higher proportion of flat- lying taxa, and a dominance of displayed and multifoliate taxa. Their position correlates with intermediate values for explanatory variables.

8.3.2.3 Morphotype variables only active When morphotype characters were set as the only active variables, three groups were supported by inertia gain. Group 1 (including Sword Point and Spaniard’s Bay) was characterised by a higher-than- average abundance of upright, non-stalked taxa, a lower-than-average proportion of stalked taxa and an absence of flat-lying taxa (Appendix 2.2). Their position in the PCA (Fig. 8.2d) correlates with a coarser average grain size of matrix and floating grains, thin beds, lower taxonomic diversity and a lower abundance of silt aggregates (Fig. 8.2c). Group 2 (including Bed B and Shingle Head) was characterised by a higher-than-average proportion of stalked taxa (Appendix 2.2). Their position on the PCA space (Fig. 8.2d) correlates with coarser average but finer maximum grain size of floating grains (Fig. 8.2c). Group 3 (including Mistaken Point E and H14) was characterised by a higher-than-average abundance of Fractofusus and a higher-than-average proportion of flat-lying taxa (Appendix 2.2). Their position on the PCA (Fig. 8.2d) space correlates with thick beds, higher taxonomic diversity, a higher proportion of silt aggregates, and finer average and maximum grain size of floating grains and matrix (Fig. 8.2c).

151 Variable 1 Variable 2 Pearson's product-moment p-value correlation Thickness of bed below Abundance of Fractofusus 0.57 0.003 Total taxa Abundance of Fractofusus 0.49 0.01 Max. size fl oating grains Total taxa -0.3 0.15 Percentage fl oating grains Abundance of Fractofusus -0.28 0.17 Thickness of bed below Total taxa 0.28 0.18 Average matrix grain size Abundance of Fractofusus -0.27 0.2 Percentage of glebules Abundance of Fractofusus 0.22 0.28 Percentage of silt lenses Abundance of Fractofusus -0.2 0.33 Max. size of fl oating grains Abundance of Fractofusus -0.19 0.38 Percentage of glebules Total taxa -0.12 0.56 Table 8.3. The ten most highly correlated pairs of variables from the linear regression analysis of single pairs of variables.

Descriptor variable Formula p-value R2 adjusted R2 Total taxa -0.1048=-0.3505*PFG+0.2544*AFG- 0.084 0.35 0.21 0.3864*MFG+0.5*TBB Proportion fl at-lying taxa -4.29e-17=-0.3683*AMG 0.077 0.136 0.096 Proportion upright, non-stalked taxa 0.1757=[0,1][0.10977,-0.88246]*SL+[0,1][- 0.012 0.533 0.403 (including Beo-Cul) 1.27509,0.09397]*G-0.52646*TBB Proportion upright, non-stalked taxa -0.1525=0.2425*AMG+[0,1][-1.0450,0.5227]*G- 0.01152 0.533 0.4033 (excluding Beo-Cul) 0.3043*MFG-0.4586*TBB, Proportion stalked taxa (excluding 2.13e-16=0.2741*PFG-0.3386*MFG 0.139 0.171 0.092 Beo-Cul) Proportion stalked taxa (including -0.02938=0.13282*AFG-0.28745*MFG 0.3344 0.1038 0.01413 Beo-Cul) Abundance of Fractofusus -0.8808=[0,1][1.1891,-0.0240]*G-0.2135*PFG- 0.004 0.593 0.480 0.2446*MFG+0.5570*TBB Table 8.4. Results from the stepwise deletion regression analyses.

8.3.2.4 All variables active When all variables contributed actively to construction of the principal components space, three groups are supported by inertia gain. Group 1 (including Spaniard’s Bay, Bristy Cove and Pigeon Cove) was characterised by a higher-than-average proportion of upright, non-stalked taxa, a absence of Fractofusus and a lower-than-average proportion of flat-lying taxa and a dominance of unifoliate, furled and rotated taxa (Appendix 2.2). Their position in the PCA space (Fig. 8.2f) also correlates with coarser average grain size of matrix and maximum size of floating grains, and with lower taxonomic diversity (Fig. 8.2e). Group 2 (including Bed B, Shingle Head and the Port Union beds) was characterised by a dominance of multifoliate, displayed and unfurled taxa, and a higher-than-average proportion of flat-lying taxa. Their position in the PCA (Fig. 8.2f) space also correlates with thin beds, a lower abundance of silt aggregates, and coarser average but finer maximum grain size of floating grain size (Fig. 8.2e) the position of the majority of this group correlates with high taxonomic diversity. Group 3 (including Mistaken Point E and H14) was characterised by thicker-than-average beds, finer-than-average average grain size of floating grains, a higher-than-average abundance of Fractofusus and a lower-than-average proportion of

152 upright, non-stalked fronds. Their position in the PCA space (Fig. 8.2f) also correlates with finer average and maximum grain size of floating grains and of the matrix (Fig. 8.2e).

8.3.3 Summary of analytical results

The only pair of single variables with a statistically significant correlation (using Pearson’s product- moment coefficient) was between the thickness of the bed beneath the fossiliferous horizon and the abundance of Fractofusus (Table 8.2); when scaled, only thickness of bed and proportion of upright, non-stalked fronds was statistically significant (p<0.05). The analyses correlating multiple explanatory variables to descriptor variables were more successful, with strong, statistically significant (p<0.05) correlations identified for the abundance of Fractofusus and the proportion of upright, non-stalked fronds: Fractofusus is more abundant on surfaces characterised by low levels of disturbance (thicker beds and smaller grain size), and non-stalked fronds with intermediate disturbance (thinner beds and coarser average grain size of floating grains, but finer maximum size of floating grains). The best-fit models for total number of taxa present and the proportion of flat-lying fronds had p-values less than 0.1 (Table 8.3, 8.4), close to statistical significance: flat-lying taxa are more dominant when the matrix is finer grained, and taxonomic diversity is higher at low to intermediate levels of disturbance (thick beds and finer maximum grain size of floating grains, but coarser average size of floating grains). The only variable for which no model close to statistical significance could be identified was the proportion of stalked taxa (Table 8.3, Table 8.4).

The principal components analysis permits investigation of correlations between all variables, even when this correlation is not statistically significant (as identified through p-values), allowing subtle trends to be discerned. For all tests where sedimentary (explanatory) variables were used to construct the principal components space (when branching characters and morphotype variables were active and supplementary; Fig. 8.2a, b), variables indicating high disturbance (thin beds, coarse grain size, high proportions of coarse grain component) correlated with high proportions of upright taxa (stalked and non-stalked), while variables indicating low disturbance (thick beds, fine grain size, low proportion of coarse grain component) correlated with high proportions of flat-lying taxa and abundance of Fractofusus, although this taxon also correlates with abundance of silt aggregates for all analyses (Fig. 8.2a, b). Unifoliate, rotated and furled taxa correlate with variables characterised by higher levels of disturbance (Fig. 8.2a). Beds characterised by high levels of sediment influx are also characterised by low taxonomic diversity, as are those with low levels of sediment influx (where a correlation is determinable for the majority of the group), while beds characterised by intermediate and intermediate—low levels of sediment influx are characterised by high taxonomic diversity (e.g. Fig. 8.2, Table 8.4).

153 8.4 Discussion

Differences in community composition between Ediacaran beds in Newfoundland have been interpreted as reflecting ecological succession, and any environmental control rejected on the basis of their “similar” sedimentology (Clapham et al. 2003). In a study of eight communities from Mistaken Point, comparison of fossil densities, species richness and species diversity between communities led to the observed differences in taxonomic composition being attributed to ecological succession (Clapham et al. 2003). A simple succession from “frond-poor”, characterised by high abundances of flat-lying forms Fractofusus and Pectinifrons, to frond-rich communities, dominated by upright fronds such as Charnia and Charniodiscus, was proposed by those authors. However, low diversity communities are not always or exclusively comprised of taxa interpreted to represent early colonisers (Mitchell et al. 2015; Chapter 9), nor are the most diverse beds necessarily taxonomically similar in their composition. This suggests that other factors are influencing community composition. Four of the seven beds from that initial study were included in the analyses presented here and, despite superficial similarities in their sedimentology, subtle but quantifiable differences can be discerned (Appendix 2.2). The results of this study indicate that ecological succession is not the only control on community composition in these ancient ecosystems.

8.4.1 Interpretation of sedimentary fabrics and the coarse grain component

Although each type of disturbance (volcanic eruption, ice scour) is characterised by a particular set of physical and/or chemical effects, organisms respond to the mechanisms and consequences of disturbance (e.g. sediment influx, variation in current speed, changes in sediment or water chemistry) rather than its type per se. The consequences include variability in the sediment grain size of the substrate (e.g. through influx of sediment of a different grain size), thickness of burial events, the volumetric proportion and composition of suspended clays (Etter & Grassle 1992, Smit et al. 2008). While coarse grains may damage the organisms, and influence the settlement of water-borne propagules, clay particles may adhere to the surface of the organism, clogging its pores or absorptive surfaces. Accordingly, some organisms may be better able to withstand or avoid clay particle influence, and will therefore be better adapted to low ambient disturbance, while others may be better able to resist disturbance events bringing comparatively coarse sediment into the system. The mechanisms of disturbance include the lateral vs vertical mode of transport. Interpretation of the delivery mechanism for the coarse grain component is, therefore, critical for ascertaining their affect on the environment and the benthic communities.

8.4.1.1 Silt lenses and laminae Lenticular lamination in mudstones is recognised to reflect deposition by lateral currents (Schieber et al. 2010). The silt lenses are interpreted to represent sediment-starved ripples, or drapes of coarser sediment in mud ripples, taken by Rebesco et al. (2014) to indicate alternating flow condititions and low- to medium strength currents. The laminae could record more continuous lenses (on the scale of the thin

154 section), or alternatively the “striped” bedding of Rebesco et al. (2014), interpreted to record low current strengths and sediment dominantly from suspension (with the coarser laminae from lateral depostion or coarser hemieplagite). Accordingly, a higher proportion of silt lenses and silt laminae reflect more frequent influx of laterally-derived sediment.

8.4.1.2 Silt aggregates and floating grains There are several possible interpretations for the “silt aggregates”. The fact that they are not connected to each other suggests that they are not burrows ( cf. Menon et al., 2013). They bear a remarkable resemblance to agglutinated foraminifera (Macquaker et al. 2010, Lazar et al. 2015) and in some cases to fecal pellets (Plint et al. 2012), but without conclusive evidence this resemblance is interpreted as superficial. Perhaps the simplest explanation is that they record the silt pseudonodules which are common in mud turbidites (T7 ; Stow & Piper 1984), which result from sinking of sediment laminae or ripples into the mud through soft sediment deformation. However, these silt pseudonodules are typically irregularly shaped and can be quite large in maximum dimension (Cox 2012). A contourite origin for the “intraclasts” described from Spaniard’s Bay was rejected on the basis that adjacent intraclasts have different mineralogical compositions (Brasier et al. 2013), and that holds for the “silt aggregates”. An intraclast interpretation requires that they comprise aggregated clasts of mud which have been eroded up- current and redeposited, and that the source mud be sufficiently cohesive or stabilised to allow transport as an aggregate and rounding of clast edges (Schieber et al. 2010, Lazar et al. 2015). The rounded edges and irregular shapes of the silt aggregates are certainly consistent with this interpretation. It is odd however that the grain size of the “silt aggregates” is consistently coarser than the matrix, and that it comprises predominantly detrital quartz and other comparatively coarse siliclastic material rather than the very fine silt to clay grade material that comprises the bulk matrix. There are additionally no beds of comparable lithology to the “silt aggregates” in the succession that could have acted as a source for such intraclasts.

Silt clusters of comparable appearance and grain size have been described from the Soom Shale, where they are also associated with outsize, floating silt grains (Gabbott et al. 2010). There, the silt clusters were interpreted as silt aggregates either accumulating on small patches of benthic microbial mat from lateral grain transport, or settling through the water column as marine snow (as sunken clots of sediment, perhaps glued together by EPS and organic matter at the surface), as organo-mineralic aggregates (Plint et al. 2012, Lazar et al. 2015). These interpretations were based on the discontinuous nature of the silt lenses and their association in that locality with organic matter (Gabbott et al. 2010). Given the >50km distance from the nearest shoreline, the source of the silt found in the Soom Shale clusters and floating grains was interpreted to be wind-derived loess blown across sea ice. A wind-derived interpretation is reasonable for the successions included in this study, which are demonstrably deep-sea slope or basin

155 (Chapter 2). The presence of sea-ice at this time is not unreasonable given the palaeolatitude and relative temporal proximity of the Gaskiers glaciation (Chapter 1), although it may not be required given the steep slopes of volcanic island arc flanks and the uncertainty of the distance of any of these localities to the shore – particles <30μm in size can travel for tens of kilometres, and larger particles can travel up to 20km from the shore (Foda et al. 1985).

Accordingly, the “silt aggregates” and floating grains are interpreted as clots of organic matter and silt that have sunk though the water column, and which may have acted in the manner of modern “marine snow”. The coarse-grain fraction of the silt aggregates and the floating grains in the sections studied here are interpreted as wind-derived or possibly even ice-rafted debris, with the rare eu-/subhedral crystals likely representing components which entered the air column following a pyroclastic eruption. Their arrangement in bands may represent wind storms or periods of more frequent ice-rafting, or alternatively from periods of higher surface productivity producing greater amounts of organic matter with which the silt grains became clotted. The “silt aggregates” and floating grains thus record a different, much gentler, mode of sediment transport than the silt lenses. Their occurrence in discrete layers may reflect different events, or differential settling of aggregates/grains from one event as their relative densities and thus sinking rates are altered as they fall through the water column.

8.4.2 Correlating environmental and biological parameters

From both the PCA and the stepwise regression analyses, it is apparent that both the composition of the community and its taxonomic diversity correlate with ambient disturbance and energy levels as characterised by sediment influx.

8.4.2.1 Morphotype and sediment influx The concentration of suspended sediment decreases with height above the substrate surface (Hill & McCave 2001), and so it may simply be that in environments with higher levels of ambient disturbance, the concentration of suspended sediment is detrimental to flat-lying organisms, smothering or clogging their absorptive surfaces. The tendency for upright taxa to dominate surfaces characterised by higher levels of sediment influx makes intuitive sense – any incoming sediment would likely smother flat- lying taxa. In modern marine ecosystems, even a thin event bed covering the microbial mat can hinder recruitment (Lambshead et al. 2001, Dahms et al. 2004, Smit et al. 2008, Gates & Jones 2012), and this may have been particularly problematic for organisms such as Fractofusus which reproduced via stolon-like strategies (Mitchell et al. 2015): not only would the potential parent organisms be smothered, the substrate could have been rendered unfavourable for this mode of colonisation until the mat became re-established. It may also be the case that flat-lying forms are better adapted to dealing with clogging by fine clay particles that are swept away in faster current speeds than are upright forms – the two morphotypes are therefore adapted to different levels of ambient disturbance and current speed. 156 The abundance of Fractofusus and the proportion of flat-lying organisms on a surface correlate with fine grain size of matrix and floating grains, thick beds and rare/absent floating grains and silt lenses, but with high proportions of silt aggregates, while the proportion of upright taxa correlate with thin beds, coarse grain sizes and rare silt aggregates. The correlation of Fractofusus abundance with silt aggregates may reflect the pull of the MPE and H14 beds on this variable, or it may reflect the comparatively minor disturbance caused by the delivery mechanism interpreted for this type of sediment influx – the arrival of clots of sediment would be sporadic (not controlled by a gravity flow). It may even have been advantageous for this taxon, acting as a way of delivery nutrients (organic matter) to the substrate surface, in the manner of modern “marine snow” (section 8.6.1.2, cf. Gabbott et al. 2010).

The failure of stalked taxa to statistically significantly (p<0.05) correlate to any sedimentary parameter has a number of possible causes. First, it may be a genuine lack of correlation: stalked taxa (as defined here) are observed on most of the bedding surfaces (Appendix 2.2), but in variable proportions. Secondly, the inclusion of taxa with even a short stalk, and taxa with variable stalk lengths (“dusters”, Beothukis, Culmofrons; Primocandelabrum; Chapter 4) in the stalked morphotype, may bias the results – it may be that there is a critical threshold for stalk length to be beneficial. Thirdly, the result may similarly be biased by the inclusion of Charniodiscus – although the taxonomy of Charniodiscus as a genus is in serious need of revision, the consensus is that this genus is not related to rangeomorphs (Grazhdankin 2014, Laflamme et al. 2013). Its mode of life, tissue rheology and resistance to environmental disturbance may be very different to that of the rangeomorphs. Finally, there is a weak correlation (p=0.098) between the total number of taxa and the proportion of stalked taxa – this may reflect a competitive advantage of stalked taxa in diverse communities (cf. Laflamme et al. 2012b). It should also be noted that there is some debate in the ecological literature as to the overuse and over-reliance on p-values, as their indiscriminate use may mask subtle trends and those obscured by noise from “real” data (reviewed in Yoccoz, 1991).

8.4.2.2 Branching architecture and ambient disturbance The dominance of organisms with particular branching characters also correlates with sediment influx. Charnia-type branching (unifoliate, rotated and furled) correlates with sedimentary features interpreted to reflect higher levels of disturbance (thin beds, coarser grain size, high proportions of silt lenses and floating grains, but low levels of silt aggregates) than Bradgatia-type branching (multifoliate, displayed and unfurled). This suggests that there is a functional driver to branching architecture. Surfaces of sediment and organisms alike are mantled by a diffusive boundary layer (DBL; Jorgensen 2001, Jumars et al. 2001), the thickness of which is affected by the rate of flow (Jorgensen & Des Marais 1990) and surface roughness (Jorgensen 2001). Because the rate of diffusion is slow, the thinner the DBL over an organism’s surface, the more efficiently it can access solutes and particles transported in the flow – until physiological processes (such as transport of particles/solutes across cell walls/membranes) become the limiting factor in nutrient acquisition (Jumars et al. 2001). 157 The low-lying form Fractofusus, which dominates many surfaces in Newfoundland but is unknown from Charnwood Forest (Wilby et al. 2011), occupies that part of the water column characterised by lower velocity flows (cf. Ghisalberti et al. 2014). Its unfurled structure would have increased the surface roughness of the organism and consequently have ameliorated the lower incidence of nutrients in more quiescent, lower-flow regimes. Contrastingly, those forms that grow taller in the water column and therefore enter regions of higher-velocity flow (Orthiokaterna), and those which are associated with environments experiencing higher levels of physical disturbance (e.g. Charnia), have furled branches, particularly at higher orders. Furling of rangeomorph branches seemingly reduces the area of surface area exposed to the water column, seemingly counter-intuitive for absorption of nutrients (Liu et al. 2015a). It is possible that adopting this strategy provided some protection to the growing tips, the advantage of which outweighed the possible disadvantage of reducing surface area. This strategy complements the observation of novel growth features observed in Orthiokaterna, which are argued to result from sediment- or current-induced damage (Chapter 5). The extreme of the furled protective strategy may be represented by the development of a protective external sheath in Undosyrus (Chapter 6).

8.4.2.3 Taxonomic diversity and disturbance When total diversity is supplementary (thought to give a more accurate indication of true correlation (Husson et al. 2010b), this parameter correlates with intermediate to low levels of disturbance: it increases almost at right angles to parameters indicating high disturbance (percentage and average grain size of floating grains, and average matrix grain size) and to ones indicating low disturbance (thick beds). However, it is strongly negatively correlated to the abundance of silt lenses and the maximum grain size of floating grains (Fig. 8.2a, b). As the silt lenses are interpreted to record small-scale turbidite events, perhaps this reflects a limit on diversity imposed by disturbance events. The correlation between taxonomic diversity and thicker beds (when biological characters were active, Fig. 8.2c, e) is interesting. It may reflect pulling of this variable towards Mistaken Point E (MPE, Fig. 8.2), which has high diversity and a thick bed, and also towards H14 (H14.1a2, Fig. 8.2). One point to note is that, although twelve taxa have been identified on H14, 98.3% of individuals in the community are Fractofusus (Mitchell et al. 2015). Therefore, although it has a high taxonomic richness, it has a low taxonomic evenness. It may also reflect differential sampling intensities – the H14 surface has been subjected to more intense scrutiny than other Avalonian bedding surface outside of the Mistaken Point surfaces, allowing for the recognition of rare taxa. Similar sampling effects have been shown to have a significant influence on studies of taxonomic diversity in the modern (Mackey & Currie 2001).

While total taxa present on the surface is a crude measure of taxonomic diversity on the bedding plane, it is highly dependent on the surface area of the bedding plane and the degree of erosion of the bedding surface (which affects the ease and accuracy of taxonomic identification; e.g. Liu et al. 2015a, Mitchell

158 et al. 2015). It is also highly dependent on the current state of taxonomy. If current taxonomic diagnoses suffer from over-splitting of taxa due to lack of recognition of intra-specific variation (Chapter 4), then current values of taxonomic abundance likely overestimate species richness. Contrastingly, for surfaces with a high proportion of rare forms or of groups awaiting thorough taxonomic description, current values are also likely to be a underestimate of true diversity (Liu et al. 2015a). The trends of morphotype preference and taxonomic diversity are highlighted by two extremes: Sword Point (SP2) is dominated by Charnia masoni and is characterised by high levels of disturbance (thin beds, coarse and abundant floating grains; Fig. 8.2), whereas H14 (H14.1a) is dominated by Fractofusus and is characterised by very low levels of disturbance (thick beds, a fine-grained matrix and rare, fine-grained floating grains; Fig. 8.2). The four beds with the highest taxonomic diversity, Port Union 4 and Port Union, Mistaken Point E and Bed B, have intermediate to high levels of disturbance (Fig. 8.2).

One implication of the correlation between thick hemipelagite beds and high diversity is that it likely reflects communities which have had an extended period of time to develop. Fractofusus is considered to be an early colonising form (Chapter 8; Clapham et al. 2003, Mitchell et al. 2015): its abundance and even dominance in diverse communities requires further explanation, as other sites with comparable or higher taxonomic diversity (e.g. NQ Bed B, PU, PU.4a) have a much lower proportion of flat-lying organisms, and thinner beds, while other sites with thick beds (SP.1, PC) have much lower taxonomic richness (Fig. 8.2).

8.5.1 The persistence of Fractofusus

In modern communities, persistence of early colonisers in communities which would be considered late in the sequence of ecological succession (or that have had a long period of development) can be achieved in a number of ways (Platt & Connell 2003). In areas where disturbance is spatially variable, early colonisers can persist in refuges – either those characterised by high disturbance (where the substrate is periodically renewed), or those where conditions are sufficiently favourable for early colonisers that they are not displaced by taxa characteristic of a later stage in the community succession (Platt & Connell 2003). These are frequently areas where resources are not limited, as early colonisers are typically adapted to high-nutrient conditions (Peet & Christensen 1980, Pickett 1976).

If “silt aggregates” do indeed represent a mode of nutrient delivery, then high nutrient requirements may explain the association between Fractofusus and the abundance of “silt aggregates”. However, the spatial distribution of Fractofusus on both H14 and MPE indicate that there is no environmental control to their spatial distribution (Mitchell et al. 2015), so it is unlikely that they persisted in such spatial refuges on those sites. Alternatively, if conditions are favourable for their continued growth and reproduction, early colonisers may inhibit recruitment of later-stage taxa by increasing the mortality of these late invaders, or simply by occupying the space onto which they would become recruited (Connell 159 1997). A detrimental effect on late colonisers attributable to Fractofusus would be revealed by spatial analysis. Space occupation may explain the abundance of Fractofusus on the H14 surface: its stolon-like reproduction is argued to have permitted the rapid colonisation of these surfaces to densities of 42.8/m2 and to 98.3% of the total population (Mitchell et al. 2015) – if left undisturbed (as above thick beds with low ambient disturbance), then these organisms may have been hard to displace. In contrast, on surfaces where disturbance high enough to be detrimental to Fractofusus, or even to limit its competitive advantage for space occupation, then other taxa may colonise the surface.

8.6 Conclusions and future work

When examined in thin section, beds which appear very similar at outcrop scale show a remarkable variety of microscopic facies and sedimentary structures, including evidence of small-scale turbidites and ash flow deposits. The degree of sedimentary influx is investigated through a number of sedimentary parameters, including grain size, proportion of coarse grain component, and bed thickness. The proportion of flat-lying taxa, especially Fractofusus, correlate with sedimentary variables indicating a low degree of disturbance. The proportion of non-stalked upright taxa correlates positively with parameters indicative higher levels of disturbance (e.g. thin beds). The total number of taxa correlates with both low (thick beds, lower abundance and finer maximum grain size of floating grains) and high disturbance parameters (larger maximum grain size of floating grains). Future work should include investigation of non- parametric relationships between disturbance and taxonomic abundance.The low statistical significance of some of these correlations (Appendix) may reflect noise in the data, or simply that only one aspect of disturbance (sedimentary influx) is being investigated – it is not possible to quantify all factors that originally operated on the community. Additional bedding surfaces would help to constrain the results and improve their accuracy. It may also be that the numeric abundance of individuals which belong to a particular taxon is more strongly affected by the degree of disturbance than their presence/absence. Only eight sites have taxonomic abundance data (Chapter 7), that is, data on the proportion of individuals belonging to a given taxon. Furthermore, only one of those, Bed B, has abundance data which reflects current taxonomic identification, see Chapter 9). Until these issues are addressed, it is not possible to include taxonomic or morphotype abundance in the dataset.

While these results demonstrate a link between sediment influx and community composition, the model could be refined by inclusion of categorisation of the chemical composition of the sediment influx, either in qualitative terms (epiclastic vs volcaniclastic; Chapter 2), or even in quantitative terms (mineral composition, trace element composition). Inclusion of taxonomic evenness in addition to taxonomic richness, and percentage abundance of each taxon within the community, would doubtless serve to refine this study.

160 9. MULTIMODALITY: DISTURBANCE AND REPRODUCTION

9.1 Introduction

In modern communities, disturbance events affect the taxonomic composition and diversity of a community (Huston 1979), its structure (Huston & DeAngelis 1987), and even the path of succession that it takes (Dale et al. 2005). In fossil communities, ecological succession is impossible to identify directly, but various properties of the community, such as its composition and the size distributions (demographics) of its component populations, can be analysed (Darroch et al., 2013; Mitchell et al., 2015). The demographics of a population reflect the processes that have operated on it during its development, including reproductive frequency and disturbance, and relative rates of growth and mortality (Huston & DeAngelis 1987). The advantage of undertaking such studies on Ediacaran communities is that the constituent organisms are sessile and, with certain caveats (Liu et al. 2015a), can be treated as intact, in situ communities.

The population structures for three rangeomorphs (Fractofusus, Beothukis, and Pectinifrons) and one non-rangeomorph (Thectardis) have been studied to date, all from bedding surfaces in Newfoundland (Darroch et al. 2013, Mitchell et al. 2015). The distributions of Thectardis, Beothukis (referred to as “Charnia” in that study) and Pectinifrons have been interpreted be unimodal, suggestive of continuous recruitment or reproduction (Darroch et al. 2013). While Fractofusus has also been interpreted to have a unimodal population structure on the majority of bedding surfaces, (Darroch et al. 2013, Mitchell et al. 2015), a trimodal population structure has been demonstrated for this taxon on one bedding surface (Mitchell et al. 2015). Giant fronds have been reported from several Ediacaran localities (Fedonkin 2003, Narbonne & Gehling 2003), but perhaps the best known are the large individuals of Charnia masoni (formerly C. grandis) known from Bed B and Bradgate Park in Charnwood Forest (Boynton & Ford 1995, Brasier et al. 2012, Wilby et al. 2015). Indeed, one of the most notable features of the community on Bed B is its unusual abundance of “outsize” specimens, but other large (>20 cm) C. masoni individuals have been identified from Hofmann’s site 4 in Newfoundland (Hofmann et al. 2008), and from partial specimens on surfaces at Brigus South in Newfoundland. While these large individuals are typically considered as mere curiosities, they have the potential to provide important information about the community. The previous chapter discussed the influence of ambient disturbance on a community. This chapter analyses the population structures of individual taxa on all surfaces for which size data is currently available, and uses this to identify the influence of periodic, higher intensity disturbance events and reproductive processes on the preserved structure of the communities.

161 9.2 Material

Bed B from Charnwood Forest and eight bedding planes from Newfoundland were analysed for size data. Details of the casts of Bed B fossils and their specimen numbers are listed in Appendix 2.1. All measurements of specimens on Bed B (Chapter 1) included in this study were made by me from casts made by me. Organism height is taken as total length to centre of holdfast (Fig. 4.2a). Bedding planes from Newfoundland analysed in this study included the Mistaken Point D, E and G (MPE, MPD, MPG) surfaces; the Pigeon Cove (PC), Bristy Cove (BC), “lower mistaken point” (LMP) and Shingle Head (SH) surfaces in the Mistaken Point Ecological Reserve and environs (Clapham & Narbonne 2002, Darroch et al. 2013; and the H14 surface on the Bonavista Peninsula (Hofmann et al. 2008). For the first seven of these, retro-deformed specimen heights were taken from the dataset made available by Matthew E. Clapham, which was used in his studies of tiering and community succession (Clapham et al. 2003, Clapham & Narbonne 2002), and more recently to investigate population structures of individual taxa (Darroch et al. 2013). The H14 dataset was provided by Emily Mitchell, based on data collected by me, A. Liu and J. Matthews in 2012 (data collection method described in Mitchell et al. 2015). The measurements for Bed B were not retro-deformed, as strain has been shown to be even across this surface (Wilby et al. 2015).

9.3 Methods

Model-based clustering algorithms are commonly used to determine the number of mixing components in a population (Darroch et al. 2013, Mitchell et al. 2015, Tarhan et al. 2015), and that approach is adopted here. Model-based clustering analysis was run on the measurements for all taxa combined (i.e. the whole community) and for each taxon separately on each of the eight bedding surfaces.The model- based clustering method used is that described in Appendix 1.3.2, using the mclust package in R (Fraley et al. 2012), with the Bayesian Information Criterion (BIC, Appendix 1.1.1), classification and density distribution plots generated by that package to investigate the structure of the data. Normality of the data distributions was ascertained using the Shapiro-Wilks (Royston 1995, Shapiro & Wilk 1965) test, where normality is indicated by p-values above 0.05 (i.e. the null hypothesis, that the data is normally distributed, is not rejected).

Taxa for which fewer than ten individuals were identified on a surface were not analysed separately, with the exception of Orthiokaterna on Bed B. In some Ediacaran studies, values of specimen height are logged before running the model-based clustering algorithm (Darroch et al. 2013, Mitchell et al. 2015), but in others this step is not followed (Tarhan et al. 2015). For all beds, one iteration of each data subset was performed on raw (non-logged) values, and one on logged values to determine the extent to which the results are influenced by this step.

162 9.4 Results

The number of constituent populations most strongly supported by BIC values, alternative numbers which are not definitely rejected by the cluster models (i.e. those within 10 BIC; Appendix 1.3.2), approximate size boundaries to the supported sub-populations, and the results from the tests for normality for both whole and sub-populations, are also discussed (Table 9.1).

9.4.1 Bed B

For the raw (non-logged) data (Fig. 9.1), 2 sub-populations were supported for the whole community (Fig. 9.1a), for Charnia (Fig. 9.1b), for Charniodiscus (Fig. 9.1c) and for Primocandelabrum (Fig. 9.1d); 5 sub-populations were supported for Orthiokaterna (Fig. 9.1e), and 9 were supported for Undosyrus (Fig. 9.1f). In all cases where more than one sub-population was strongly supported, the largest few individuals (>25 cm) were placed into the second mixing component (Fig. 9.1), with the uncertainty of classification limited to a narrow range. A single individual of both Undosyrus and Orthiokaterna measures over 25 cm, and is placed in its own group. When whole populations are considered, Orthiokaterna is the only taxon whose size distribution passes the test for normality. For three of the other taxa (Charnia, Charniodiscus, and Undosyrus), the main group identified by the cluster analyses pass the test for normality; the second groups have only two specimens and the normality of their distribution cannot be analysed. When the Primocandelabrum population is split into two groups, the main group fails the test for normality and the second is too small to analyse. However, there is a certain amount of uncertainty over group assignment by the model at the junction between the groups because of their small size number. When group assignment is adjusted so that the seven largest specimens are excluded from the population, the main group passes the test for normality; the second one (comprising the seven largest specimens) does not, but as three groups are not definitely rejected it is possible that this group encompasses two sub-populations, with the largest specimen comprising its own sub-population. The individuals for this taxon cluster most strongly either side of a central depauperate area, producing a long tail towards larger specimens that is not seen in the other taxa (Fig. 9.1).

For the logged data (Fig. 9.2), 1 sub-population was supported for the community (Fig. 9.2a), Charnia (Fig. 9.2b) and Primocandelabrum (Fig. 9.2c); 2 were supported for Charniodiscus (Fig. 9.2d); 5 for Orthiokaterna (Fig. 9.2e); and 8 for Undosyrus (Fig. 9.2f). When logged, the whole population size distributions for Charnia, Primocandelabrum, Orthiokaterna and Undosyrus pass the tests for normality, but those for Charniodiscus do not. When Charniodiscus is divided into the two groups supported by BIC values, the main group distribution passes normality (the second group is too small to analyse).

163 9.4.2 Lower Mistaken Point

One specimen, measuring 189.5 cm, was excluded from the analysis as it is considerably (>1.4 m) larger than the next largest individual and would strongly bias the results – this individual is also not a rangeomorph. On raw data (Fig. 9.3), BIC values support 3 sub-populations for the community as a whole (<7 cm, 715 cm, 1535 cm; Fig. 9.3a), 2 for “Charnia 2” (<11 cm and 1120 cm; Fig. 9.3c) and for “ostrich feathers” (< 8 cm and 817 cm; Fig. 9.3d; 4 for “Charnia” (<2 cm, 27 cm, 720 cm, 2032 cm; Fig. 9.3b; and 1 for Charniodiscus (<22; Fig. 9.3e; Table 9.1) – other taxa had only a few specimens and so were not analysed. Charniodiscus is the only taxon whose whole population size distribution passes normality. When “Charnia” is divided into the four supported sub- populations, the size distributions for two of these pass normality but the other two fail . When “Charnia 2” is divided into two sub-populations, only one passes the test for normality. When “ostrich feathers” are divided into two sub-populations, both pass the test for normality.

On logged data (Fig. 9.4), BIC values support 2 sub-populations for the whole community (Fig. 9.4a), for “Charnia 2” (Fig. 9.4c) and for “ostrich feathers” (Fig. 9.4d); 1 is supported for “Charnia” (Fig. 9.4b) and for Charniodiscus (Fig. 9.4e). The three most abundant taxa on this surface, “Charnia”, “Charnia 2” and “Ostrich feathers”, had smooth distribution curves when logged (Fig. 9.4b—d). On density distribution plots for logged data, the presence of multiple sub-populations is particularly evident from the bimodal distributions of “Charnia 2” (Fig. 9.4c) and “ostrich feathers” (Fig. 9.4d), and from the hump on the shoulder of the distribution for all individuals (Fig. 9.4a). The logged size distributions for the whole populations of “Charnia” and Charniodiscus pass the test for normality; the others fail. When “Charnia 2” and “ostrich feathers” are divided into two sub-populations, the size distributions for each sub-population pass normality.

9.4.3 Mistaken Point D

On raw values (Fig. 9.5), 1 population was supported for Bradgatia (<19 cm; Fig. 9.5b) and “Charnia” (<27 cm; Fig. 9.5c); 2 sub-populations were supported for Pectinifrons (<30 cm, 3046 cm; Fig. 9.5d; and 3 were supported for Fractofusus (<7 cm, 714 cm, 1432 cm; Fig. 9.5e) and for the whole community (<9 cm, 916 cm, 1646 cm; Fig. 9.5a).Whole population size distributions for all taxa fail normality. When Fractofusus is divided into the three supported sub-populations, the size distributions for each sub-population fail the test for normality. When Pectinifrons is divided into two sub-populations, the larger one fails the test for normality but the smaller one (comprised of the few largest individuals, Fig. 9.5d) passes. The profile shape – with “humps” on the side of the main population indicates that the distributions of the various groups for Pectinifrons, for Fractofusus, and for all individuals overlap, rather than forming discrete populations.

164 When logged, 1 population was supported for the whole community, for Pectinifrons (the most numerically abundant taxon on the surface), for “Charnia”, and for Fractofusus; 2 sub-populations were supported for Bradgatia. The size distribution plots for the communtiy (Fig. 9.3a), “Charnia” (Fig. 9.3c), Pectinifrons (Fig. 9.3d) and Fractofusus (Fig. 9.3e) were smooth and symmetrical, and pass the test for normality. In contrast, the Bradgatia distribution has a clear bimodality (Fig. 9.3b), with the majority of individuals falling into one group comprising large-size individuals, and a second group comprising comparatively small-size individuals, forming a second peak on the left-hand flank. When divided into the two supported groups, both pass the test for normality.

9.4.4 Mistaken Point E

For raw values, 1 population was supported for Thectardis (>17 cm; Fig. 9.7i); 2 sub-populations were supported for Bradgatia (<8 cm, 812 cm; Fig. 9.7b), for “Charnia” (<11 cm, 1121 cm; Fig. 9.7c) and for “lobate discs” (<20 cm; 2030 cm; Fig. 9.7h); 3 for Fractofusus (<7 cm, 715 cm; 1533.5 cm; Fig. 9.7g) , “dusters” (<3 cm; 36 cm; 615 cm; Fig. 9.7f) and Charniodiscus (frond length, to account for the disparate stalk lengths of the species of Charniodiscus; Laflamme et al. 2004; <3 cm, 35 cm; 521 cm; Fig. 9.7d; and 4 for Charniodiscus (total height; <3 cm, 35 cm, 512 cm, 1229 cm; Fig. 9.7e) and the whole community (Fig. 9.7a). Thectardis is the only taxon whose whole population size distribution passes the test for normality. When divided into the supported number of sub-populations, “Charnia” is the only taxon for which all sub-population size distributions pass the test for normality. For “lobate discs”, only the larger of the two sub-populations (comprising the smaller individuals) passes the test for normality. For all populations where multiple populations are supported, the distributions of the sub-populations overlap (except for the “lobate disc” population, where the sub-population of larger individuals appears discrete).

When values where logged, 2 sub-populations were supported for the whole community, for “dusters”, for “lobate discs” and for Charniodiscus (both for total height and frond length); 1 population was supported for Fractofusus (the most numerically abundant taxon), for “Charnia”, and for Thectardis. The peak for the distribution curve for the community is skewed to the right, with a second peak on the left flank indicated by the change in slope of the distribution curve (Fig. 9.8a). The size distribution curves for Bradgatia (Fig. 9.8b), “Charnia” (Fig. 9.8b), “lobate discs” (Fig. 9.8f), Fractofusus (Fig. 9.8g) and Thectardis (Fig. 9.8h) are smooth and approximately symmetrical; the curve for “lobate discs” is additionally very tight around the peak. The size distribution for the whole populations of Bradgatia, Fractofusus and Thectardis all pass the test for normality. For “lobate discs”, the size distributions for the whole population and for both supported sub-populations all fail normality. The distribution curve for “dusters” (Fig. 9.8e) is smooth, but is skewed to the left, with a pronounced tail towards the right; the distributions for the whole population and for each of the two supported sub-populations fail the test for

165 normality. The density distribution curve for Charniodiscus (Fig. 9.8d) is bimodal, with the majority of individuals comprising a group of small-size individuals, and the remainder comprising a second group of larger-size individuals, visible as a second peak on the right-hand flank. When divided into the two supported sub-populations, the distributions for each sub-population fail for the total length, but pass when only frond length was considered.

9.4.5 Mistaken Point G

For raw values, 3 sub-populations were supported for the community (<7 cm; 720 cm; 2031 cm; Fig. 9.9a); 2 were supported for Bradgatia (<5 cm; 512 cm; Fig. 9.9b) and for the “unnamed frond” (<10 cm; 1014 cm; Fig. 9.9c). For both Bradgatia and the “unnamed fromd”, the whole population size distributions failed the test for normality, but those for each of the two supported sub-populations passed the test for normality. The two sub-populations for Bradgatia contain a similar number of individuals, but only the largest two individuals of the “unnamed frond” were placed in their own sub- population. The first and second sub-populations of Bradgatia and of the community overlap, but the third sub-population of the community (comprised of the two largest individuals on the surface, both Charniodiscus) and the second sub-population of the “unnamed frond” (comprised of the two largest individuals of that taxon) comprise a discrete population.

When all values were logged, 1 population was supported for the community (Fig. 9.10a), for Bradgatia (Fig. 9.10b), and for the “unnamed frond” (Fig. 9.10c). The density distribution curves for all individuals, Bradgatia and the “unnamed frond” are all smooth and symmetrical, and pass the test for normality.

9.4.6 Pigeon Cove

On raw values, 2 sub-populations were supported for the community (<17 cm, 1723 cm; Fig. 9.9d) 1 population was supported for Thectardis (<16 cm; Fig. 9.9e); 4 were supported for “Charnia” (<9 cm, 911 cm, 1120 cm; 2023 cm; Fig. 9.9f). The whole population size distributions for all individuals fail the test for normality, but the distributions for each individual taxa pass the test for normality. The shape of the “Charnia” distribution is irregular, but comprises three main peaks. When “Charnia” specimens are divided into the four supported populations, the size distributions of three of these pass the test for normality. The second sub-population for all individuals (comprised of the largest 5 individuals) is discrete. The Thectardis population distribution is smooth and symmetrical.

When values were logged, 1 population was supported for the community (Fig. 9.10d); 2 sub-populations were supported for Thectardis (Fig. 9.10e); and 4 for “Charnia” (Fig. 9.10f). The density distribution curve for the community is smooth and symmetrical. The curve for “Charnia” is erratic in its profile, with three peaks; it is not smooth, but it does pass the test for normality. When divided into the four

166 supported sub-populations, the size distributions for three of these pass the test for normality. The size distribution curve for Thectardis (Fig. 9.5f) is smooth and overall right-skewed; it is bimodal, with most individuals comprising a population of large-size individuals, and a second group comprising small-size individuals reflected by the second peak on the left-hand flank; it fails the test for normality.

9.4.7 Bristy Cove

On raw values, 2 sub-populations were supported for the community (<4 cm, 411 cm; Fig. 9.11a) and for Fractofusus (<4 cm, 411 cm; Fig. 9.11b); the only taxon on the surface with more than ten individuals). The whole size distribution for the whole Fractofusus population fails the test for normality. When divided into the two supported sub-populations, the size distribution for the largest population (comprised of the smaller-size individuals) just fails the test for normality, but that for the smaller population (the seven largest individuals) passes the test for normality. The populations appear discrete, but the distribution for the second sub-population of Fractofusus (the larger individuals) is smeared and does not form a prominent peak.

When values were logged, 2 sub-populations were supported for the community (Fig. 9.12a) and for Fractofusus (Fig. 9.12b). The density distribution curves for all individuals, and for Fractofusus, are smooth, and are left-skewed, with a small second peak on the right-hand flank. The whole population size distribution for Fractofusus fails the test for normality. When divided into the two supported sub- populations, the size distributions for both sub-populations pass the test for normality. Three Charnia individuals and one unidentified frond are also included in the second population for all individuals.

9.4.8 Shingle Head.

On raw values, 3 populations were supported for the community (<12 cm, 1242 cm, 4247 cm; Fig. 9.11c), and 2 populations were supported for Pectinifrons (<12 cm; 1222 cm; Fig. 9.11d), the only taxon on the surface with more than ten individuals). The size distribution for the whole Pectinifrons population and for each of the two supported sub-populations failed the test for normality. The second sub-population for both all individuals and for Pectinifrons overlap that of the first sub-population (the smaller individuals), whereas the third population for all individuals (the largest two individuals, both unidentified in the Clapham dataset, but likely to be Culmofrons and Primocandelabrum, pers. obs., 2012 field season) forms a discrete peak.

When values were logged, 1 population was supported for the community as a whole, and for Pectinifrons. The size distribution for the whole Pectinifrons population passes the test for normality. The density distribution curves for all individuals (Fig. 9.6c) and for Pectinifrons (Fig. 9.6d) are both smooth and approximately symmetrical, although the curve for all individuals does have a longer tail to the right hand side, reflecting two large specimens (one “Charnia”, and one unidentified frond).

167 9.4.9 H14 surface

On raw values, 3 populations are supported for Fractofusus (<5 cm, 511 cm, 1142 cm; Fig. 9.11e). The size distributions for both the whole population and the supported sub-populations fail the test for normality. The distributions for the first and second populations (small and medium sized individuals) overlap, but the third population (the largest individuals) form a discrete peak.

When logged, 2 populations were supported for Fractofusus (Fig. 9.12e). The size distributions for both the whole population and the supported population fail the test for normality. Two clear peaks are observed in the size distribution, but the flanks overlap.

9.4.10 Summary of results for each taxon

Thecardis is the only taxon for which only unimodal distributions are observed (MPE, PC); multimodal populations are most strongly supported for all other taxa for all (or the majority of) their populations on the studied bedding planes. Fractofusus most commonly forms three overlapping sub-populations (MPD, MPE and H14), but on one surface it has two discrete populations (BC); one point to note is that, on H14, the distributions for the small- and medium-sized individuals of Fractofusus are more strongly overlapping than the medium- and large-sized individuals (Fig. 9.11e). Bradgatia forms two overlapping populations (MPE, MPG) or one single population (MPD); Pectinifrons forms two overlapping sub- populations (MPD, SH); Primocandelabrum forms two discrete populations (Bed B) and “dusters” (which is dominantly Primocandelabrum; Mitchell et al. 2015) forms three overlapping populations (MPE); “ostrich feathers” form two discrete populations (LMP); Charnia forms two populations (Bed B; “Charnia” (Beothukis and/or Culmofrons) forms variable numbers of populations, with one (MPD), two (MPE) and four (PC), although the true taxonomic affinity is uncertain on this bed), but the sub- populations are always discrete; “Charnia 2” (Trepassia wardae, Mitchell et al. 2015) forms two discrete populations (LMP); Charniodiscus forms two discrete populations on Bed B, one on LMP and either three (frond only) or four (total length) overlapping populations on MPE.

There is no correlation between the number of populations and the maximum size of individuals, whether all individuals (Pearson’s product-moment correlation = 0.22, p=0.6036), all populations (Pearson’s product-moment correlation = 0.1, p=0.6277), or individual taxa (“Charnia”, Pearson’s product-moment correlation = 0.22, p=0.78) are considered. Fractofusus is the only exception, where the correlation is close to significance (Pearson’s product-moment correlation = 0.95, p=0.052), but as three of the four bedding planes (MPD, MPE and H14) are very similar in terms of their maximum size and have the same number of populations, the reliability of this correlation is uncertain.

168 9.5 Discussion: The myriad causes of multimodality

Multimodal size distributions can be generated through a number of different mechanisms, both inherent and imposed (Huston & DeAngelis 1987, Barry & Tegner 1990). Inherent multimodality reflects processes such as sexual dimorphism, genetic variation within a single cohort, polymorphism, and when there is a maximum size limit to growth (Power 1978, Huston & DeAngelis 1987). Multimodality may be imposed on a population through the coincidence of sequential cohorts, random environmental factors (e.g. substrate suitability), ecological effects (e.g. competition, neighbour-shading), and differential mortality as a result of varying vitality or of a random environmental event (Huston & DeAngelis 1987). Additionally, if the rate of growth is equalled by that of mortality, bimodality can be induced in a population through a “pileup” of individuals at the larger-size end (Barry and Tegner, 1990), especially if growth slows with age (Power 1978, DeAngelis & Mattice 1979). The various processes rarely occur in isolation, and frequently act in concert: for example, an existing bimodal signal may be enhanced if the size difference leads to a competitive advantage (Harper 1967, Harper 1977), and conversely may be suppressed if mortality is low and growth slows sufficiently with age for the two sub-populations to merge (Power 1978, DeAngelis & Mattice 1979, Barry & Tegner 1990).

There are two types of multimodality observed in the populations studied here: ones where the sub- populations overlap (Figs.; 9.6d, e; 9.8e, f, g; 9.10b; 9.12d), and ones where the sub-populations are distinct, with the modes producing clear, separate peaks (Figs. 9.2; 9.4d,e; 9.8h; 9.10c). Before the likely processes behind the population demographics can be ascertained, certain aspects of the data treatment must be investigated.

9.5.1 To log or not to log

The question of whether to log data before analysis is contentious in the literature (e.g. Keene 1995, Xiao et al. 2011, Changyong et al. 2014). Many biological phenomena, for instance the exponential growth of bacteria (Limpert et al. 2001), produce log-normal data, but most statistical analyses require the input data to have a normal distribution. Accordingly, biological data with skewed or otherwise non-normal distributions are often converted to produce normal distributions prior to analysis (Bak & Meesters 1998, Limpert et al. 2001, Meesters et al. 2001), and implementation of this step has become common practice in Ediacaran literature (Darroch et al. 2013, Zakrevskaya 2014, Mitchell et al. 2015), although not by Tarhan et al. (2015). However, the case of multimodal population distributions is a unique one: while the constituent sub-populations may have normal distributions, the whole population by definition does not – this is the primary assumption of the Mclust function which is used in these analyses, and the basis for recognition of sub-populations (Fraley & Raftery 2002, Fraley & Raftery 2007, Fraley et al. 2012). Indeed, in studies of modern communities investigating the extent and source

169 A B Bed B, All individuals Bed B, Charnia BIC BIC -960 -940 -920 -900 -880 -330 -310 -290 -270 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 density Number of components density Number of components 0.00 0.02 0.04 0.06 0.08 |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||| | || ||| | | | 0.00|| 0.02||||||||||||||| 0.04|||||||| |||| 0.06||| | | | 0.08 | | | | ||

0 10203040 0 5 10 15 20 25 30 35 data[, 3] data[1:41, 3]

C D Bed B, Charniodiscus Bed B, Primocandelabrum BIC BIC -190 -180 -170 -160 -420 -410 -400 -390 -380 -370

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 density Number of components density Number of components 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 | |||||||||| |||||||| | | | || 0.00|||||||||||||| 0.02|||||||||||||||| ||||||| || | | | 0.04| |||| | | | | | 0.06 | | |

10 15 20 25 30 10 20 30 40 data[42:68, 3] data[69:123, 3]

E F Bed B, Orthiokaterna Bed B, Undosyrus BIC BIC

-76 -74 -72 -70 -68 1 2 3 4 5 6 7 8 9

Number of components -68 -66 -64 -62 -60 -58 0.08 0.10 0.12 0.14

1 2 3 4 5 6 7 8 9 density density Number of components 0.0 0.2 0.4 0.6 0.00 0.02 0.04| 0.06 | ||| | || | || | | | | || | |

5 101520253035 4 6 8 10121416 data[126:134, 3] data[137:146, 3]

Figure 9.1. The population structures on Bed B, Charnwood Forest for raw (non-logged) values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, two populations (n = 146); B) Charnia, two populations (n = 41); C) Charniodiscus, two populations (n = 17); D) Primocandelabrum, two populations (n = 55); E) Orthiokaterna, five populations (n = 9); F) Undosyrus, nine populations (n = 10).

170 A B Bed B, All individuals Bed B, Charnia BIC BIC -360 -340 -320 -300 -130 -120 -110 -100 -90 -130 -120 -110 1 2 3 4 5 6 7 8 9 Number of components 1 2 3 4 5 6 7 8 9 Number of components density density 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

||||||||||||||||| || | || ||||| ||||||||||||||| |||| ||||||||||||||||||||||| ||||| | ||| | | |||||||||||| |||||| |||| | ||||||| | ||||||| || || | |||||| | || | | | | | | || | | | | 0.0 0.1 0.2 0.3||||| 0.4 0.5 0.6 0.7 |||||||||||||||||||||||||||||||||| | |

1234 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 datal1[1:145, ] data[1:41, ]

C Bed B, Charniodiscus D Bed B, Primocandelabrum BIC BIC -170 -150 -130 -60 -50 -40 -30 -20

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.5 1.0 1.5 | ||||||||||| |||||||| || | || 0.0||||| 0.1 0.2 0.3 |||| 0.4 || ||||||||||||||||| 0.5||| 0.6 ||| | | | | | ||| || | ||| | | | | |

2.0 2.5 3.0 3.5 1.01.52.02.53.03.54.04.5 data[42:68, ] data[69:123, ]

E F Bed B, Orthiokaterna Bed B, Undosyrus BIC BIC -26 -24 -22 -20 -18 -16 -14 -35 -30 -25 -20 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.2 0.3 0.4 0.5 0.6 0.8 0.7 012345 || |||| || | || | | | | || |

2.0 2.5 3.0 3.5 4.0 1.5 2.0 2.5 3.0 3.5 4.0 data[126:134, ] data[137:145, ]

Figure 9.2. The population structures on Bed B, Charnwood Forest for logged values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, one population (n = 145); B) Charnia, one population (n = 41); C) Charniodiscus, two populations (n = 17); D) Primocandelabrum, one population (n = 55); E) Orthiokaterna, one population (n = 9); F) Undosyrus, eight populations (n = 10).

171 A B LMP, All individuals LMP, “Charnia” BIC BIC -900 -880 -860 -840 -1700 -1650 -1600 -1550 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.00 0.05 0.10 0.15 0.20 |||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||| |||||||||||||||||||||| ||||||||||| |||||||||||||||||||||||| || ||||||| ||||||||| ||||||||| ||||||||||||||||||||||||||||||||| ||| |||||||||||| | | ||||||| | ||||||| ||| | |||| ||||||||||||||| || | | | 0.00|||||||||||||||||||||||||| 0.05|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 0.10|||||||||||||||||||| |||||| | ||||| 0.15 || | ||| | || || | |

0 5 10 15 20 25 30 35 5 1015202530 data[1:284, 2] data[1:151, 2]

C D LMP, “Charnia 2” LMP, “Ostrich feathers” BIC BIC -340 -320 -300 -280 -260 -240 -220 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.02 0.04 0.06 0.08 0.10 0.1 0.14

|| |||||||||||||| |||| ||| ||||| | ||||||| || |||||||||||| 0.00 0.05||| 0.10||||| 0.15||||||||||||||| 0.20|||||||||| 0.25 |||| |||| 0.30 || | |||| ||| |

5101520 246810121416 data[152:202, 2] data[225:278, 2]

LMP, Charniodiscus

Figure 9.3. The population structure on Lower Mistaken Point for raw (non-logged) values, showing BIC the density distribution plots (main), classification of clusters where identified (coloured lines, showing the -100 -95 -90 -85 -80 1 2 3 4 5 6 7 8 9 Number of components position of each specimen), and Bayesian Information

density Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating

0.02 0.03 0.04 0.05 0.06 component clusters have equal variance). A) All taxa, three populations (n = 284); B) “Charnia”, four ||||| | | || | | populations (n= 151); C) “Charnia 2”, two populations

5101520 (n = 51); D) “ostrich feather”, two populations (n = data[203:213, 2] 54); E) Charniodiscus, one population (n = 11).

172 A B LMP, All individuals LMP, “Charnia” BIC BIC -340 -320 -300 -280 -600 -580 -560 -540 -520 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.2 0.4 0.6 || |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||| |||| | |||| || ||| |||||| | | | | | 0.0 0.1||||| 0.2|||||||||||| 0.3|||||||||| 0.4||||||||||| 0.5||||||||||||||||||| 0.6|||||||||||||||||| 0.7 ||| ||||||||||||||||||||||||||||||||||| |||||| | ||||||| | ||| | || || | |

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 1.01.52.02.53.03.5 datal2[1:284, ] datal1[1:151, ]

C D LMP, “Charnia 2” LMP, “ostrich feather” BIC BIC -100 -90 -80 -70 -60 -50

-160 -140 -120 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.1 0.2 0.3 0.4 0.5 0.6 0.7

|| |||||||||||||| |||| ||| ||| | | | | ||||| ||||||||||||||| 0.0 0.2||| 0.4||||| 0.6 0.8||||||||| 1.0||||||||||| 1.2||||| || 1.4 || |||||| | |||| ||| |

1.0 1.5 2.0 2.5 3.0 1.0 1.5 2.0 2.5 datal1[152:202, ] datal1[225:278, ]

E LMP, Charniodiscus

Figure 9.4. The population structure on Lower

BIC Mistaken Point for logged values, showing the density distribution plots (main), classification of -40 -35 -30 -25

1 2 3 4 5 6 7 8 9 clusters where identified (coloured lines, showing the Number of components position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V”

density models, indicating component clusters have variable variance; grey triangles = “E” models, indicating

0.3 0.4 0.5 0.6 0.7 component clusters have equal variance). A) All taxa, two populations (n = 284); B) Charnia”, one population |||||||||| (n = 151); C) “Charnia 2”, two populations (n= 51);

1.5 2.0 2.5 3.0 D) “ostrich feather”, two populations (n = 54); E) datal1[203:213, ] Charniodiscus, one population (n = 11).

173 A B LMP, All individuals LMP, Bradgatia BIC BIC -16300 -16100 -15900 -800 -780 -760 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.002 0.004 0.006 0.008 0.010 0.01 0.014

0.000 0.002|||||||||||||||||||||||||||||||||||||| 0.004||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||| 0.006||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||| 0.008||| ||||||||||||||||||||||||||||||| |||||||||||||||| |||||| |||||| |||||||||| |||||| ||||||||| ||||||||||||||||||||||||||||||||||| |||||||||| 0.010 ||||||||||||||||||||| |||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||| |||||||||||| |||||||||||||||||||||| |||||| |||||| |||||||||| | ||||||| ||||| ||||| | | ||| | | | | |||||||||||||||||| | ||| | |||||||||||||||||||||||||||||||| | ||||||||||||||||| | |

100 200 300 400 60 80 100 120 140 160 180 200 data[1:1487, 2] data[1:76, 2]

C D LMP, “Charnia” LMP, Pectinifrons BIC BIC -220 -210 -200 -190

1 2 3 4 5 6 7 8 9 -2120 -2080 -2040 1 2 3 4 5 6 7 8 9 Number of components Number of components density density

0.000|||||| 0.002 0.004 ||||| 0.006 || | | | | |||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||| |||||| ||||||||||||| ||||||| | ||| | | | ||| | | | | 0.000 0.001 0.002 0.003 0.004 0.005 0.006

100 150 200 250 100200300400 data[77:93, 2] data[105:279, 2]

E LMP, Fractofusus

Figure 9.5. The population structures on Mistaken Point D surface for raw (non-logged) values, showing BIC the density distribution plots (main), classification of clusters where identified (coloured lines, showing the

-12250 -12150 -12050 1 2 3 4 5 6 7 8 9 Number of components position of each specimen), and Bayesian Information

density Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, three populations (n = 1487); B) Bradgatia, ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||| || ||||||||||||||||||| | ||||| 0.000 0.002 0.004 0.006 0.008 0.010 0.012 one population (n = 76); C) “Charnia”, one population 50 100 150 200 250 300 (n = 17); D) Pectinifrons, two populations (n = 175); E) data[280:1448, 2] Fractofusus, three populations (n = 1169).

174 A B MPD, All individuals MPD, Bradgatia BIC BIC -2250 -2150 -70 -50 -30

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.5 1.0 1.5 2.0

0.0||||||||||| 0.2|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 0.4|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||| |||||||||||||||||||||| ||||||||||||||||||||| |||||||||||| |||||||||| || 0.6||||||||| ||||||| |||||||| |||||||||||||||||||||||||||||||||| ||||||||||||||||||||| ||||||||||||||| ||||||||| ||||||||| ||||||||| |||||| | |||||| ||| ||||||||| |||||| |||||||||||| |||||| || |||||||| ||||||||||||||||| |||||||||| 0.8 1.0 ||||||| | |||||| |||||| ||| |||||||| | | |||||| || || | |||||| ||| ||| ||| |||| ||||||||| |||| ||| || | || | |||||||||| | | ||||||| || | ||| |||| || |||| || | ||| | | || || | | | || || ||| | | | | | |||||||||| |||||||||||| ||||||||||||||||||||||||||||||||||||||| | ||| ||| | |||

3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 1.82.02.22.42.62.8 datal1[1:76, 1] datal1[1:1487, 1]

C D MPD, “Charnia” MPD, Pectinifrons BIC BIC -50 -40 -30 -20 -300 -260 -220

1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 ||||||||||||| | | | | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||| ||||||||||| |||||||||||||| ||| ||||| || | | || ||| || |

2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 0.5 1.0 1.5 2.0 2.5 3.0 datal1[77:93, 1] datal1[105:279, 1]

E MPD, Fractofusus

Figure 9.6. The population structures on Mistaken Point D surface for logged values, showing the density BIC distribution plots (main), classification of clusters where identified (coloured lines, showing the position -1540 -1500 -1460 -1420 1 2 3 4 5 6 7 8 9 of each specimen), and Bayesian Information Criterion Number of components

density (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, one population (n = 1487); B) Bradgatia, two populations | 0.0 0.2 0.4 0.6 0.8 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||| |||||||||||||||||| ||||||||||||||||||||| |||||||||||| |||||||||||| ||||||||| ||| |||||||||| ||||||||||||||||||||| |||||||||||||||||||||||| |||||||||||| ||||||||| ||||| |||| |||||| ||| ||| |||| ||| ||||| || |||| ||| | |||||| ||||||| |||||||| | |||| | || ||| || | | ||| |||| | | ||| | |||| || | || | | | | | | ||| | | | | | | || | | | | | | | | | | | | | | | | (n = 76); C) “Charnia”, one population (n = 17); D)

3.54.04.55.05.56.06.5 Pectinifrons, one population (n = 175); E) Fractofusus, datal1[280:1448, 1] one population (n = 1169).

175 A B MPE, All individuals MPE, Bradgatia BIC BIC -2020 -1980 -1940 -33600 -33200 -32800 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.000 0.005 0.010 0.015 0.020 0.025

0.000 0.002 0.004||||||||||| ||||||||||||||||||||||| 0.006||||||||||||||||||||| ||||||||||||||| |||||||||||||||| 0.008|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 0.010|||||||||||||||||||||||||||||||||||||||||||||||||||| |||||| 0.012 |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||| |||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||| ||| |||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||| |||||||||||||||| ||||||||| ||||||||||| ||||||| ||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||| |||||||| |||||||||||||||||||||||| |||||||||||| | ||||||||||| |||||| ||||||||||||||||| |||||| | | |||| || |||||| |||| | || | || | ||| | | ||| ||| || || ||| | | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||| | || || |

0 50100150200250300 20 40 60 80 100 data[1:3231, 2] data[1:226, 2]

C D MPE, “Charnia” MPE, Charniodiscus BIC BIC -8400 -8200 -8000 -7800

-880 -860 -840 -820 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density

0.000 0.005|||||||||||||||||||| 0.010||||||||||||||||||||||||||||||||||||||||||||||||| 0.015|||||||||||||||||||||||||||||||||||||||||||||||||||| 0.020|||||||||||||||||||||||||||||||||||||||||||||||| 0.025 |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||| | |||||||||||||||||||||| | | |||

0.000||||| 0.005|| ||| | 0.010||| |||||||||||| |||| ||||||||| 0.015|||||||||| ||||||||||||||| 0.020 ||| ||||||||||||

50 100 150 200 0 50 100 150 200 250 300 data[227:310, 2] data[311:1135, 2]

E E MPE, Charniodiscus (frond only) MPE, “dusters” BIC BIC -7600 -7400 -7200 -7000 -3400 -3350 -3300 -3250 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.00 0.01 0.02 0.03 0.04 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||| | | || 0.000 0.005 0.010||||| 0.015|||||||||||||||||| 0.020|||||||||||||||||| 0.025|||||||||||||||| 0.030 |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||| ||||||||||| | ||||||||||||| || ||||||||||| | ||| ||||| |

0 50 100 150 200 20 40 60 80 100 120 140 data[311:1135, 4] data[1136:1497, 2]

Continued overleaf

176 G H MPE, Fractofusus MPE, “lobate discs” BIC BIC -1520 -1480 -15500 -15300 -15100 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.00 0.002 0.004 0.006 0.008 0.010 0.012

||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| | |||||| |||| | | | | ||| ||| || ||| | | 0.000 0.002 0.004||||||||||||||| 0.006 0.008 0.010 0.012 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||| |||||||||||||||| | || |||| | ||

50 100 150 200 250 300 50 100 150 200 250 300 data[1641:3137, 2] data[1498:1639, 2]

I MPE, Thectardis BIC -270 -260 -250 -240 1 2 3 4 5 6 7 8 9 Number of components density 0.005 0.010 0.015

||||||||||||||||||| | |||| |

60 80 100 120 140 160 data[3140:3164, 2]

Figure 9.7. The population structures on Mistaken Point E surface for raw (non-logged) values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, four populations (n = 3231); B) Bradgatia, two populations (n = 226); C) “Charnia”, two populations (n = 84); D) Charniodiscus, four populations (n = 825); E) Charniodiscus, frond only, three populations (n=825); F) “dusters”, three populations (n = 362); G) Fractofusus, three populations (n = 1496); H) “lobate discs”, two populations (n = 142); I) Thectardis, one population (n = 26).

177 A B MPE, All individuals MPE, Bradgatia BIC BIC -200 -160 -120

-6100 -6050 -6000 -5950 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4

0.0 0.1 0.2|| 0.3||||||||||||||||||| 0.4||||||||||||||||||||||||||||||||||||||||||||||| 0.5||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||| |||||||||| 0.6 |||||||||| |||||| |||||| |||||||||| |||||| |||||| |||||| ||||| ||||||| ||||||| |||||||||||||| ||| | || |||| ||||| || ||||||| | || | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||| ||||||||||||||||||| |||||||||| |||||||||||||||| |||||| ||||||||||||||||| |||||||||||| |||||| ||||||||||||| ||| | |||||||||| ||||| |||||| ||| ||| ||| | |||||||||||||| ||||||||||||| ||| |||||| ||| ||| ||||||||||||| |||| |||||||| |||||||||||||||||| ||| |||||| | ||||| | ||| ||||||||| ||||||||| | ||||||||| ||| |||||| |||||| || || | | || ||| |||| |||| ||||| | |||| || | | | || | | || || ||| ||| | || || || | | | | | | | | | | | ||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||| || ||||||||| ||||| ||||||||||| || | | || | | | | | | | | | | |

2345 3.0 3.5 4.0 4.5 datal1[1:3231, 1] datal1[1:226, 1]

C D MPE, “Charnia” MPE, Charniodiscus BIC BIC -1580 -1540 -1500 -140 -120 -100 -80 1 2 3 4 5 6 7 8 9 Number of components 1 2 3 4 5 6 7 8 9 Number of components density density 0.0 0.2 0.4 0.6 0.8 1.0 |||||||||||| ||||||||||||||||||||||||||||||||||| ||||||||| |||||| || | | || | 0.0 0.2 0.4 0.6 |||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||| |||||||||||||| ||||||||| || |||||||||| |||||| | ||||| ||| ||| |||||| || |||||| |||||| |||| ||| | | | || | | | | ||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||| |||| |||||| |||||||||||| ||||||| | |||| ||| || | | | | | || || | ||| | |

3.0 3.5 4.0 4.5 5.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 datal1[227:310, 1] datal1[311:1135, 1]

E E MPE, Charniodiscus (frond only) MPE, “dusters” BIC BIC -680 -660 -640

-1700 -1660 -1620 -1580 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 ||||||||||||||| ||||||||||||||||||||||||||||| |||||||||| ||||||||| ||||||||||| |||| |||||||||||||| |||||||| ||| ||||||| |||||| |||||||||||||||| ||||||||| |||||| ||| ||||||| ||||||||||||||||||| |||||||| | |||| || || ||||||||||||||||||||||| |||||||||||||||||||||| |||||||||||||||| ||| ||||| |||| || | |||||| | | |||| |||||||| | |||||| | 0.0|||| 0.2|||||||||||||||||||||||||||||||||||||||| |||||| 0.4|||||| ||||||||||||||||||||||||||||| ||||||||| |||||||||| 0.6 |||||| ||| ||| |||||| ||||| | | || || ||||||||||||||||||| ||||||||||||||||||||||||| ||||||| || ||| | | | ||

2345 2.5 3.0 3.5 4.0 4.5 5.0 datal2[1:825, ] datal1[1136:1497, 1]

Continued overleaf

178 G H MPE, Fractofusus MPE, “lobate discs” BIC -180 -160 -140 -120 -100 -2200 -2150 -2100 -2050 1 2 3 4 5 6 7 8 9 Number of components density density 0.0 0.2 0.4 0.6 0.8 0.0 0.5 1.0 1.5 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||| ||||||||||||||| | ||||||||||||| |||||||| ||||||||||| ||||||| |||||||||||||||||| ||| ||||||||||||| |||||||||||| ||| || ||||||||| ||||| |||||||||| ||||| ||| |||||| |||||||| ||||||||||||||||| | |||||| || ||| ||| |||||| ||||||| |||| |||| | |||| |||| ||| || | ||| || |||| || | | || | ||| | | | | | | | | ||| || | | | | | | ||||||||||||||| |||| |||||| ||||||||||||||||||||||||||||||||||||||||||| ||||||||||||| ||||||| |||||||||||||| | ||||| || || || || | | |

3.0 3.5 4.0 4.5 5.0 5.5 6.0 3.5 4.0 4.5 5.0 5.5 datal1[1641:3137, 1] datal1[1498:1639, 1]

C MPE, Thectardis BIC -60 -50 -40 -30 -20 1 2 3 4 5 6 7 8 9 Number of components density

0.0 0.5|||||| 1.0||||||| || 1.5 | ||||| | |

3.4 3.8 4.0 4.2 4.4 4.6 4.8 5.0 datal1[3140:3165, 1]

Figure 9.8. The population structures on Mistaken Point E surface for logged values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, two populations (n = 3231); B) Bradgatia, one population (n = 226); C) “Charnia”, one population (n = 84); D) Charniodiscus, two populations (n = 825); E) Charniodiscus, frond only, two populations (n=825); F) “dusters” two populations (n = 362); G) Fractofusus, one population (n = 1496); H) “lobate discs”, two populations (n = 142); I) Thectardis, one population (n = 26).

179 A B MPG, All individuals MPG, Bradgatia BIC -520 -510 -500 -490

1 2 3 4 5 6 7 8 9 Number of components density density 0.000 0.005 0.010 0.015 0.020 ||| ||||||||| || ||||| ||||||||||||||| |||||||||||||||||| ||||| |||||| ||||| |||| | | |||||||||||| ||||||| | | ||| | | | | | | || 0.000|||| 0.010|||||||||||||||||||| 0.020|||||||||||||| | 0.030 ||||||||| | | |

50 100 150 200 250 300 40 60 80 100 120 data[1:120, 2] data[1:55, 2]

C D MPG, “unnamed frond” PC, All individuals BIC BIC -1700 -1660 -1620 -290 -280 -270 -260 -250 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density

0.000 0.005 0.010|| | 0.015 ||| 0.020 || 0.025 | || 0.030 |||| |||||| | || | || || | || |||||||||||||| | ||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||| |||| |||| |||||||| | ||||| | | | || | || 0.000 0.002 0.004 0.006 0.008 0.010 0.012

20 40 60 80 100 120 50 100 150 200 data[73:100, 2] data[1:160, 2]

E F PC, “Charnia” PC, Thectardis BIC BIC -225 -220 -215 -210 -205 1 2 3 4 5 6 7 8 9 -1460 -1420 -1380 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.002 0.004 0.006 0.008 0.010 0.014 0.012

|||| | ||| ||| |||| ||| || | || |||||| |||||| | ||||| | |||||| |||| ||||||||||||||||||||||||||||||||||||||||||||||||| |||| ||||||||||| | |||||||||| |||| ||||||||| | |||| | | 0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014

50 100 150 200 20 40 60 80 100 120 140 160 data[1:18, 2] data[19:158, 2]

Figure 9.9. The population structures on Mistaken Point G (MPG) and Pigeon Cove (PC) for raw (non-logged) values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, MPG, three populations (n = 120); B) Bradgatia, MPG, two populations (n = 55); C) “unnamed frond”, MPG, two populations (n = 28); D) All taxa, PC, two populations (n = 160); E) “Charnia”, PC, four populations (n= 18); F) Thectardis, PC, one population (n = 140). 180 A B MPG, All individuals MPG, Bradgatia BIC -80 -70 -60 -50 -40 -30

-260 -240 -220 -200 1 2 3 4 5 6 7 8 9 Number of components density density 0.0 0.2 0.4 0.6 0.8 ||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||| ||| || || || | | | 0.0 0.2|| 0.4|| 0.6||||||||||||| 0.8||||||||||||| 1.0|||||| | 1.2| 1.4 | ||||| |

3.0 3.5 4.0 4.5 5.0 5.5 3.4 3.6 3.8 4.0 4.2 4.4 4.6 4.8 datal1[1:120, 1] datal1[1:55, 1]

C D MPG, “unnamed frond” PC, All individuals BIC BIC -80 -70 -60 -50

1 2 3 4 5 6 7 8 9 -300 -260 -220 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.2 0.4 0.6 0.8 |||||||||||| ||| |||||||| || | | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| | ||||||||||||||||||| |||||||| |||||||||||||||||||||||||| ||||| ||| | | | | 0.0 0.2 0.4 0.6 0.8

3.0 3.5 4.0 4.5 5.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 datal1[73:100, 1] datal1[1:160, 1]

E F PC, “Charnia” PC, Thectardis BIC BIC -55 -50 -45 -40 -35 -30

1 2 3 4 5 6 7 8 9 -240 -220 -200 -180 -160 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.0 0.2 0.4 0.6 0.8 1.0 0.00.51.01.5 ||||| ||| ||||||| | || ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||| ||||||||||||| |||| ||||||||||||||||||||||||||||| ||||||| ||

4.0 4.5 5.0 3.0 3.5 4.0 4.5 5.0 datal1[1:18, 1] datal1[19:158, 1]

Figure 9.9. The population structures on Mistaken Point G (MPG) and Pigeon Cove (PC) for logged values, showing the density distribution plots (main), classification of clusters where identified (coloured lines, showing the position of each specimen), and Bayesian Information Criterion (BIC) plots (inset, black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All individuals, MPG, one population (n = 120); B) Bradgatia, MPG, one population (n = 55); C) “unnamed frond”, MPG, one population (n = 28); D) All taxa, PC, one population (n = 160); E) “Charnia”, PC, four populations (n= 18); F) Thectardis, PC, two populations (n = 140). 181 A B BC, All individuals BC, Fractofusus BIC BIC -720 -700 -680 -640 -620 -600 -580 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.00 0.01 0.02 0.03 0.04 0.05 0.00 0.01|||||||||||||||||| 0.02|||||||||||||||||||| 0.03||||||||||||||| 0.04||||||| 0.05 || | | | | | ||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||| || |

20 40 60 80 100 20 40 60 80 100 data[1:84, 2] data[1:76, 2]

C D SH, All individuals SH, Pectinifrons BIC BIC -2980 -2940 -2900 -3550 -3450 -3350 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.000 0.005 0.010 0.015 ||||||||||||||| || |||||| |||| |||||||||| |||||||||||||||||||||||||||| |||||||||||| |||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||| | | ||||| | | | | 0.000||| |||||| 0.005||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 0.010|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| 0.015 ||||||||||||||||||||||||||||||||||||||| ||||||||||||| ||||| | ||||| | |

100 200 300 400 50 100 150 200 data[1:340, 2] data[14:316, 2]

E H14, Fractofusus Figure 9.11. The population structure on Bristy Cove (BC), Shingle Head (SH) and H14 for raw (non-logged) values, showing the density distribution plots (main), BIC classification of clusters where identified (coloured lines, showing the position of each specimen), and -8000 -7600 -7200 1 2 3 4 5 6 7 8 9 Number of components Bayesian Information Criterion (BIC) plots (inset,

density black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All taxa, BC, two populations (n = 84); B) Fractofusus, BC, two populations (n = 76); C) All 0.00 0.05 0.10 0.15 |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||| ||| |||| | taxa, SH, three populations (n = 340); D) Pectnifrons,

0 10203040 SH, two populations (n = 313); E) Fractofusus, H14, data[1:1214, 1] three populations (n = 1214).

182 A B BC, All individuals BC, Fractofusus BIC BIC -160 -140 -120

-190 -170 -150 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components Number of components density density 0.2 0.4 0.6 0.8

||||||||||||||||||| ||| ||||||| || |||||||| ||||||| | |||| | |||| | | |||| || | ||||||||||||||||||| ||| ||||| |||||||| ||||||| | ||| | | ||||| | | | || | 0.0 0.2 0.4 0.6 0.8 1.0

2.0 2.5 3.0 3.5 4.0 4.5 2.0 2.5 3.0 3.5 4.0 4.5 datal1[1:84, 1] datal1[1:76, 1]

C D SH, All individuals SH, Pectinifrons BIC BIC -360 -320 -280 -480 -460 -440 -420 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Number of components density density 0.0 0.2 0.4 0.6 0.8 |||||||||||||||||||||||||| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||| |||||||||||||||||||||| ||||||||||||||||||||||||||||| |||||||| ||||| |||||||||| ||||||| | |||||| |||| | || | | | | | | | 0.0||||| 0.2|||||| 0.4||||||||||||||||||||||||||||||||||||||||| 0.6||||||||||||||||||||||| |||| ||||||| 0.8|||||||||||||||||||||||| ||||||||||||||||||| |||||| 1.0 ||| |||||| ||||||| ||| |||||||| |||||| | |||| | || || | || | | | | | | | | | |

3.0 3.5 4.0 4.5 5.0 5.5 6.0 3.5 4.0 4.5 5.0 5.5 datal1[1:340, 1] datal1[14:316, 1]

E H14, Fractofusus Figure 9.12. The population structures on Bristy Cove (BC), Shingle Head (SH), and H14 for logged values, showing the density distribution plots (main), BIC classification of clusters where identified (coloured -2600 -2550 -2500 lines, showing the position of each specimen), and 1 2 3 4 5 6 7 8 9 Number of components Bayesian Information Criterion (BIC) plots (inset,

density black triangles = “V” models, indicating component clusters have variable variance; grey triangles = “E” models, indicating component clusters have equal variance). A) All taxa, BC, two populations (n = 84); B) Fractofusus, BC, two populations (n = 76); C) All 0.0 0.1|||||| 0.2||||||||||||||| 0.3||||||||||||||| 0.4|||||||||||||||||| ||| 0.5|||||||||||||||||||||||||||||| 0.6 |||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||| |||| | taxa, SH, one population (n = 340); D) Pectnifrons,

0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 SH, one population (n = 313); E) Fractofusus, H14, datal1[1:1214, 1] two populations (n = 1214).

183 of multimodality, it is the true (unmanipulated) size of the organism that is included in the analysis (e.g. Huston & DeAngelis 1987, Barry & Tegner 1990, Stohlgren 1992, Alessandrini et al. 2011, Portilla- Alonso & Martorell 2011). The data are only logged where it is explicitly appropriate to do so, for example in certain colonial corals: it has been argued that the size-frequencies of these communities can be modelled by log-normal functions, based on assumptions regarding the semi-spherical shape of the colonies and their attendant power-law increase in size (Bak & Meesters 1998). The same assumptions cannot be made for rangeomorphs.

When the results of the cluster analyses run on both raw and logged data are compared, it is apparent that including this step in the analyses (without separate analysis to determine size classes; Mitchell et al. 2015), strongly biases the results and the subsequent interpretations. For the majority of the populations, the size distributions generated from raw values are multimodal, but this structure is suppressed when the values are logged. Even though one population is supported when the values are logged, multimodality is frequently not definitively rejected from the BIC values – for most cases, the number of sub-populations next most strongly supported (within 10 BIC) is the same as the number of sub-populations identified from the raw values (Table 9.1). The reason behind this suppression of normality is clear – normality of distribution has been induced into the population(s) by logging the values, and so one population is identified by the Mclust algorithm. In cases where the raw values produce population structures that are unimodal and have a normal distribution, bimodality can be induced when the data are logged (e.g. Bradgatia on Mistaken Point D, Fig. 9.5b, 9.7b; Thectardis on Pigeon Cove, Fig. 9.9e, 9.11e). Again, however, one population is not decisively rejected based on the BIC values – the original structure is present, but is simply muted by the extra (unnecessary and inapproporiate) data manipulation. It is concluded here that the analyses run on raw values are most representative of the original demographic, and it is those values that are considered further.

9.5.2 The normality of the distributions

When analysed independently, the distributions for many of the sub-populations identified do not pass the test for normality (Table 9.1; e.g. Fractofusus on H14). This is most common for those populations where the peaks in the distributions of the less numerous sub-population(s) are expressed as humps or deflections on the shoulder of the distribution of the dominant population (i.e. that containing the greatest number of individuals), indicating a strong overlap of the sub-population distributions. This overlap induces a degree of uncertainty to the group assignment of individuals at the boundaries between the sub-populations. Consequently, the smallest individuals of the population comprising the larger-sized individuals will be included in the sub-population of smaller-sized individuals, and correspondingly the largest individuals of the sub-population of smaller-sized individuals will be included in the sub-population comprised

184 Shapiro-Wilks p (group 4) % (group 4) n (group 4) Shapiro- Wilks p (group 3) approx. group 3/4 division % (group 3) or raw values. or n (group 3) Shapiro- Wilks p (group 2) approx. group 2/3 division % (group 2) n (group 2) Shapiro-Wilks p (group 1) approx. group 1/2 division % (group 1) n (group 1) Shapiro-Wilks p (all) Shapiro-Wilks W (all) size (cm) <6 <10 n (all) max. populations 3 4 825 20.8 0.746 2.20E-16 495 60 3 8.95E-8 204 24.727 5 2.17E-8 126 0.153 3.60E-11 Taxon No. of Charnia 2CharniodiscusOrthiokaterna 2Primocandelabrum 5 2Undosyrus 3"Charnia" 9 6"Charnia 2" 41 3Charniodiscus 5 2 4 3 35"Ostrich feathers" 4 4 7 2 4 27 1 1 8 2 0.848 6 9Bradgatia 32.5 53 5 2"Charnia" 3 2 37.6 0.722 46 6.57E-5 2 9Fractofusus 3 3 0.861 1 4Pectinifrons 5 0.783 39 17 7.99E-6 1 4 3 3 0.098 151Bradgatia 95.122 0.651 2 2.02E-7 25 2 51 11 25 31.6"Charnia" 2 54 19.8 21.6Charniodiscus 92.593 3 51 3 0.830 0.000 2 30 16.8Charniodiscus frond 0.832 0.905 0.773 2only 96.226 4 4 0.723 5.66E-12 28"Dusters" 76 1 0.634 4.34E-6 0.212 2Fractofusus 17 17 18.9 9.02E-9 1169Thectardis 0.008 39 27 4.878 2 0.970 3 32 175 11.258"lobate disc" 3 50 4 2 45.8 76.471 0.883 2 0.907 1 7.407 0.071 11Bradgatia 92.593 2 0.956 226"unnamed frond" 8 3.51E-2 84 0.005 3.774 ts 2.20E-16 11.6 825 0.024 2 2.91E-5 21.2 2 29.1 0.967"Charnia" 421 75 0.429 ts 0.946Thectardis 0.765 2 158 12 36.014 3 4.05E-5 49.669 ts 362 1 7 4 90.286 4 1497Fractofusus 6.96E-8 23.529 2.20E-16 7 15.3 1 30 33.5 198 25 3 3 142 7.407 0.817 225 69 1.54E-9Pectinifrons 0.892 16.5 2 87.611 1 30 0.065 0.042 2 8 2 27.273 0.985 82.143 607 28 55 0.427 5 2.20E-16 3 2.20E-16 2 0.896 55 11 3 3 17 13.5 12.1 51.925 158 0.314 0.964 0.156 613 0.364 18 0.715 0.922 1.64E-8 14 3.12E-7 0.227 9.714 43.646 20 40.949 22.7 3 28 140 7 135 323 4.69E-6 1.06E-14 0.002 15 3 15.7 0.939 0.011 141 12.389 76 95.070 39.152 0.986 1.42E-5 26 20 17.857 2.74E-11 0.066 29 4 0.121 10.5 5 0.273 303 92.857 141 767 0.887 0.671 0.150 52.727 0.026 22.4 0.304 10 0.008 5 5 5.91E-9 38.950 51.236 0.112 0.912 0.122 9.22E-12 217 6 15 7 0.650 27.778 0.221 69 0.263 9 2.31E-5 4.930 2.20E-16 0.000 12 2 90.789 247 117 26 3.8 0.448 63 7.143 0.078 81.518 2.55E-7 47.273 0.174 12 60 0.047 3 2.64E-8 0.073 7.77E-5 16.667 7 ts 1.06E-10 5.17E-6 0.006 11 5.18E-6 56 9.211 2.20E-16 18.482 7 0.145 0.389 1.12E-5 20 0.645 3 0.167 0.424 Bedding Surface Bed B All 2LMP AllMPD 3 All 146 46MPE 3 All 285 188.5 4 1487 45.8MPG 25 AllPC 33.5 All 3 7BC AllSH 2 AllH14 9 Fractofusus 2 120 3 3 31.4 160 15 4 4 22.7 5 6 84 10.5 1214 16 340 41.6 47 0.838 2.20E-16 7 434 7 35.750 0.957 17 5 3.8 361 12 29.736 20 14 0.929 11 419 0.345 0.916 42 Table 9.1. Number of populations supported through model-based clsutering algorithms and normality of constituent populations f of populations supported through 9.1. Number Table 185 of larger-sized individuals. This mistaken assignment means that the distributions are artificially truncated, and one tail is missing from the distribution profile – consequently, the distributions fail the tests for normality. This can be demonstrated by manual adjustment of the group assignment of individuals, as has been done here for Primocandelabrum on Bed B. Where the distributions of the sub-populations are sufficiently discrete, forming a clear second peak with flanks either side of the peak, they pass the test for normality (e.g. Charnia on Bed B, Figs. 9.2b, 9.3b; “ostrich feathers” on Lower Mistaken Point, Figs. 9.4d, 9.5d; “Charnia” on Mistaken Point E, Figs. 9.8b, 9.9b; and the “unnamed frond” on Mistaken Point G, Figs. 9.10c, 9.11c).

9.5.3 Multiple, discrete populations

Several processes can result in a demographic where the distributions of the sub-populations have little to no overlap. Firstly, it may result from exaggeration of existing (overlapping) multimodality. This is the most likely explanation for the observation that the large-size sub-population on H14 has less overlap with the medium-sized sub-population than the medium- has with the small-sized sub- population; explained by the larger individuals having some competitive advantage over the smaller individuals. Secondly, it could reflect inherent dimorphism. Although there is no recorded evidence for discrete sexes or indeed of any reproductive structure in Ediacaran fossils (Mitchell et al. 2015), it is impossible to definitively rule out sexual dimorphism in these organisms. The possibility that the observed bimodality could reflect polymorphism (Huston & DeAngelis 1987) is intriguing, but there is no discernible morphological change associated with size, or of any taxon which is only represented by large-size individuals – at least, not since Charnia grandis has been conclusively demonstrated to represent larger individuals of Charnia masoni (Brasier et al. 2012, Wilby et al. 2015). Thirdly, it may result from interrupted or pulsed reproduction, where the time interval between generations (or the rate of growth) is sufficiently great. Finally, it could be explained by removal of part of an original population. This could be through selective removal of the medium-sized individuals or, more parsimoniously, through removal of the majority of an original population followed by recruitment of a new population to the surface – a combination of population removal and interrupted reproduction/ recruitment. If the reproductive events were pulsed or intermittent, there was a time interval before production of new recruits, perhaps due to the surviving individuals having to reach a critical maturity or size level (cf. Huston & DeAngelis 1987) – this is interrupted reproduction. Alternatively, if reproduction was continuous before and after the disturbance, then the break represents a lag before recruits could successfully settle on the surface – i.e. it is interrupted recruitment. There is no reason why the same process should be responsible for producing a similar demographic in populations in

186 different communities, which experienced differences in local environmental context (Chapter 7) – is it possible to support one process over another for any of the studied bedding planes?

9.5.3.1 Bimodality on Bed B The three most abundant taxa on Bed B have clear, discrete, bimodal size distributions, with only a very small proportion of the individuals comprising the large-size class (Table 1). This structure is even discernible in those taxa with only a few individuals (Undosyrus, Orthiokaterna): one is notably larger than the bulk of the population (plotting far from the others, Fig. 9.1e, f). Highly asymmetrical bimodality such as this is typically observed in the recovery stages of extant communities subjected to comparatively high rates of periodic disturbance (where these are associated with high mortality) – the few largest individuals are those which have survived one or more disturbance events, and the smaller individuals comprise the secondarily-recruited, recovery population (Stohlgren 1992). That the boundary between the small- and large-size sub-populations occurs at very similar values for each taxa, that a thin microtuff is observed 0.8 cm below the fossiliferous horizon (Appendix 2.2) and, crucially, that this microtuff overlays the holdfast disc of one of the largest individuals on the surface (Wilby et al. 2015), all support a single disturbance event interrupting reproduction or recruitment in the community on Bed B. The few individuals comprising the large-size sub-population are interpreted to be the survivors of the disturbance event, while the isolated discs observed on the surface are likely to be the remaining evidence of the culled individuals. The sequence of events is summarised for one taxon (Charnia masoni) in Figure 9.14. The small differences between taxa, in terms of the size of the largest individuals and of the size range of the smaller population, likely reflect taxon-dependent growth and/or recruitment rates.

The much larger size of the survivor individuals compared to the recovery population could represent a number of processes. Firstly, it is possible that if competition for nutrients was sufficiently intense in the recovery population (suggested by the high degree of overlap between taxa in that part of the water column, Chapter 7), or if nutrient availability was sufficiently restricted following the disturbance, this could induce the individuals affected by competition to maintain smaller sizes than they could optimally obtain (Sebens 1983, Johnson 1990). The survivor individuals may have grown much faster during the period when population density was much reduced following the disturbance event (Fig. 9.13c). An increase in growth rate following reduction in competition such as this has been interpreted to indicate a degree of plasticity to rangeomorph growth (Wilby et al. 2015). Secondly, it is possible that there was a considerable time delay between the disturbance event and recruitment of the recovery population. Substrate sensitivity of new recruits is common in modern communities (Lambshead et al. 2001, Smit et al. 2008), especially those in which interstitial microbial communities, themselves adversely affected by even thin burial (Gates & Jones 2012), are influential in aiding recruitment to the surface (Kirchman et al. 1981, Dahms et al. 2004). Finally, any height difference likely afforded the survivor individuals

187 A

B

C

D

E

Figure 9.13. Sequence of events leading to discrete bimodal population structures (e.g. on Bed B and LMP), shown for Charnia masoni. A) An incumbent population develops on an epiclastic silt lamina (green) and surficial biomat (plae blue); B) the population is smothered by a microtuff (orange), and all but a few individuals are killed - the holdfasts of those killed collapse and partially fill with ash; C) the survivors grow following the disturbance event, growing high into the water column; D) the biomat has re-established and a new cohort (recovery population) is recruited to the surface, forming a densely-populated understory; E) burial and in situ preservation of the community beneath the major event bed that felled all inhabitants.

188 a competitive advantage over the recovery population, as their fronds would have been in the higher, faster-moving and comparatively underexploited parts in the water column (Fig. 7.2), exaggerating the original bimodality.

9.5.4 Multiple, overlapping populations

The overlapping sub-populations can be produced in a number of different ways. It is possible to produce this population structure from a single cohort (i.e. resulting from a single reproductive event), either when subsets of the population have different growth rates, or where growth rates are homogenous across the population but a small proportion of the population was originally larger in size (Huston & DeAngelis 1987). It can also arise if the population is examined a short time after other processes, such as differential access to resources, have begun to operate – the bimodality becomes more pronounced over time (Huston & DeAngelis 1987). Differential access to resources from substrates can be rejected by spatial analysis (Mitchell et al. 2015), but this technique cannot identify differential access to resources due to, for example, canopy flow dynamics (cf. Ghisalberti et al. 2014). One particular aspect of this is asymmetric competition, where some individuals are under more intense competitive stress and so their growth rate is reduced, as is common in plant stands, particularly multi-generational ones – for example, individuals shaded by a larger neighbour are under more intense competition than those in more open areas and the larger individual itself, and so are smaller (Diggle 1976, Huston & DeAngelis 1987). In benthic marine communities, this could be achieved through preferential resource depletion in more densely-populated areas, deflection of the flow around a larger individual, or interception of a (particulate) resource falling through the water column by larger individuals. While such asymmetric competition can be rejected for the Fractofusus individuals on H14, where the medium- rather than the small-sized individuals cluster around the largest individuals (Mitchell et al. 2015), it cannot yet be rejected for other communities and populations.

Perhaps the simplest way of producing this demographic (overlapping sub-populations) is through pulsed reproduction, where the reproductive events are sufficiently closely spaced for the distributions of the individual cohorts to overlap – by combining size distribution data with spatial analysis, this has been neatly demonstrated in the Fractofusus population on H14, where the double-cluster pattern (clusters of clusters) was interpreted to record stolon-like reproduction (Mitchell et al. 2015). One issue arising from that study was that, while the same double cluster patterns observed on H14 were also observed on MPE and single clusters where observed on MPD, the population distributions on the latter two surfaces were found to be unimodal. It is possible to explain this disparity as the stolon-like reproduction on MPD and MPE being continuous in contrast to the pulsed reproduction on H14, perhaps due to more continuously favourable conditions. However, when the raw data is analysed (section 9.4), Fractofusus has three sub-populations on MPE, and two populations on MPD. Additionally, the size class boundaries and

189 maximum size reached on MPE are very similar to those of H14 (identified in this study and in Mitchell et al. 2015). Combined with the correlation between maximum size and the number of sub-populations, this similarity of absolute sizes could feasibly represent a threshold size/maturity to reproduction in this taxon. In this scenario, the populations on H14 and MPE are comprised of three generations (the three sub-populations, (Mitchell et al. 2015), and MPD is one stage earlier in the succession, with only two generations (cf. Mitchell et al. 2015).

On MPE, there is no evidence (demographic or sedimentological) for a disturbance event or other significant temporal interruption to reproduction – the sub-populations for all taxa with multimodal distributions overlap. Some of these taxa reach comparatively small maximum sizes (“dusters” and Bradgatia on MPE) and so, rather than representing post-disturbance recruitment, their comparatively small size and small population numbers may be explained by slower growth and delayed onset of reproductive maturity – this is typical of species which characterise and persist into late-stage successions where disturbance is low (Platt & Connell 2003).

9.5.5 Surfaces with multiple types of multimodality

While a disturbance event may be responsible for the discrete bimodality observed in the structure of populations on other surfaces, it is difficult to rule out other processes without sedimentological evidence of the event or the shared size boundary between populations observed on Bed B. Assuming that disturbance is responsible for the discrete bimodality on LMP (e.g. in “Charnia”), the fact that a greater proportion of the population comprises the large-size sub-population than on Bed B could indicate that a smaller proportion of the initial community was culled (by, for instance, a less intense disturbance event), or that the community was better able to recover from the disturbance event. Other taxa on this surface have multiple overlapping sub-populations (“ostrich feathers” and “Charnia 2”), and are also of comparatively small maximum sizes and low population numbers. It is possible that they reached reproductive maturity at a smaller size (younger age) than those with discrete sub-populations – this is a common trait of post-disturbance colonisers in modern environments (Platt & Connell 2003). On MPG, only two taxa have sufficient numbers of individuals to analyse. Bradgatia has overlapping populations, and the “unnamed frond” has discrete populations. Both reach similar maximum sizes, but are significantly smaller than the two Charniodiscus individuals on the surface (Fig. 9.9) – while the overlapping populations of Bradgatia may represent pulsed reproduction post-disturbance (as on LMP), the discrete populations of the “unnamed” frond may indicate that this taxon grew at a slower rate than the Charniodiscus individuals following the disturbance. The small numbers of Charniodiscus on this surface suggests that this taxon is less able to recover and reproduce from the disturbance event than, for example, Charnia.

190 While manual adjustment of group assignment can be used to counter mistaken assignment, and therefore improve the normality of distribution, there are a number of biological processes which could be involved. In Primocandelabrum, the subtle clustering of individuals in the bulk population (Fig. 9.1b) could reflect a subtle or incipient bimodality – perhaps pulsed reproduction. There is no evidence for this structure in other taxa or for a second event bed, and so it is unlikely that this subtle bimodality reflects a second culling event. There is a great deal of variation in the morphology of Primocandelabrum on this surface, particularly in stalk length – the non-normality of the distribution of the main Primocandelabrum sub- population likely represents the intra-specific variation documented in this taxon (Chapter 4). Unless morphologic studies of sufficient rigour are undertaken on other taxa, it is impossible to assign the cause of size disparity to such processes.

9.5.6 Unimodal distributions

The unimodal distributions of some taxa has been interpreted previously to record continuous reproduction (Darroch et al. 2013). However, it is also possible (and is indeed relatively common) for this demographic to result from a single reproductive or recruitment event (Huston & DeAngelis 1987). Original disparity in size of individuals can be genetic (some propagules being smaller upon their release or settlement), or simply due to subtle differences in growth rates reflecting either a genetics or an environmental control (Huston & DeAngelis 1987). Thus, the normal distributions of the recovery populations (e.g. on Bed B) could reflect a single recruitment event following the disturbance event. The same can be said of the unimodal distributions of the taxa for which the multimodality has been taken to record a reproductive signal – they may be continuous reproduction within each pulse, or they may record single (seasonal?) reproductive events. That Thectardis has only unimodal populations, even on beds where other taxa have multimodal distributions, highlights the difference between this taxon and all others on the surface. This taxon additionally does not appear to survive disturbance events. However, even where these events are not observed (e.g. MPE, Chapter 8, Appendix 8.2), there is no indication in Thectardis of the pulsed reproduction observed in other taxa.

9.5.7 The influence of disturbance on community composition

In modern communities, the trajectory followed during recovery (secondary succession) is determined by the magnitude and frequency of the disturbance event(s): where the spatial extent of disturbance is large relative to the recruitment ability of the community, recovery may be slow; where the intervals between disturbance events are short, the community may be held at an early succession stage and exhibit a high degree of variance with time; and where disturbance is intense, the reinstated community may differ from the original one (Turner et al. 1993, O‘Neill 1998, Dale et al. 2005). However, the composition of modern communities recovering from disturbance typically reflects the composition of any survivors (Smith & Hessler 1987) – the “biological legacy” of the disturbance (Franklin 1990, Dale

191 et al. 2005). This is in part due to the mediation of flow around survivors and the beneficial effect this has on recruitment (Gallagher et al. 1983, Eckman 1990, Eckman 1983), but also due to the fact that the survivors will be able to release propagules into the immediate vicinity of the community straight away, whereas other taxa can only be recruited to the surface from distant communities. Biological legacy not only refers to the influence of the survivors on the secondary community, but also of the remnants of the culled population. In Ediacaran systems, this could include some taxa preferentially using the decaying remnants of individuals (represented by iveshediomorphs, Liu et al. 2011) as a food source (Liu et al. 2015a). Such a preference is perhaps reflected in the positive correlation between “Charnia” (Beothukis) and iveshediomorphs on MPE (Mitchell, unpublished thesis). On MPD, while some taxa have discrete bimodality, others (Bradgatia and “Charnia”) have unimodal distributions – these taxa also have a smaller maximum size than those which display the multimodal distribution. The fact that these taxa have multimodal distributions on other surfaces indicates that they are capable of producing this structure under appropriate circumstances. It is possible that the entirety of an original population of Bradgatia and “Charnia” on MPD was removed by the disturbance event, or simply that they only arrived in this community post-disturbance. In either case, they would form part of the secondary succession community.

9.6 Conclusions

Several taxa are found to have multimodal population distributions, contrary to the findings of other studies (Darroch et al. 2013). In Fractofusus, this has been attributed to pulsed reproduction and multiple generations recorded in the same bed (Mitchell et al. 2015). This study also identifies the multiple overlapping sub-populations indicative of this process in Fractofusus on other bedding planes (MPE and MPD), but also in several other taxa, including Pectinifrons and Bradgatia. In other populations, the distributions of the sub-populations are discrete. There are a number of possible explanations for this, including exaggeration of multimodality from originally overlapping sub-population distributions through differential vitality with increasing size. However, the combination of sedimentological evidence and population structure on Bed B indicates that part of an incumbent population was culled by a disturbance event, and that a second, recovery population colonised the surface following the disturbance event. The underlying driver of multimodality on other populations may be distinguished in future through spatial analysis, to identify effects such as differential competition and/or neighbour shading, and/or through sedimentological analysis, to identify disturbance events similar to those which affected the Bed B community. In some communities, it is possible to identify species as components of a secondary or late-stage succession, through comparing their overall size distributions and the number of component populations of different taxa on the surface. The influence of the life history traits and ecological strategies of the organisms on the structure of their community are discussed in the final chapter of this thesis. 192 10. THE PALAEOECOLOGY OF AVALONIAN RANGEOMORPH COMMUNITIES

The ecology of an organism – how it interacts with its community and its physical environment – is a fundamental component of its biology. In modern systems, biology and environment interplay to produce the community structures and phenotypes observed, and there is no reason to assume that Ediacaran systems would have behaved differently. While predation has played a critical role in structuring the ecosystems of the Phanerozoic, it seemingly only appeared in the terminal Ediacaran (Hua et al. 2003; at least for macroscopic organisms (cf. Porter 2011). In addition to biological interactions, however, modern communities are affected by ambient disturbance and by discrete disturbance events, and these physical processes would have operated in Ediacaran ecosystems just as they did in Phanerozoic ones.

In modern communities, the reproductive mode and frequency of an organism reflect particular ecological strategies (Platt & Connell 2003). Organisms adopt one of two strategies: adaptation to competition, or tolerance to disturbance (White & Pickett 1985). Modern organisms in low disturbance environments are adapted to competition for resources. They typically favour quick growth and maturity, and rapid production of large amounts of offspring – especially the pioneering (colonising) and early-stage successional species. In contrast, organisms adapted to thrive in environments facing high or variable levels of disturbance typically favour wide dispersal ability of propagules, slow reproduction, and reach reproductive maturity slowly – especially late-stage successional species. These disparate strategies leave a distinctive signal in the demographic structure of a population (Chapter 9). Accordingly, by combining multiple lines of evidence, gleaned from petrographic study, detailed study of morphology and population ecology, it is possible to apply these same principles to ancient ecosystems. The Ediacaran ecosystems of Avalonia represent the first record of complex macrofossil communities. They provide a window onto the beginnings of the ecological structuring that characterises the Phanerozoic.

10.1 The ecological strategies of rangeomorphs and the succession of their communities

The successional pathway of a community is controlled by a number of factors, including the life histories and ecological strategies of the constituent organisms, and whether they facilitate or inhibit the recruitment of the taxa that come after them. Subjecting the community to a disturbance event can trigger a different (secondary) succession (Platt & Connell 2003, Dale et al. 2005). The influence of such events is determined by their severity, but also by the environmental resilience of the taxa within the community. This depends both on the stage in the primary succession that the community is in, but also on the ambient disturbance regime. Regions with higher levels of ambient disturbance will be populated by more environmentally resilient taxa, themselves better able to survive and recover from disturbance events than species adapted to low disturbance environments. Ambient physical disturbance has been 193 demonstrated to have a significant influence on the composition of Avalonian communities, in terms of the taxa present, the type of branching architecture that is dominant, and the taxonomic diversity of the community (Chapter 8).

10.1.1 Low-disturbance communities and adaptation of rangeomorphs to competition

One of the most recognisable rangeomorphs is Fractofusus. This genus is enormously abundant in Newfoundland, but is seemingly endemic to that region (Mitchell et al., 2015), barring a single specimen reported from NW Canada (Narbonne et al. 2014). Unusual among rangeomorphs, Fractofusus lacks a holdfast disc and adopted a flat-lying habit; it occupies the lowest levels of the water column (Chapter 7) and dominates communities in low-disturbance environments, such as H14 (Figs. 10.1, Fig. 10.2). Its unfurled branching architecture is interpreted to have increased its surface roughness, consequently reducing the thickness of the diffusive boundary layer over the organism and maximising its exchange efficiency with the water column (section 4.4.1). This would have ameliorated the lower incidence of nutrients in those parts of the water column characterised by lower flow speeds (Chapter 7, 8), enabling Fractofusus to thrive in low-flow environments. The comparatively rapid and prolific reproductive strategy of Fractofusus (Chapter 9; Mitchell et al. 2015) enabled it to out-compete other taxa for space, thereby preventing their colonisation of local surfaces.

Figure 10.1. Artist’s reconstruction of colonies of Fractofusus, with three generations depicted. Spatial patterns are those of the H14 surface, data provided by E. Mitchell. Its profilic and comparatively rapid reproduction allowed Fractofusus to thrive in early-stage successional communities, and especially in low-disturbance environments where they out-competed other taxa for space. Painting figured in Mitchell et al. 2015. 194 In environments characterised by low to intermediate physical disturbance (as characterised by proportion and grain size of sediment influx; Chapter 8), and those that lack discrete disturbance event(s) (Chapter 8, 9), such as the Mistaken Point E surface, Fractofusus was able to persist relatively late in the development of the community. This allowed a diverse community to develop that includes large populations of both early (Fractofusus) and middle—late-stage colonisers (Charnia/Beothukis; Figs. 10.2, 10.3). The upright fronds that are common in these environments, such as Beothukis and Bradgatia, typically have unfurled and displayed branching, and are rare in higher disturbance environments (like those recorded by Bed B). They likely gained the same increase in exchange efficiency with the water column as Fractofusus, perhaps giving them a competitive edge over fronds with furled architecture in these comparatively slow- flow, low-disturbance environments. While the dense space occupation of Fractofusus may have impeded settling of later-stage colonisers (stalked, multifoliate taxa – the “dusters” and Primocandelabrum), its dense occupation of the lowest levels in the water column may have additionally impeded their growth, preventing these late-stage taxa from reaching their usual heights (Chapter 7).

10.1.2 Higher disturbance environments and resilience to disturbance

In environments characterised by higher ambient disturbance (proportionally abundant and coarse influx of sediment, Chapter 8), the communities were dominated by upright taxa. This indicates that these were better able to persist in higher-disturbance environments than their low-lying counterparts (Chapter 8, 9). In environments characterised by a higher sediment influx and more frequent small-scale turbidite events, adopting an upright habit would have lifted the delicate frondose parts out of the lower parts of the water column where concentrations of suspended sediment would have been highest (Adams & Weatherly 1981). These higher energy environments are also characterised by rangeomorphs with furled and/or rotated architectures. In these organisms, placing the frond in a region of higher flow would have served to reduce the diffusive boundary layer over the frond. However, in doing so, the fronds may have been more susceptible to damage from current or neighbour abrasion (Chapter 5). Accordingly, any advantage afforded through protection of the growing tips by furling likely outweighed the increase in nutrient uptake that would have resulted from exposure of the furled parts of the frond to the water column.

Two rangeomorphs from Charnwood Forest, Orthiokaterna fordi gen. et sp. nov. and Undosyrus nemoralis gen. et sp. nov., reveal unique insight into the influence of ambient disturbance on the rangeomorph phenotype, and record adaptations that would have enhanced their tolerance to damage. “Eccentric branches” were interpreted as a growth response to damage, and are only observed in multifoliate rangeomorphs such as Primocandelabrum and Orthiokaterna (Chapter 5). This ability to recover from damage, whether caused by currents, sediment or abrasion by neighbouring branches, would have been especially beneficial in these higher-disturbance environments, serving to fill in space created in the

195 stolon-like reproduction reproduction

low high ambient ambient disturbance disturbance

disturbance event, culls part of existing population

disturbance event, culls all taxa except one or two

Figure 10.2. Ecological succession for rangeomorph communities. The blue to yellow arrow indicates increasing levels of ambient disturbance. The first row shows the initial colonising community that settles on the substrate, and the second row reflects the community that develops after a period of time in an environment of a given level of ambient disturbance. The third row describes the secondary succession community that develops after a disturbance event that culls a major part of the community in row 2. The composition of the recovery community reflects that of the survivors of the initial community. crown and increasing the surface area for absorption. Comparable growth responses are known from colonial organisms such as gorgonian octocorals, and recognition of this type of growth response in rangeomorphs likely indicates a similar level of modular integration and transport (Chapter 4). Based on comparison of its morphology with that of the branching architecture of the associated rangeomorph, the brush structure of Undosyrus (Chapter 6) was interpreted to have arisen through modification of rangeomorph elements. This adaptation of an existing structure is the first evidence of such developmental complexity in rangeomorphs, and the presence of both original and modified rangeomorph elements in a single individual indicates an unexpected complexity to their growth programmes. The brush structure was interpreted to have functioned as an external sheath, serving to protect the frondose parts within from current- or sediment-induced damage.

In intermediate—high disturbance environments (Chapter 8), the initial colonisation of Fractofusus and other low-lying taxa was likely prevented or held in check, allowing upright taxa to colonise the surfaces (Chapter 8; Fig. 10.2). Alternatively, where a disturbance event such as an ash-flow culled part of an

196 existing population (Chapter 9), any Fractofusus present on the surface would have been preferentially smothered while those with an upright morphology survived (Fig. 10.2). (Re-)colonisation of these surfaces by Fractofusus would necessarily have relied on recruitment of water-borne propagules sourced from other communities. In contrast, taxa with representatives that survived the disturbance event would have had a source of propagules or stolons already within the community, increasing their reproductive and re-population potential relative to those organisms that did not survive the event. The pause between the disturbance event and recruitment of the recovery population produces discrete bimodal population structures in the survivor taxon, and may reflect the time taken for the survivors to reach reproductive maturity, or for the substrate to become suitable for colonisation. Consequently, the succession of the community following the disturbance event is biased towards the composition of any upright fronds that survived the disturbance event (Lower Mistaken Point, Bed B; Fig. 10.2). Accordingly, high-diversity communities of upright taxa with only rare to absent flat-lying taxa (Chapter 8) developed in these environments (Mistaken Point G, Lower Mistaken Point; Fig. 10.2).

If ambient disturbance was sufficiently severe, then only the most environmentally resilient taxa (e.g. Charnia) would have persisted, giving rise to low-diversity communities of upright fronds (e.g. Sword Point; Fig. 10.2). In his synthesis of ecological occurrences of Ediacaran organisms, Grazdhankin (2004) noted that the wide ecological distribution of Charnia could reflect either transport into deep-water basins, or a wide ecological tolerance. The environmental resilience identified for this taxon (Chapter 8) may explain its persistence in higher-energy communities – it is an environmental generalist (cf. Platt & Connell 2003).

10.2 The evolution of a stem: response to competition or reproduction?

Inter-specific competition in these dense communities (Fig. 10.3) has been argued to have driven the morphological diversity of fronds, their elongation and the development of a stalk (Laflamme et al. 2012b, Hoyal Cuthill & Conway Morris 2014). However, the structures of most communities studied in this thesis suggest that competition for water-borne resources was minimal (Chapter 7). While the upright vs. flat-lying morphotype, and different branching architectures, are convincingly correlated to environmental parameters (Chapter 8), the advantage of a stalk is not so readily apparent. This is especially intriguing given the diversity in stem length in the stalked, multifoliate rangeomorphs of Charnwood Forest (Chapter 4, 5). While this could reflect indirect competition through downstream and neighbour effects (Chapter 7), an intriguing alternative may be found through comparison with modern communities. One explanation for increased height in both plant communities and the modern marine benthos argues that their height gives them a reproductive advantage (Gaylord et al. 2002, Thomson et al. 2011). By releasing their propagules into regions of higher flow speed, the propagules travel greater distances before settling, increasing the geographic spread of the taxon. The cosmopolitan geographic

197 Figure 10.3. Middle- to late-stage communities in low to intermediate disturbance environments, characterised by a diverse community in which early stage species persist. Community composition, densities and size distributions modelled after Mistaken Point E surface, data provided by E. Mitchell. Painting figured in Liu et al. 2015a. distribution of Charnia has been argued to support dispersal by water-borne propagules (Darroch et al. 2013), and its spatial distribution is different to those of taxa interpreted to have had a stoloniferous mode of reproduction (Mitchell et al. 2015). Therefore, rather than a method of avoiding inter-specific competition, it is possible that frond elongation and the development of a stem reflects the drive to release reproductive propagules higher into the water column, thereby increasing their dispersal distances.

10.3 The influence of rangeomorphs on their environment

The Ediacaran to Cambrian transition is marked by the development of animals with the power to change their ecosystems. Having established how environment influenced rangeomorphs, did rangeomorphs in turn have any impact on their environment? Whether they made their living through osmotrophy or filter feeding, they extracted dissolved or particulate organic matter, originally produced in the surface water, from the water column. This filtering of the water column has been argued to have played a critical role in oxygenation of the water column, and rangeomorphs may have enhanced this. By having their holdfasts in or beneath the mat, they would have functioned as a pathway between the sediment pore- waters and the water column.

Sessile organisms have an appreciable influence on the flow of water over the substrate (Ghisalberti & Nepf 2009). Where they occur in dense communities, upright and stalked organisms form a canopy,

198 and have a considerable influence on the flow profile within the community. The vortices they spall increase mixing between community and water column, and the waving of the canopy in response to the vortices serves to increase the flow velocity and shear stress within the canopy (Ghisalberti & Nepf 2009). Propagules often display taxon-specific preferences to flow profile, and the influence of existing organisms in the community can inhibit or facilitate the recruitment of other individuals on the surface, (Eckman 1983). Fractofusus was an early-stage coloniser, and its branching habit and rapid reproductive mode enabled it to dominate communities in environments characterised by low disturbance and low flow speeds (Chapter 7, 8). By acting as an obstacle and increasing surface roughness, small eddies would have been induced down-stream of Fractofusus, and the flow speeds would have locally been reduced even further. While its efficient colonisation of space would have impeded colonisation of the surface by other taxa, its influence on the flow profile may potentially have enhanced settlement of other taxa in those areas not yet colonised, allowing small but diverse communities of upright fronds to develop on the surfaces where Fractofusus was present (Chapter 7). In contrast, in dense (recovery?) communities of upright fronds, the flow profile created by the upright fronds would potentially have prevented the colonisation of the surface by Fractofusus, at least in large numbers (Lower Mistaken Point; Fig. 10.2). This inhibition vs. facilitation of succession could be identified through analysis of spatial relationships (Mitchell, unpublished thesis).

In addition to flow speeds, the propagules of many modern marine benthic organisms are sensitive to changes in substrate composition (Gates & Jones 2012), and their recruitment is strongly influenced by the microbial community present (Kirchman et al. 1981, Dahms et al. 2004). The upright/flat-lying composition of the community, organism density, and especially the proportion of iveshediomorphs (the decaying remnants of rangeomorphs/arboreomorphs) would potentially have influenced the microbial community present, through their interaction with the flow profile and the chemistry of the substrate pore-waters. This, in turn, may have affected the composition and the successional path of the community beyond the simple presence of a source of propagules in the surface.

The Avalonian communities are among the oldest known examples of complex macroscopic life. While much remains to be done to further elucidate the phylogenetic relationships within the groups and to extant phyla, we are slowly gaining some appreciation of their ecology, of their interaction with their environment and with each other. The correlation between environmental parameters and the taxonomic composition of a community, its diversity and the morphology of its constituent organisms highlights the importance of taking sedimentological context into account when interpreting rangeomorph ecology. While they represent but a step on the path to the complexities of modern ecosystems, the morphological and ecological complexity displayed by these ancient and enigmatic organisms is truly remarkable.

199 REFERENCES

ADAMS, C.E. & WEATHERLY, G.L. 1981. Some effects of suspended sediment stratification on an oceanic bottom boundary layer. Journal of Geophysical Research: Oceans, 86, 4161–4172, doi: 10.1029/ JC086iC05p04161. ADAMS, C.E. & WEATHERLY, G.L. 1981. Suspended-sediment transport and benthic boundary-layer dynamics. Marine Geology, 42, 1–18, doi: 10.1016/0025-3227(81)90155-9. ALESSANDRINI, A., BIONDI, F., DI FILIPPO, A., ZIACO, E. & PIOVESAN, G. 2011. Tree size distribution at increasing spatial scales converges to the rotated sigmoid curve in two old-growth beech stands of the Italian Apennines. Forest Ecology and Management, 262, 1950–1962, doi: 10.1016/j.foreco.2011.08.025. ALLEN, P.A. & ETIENNE, J.L. 2008. Sedimentary challenge to Snowball Earth. Nature Geosci, 1, 817–825, doi: 10.1038/ngeo355. ALLER, R.C. 2001. Transport and reactions in the bioirrigated zone. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 269–301. ANTCLIFFE, J.B. & BRASIER, M.D. 2007. Charnia and sea pens are poles apart. Journal of the Geological Society, 164, 49–51, doi: 10.1144/0016-76492006-080. ANTCLIFFE, J.B., CALLOW, R.H.T. & BRASIER, M.D. 2014. Giving the early fossil record of sponges a squeeze. Biological Reviews, 89, 972–1004, doi: 10.1111/brv.12090. ANTCLIFFE, J.B., HANCY, A.D. & BRASIER, M.D. 2015. A new ecological model for the 565 Ma Ediacaran biota of Mistaken Point, Newfoundland. Precambrian Research, 268, 227–242, doi: 10.1016/ j.precamres.2015.06.015. ÁRKAI, P., SASSI, F.P. & DESMONS, J. 2003. Very low-and low-grade metamorphism. In: Recommendations by the IUGS Subcommission on the Systema Tics of Metamorphic Rocks. BAK, R.P. & MEESTERS, E.H. 1998. Coral population structure: the hidden information of colony size-frequency distributions. Marine Ecology Progress Series, 162, 301–306. BAMFORTH, E.L. & NARBONNE, G.M. 2009. New Ediacaran Rangeomorphs from Mistaken Point, Newfoundland, Canada. Journal of Paleontology, 83, 897–913, doi: 10.1666/09-047.1. BARNES, D.K.. & CONLAN, K.E. 2007. Disturbance, colonization and development of Antarctic benthic communities. Philosophical Transactions of the Royal Society B: Biological Sciences, 362, 11–38, doi: 10.1098/ rstb.2006.1951. BARRIE, C.D., BOYCE, A.J., ET AL. 2009. On the growth of colloform textures: a case study of sphalerite from the Galmoy ore body, Ireland. Journal of the Geological Society, 166, 563–582, doi: 10.1144/0016-76492008- 080. BARRY, J.D. & TEGNER, M.J. 1990. Inferring Demographic Processes from Size-Frequency Distributions: Simple Models Indicate Specific Patterns of Growth and Mortality. Fishery bulletin, 88, 13–19. BAUMILLER, T.K. 1997. Crinoid functional morphology. In: Geobiology of Echinoderms. Paleontological Society Papers, 45–68. BENUS, A.P. 1988. Sedimentological context of a deep-water Ediacaran fauna (Mistaken Point, Avalon Zone, eastern Newfoundland). Trace Fossils, Small Shelly Fossils and the Precambrian-Cambrian Boundary, 463, 8–9. BILLINGS, E. 1872. Fossils in Huronian Rocks. Canadian Naturalist and Quarterly Journal of Science, 6, 478. BLACKWOOD, J.C., HASTINGS, A. & MUMBY, P.J. 2010. The effect of fishing on hysteresis in Caribbean coral reefs. Theoretical Ecology, 5, 105–114, doi: 10.1007/s12080-010-0102-0.

200 BLAND, B.H. 1984. Arumberia Glaessner & Walter, a review of its potential for correlation in the region of the Precambrian–Cambrian boundary. Geological Magazine, 121, 625–633, doi: 10.1017/ S0016756800030776. BLAND, B.H. 1994. Trace fossils in the Swithland Formation. Transactions of the Leicester Literary and Philosophical Society, 8, 27. BLAND, B.H. & GOLDRING, R. 1995. ‘Teichichnus SEILACHER 1955’ and other trace fossils (Cambrian?) from the Charnian of Central England. Neues Jahrbuch fur Geologie und Paleontologie, 195, 5–23. BOLES, J.R. & COOMBS, D.S. 1977. Zeolite facies alteration of sandstones in the Southland Syncline, New Zealand. American Journal of Science, 277, 982–1012, doi: 10.2475/ajs.277.8.982. BONDER, M.J., ABELN, S., ZAURA, E. & BRANDT, B.W. 2012. Comparing clustering and pre-processing in taxonomy analysis. Bioinformatics, 28, 2891–2897, doi: 10.1093/bioinformatics/bts552. BOTTJER, D.J. & AUSICH, W.I. 1986. Phanerozoic Development of Tiering in Soft Substrata Suspension-Feeding Communities. Paleobiology, 12, 400–420. BOUDREAU, B.P. 2001. Solute Transport above the Sediment-Water Interface. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 104–126. BOWMAN, A.W. & AZZALINI, A. 2014. R Package ‘Sm’: Nonparametric Smoothing Methods (version 2.3-5.4). BOYNTON, H.E. & FORD, T.D. 1979. Pseudovendia charnwoodensis – a new Precambrian arthropod from Charnwood Forest, Leicestershire. Mercian Geologist, 7, 175–177. BOYNTON, H.E. & FORD, T.D. 1995. Ediacaran fossils from the Precambrian (Charnian Supergroup) of Charnwood Forest, Leicestershire, England. Mercian Geologist, 13, 165–182. BOZDOGAN, H. 1987. Model selection and Akaike’s Information Criterion (AIC): The general theory and its analytical extensions. Psychometrika, 52, 345–370, doi: 10.1007/BF02294361. BRASIER, M. & ANTCLIFFE, J. 2004. Decoding the Ediacaran Enigma. Science, 305, 1115–1117, doi: 10.2307/3837606. BRASIER, M.D. & ANTCLIFFE, J.B. 2009. Evolutionary relationships within the Avalonian Ediacara biota: new insights from laser analysis. Journal of the Geological Society, 166, 363–384, doi: 10.1144/0016-76492008-011. BRASIER, M.D., ANTCLIFFE, J.B. & LIU, A.G. 2012. The architecture of Ediacaran Fronds. Palaeontology, 55, 1105– 1124, doi: 10.1111/j.1475-4983.2012.01164.x. BRASIER, M.D., LIU, A.G., MENON, L., MATTHEWS, J.J., MCILROY, D. & WACEY, D. 2013. Explaining the exceptional preservation of Ediacaran rangeomorphs from Spaniard’s Bay, Newfoundland: A hydraulic model. Precambrian Research, 231, 122–135, doi: 10.1016/j.precamres.2013.03.013. BRETT, C.E. & LIDDELL, W.D. 1978. Preservation and paleoecology of a Middle Ordovician hardground community. Paleobiology, 4, 329–348. BRIGGS, D.E.G. 2003. THE ROLE OF DECAY AND MINERALIZATION IN THE PRESERVATION OF SOFT- BODIED FOSSILS. Annual Review of Earth and Planetary Sciences, 31, 275–301, doi: 10.1146/annurev. earth.31.100901.144746. BRIGGS, D.E.G., BOTTRELL, S.H. & RAISWELL, R. 1991. Pyritization of soft-bodied fossils: Beecher’s Trilobite Bed, Upper Ordovician, New York State. Geology, 19, 1221–1224, doi: 10.1130/0091-7613(1991)019<1221: POSBFB>2.3.CO;2. BRIGGS, D.E.G., RAISWELL, R., BOTTRELL, S.H., HATFIELD, D. & BARTELS, C. 1996. Controls on the pyritization of exceptionally preserved fossils; an analysis of the Lower Devonian Hunsrueck Slate of Germany. American Journal of Science, 296, 633–663, doi: 10.2475/ajs.296.6.633. BRISTOW, T.F. & KENNEDY, M.J. 2008. Carbon isotope excursions and the oxidant budget of the Ediacaran atmosphere and ocean. Geology, 36, 863–866, doi: 10.1130/G24968A.1.

201 BROWNE, N.K., PRECHT, E., LAST, K.S. & TODD, P.A. 2014. Photo-physiological costs associated with acute sediment stress events in three near-shore turbid water corals. Marine Ecology Progress Series, 502, 129–143, doi: 10.3354/meps10714. BUATOIS, L.A., NARBONNE, G.M., MÁNGANO, M.G., CARMONA, N.B. & MYROW, P. 2014. Ediacaran matground ecology persisted into the earliest Cambrian. Nature Communications, 5, 3544, doi: 10.1038/ncomms4544. BULL, E.E. 1996. Implications of normal and abnormal growth structures in a Scottish Silurian dendroid graptolite. Palaeontology, 39, 219–240. BURGE, D.O. & ZHUKOVSKY, K. 2013. Taxonomy of the Ceanothus vestitus Complex (Rhamnaceae). Systematic Botany, 38, 406–417, doi: 10.1600/036364413X666750. BURZYNSKI, G. & NARBONNE, G.M. 2015. The discs of Avalon: Relating discoid fossils to frondose organisms in the Ediacaran of Newfoundland, Canada. Palaeogeography, Palaeoclimatology, Palaeoecology, 434, 34–45, doi: 10.1016/j.palaeo.2015.01.014. BUTTERFIELD, N.J. 2009a. Macroevolutionary turnover through the Ediacaran transition: ecological and biogeochemical implications. Geological Society, London, Special Publications, 326, 55–66, doi: 10.1144/ SP326.3. BUTTERFIELD, N.J. 2009b. Modes of pre-Ediacaran multicellularity. World Summit on Ancient Microscopic FossilsWorld Summit on Ancient Microscopic Fossils, 173, 201–211, doi: 10.1016/j.precamres.2009.01.008. CAI, Y., SCHIFFBAUER, J.D., HUA, H. & XIAO, S. 2012. Preservational modes in the Ediacaran Gaojiashan Lagerstätte: Pyritization, aluminosilicification, and carbonaceous compression. Palaeogeography, Palaeoclimatology, Palaeoecology, 326–328, 109–117, doi: 10.1016/j.palaeo.2012.02.009. CALLOW, R.H.T. & BRASIER, M.D. 2009. Remarkable preservation of microbial mats in Neoproterozoic siliciclastic settings: Implications for Ediacaran taphonomic models. Earth-Science Reviews, 96, 207–219, doi: 10.1016/j.earscirev.2009.07.002. CANFIELD, D.E. 1989. Reactive iron in marine sediments. Geochimica et Cosmochimica Acta, 53, 619–632, doi: 10.1016/0016-7037(89)90005-7. CANFIELD, D.E., RAISWELL, R. & BOTTRELL, S.H. 1992. The reactivity of sedimentary iron minerals toward sulfide. American Journal of Science, 292, 659–683, doi: 10.2475/ajs.292.9.659. CANFIELD, D.E., POULTON, S.W. & NARBONNE, G.M. 2007. Late-Neoproterozoic Deep-Ocean Oxygenation and the Rise of Animal Life. Science, 315, 92–95, doi: 10.1126/science.1135013. CANFIELD, D.E., SIMON W. P OULTON, KNOLL, A.H., NARBONNE, G.M., ROSS, G., GOLDBERG, T. & STRAUSS, H. 2008. Ferruginous Conditions Dominated Later Neoproterozoic Deep-Water Chemistry. Science, 321, 949–952, doi: 10.2307/20144603. CANNELL, M.G.R. & GRACE, J. 1993. Competition for light: detection, measurement, and quantification. Canadian Journal of Forest Research, 23, 1969–1979, doi: 10.1139/x93-248. CAREY, D.A. 1983. Particle Resuspension in the Benthic Boundary Layer Induced by Flow around Polychaete Tubes. Canadian Journal of Fisheries and Aquatic Sciences, 40, s301–s308, doi: 10.1139/f83-291. CARNEY, J.N. 1999. Revisiting the Charnian Supergroup: new advances in understanding old rocks. Geology Today, 15, 221–229. CARNEY, J.N. 2000. Charnwood Forest. In: Carney, J. N., Horak, J. M. & Pharaoh, T. C. (eds) Precambrian Rocks of England and Wales. Peterborough, Joint Nature Conservation Committee, 19–52. CARNEY, J.N. 2000. Igneous processes within late Precambrian volcanic centres near Whitwick, north-western Charnwood Forest. Mercian Geologist, 15, 7–28. CARNEY, J.N., ALEXANDRE, P., PRINGLE, M.S., PHARAOH, T.C., MERRIMAN, R.J. & KEMP, S.J. 2008. 40Ar–39Ar isotope constraints on the age of deformation in Charnwood Forest, UK. Geological Magazine, 145, 702–713, doi: 10.1017/S0016756808004950.

202 CARNEY, R.S. 2005. Zonation of Deep Biota on Continental Margins. In: Gibson, R. N., Gordon, J. D. M. & Atkinson, R. J. A. (eds) Oceanography and Marine Biology. CRC Press, Oceanography and Marine Biology - An Annual Review, 211–278. CAVALAZZI, B., AGANGI, A., BARBIERI, R., FRANCHI, F. & GASPAROTTO, G. 2014. The formation of low-temperature sedimentary pyrite and its relationship with biologically-induced processes. Geology of Ore Deposits, 56, 395–408, doi: 10.1134/S107570151405002X. CHAMBERS, J.M., CLEVELAND, W.S., KLEINER, B. & TUKEY, P.A. 1983. Graphical Methods for Data Analysis. Wadsworth International Group. CHANGYONG, F., HONGYUE, W., NAIJI, L.U., TIAN, C., HUA, H.E., YING, L.U. & OTHERS. 2014. Log-transformation and its implications for data analysis. Shanghai Archives of Psychiatry, 26, 105. CHAPMAN, G. 1981. Individuality and modular organisms. Biological Journal of the Linnean Society, 15, 177–183, doi: 10.1111/j.1095-8312.1981.tb00757.x. CLAPHAM, M.E. & NARBONNE, G.M. 2002. Ediacaran epifaunal tiering. Geology, 30, 627–630, doi: 10.1130/0091- 7613(2002)030<0627:EET>2.0.CO;2. CLAPHAM, M.E., NARBONNE, G.M. & GEHLING, J.G. 2003. Paleoecology of the Oldest Known Animal Communities: Ediacaran Assemblages at Mistaken Point, Newfoundland. Paleobiology, 29, 527–544. CLAPHAM, M.E., NARBONNE, G.M., GEHLING, J.G., GREENTREE, C. & ANDERSON, M.M. 2004. Thectardis avalonensis: a new Ediacaran fossil from the Mistaken Point biota, Newfoundland. Journal of Paleontology, 78, 1031– 1036, doi: 10.1666/0022-3360(2004)078<1031:TAANEF>2.0.CO;2. COCKS, L.R.M., MCKERROW, W.S. & VAN STAAL, C.R. 1997. The margins of Avalonia. Geological Magazine, 134, 627–636. COGHLAN, A. 2014. A Little Book of R For Time Series. Wellcome Trust Sanger Institute: Cambridge, UK. COLLINS, S.L. & GLENN, S.M. 1997. Intermediate Disturbance and its Relationship to Within- and Between-patch Dynamics. New Zealand Journal of Ecology, 21, 103–110, doi: 10.2307/24054530. CONNELL, J.H. 1978. Diversity in tropical rain forests and coral reefs. Science, 199, 1302–1310. CONNELL, J.H. 1997. Disturbance and recovery of coral assemblages. Coral Reefs, 16, S101–S113, doi: 10.1007/ s003380050246. COOMBS, D.S., ELLIS, A.J., FYFE, W.S. & TAYLOR, A.M. 1959. The zeolite facies, with comments on the interpretation of hydrothermal syntheses. Geochimica et Cosmochimica Acta, 17, 53–107, doi: 10.1016/0016- 7037(59)90079-1. COPE, J.C.W. 1977. An Ediacara-type fauna from South Wales. Nature, 268, 624–624, doi: 10.1038/268624a0. COX, R.T. 2012. Recent Advances in North American Paleoseismology and Neotectonics East of the Rockies. Geological Society of America. DADE, W.B., HOGG, A.J. & BOUDREAU, B.P. 2001. Physics of Flow above the Sediment-Water Interface. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 4–43. DAHMS, H.-U., DOBRETSOV, S. & QIAN, P.-Y. 2004. The effect of bacterial and diatom biofilms on the settlement of the bryozoan Bugula neritina. Journal of Experimental Marine Biology and Ecology, 313, 191–209, doi: 10.1016/j.jembe.2004.08.005. DALE, V.H., SWANSON, F.J. & CRISAFULLI, C.M. 2005. Disturbance, Survival, and Succession: Understanding Ecological Responses to the 1980 Eruption of Mount St. Helens. In: Dale, V. H., Swanson, F. J. & Crisafulli, C. M. (eds) Ecological Responses to the 1980 Eruption of Mount St. Helens. New York, NY, Springer New York. DARROCH, S.A.F., LAFLAMME, M., SCHIFFBAUER, J.D. & BRIGGS, D.E.G. 2012. Experimental formation of a microbial death mask. PALAIOS, 27, 293–303, doi: 10.2110/palo.2011.p11-059r.

203 DARROCH, S.A.F., LAFLAMME, M. & CLAPHAM, M.E. 2013. Population structure of the oldest known macroscopic communities from Mistaken Point, Newfoundland. Paleobiology, 39, 591–608, doi: 10.1666/12051. DASSOW, M. VON, ODELL, G.M. & MANDOLI, D.F. 2001. Relationships between growth, morphology and wall stress in the stalk of Acetabularia acetabulum. Planta, 213, 659–666, doi: 10.1007/s004250100538. DEAN, B. & BHUSHAN, B. 2010. Shark-skin surfaces for fluid-drag reduction in turbulent flow: a review. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 368, 4775–4806, doi: 10.1098/rsta.2010.0201. DEANGELIS, D.L. & MATTICE, J.S. 1979. Implications of a partial-differential-equation cohort model. Mathematical Biosciences, 47, 271–285, doi: 10.1016/0025-5564(79)90042-7. DE KROON, H., HUBER, H., STUEFER, J.F. & VAN GROENENDAEL, J.M. 2005. A modular concept of phenotypic plasticity in plants: Research review. New Phytologist, 166, 73–82, doi: 10.1111/j.1469-8137.2004.01310.x. DEWEL, R.A., DEWEL, W.C. & MCKINNEY, F.K. 2001. Diversification of the Metazoa: Ediacarans, colonies, and the origin of eumetazoan complexity by nested modularity. Historical Biology, 15, 193–218, doi: 10.1080/1 0292380109380592. DIGGLE, P.J. 1976. A Spatial Stochastic Model of Inter-Plant Competition. Journal of Applied Probability, 13, 662–671, doi: 10.2307/3212521. DILLON, W.R. & GOLDSTEIN, M. 1984. Multivariate Analysis: Methods and Applications. Wiley. DROSER, M.L., HUGHES, N.C. & JELL, P.A. 1994. Infaunal communities and tiering in Early Palaeozoic nearshore clastic environments: trace-fossil evidence from the Cambro-Ordovician of New South Wales. Lethaia, 27, 273–283, doi: 10.1111/j.1502-3931.1994.tb01574.x. DROSER, M.L., GEHLING, J.G. & JENSEN, S.R. 2006. Assemblage palaeoecology of the Ediacara biota: The unabridged edition? Palaeogeography, Palaeoclimatology, Palaeoecology, 232, 131–147, doi: 10.1016/ j.palaeo.2005.12.015. DUNN, G. & EVERITT, B.S. 2004. An Introduction to Mathematical Taxonomy. Courier Corporation. DZIK, J. 2002. Possible ctenophoran affinities of the precambrian ‘sea-pen’ Rangea. Journal of Morphology, 252, 315–334, doi: 10.1002/jmor.1108. DZIK, J. 2003. Anatomical Information Content in the Ediacaran Fossils and Their Possible Zoological Affinities. Integrative and Comparative Biology, 43, 114–126, doi: 10.2307/3884846. ECKMAN, J.E. 1983. Hydrodynamic processes affecting benthic recruitment1. Limnology and Oceanography, 28, 241–257, doi: 10.4319/lo.1983.28.1.0241. ECKMAN, J.E. 1990. A model of passive settlement by planktonic larvae onto bottoms of differing roughness. Limnology and Oceanography, 35, 887–901, doi: 10.4319/lo.1990.35.4.0887. ECKMAN, J.E., NOWELL, A.R. & JUMARS, P.A. 1981. Sediment destabilization by animal tubes. Journal of Marine Research, 39, 361–374. ELLIOTT, D.A., VICKERS-RICH, P., TRUSLER, P. & HALL, M. 2011. New Evidence on the Taphonomic Context of the Ediacaran Pteridinium. Acta Palaeontologica Polonica, 56, 641–650, doi: 10.4202/app.2010.0060. ENRÍQUEZ, S. & PANTOJA-REYES, N.I. 2005. Form-function analysis of the effect of canopy morphology on leaf self-shading in the seagrass Thalassia testudinum. Oecologia, 145, 234–242, doi: 10.1007/s00442-005- 0111-7. ERWIN, D.H., LAFLAMME, M., TWEEDT, S.M., SPERLING, E.A., PISANI, D. & PETERSON, K.J. 2011. The Cambrian Conundrum: Early Divergence and Later Ecological Success in the Early History of Animals. Science, 334, 1091–1097, doi: 10.1126/science.1206375. ETTER, R.J. & GRASSLE, J.F. 1992. Patterns of species diversity in the deep sea as a function of sediment particle size diversity. Nature, 360, 576–578, doi: 10.1038/360576a0. EVANS, S.D., DROSER, M.L. & GEHLING, J.G. 2015. Dickinsonia liftoff: Evidence of current derived morphologies. Ediacaran Environments and Ecosystems, 434, 28–33, doi: 10.1016/j.palaeo.2015.02.006.

204 FARRELL, Ú.C. 2014. Pyritization of soft-tissues in the fossil record: an overview. In: Laflamme, M., Schiffbauer, J. D. & Darroch, S. A. F. (eds) Reading and Writing of the Fossil Record: Preservational Pathways to Exceptional Fossilization. The Paleontological Society Papers, 313. FARRELL, Ú.C., BRIGGS, D.E., HAMMARLUND, E.U., SPERLING, E.A. & GAINES, R.R. 2013. Paleoredox and pyritization of soft-bodied fossils in the Ordovician Frankfort Shale of New York. American Journal of Science, 313, 452–489. FEDONKIN, M.A. 2003. The origin of the Metazoa in the light of the Proterozoic fossil record. Paleontological Research, 7, 9–41, doi: 10.2517/prpsj.7.9. FEDONKIN, M.A. & IVANTSOV, A.Y. 2007. Ventogyrus, a possible siphonophore-like trilobozoan coelenterate from the Vendian Sequence (late Neoproterozoic), northern Russia. Geological Society, London, Special Publications, 286, 187–194, doi: 10.1144/SP286.14. FEDONKIN, M.A. & WAGGONER, B.M. 1997. The Late Precambrian fossil Kimberella is a mollusc-like bilaterian organism. Nature, 388, 868–871. FISHER, R.V. & SCHMINCKE, H.-U. 1984. Pyroclastic Rocks. Berlin ;New York, Springer-Verlag,. FISHER, R.V. & SCHMINCKE, H.-U. 1984. Pyroclastic Rocks. New York, Springer-Verlag. FLUDE, L.I. & NARBONNE, G.M. 2008. Taphonomy and ontogeny of a multibranched Ediacaran fossil: Bradgatia from the Avalon Peninsula of Newfoundland. Canadian Journal of Earth Sciences, 45, 1095–1109, doi: 10.1139/E08-057. FODA, M.A., KHALAF, F.I. & AL-KADI, A.S. 1985. Estimation of dust fallout rates in the northern Arabian Gulf. Sedimentology, 32, 595–603, doi: 10.1111/j.1365-3091.1985.tb00473.x. FORD, T.D. 1958. Pre-Cambrian Fossils from Charnwood Forest. Proceedings of the Yorkshire Geological and Polytechnic Society, 31, 211–217, doi: 10.1144/pygs.31.3.211. FOSTER, M.S. 1975. Algal succession in a Macrocystis pyrifera forest. Marine Biology, 32, 313–329, doi: 10.1007/ BF00388989. FRALEY, C. & RAFTERY, A.E. 2002. Model-Based Clustering, Discriminant Analysis, and Density Estimation. Journal of the American Statistical Association, 97, 611–631, doi: 10.1198/016214502760047131. FRALEY, C. & RAFTERY, A.E. 2007. Bayesian regularization for normal mixture estimation and model-based clustering. Journal of Classification, 24, 155–181. FRALEY, C., RAFTERY, A.E., MURPHY, T.B. & SCRUCCA, L. 2012. Mclust Version 4 for R: Normal Mixture Modeling for Model-Based Clustering, Classification, and Density Estimation. Seattle, Washington, Department of Statistics, University of Washington, technical 597, Tecnhincal Reports. FRANCO, M. 1986. The Influences of Neighbours on the Growth of Modular Organisms with an Example from Trees. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 313, 209–225. FRANKLIN, J.F. 1990. Biological legacies: A critical management concept from Mount St. Helens. In: Trans. North American Wildlands Natural Resource Conference. 216–219. GABBOTT, S.E., ZALASIEWICZ, J., ALDRIDGE, R.J. & THERON, J.N. 2010. Eolian input into the Late Ordovician postglacial Soom Shale, South Africa. Geology, 38, 1103–1106, doi: 10.1130/G31426.1. GAGE, J.D. & TYLER, P.A. 1992. Deep-Sea Biology: A Natural History of Organisms at the Deep-Sea Floor. Cambridge, UK, Cambridge University Press. GALLAGHER, E.D., JUMARS, P.A. & TRUEBLOOD, D.D. 1983. Facilitation of Soft-Bottom Benthic Succession By Tube Builders. Ecology, 64, 1200–1216, doi: 10.2307/1937829. GATES, A.R. & JONES, D.O.B. 2012. Recovery of Benthic Megafauna from Anthropogenic Disturbance at a Hydrocarbon Drilling Well (380 m Depth in the Norwegian Sea). PLoS ONE, 7, e44114, doi: 10.1371/ journal.pone.0044114.

205 GAYLORD, B., REED, D.C., RAIMONDI, P.T., WASHBURN, L. & MCLEAN, S.R. 2002. A physically based model of macroalgal spore dispersal in the wave and current-dominated nearshore. Ecology, 83, 1239–1251. GEHLING, J.G. 1991. The case for Ediacaran fossil roots to the metazoan tree. Geological Society of India Memoir, 20, 181–224. GEHLING, J.G. 1999. Microbial Mats in Terminal Proterozoic Siliciclastics: Ediacaran Death Masks. PALAIOS, 14, 40, doi: 10.2307/3515360. GEHLING, J.G. & DROSER, M.L. 2013. How well do fossil assemblages of the Ediacara Biota tell time? Geology, 41, 447–450, doi: 10.1130/G33881.1. GEHLING, J.G. & NARBONNE, G.M. 2007. Spindle-shaped Ediacara fossils from the Mistaken Point assemblage, Avalon Zone, Newfoundland. Canadian Journal of Earth Sciences, 44, 367–387, doi: 10.1139/e07-003. GEHLING, J.G., NARBONNE, G.M. & ANDERSON, M.M. 2000a. The first named Ediacaran body fossil, Aspidella Terranovica. Palaeontology, 43, 427–456, doi: 10.1111/j.0031-0239.2000.00134.x. GEHLING, J.G., NARBONNE, G.M. & ANDERSON, M.M. 2000b. The first named Ediacaran body fossil, Aspidella Terranovica. Palaeontology, 43, 427–456, doi: 10.1111/j.0031-0239.2000.00134.x. GEHLING, J.G., DROSER, M.L., JENSEN, S.R. & RUNNEGAR, B.N. 2005. Ediacaran organisms: relating form and function. In: Briggs, D. E. G. (ed.) Evolving Form and Function: Fossils and Development: Proceedings of a Symposium Honoring Adolf Seilacher for His Contributions to Paleontology, in Celebration of His 80th Birthday. New Haven, Peabody Museum of Natural History, Yale University, 43–67. GERVAIS, F. 1997. Light dependent growth, dark survival and glucose uptake by cryptophytes isolated from a freshwater chemocline. Journal of Phycology, 33, 18–25, doi: 10.1111/j.0022-3646.1997.00018.x. GHISALBERTI, M. & NEPF, H. 2009. Shallow flows over a permeable medium: the hydrodynamics of submerged aquatic canopies. Transport in porous media, 78, 309–326. GHISALBERTI, M. & NEPF, H.M. 2002. Mixing layers and coherent structures in vegetated aquatic flows. Journal of Geophysical Research: Oceans, 107, 3–1, doi: 10.1029/2001JC000871. GHISALBERTI, M., GOLD, D.A., ET AL. 2014. Canopy Flow Analysis Reveals the Advantage of Size in the Oldest Communities of Multicellular Eukaryotes. Current Biology, 24, 305–309, doi: 10.1016/ j.cub.2013.12.017. GLAESSNER, M.F. 1979. Precambrian. In: Robinson, R. A. & Teichert, C. (eds) Treatise on Invertebrate Paleontology, Part A. Boulder, CO/Lawrence, KS, Geological Society of America, Kansas Press, A79–A118. GLAESSNER, M.F. 1985. The Dawn of Animal Life: A Biohistorical Study. Cambridge University Press. GLAESSNER, M.F. & WADE, M. 1966. The late Precambrian fossils from Ediacara, South Australia. Palaeontology, 9, 599–628. GLAESSNER, M.F. & WALTER, M.R. 1975. New Precambrian fossils from the Arumbera Sandstone, Northern Territory, Australia. Alcheringa: An Australasian Journal of Palaeontology, 1, 59–69, doi: 10.1080/0311 5517508619480. GLASS, J.B., AXLER, R.P., CHANDRA, S. & GOLDMAN, C.R. 2012. Molybdenum limitation of microbial nitrogen assimilation in aquatic ecosystems and pure cultures. Frontiers in Microbiology, 3, doi: 10.3389/ fmicb.2012.00331. GLIBERT, P.M. & LEGRAND, C. 2006. The Diverse Nutrient Strategies of Harmful Algae: Focus on Osmotrophy. In: Granéli, E. & Turner, J. (eds) Ecology of Harmful Algae. Springer Berlin Heidelberg, Ecological Studies, 163–175. GLYNN, P.W. 1973. Ecology of a Caribbean coral reef. The Porites reef-flat biotope: Part II. Plankton community with evidence for depletion. Marine Biology, 22, 1–21, doi: 10.1007/BF00388905. GOODAY, A.J. 2003. Benthic foraminifera (protista) as tools in deep-water palaeoceanography: Environmental influences on faunal characteristics. In: Advances in Marine Biology. Academic Press, 1–90.

206 GORELICK, R. 2015. Why Vegetative Propagation of Leaf Cuttings is Possible in Succulent and Semi-Succulent Plants. Haseltonia, 51–57, doi: 10.2985/026.020.0109. GRANT, S.W. 1990. Shell structure and distribution of Cloudina, a potential index fossil for the terminal Proterozoic. American Journal of Science, 290-A, 261–294. GRAZHDANKIN, D. 2004. Patterns of distribution in the Ediacaran biotas: facies versus biogeography and evolution. Paleobiology, 30, 203–221. GRAZHDANKIN, D. 2014. Patterns of Evolution of the Ediacaran Soft-Bodied Biota. Journal of Paleontology, 88, 269–283, doi: 10.1666/13-072. GRAZHDANKIN, D. & GERDES, G. 2007. Ediacaran microbial colonies. Lethaia, 40, 201–210, doi: 10.1111/j.1502- 3931.2007.00025.x. GRAZHDANKIN, D. & SEILACHER, A. 2005. A re-examination of the Nama-type Vendian organism Rangea schneiderhoehni. Geological Magazine, 142, 571–582, doi: 10.1017/S0016756805000920. GRAZHDANKIN, D.V., BALTHASAR, U., NAGOVITSIN, K.E. & KOCHNEV, B.B. 2008. Carbonate-hosted Avalon-type fossils in arctic Siberia. Geology, 36, 803–806, doi: 10.1130/G24946A.1. GREY, K., HILL, A.C. & CALVER, C. 2011. Chapter 8 Biostratigraphy and stratigraphic subdivision of Cryogenian successions of Australia in a global context. Geological Society, London, Memoirs, 36, 113–134, doi: 10.1144/M36.8. GRIME, J.P. 1973. Competitive Exclusion in Herbaceous Vegetation. Nature, 242, 344–347, doi: 10.1038/242344a0. GROTZINGER, J.P., WATTERS, W.A. & KNOLL, A.H. 2000. Calcified Metazoans in Thrombolite-Stromatolite Reefs of the Terminal Proterozoic Nama Group, Namibia. Paleobiology, 26, 334–359, doi: 10.2307/2666114. HALVERSON, G.P. & HURTGEN, M.T. 2007. Ediacaran growth of the marine sulfate reservoir. Earth and Planetary Science Letters, 263, 32–44, doi: 10.1016/j.epsl.2007.08.022. HALVERSON, G.P., HOFFMAN, P.F., SCHRAG, D.P., MALOOF, A.C. & RICE, A.H.N. 2005. Toward a Neoproterozoic composite carbon-isotope record. Geological Society of America Bulletin, 117, 1181–1207, doi: 10.1130/ B25630.1. HARPER, J.L. 1967. A Darwinian Approach to Plant Ecology. Journal of Ecology, 55, 247–270, doi: 10.2307/2257876. HARPER, J.L. 1977. Population Biology of Plants. Academic Press. HEBERT, C.L., KAUFMAN, A.J., PENNISTON-DORLAND, S.C. & MARTIN, A.J. 2010. Radiometric and stratigraphic constraints on terminal Ediacaran (post-Gaskiers) glaciation and metazoan evolution. Precambrian Research, 182, 402–412, doi: 10.1016/j.precamres.2010.07.008. HENRY, H.A.L. & AARSSEN, L.W. 1997. On the Relationship between Shade Tolerance and Shade Avoidance Strategies in Woodland Plants. Oikos, 80, 575–582, doi: 10.2307/3546632. HENRY, I.M., WILKINSON, M.D., HERNANDEZ, J.M., SCHWARZ-SOMMER, Z., GROTEWOLD, E. & MANDOLI, D.F. 2004. Comparison of ESTs from juvenile and adult phases of the giant unicellular green alga Acetabularia acetabulum. BMC Plant Biology, 4, 3, doi: 10.1186/1471-2229-4-3. HILL, P.S. & MCCAVE, I.N. 2001. Suspended Particle Transport in Benthic Boundary Layers. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 78–103. HOFFMAN, P.F., KAUFMAN, A.J., HALVERSON, G.P. & SCHRAG, D.P. 1998. A Neoproterozoic Snowball Earth. Science, 281, 1342–1346, doi: 10.1126/science.281.5381.1342. HOFFMAN, P.F. & LI, Z.-X. 2009. A palaeogeographic context for Neoproterozoic glaciation. Palaeogeography, Palaeoclimatology, Palaeoecology, 277, 158–172, doi: 10.1016/j.palaeo.2009.03.013. HOFMANN, H.J., O’BRIEN, S.J. & KING, A.F. 2008. Ediacaran Biota on Bonavista Peninsula, Newfoundland, Canada. Journal of Paleontology, 82, 1–36, doi: 10.1666/06-087.1.

207 HOWE, M.P.A., EVANS, M., CARNEY, J.N. & WILBY, P.R. 2012. New perspectives on the globally important Ediacaran fossil discoveries in Charnwood Forest, UK: prequel to. Proceedings of the Yorkshire Geological Society, 59, 137–144, doi: 10.1144/pygs2012-321. HOYAL CUTHILL, J.F. & CONWAY MORRIS, S. 2014. Fractal branching organizations of Ediacaran rangeomorph fronds reveal a lost Proterozoic body plan. Proceedings of the National Academy of Sciences, 111, 13122– 13126, doi: 10.1073/pnas.1408542111. HUA, H., PRATT, B.R. & ZHANG, L.-Y. 2003. Borings in Cloudina Shells: Complex Predator-Prey Dynamics in the Terminal Neoproterozoic. PALAIOS, 18, 454–459, doi: 10.1669/0883-1351(2003)018<0454: BICSCP>2.0.CO;2. HULTH, S., TENGBERG, A., LANDÉN, A. & HALL, P.O.J. 1997. Mineralization and burial of organic carbon in sediments of the southern Weddell Sea (Antarctica). Deep Sea Research Part I: Oceanographic Research Papers, 44, 955–981, doi: 10.1016/S0967-0637(96)00114-8. HUSSON, F., LÊ, S. & PAGÈS, J. 2010a. Exploratory Multivariate Analysis by Example Using R. CRC Press, Chapman & Hall/CRC Computer Science & Data Analysis. HUSSON, F., JOSSE, J. & PAGES, J. 2010b. Principal component methods - hierarchical clustering - partitional clustering: why would we need to choose for visualizing data? Technical Report - Agrocampus, September 2010. HUSTON, M. 1979. A General Hypothesis of Species Diversity. The American Naturalist, 113, 81–101. HUSTON, M.A. & DEANGELIS, D.L. 1987. Size Bimodality in Monospecific Populations: A Critical Review of Potential Mechanisms. The American Naturalist, 129, 678–707. ICHASO, A.A., DALRYMPLE, R.W. & NARBONNE, G.M. 2007. Paleoenvironmental and basin analysis of the late Neoproterozoic (Ediacaran) upper Conception and St. John’s groups, west Conception Bay, Newfoundland. Canadian Journal of Earth Sciences, 44, 25–41, doi: 10.1139/e06-098. JARDINE, N. & SIBSON, R. 1971. Mathematical Taxonomy. Wiley. JENKINS, R.J.F. & GEHLING, J.G. 1978. A review of the frond-like fossils of the Ediacara assemblage. Records of The South Australian Museum (Adelaide), 17, 347–359. JENNINGS, R.M., ETTER, R.J. & FICARRA, L. 2013. Population Differentiation and Species Formation in the Deep Sea: The Potential Role of Environmental Gradients and Depth. PLoS ONE, 8, e77594, doi: 10.1371/ journal.pone.0077594. JIN, L. & NEI, M. 1990. Limitations of the evolutionary parsimony method of phylogenetic analysis. Molecular Biology and Evolution, 7, 82–102. JOHNSON, A.S. 1990. Flow around phoronids: Consequnces of a neighbor to suspension feeders. Limnology and Oceanography, 35, 1395–1401. JOHNSTON, D.T., POULTON, S.W., ET AL. 2012. Late Ediacaran redox stability and metazoan evolution. Earth and Planetary Science Letters, 335–336, 25–35, doi: 10.1016/j.epsl.2012.05.010. JOHNSTON, D.T., POULTON, S.W., TOSCA, N.J., O’BRIEN, T., HALVERSON, G.P., SCHRAG, D.P. & MACDONALD, F.A. 2013. Searching for an oxygenation event in the fossiliferous Ediacaran of northwestern Canada. Special Issue dedicated to H.D. Holland: Evolution of the atmosphere and ocean through time, 362, 273–286, doi: 10.1016/j.chemgeo.2013.08.046. JORGENSEN, B.B. 2001. Life in the Diffusive Boundary Layer. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 348–373. JORGENSEN, B.B. & BOUDREAU, B.P. 2001. Diagenesis and Sediment-Water Exchange. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 211–244. JORGENSEN, B.B. & DES MARAIS, D.J. 1990. The diffusive boundary layer of sediments: oxygen microgradients over a microbial mat. Limnology and Oceanography, 35, 1343–1355.

208 JOSSE, J. & HUSSON, F. 2012. Handling missing values in exploratory multivariate data analysis methods. Journal de la Société Française de Statistique, 153, 79–99. JOSSE, J., CHAVENT, M., LIQUET, B. & HUSSON, F. 2012. Handling Missing Values with Regularized Iterative Multiple Correspondence Analysis. Journal of Classification, 29, 91–116, doi: 10.1007/s00357-012-9097-0. JUMARS, P.A., ECKMAN, J.E. & KOCH, E. 2001. Macroscopic Animals and Plants in Benthic Flows. In: Boudreau, B. P. & Jorgensen, B. B. (eds) The Benthic Boundary Layer : Transport Processes and Biogeochemistry: Transport Processes and Biogeochemistry. Oxford University Press, USA, 320–347. KEENE, O.N. 1995. The log transformation is special. Statistics in Medicine, 14, 811–819, doi: 10.1002/ sim.4780140810. KENCHINGTON, C.G. & WILBY, P.R. 2014. Of time and taphonomy: preservation in the Ediacaran. In: Laflamme, M., Schiffbauer, J. D. & Darroch, S. A. F. (eds) Reading and Writing of the Fossil Record: Preservational Pathways to Exceptional Fossilization. The Paleontological Society Papers, 20, 313. KIRCHMAN, D., GRAHAM, S., REISH, D. & MITCHELL, R. 1981. Bacteria induce settlement and metamorphosis of Janua (Dexiospira) brasiliensis Grube (Polychaeta:Spirprbidae). Journal of Experimental Marine Biology and Ecology, 56, 153–163, doi: 10.1016/0022-0981(81)90186-6. KOEHL, M.A.R. 1977. Mechanical diversity of connective tissue of the body wall of sea anemones. Journal of Experimental Biology, 69, 107–125. KOHLMEYER, J. & KOHLMEYER, E. 1979. Marine Mycology: The Higher Fungi. New York; London, Academic Press. KÜES, U. & LIU, Y. 2000. Fruiting body production in basidiomycetes. Applied Microbiology and Biotechnology, 54, 141–152, doi: http://dx.doi.org/10.1007/s002530000396. LAFLAMME, M. & NARBONNE, G.M. 2008. Ediacaran fronds. Palaeogeography, Palaeoclimatology, Palaeoecology, 258, 162–179, doi: 10.1016/j.palaeo.2007.05.020. LAFLAMME, M., NARBONNE, G.M. & ANDERSON, M.M. 2004. Morphometric Analysis of the Ediacaran Frond Charniodiscus from the Mistaken Point Formation, Newfoundland. Journal of Paleontology, 78, 827– 837. LAFLAMME, M., NARBONNE, G.M., GREENTREE, C. & ANDERSON, M.M. 2007. Morphology and taphonomy of an Ediacaran frond: Charnia from the Avalon Peninsula of Newfoundland. Geological Society, London, Special Publications, 286, 237–257, doi: 10.1144/SP286.17. LAFLAMME, M., XIAO, S. & KOWALEWSKI, M. 2009. Osmotrophy in modular Ediacara organisms. Proceedings of the National Academy of Sciences of the United States of America, 106, 14438–14443, doi: 10.1073/ pnas.0904836106. LAFLAMME, M., SCHIFFBAUER, J.D., NARBONNE, G.M. & BRIGGS, D.E.G. 2011. Microbial biofilms and the preservation of the Ediacara biota. Lethaia, 44, 203–213, doi: 10.1111/j.1502-3931.2010.00235.x. LAFLAMME, M., SCHIFFBAUER, J.D. & NARBONNE, G.M. 2012a. Deep-water microbially induced sedimentary structures (MISS) in deep time; the Ediacaran fossil Ivesheadia. Special Publication - Society for Sedimentary Geology, 101, 111–123. LAFLAMME, M., FLUDE, L.I. & NARBONNE, G.M. 2012b. Ecological Tiering and the Evolution of a Stem: The Oldest Stemmed Frond from the Ediacaran of Newfoundland, Canada. Journal of Paleontology, 86, 193–200, doi: 10.1666/11-044.1. LAFLAMME, M., DARROCH, S.A.F., TWEEDT, S.M., PETERSON, K.J. & ERWIN, D.H. 2013. The end of the Ediacara biota: Extinction, biotic replacement, or Cheshire Cat? Gondwana Research, 23, 558–573, doi: 10.1016/ j.gr.2012.11.004. LAJUS, D., SUKHIKH, N. & ALEKSEEV, V. 2015. Cryptic or pseudocryptic: can morphological methods inform copepod taxonomy? An analysis of publications and a case study of the Eurytemora affinis species complex. Ecology and Evolution, 5, 2374–2385, doi: 10.1002/ece3.1521.

209 LAMBERT, R.S.J. & MCKERROW, W.S. 1976. The Grampian Orogeny. Scottish Journal of Geology, 12, 271–292, doi: 10.1144/sjg12040271. LAMBSHEAD, P.J.D., TIETJEN, J., GLOVER, A., FERRERO, T., THISTLE, D. & GOODAY, A.J. 2001. Impact of large-scale natural physical disturbance on the diversity of deep-sea North Atlantic nematodes. Marine Ecology Progress Series, 214, 121–126, doi: 10.3354/meps214121. LAN, Z.-W. & CHEN, Z.-Q. 2012. Exceptionally preserved microbially induced sedimentary structures from the Ediacaran post-glacial successions in the Kimberley region, northwestern Australia. Precambrian Research, 200–203, 1–25, doi: 10.1016/j.precamres.2012.01.006. LAPWORTH, C. 1888. On the Discovery of the Olenellus Fauna in the Lower Cambrian Rocks of Britain. Geological Magazine, 5, 484–487, doi: 10.1017/S0016756800182731. LAZAR, O.R., BOHACS, K.M., MACQUAKER, J.H.S., SCHIEBER, J. & DEMKO, T.M. 2015. Capturing Key Attributes of Fine-Grained Sedimentary Rocks In Outcrops, Cores, and Thin Sections: Nomenclature and Description Guidelines. Journal of Sedimentary Research, 85, 230–246, doi: 10.2110/jsr.2015.11. LE BAS, M.J. 1984. Geological evidence from Leicestershire on the crust of southern Britain. Transactions of the Leicester Literary and Philosophical Society, 76, 54–67. LENTON, T.M., BOYLE, R.A., POULTON, S.W., SHIELDS-ZHOU, G.A. & BUTTERFIELD, N.J. 2014. Co-evolution of eukaryotes and ocean oxygenation in the Neoproterozoic era. Nature Geoscience, 7, 257–265, doi: 10.1038/ngeo2108. LI, Z.X., BOGDANOVA, S.V., ET AL. 2008. Assembly, configuration, and break-up history of Rodinia: A synthesis. Testing the Rodinia Hypothesis: Records in its Building Blocks, 160, 179–210, doi: 10.1016/ j.precamres.2007.04.021. LIMPERT, E., STAHEL, W.A. & ABBT, M. 2001. Log-normal Distributions across the Sciences: Keys and Clues On the charms of statistics, and how mechanical models resembling gambling machines offer a link to a handy way to characterize log-normal distributions, which can provide deeper insight into variability and probability—normal or log-normal: That is the question. BioScience, 51, 341–352, doi: 10.1641/0006- 3568(2001)051[0341:LNDATS]2.0.CO;2. LIRMAN, D. 2000. Fragmentation in the branching coral Acropora palmata (Lamarck): growth, survivorship, and reproduction of colonies and fragments. Journal of Experimental Marine Biology and Ecology, 251, 41–57, doi: 10.1016/S0022-0981(00)00205-7. LIU, A.G., MCLLROY, D. & BRASIER, M.D. 2010. First evidence for locomotion in the Ediacara biota from the 565 Ma Mistaken Point Formation, Newfoundland. Geology, 38, 123–126, doi: 10.1130/G30368.1. LIU, A.G., MCILROY, D., ANTCLIFFE, J.B. & BRASIER, M.D. 2011. Effaced preservation in the Ediacara biota and its implications for the early macrofossil record: EDIACARAN TAPHOMORPHS. Palaeontology, 54, 607–630, doi: 10.1111/j.1475-4983.2010.01024.x. LIU, A.G., MCILROY, D., MATTHEWS, J.J. & BRASIER, M.D. 2012. A new assemblage of juvenile Ediacaran fronds from the Drook Formation, Newfoundland. Journal of the Geological Society, 169, 395–403, doi: 10.1144/0016-76492011-094. LIU, A.G., MATTHEWS, J.J., MENON, L.R., MCILROY, D. & BRASIER, M.D. 2014. Haootia quadriformis n. gen., n. sp., interpreted as a muscular cnidarian impression from the Late Ediacaran period (approx. 560 Ma). Proceedings of the Royal Society of London B: Biological Sciences, 281, 20141202, doi: 10.1098/ rspb.2014.1202. LIU, A.G., KENCHINGTON, C.G. & MITCHELL, E.G. 2015a. Remarkable insights into the paleoecology of the Avalonian Ediacaran macrobiota. Gondwana Research, 27, 1355–1380, doi: 10.1016/j.gr.2014.11.002. LIU, A.G., MATTHEWS, J.J. & MCILROY, D. 2015b. The Beothukis/Culmofrons problem and its bearing on Ediacaran macrofossil taxonomy: evidence from an exceptional new fossil locality. Palaeontology, n/a – n/a, doi: 10.1111/pala.12206.

210 LI, Z.X., BOGDANOVA, S.V., ET AL. 2008. Assembly, configuration, and break-up history of Rodinia: A synthesis. Testing the Rodinia Hypothesis: Records in its Building Blocks, 160, 179–210, doi: 10.1016/ j.precamres.2007.04.021. LOVE, L.G. 1971. Early diagenetic polyframboidal pyrite, primary and redeposited, from the Wenlockian Denbigh Grit Group, Conway, North Wales, U.K. Journal of Sedimentary Research, 41, 1038–1044, doi: 10.1306/74D723EC-2B21-11D7-8648000102C1865D.

MACGABHANN, B.A. 2007. Discoidal fossils of the Ediacaran biota: a review of current understanding. Geological Society, London, Special Publications, 286, 297–313, doi: 10.1144/SP286.21. MACGABHANN, B.A. 2014. There is no such thing as the ‘Ediacara Biota’. Geoscience Frontiers, 5, 53–62, doi: 10.1016/j.gsf.2013.08.001. MACKEY, R.L. & CURRIE, D.J. 2001. The Diversity-Disturbance Relationship: Is It Generally Strong and Peaked? Ecology, 82, 3479–3492, doi: 10.2307/2680166. MACQUAKER, J.H.S., KELLER, M.A. & DAVIES, S.J. 2010. Algal Blooms and ‘Marine Snow’: Mechanisms That Enhance Preservation of Organic Carbon in Ancient Fine-Grained Sediments. Journal of Sedimentary Research, 80, 934–942, doi: 10.2110/jsr.2010.085. MADSEN, J.D., CHAMBERS, P.A., JAMES, W.F., KOCH, E.W. & WESTLAKE, D.F. 2001. The interaction between water movement, sediment dynamics and submersed macrophytes. Hydrobiologia, 444, 71–84. MAPSTONE, N.B. & MCILROY, D. 2006. Ediacaran fossil preservation: Taphonomy and diagenesis of a discoid biota from the Amadeus Basin, central Australia. Precambrian Research, 149, 126–148, doi: 10.1016/ j.precamres.2006.05.007. MARFENIN, N.N. 1997. Adaptation capabilities of marine modular organisms. Hydrobiologia, 355, 153–158, doi: 10.1023/A:1003069831832. MARTIN, M.W., GRAZHDANKIN, D.V., BOWRING, S.A., EVANS, D. A. D., FEDONKIN, M.A. & KIRSCHVINK, J.L. 2000. Age of Neoproterozoic Bilatarian Body and Trace Fossils, White Sea, Russia: Implications for Metazoan Evolution. Science, 288, 841–845, doi: 10.1126/science.288.5467.841. MASON, S.J., NARBONNE, G.M., DALRYMPLE, R.W. & O’BRIEN, S.J. 2013. Paleoenvironmental analysis of Ediacaran strata in the Catalina Dome, Bonavista Peninsula, Newfoundland. Canadian Journal of Earth Sciences, 50, 197–212, doi: 10.1139/cjes-2012-0099. MASON, S.J. & NARBONNE, G.M. 2016. Two new Ediacaran small fronds from Mistaken Point, Newfoundland. Journal of Paleontology, 90, 183–194, doi: 10.1017/jpa.2016.14. MCILROY, D. & LOGAN, G.A. 1999. The Impact of Bioturbation on Infaunal Ecology and Evolution during the Proterozoic-Cambrian Transition. PALAIOS, 14, 58–72, doi: 10.2307/3515361. MCILROY, D. & WALTER, M.R. 1997. A reconsideration of the biogenicity of Arumberia banksi Glaessner & Walter. Alcheringa: An Australasian Journal of Palaeontology, 21, 79–80, doi: 10.1080/03115519708619187. MCILROY, D., BRASIER, M.D. & MOSELEY, J.B. 1998. The Proterozoic–Cambrian transition within the ‘Charnian Supergroup’ of central England and the antiquity of the Ediacara fauna. Journal of the Geological Society, 155, 401–411, doi: 10.1144/gsjgs.155.2.0401. MCILROY, D., CRIMES, T.P. & PAULEY, J.C. 2005. Fossils and matgrounds from the Neoproterozoic Longmyndian Supergroup, Shropshire, UK. Geological Magazine, 142, 441–455. MCILROY, D., BRASIER, M.D. & LANG, A.S. 2009. Smothering of microbial mats by macrobiota: implications for the Ediacara biota. Journal of the Geological Society, 166, 1117–1121, doi: 10.1144/0016-76492009-073. MCMENAMIN, M.A.S. 1986. The Garden of Ediacara. PALAIOS, 1, 178–182, doi: 10.2307/3514512. MCMENAMIN, M.A.S. 1998. The Garden of Ediacara. New York, Columbia University Press. MEESTERS, E.H., HILTERMAN, M., KARDINAAL, E., KEETMAN, M., DE VRIES, M. & BAK, R.P.M. 2001. Colony size- frequency distributions of scleractinian coral populations : spatial and interspecific variation. Marine ecology. Progress series, 209, 43–54. 211 MENON, L.R., MCILROY, D. & BRASIER, M.D. 2013a. Evidence for Cnidaria-like behavior in ca. 560 Ma Ediacaran Aspidella. Geology, 41, 895–898, doi: 10.1130/G34424.1. MENON, L.R., MCILROY, D. & BRASIER, M.D. 2013b. Evidence for Cnidaria-like behavior in ca. 560 Ma Ediacaran Aspidella. Geology, 41, 895–898, doi: 10.1130/G34424.1. MENON, L.R., MCILROY, D., LIU, A.G. & BRASIER, M.D. 2015. The dynamic influence of microbial mats on sediments: fluid escape and pseudofossil formation in the Ediacaran Longmyndian Supergroup, UK. Journal of the Geological Society, doi: 10.1144/jgs2015-036. MERRIMAN, R.J. & FREY, M. 1998. Patterns of Very Low-Grade Metamorphism in Metapelitic Rocks. In: Frey, rtin & Robinson, D. (eds) Low-Grade Metamorphism. Blackwell Publishing Ltd., 61–107. MERZ, R.A. 1984. Self-Generated versus Environmentally Produced Feeding Currents: A Comparison for the Sabellid Polychaete Eudistylia vancouveri. Biological Bulletin, 167, 200–209, doi: 10.2307/1541348. MEYER, M., ELLIOTT, D., ET AL. 2014. Three-dimensional microCT analysis of the Ediacara fossil Pteridinium simplex sheds new light on its ecology and phylogenetic affinity. Precambrian Research, 249, 79–87, doi: 10.1016/j.precamres.2014.04.013. MIKULAS, R. & PEK, I. 1995. Tertiary borings in Devonian limestone rockgrounds from Hranice (northern Moravia, Czech Republic). Vestnik Ceskeho Geologickeho Ustavu, 70, 19–26. MISRA, S.B. 1969. Late Precambrian (?) fossils from southeastern Newfoundland. Geological Society of America Bulletin, 80, 2133–2140. MITCHELL, E.G., KENCHINGTON, C.G., LIU, A.G., MATTHEWS, J.J. & BUTTERFIELD, N.J. 2015. Reconstructing the reproductive mode of an Ediacaran macro-organism. Nature, 524, 343–346, doi: 10.1038/nature14646. MITCHELL, E.G. Unpublished thesis. Quantitative Approaches to Understanding Ediacaran Ecology. University of Cambridge, 2013. MOSELEY, J.B. & FORD, T.D. 1985. A stratigraphic revision of the Late Precambrian rocks of the Charnwood Forest, Leicestershire. Mercian Geologist, 10, 1–18. MOSELEY, J.B. & FORD, T.D. 1989. The sedimentology of the Charnian Supergroup. Mercian Geologist, 11, 251– 274. NARBONNE, G.M. 2004. Modular Construction of Early Ediacaran Complex Life Forms. Science, 305, 1141– 1144. NARBONNE, G.M. 2005. THE EDIACARA BIOTA: Neoproterozoic Origin of Animals and Their Ecosystems. Annual Review of Earth and Planetary Sciences, 33, 421–442, doi: 10.1146/annurev.earth.33.092203.122519. NARBONNE, G.M. & GEHLING, J.G. 2003. Life after snowball: The oldest complex Ediacaran fossils. Geology, 31, 27–30, doi: 10.1130/0091-7613(2003)031<0027:LASTOC>2.0.CO;2. NARBONNE, G.M., DALRYMPLE, R.W. & GEHLING, J.G. 2001. Neoproterozoic Fossils and Environments of the Avalon Peninsula, Newfoundland, Guidebook, Trip B5. Geological Association of Canada/Mineralogical Association of Canada Annual Meeting, St John’s, Newfoundland. NARBONNE, G.M., LAFLAMME, M., GREENTREE, C. & TRUSLER, P. 2009. Reconstructing a Lost World: Ediacaran Rangeomorphs from Spaniard’s Bay, Newfoundland. Journal of Paleontology, 83, 503–523, doi: 10.2307/29739123. NARBONNE, G.M., LAFLAMME, M., TRUSLER, P.W., DALRYMPLE, R.W. & GREENTREE, C. 2014. Deep-Water Ediacaran Fossils from Northwestern Canada: Taphonomy, Ecology, and Evolution. Journal of Paleontology, 88, 207–223, doi: 10.1666/13-053. NOBLE, S.R., CONDON, D.J., CARNEY, J.N., WILBY, P.R., PHARAOH, T.C. & FORD, T.D. 2014. U-Pb geochronology and global context of the Charnian Supergroup, UK: Constraints on the age of key Ediacaran fossil assemblages. Geological Society of America Bulletin, doi: 10.1130/B31013.1. NOFFKE, N. 2009. The criteria for the biogeneicity of microbially induced sedimentary structures (MISS) in Archean and younger, sandy deposits. Microbial Mats in Earth’s Fossil Record of Life: Geobiology, 96, 173–180, doi: 10.1016/j.earscirev.2008.08.002. 212 NOFFKE, N., GERDES, G., KLENKE, T. & KRUMBEIN, W.E. 2001. Microbially Induced Sedimentary Structures: A New Category within the Classification of Primary Sedimentary Structures. Journal of Sedimentary Research, 71, 649–656, doi: 10.1306/2DC4095D-0E47-11D7-8643000102C1865D. NOFFKE, N., KNOLL, A.H. & GROTZINGER, J.P. 2002. Sedimentary Controls on the Formation and Preservation of Microbial Mats in Siliciclastic Deposits: A Case Study from the Upper Neoproterozoic Nama Group, Namibia. PALAIOS, 17, 533–544, doi: 10.2307/3515692. NORIN, E. 1958. The Sediments of the Central Tyrrhenian Sea. In: Pettersson, H. (ed.) Sediment Cores from the Mediterranean Sea and the Red Sea. Goteborg: Elanders, Reports of the Swedish Deep-Sea Expedition, 1947-1948, 1–136. NYLANDER, J.A.A., RONQUIST, F., HUELSENBECK, J.P. & NIEVES-ALDREY, J. 2004. Bayesian Phylogenetic Analysis of Combined Data. Systematic Biology, 53, 47–67, doi: 10.1080/10635150490264699. O’BRIEN, L.J. & CARON, J.-B. 2012. A New Stalked Filter-Feeder from the Middle Cambrian Burgess Shale, British Columbia, Canada. PLoS ONE, 7, doi: 10.1371/journal.pone.0029233. O’BRIEN, S.J. & KING, A.F. 2005. Late Neoproterozoic (Ediacaran) stratigraphy of the Avalon Zone sedimentary rocks, Bonavista Peninsula, Newfoundland. In: Current Research, Newfoundland and Labrador Department of Natural Resources, Geological Report. 101–113. OCH, L.M. & SHIELDS-ZHOU, G.A. 2012. The Neoproterozoic oxygenation event: Environmental perturbations and biogeochemical cycling. Earth-Science Reviews, 110, 26–57, doi: 10.1016/j.earscirev.2011.09.004. OHFUJI, H. & RICKARD, D. 2005. Experimental syntheses of framboids—a review. Earth-Science Reviews, 71, 147–170, doi: 10.1016/j.earscirev.2005.02.001. OKAMURA, B. 1984. The effects of ambient flow velocity, colony size, and upstream colonies on the feeding success of bryozoa. I. Bugula stolonifera Ryland, an arborescent species. Journal of Experimental Marine Biology and Ecology, 83, 179–193, doi: 10.1016/0022-0981(84)90044-3. O‘NEILL, R. 1998. Recovery in complex ecosystems. Journal of Aquatic Ecosystem Stress and Recovery, 6, 181– 187, doi: 10.1023/A:1009996332614. PAGNI, M., NICULITA-HIRZEL, H., ET AL. 2013. Density-based hierarchical clustering of pyro-sequences on a large scale—the case of fungal ITS1. Bioinformatics, 29, 1268–1274, doi: 10.1093/bioinformatics/btt149. PAPADIMITRIOU, S., KENNEDY, H., BENTALEB, I. & THOMAS, D.N. 2002. Dissolved organic carbon in sediments from the eastern North Atlantic. Marine Chemistry, 79, 37–47. PAPEZIK, V.S. 1974. Prehnite-pumpellyite facies metamorphism of late Precambrian rocks of the Avalon Peninsula, Newfoundland. The Canadian Mineralogist, 12, 463–468. PATERSON, G.L.J., GLOVER, A.G., ET AL. 2011. Disturbance, productivity and diversity in deep-sea canyons: A worm’s eye view. Deep Sea Research Part II: Topical Studies in Oceanography, 58, 2448–2460, doi: 10.1016/j.dsr2.3011.04.008. PEET, R.K. & CHRISTENSEN, N.L. 1980. Succession: A Population Process. Vegetatio, 43, 131–140, doi: 10.2307/20145824. PENNY, A.M., WOOD, R., CURTIS, A., BOWYER, F., TOSTEVIN, R. & HOFFMAN, K.-H. 2014. Ediacaran metazoan reefs from the Nama Group, Namibia. Science, 344, 1504–1506, doi: 10.1126/science.1253393. PETERSON, K.J., WAGGONER, B. & HAGADORN, J.W. 2003. A fungal analog for Newfoundland Ediacaran fossils? Integrative and Comparative Biology, 43, 127–136. PFLUG, H.-D. 1972. Systematik der jung-präkambrischen Petalonamae Pflug 1970. Paläontologische Zeitschrift, 46, 56–67, doi: 10.1007/BF02989552. PHARAOH, T.C., WEBB, P.C., THORPE, R.S. & BECKINSALE, R.D. 1987. Geochemical Evidence for the Tectonic Setting of Late Proterozoic Volcanic Suites in Central England. Geological Society, London, Special Publications, 33, 541–552, doi: 10.1144/GSL.SP.1987.033.01.36. PICKETT, S.T.A. 1976. Succession: An Evolutionary Interpretation. The American Naturalist, 110, 107–119.

213 PLATT, W.J. & CONNELL, J.H. 2003. Natural Disturbances and Directional Replacement of Species. Ecological Monographs, 73, 507–522, doi: 10.1890/01-0552. PLINT, A.G., MACQUAKER, J.H.S. & VARBAN, B.L. 2012. Bedload Transport of Mud Across A Wide, Storm- Influenced Ramp: Cenomanian–Turonian Kaskapau Formation, Western Canada Foreland Basin. Journal of Sedimentary Research, 82, 801–822, doi: 10.2110/jsr.2012.64. PORTER, D. 1986. In: The Biology of Marine Fungi. Cambridge, UK, Cambridge University Press, 141–154. PORTER, S. 2011. The rise of predators. Geology, 39, 607–608, doi: 10.1130/focus062011.1. PORTILLA-ALONSO, R.M. & MARTORELL, C. 2011. Demographic consequences of chronic anthropogenic disturbance on three populations of the endangered globose cactus Coryphantha werdermannii. Journal of Arid Environments, 75, 509–515, doi: 10.1016/j.jaridenv.2011.01.015. POWER, G. 1978. Fish Population Structure in Arctic Lakes. Journal of the Fisheries Research Board of Canada, 35, 53–59, doi: 10.1139/f78-008. PYZIK, A.J. & SOMMER, S.E. 1981. Sedimentary iron monosulfides: Kinetics and mechanism of formation. Geochimica et Cosmochimica Acta, 45, 687–698, doi: 10.1016/0016-7037(81)90042-9. RAFF, R.A. & RAFF, E.C. 2014. The role of biology in the fossilization of embryos and other soft-bodied organisms: microbial biofilms and lagerstatten. Geological Society of America, Special Publication. RAISWELL, R. 1993. Kinetic controls on depth variations in localised pyrite formation. Chemical Geology, 107, 467–469, doi: 10.1016/0009-2541(93)90233-9. RAISWELL, R., WHALER, K., DEAN, S., COLEMAN, M.L. & BRIGGS, D.E.G. 1993. A simple three-dimensional model of diffusion-with-precipitation applied to localised pyrite formation in framboids, fossils and detrital iron minerals. Marine Geology, 113, 89–100, doi: 10.1016/0025-3227(93)90151-K. RAISWELL, R., NEWTON, R., BOTTRELL, S.H., COBURN, P.M., BRIGGS, D.E.G., BOND, D.P.G. & POULTON, S.W. 2008. Turbidite depositional influences on the diagenesis of Beecher’s Trilobite Bed and the Hunsrück Slate; sites of soft tissue pyritization. American Journal of Science, 308, 105–129, doi: 10.2475/02.3008.01. RAUPACH, M.R., FINNIGAN, J.J. & BRUNEI, Y. 1996. Coherent eddies and turbulence in vegetation canopies: The mixing-layer analogy. Boundary-Layer Meteorology, 78, 351–382, doi: 10.1007/BF00120941. R CORE TEAM. 2014. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. REBESCO, M., HERNÁNDEZ-MOLINA, F.J., VAN ROOIJ, D. & WÅHLIN, A. 2014. Contourites and associated sediments controlled by deep-water circulation processes: State-of-the-art and future considerations. 50th Anniversary Special Issue, 352, 111–154, doi: 10.1016/j.margeo.2014.03.011. RETALLACK, G.J. 1994. Were the Ediacaran Fossils Lichens? Paleobiology, 20, 523–544, doi: 10.2307/2401233. RETALLACK, G.J. 2014. Volcanosedimentary paleoenvironments of Ediacaran fossils in Newfoundland. Geological Society of America Bulletin, 126, 619–638, doi: 10.1130/B30892.1. RIBEIRO FILHO, E. & MUSSA, D. 1977. Histological and framboidal textures in diagenetic pyrite from the gondwana coal beds, Brazil. Boletim IG-USP, 8, 01–20. RONQUIST, F. & HUELSENBECK, J.P. 2003. MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics, 19, 1572–1574, doi: 10.1093/bioinformatics/btg180. ROTHMAN, D.H., HAYES, J.M. & SUMMONS, R.E. 2003. Dynamics of the Neoproterozoic carbon cycle. Proceedings of the National Academy of Sciences, 100, 8124–8129, doi: 10.1073/pnas.0832439100. ROYSTON, P. 1995. Remark AS R94: A Remark on Algorithm AS 181: The W-test for Normality. Applied Statistics, 44, 547, doi: 10.2307/2986146. RUNNEGAR, B.N. & FEDONKIN, M.A. 1992. Proterozoic metazoan body fossils. In: Schopf, J. W. & Klein, C. (eds) The Proterozoic Biosphere, A Multidisciplinary Study. New York, Cambridge University Press, 369– 387.

214 RYALAND, J.S. & WARNER, G.F. 1986. Growth and Form in Modular Animals: Ideas on the Size and Arrangement of Zooids. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 313, 53–76, doi: 10.1098/rstb.1986.0025. SALTER, J.W. 1856. On Fossil Remains in the Cambrian Rocks of the Longmynd and North Wales. Quarterly Journal of the Geological Society, 12, 246–251, doi: 10.1144/GSL.JGS.1856.012.01-02.31. SÁNCHEZ, J.A. & LASKER, H.R. 2004. Do multi–branched colonial organisms exceed normal growth after partial mortality? Proceedings of the Royal Society of London B: Biological Sciences, 271, S117–S120. SÁNCHEZ, J.A., ZENG, W., COLUCI, V.R., SIMPSON, C. & LASKER, H.R. 2003. How similar are branching networks in nature? A view from the ocean: Caribbean gorgonian corals. Journal of Theoretical Biology, 222, 135–138, doi: 10.1016/S0022-5193(03)00017-1. SANSOM, R.S. 2014. Experimental Decay of Soft Tissues. In: Laflamme, M., Schiffbauer, J. D., Darroh, S. A. F, editor(s). Reading and Writing of the Fossil Record: Preservational Pathways to Exceptional Fossilization: The Paleontological Society Papers, Volume 20. Paleontological Research Institution; 2014. p. 259-274. World Wide Web Address: https://www.escholar.manchester.ac.uk/uk-ac-man-scw:240973. SANSÓN, L.Z. & VAN HEIJST, G.-J. 2015. Laboratory Experiments on Flows Over Bottom Topography. In: von Larcher, T. & Williams, P. D. (eds) Modeling Atmospheric and Oceanic Flows: Insights from Laboratory Experiments and Numerical Simulations. American Geophysical Union; Wiley, Geophysical Monograph Series, 205, 139–158. SCHIEBER, J. & RICIPUTI, L. 2004. Pyrite ooids in Devonian black shales record intermittent sea-level drop and shallow-water conditions. Geology, 32, 305–308, doi: 10.1130/G20202.1. SCHIEBER, J. & RICIPUTI, L. 2005. Pyrite and Marcasite Coated Grains in the Ordovician Winnipeg Formation, Canada: An Intertwined Record of Surface Conditions, Stratigraphic Condensation, Geochemical ‘Reworking,’ and Microbial Activity. Journal of Sedimentary Research, 75, 907–920, doi: 10.2110/jsr.2005.070. SCHIEBER, J., SOUTHARD, J.B. & SCHIMMELMANN, A. 2010. Lenticular Shale Fabrics Resulting from Intermittent Erosion of Water-Rich Muds—Interpreting the Rock Record in the Light of Recent Flume Experiments. Journal of Sedimentary Research, 80, 119–128, doi: 10.2110/jsr.2010.005. SCOTESE, C.R. 2009. Late Proterozoic plate tectonics and palaeogeography: a tale of two supercontinents, Rodinia and Pannotia. Geological Society, London, Special Publications, 326, 67–83, doi: 10.1144/SP326.4. SCOTT, C., LYONS, T.W., BEKKER, A., SHEN, Y., POULTON, S.W., CHU, X. & ANBAR, A.D. 2008. Tracing the stepwise oxygenation of the Proterozoic ocean. Nature, 452, 456–459, doi: 10.1038/nature06811. SCOTT, R.J., MEFFRE, S., WOODHEAD, J., GILBERT, S.E., BERRY, R.F. & EMSBO, P. 2009. Development of Framboidal Pyrite During Diagenesis, Low-Grade Regional Metamorphism, and Hydrothermal Alteration. Economic Geology, 104, 1143–1168, doi: 10.2113/gsecongeo.104.8.1143. SEBENS, K.P. 1983. The larval and juvenile ecology of the temperate octocoral Alcyonium siderium Verrill. II. Fecundity, survival, and juvenile growth. Journal of Experimental Marine Biology and Ecology, 72, 263– 285, doi: 10.1016/0022-0981(83)90111-9. SEILACHER, A. 1984. Late Precambrian and Early Cambrian Metazoa: Preservational or Real Extinctions? In: Holland, H. D. & Trendall, A. F. (eds) Patterns of Change in Earth Evolution. Springer Berlin Heidelberg, Dahlem Workshop Reports Physical, Chemical, and Earth Sciences Research Reports, 5, 159–168. SEILACHER, A. 1992. Vendobionta and Psammocorallia: lost constructions of Precambrian evolution. Journal of the Geological Society, 149, 607–613, doi: 10.1144/gsjgs.149.4.0607. SEILACHER, A. 1999. Biomat-Related Lifestyles in the Precambrian. PALAIOS, 14, 86–93, doi: 10.2307/3515363. SEILACHER, A., GRAZHDANKIN, D. & LEGOUTA, A. 2003. Ediacaran biota: The dawn of animal life in the shadow of giant protists. Paleontological Research, 7, 43–54, doi: 10.2517/prpsj.7.43.

215 SEILACHER, A., BUATOIS, L.A. & GABRIELA MÁNGANO, M. 2005. Trace fossils in the Ediacaran–Cambrian transition: Behavioral diversification, ecological turnover and environmental shift. Palaeogeography, Palaeoclimatology, Palaeoecology, 227, 323–356, doi: 10.1016/j.palaeo.2005.06.003. SEREZHNIKOVA, E.A. 2010. Colonization of substrates: Vendian sedentary benthos. Paleontological Journal, 44, 1560–1569, doi: 10.1134/S0031030110120117. SHAISH, L. & RINKEVICH, B. 2009. Critical Evaluation of Branch Polarity and Apical Dominance as Dictators of Colony Astogeny in a Branching Coral. PLoS ONE, 4, e4095, doi: 10.1371/journal.pone.0004095. SHAISH, L., ABELSON, A. & RINKEVICH, B. 2007. How Plastic Can Phenotypic Plasticity Be? The Branching Coral Stylophora pistillata as a Model System. PLoS ONE, 2, doi: 10.1371/journal.pone.0000644. SHAPIRO, S.S. & WILK, M.B. 1965. An Analysis of Variance Test for Normality (Complete Samples). Biometrika, 52, 591, doi: 10.2307/2333709. SHILLING, F.M. & MANAHAN, D.T. 1990. Energetics of early development for the sea urchinsStrongylocentrotus purpuratus andLytechinus pictus and the crustaceanArtemia sp. Marine Biology, 106, 119–127, doi: 10.1007/BF02114682. SHU, D.-G., MORRIS, S.C., ET AL. 2006. Lower Cambrian Vendobionts from China and Early Diploblast Evolution. Science, 312, 731–734, doi: 10.2307/3846094. SIERRA, M.M.D., DONARD, O.F.X., ETCHEBER, H., SORIANO-SIERRA, E.J. & EWALD, M. 2001. Fluorescence and DOC contents of pore waters from coastal and deep-sea sediments in the Gulf of Biscay. Organic Geochemistry, 32, 1319–1328. SINGER, A., PLOTNICK, R. & LAFLAMME, M. 2012. Experimental fluid mechanics of an Ediacaran frond. Palaeontologia Electronica, 15, 19A. SMITH, C.R. & HESSLER, R.R. 1987. Colonization and succession in deep-sea ecosystems. Trends in Ecology & Evolution, 2, 359–363, doi: 10.1016/0169-5347(87)90136-4. SMIT, M.G.D., HOLTHAUS, K.I.E., ET AL. 2008. Species sensitivity distributions for suspended clays, sediment burial, and grain size change in the marine environment. Environmental Toxicology and Chemistry, 27, 1006–1012, doi: 10.1897/07-339.1. SPERLING, E.A., PETERSON, K.J. & LAFLAMME, M. 2011. Rangeomorphs, Thectardis (Porifera?) and dissolved organic carbon in the Ediacaran oceans. Geobiology, 9, 24–33, doi: 10.1111/j.1472-4669.2010.00259.x. STANLEY, D.J. 1983. Parallel Laminated Deep-Sea Muds and Coupled Gravity Flow-Hemipelagic Settling in the Mediterranean. Smithsonian Contributions to the Marine Sciences, 19, 1–19. STEINER, M. & REITNER, J. 2001. Evidence of organic structures in Ediacara-type fossils and associated microbial mats. Geology, 29, 1119–1122, doi: 10.1130/0091-7613(2001)029<1119:EOOSIE>2.0.CO;2. STEPHENS, G.C. 1988. Epidermal amino acid transport in marine invertebrates. Biochimica et Biophysica Acta (BBA) - Reviews on Biomembranes, 947, 113–138, doi: 10.1016/0304-4157(88)90022-6. STOHLGREN, T.J. 1992. Resilience of a heavily logged grove of giant sequoia (< i> Sequoiadendron giganteum) in Kings Canyon National Park, California. Forest ecology and management, 54, 115–140. STOW, D.A. & WETZEL, A. 1990. 3. Hemiturbidite: A new type of deep-water sediment. Proceedings of the Ocean Drilling Project, Leg, 116, 25–34. STOW, D.A.V. & PIPER, D.J.W. 1984. Deep-water fine-grained sediments: facies models. Geological Society, London, Special Publications, 15, 611–646, doi: 10.1144/GSL.SP.1984.015.01.38. TARHAN, L.G., DROSER, M.L. & GEHLING, J.G. 2010. Taphonomic Controls on Ediacaran Diversity: Uncovering the Holdfast Origin of Morphologically Variable Enigmatic Structures. PALAIOS, 25, 823–830, doi: 10.2110/ palo.2010.p10-074r. TARHAN, L.G., DROSER, M.L., GEHLING, J.G. & DZAUGIS, M.P. 2015. Taphonomy and morphology of the Ediacara form genus Aspidella. Precambrian Research, 257, 124–136, doi: 10.1016/j.precamres.2014.11.026.

216 THOMSON, F.J., MOLES, A.T., AULD, T.D. & KINGSFORD, R.T. 2011. Seed dispersal distance is more strongly correlated with plant height than with seed mass. Journal of Ecology, 99, 1299–1307, doi: 10.1111/j.1365- 2745.2011.01867.x. THRUSH, S.F. & DAYTON, P.K. 2002. Disturbance to Marine Benthic Habitats by Trawling and Dredging: Implications for Marine Biodiversity. Annual Review of Ecology and Systematics, 33, 449–473. TURNER, M.G., ROMME, W.H., GARDNER, R.H., O’NEILL, R.V. & KRATZ, T.K. 1993. A revised concept of landscape equilibrium: disturbance and stability on scaled landscapes. Landscape Ecology, 8, 213–227. TURNER, M.G., DALE, V.H. & EVERHAM, E.H. 1997. Fires, hurricanes, and volcanoes: Comparing large disturbances. Bioscience, 47, 758–768. VALLADARES, F. & PEARCY, R.W. 2000. The role of crown architecture for light harvesting and carbon gain in extreme light environments assessed with a structurally realistic 3-D model. In: Anales Del Jardín Botánico de Madrid. 3–16. VANDENBRINK, J.P., BROWN, E.A., HARMER, S.L. & BLACKMAN, B.K. 2014. Turning heads: The biology of solar tracking in sunflower. Plant Science, 224, 20–26, doi: 10.1016/j.plantsci.2014.04.006. VAN KRANENDONK, M.J., GEHLING, J.G. & SHIELDS, G.A. 2008. Precambrian. In: Ogg, J. G., Ogg, G. M. & Gradstein, F. M. (eds) The Concise Geological Timescale. Cambridge, UK, Cambridge University Press, 23–36. VARGA, Á., GYETVAI, G., NAGY, L. & NAGY, G. 2009. Electrochemical time of flight method for determination of diffusion coefficients of glucose in solutions and gels. Analytical and Bioanalytical Chemistry, 394, 1955–1963, doi: 10.1007/s00216-009-2859-9. VICKERS-RICH, P., IVANTSOV, A.Y., ET AL. 2013. Reconstructing Rangea: New Discoveries from the Ediacaran of Southern Namibia. Journal of Paleontology, 87, 1–15, doi: 10.1666/12-074R.1. VINCENT, W.F. & GOLDMAN, C.R. 1980. Evidence for algal heterotrophy in Lake Tahoe, California-Nevada. Limnol. Oceanogr., 25, 89–99. VISAGIE, C.M., ROETS, F. & JACOBS, K. 2009. A new species of Penicillium, P. ramulosum sp. nov., from the natural environment. Mycologia, 101, 888–895, doi: 10.3852/08-149. VOGEL, S. 1977. Current-induced flow through living sponges in nature. Proceedings of the National Academy of Sciences of the United States of America, 74, 2069–2071. WACEY, D., KILBURN, M.R., ET AL. 2015. Uncovering framboidal pyrite biogenicity using nano-scale CNorg mapping. Geology, 43, 27–30, doi: 10.1130/G36048.1. WAGGONER, B. 2003. The Ediacaran biotas in space and time. Integrative and Comparative Biology, 43, 104– 113. WAINWRIGHT, S.A. & DILLON, J.R. 1969. On the Orientation of Sea Fans (Genus Gorgonia). Biological Bulletin, 136, 130–139, doi: 10.2307/1539674. WANG, Y. & LI, J. 2008. Molecular Basis of Plant Architecture. Annual Review of Plant Biology, 59, 253–279, doi: 10.1146/annurev.arplant.59.032607.092902. WATKINS, R. 1991. Guild structure and tiering in a high-density Silurian community, Milwaukee County, Wisconsin. PALAIOS, 6, 465–478, doi: 10.2307/3514985. WATTS, W.W. 1903. Charnwood Forest: A Buried Triassic Landscape. The Geographical Journal, 21, 623–633, doi: 10.2307/1775653. WHITE, P.S. & PICKETT, S.T.A. 1985. Natural disturbance and patch dynamics: an introduction. In: Pickett, S. T. A. & White, P. S. (eds) The Ecology of Natural Disturbance and Patch Dynamics. Elsevier, 3–16. WICKHAM, H. 2009. ggplot2: Elegant Graphics for Data Analysis. New York, Springer. WILBY, P.R., CARNEY, J.N. & HOWE, M.P.A. 2011. A rich Ediacaran assemblage from eastern Avalonia: Evidence of early widespread diversity in the deep ocean. Geology, 39, 655–658, doi: 10.1130/G31890.1.

217 WILBY, P.R., KENCHINGTON, C.G. & WILBY, R.L. 2015. Role of low intensity environmental disturbance in structuring the earliest (Ediacaran) macrobenthic tiered communities. Palaeogeography, Palaeoclimatology, Palaeoecology, 434, 14–27, doi: 10.1016/j.palaeo.2015.03.033. WILKIN, R.T. & BARNES, H.L. 1997. Formation processes of framboidal pyrite. Geochimica et Cosmochimica Acta, 61, 323–339, doi: 10.1016/S0016-7037(96)00320-1. WILKIN, R.T., BARNES, H.L. & BRANTLEY, S.L. 1996. The size distribution of framboidal pyrite in modern sediments: An indicator of redox conditions. Geochimica et Cosmochimica Acta, 60, 3897–3912, doi: 10.1016/0016- 7037(96)00209-8. WILKIN, R.T., ARTHUR, M.A. & DEAN, W.E. 1997. History of water-column anoxia in the Black Sea indicated by pyrite framboid size distributions. Earth and Planetary Science Letters, 148, 517–525, doi: 10.1016/ S0012-821X(97)00053-8. WILLIAMS, H. & KING, A.F. 1979. Trepassey Map Area, Newfoundland. Geological Survey of Canada, Memoir, 389. WILSON, L.A. & BUTTERFIELD, N.J. 2014. Sediment Effects on the Preservation of Burgess Shale–Type Compression Fossils. PALAIOS, 29, 145–154, doi: 10.2110/palo.2013.075. WILSON, N.S.F., ZENTILLI, M. & SPIRO, B. 2003. A Sulfur, Carbon, Oxygen, and Strontium Isotope Study of the Volcanic-Hosted El Soldado Manto-Type Copper Deposit, Chile: The Essential Role of Bacteria and Petroleum. Economic Geology, 98, 163–174, doi: 10.2113/gsecongeo.98.1.163. WOOD, D.A., DALRYMPLE, R.W., NARBONNE, G.M., GEHLING, J.G. & CLAPHAM, M.E. 2003. Paleoenvironmental analysis of the late Neoproterozoic Mistaken Point and Trepassey formations, southeastern Newfoundland. Canadian Journal of Earth Sciences, 40, 1375–1391, doi: 10.1139/e03-048. WOOTTON, J.T. 1998. Effects of Disturbance on Species Diversity: A Multitrophic Perspective. The American Naturalist, 152, 803–825, doi: 10.1086/286210. XIAO, S. & LAFLAMME, M. 2009. On the eve of animal radiation: phylogeny, ecology and evolution of the Ediacara biota. Trends in Ecology & Evolution, 24, 31–40, doi: 10.1016/j.tree.2008.07.015. XIAO, S., SHEN, B., ZHOU, C., XIE, G. & YUAN, X. 2005. A uniquely preserved Ediacaran fossil with direct evidence for a quilted bodyplan. Proceedings of the National Academy of Sciences of the United States of America, 102, 10227–10232, doi: 10.1073/pnas.0502176102. XIAO, S., ZHOU, C., LIU, P., WANG, D. & YUAN, X. 2014. Phosphatized Acanthomorphic Acritarchs and Related Microfossils from the Ediacaran Doushantuo Formation at Weng’an (South China) and their Implications for Biostratigraphic Correlation. Journal of Paleontology, 88, 1–67, doi: 10.1666/12-157R. XIAO, X., WHITE, E.P., HOOTEN, M.B. & DURHAM, S.L. 2011. On the use of log-transformation vs. nonlinear regression for analyzing biological power laws. Ecology, 92, 1887–1894, doi: 10.1890/11-0538.1. YANG, Z. 1996. Phylogenetic analysis using parsimony and likelihood methods. Journal of Molecular Evolution, 42, 294–307, doi: 10.1007/BF02198856. YOCCOZ, N.G. 1991. Use, Overuse, and Misuse of Significance Tests in Evolutionary Biology and Ecology. Bulletin of the Ecological Society of America, 72, 106–111. YUAN, X., CHEN, Z., XIAO, S., ZHOU, C. & HUA, H. 2011. An early Ediacaran assemblage of macroscopic and morphologically differentiated eukaryotes. Nature, 470, 390–393, doi: 10.1038/nature09810. ZAKREVSKAYA, M. 2014. Paleoecological reconstruction of the Ediacaran benthic macroscopic communities of the White Sea (Russia). Palaeogeography, Palaeoclimatology, Palaeoecology, 410, 27–38, doi: 10.1016/ j.palaeo.2014.05.021. ZHU, M., GEHLING, J.G., XIAO, S., ZHAO, Y. & DROSER, M.L. 2008. Eight-armed Ediacara fossil preserved in contrasting taphonomic windows from China and Australia. Geology, 36, 867–870, doi: 10.1130/ G25203A.1.

218 APPENDIX 1: MULTIVARIATE STATISTICAL TECHNIQUES

1.1 Data treatment

The challenging nature of the dataset required several data treatments to be applied to the data before the main statistical tests could be run. The methods behind these treatments are discussed below, together with the ways in which their influence on the data was ascertained.

1.1.1 Iterations of the data

Tests were run on all specimens which fit into the broad “multifoliate, stalked” morphospace for which both branching and morphological characters could be recorded (Undosyrus was excluded from the analysis because branching characters were difficult to obtain for the majority of specimens, see Chapter 6). The tests were run primarily on a filtered dataset (containing 40 Primocandelabrum, 2 dusters, 7 Orthiokaterna; Appendix), which excluded those specimens for whom either no morphologic characters or no branching characters could be determined. This removed individuals whose data was strongly biased by the need to replace missing data and the potential bias that these individuals could impart to the analyses. Third order branching characters were excluded on the basis that they could not be obtained for the majority of specimens, and even then in only rare cases could all characters be determined. They have also been excluded from previous taxonomic frameworks on this basis (Brasier et al. 2012). Each dataset was analysed with respect to all characters, to branching characters only and to morphologic characters only. To determine the influence of removing one branching characteristic, one set of iterations excluded concealed characters, as these showed little variance and are not obviously accounted for in the branching scheme of (Narbonne et al. 2009). One round of iterations on both the total set excluded out-size specimens (“Big Bertha” and “Big Sue”) because clustering mechanisms are very sensitive to outliers (Dillon & Goldstein 1984), thereby removing the possibility that the results are biased by ontogenetic effects, or ones arising from their complex life history of these individuals (Wilby et al. 2015).

All the measurements used (Fig. 4.2a) describe unique aspects of the shape of the individual specimens. However, some measurements correlate with each other to a degree (Fig. 4.2a), such as stem length to inflection point and stem length to base of branches; individuals with overall longer stems will have both of these measurements longer than those with overall small stems. Similarly, individuals with overall larger crowns will tend to have larger left and right branches than those with small crowns. Accordingly, one set of iterations was run with non-correlating morphological characters.

As the variables used to construct the principal component space reflect the variance in the specimens, including Orthiokaterna must bias the space construction and consequently the groupings within Primocandelabrum sp. Consequently, analyses were additionally run on a dataset excluding Orthiokaterna. 219 These were run only the filtered dataset, with runs using all characters, branching characters only and morphological characters only, both including and excluding the outsize specimen, “Big Bertha”. As the first order branching characters rotated/displayed and furled/unfurled had only one category recorded from the non-Orthiokaterna specimens, they were necessarily excluded from these analyses. A further round removed “dusters” from these analyses in addition to Big Bertha, to remove from the analysis the influence of these two individuals, which have a different crown construction which is not accounted for in the character matrix.

1.1.2 Scaling data

Scaling data is an important step when using Euclidean distance matrices in these types of analyses (the standard, although not only available option), as this distance matrix is not scale-invariant, and preserves relative distances when constructing a distance matrix (Dillon & Goldstein 1984). Scaling is performed during the principal components analyses by the techniques used (Husson et al. 2010a). To determine the influence of different scaling algorithms, one iteration was run in FAMD on data standardised by the scale() function and non-standardised data (Coghlan 2014, Dillon & Goldstein 1984). Due to the necessity to convert the imputed data into a dataframe before scaling, the iteration run on standardised data treats each category as a separate variable, with each individual assigned a score indicating the degree to which it belongs to each category variable. However, the next steps no longer read the dataframe as imputed data, and so construct the principal components space using all data (including imputed values). It also treats all variables as continuous (numeric) because of this.

1.1.3 Dealing with missing data

The majority of analyses cannot be run on tests with missing data. There are two approaches to dealing with missing data: replace the missing value, or delete the individual which has missing values. The latter approach is unfeasible in this case, as only five specimens have no missing values. Missing values can be replaced by the average value for the given criterion, but a more accurate way is to first impute the values. There are a number of different approaches to this problem, but I used the principal component method implemented in the MissMDA package in R outlined in (Josse & Husson 2012a). This method ensured consistency with the subsequent tests used, and has the added benefit of the option to impute both categorical and numeric characters together as well as separately, as it uses models based on a principal components analysis (PCA), multiple correspondence analysis (MCA) or factorial analysis of mixed data (FAMD). When imputing missing values using FAMD, the categorical values are first coded into dummy variables, and replacing missing values with the mean value of each variable. FAMD is then run on this table, and missing values are replaced with predicted ones based on the FAMD, the specified number of components to retain, and the regularised reconstruction formulae based on this number of

220 components: these steps are repeated until convergence of predicted values (Audigier et al. 2013). The predicted values represent a “degree of belonging” to each category. The protocol was as follows:

1) Estimate the number of components to retain in the imputation step. This is essential because the number of principal components retained in the next step alter all subsequent results. The data are imputed multiple times by cross-validation, each time using between one and five principal components (or otherwise specified minimum and maximum components), and returns the value which produces the smallest mean square error prediction. This value is then input into the next step. The estim_ncpMCA and estim_ncpPCA functions were used (Josse & Husson 2012b).

2) Impute values to construct a complete dataset. The regularised iterative algorithm was used in preference to the older “EM” iterative algorithm to avoid overfitting, which are particularly problematic when, as in this dataset, there are large amounts of missing values (Josse et al. 2012). The “leave-one- out” method was used, in which each known value is systematically removed and predicted using the model generated, then cross-validated to the real value (Bro et al. 2008). The model is revised with each subsequent prediction to refine its predictive power, and is then used to predict the missing values (Josse & Husson 2012a). These steps were run on each of the data iterations described above, to remove any influence of outliers and different groupings on the model used to predict values. The imputed dataset was subsequently read into all subsequent analyses.

1.2 Multivariate analysis and reduction of dimensionality

Component analyses were adopted in preference to other factor analysis methods such as maximum likelihood and parsimony, as they do not rely on the assumption of common or unique factors in the data (Dillon & Goldstein 1984). A variety of packages exist for analysing multivariate data, but very few packages permit analysis of both continuous (i.e. quantitative or numeric, such as size data) and categorical (i.e. qualitative or binary, such as present/absent characters) factors in tandem without significant and often misleading treatment of the data (Hill & Smith 1976, Husson et al. 2010a). For this reason, the FactoMineR package (Lê et al. 2008) was used in this analysis, as this package contains functions which can analyse numerical only (PCA), categorical only (MCA) and numerical and categorical factors combined (FAMD). It also permits the inclusion of supplementary variables and/or individuals into analysis: these do not contribute to the construction of the dimensions, but their covariance or correlation with the subject variables or individuals can be ascertained in this manner.

1.2.1 Principal Components Analysis (PCA)

Principal components analysis (PCA) is a commonly used approach to reduce dimensionality in a multivariate dataset, that is, to convert a multidimensional space into a two (or three) dimensional representation by constructing axes (the principal components) which are a linear combination of the

221 variables, and which account for most of the variance in the data (Dillon & Goldstein 1984, Jolliffe 2002, Wold et al. 1987). PCA is an example of an interdependence method of multivariate data exploration: it assumes equal association or disassociation across all variables, and does not require distinction of variable types. This is distinct from dependence methods such as multiple regression analysis, canonical correlation analysis, and (multiple) analysis of variance, which investigates the dependence of one (or a set of variables) on a predictor variable (Dillon & Goldstein 1984). In the FactoMineR package, there are options for exploring the degree to which each character has contributed to the construction of each dimension (by examining the “dimdesc” output), to identify those which exert a strong control on the co-ordinates of the individuals in the PCA space.

1.2.2 Multiple Correspondence Analysis (MCA)

Multiple correspondence analysis (MCA) is similar to PCA, but it is run on categorical variables, and is frequently used in the psychology literature (Greenacre & Blasius 2006, Tenenhaus & Young 1985). It is essentially the extension of the more commonly used correspondence analysis on to a data matrix of more than two categorical variables, with The individuals and variables are scaled in a multidimensional fashion, as in PCA (Husson & Josse 2014). In the FactoMineR MCA function (Husson et al. 2010a), missing values may be treated as a separate variable, or replaced by imputed values.

1.2.3 Factor Analysis of Mixed Data (FAMD)

FAMD is effectively a combination of PCA and MCA. In all FAMD analyses, categorical variables are converted to dummy variables by scaling them to unit variance and converting them to a disjunctive data table, which allows continuous and categorical characters to influence the analysis equally, enabling both sets of characters to be used in conjunction, and permits appropriate treatment of multistate (i.e. non- binary) categorical characters (Hill & Smith 1976, Pagès 2004). By adopting this approach, any signal reduction imposed by artificial scaling, or errors extrapolated through treating categorical variables as continuous ones, is removed. The FactoMineR function also permits extensive analysis of the dimensions, as well as the correlation of variables with the dimensions and, importantly, with each other (Dillon & Goldstein 1984, Husson et al. 2010a).

1.3 Clustering

Clustering analysis seeks to group objects in a data matrix together based on their shared similarity. Clusters themselves are defined by inter-object similarities, which should have a smaller variance within clusters than between them (Dillon & Goldstein 1984). A number of techniques can be used to identify clusters, the first step of which is to transform the data matrix into a similarity or distance matrix, by calculating the distances between pairs of individuals for each of the variables. Several methods can then be used to cluster the individuals together based on this matrix, but three of the most common are

222 1) hierarchical/agglomerative, 2) partitioning/divisive and 3) model-based. K-means clustering is the most popular type of partitioning method, but requires the number of clusters to be known beforehand: this can be identified using a scree plot (Everit & Hothorn 2002). This method divides the group into the specified number of clusters by moving individuals between clusters until there is no greater reduction in the error component by additional movements (Dillon & Goldstein 1984). Its main advantage over hierarchical clustering is that individuals can be reassigned to a cluster, that is, it is not irrevocable. However, hierarchical clustering is considered most appropriate if the groupings correspond to natural clustering, such as biological species (e.g. Dillon & Goldstein 1984), and so was applied here. Model- based clustering identifies the optimum number of mixing components within a dataset by comparing the data to generated models, and is used to examine the number of clusters contained within the data in an unbiased way, employing the widely-used Mclust function.

1.3.1 Hierarchical clustering

Hierarchical clustering techniques initially treat each individual as an individual group, and seek to combine individuals into larger clusters. It is considered to be the most appropriate method for natural groupings, such as taxonomic ones (Dillon & Goldstein 1984). The most widely applied technique is Ward’s sum of error squares method (Dillon & Goldstein 1984, Ward 1963). This method groups individuals based on the minimum loss of information arising from grouping each pair, based on the increase in the error sum of squares induced by each possible pair combination (an objective function). The information loss is measured by the total sum of squared deviations of each observation from the mean of the cluster. The goal of this clustering method is to optimise this objective function, by identifying the group amalgamation which results in the lowest impairment of the function: the first cluster is formed from the pair of individuals that give the lowest increase in the objective function, and subsequent individuals are assigned either to the existing cluster or to a new cluster, whichever gives the smallest increase in the objective function. The next new cluster is again defined by identifying the pair which, when clustered, gives the lowest increase in the objective function (Dillon & Goldstein 1984). This method may also be referred to in terms of the inertia (variance) of the clusters (Fig. 4-6 a,c,e), where each aggregation of clusters seeks to minimise the growth of within-group inertia (the homogeneity of a cluster) and, correspondingly, to minimise the reduction in between-group inertia (Husson et al. 2010b). Changes in the value of this function indicate the number of natural groupings, as defined by the elbow in a scree plot (Everit & Hothorn 2002), or by the greatest gain in within-group inertia (Husson et al. 2010b).

The HCPC function in FactoMineR (Husson et al. 2010b) was used to undertake clustering analysis, as this function permits incorporation of both continuous and categorical variables into the analysis whilst bestowing equal weight to each variable. It also takes into account imputed values, and does not

223 incorporate these values into construction of the principal components space. This method uses Ward’s criterion to aggregate individuals into clusters, producing a cluster dendrogram (Fig. 4-6 b,d,f). The function recommends an optimal number of clusters (Q) based on the inertia gain at each aggregation, based on the criterion Δ(Q)/ Δ(Q + 1), where Δ(Q) is the increase of between-group inertia moving from Q – 1 to Q clusters. By default, the minimum number of clusters is set at 3, as the greatest increase in inertia gain is often between 2 and 3 groups (Husson et al. 2010b). Visual inspection of the inertia gain plot can confirm the most appropriate number of clusters. This function also permits K-means analysis to be conducted on the data before hierarchical clustering, where the number of clusters can be pre-set (Husson et al. 2010b). In these analyses, the K-means integer was left as infinite to avoid imparting bias to the clustering via pre-processing, and to permit consolidation of the clusters.

The assignment of individuals to specific groups was compared by determining the percentage of individuals for whom group assignment matched for each pair of analyses and the percentage of analyses resulting in assignment of an individual to a specific group was calculated to aid in broader comparisons. The assignment of a number label to a cluster is based on the position of the constituent individuals along the first principal component, and so is dependent on the variables used to construct the principal component space (Husson et al. 2010b). The differences in composition of the clusters between iterations is more interesting than their shared label, and so, for comparisons resulting in a low percentage match (less than 33% for 3 clusters, less than 50% for 2 clusters), cluster labels were switched and the comparison re-run. In order to compare group assignment from tests including Orthiokaterna from those excluding these specimens, the non-Orthiokaterna group is forced into two clusters (p.?).

1.3.2 Model-based clustering and determination of the number of mixing components

The Mclust function of the mclust package in R was used (Fraley et al. 2012) as it is widely used and offers numerous options for exploring the density distribution of individuals and the clustering of the data. This algorithm determines the optimal number and shape (distribution, ellipticity, volume and orientation) of groupings within data (Fraley et al. 2012) by comparing it to probability models via iterative Expectation-Maximization (EM) methods, an approach to maximium likelihood estimation (described in (Dempster et al. 1977, Fraley & Raftery 2002, McLachlan & Krishnan 2008). The output examined here was the plot of the Bayesian Information Criterion (BIC) for different models each with a different number of components, in which the higher the BIC value, the greater the likelihood that the data fit the model of a given shape and number of components. A difference of more than >10 indicates strong supported for one model over another, while a difference of <6 indicates only weak support for one model over another (Fraley & Raftery 2007).

224 1.4 References unique to this section AUDIGIER, V., HUSSON, F. & JOSSE, J. 2013. A principal components method to impute missing values for mixed data. arXiv:1301.4797. BRO, R., KJELDAHL, K., SMILDE, A.K. & KIERS, H. A. L. 2008. Cross-validation of component models: a critical look at current methods. Analytical and Bioanalytical Chemistry, 390, 1241–1251, doi: 10.1007/ s00216-007-1790-1. DEMPSTER, A.P., LAIRD, N.M. & RUBIN, D.B. 1977. Maximum likelihood from incomplete data via the EM algorithm. Journal of the royal statistical society. Series B (methodological), 1–38. EVERIT, B.S. & HOTHORN, T. 2002. A Handbook of Statistical Analyses Using R. Boca Raton, FL, CRC Press. GREENACRE, M. & BLASIUS, J. 2006. Multiple Correspondence Analysis and Related Methods. CRC Press. HILL, M.O. & SMITH, A.J.E. 1976. Principal Component Analysis of Taxonomic Data with Multi-State Discrete Characters. Taxon, 25, 249–255, doi: 10.2307/1219449. HUSSON, F. & JOSSE, J. 2014. multiple correspondence analysis. In: Visualization and Verbalization of Data. Boca Raton, FL, CRC Press, 165–184. JOLLIFFE, I. 2002. Principal Component Analysis, 2nd ed. John Wiley & Sons, Ltd. JOSSE, J. & HUSSON, F. 2012a. Handling missing values in exploratory multivariate data analysis methods. Journal de la Société Française de Statistique, 153, 79–99. JOSSE, J. & HUSSON, F. 2012b. Selecting the number of components in principal component analysis using cross-validation approximations. Computational Statistics & Data Analysis, 56, 1869–1879, doi: 10.1016/ j.csda.2011.11.012. LÊ, S., JOSSE, J. & HUSSON, F. 2008. FactoMineR: an R package for multivariate analysis. Journal of statistical software, 25, 1–18. MCLACHLAN, G.J. & KRISHNAN, T. 2008. Basic Theory of the EM Algorithm. In: The EM Algorithm and Extensions. John Wiley & Sons, Inc., 77–103. PAGÈS, J. 2004. Analyse factorielle de données mixtes. Revue de Statistique Appliquée, 52, 93–111. TENENHAUS, M. & YOUNG, F.W. 1985. An analysis and synthesis of multiple correspondence analysis, optimal scaling, dual scaling, homogeneity analysis and other methods for quantifying categorical multivariate data. Psychometrika, 50, 91–119, doi: 10.1007/BF02294151. WARD, J.H.W. 1963. Hierarchical Grouping to Optimize an Objective Function. Journal of the American Statistical Association, 58, 236–244, doi: 10.1080/01621459.1963.10500845. WOLD, S., ESBENSEN, K. & GELADI, P. 1987. Principal component analysis. Chemometrics and Intelligent Laboratory Systems, 2, 37–52, doi: 10.1016/0169-7439(87)80084-9.

225 226 APPENDIX 2: DATA TABLES

A.2.1 Specimens from Charnwood Forest Species Field Specimen GSM Height Species Field Specimen GSM Height no. (cm) no. (cm) Charnia 7C2 105983 2 Charniodiscus 5F1a (upper) 106011 12.2 Charnia 7C3 105973 28.6 Charniodiscus 5F1b (lower) 106011 17.55 Charnia 6A6 105982 15.2 Charniodiscus 10B9 106068 11 Charnia 3B4 105994 1.2 Charniodiscus 4B6 106114 10.75 Charnia 3E5 105972 13.7 Charniodiscus 9D 105881 11.1 Charnia 6A3 105980 12.5 Charniodiscus 4A2a (left) 106025 14.3 Charnia 5C9 105988 9.2 Charniodiscus 4A2c (right) 106025 11.2 Charnia 7C1 105985 12 Charniodiscus 6A4 106021 12.6 Charnia 3(3) 105975 14.3 Charniodiscus 13A5 106027 10.2 Charnia 7B4 105977 5.5 Charniodiscus 20Bb&c 106039 9.25 Charnia 13B2 105984 7.5 Charniodiscus 7A2 106013 13.9 Charnia 8B2 106079 8.8 Charniodiscus 7A7 106121 8.2 Charnia 6A5 105979 10.5 Charniodiscus 4A4 106023 10.5 Charnia 8B1 106078 11.6 Charniodiscus 7A3 106018 13.6 Charnia 4A5 105995 4.5 Charniodiscus 13A2 105960 30.6 Charnia 3B3 105987 11.4 Charniodiscus 3C2 106017 13 Charnia 10A3 106081 6.5 Charniodiscus 5A1 106022 8.1 Charnia 3B6 105991 2.8 Charniodiscus 2B3 106019 13 Charnia 6B1 105990 4.5 Charniodiscus 5G1 106002 18.2 Charnia 10C4 106086 9.1 Charniodiscus 7 105876 11.2 Charnia 1A1 105997 17.2 Charniodiscus 10C9 106069 14.95 Charnia 10A5 106082 6.9 Charniodiscus 1B2b 106015 8.6 Charnia 8D1 106080 9.4 Charniodiscus 5F4 105976 10.7 Charnia 3D3 105992 16.7 Charniodiscus 3B7 106028 12.5 Charnia 5C8 105996 15.1 Charniodiscus 3B7 106028 6.95 Charnia 1B2 105978 14.2 Charniodiscus 13B1 105968 32.5 Charnia 5B4 105971 6.2 Primocandelabrum 3D8 105963 11.6 Charnia 10B1 106084 3.3 Primocandelabrum 10C8 106052 4.3 Charnia 3C1a`(right) 105974 6.4 Primocandelabrum 4A3 105944 16.9 Charnia 3C1b (left) 105974 5.4 Primocandelabrum 10A7 106047 5.9 Charnia 3B5 105986 11.5 Primocandelabrum 1B1 105950 17 Charnia 3D? 105992 6.3 Primocandelabrum 5E2 105947 9 Charnia 2B2 105993 8.7 Primocandelabrum 13A8 105946 8.5 Charnia 3B2 105981 7.1 Primocandelabrum 6A1 105998 1.9 Charnia 5F4 105976 7.2 Primocandelabrum 5F3 105941 5.35 Charnia 19A2 106040 11 Primocandelabrum 13A4 105955 9.4 Charnia 20Bd 106039 10.7 Primocandelabrum 5C5 105961 8.2 Charnia 5C2 105966 10.2 Primocandelabrum 5C1 106026 8.7 Charnia A4 106000 14 Primocandelabrum 5G3 105942 9.2 Charnia 5C6 105959 11.8 Primocandelabrum 10B8 106049 12 Charnia A5 106001 35 Primocandelabrum 3(2) 105948 14.1

227 Species Field Specimen GSM Height Species Field Specimen GSM Height no. (cm) no. (cm) Primocandelabrum 10B3 106048 6.35 Orthiokaterna 5C6 105959 21 Primocandelabrum 4B3 105970 7.8 Orthiokaterna 19A1 106040 22.4 Primocandelabrum 7A4 105949 15.3 Bradgatia 10B4 106067 6 Primocandelabrum 5E3 105940 16.75 Bradgatia 5E1 106010 16 Primocandelabrum 5F5 106116 2.3 Undosyrus 5E4 105962 12 Primocandelabrum 2B1 105943 8.5 Undosyrus 5F6 106006 4 Primocandelabrum 3(1) 105945 15 Undosyrus 3D4 106007 10.2 Primocandelabrum 3(2) 105948 13.8 Undosyrus 3A1 106009 17 Primocandelabrum 10C3a 106051 5.4 Undosyrus 10A4 106066 8.9 Primocandelabrum 10C3b 106051 4.2 Undosyrus 20Ag 106039 4.1 Primocandelabrum 10C3c 106051 2.6 Undosyrus 11D2 105885 5.2 Primocandelabrum 8B4 106042 4.3 Undosyrus 11D1 105885 5.8 Primocandelabrum 8B3b 106041 4.4 Undosyrus 12A1 105886 10.6 Primocandelabrum 3D6 106016 10 Undosyrus 3bX 105872 9.7 Primocandelabrum 7A5 105952 3.4 Primocandelabrum 19B3 106040 Primocandelabrum 19C4 106049 6.85 Primocandelabrum 20Ba 106039 14.3 Primocandelabrum 19B1 106040 13.1 Primocandelabrum 10A2 106046 17.5 Primocandelabrum 7C6 105951 Primocandelabrum 10C4 106086 7 Primocandelabrum 19A4 106040 8 Primocandelabrum 19A3 106040 11.1 Primocandelabrum 4A1a (left) 105953 20.2 Primocandelabrum 4A1b (right) 105953 7.6 Primocandelabrum 10C5a 106073 5.05 Primocandelabrum 10C5b 106073 2.5 Primocandelabrum 6A2a (left) 105969 15.95 Primocandelabrum 6A2b (top left) 105969 7.3 Primocandelabrum 6A2c (right) 105969 22.7 Primocandelabrum 7A7 106121 7.9 Primocandelabrum Big Bertha 105872 46 Primocandelabrum 10A7 106047 5.8 Primocandelabrum 5F3 105941 5 Primocandelabrum 3C3 106111 5.9 Primocandelabrum 3F2 105954 28.9 duster 8C1 106043 8.8 duster 8C3 106045 7.2 Orthiokaterna 3D7 106034 12 Orthiokaterna 5A4 106012 7.55 Orthiokaterna 3E1(1) 105872 11.4 Orthiokaterna 6 105875 37.6 Orthiokaterna 6B2b 105875 12.4 Orthiokaterna 5G2 106113 7.6

228 A.2.2 Thin Section Data oating grains in Max. size fl bed below? oating fl Average Average grain size oating grains in fl bed below Glebules? P e r c e n t a g e Silt lenses in bed below? grain size Bradgate 0.006 1 0 5 na 0.1 Charnwood Charnwood Forest Location Formation matrix Average number Bed BSP.1SP.2 R003 BS.1BS.2 15105.1MC.1a 15105.2PU.V.3 16105.1 MPERPU.4a 101912 MPER 19105.1HF14.1a 19105.3 Brigus SouthHF14.1a2 Trepassey 19105.4 Brigus SouthSPS.1 Trepassey 20105.1a Bonavista TrepasseyPU.V.4 20105.1a2 Bonavista TrepasseyPU.1 0.006 Bonavista Mistaken Point Bonavista 0.008 21105.1PU.2 Bonavista 0.006 Mistaken Point 21105.4BC 0.006 0.008 Trepassey TrepasseyHS 0.008 Trepassey 22105.1 Bonavista 1STB 1 22105.2 Bonavista 0.5PU 0.5 0.012 0.5 0.008 Trepassey 22105.4 0.006 BonavistaSB 0 Trepassey 22105.7 BonavistaGHSB na 1 TrepasseyPC 0 0.006 1 Bonavista 0.5 0 na 1 0.5 TrepasseyCA 0 0.006 Bonavista na naMP 0 Fermeuse 0.006 0.008SH Bonavista 0.5 Fermeuse 0.01 0.006 na 0.05 0 0.5 2.5 0.5 Bonavista 0.5 0.5 na 0.25 0.5 Trepassey 1 Spaniard's Bay na 0.5 Spaniard's Bay 2.5 Trepassey 0.5 Trepassey Trepassey na 1 1 0.015 MPER 1 0.008 0.035 0.015 0.5 0.05 0.1 0.03 MPER 0.036 0.012 0.004 0.006 0 MPER 0 0.035 1 0.5 Drook MPER 0 0 0.17 0.5 0.04 0.015 0.08 0.035 0.136 1 0.03 0.5 0 Mistaken Point 0.5 0.15 Trepassey 0.05 0.1 0.006 0.012 0.05 0.02 0.035 0.1 0.5 1 0.12 1 0.004 0 0.035 1 0.014 0.025 0.5 0.5 0.01 0.025 0.08 0.035 0 0.05 0.5 0 0.05 0.1 0.1 0.1 1 0.03 1 0.06 0.05 0.036 0 0.03 0 0.03 0.015 0.1 5 0.068 0.15 0.07 0.05 0.02 2.5 0.01 0.035 0.035 0.02 0.07 0.12 0.05 0.12 Surface CGK sample

229 abundant? at Fractofusus fl Proportion Proportion Proportion fronds exc culmo and beo 0100 0001 Proportion Proportion stalked inc beo and culmo Proportion Proportion fronds Proportion Proportion stalked unifoliate or multifoliate? displayed or rotated (exc C. discid) furled? (exc C.discid) Total taxaTotal unfurled or Thickness Thickness bed below (mm) Bed BSP.111.810.50.50.5000000 0.82SP.2BS.1BS.2 1 14MC.1a 3.12PU.V.3 0.92 3.92 1PU.4a 3 7.4 6HF14.1a 1 2.48HF14.1a2 3.6 4 14.08SPS.1 0 12 1PU.V.4 17 0 5.9PU.1 0 12 12 6.9 0PU.2 1BC 1.04 0 0.5 8 1 1HS 6.12 0 12 0STB 1 4 0PU 0.5 1.32 6 1 0SB 0 0.5 3.92 1 0.214 0GHSB 0 1 2.64 0 13PC 9 1 0.68 0 1.4 12 0.5CA 0.357 0.5 0.5 0MP 15 0 0 0 11.88 1SH 0 7 1 1 0 3.72 8 0 0.214 0 1 30 0 4 0.083 0.167 0.118 0.88 0 0.667 0 3 0 1 0 0.5 1 1 0.357 0.5 0.5 0.417 17 0.5 0.5 0.333 0.412 5 0.5 0.125 0.5 0 0 0 0.083 0 1 0 0 0 1 0.083 0.25 0.5 1 0.118 1 0.25 0.5 0.5 0 0 0 0.417 0.667 0 0.417 0.412 0 0.25 0.5 0 1 0.5 1 0.231 0.167 0 0.25 1 0.111 0.5 0.083 0 0.333 0 0.133 0.059 0 0.385 0.083 0.417 0.25 0.444 0.167 0.375 0 0 0.417 0.125 0 0 0.25 0.533 0.5 0 0.5 0.308 1 0.167 0 0.111 0.5 0.5 0.125 0 0.5 0 0.857 0.267 0.25 0.118 0 0.333 0.2 0.308 0.417 0.444 0.5 0.25 0.143 0.4 0.125 0.667 0.353 0.25 0 0.167 0 0 0.6 0 0 0.714 0.5 0 0.176 0.067 0.25 0.5 0.4 0 0 0 0 0.294 0.5 0.25 0 0.667 0.4 0.176 1 0 0 0 1.5 0.2 0 0 Surface

230 A.2.3 Framboid maximum dimension counts

20105.1b 22105.3 15105.2 10.9.12 20105.1b 22105.3 15105.2 10.9.12 7.737 8.84 8.65 8.63 3.8 5.14 8.56 6.46 9.687 6.51 9.55 6 6.5 6.69 8.58 10.31 8.7 6.3 4.05 7.53 6.66 9.6 6.42 5.18 6.008 7.5 7.33 3.29 7.26 6.51 8.96 7.26 6.496 8.5 7.68 9.91 6.07 5.69 5.69 6.59 4.49 8.85 6.66 6.33 11.02 6.07 6.21 6.78 6.63 9.17 8.84 6.35 7.42 11.57 6.13 5.84 4.7 5.5 7.15 5.51 8.53 8.73 4.29 6.35 9.5 10.18 6.23 7.3 4.71 6.69 3.79 7.79 8.4 12 4.66 7.66 8.71 7.26 12.63 6.42 7.1 8.05 4.49 4.84 8.05 8.05 7.58 8.53 7.77 8.85 3.5 6.92 5.14 11.57 5.16 9.37 4.9 3.53 6.6 6.23 11.73 10.58 5.18 4.85 7.03 8.53 5.51 9.15 3.49 11.97 3.83 4.32 5.87 5.65 11.94 7.36 7.1 7.5 5.51 6.33 9.15 5.09 11 2.86 5.7 11.49 9.41 8.48 4.9 4.94 7.82 4.76 6.5 5.82 7.15 7.68 2.55 7.26 5.16 12.15 7.1 7.63 7.22 7.63 6.17 6.03 9.37 4.54 2.8 6.03 5.81 8.23 7.1 4.7 5.61 9.81 4.6 9.6 7.58 10.18 3.53 4.52 4.11 7.91 9.39 9.83 8.19 8.85 7.77 5.5 9.98 9.46 6.008 4.52 5.14 7.63 7.998 8.6 8.01 8.32 9.47 10.42 7.74 15.4 5.51 15.56 10.33 3.73 5.01 10.03 3.53 13.7 2.75 16.09 6.74 6.46 9.88 11.74 4.85 7.89 8.8 8.47 8.65 7.91 6.31 4.11 9.15 8.93 6.07 7.06 5.05 7.08 4.31 8.05 7.17 7.96 8.54 9.47 8.56 9.7 9.02 11.57 10.64 15.67 3.5 7.79 3.37 10.12 6.79 4.11 11.35 6.78 7.998 4.73 3.72 7.05 8.82 6.07 9.52 2.23 5.2 12.7 8.75 10.64 7.5 4.7 6.82 4.8 5.997 6.99 9.2 6.47 5.2 5.69 8.11 8.93 4.4 7.06 8.41 6.84 6.17 8.5 11.58 7.63 5.5 11.38 4.67 18.08 5.93 8.23 11.81 13.28 9.3 3.99 3.07 7.18 4.75 7.89 3.83 6.04 6.2 7.99 5.81 12.41 4.8 11.56 12.09 9.69 5.79 10.77 6.54 11.43 5.3 12.34 10.33 8.15 6.17 11.16 8.04 4 4.71 8.05 3.72 10.96 4.13 10.03 4.52 8.27 4.59 13.01 10.07 8.42 8.76 13.26 4.8 4.76 4.36 13.92 7.08 12.23 6.03 10.18 10.05 11.42 7.03 16.96 6.33 10.3 2.91 8.5 6.51 6.4 7.1 5.51 3.07 7.37 7.6 5.99 6.07 7.68 7.24 10.03 10.31 8.18 4.6 9.17 7.59 7.38 7.96 11.57 4.54 9.1 4 13.26 8.94 7.86

231 20105.1b 22105.3 15105.2 10.9.12 9.57 6.07 4.38 3.75 4.42 13.97 8.5 10.89 13 6.63 10.49 8.6 11.18 9.71 6.03 12.66 6.01 4.52 10.58 11.83 7.89 6.74 3.16 7.98 9.3 11.25 8.528 7.272 8.022 8.05 5.81 6.33 2.920 2.460 2.790

232 A.2.4 Raw measurements of Primocandelabrum and Orthiokaterna

Specimen Disc Disc No. D iam eter disc rings Stem Stem Crown Crown Length Length Total width height disc Inner Middle Outer height (c-bb)width at top at width height (i) left right height rings (c-i) (atbase) flare i) branch (i) branch (i) (c) 3D8 4.05 2.9 3 0.3 0.5 0.55 0.45 2.3 2.9 2.1 2.2 2 16.2 8.6 9.2 10.2 11.6 10C8 1.1 1 0 0.5 1.1 0.4 0.25 0.3 3.7 3.3 3.2 3.2 4.3 4A3 3.9 3.1 2 0.55 0.5 0.1 0.9 3.5 5.3 3.2 2.2 2.1 16.2 13.4 13.2 8.1 16.9 10A7 2.1 1.4 1 0.3 0.1 0.9 2 2.5 0.6 1.1 0.7 3.6 4.1 4 3.6 5.9 1B1 4.8 3.6 3 0.4 0.5 0.6 0.2 2.8 4.1 1.5 2.8 1.4 15.2 10.2 14.8 11.8 17 5E2 2.9 1.5 3 0.3 0.3 0.3 0.1 0.2 0.1 3.2 4.7 0.4 2 0.4 8.4 6.2 6.3 6.4 9 13A8 2.6 1.7 3 0.4 0.4 0.6 0.2 3.1 3.6 0.9 1.9 1.1 10.6 5.9 6.4 6.4 8.5 6A1 0.5 0.4 0 0.45 1 0.6 0.3 2.1 1.5 2.3 1.7 1.9 5F3 1 0.6 1 0.2 0.1 1.7 2.35 0.2 0.9 0.35 4.3 3.6 2.9 2.4 5.35 13A4 3.6 2.7 5 0.4 0.75 0.25 0.25 0.2 4 4.8 0.3 1.7 1.1 8.8 4.9 3.5 7 9.4 5C5 2.6 1.7 4 0.1 0.3 0.6 0.3 1.4 2.2 0.55 2.5 0.65 8.45 7 6.6 6.3 8.2 5C1 3.1 2.9 3 0.45 0.2 0.8 0.2 3.5 5.4 0.4 2.5 0.5 8.3 5 6.5 6.5 8.7 5G3 1.2 1.1 2 0.3 0.25 2 3.4 1.8 0.6 8.2 7.1 7.8 7.8 9.2 10B8 3.3 2.6 2 0.8 0.2 0.6 3.1 4.6 1.15 2.6 1.4 16.5 9.3 9.6 11.3 12 8C2 3.4 2.4 3 0.5 0.1 0.15 1.1 2.4 1.2 2.1 0.9 8.2 3.5 5 4.3 5.4 10B3 0.6 0 1.6 2.3 0.6 1.5 0.5 6 5 3.4 4 6.35 4B3 0.8 0.9 2 0.3 0.05 0.05 1.55 2.7 0.4 1.3 0.4 5.9 6.2 5.7 5.4 7.8 7A4 4.6 4.1 1 1.6 0.5 2.9 4.5 0.9 3.1 0.9 17.1 12.45 12.55 12 15.3 5E3 4.5 2.9 2 0.05 0.05 0.05 0.4 1.4 5.4 6 1.25 2.9 1.8 23.1 11.6 11.5 11.3 16.75 5F5 0.4 0.4 2 0.05 0.05 0.1 0.85 1.7 0.2 0.9 2.85 1.3 1.65 2 2.3 2B1 2.1 1.4 1 0.45 0.2 1.4 2.8 1 2.4 0.9 10.6 7 6.4 6.75 8.5 3(1) 8.1 6 2 1.7 1.3 0.2 1.2 4.3 5.1 2.9 3.8 2.1 27.5 10.8 17.8 15.5 15 3(2) 4 2.9 5 1.4 0.1 0.1 0.3 0.1 0.1 2.7 4 2.9 3.3 2.4 20 11.1 13.2 15.5 13.8 10C3a 1.8 1.1 1 0.35 0.5 1.1 2.1 0.7 2.1 0.7 8.2 4.8 4.8 5.3 5.4 10C3b 1 0.6 1 0.2 0.3 0.6 1.5 0.4 1.2 0.4 5.1 3.4 3.6 3.4 4.2 10C3c 0.6 0.4 0 0.75 1.1 0.6 0.8 0.6 3.3 1.8 2 1.9 2.6 8B4 1.45 1 2 0.2 0.2 0.1 0.3 0.45 1.4 0.45 1.25 0.2 5 3.85 2.65 3.65 4.3 8B3b 1.5 1.5 1 0.45 0.3 1.3 2.45 0.8 2.1 0.5 4.95 3.1 4.3 3.9 4.4 3D6 3 3 2 0.6 0.55 0.1 0.3 3.75 4.65 1 3.1 0.45 10.2 6.6 8 5.5 10 7A5 0.75 1.7 0.25 1.65 0.2 2.3 2.7 2.2 1.8 3.4 19B3 24.2 12 15.6 13 19C4 2.6 2.1 1 0.9 0.25 2.3 2.8 1.2 2.25 1.3 5.6 4 3.5 4.1 6.85 20Ba 5.5 4.8 4 0.05 0.7 0.9 0.25 0.5 5.4 6.9 0.2 2.1 0.4 10.2 9.3 11.2 9.3 14.3 19B1 6.5 4.6 1 1.8 1.1 3.5 5.1 2.6 3.3 2 22.8 9.9 13.5 9.2 13.1 10A2 5.5 4.3 1 3.2 0.1 4 5.9 3.7 3.7 2 18.8 13.3 16.8 14.1 17.5 7C6 3.3 2.6 5 0.05 0.6 0.15 0.35 0.3 0.4 2.9 3.8 1 10C4 2 1.4 2 0.25 0.2 0.75 1.65 3.25 0.7 2.3 0.7 7.2 5.4 5.1 5 7 19A4 2.7 1.9 2 0.25 0.2 0.7 2.3 3.6 0.7 1.6 0.35 5.8 5.85 5.6 5.3 8 19A3 4 3.2 2 1.4 0.2 0.3 2.55 3.65 1.3 2.4 1.2 9.8 8.6 8.5 8.1 11.1 4A1a (left) 3.1 2.5 2 0.35 0.45 0.5 2.5 4.3 0.3 2.7 0.4 6.7 5.5 4.4 5 7.6 4A1b (right) 10.5 9.1 3 2.2 1.5 1.2 0.8 9.4 12.9 4 4.7 0.75 18.1 11.9 13.3 12.9 20.2 10C5a 1.3 1.1 2 0.05 0.3 0.3 0.95 1.8 0.75 1.7 0.75 4.3 4.2 3 3.9 5.05 10C5b 1.1 0.9 1.1 1.2 0.1 0.2 0.1 1.8 1.4 1 1.1 2.5 10b10 2.9 2 1 0.9 0.5 3.1 3.8 1 1.4 0.6 5.3 4.9 5.2 5.2 7.8 3E4 3.8 2.8 0 3.5 5.5 1.4 0.4 7.15 7.8 6 5.4 11.2 7A6 1.3 0.9 1 0.3 0.25 1.4 2.6 0.35 1.4 0.4 5.8 5.1 3.8 4.2 6.7 7A7 2.5 1.9 1 0.6 0.75 2.8 3.9 0.6 1.5 1 4.6 5 3.8 4.1 7.9 3F2 17.4 13.3 3 13.3 1.7 25.2 16 16.3 15.6 28.9 3C3 1.6 1.3 1 1.7 2.65 0.6 0.5 4.1 3.4 2.4 2.5 5.9 6a2 l 5.4 4.4 3.5 5.5 2.25 2.7 2 21.6 15.85 14.8 12.8 15.95 6a2 r 4.9 4.1 4.6 6.7 2.3 3.5 1.5 25.5 17.1 16 15.5 22.7 Big Bertha 41.3 30.3 20.6 27.4 6.1 15.5 4.6 63 27.2 31.5 39.5 46 5A4 3.6 3 3.7 4.9 0.5 0.5 3.9 3.8 2.9 3 7.55 5G2 2.7 2.3 4.4 0.9 3 3.9 3.7 4.2 7.6 3E1(1) 3.1 2.3 7.6 7.9 0.6 3.7 3.7 2.3 2.9 11.4 3D7 5.6 5.4 6.8 7.6 1.7 1.4 4 5.4 3 3.1 12 6B2b 7.7 6.7 7.5 7.9 1.3 4.2 4.05 3.85 3 12.4 5C4 5.8 4.4 6.5 2.7 4.6 6 4.4 3.9 12.9 5C6 11.2 9.7 13.5 14.5 0.9 7.3 8 6.4 7.2 21 19A1 11.4 9.9 14.2 15.2 2.5 8.7 8.7 7 7.3 22.4 6 27 21.9 21.1 21.9 3.5 17.8 16.2 13.15 12.3 37.6 8C1 4.15 3.4 1 1.1 1.2 4.7 6.2 1.8 4.1 1.1 4.9 4.2 3.9 3.8 8.6 8C3 3.1 2.5 3 0.5 0.1 0.2 0.9 1.75 3.75 1.6 3.3 0.7 6.1 5.7 3.2 3.85 7.4 6a2 m 2.8 2.3 1.8 2.1 2.3 0.6 7.8 5.9 6 5.4 7.3 average all 4.71 3.76 1.94 0.64 0.39 0.39 0.28 0.20 0.49 4.03 4.99 1.23 2.40 1.07 10.55 7.22 7.34 7.15 11.09 std. dev. all 6.32 4.89 1.29 0.67 0.36 0.34 0.10 0.10 0.36 4.30 4.70 1.18 2.11 0.86 9.63 4.66 5.53 5.76 7.97 average 4.08 3.17 1.93 0.63 0.40 0.40 0.28 0.20 0.46 3.12 4.15 1.21 2.35 0.99 11.54 7.43 7.89 7.66 10.39 Prim ocandelabrum std. dev. 6.08 4.56 1.31 0.68 0.36 0.34 0.10 0.10 0.34 3.35 3.92 1.22 2.16 0.82 10.30 4.90 5.86 6.18 7.64 Prim ocandelabrum average 8.68 7.29 na na na na na na na 9.48 11.41 1.10 na 1.59 6.36 6.64 5.19 5.21 16.09 Orthiokaterna std. dev 7.59 6.19 na na na na na na na 5.66 5.99 0.85 na 1.06 4.68 4.04 3.38 3.18 9.58 Orthiokaterna

233 A.2.5 Proportions of Primocandelabrum and Orthiokaterna (divided by total height)

Specimen Disc Disc Stem Stem Crown Crown Length Length width height height width at top at width height (i) left right (c-i) (c-bb) (atbase) flare i) branch (i) branch (i) 3D8 0.35 0.25 0.20 0.25 0.18 0.19 0.17 1.40 0.74 0.79 0.88 10C8 0.26 0.23 0.12 0.26 0.09 0.06 0.07 0.86 0.77 0.74 0.74 4A3 0.23 0.18 0.21 0.31 0.19 0.13 0.12 0.96 0.79 0.78 0.48 10A7 0.36 0.24 0.34 0.42 0.10 0.19 0.12 0.61 0.69 0.68 0.61 1B1 0.28 0.21 0.16 0.24 0.09 0.16 0.08 0.89 0.60 0.87 0.69 5E2 0.32 0.17 0.36 0.52 0.04 0.22 0.04 0.93 0.69 0.70 0.71 13A8 0.31 0.20 0.36 0.42 0.11 0.22 0.13 1.25 0.69 0.75 0.75 6A1 0.26 0.21 0.24 0.53 na 0.32 0.16 1.11 0.79 1.21 0.89 5F3 0.19 0.11 0.32 0.44 0.04 0.17 0.07 0.80 0.67 0.54 0.45 13A4 0.38 0.29 0.43 0.51 0.03 0.18 0.12 0.94 0.52 0.37 0.74 5C5 0.32 0.21 0.17 0.27 0.07 0.30 0.08 1.03 0.85 0.80 0.77 5C1 0.36 0.33 0.40 0.62 0.05 0.29 0.06 0.95 0.57 0.75 0.75 5G3 0.13 0.12 0.22 0.37 na 0.20 0.07 0.89 0.77 0.85 0.85 10B8 0.28 0.22 0.26 0.38 0.10 0.22 0.12 1.38 0.78 0.80 0.94 8C2 0.63 0.44 na 0.44 0.22 0.39 0.17 1.52 0.65 0.93 0.80 10B3 0.09 na 0.25 0.36 0.09 0.24 0.08 0.94 0.79 0.54 0.63 4B3 0.10 0.12 0.20 0.35 0.05 0.17 0.05 0.76 0.79 0.73 0.69 7A4 0.30 0.27 0.19 0.29 0.06 0.20 0.06 1.12 0.81 0.82 0.78 5E3 0.27 0.17 0.32 0.36 0.07 0.17 0.11 1.38 0.69 0.69 0.67 5F5 0.17 0.17 0.37 0.74 0.09 0.39 0.00 1.24 0.57 0.72 0.87 2B1 0.25 0.16 0.16 0.33 0.12 0.28 0.11 1.25 0.82 0.75 0.79 3(1) 0.54 0.40 0.29 0.34 0.19 0.25 0.14 1.83 0.72 1.19 1.03 3(2) 0.29 0.21 0.20 0.29 0.21 0.24 0.17 1.45 0.80 0.96 1.12 10C3a 0.33 0.20 0.20 0.39 0.13 0.39 0.13 1.52 0.89 0.89 0.98 10C3b 0.24 0.14 0.14 0.36 0.10 0.29 0.10 1.21 0.81 0.86 0.81 10C3c 0.23 0.15 0.29 0.42 0.23 0.31 0.23 1.27 0.69 0.77 0.73 8B4 0.34 0.23 0.10 0.33 0.10 0.29 0.05 1.16 0.90 0.62 0.85 8B3b 0.34 0.34 0.30 0.56 0.18 0.48 0.11 1.13 0.70 0.98 0.89 3D6 0.30 0.30 0.38 0.47 0.10 0.31 0.05 1.02 0.66 0.80 0.55 7A5 na na 0.22 0.50 0.07 0.49 0.06 0.68 0.79 0.65 0.53 19B3 na na na na na na na na na na na 19C4 0.38 0.31 0.34 0.41 0.18 0.33 0.19 0.82 0.58 0.51 0.60 20Ba 0.38 0.34 0.38 0.48 0.01 0.15 0.03 0.71 0.65 0.78 0.65 19B1 0.50 0.35 0.27 0.39 0.20 0.25 0.15 1.74 0.76 1.03 0.70 10A2 0.31 0.25 0.23 0.34 0.21 0.21 0.11 1.07 0.76 0.96 0.81 7C6 na na na na na na na na na na na 10C4 0.29 0.20 0.24 0.46 0.10 0.33 0.10 1.03 0.77 0.73 0.71 19A4 0.34 0.24 0.29 0.45 0.09 0.20 0.04 0.73 0.73 0.70 0.66 19A3 0.36 0.29 0.23 0.33 0.12 0.22 0.11 0.88 0.77 0.77 0.73 4A1a (left) 0.41 0.33 0.33 0.57 0.04 0.36 0.05 0.88 0.72 0.58 0.66 4A1b (right) 0.52 0.45 0.47 0.64 0.20 0.23 0.04 0.90 0.59 0.66 0.64 10C5a 0.26 0.22 0.19 0.36 0.15 0.34 0.15 0.85 0.83 0.59 0.77 10C5b 0.44 0.36 0.44 0.48 0.04 0.08 0.04 0.72 0.56 0.40 0.44 10b10 0.37 0.26 0.40 0.49 0.13 0.18 0.08 0.68 0.63 0.67 0.67 3E4 0.34 0.25 0.31 0.49 na 0.13 0.04 0.64 0.70 0.54 0.48 7A6 0.19 0.13 0.21 0.39 0.05 0.21 0.06 0.87 0.76 0.57 0.63 7A7 0.32 0.24 0.35 0.49 0.08 0.19 0.13 0.58 0.63 0.48 0.52 3F2 0.60 0.46 0.46 na na na 0.06 0.87 0.55 0.56 0.54 3C3 0.27 0.22 0.29 0.45 0.10 na 0.08 0.69 0.58 0.41 0.42 6a2 l 0.34 0.28 0.22 0.34 0.14 0.17 0.13 1.35 0.99 0.93 0.80 6a2 r 0.22 0.18 0.20 0.30 0.10 0.15 0.07 1.12 0.75 0.70 0.68 Big Bertha 0.90 0.66 0.45 0.60 0.13 0.34 0.10 1.37 0.59 0.68 0.86 55A4A4 0.480.48 0.400.40 0.490.49 0.650.65 0.070.07 nana 0.070.07 0.520.52 0.500.50 0.380.38 0.400.40 55G2G 2 0.360.36 0.300.30 0.580.58 nana nana nana 0.120.12 0.390.39 0.510.51 0.490.49 0.550.55 33E1(1)E1(1) 0.270.27 0.200.20 0.670.67 0.690.69 nana nana 0.050.05 0.320.32 0.320.32 0.200.20 0.250.25 33D7D 7 0.470.47 0.450.45 0.570.57 0.630.63 0.140.14 nana 0.120.12 0.330.33 0.450.45 0.250.25 0.260.26 66B2bB2b 0.620.62 0.540.54 0.600.60 0.640.64 nana nana 0.100.10 0.340.34 0.330.33 0.310.31 0.240.24 55C4C 4 0.450.45 0.340.34 0.500.50 nana nana nana 0.210.21 0.360.36 0.470.47 0.340.34 0.300.30 55C6C 6 0.530.53 0.460.46 0.640.64 0.690.69 nana nana 0.040.04 0.350.35 0.380.38 0.300.30 0.340.34 119A19A1 0.510.51 0.440.44 0.630.63 0.680.68 nana nana 0.110.11 0.390.39 0.390.39 0.310.31 0.330.33 6 0.720.72 0.580.58 0.560.56 0.580.58 nana nana 0.090.09 0.470.47 0.430.43 0.350.35 0.330.33 88C1C 1 0.480.48 0.400.40 0.550.55 0.720.72 0.210.21 0.480.48 0.130.13 0.570.57 0.490.49 0.450.45 0.440.44 88C3C 3 0.420.42 0.340.34 0.240.24 0.510.51 0.220.22 0.450.45 0.090.09 0.820.82 0.770.77 0.430.43 0.520.52 66a2a2 m 0.380.38 0.320.32 0.250.25 0.290.29 nana 0.320.32 0.080.08 1.071.07 0.810.81 0.820.82 0.740.74 average all 0.36 0.28 0.33 0.45 0.12 0.26 0.10 0.93 0.67 0.67 0.66 std. dev. all 0.15 0.12 0.14 0.13 0.06 0.10 0.05 0.36 0.15 0.22 0.20 average 0.33 0.26 0.28 0.42 0.11 0.25 0.10 1.04 0.72 0.74 0.72 Prim ocandelabrum std. dev. 0.14 0.10 0.09 0.11 0.06 0.09 0.05 0.30 0.10 0.18 0.15 Prim ocandelabrum aaverageverage 0.490.49 0.410.41 0.580.58 0.650.65 0.100.10 na 0.100.10 0.390.39 0.420.42 0.330.33 0.330.33 OOrthiokaternarthiokaterna sstd.td. dedevv 0.130.13 0.120.12 0.060.06 0.040.04 0.050.05 na 0.050.05 0.070.07 0.070.07 0.080.08 0.100.10 OOrthiokaternarthiokaterna

234 A.2.6 Branching architecture of Primocandelabrum and Orthiokaterna.

Specimen 1st 1st Furled? 1st 1st 2nd 2nd 2nd 2nd 2nd Displayed? Infl ation? Concealed? Displayed? Furled? Radiating? Infl ation? Concealed? 3D8 rotated unfurled proximal concealed displayed unfurled subparallel median concealed 10C8 rotated unfurled proximal concealed displayed unfurled subparallel median concealed 4A3 rotated unfurled proximal unconcealed displayed unfurled subparallel concealed 10A7 rotated unfurled proximal concealed displayed unfurled subparallel median concealed 1B1 rotated unfurled proximal unconcealed displayed unfurled radiating median concealed 5.00E+2 rotated unfurled proximal concealed displayed subparallel 13A8 unfurled proximal unconcealed 6A1 furled radiating concealed 5F3 rotated unfurled median concealed displayed unfurled subparallel median concealed 13A4 rotated unfurled proximal furled radiating 5C5 rotated unfurled proximal 5C1 rotated unfurled proximal 5G3 rotated unfurled proximal unfurled 10B8 rotated unfurled proximal concealed displayed unfurled median concealed 8C2 rotated unfurled unconcealed unfurled radiating unconcealed 10B3 rotated unfurled concealed displayed unfurled radiating median concealed 4B3 unfurled proximal concealed displayed unfurled radiating median concealed 7A4 rotated unfurled proximal concealed displayed unfurled concealed 5.00E+3 rotated unfurled proximal concealed unfurled radiating concealed 5F5 rotated unfurled proximal concealed displayed unfurled 2B1 rotated unfurled proximal unconcealed concealed 3(1) rotated unfurled unconcealed 3(2) rotated unfurled proximal unconcealed displayed unfurled radiating concealed 10C3a rotated unfurled proximal unconcealed displayed unfurled subparallel unconcealed 10C3b rotated proximal unconcealed subparallel 10C3c unconcealed 8B4 rotated unfurled proximal unconcealed displayed unfurled subparallel proximal concealed 8B3b rotated unfurled proximal concealed radiating 3D6 7A5 rotated unfurled proximal unconcealed displayed subparallel concealed 19B3 rotated unfurled proximal concealed displayed furled subparallel median concealed 19C4 rotated unfurled proximal unconcealed displayed unfurled radiating proximal concealed 20Ba rotated median concealed radiating median concealed 19B1 rotated proximal unconcealed subparallel concealed 10A2 rotated unfurled proximal unconcealed displayed unfurled subparallel concealed 7C6 rotated unconcealed 10C4 poor 19A4 rotated proximal concealed median concealed 19A3 rotated proximal unconcealed rotated unfurled median concealed 4A1a (left) 4A1b (right) rotated unfurled proximal concealed rotated unfurled subparallel median concealed 10C5a very poor 10C5b very poor 10b10 rotated unfurled proximal concealed rotated unfurled radiating median concealed 3.00E+4 poor 7A6 poor 7A7 poor 235 Specimen 1st 1st Furled? 1st 1st 2nd 2nd 2nd 2nd 2nd Displayed? Infl ation? Concealed? Displayed? Furled? Radiating? Infl ation? 3F2 jumble 3C3 unfurled proximal concealed displayed unfurled radiating median concealed 6a2 l rotated unfurled proximal unconcealed displayed unfurled radiating median concealed 6a2 r 5A4 displayed unfurled na concealed displayed unfurled na na concealed 5G2 na na na concealed na na na na concealed 3E1(1) displayed unfurled na concealed displayed furled radiating proximal concealed 3D7 na na na concealed displayed na na na concealed 6B2b displayed furled na concealed displayed furled radiating na concealed 5C4 na na na concealed na na na na concealed 5C6 displayed furled na concealed displayed furled radiating na concealed 19A1 displayed furled na concealed displayed furled radiating proximal concealed 8C1 displayed unfurled median unconcealed rotated unfurled subparallel median concealed 8C3 displayed unfurled median concealed displayed unfurled subparallel median concealed 6a2 m 6 displayed unfurled distal concealed displayed furled radiating proximal na Big Bertha unfurled displayed unfurled subparallel distal concealed

236 A.2.7 Taxonomic assignments of Primocandelabrum specimens.

Specimen Taxon 10A7 P. anatonos 5E2 P. anatonos 5F3 P. anatonos 13A4 P. anatonos 5C1 P. anatonos 5F5 P. anatonos 19C4 P. anatonos 20Ba P. anatonos 19A4 P. anatonos 4A1b P. anatonos (right) 10b10 P. anatonos 3C3 P. anatonos 10C8 P. boyntoni 4A3 P. boyntoni 1B1 P. boyntoni 13A8 P. boyntoni 5C5 P. boyntoni 5G3 P. boyntoni 10B8 P. boyntoni 10B3 P. boyntoni 4B3 P. boyntoni 7A4 P. boyntoni 5E3 P. boyntoni 2B1 P. boyntoni 10C3b P. boyntoni 8B4 P. boyntoni 7A5 P. boyntoni 10A2 P. boyntoni 19A3 P. boyntoni 6a2 l P. boyntoni 3D8 P. katatonos 6A1 P. katatonos 8C2 P. katatonos 3(1) P. katatonos 3(2) P. katatonos 10C3a P. katatonos 10C3c P. katatonos 8B3b P. katatonos 19B1 P. katatonos

237 A.2.8 Regression calculations for size data for Orthiokaterna dence fi 95% con dence fi con p-value slope 95% 2 adjusted R 2 R dence fi 95% con dence fi con -value slope 95% p 2 adjusted R 2 0.990.97 0.99 0.97 4.90E-50.97 0.21 8.69E-7 0.97 0.61 0.18 7.34E-7 0.52 0.24 0.85 0.7 0.94 0.73 0.85 0.92 0.97 0.83 0.006 0.89 0.22 0.001 0.87 0.53 0.12 0.000 0.31 0.32 0.94 0.75 0.61 1.27 all specimensR excluding largest specimen ection 0.58 0.52 0.018 0.08 0.02 0.15 0.23 0.1 ection 0.01 0.23 -0.13 0.07 0.77 -0.06 -0.001 0.2 -0.005 0.004 0.01 -0.15 0.79 -0.001 -0.01 0.008 fl fl Absolute measurements plotted agaist total height Disc widthCombined (n = 6)Stem height (c-i)Stem width at in 0.85 0.96Crown width 0.98Crown height (i) 0.82 0.95Length left (i) 0.97Length right (i) 0.96 0.008 0.96 4.45E-6Absolute measurements 5.88E-7Disc width vs disc height 0.17 0.94 0.78 0.95 0.96 1.00 0.58 0.93disc width vs triangle width 0.07 0.93 0.63 4.71E-6Disc width vs crown 3.41E-6 0.995 0.5 0.92width 0.48 0.41length left vs right 0.27 1.59E-5 0.92 1.60E-9 0.97 3.05E-5Crown width vs crown 0.66 0.34 0.81 0.39height 0.34 0.32 0.47 0.89Proportions of absolute measurements to total height plotted against 0.96 0.97 0.36Disc width 0.77 0.57 0.49 0.24 0.29 0.88Stem height (c-i) 1.67E-6 0.97 0.42Stem width at in 0.86 0.92 0.88 0.93 0.4 0.2 0.000Crown width 0.03 5.35E-6 0.51 0.8Crown height (i) 0.99 0.9 0.86 0.59 0.78 0.69 0.77 0.1Length left (i) -0.11 0.45Length right (i) 0.99 0.76 0.06 0.001 0.000 0.04 0.39 1.08 0.58 0.73Crown width/disc width -0.09 0.65 0.03 0.29 0.01 0.34 5.02E-7 0.35 0.003 -0.07 -0.09 0.79 0.91 0.04 0.8 0.004 0.29 0.001 0.89 0.01 0.18 0.27 -0.13 0.51 0.21 0.25 0.58 0.9 0.3 -0.1 -0.004 0.79 0.001 0.79 0.14 0.14 0.001 0.46 0.45 -0.002 0.007 0.000 0.62 0.13 0.98 0.018 -0.004 -0.001 -0.013 -0.008 0.41 0.3 1.06 -0.002 0.46 0.17 0.008 -0.008 -0.03 0.005 -0.01 0.73 0.18 0.007 0.12 0.23 0.03 0.005 0.007 1.39 0.16 0.13 0.33 -0.03 0.01 0.31 0.1 0.006 -0.016 0.41 0.22 0.23 0.01 -0.05 -0.003 0.38 -0.004 0.14 -0.007 -0.01 0.016 0.44 -0.006 -0.014 0.005 -0.02 0.03 -0.02 -0.006 0.007 -0.02 -0.06 -0.02 0.009 0.01 0.01

238 239 A.2.9 Raw measurements and branching architecture of Undosyrus rst fi Groove to division (mm) 15.5, 7.75 15.8 of division obscured incomplete Groove (cm) (straight line connecting base to tip) 3,2 no. grooves: projected Grooves (pairs): observed Length right (i) Length left (i) Crown height Crown width Stem width at i Proportions (divided by total height) No. Stem height (c-i) Disc width 0.02 0.09 0.04 0.11 0.05 0.09 0.11 Total Total height to c Length right (i) Length left (i) Crown height (i) Crown width Stem width at i Stem height (c-i) width 5.00E+4 1.35F6 4.33D4 0.9 0.6 11.33A1 1.1 1.3 910A4 4.6 0.320Ag 2.7 1.3 1.9 9.1 1.6 8.5 0.8 2.9 10.1 0.9 1.2 7.4 12 3.1 7.6 0.7 7.2 0.11 2.5 11.2 4.1 7.3 0.36 8.6 9.4 4 4.3 0.08 5.6 0.94 10.2 11.5 3.9 0.15 6.9 0.75 0.13 9 0.28 0.45 3.3 0.76 8.9 0.08 0.19 0.84 0.68 4.1 0.15 0.83 50 0.73 0.18 0.16 0.73 0.78 0.10 0.24 0.71 0.85 65 0.63 0.14 0.84 0.82 0.80 15 56 0.63 0.84 5.1 0.78 0.76 31 0.65 7.9, 7.8, 27 1.7 44 5.6, 7.4 32 20.5 3.1 n/a 2.8, 3.4, 16 46 62 7.3, 10 17, 21 Specimen Disc 11D211D1 0.9 1.2 0.712A1 0.5 1.8 5.2 0.6 1.53Bx 4.5 2 3.5average 1.1 4.5standard 2.9 0.7 2deviation 10.5 3.8 3.3 0.7 8.5 3.4 5.2 7.7 9.1 5.8 0.17 7.2 0.23 9.3 0.12 0.10 0.31 7 10.6 1.00 0.10 0.14 0.67 6.2 0.78 0.19 0.56 0.78 0.07 9.7 0.63 0.66 0.99 0.11 11 0.59 0.80 0.21 20 0.86 0.07 33 0.88 0.79 34 0.74 44 1.5 0.72 0.14 1.9 0.27 0.64 56 0.10 61 n/a 0.85 3.4, 2.4 n/a - base 0.76 67 too 0.71 0.72 5.5 21.3

240 3rd Concealed? ation? fl 3rd In 3rd Radiating? 3rd Furled? 3rd Displayed? rotated furled radiating concealed 2nd Concealed? ation? fl 2nd In median rotated concealed 2nd Radiating? 2nd Furled? 2nd Displayed? Concealed? ation? 1st fl 1st In 1st Furled? Displayed? Specimen 1st 5.00E+4 rotated5F6 unfurled3D4 rotated 3A1 rotated10A4 rotated concealed20Ag rotated displayed unfurled furled11D2 radiating11D1 median12A1 concealed concealed rotated displayed3Bx furled furled concealed radiating rotated displayed subparallel displayed no furled radiating concealed concealed rotated displayed furled concealed unfurled concealed radiating distal displayed displayed furled concealed furled concealed median concealed concealed furled radiating rotated furled concealed

241 A.2.10 Peak heights and overlap distributions peak overlaps with Orthiokaterna and "dusters" Orthiokaterna Orthiokaterna and Charnia and Primocandebrum and Primocandebrum "dusters" and Primocandebrum Bradgatia Fractofusus Fractofusus Fractofusus range frondose mass upper range frondose mass lower peak frondose mass upper frondose mass lower 07 65 8 05 6 5 33 6 1 everything except 385 0 nothing 46 6 everything except 17 0 Bradgatia, Charniodiscus 46 Bradgatia, Charniodiscus, 5 80 4 2 9 0 Bradgatia, Charniodiscus 2 352 3 Charnia, Charniodiscus, 3 2 0 22 everything except 16 everything except Undosyrus and "duster" Undosyrus and "duster" Charniodiscus and Orthiokaterna Charniodiscus and Orthiokaterna Orthiokaterna and "duster" everything except Charniodiscus and Orthiokaterna Orthiokaterna and Charniodiscus everything 0everything except "frond" and Fractofusus 2 0 33 everything C'discus "frond" and "ostrich feather" "frond" and "ostrich feather" "Charnia" and "frond" 95% 95% overlaps with IQR overlaps with peak 95% quantile lower total height 41 9.4 8 11 stalked everything except Taxon n median Charnia 41Charniodiscus 9.4 27 8Orthiokaterna 12.2 9 11 11.4Primocandelabrum 53 12.4 everything 13.5 8.2Undosyrus 7.4 Orthio, Brad, Char 6.6 everything except "dusters" 17.8 everything 9 10.2 everything 9.3Stalked everything 2 6.8 0Upright, non- everything except 8stalked 12 everything except 103 2 10.2 7.2 everything 9.3 9.2 0 11.4 everything except Bradgatia everything except upright 35Charniodiscus 11 Charniodiscus, "Ostrich feathers" 9.4 54 everything except 4.4 5 4 13.7 everything except 4.9 everything except Bradgatia 2 11 5.5 17 everything everything 0 5"Charnia" 0 151 5.3 15 4.6 everything except 5.8 everything except "Charnia 2" 51 5.5 4.1 6.7 everything everything 2 3 0 20 everything except Bedding Surface Bed B All 220 9.2 8.5 10.1LMP All 5 285 6 5 0 4.7 5.5 46 2 3 0 33

242 peak overlaps with Fractofusus Fractofusus Fractofusus range frondose mass upper range frondose mass lower peak frondose mass upper 20 31 1 0 3 0 12 0 9 everything except 60 "Charnia", "frond" 0 everything 60 7 00 2 0 19 4 00 27 everything 0 60 everything 30 23 everything 6 0 everything 2 0 20 0 everything 27 23 everything everything frondose mass lower Pectinifrons and "duster" "unknown" Pectinifrons and "duster" Pectonifrons, and "dusters" "frond" and Fractofusus Charniodiscus Fractofusus, "unknown" Bradgatia, Pectinifrons, Fractofusus, "unknown" everything except "dusters", "frond" "Charnia" and "frond" "unknown" Bradgatia, Charnia 95% overlaps with IQR overlaps with peak 95% quantile upper 95% quantile lower total height 124 11.8 11.1 12.6 "unknown", 161 5.1 4.5 5.7 everything everything 1 2 0 33 everything "frond""unknown" 7 3 3.8 18.1 3 0.3 4.7"Charnia" 36 everything except except Fractofusus everything 17Pectinifrons 1169 11.5 8.7 Charniodiscus 0"dusters" 9.4 175"frond" 8.5 18.1"unknown" 33 13.7 9 8 17 Bradgatia, everything Upright, non- 2 29 4.1stalked "unknown" 19.3Flat-lying 9.8 4.7 2.9 nothing everything except 8.2 1.6 5.3 1345 9.4 11.4 7.7 "frond" Bradgatia, "Charnia", "dusters" 9.1 9.7 "frond" "dusters", Fractofusus "unknown" 0 everything except everything except 1 2 0 2 0 8 everything 7 everything "dusters"Fractofusus 4 4 5.3 2.8 3.3Stalked 0 7.3Upright, non- stalked 5.8 everything 120 everything 4.9 everything except 4.2 everything except 5.4 everything everything 2 3 0 22 everything except Bradgatia 76 12.2 11.4Pectinifrons, 13 "Charnia", Taxon n median MPE All 3231 6.3 6.1 6.4 0 1 0 22 MPD All 1457 9.6 9.2 9.8 0 2 0 27 Bedding Surface

243 peak overlaps with Charniodiscus and "dusters" Charniodiscus and "dusters" Charniodiscus and "dusters" Charniodiscus and "dusters" Charniodiscus and "dusters" Charniodiscus and "dusters" range frondose mass upper range frondose mass lower peak frondose mass upper frondose mass lower 11 20 2 0 1 0 21 0 everything except Fractofusus 22 11 everything except Fractofusus everything except 011013 10 2 3 00 0 22 4 everything except Fractofusus 22 0 everything 13 everything except 2 3 0 14 everything except 0 40 02 3 70 3 0 everything except 3 0 12 0 32 everything except everything 13 everything except "duster", Fractofusus, "unknown" Charniodiscus, "unknown" "dusters" "Charnia", Fractofusus, Thectardis, "unknown" everything except Thectardis, "spoon frond" and "Charnia" everything except Thectaridis, "spoonf frond" and "dusters" everything except Charniodiscus and "duster" Charniodiscus and "duster" everything except Charniodiscus and "duster" eveything except Charniodiscus "upright, non-stalked" , "duster" and Charniodiscus everything except Charniodiscus 95% overlaps with IQR overlaps with peak Thectardis, "Charnia" "Charnia", "unknown" "dusters", "dusters", Charniodiscus "unknown", Bradgatia Charniodiscus Charniodiscus "duster" "duster", "unknown" 95% quantile upper 95% quantile lower total height TaxonCharniodiscus n 825"Dusters" median 4Fractofusus 362Thectardis 3.7 1497 3.4"spoon frond" 7.8 4.2 25 3.2 2 "Dusters" 7.5 10.7 3.6 12.9 8 9.6except 7.9 Charniodiscus Bradgatia, "Charnia", 11.7 "Charnia" everything 17.7 "spoon frond" "Charnia", "spoon frond" 0 5 0 17 everything Bradgatia 55 5.2 4.6 5.8 "unknown"Stalked 67Upright, non-stalked 379 5.7 1187 6.2 3.8 4.8 5.9 3.7 6.7 6.4 4 Bradgatia everything "unnamed frond" 28"Charnia" 3.4Charniodiscus 2.9 4"dusters" 8"unknown" 4 4.6Stalked 11.1 5 "Charnia" 3.2 20 6.6 7.9 5.4 6 13 15.5 everything 4.9 4.1 10.1 everything except "duster" except 10.9 6.6 7.4 everything "frond" everything excpet 12.8 "duster" 5 Charniodiscus Chariodiscus and 11 2 0 3 32 1 nothing 12 nothing Upright, non-stalked 107 4.9 4.5 5.3 Bradgatia, "Charnia", Bedding Surface MPG All 120 5 4.5 5.4 2 3 0 32

244 peak overlaps with range frondose mass upper range frondose mass lower 047 15 peak frondose mass upper frondose mass lower "Charnia" 0 2everything 0 0 8 3 Fractofusus 0 47 everything 95% overlaps with IQR overlaps with peak Fractofusus Bradgatia, Bradgatia, "unknown" 95% quantile upper 95% quantile lower total height Taxon n median "Charnia"Thectardis"unknown"Upright, non-stalked 18 20 140 13.3 2 7.9 11.8Fractofusus"Charnia" 9.6 6.6 8"unknown" 7.2 76 16.9 3.4 8.6 1 15.5 5Bradgatia "unknown" 3 9.9 everything"Charnia" "unknown" 4.7 3.9 0.9 everything Thectardis 10 3.5 1.1 everything 0.9 everything 1 5 5.9 "unknown" Thectardis 6.9 47 0 "unknown" 4.2 "Charnia" and 0 0 nothing 5.7 0 6 "unknown" "unknown" 6 2 0 5 0 0 0 "unknown" 2 0 23 1 23 16 3 everything 0 10 everything everything 0 0 everything 3 8 8 0 "unknown" nothing 10 everything CharniodiscusPectinifrons 1"unknown"Upright, non-stalked 303 35 4.4 7.4 24 4.6 4 7.1 3.7 7.7 2.8 5.4 nothing 5 Bradgatia "unknown" Bradgatia, Pectinifrons, 0 2 0 12 everything everything Bedding Surface BC All 84 1 0.9 1.1 0 1 0 8 PC All 160 8.2SH 7.5 All 8.9 339H14 7.1 Fractofusus 6.7 7.4 0 2 0 23 1 2 0 47

245