PreprintPDF Available

Variable responses of individual species to tropical forest degradation

February 2024

DOI:10.1101/2024.02.09.576668

Authors:

David Orme

Imperial College London

Show all 126 authorsHide

Preprints and early-stage research may not have been peer reviewed yet.

The functional stability of ecosystems depends greatly on interspecific differences in responses to environmental perturbation. However, responses to perturbation are not necessarily invariant among populations of the same species, so intraspecific variation in responses might also contribute. Such inter-population response diversity has recently been shown to occur spatially across species ranges, but we lack estimates of the extent to which individual populations across an entire community might have perturbation responses that vary through time. We assess this using 524 taxa that have been repeatedly surveyed for the effects of tropical forest logging at a focal landscape in Sabah, Malaysia. Just 39 % of taxa - all with non-significant responses to forest degradation - had invariant responses. All other taxa (61 %) showed significantly different responses to the same forest degradation gradient across surveys, with 6 % of taxa responding to forest degradation in opposite directions across multiple surveys. Individual surveys had low power (< 80 %) to determine the correct direction of response to forest degradation for one-fifth of all taxa. Recurrent rounds of logging disturbance increased the prevalence of intra-population response diversity, while uncontrollable environmental variation and/or turnover of intraspecific phenotypes generated variable responses in at least 44 % of taxa. Our results show that the responses of individual species to local environmental perturbations are remarkably flexible, likely providing an unrealised boost to the stability of disturbed habitats such as logged tropical forests.

Broad taxonomic patterns in the replicability of single-year biodiversity surveys. (A) Number

…

The probability of an analysis returning the correct sign (direction) of a taxon's response to forest degradation as a function of the number of single-year surveys in which that taxon was detected. (A) Each grey line represents an analysis for one taxon (n = 494). (B) Violin plots showing the distribution of probabilities of a single-year survey returning the correct sign for seven taxonomic groups. Points indicate the probability for individual taxa within each group.

…

Figure S3. Bootstrapped estimates of the proportion of taxa with invariant response patterns with respect to (A) year of survey, (B) El Niño events and (C) logging events. Thick line represents the median, boxes the 1 st and 3 rd quantiles, and whiskers the range. In panels (B) and (C), data were categorised into pairwise comparisons of taxon responses to forest degradation in which both surveys were conducted in years during which the event occurred (within), when one survey occurred during an event and the other occurred outside of the event (straddling), and when both surveys occurred outside of the event (outside).

…

Figures - uploaded by David Hemprich-Bennett

Content may be subject to copyright.

Content uploaded by David Hemprich-Bennett

Content may be subject to copyright.

Variable responses of individual species to

tropical forest degradation

Robert M. EWERS 1; William D. PEARSE 1; C. David L. ORME 1; Priyanga AMARASEKARE 2; Tijmen DE

LORM 1; Natasha GRANVILLE 1; Rahayu ADZHAR 1,3; David C. ALDRIDGE 4; Marc ANCRENAZ 5,6;

Georgina ATTON 7; Holly BARCLAY 8; Maxwell V. L. BARCLAY 9; Henry BERNARD 10; Jake E. BICKNELL

11; Tom R. BISHOP 1,12,13; Joshua BLACKMAN 14; Sabine BOTH 15; Michael J. W. BOYLE 1,16; Hayley

BRANT 1; Ella BRASINGTON 1; David F.R.P. BURSLEM 17; Emma R. BUSH 18; Kerry CALLOWAY 9; Chris

CARBONE 19; Lauren CATOR 1; Philip M. CHAPMAN 1,20; Vun Khen CHEY 21; Arthur CHUNG 21; Elizabeth

L. CLARE 14,22; Jeremy CUSACK 1,23; Martin DANČÁK 24; Zoe G. DAVIES 11; Charles W. DAVISON 1,25,26;

Mahadimenakbar M. DAWOOD 10; Nicolas J. DEERE 11; Katharine J. M. DICKINSON 27; Raphael K.

DIDHAM 28,29; Timm F. DÖBERT 28,29,30; Rory A. DOW 31,32; Rosie DRINKWATER 14; David P. EDWARDS

33,63; Paul EGGLETON 9; Aisyah FARUK 34; Tom M. FAYLE 14,35; Arman Hadi FIKRI 10; Robert J. FLETCHER

36; Hollie FOLKARD-TAPP 1; William A. FOSTER 4; Adam FRASER 1; Richard GILL 1; Ross E. J. GRAY 1;

Ryan GRAY 37; Nichar GREGORY 1,38; Jane HARDWICK 39; Martina F. HARIANJA 4; Jessica K. HAYSOM 11;

David R. HEMPRICH-BENNETT 14,40; Sui Peng HEON 1,37; Michal HRONEŠ 41; Evyen W. JEBRAIL 10; Nick

JONES 42; Palasiah JOTAN 43; Victoria A. KEMP 14; Lois KINNEEN 44; Roger KITCHING 45; Oliver KONOPIK

46; Boon Hee KUEH 10; Isolde LANE-SHAW 1,47; Owen T. LEWIS 40; Sarah H. LUKE 4,48,49; Emma

MACKINTOSH 1,50; Catherine S. MACLEAN 1; Noreen MAJALAP 21; Yadvinder MALHI 51; Stephanie

MARTIN 1,52; Michael MASSAM 1,53; Radim MATULA 43; Sarah MAUNSELL 45; Amelia R. MCKINLAY 1;

Simon MITCHELL 11; Katherine E. MULLIN 11; Reuben NILUS 21; Ciar D. NOBLE 1,54; Jonathan M.

PARRETT 55; Marion PFEIFER 56; Annabel PIANZIN 10; Lorenzo PICINALI 57; Rajeev PILLAY 36,58; Frederica

POZNANSKY 1,59; Aaron PRAIRIE 1,60; Lan QIE 1,61; Homathevi RAHMAN 10; Terhi RIUTTA 1,51,62; Stephen

J. ROSSITER 14; J. Marcus ROWCLIFFE 19; Gabrielle Briana ROXBY 1; Dave J. I. SEAMAN 11; Sarab S.

SETHI 1,63; Adi SHABRANI 64,65; Adam SHARP 1,66; Eleanor M. SLADE 67; Jani SLEUTEL 68; Nigel STORK 69;

Matthew STRUEBIG 11; Martin SVÁTEK 70; Tom SWINFIELD 4; Heok Hui TAN 71; Yit Arn TEH 56; Jack

THORLEY 4; Edgar C. TURNER 1,4; Joshua P. TWINING 1,72; Maisie VOLLANS 1,40; Oliver WEARN 1,73;

Bruce L. WEBBER 28,29; Fabienne WIEDERKEHR 1,74; Clare L WILKINSON 1,75; Joseph WILLIAMSON 14,76;

Anna WONG 77; Darren C. J. YEO 71,75; Natalie YOH 11,78; Kalsum M. YUSAH 10,79; Genevieve YVON-

DUROCHER 80; Nursyamin ZULKIFLI 81; Olivia DANIEL 1; Glen REYNOLDS 37; Cristina BANKS-LEITE 1

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

1 Georgina Mace Centre for the Living Planet, Department of Life Sciences, Imperial College London,

Silwood Park Campus, Buckhurst Road, Ascot SL5 7PY, UK

2 Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA, United

States

3 Department of Geographical Sciences, University of Maryland, College Park, MD, USA

4 Department of Zoology, The David Attenborough Building, University of Cambridge, Cambridge CB2

3QZ

5 Borneo Futures, Bandar Seri Begawan, Brunei

6 Kinabatangan Orang-Utan Conservation Programme, Kota Kinabalu, Sabah, Malaysia

7 Faculty of Health Sciences, University of Bristol, BS8 1UD, UK

8 School of Science, Monash University, Jalan Lagoon Selatan, 47500 Subang Jaya, Selangor Darul

Ehsan, Malaysia

9 Department of Life Sciences, The Natural History Museum London, SW7 5BD, UK

10 Institute for Tropical Biology and Conservation, Universiti Malaysia Sabah, Jalan UMS, 88400 Kota

Kinabalu, Sabah, Malaysia

11 Durrell Institute of Conservation and Ecology (DICE), School of Anthropology and Conservation,

University of Kent, Canterbury, CT2 7NR, UK

12 Department of Zoology and Entomology, University of Pretoria, Pretoria, South Africa

13 School of Biosciences, Cardiff University, Cardiff, UK

14 School of Biological and Behavioural Sciences, Queen Mary University of London, London, E1 4NS,

15 School of Environmental and Rural Science, Faculty of Science, Agriculture, Business and Law, Uni-

versity of New England, Armidale NSW 2351, Australia

16 School of Biological Sciences, The University of Hong Kong, Hong Kong City, Hong Kong

17 School of Biological Sciences, University of Aberdeen, Aberdeen, AB24 3UU UK

18 Royal Botanic Gardens Edinburgh, Arboretum Place, Edinburgh EH3 5NZ, UK

19 Institute of Zoology, Zoological Society of London, NW1 4RY, UK

20 BSG Ecology, Worton Park, Worton, Witney, Oxfordshire, OX29 4SX, UK

21 Forest Research Centre, Sabah Forestry Department, PO Box 1407, 90715 Sandakan, Sabah, Malay-

sia

22 Department of Biology, York University, Toronto, Ontario, M3J 1P3, Canada

23 Centro de Modelación y Monitoreo de Ecosistemas, Universidad Mayor, Chile

24 Department of Ecology and Environmental Sciences, Faculty of Science, Palacký University,

Šlechtitelů 27, CZ-78371, Czech Republic

25 Center for Biodiversity Dynamics in a Changing World (BIOCHANGE), Department of Biology, Aar-

hus University, Ny Munkegade 114, DK-8000 Aarhus C, Denmark

26 Center for Ecological Dynamics in a Novel Biosphere (ECONOVO), Department of Biology, Aarhus

University, Ny Munkegade 114, DK-8000 Aarhus C, Denmark

27 Department of Botany, University of Otago, PO Box 56, Dunedin, 9054, New Zealand

28 CSIRO Health and Biosecurity, Centre for Environment and Life Sciences, 147 Underwood Ave, Flo-

reat, WA, 6014 Australia

29 School of Biological Sciences, The University of Western Australia, 35 Stirling Highway, Crawley,

WA 6009, Australia

30 Faculty of Science, University of Alberta, AB, T6G 2E1, Canada

31 Institute of Biodiversity and Environmental Conservation, Universiti Malaysia Sarawak, Malaysia

32 Naturalis Biodiversity Centre, PO Box 2300 RA Leiden, The Netherlands

33 Ecology and Evolutionary Biology, School of Biosciences, University of Sheffield, S10 2TN, UK

34 Royal Botanic Gardens, Kew, Wakehurst, Ardingly, Haywards Heath, West Sussex RH17 6TN, UK

35 Biology Centre of the Czech Academy of Sciences, Institute of Entomology, Branisovska 31, 370 05

Ceske Budejovice, Czech Republic

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

36 Department of Wildlife Ecology and Conservation, University of Florida, 110 Newins-Ziegler Hall,

Gainesville, FL 32611, USA

37 South East Asia Rainforest Research Partnership, Danum Valley Field Centre, 91112 Lahad Datu,

Sabah, Malaysia

38 EcoHealth Alliance, 520 Eighth Ave., Ste 1200, New York, NY 10018, USA

39 Marine Resources Unit, Department of Environment, Grand Cayman, Cayman Islands

40 Department of Biology, University of Oxford, Oxford OX1 3PS, UK

41 Department of Botany, Faculty of Science, Palacký University, Šlechtitelů 27, CZ-78371, Czech Re-

public

42 Department of Mathematics, Imperial College London, South Kensington Campus, London SW7

2AZ, UK

43 Department of Forest Ecology, Faculty of Forestry and Wood Sciences, Czech University of Life Sci-

ences Prague, Czech Republic

44 Department of Sustainable Land Management, School of Agriculture, Policy and Development,

University of Reading, Berkshire, UK

45 School of Environmental and Natural Sciences, Griffith University, Brisbane, Queensland 4111,

Australia

46 Department of Animal Ecology and Tropical Biology, Biocenter, University of Wuerzburg, Am Hub-

land, 97074 Würzburg, Germany

47 Department of Wood and Forest Science, Laval University, Quebec, Quebec, G1V 0A6, Canada

48 School of Biosciences, University of Nottingham, Sutton Bonington Campus, Nr Loughborough,

Leicestershire, LE12 5RD, UK

49 School of Biological Sciences, University of East Anglia, Norwich, Norfolk, NR4 7TJ, UK

50 Forest Research Institute, University of the Sunshine Coast, Sippy Downs, Queensland, Australia

51 Environmental Change Institute, School of Geography and the Environment, University of Oxford,

Oxford OX1 3QY, UK

52 Field Programmes Department, Durrell Wildlife Conservation Trust, Les Augres Manor La Profonde

Rue Trinity Jersey JE3 5BP

53 School of Biosciences, The University of Sheffield, Western Bank, Sheffield, S10 2TN, UK

54 University of East Anglia, School of Environmental Sciences, Norwich Research Park, Norwich, Nor-

folk, NR4 7TJ, UK

55 Evolutionary Biology Group, Faculty of Biology, Adam Mickiewicz University, Poznań, Poland

56 School of Natural and Environmental Sciences, Newcastle University, Newcastle upon Tyne, NE1

7RU, UK

57 Dyson School of Design Engineering, Imperial College London, South Kensington Campus, London

SW7 2AZ, UK

58 Natural Resources and Environmental Studies Institute, University of Northern British Columbia,

Prince George, British Columbia, Canada

59 Centre for Ecology and Conservation, School of Biosciences, University of Exeter, Penryn Campus

TR10 9FE, UK

60 Department of Soil and Crop Sciences, Colorado State University, Fort Collins, CO, USA

61 Department of Life Sciences, School of Life and Environmental Sciences, University of Lincoln, Bray-

ford Pool, Lincoln, LN6 7TS, UK

62 Department of Geography, University of Exeter, Streatham Campus, Exeter, EX4 4QJ, UK

63 Department of Plant Sciences, University of Cambridge, Cambridge, UK

64 Wildlife Biology Program, Division of Biological Sciences, University of Montana, Missoula, MT,

59812, USA

65 WWF-Malaysia, Sabah Office, 88000 Kota Kinabalu, Sabah, Malaysia

66 Conservation & Fisheries Directorate, Ascension Island Government, Georgetown, Ascension Island

67 Asian School of the Environment, Nanyang Technological University, Singapore City, Singapore

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

68 Ecology and Biodiversity Research Group, Departement of Biology, Vrije Universiteit Brussel, Brus-

sels, Belgium

69 Centre for Planetary Health and Food Security, Griffith University, Brisbane, Queensland 4111,

Australia

70 Department of Forest Botany, Dendrology and Geobiocoenology, Faculty of Forestry and Wood

Technology, Mendel University in Brno, Brno, Czech Republic

71 Lee Kong Chian Natural History Museum, National University of Singapore, Singapore

72 New York Cooperative Fish and Wildlife Research Unit, Department of Natural Resources, 321 Fer-

now Hall, Cornell University, Ithaca, NY 14853-3001

73 Fauna & Flora International – Vietnam Programme, 46 Tran Kim Xuyen Lane, Yen Hoa Ward, Cau

Giay, Hanoi, Vietnam

74 Institute of Microbiology, Department of Biology, ETH Zürich, Zürich, Switzerland

75 Department of Biological Sciences, National University of Singapore, Singapore

76 Centre for Biodiversity and Environment Research, Department of Genetics, Evolution and Environ-

ment, University College London, London, UK

77 Malaysian Nature Society, Sabah Branch, JKR 641 Jalan Kelantan, Bukit Persekutuan 50480, Kuala

Lumpur, Malaysia

78 The Nelson Institute for Environmental Studies, University of Wisconsin-Madison, Madison, WI,

USA

79 Royal Botanic Gardens, Kew, Richmond, London, TW9 3AE, UK

80 School of Physiology, Pharmacology and Neuroscience, University of Bristol, Bristol BS8 1QU, UK

81 Faculty of Forestry and Environment, Universiti Putra Malaysia, Seri Kembangan, 43400 Selangor,

Malaysia

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Abstract

The functional stability of ecosystems depends greatly on interspecific differences in responses to

environmental perturbation. However, responses to perturbation are not necessarily invariant

among populations of the same species, so intraspecific variation in responses might also contribute.

Such inter-population response diversity has recently been shown to occur spatially across species

ranges, but we lack estimates of the extent to which individual populations across an entire

community might have perturbation responses that vary through time. We assess this using 524 taxa

that have been repeatedly surveyed for the effects of tropical forest logging at a focal landscape in

Sabah, Malaysia. Just 39 % of taxa – all with non-significant responses to forest degradation – had

invariant responses. All other taxa (61 %) showed significantly different responses to the same forest

degradation gradient across surveys, with 6 % of taxa responding to forest degradation in opposite

directions across multiple surveys. Individual surveys had low power (< 80 %) to determine the

correct direction of response to forest degradation for one-fifth of all taxa. Recurrent rounds of

logging disturbance increased the prevalence of intra-population response diversity, while

uncontrollable environmental variation and/or turnover of intraspecific phenotypes generated

variable responses in at least 44 % of taxa. Our results show that the responses of individual species

to local environmental perturbations are remarkably flexible, likely providing an unrealised boost to

the stability of disturbed habitats such as logged tropical forests.

Introduction

Species differ in their traits, with the implication that they will respond differently to the same

environmental perturbation (1, 2). This interspecific “response diversity” has been identified as a key

determinant of community and ecosystem stability for several decades (3, 4). Yet there is newly

emerging evidence that perturbation responses within species are surprisingly variable, because any

given species might respond differently to the same perturbation depending on where it experiences

that perturbation (5): the population-level responses of individual species vary according to position

within their geographic range (6), climate envelope (7) and macroclimatic conditions (8).

Consequently, the “response traits” that purportedly define how a given species will respond to

environmental change (9) may not have a fixed relationship with species’ morphological traits (10),

despite this being widely assumed in many analyses (2, 11, 12). What we don’t yet know, however, is

whether the population-level responses at a single location are fixed and invariant through time, or

whether those responses might also vary. Such variation might arise through population-level

turnover in the phenotypes of individuals, and in response to the ever-changing environmental

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

conditions in which local ecosystems are embedded. If it exists, any such local, intra-population

response diversity through time might be expected to boost the stability of disturbed ecosystems (3,

13).

Here, we directly quantified the degree of intra-population response diversity by comparing taxon-

specific occurrence patterns described in multiple surveys that were collected within a single

landscape: the SAFE Project in Sabah, Malaysia (14). Data were collected along a forest degradation

gradient defined by variation in logging intensity, which we quantify as the percentage of biomass

reduction. Logging in tropical rainforests is often a recurrent perturbation, and our sites are no

different. Sites were logged between zero and four times between 1970 and 2008 (15), and many

underwent an additional round of salvage logging between 2012 and 2015. Sites at SAFE therefore

extend across all levels of biomass reduction, encompassing primary forest with no (0 %) biomass

removal, areas of light and moderate selective harvesting of trees, through to salvage logged and

clear-felled sites with virtually all (100 %) above-ground biomass removed. Habitat degradation

gradients of this magnitude should generate strong, predictable impacts on the occurrence patterns

of individual taxa, and as such it represents a good system in which to quantify the degree of intra-

population response diversity.

Our analysis encompassed 119 single-year surveys that included at least one taxon that was also

sampled in at least one other survey. Individual surveys – including those conducted on the same

taxon – varied in one or more dimensions of survey method, sample sites and year (Table S1),

reflecting variation in the study design process of individual researchers. While all surveys were

conducted within a single year, we stress that our definition of single-year survey does not imply

that each survey conducted just a single site visit. Sixty-six of the 119 surveys (55 %) conducted

repeat site visits within the year, meaning the researchers had sampled more intensively than a

snapshot survey in which a site is visited just a single time. Data from multiple site visits within a

year were aggregated to represent a single survey for analysis.

There were 1,258 taxa observed in two or more of these spatially overlapping, single-year surveys, of

which 524 had high enough occurrences to be modelled in multiple surveys (n ≥ 5 occurrences in

each of ≥ 2 surveys). Sensitivity analysis demonstrated that this choice of occurrence threshold

ensured the results and conclusions we present are conservative estimates of response diversity (SI

Appendix; Fig. S1). The 524 taxa included 122 plants, 205 invertebrates, 17 fish, 2 reptiles, 17

amphibians, 100 birds and 61 mammals. We focus our analysis on patterns of taxon occurrence.

Taxon occurrence is a simple, commonly employed analysis of biodiversity patterns that should be

more robust to among-year variation in population sizes than analyses based on abundance (16, 17),

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

and therefore represents a conservative test of response diversity. We fitted binomial general linear

models to presence-absence data for 1,942 taxa × survey combinations, from which we recorded

two metrics: (1) the statistical significance of the occurrence pattern, categorised into significant (p <

0.05) or non-significant (p ≥ 0.05) (sensitivity analysis demonstrated little impact of the choice of p-

value on our conclusions; SI Appendix and Fig. S1); and (2) the slope and intercept of the occurrence

pattern. These two metrics can be combined to define four types of response diversity in empirical

data (18) (Fig. 1): (A) invariant, where responses are statistically indistinguishable; (B) magnitude,

where all responses are statistically significant and have a common direction, but have variable slope

estimates; (C) uncertainty, where some responses are statistically significant but others are not; and

(D) sign changes, where statistically significant responses occur but in opposite directions.

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Figure 1: Examples of intra-population response diversity to a gradient of forest degradation. In all

panels, forest degradation is represented as a percentage reduction in aboveground biomass, where

zero represents the median biomass in unlogged forest. Each line displays a fitted model generated

to empirical data collected from a different survey of that particular taxon. Shaded polygons

represent the 95 % confidence intervals. Statistically non-significant relationships are displayed as an

intercept-only fitted model. (A) Invariant: all observed responses of that taxon to forest degradation

in different surveys are statistically indistinguishable (e.g. the tree genus Vitex). (B) Magnitude:

responses of that taxon are all statistically significant and have the same direction of effect, but have

slope and/or intercept estimates that differ (e.g. Bornean yellow muntjac Muntiacus atherodes). (C)

Uncertainty: responses of that taxon vary in their statistical significance (e.g. Yellow breasted

flowerpecker Prionochilus maculatus). (D) Sign: responses of that taxon are statistically significant,

but have response patterns in opposing directions (e.g. dung beetle Onthophagus mulleri).

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

High intra-population response diversity of tropical forest taxa

Almost two-thirds (61 %) of all taxa exhibited response diversity, and our consistently conservative

approach to the analysis means this estimate represents a lower bound (Fig. S1). We found 154 taxa

(29 %) with magnitude response diversity, having response patterns that were consistent in terms of

statistical significance and directions of effect, but where the slope or intercept of the observed

effect varied significantly among surveys (Fig. 1B). Most commonly, response diversity was in the

form of uncertainty, (n = 254, 48 %; Fig. 2A), with some surveys finding non-significant response

patterns but others finding statistically significant trends (Fig. 1C). This proportion is higher than

would be expected by chance (12-41 %, SI Appendix), indicating this result is not a spurious one

emerging from the combination of Type I (false positive) and Type II (false negative) sampling errors

across multiple surveys. Finally, 6 % of taxa (n = 32) displayed the most extreme form of intra-

population response diversity – sign diversity – where repeated surveys detected statistically

significant response patterns in opposing directions (Fig. 1D). The latter three classes are not

mutually exclusive, and 23 % of taxa (n = 122) exhibited multiple forms of response diversity (Fig.

2A). For example, a taxon with three surveys might have one with a non-significant result and two

with statistically significant responses in opposite directions, demonstrating both uncertainty and

sign class. Our results provide quantitative insight into the intra-population response diversity of

individual taxa to an environmental gradient at a single location, and indicate that more than one

half of tropical forest taxa might demonstrate remarkably flexible responses to habitat degradation.

Less than half of all taxa (39 %; n = 206) had invariant responses to the forest degradation gradient

(Fig. 2A), in which all surveys gave results that were indistinguishable in terms of their statistical

significance, direction and magnitude of effect (Fig. 1A). The proportion of taxa with fully invariant

results varied across broad taxonomic groupings, varying from zero in fish to more than half for

plants (Fig. 2B). However, all of the 206 taxa (100 %) with invariant response patterns also had no

significant response to the forest degradation gradient in any survey. By contrast, there were 318

taxa that exhibited a significant response in at least one survey, and not one of those (0 %)

responded in a fully invariant manner in all surveys, suggesting intra-population response diversity is

the norm among taxa with meaningful habitat preferences.

Some of the intra-population response diversity we observed can be ascribed to life history

characteristics of the taxa. The number of taxa exhibiting each of the four response diversity classes

differed from a null expectation for all taxonomic groups (Fig. S2;  Goodness of fit test, 

 > 15.9,

p < 0.015). Plants were more likely to have invariant responses than expected by chance, reflecting

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

the fact that trees are long-lived and stationary organisms. Repeated surveys of vegetation plots are

always likely to detect the same individuals and thereby generate invariant response patterns.

However, different studies with alternative spatial designs, or that examine different life history

stages such as adults versus seedlings, still generated variable results. Mammals, by contrast, were

more likely to exhibit response diversity in the form of both the magnitude and the sign classes. This

pattern probably reflects the highly mobile nature of large mammals, which might allow them to

rapidly re-distribute in response to continuously changing local conditions, including the spatial and

temporal patchiness of fruiting events.

Figure 2. Broad taxonomic patterns in the replicability of single-year biodiversity surveys. (A) Number

of taxa exhibiting each of the four classes of intra-population response diversity (see Fig. 1 caption

for definitions). Classes are not all mutually exclusive, which is displayed with partially overlapping

bars. (B) Bootstrapped estimates of the proportion of taxa with invariant, statistically

indistinguishable response patterns across multiple surveys. Thick line represents the median, boxes

the 1st and 3rd quantiles respectively, and whiskers the range.

Causes of intra-population response diversity

Intra-population response diversity can arise through spurious or ecologically meaningful

mechanisms. First, variation in study design and field methods might generate spurious differences

in observed response patterns that have nothing to do with the ecology of the species themselves.

By contrast, ecologically meaningful response diversity might arise in one of two ways. First,

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

variation in the phenotype of individuals can influence how they respond to disturbances. Any

temporal turnover in the phenotypic composition of populations, which could be driven by selection

forces linked to the forest degradation or varying environmental conditions, might contribute to the

intra-population response diversity we have observed. We have no repeated morphological or

genomic measurements that would allow us to detect or quantify this effect. And second,

meaningful response diversity that might arise from uncontrollable and unmeasured environmental

variation (18). Specific examples might include human disturbance such as hunting intensity and

logging activity, and year-to-year variation in animal population sizes and movements driven by

phenological events such as fruiting, or by climatic variation and extremes such as El Niño

oscillations. When comparing the occurrence patterns from two surveys, we can only definitively

rule out spurious response diversity, and by elimination confirm the presence of meaningful

response diversity, if a pair of surveys used identical sampling methods in identical sampling sites. In

these cases, any inconsistency in the taxon-specific response patterns between surveys must be due

to either unaccounted for environmental variation or phenotypic turnover.

We confirmed the presence of meaningful response diversity in 44 % of taxa (174 out of 397 that

had overlapping methods and sampling sites). A roughly equal number of taxa (n = 168, 42 %) had

variable response patterns that could not be definitively confirmed as being meaningful due to

variations in sampling sites and/or methods, and we therefore consider them to be confirmed

examples of spurious response diversity. Both estimates should be considered to represent the

lower bound of the actual values, however, as any survey pair could simultaneously exhibit both

meaningful and spurious response diversity and there is no statistical method that can quantify this

overlap.

We investigated three potential hypotheses that might explain the presence of meaningful intra-

population response diversity in our data. There was no effect of the number of years between

surveys on the probability of two surveys giving invariant results (Fig S3A; binomial GLM: 󰇛󰇜

 = 1.05,

p = 0.30), nor was there a discernible effect of El Niño events (Fig. S3B; 󰇛󰇜

 = 0.84, p = 0.66), despite

the latter having impacted forest growth patterns both through time and across space at our study

site (19). We did, however, find that an extended logging event that occurred in the middle of our

decade of observations, and that further reduced the biomass across large portions of the SAFE

Project study area, influenced the pattern of survey results (Fig. S3C; 󰇛󰇜

 = 37.4, p < 0.001). We

found increased intra-population response diversity for survey pairs occurring within the logging

events relative to survey pairs occurring in non-disturbed years, indicating that species responses are

more variable during extreme land use change events. This increased response diversity did not

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

occur equally across taxonomic groups (logging x taxonomic group interaction effect: 󰇛󰇜

 = 41.5, p

< 0.001), but was instead driven by a reduction in the proportion of invariant responses of bird and

mammal taxa. These taxa both have relatively high mobility, so their increased response diversity

may reflect a tendency to move away from active disturbances.

Spurious response diversity, meanwhile, arose from variation in the methodology and fine details of

individual studies, with the exact effect varying almost species-by-species (SI Appendix). One

possible implication is that variation in the way sample sites are distributed along an environmental

gradient may exert considerable, undetected influence over the outcome of a biodiversity survey.

Even unconscious bias in the choices of individual researchers about exactly where in the vicinity of a

planned sample point to set a quadrat or place a trap could be affecting the outcome of surveys (20).

Generating reliable results from ecological surveys

Much of ecology and conservation relies on single-year and snapshot surveys. Close to half (44 %) of

the studies published in the journal Ecology present results based on data from a single year, and

many of the world’s largest conservation NGOs rely on rapid biodiversity assessments to prioritise

their actions (21). Drawing general inference from single-year studies on taxa with intra-population

response diversity could easily lead to misleading conclusions about biodiversity patterns (17), and

these could, in turn, lead to poor management and conservation decision-making.

We estimate that any given taxon needs to be analysed in each of three surveys to gain reliable

insight into the impacts of forest degradation (Fig. 3A). Of the four response diversity classes, getting

the direction (sign) of an effect wrong is the most immediately problematic: it could lead directly to

management decisions that are the exact opposite of what is needed. We therefore used Bayesian

hierarchical models to estimate the observed variation within and among taxon-specific surveys, and

so determine the probability that analysis of a single-year survey will return the correct sign. This

probability was less than 80 % – the assumed power of standard statistical tests – for one fifth of all

taxa (n = 109; 21 %), but varied significantly among taxonomic groups (Fig. 3B, Kruskal-Wallis: 󰇛󰇜

 =

78.9, p < 0.001). Birds were most likely to return correct signs from a single survey, while

invertebrates were the least likely. On average, three surveys were needed to ensure a 90 %

probability of getting the right direction of effect for 90 % of taxa (Fig. 3).

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Figure 3. The probability of an analysis returning the correct sign (direction) of a taxon’s response to

forest degradation as a function of the number of single-year surveys in which that taxon was

detected. (A) Each grey line represents an analysis for one taxon (n = 494). (B) Violin plots showing

the distribution of probabilities of a single-year survey returning the correct sign for seven taxonomic

groups. Points indicate the probability for individual taxa within each group.

Implications for ecology and conservation management

Our analyses raise the uncomfortable possibility that three out of every five taxon-specific results

published in single-year studies may be unreliable representations of biodiversity patterns, whether

that be through spurious methodological issues or due to meaningful intra-population response

diversity. Yet our results should not be a complete surprise and do not appear to be unique to our

study site. Inter-annual variation in the spatial distribution of species has been reported for taxa as

diverse as birds in Australia (17), stream invertebrates in Finland (22) and plants in China (23).

Similarly, it is becoming apparent that data spanning a decade or more are needed to obtain

consistent results in ecological field experiments (24, 25), and time lags in the responses of species

to environmental impacts (26) can mean the results of short-term studies can generate

unrepresentative results.

Given such high levels of intra-population variation in response patterns, how can we generate

conclusive insights in ecology and conservation? Intra-population response diversity means clear,

definitive answers to biodiversity and conservation questions cannot be obtained from short funding

cycles. Grants to collect new field data need to be awarded for longer durations, the continuation of

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

multi-taxa biodiversity time series along environmental gradients should be prioritised, and studies

specifically targeted at repeating earlier work should be supported. Until then, the most efficient

way forward will likely be to find new ways that reliably transfer results among study sites. Early

indications are that such transferability may be low (5), but for predictable reasons including the

location of a site relative to species’ geographic ranges (6) and climatic tolerances (7). Clearer

understanding of how these macroecological patterns influence site- and species- specific

biodiversity patterns will be key to developing new predictive frameworks to extrapolate findings

from heavily studied sites to regions lacking equivalent data (5). Such frameworks will provide a

means to maximise the utility of the data that do exist, help contextualise the generality of

ecological conclusions arising from single-year studies, and give the empirical evidence needed to

support urgent decision making at national and regional scales.

The functional stability of ecosystems can be promoted through the diversity of relationships

between species and the environment (3), with communities containing suites of species that vary in

their responses to environmental gradients better able to stabilise ecosystem properties than

communities lacking equivalent response diversity (4). Temporal or spatial variation in the responses

of individual populations is similarly expected to contribute to the stability of ecosystems (27), but

that contribution is unlikely to be equal among the different classes of response variability: we

expect variation in the sign of a response will make the strongest contribution to stability, followed

by the uncertainty class and with variable response magnitudes having the smallest impact. Our data

demonstrate intra-population response diversity may be the norm rather than the exception, and

that year-to-year variation in ecological context might mediate, or even reverse, species responses

to environmental gradients. This flexibility in the responses of individual species to environmental

changes might ensure human-modified environments, like logged and degraded tropical rainforests,

are more stable and more resilient ecosystems than expected.

Acknowledgements

The SAFE Project was supported by the Sime Darby Foundation. Site access and research permits

were provided by the Maliau Basin Management Committee, Sabah Foundation, Benta Wawasan,

Sabah Softwoods, Innoprise Foundation, Sabah Forestry Department, and the Sabah Biodiversity

Centre. RME is supported by the NOMIS Foundation. Data collection was funded by Australian

Research Council grant DP140101541; Bat Conservation International; British Council Newton-Ungku

Omar Fund; British Ecological Society grant 3256/4035; Cambridge University Commonwealth Fund;

Cambridge Trust; Imperial College London SSCP DTP grant NE/L002515/1; Jardine Foundation;

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Malaysia Industry Group for High Technology; Panton Trust; Ministry of Education, Youth and Sports

of the Czech Republic grant INTER-TRANSFER LTT19018; Primate Society of Great Britain; ProForest;

Royal Society of London grant RG130793; Sime Darby Foundation; S.T. Lee Fund; Tim Whitmore

Fund; University of Kent; Universiti Malaysia Sabah; UK Natural Environment Research Council grants

NE/K016253/1, NE/K016407/1, NE/K016148/1, NE/K0106261/1, NE/K015377/1, NE/L002582/1,

NE/P00363X/1, and studentship 1122589; UK Research and Innovation; Universiti Malaysia Sabah,

University of Florida Institute of Food and Agricultural Sciences; Varley Gradwell Travelling

Fellowship; World Wildlife Fund for Nature. Data collection was supported by Saloni Barsrur, Susan

Benedick, Victoria Bignet, Stephen Brooks, Keiron Brown, Stephen Butler, Daniel Carpenter, Kristina

Graves, Herry Heroin, Alex Kendall, Darren Mann, Sol Milne, John Mumford, Derek Shapiro, Kathryn

Sieving, John Sugau, Elizabeth Telford, Bradley Udell and Bakhtiar Effendy Yahya.

Author contributions

RME designed the study, conducted the analyses and drafted the manuscript. WDP, CDLO, PA, TdL,

GR and CBL supported the data analysis, helped interpret the results and edited the manuscript. All

other authors contributed field data and checked the manuscript.

Materials and methods

Data analysis and construction of figures were conducted in the R v4.02 computing environment

(28), using the packages arm (29), dplyr (30), lme4 (31), MuMIn (32), paletteer (33), rstanarm (34,

35), safedata (36) and scales (37).

Taxa occurrence data

We compiled taxa records from 47 data sources, of which 45 were published on the SAFE Project

data repository (38) and the remaining 2 were presented in published papers (39, 40) (Table S1).

Data were collected in the eleven-year period 2010 to 2020. We accepted both presence-absence

and abundance data for the analysis, but restricted it to records where the sampling locations had

known geographic coordinates. Where data sources contained data generated by multiple sampling

methods, they were split to consider each method as a separate survey. Similarly, data sources

comprising samples collected from multiple years were also split to consider each year as a different

survey.

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Only identified taxa were analysed, although not all taxa were identified to species. We rejected any

taxa identified to less than ordinal level. Morphospecies represent a particularly difficult challenge

for analyses of replicability, because the within-survey codes used to identify them are not

consistent among surveys. This means taxa cannot be accurately matched and therefore compared

among surveys. To surmount this challenge, we aggregated morphospecies by genus within

individual surveys, which allowed us to match taxa among surveys. Genus is a commonly used level

of taxonomic resolution used in tropical analyses of diverse taxa such as trees (41) and ants (42), and

taxa aggregated in this way accounted for just 6 % (n = 32) of all taxa in our analysis.

The median number of surveys per taxon was 3.0 (range: 2-15), and the median number of pairwise

comparisons per taxon was also 3.0 (range: 1-105).

Quantifying forest degradation

Taxa at the SAFE Project were collected along a tropical rainforest degradation gradient that runs

from unlogged, old growth forest in strictly protected areas, along a gradient of logging damage

going from low-level, selective extraction of individual trees within water catchments through to

high-intensity, salvage logged forest with no restrictions on tree harvesting, and ends with oil palm

plantation with palms that ranged in age from 5 – 20 years old. We used Aboveground Carbon

Density (ACD, Mg.ha-1) derived from airborne LiDAR data (43, 44) as a base metric from which to

quantify habitat degradation. ACD values ranged from 273 Mg.ha-1 in unlogged forest through to just

1 Mg.ha-1 in deforested patches. For ease of interpretation, we converted these to a percentage

reduction in biomass density relative to the median biomass density observed in unlogged forest

(230 Mg.ha-1).

We used maps of above-ground carbon density generated from LiDAR data collected in November

2014 (43) and again in April 2016 (44). These dates approximately bracket a salvage logging round

during which forest quality was greatly reduced across much of the study area. Each survey was

assigned the forest quality metrics that were collected closest in time to the date of the taxa record.

Sampling units varied among the individual surveys, with taxa recorded either at specific point

locations (e.g. insect traps), along transects (e.g. fish censes), or within a polygon sampling area (e.g.

tree plots). We followed Pfeifer et al. (45) by implementing a 1 km buffer area around each sampling

unit and averaging the forest quality metrics over all 1-ha pixels within that buffer. We implemented

a Gaussian distance weighting that weighted pixels by distance from sampling unit, ensuring pixels

located far from the sampling units carried less weight than those immediately adjacent.

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

We opted to restrict all taxa to be analysed in response to forest quality at a single spatial scale to

keep our estimates of response diversity as simple and as conservative as possible. Allowing taxa to

vary in the spatial scale of their response among surveys introduces an additional way in which they

might have a diversity of response patterns, making it more likely that they would be classified as

demonstrating intra-population response diversity. Restricting all taxa to a single spatial scale

therefore represents a conservative estimate of the proportion of taxa exhibiting response diversity.

Quantifying and summarising intra-population response diversity

Occurrence models: To combine such disparate sampling methods across such a wide range of taxa,

we standardised all taxa records to presence-absence data for analysis. We used univariate, binomial

generalised linear models (GLMs) to model taxon occurrence in response to forest degradation, with

models fitted to individual surveys. We only fitted models to taxon × survey combinations where the

taxon had n ≥ 5 presence records within that particular survey. For each taxon x survey combination,

we fitted a model containing a single linear predictor allowing for a sigmoid pattern of occurrence

along the forest quality gradient, and estimated the statistical significance of that model using a log-

likelihood ratio test comparing the fitted model to a null model.

Controlling for detectability in analyses of species occurrence patterns is often recommended, but

such analyses have substantially higher data requirements than typically exist for all taxa within

biodiversity surveys (46). In exploratory analyses we found such models routinely failed to converge.

This is commonly the case for taxa in the tropics where communities have large numbers of

predominantly rare species with low detection probabilities, which are known issues that prevent

detectability-based models from converging (46). Moreover, statistical estimates of tropical species

responses to ecological gradients do not notably vary between models that incorporate or ignore

detection probability (46), so there is no reason to expect the use of detectability-corrected

modelling approaches to influence our main conclusions.

Categorising response diversity: For the 524 taxa that were detected and modelled from ≥ 2 surveys,

we compared model results and grouped taxa according to four classes of response diversity, based

on the four classes of ecological context-dependence described by Catford et al (18).

A. Invariant: Taxa were considered to have invariant, fully replicable occurrence patterns if all

models of that taxa agreed on two criteria:

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

i. the pattern of significance; binarized into significant (p < 0.05) or non-significant (p ≥

0.05); and

ii. the slope and intercept of the fitted model: determined by using the estimate and

accompanying standard error to statistically test for a difference among all pairwise

combinations of models of the same taxa (47).

B. Magnitude: Taxa were considered to vary only in the magnitude of the occurrence pattern if

they:

i. had a consistent pattern of significance, but

ii. had slope and/or intercept estimates that varied significantly among surveys.

C. Uncertainty: Taxa were considered to vary in their statistical certainty if they had an

inconsistent pattern of statistical significance, exhibiting both a statistically significant and

statistically insignificant response in one or more surveys respectively.

D. Sign: Taxa were considered to vary in sign if they had statistically significant response

patterns with opposing signs, meaning in some surveys their occurrence increased with

increasing forest degradation but in others their occurrence decreased.

To compare response diversity among broad taxonomic groupings, we categorised all taxa as

belonging to one of plant, invertebrate, fish, amphibian, reptile, bird or mammal groups. For each

taxonomic group, we took 1,000 bootstrapped samples of the binary responses representing

whether the individual taxa within that group lacked (statistical class 1) or exhibited response

diversity (one or more of statistical classes 2, 3 or 4). These bootstrapped samples were used to

generate distributions describing the proportion of taxa per group with response diversity.

Setting a null expectation for the proportion of taxa with response diversity: Statistical techniques,

such as the binomial GLM we used to classify response patterns as significant or non-significant,

have standard Type I and Type II error rates of 5 and 20 % respectively, meaning the probability of

true detection (statistical power) is 80 %. We used these values to set bounds on the proportion of

taxa that could have statistically inconsistent results (‘Uncertainty’ response diversity class; Fig. 1C)

for spurious reasons. The lower bound is estimated by assuming that all taxa have a “true” response

in which they are not impacted by habitat degradation. In this scenario, the probability of correctly

reporting a non-significant result in one survey (80 %) is multiplied by the probability of incorrectly

detecting a significant result where none exists (Type I error; 5 %), meaning 4 % of the pairwise

comparisons we analyse would be expected to spuriously report response diversity. Most taxa we

examined had > 2 pairwise comparisons, and the probability that at least one pair of surveys for a

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

given taxon returns a spurious result () scales according to the number of pairwise comparisons

(), such that 󰇛󰇜   . The upper bound is estimated by assuming that all taxa have

a “true” response in which they are significantly impacted by the habitat degradation gradient. In

this case, we multiply the probability of one survey accurately detecting the significant pattern

(power; 80 %) by the probability of the second survey failing to detect the significant pattern that

exists (Type II error; 20 %), suggesting 16 % of pairwise comparisons might be expected to spuriously

report response diversity. Scaled up to taxon level, this gives a probability of generating a spurious

result of 󰇛󰇜   .

For each taxon, we estimated 󰇛󰇜 and 󰇛󰇜, and then took the median of each of those

two probabilities across all taxa as a null expectation for the proportion of taxa that would exhibit

uncertainty response diversity from the spurious accumulation of Type I and Type II errors.

Sensitivity analyses

Minimum number of presence records: We used sensitivity analysis to examine the extent to which

our arbitrary threshold of requiring n ≥ 5 presence records per taxon per survey might influence our

key results. We progressively restricted our analysis to subsets of taxa that had at n ≥ 5, 10, 15, …

100 presence records within at least two separate surveys, and for each cut-off value calculated the

proportion of taxa assigned to each of the four response diversity classes.

Critical p-value for statistical significance: Our categorisation of results into uncertainty response

diversity (Fig. 1C) is dependent on the p-value used to denote statistical significance. We tested the

extent to which our results are sensitive to our choice of p = 0.05 by repeating the analysis for values

of 0.005 ≤ p ≤ 0.1 in increments of 0.005.

Taxonomic bias in the response diversity classes

We used  goodness of fit tests to determine whether the distribution of taxa within each of the

four response diversity classes were random samples of the seven taxonomic groups. We used the

proportion of taxa within the full dataset belonging to each of the seven groups as an expected

distribution, and assessed the probability that the observed distribution deviated from this. We then

used bootstrapping to determine which taxonomic groups were under- or over-represented in each

of the four classes. For each class, we took 1000 random draws of the taxa exhibiting that class

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

(sampling with replacement), from which we quantified an expected null distribution of the number

of taxa belonging to each of the seven taxonomic groups (Fig. S2). Groups where the observed

number of taxa fell below the 2.5 % or above the 97.5 % quantiles of the null distribution were

considered to be significantly under- or over-represented respectively.

Meaningful versus spurious intra-population response diversity

We conducted a secondary analysis to classify intra-population response diversity into meaningful or

spurious causes. Meaningful response diversity can only be definitively demonstrated when

response patterns vary among two surveys that share an exact sampling design and survey method,

in which case the variation in response pattern must arise due to unaccounted variation in the

environment. To identify cases of confirmed, meaningful response diversity for individual taxa, we

first identified all pairs of surveys that contained data on that particular taxon. For each survey pair,

we extracted the subset of sampling locations that were present in both surveys and repeated the

fitting of binomial GLMs and the categorisation of GLM results into the four response diversity

classes. Survey pairs that exhibited one or more of statistical classes B-D (Fig. 1) were confirmed as

having meaningful response diversity. This represents a lower bound on the true number of

meaningful response diversity cases, as surveys with different spatial designs could also exhibit

meaningful response diversity but we are unable to test for it. Taxa that were observed to have

response diversity in our main analysis, but for which we could not positively determine the

presence of meaningful response diversity in this secondary analysis, were considered to

demonstrate spurious response diversity.

We examined three potential causes of meaningful response diversity at our study site: a general

year effect; El Niño events that occurred in 2010 and again in 2015/16; and a salvage logging

operation that impacted large areas of the landscape over the years 2012-2015. The first of these

was tested by quantifying, for each year, the proportion of pairwise comparisons involving that year

that exhibited response diversity (Fig. S3A). We quantified this proportion for each of the years for

which we had survey data (2010 – 2020 inclusive), and tested for an effect of the number of years

separating the two surveys on the whether they had invariant or different response patterns. This

was tested using using a binomial GLMM, including taxon identity as a random effect. To test the

remaining two hypotheses, we divided pairwise survey comparisons into three groups for each of

the two events: (1) outside: when each of the individual surveys occurred outside of an El Niño or

logging event respectively; (2) straddling: when one of the surveys occurred within an El Niño or

logging event and the other occurred outside of such events; and (3) within: when both surveys

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

occurred within an El Niño or logging event. For each group, we quantified the proportion of

pairwise comparisons that exhibited invariant responses (Fig S3B,C). If either of these events were a

strong driver of meaningful response diversity, we would expect to see a higher proportion of taxa

exhibiting response diversity in survey pairs either straddling or embedded within these disturbance

events.

To explore the potential drivers of spurious response diversity, we used generalised linear mixed

effects models to examine the effect that researcher-controlled decisions about study design exert

on response diversity. We categorised all pairwise combinations of taxon-specific comparisons as

exhibiting response diversity or not, and used that as a response variable with taxon identity as a

random variable. As fixed effect predictor variables we included the minimum number of

occurrences per survey pair (log10-transformed), the minimum number of sample sites per survey

pair (log10-transformed), the taxonomic resolution at which that taxa had been identified (species,

genus, family or ordinal level), and whether the pair of surveys used the same or different sampling

methods. We used minimums rather than the mean or maximum because analyses based on low

sample size are more prone to generating spurious and imprecise results, and therefore a pair of

studies where one or both have a particularly low sample size are more likely to return inconsistent

results. We fitted all variables in a single model and used backwards stepwise model selection using

log-likelihood ratio tests to identify significant variables. We calculated the marginal and conditional

coefficient of determination (pseudo-r2) for the fixed and random effects respectively using the

method outlined by Nakagawa and Schielzeth (48).

Single-survey data in ecology

We reviewed all papers published in Ecology, the flagship journal of the Ecological Society of

America, in 2021, recording the number of years of empirical data presented in each publication.

Out of a total of 263 papers, 246 presented empirical data, of which 100 (41 %) presented data

collected from a single-year survey only. A further nine studies (4 %) presented data from multiple

datasets, of which one or more was a single-year survey.

Estimating the probability of detecting a true response

To estimate the probability that we would correctly detect the correct sign of a taxon’s response to

the forest degradation gradient, we re-fitted all models using Bayesian hierarchical models in

rstanarm (34, 35), fitting each taxon with hierarchically-drawn slopes of biomass and intercepts. We

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

used package defaults for model sampling (1000 warm-up iterations and 1000 subsequent sampling

iterations across each of 4 chains) with default priors apart from where specified below, and found

no evidence of model mis-specification or a lack of convergence (i.e., all  values <1.1 and no

divergent transitions in the sampling steps). Specifically, our model was defined as:

  󰇡󰇢 (1)

      (2)

 󰇛󰇜 (3)

 󰇛󰇜 (4)

where  is the presence/absence (1/0) of a taxon in a given survey () at a given site (), and  is

logit-transformed and is proportional to the predicted probability of presence of a species in a given

survey (i.e., is a standard Binomial Generalised Linear Model term).  itself is then defined by ,

the overall mean across all years, , the contrasts (difference) in the intercept for each survey year,

and , each survey year’s estimated biomass removal effect (itself multiplied by , the biomass

removal at a given site). The terms  and  are themselves hierarchically drawn from distributions

centred at 0 and  and with standard deviations  and , respectively. The Bayesian hierarchical

formulation of our approach is central to our method since, for each taxon, it allows us to estimate

variation in responses to biomass removal across surveys and also to directly parameterise the

distribution of estimated responses in equation 3. From this, we can directly estimate the probability

of observing a consistent response across years.

To ensure that our prior specifications were not biasing model results, we repeated our model fits

across two extreme prior definitions: ‘correlated’ and ‘variable’ priors. These were fitted across 4

chains, each with 3000 warm-up iterations and 1000 sampling iterations (more samples were

needed due to the priors being extreme and therefore slowing model convergence). ‘Correlated’

priors specified regularisation parameters () of 0.5 for the hierarchical terms, biasing the model

such that slopes and intercepts should be more consistent across surveys. ‘Variable’ priors set  = 5,

such that survey slopes and intercepts were more independent. These two specifications form

extremes of a continuum. Our model results were qualitatively identical to those reported in the

main text (with  set to 1, the default), suggesting that our conclusions are robust to model

specification and fitting method.

SI Appendix

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Sensitivity analyses

Minimum number of presence records: Our results are largely robust to the choice of cut-off value of

requiring n ≥ 5 presence records per taxon per survey (Fig. S1A), and our specific cut-off value of n ≥

5 resulted in the highest proportion of taxa lacking response diversity (having fully invariant

responses). This ensures our choice of n ≥ 5 ensures we present results that emphasise the most

conservative estimate of response diversity.

Critical p-value for statistical significance: Using values of p that were lower than 0.05 resulted in a

slight reduction in the proportion of taxa demonstrating uncertainty response diversity and a

corresponding increase in the proportion of taxa lacking response diversity (Fig. S1B). This effect was

most apparent at highly conservative estimates of statistical significance (p ≤ 0.02), above which the

choice of p exerted little influence on our qualitative conclusions.

Figure S1. Sensitivity of results to variation in (A) the minimum number of occurrence records

required for a taxon to be analysed; and (B) the critical p-value used to denote statistical significance.

Values represent the proportion of taxa categorised into each of the four classes of response diversity

(Fig. 1).

Taxonomic bias in the response diversity classes

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Figure S2. Expected and observed number of taxa within seven broad taxonomic groups exhibiting

the four classes of response diversity. Distributions represent, for each taxonomic group and class, a

null expectation of the number of taxa that will exhibit that particular class. Points represent the

observed number of taxa displaying each class. Significance denotes whether the point sits outside

the 95 % quantile of the null distribution.

Spurious response diversity

Spurious response diversity is likely driven by methodological differences among surveys of the same

taxon in the same landscape, although it’s not clear from our data exactly what researchers should

do to eliminate this. The probability that a taxon-specific, pairwise comparison of responses were

invariant increased as taxonomic resolution increased ( = 24.1, df = 3, p < 0.001), and there was a

paradoxical effect where taxa that were more common were less likely to demonstrate invariant

responses ( = 15.2, df = 1, p < 0.001). This likely occurs because rare taxa were less likely to have a

significant response to forest degradation (beta regression; z = -17.1, p < 0.001), and we only found

fully invariant responses within generalist taxa that were not impacted by the degradation gradient

(e.g. Fig. 1A). Increasing our threshold of n ≥ 5 for inclusion in the analysis further reduced the

proportion of taxa with invariant responses (Fig. S1), demonstrating that excluding uncommon

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

species from the analysis would generate even more extreme estimates of response diversity. The

probability of responses varying was also increased when pairs of surveys used different sampling

methods ( = 5.25, df = 1, p = 0.022), but was not impacted by the minimum sample size of a pair of

surveys ( = 1.67, df = 1, p = 0.20). Together, these fixed effects explained just 7 % of the variation

in probability of obtaining two invariant surveys, meaning spurious response diversity was not

strongly driven by variation in general study design parameters that researchers can exert direct

control over. By contrast, the random effect of taxon identity explained 66 % of the variation,

indicating the cause of spurious response diversity is almost species-specific. These results indicate

that the spurious response diversity we have detected is not obviously generated by high-level

design features of studies, and instead appears to be caused by the fine details of individual studies.

Figure S3. Bootstrapped estimates of the proportion of taxa with invariant response patterns with

respect to (A) year of survey, (B) El Niño events and (C) logging events. Thick line represents the

median, boxes the 1st and 3rd quantiles, and whiskers the range. In panels (B) and (C), data were

categorised into pairwise comparisons of taxon responses to forest degradation in which both

surveys were conducted in years during which the event occurred (within), when one survey occurred

during an event and the other occurred outside of the event (straddling), and when both surveys

occurred outside of the event (outside).

References

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

1. N. M. Williams et al., Ecological and life-history traits predict bee species responses to

environmental disturbances. Biol. Conserv. 143, 2280-2291 (2010).

2. J. P. Suraci et al., Disturbance type and species life history predict mammal responses to

humans. Global Change Biol. 27, 3718-3731 (2021).

3. S. R. P.-J. Ross, O. L. Petchey, T. Sasaki, D. W. Armitage, How to measure response diversity.

Methods Ecol Evol 14, 1150-1167 (2023).

4. D. Tilman, J. A. Downing, Biodiversity and stability in grasslands. Nature 367, 363-365 (1994).

5. C. Banks-Leite, M. G. Betts, R. M. Ewers, C. D. L. Orme, A. L. Pigot, The macroecology of

landscape ecology. Trends Ecol. Evol. 37, 480-487 (2022).

6. C. D. L. Orme et al., Distance to range edge determines sensitivity to deforestation. Nat. Ecol.

Evol. 3, 886-891 (2019).

7. J. J. Williams, T. Newbold, Vertebrate responses to human land use are influenced by their

proximity to climatic tolerance limits. Divers. Dist. 10.1111/ddi.13282 (2021).

8. M. Tourani et al., Maximum temperatures determine the habitat affiliations of North

American mammals. Proc. Natl. Acad. Sci. U.S.A. 120, e2304411120 (2023).

9. K. N. Suding et al., Scaling environmental change through the community-level: a trait-based

response-and-effect framework for plants. Global Change Biol. 14, 1-16 (2008).

10. J. H. Hatfield, C. D. L. Orme, J. A. Tobias, C. Banks-Leite, Trait-based indicators of bird species

sensitivity to habitat loss are effective within but not across data sets. Ecol. Appl. 28, 28-34

(2018).

11. M. Pacifici et al., Species’ traits influenced their response to recent climate change. Nature

Climate Change 7, 205-208 (2017).

12. G. S. Kandlikar, A. R. Kleinhesselink, N. J. B. Kraft, Functional traits predict species responses

to environmental variation in a California grassland annual plant community. J. Ecol. 110,

833-844 (2022).

13. J. Leps, F. Bello, P. Šmilauer, J. Doležal, Community trait response to environment:

Disentangling species turnover vs intraspecific trait variability effects. Ecography 34, 856-863

(2011).

14. R. M. Ewers et al., A large-scale forest fragmentation experiment: the Stability of Altered

Forest Ecosystems Project. Phil. Trans. Roy. Soc. B 366, 3292-3302 (2011).

15. T. Riutta et al., Logging disturbance shifts net primary productivity and its allocation in

Bornean tropical forests. Global Change Biol. 24, 2913-2928 (2018).

16. L. N. Joseph, S. A. Field, S. Wilcox, H. P. Possingham, Presence–absence versus abundance

data for monitoring threatened species. Conserv. Biol. 20, 1679-1687 (2006).

17. M. Maron, A. Lill, D. M. Watson, R. Mac Nally, Temporal variation in bird assemblages: how

representative is a one-year snapshot? Austral Ecol. 30, 383-394 (2005).

18. J. A. Catford, J. R. U. Wilson, P. Pyšek, P. E. Hulme, R. P. Duncan, Addressing context

dependence in ecology. Trends Ecol. Evol. 37, 158-170 (2022).

19. M. H. Nunes et al., Recovery of logged forest fragments in a human-modified tropical

landscape during the 2015-16 El Niño. Nat. Comm. 12, 1526 (2021).

20. Z. Mehrabi, E. M. Slade, A. Solis, D. J. Mann, The importance of microhabitat for biodiversity

sampling. PLoS ONE 9, e114015 (2014).

21. C. Stem, R. Margoluis, N. Salafsky, A. Brown, Monitoring and evaluation in conservation: a

review of trends and approaches. Conserv. Biol. 19 (2005).

22. R. Sarremejane et al., Climate-driven hydrological variability determines inter-annual

changes in stream invertebrate community assembly. Oikos 127, 1586-1595 (2018).

23. H. Ganjurjav et al., Temperature leads to annual changes of plant community composition in

alpine grasslands on the Qinghai-Tibetan Plateau. Environ. Monit. Assess, 190, 585 (2018).

24. S. Cusser, C. Bahlai, S. M. Swinton, G. P. Robertson, N. M. Haddad, Long-term research

avoids spurious and misleading trends in sustainability attributes of no-till. Global Change

Biol. 26, 3715-3725 (2020).

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

25. D. G. T. Naidu, S. Roy, S. Bagchi, Loss of grazing by large mammalian herbivores can

destabilize the soil carbon pool. Proc. Natl. Acad. Sci. U.S.A. 119, e2211317119 (2022).

26. N. M. Haddad et al., Habitat fragmentation and its lasting impact on Earth’s ecosystems. Sci.

Adv. 1, e1500052 (2015).

27. A. E. Noto, T. C. Gouhier, The effects of intraspecific and interspecific diversity on food web

stability. Theor. Ecol. 13, 399-407 (2020).

28. R Core Team (2020) R: A language and environment for statistical computing. (R Foundation

for Statistical Computing, Vienna, Austria).

29. A. Gelman, S. Yu-Sung (2021) arm: Data analysis using regression and multilevel/hierarchical

models. R package version 1.12-2.

30. H. Wickham, R. Francois, L. Henry, K. Muller (2021) dplyr: A grammar of data manipulation.

31. D. Bates, M. Maechler, B. Bolker (2011) lme4: Linear mixed-effects models using S4 classes.

R package version 0.999375-39.

32. K. Barton (2020) MuMIn: Multi-Model Inference.

33. E. Hvitfeldt (2021) paletteer: Comprehensive collection of color palettes.

34. B. Goodrich, J. Gabry, I. Ali, S. L. Brilleman (2022) rstanarm: Bayesian applied regression

modeling via Stan. R package version 2.21.3.

35. Stan Development Team (2021) RStan: the R interface to Stan. R package version 2.21.3.

36. A. Aldersley, C. D. L. Orme (2019) safedata: Interface to Data from the SAFE Project.

37. H. Wickham, D. P. Seidel (2020) scales: Scale functions for visualization.

38. C. D. L. Orme, R. M. Ewers (2018) The SAFE Project Community.

39. O. R. Wearn, C. Carbone, J. M. Rowcliffe, H. Bernard, R. M. Ewers, Grain-dependent

responses of mammalian diversity to land use and the implications for conservation set-

aside. Ecol. Appl. 26, 1409-1420 (2016).

40. S. L. Mitchell et al., Spatial replication and habitat context matters for assessments of

tropical biodiversity using acoustic indices. Ecol. Indic. 119, 106717 (2020).

41. W. F. Laurance et al., Rapid decay of tree-community composition in Amazonian forest

fragments. Proc. Natl. Acad. Sci. U.S.A. 103, 19010-19014 (2006).

42. M. J. W. Boyle et al., Localised climate change defines ant communities in human‐modified

tropical landscapes. Func. Ecol. 35, 1094-1108 (2021).

43. T. Jucker et al., Topography shapes the structure, composition and function of tropical forest

landscapes. Ecol. Lett. 21, 989-1000 (2018).

44. T. Jucker et al., Estimating aboveground carbon density and its uncertainty in Borneo's

structurally complex tropical forests using airborne laser scanning. Biogeosciences 15, 3811-

3830 (2018).

45. M. Pfeifer et al., Creation of forest edges has a global impact on forest vertebrates. Nature

551, 187-191 (2017).

46. C. Banks-Leite et al., Assessing the utility of statistical adjustments for imperfect detection in

tropical conservation science. J. Appl. Ecol. 51, 849-859 (2014).

47. C. C. Clogg, E. Petkova, A. Haritou, Statistical methods for comparing regression coefficients

between models. Am. J. Sociol. 100, 1261-1293 (1995).

48. S. Nakagawa, H. Schielzeth, A general and simple method for obtaining R2 from generalized

linear mixed-effects models. Methods Ecol Evol 4, 133-142 (2013).

49. H. Bernard, K. B. Hee, A. Wong, Importance of riparian reserves and other forest fragments

for small mammal diversity in disturbed and converted forest landscapes.

https://zenodo.org/record/3908128. (Zenodo, 2020).

50. T. Bishop, R. Ewers, Abundance and morphometrics of ant genera.

https://zenodo.org/record/1198839. (Zenodo, 2018).

51. S. Both et al., Functional traits of tree species in old-growth and selectively logged forest.

https://zenodo.org/record/3247631. (Zenodo, 2019).

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

28 
 
52.  H. Brant, J. Mumford, R. Ewers, S. Benedick, Mosquito Data At Safe 2012-2014. 
https://zenodo.org/record/1198846. (Zenodo, 2018). 
53.  D. Carpenter et al., The Maliau Quantitative Inventory. https://zenodo.org/record/5562260. 
(Zenodo, 2021). 
54.  P. M. Chapman, C. Davison, Small mammals at forest-oil palm edges raw datasets. 
https://zenodo.org/record/2579792. (Zenodo, 2019). 
55.  N. J. Deere, Maximizing the value of forest restoration for tropical mammals by detecting 
three-dimensional habitat associations. https://zenodo.org/record/4010757. (Zenodo, 
2020). 
56.  T. Döbert, B. L. Webber, J. B. Sugau, K. J. M. Dickinson, R. K. Didham, Landuse change and 
species invasion. https://zenodo.org/record/2536270. (Zenodo, 2019). 
57.  R. Drinkwater, R. Drinkwater, T. Swinfield, N. J. Deere, Occurrence of blood feeding 
terrestrial leeches in a degraded forest ecosystem. https://zenodo.org/record/3476542. 
(Zenodo, 2019). 
58.  R. M. Ewers, R. Gray, The importance of vertebrates in regulating insect herbivory pressure 
along a gradient of logging intensity in Sabah, Borneo. https://zenodo.org/record/3975973. 
(Zenodo, 2020). 
59.  A. Faruk, Leaf Litter Amphibian Communities. https://zenodo.org/record/1303010. (Zenodo, 
2018). 
60.  T. M. Fayle, K. M. Yusah, R. M. Ewers, M. J. W. Boyle, How does forest conversion and 
fragmentation affect ant communities and the ecosystem processes that they mediate? 
https://zenodo.org/record/3876227. (Zenodo, 2020). 
61.  A. Fraser et al., Amphibian survey of riparian buffer zones at SAFE Project, Borneo. 
https://zenodo.org/record/3973551. (Zenodo, 2020). 
62.  A. Fraser, H. Bernard, E. Mackintosh, R. M. Ewers, C. Banks-Leite, Effects of habitat 
modification on a tritrophic cascade in a lowland tropical rainforest. 
https://zenodo.org/record/3981222. (Zenodo, 2020). 
63.  R. Gray, R. Gill, R. Ewers, The Role Of Competition In Structuring Ant Community 
Composition Across A Tropical Forest Disturbance Gradient. 
https://zenodo.org/record/1198302. (Zenodo, 2018). 
64.  R. Gray, E. Slade, A. Chung, O. Lewis, Riparian_Invertebrate_Movement_Data_SAFE. 
https://zenodo.org/record/3475406. (Zenodo, 2019). 
65.  N. Gregory, R. M. Ewers, L. Cator, A. Chung, Vectorial capacity of Aedes albopictus across an 
environmental gradient. https://zenodo.org/record/3994260. (Zenodo, 2020). 
66.  J. Hardwick et al., The effects of habitat modification on the distribution and feeding ecology 
of Orthoptera 2015. https://zenodo.org/record/4275386. (Zenodo, 2020). 
67.  D. Hemprich-Bennett et al., Impacts of rainforest degradation on the diets of the 
insectivorous bats of Sabah. https://zenodo.org/record/3247465. (Zenodo, 2019). 
68.  S. Heon, P. M. Chapman, O. R. Wearn, H. Berhard, R. M. Ewers, Core SAFE project small 
mammal trapping data. https://zenodo.org/record/3955050. (Zenodo, 2020). 
69.  S. Heon, P. Chapman, H. Bernard, R. M. Ewers, Do Logging Roads Impede Small Mammal 
Movement In Borneo'S Tropical Rainforests? https://zenodo.org/record/1304117. (Zenodo, 
2018). 
70.  E. W. Jebrail, M. Dahwood, A. H. Fikri, B. Yahya, The effects of progressive land use changes 
on the distribution, abundance and behavior of vector mosquitoes in Sabah, Malaysia. 
https://zenodo.org/record/3475408. (Zenodo, 2019). 
71.  A. Kendall, R. M. Ewers, The Effect Of Forest Modification On Ectoparasite Density And 
Diversity. https://zenodo.org/record/1237736. (Zenodo, 2018). 
72.  O. Konopik, Functional diversity of amphibian assemblages along a disturbance gradient. 
https://zenodo.org/record/1995439. (Zenodo, 2018). 
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 
The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint 

29 
 
73.  I. Lane Shaw, R. M. Ewers, Microclimate Change, Forest Disturbance And Twig-Dwelling Ants. 
https://zenodo.org/record/1237732. (Zenodo, 2018). 
74.  S. H. Luke et al., Freshwater invertebrates - diversity and function of stream 
macroinvertebrates: Effects of habitat conversion and strategies for conservation. 
https://zenodo.org/record/5710509. (Zenodo, 2021). 
75.  S. Luke, Ant And Termite Assemblages Along A Tropical Forest Disturbance Gradient In 
Sabah, Malaysia: A Study Of Co-Variation And Trophic Interactions. 
https://zenodo.org/record/1198833. (Zenodo, 2018). 
76.  E. Mackintosh, A. Fraser, C. Banks-Leite, R. M. Ewers, A. Chung, Effect of vertebrate 
exclusion on ecosystem functioning. https://zenodo.org/record/4630980. (Zenodo, 2021). 
77.  K. Mullin et al., Bat activity in riparian reserves in forest and oil palm plantations. 
https://zenodo.org/record/3971012. (Zenodo, 2020). 
78.  C. Noble, Impacts of habitat disturbance on population health of Bornean frogs. 
https://zenodo.org/record/3485086. (Zenodo, 2019). 
79.  A. Pianzin, A. Wong, H. Bernard, M. Struebig, Investigating the distribution and occupancy of 
otter species across human-modified landscapes in Sabah, Malaysia. 
https://zenodo.org/record/3897377. (Zenodo, 2020). 
80.  R. Pillay, R. J. Fletcher, K. E. Sieving, B. J. Udell, H. Bernard, Bioacoustic monitoring reveals 
shifts in breeding songbird populations and singing behaviour with selective logging in 
tropical forests. https://zenodo.org/record/3366104. (Zenodo, 2019). 
81.  L. Qie, E. Telford, R. Nilus, R. Ewers, Increased importance of terrestrial vertebrate seed 
dispersal in tropical logged forests. https://zenodo.org/record/3901735. (Zenodo, 2020). 
82.  L. Qie, E. Telford, M. Massam, R. Ewers, Impact Of El Nino Drought On Seedling Dynamics. 
https://zenodo.org/record/1400564. (Zenodo, 2018). 
83.  D. Seaman, M. Struebig, H. Bernard, M. Ancrenaz, R. M. Ewers, The effect of tropical forest 
modification on primate population density and diversity. 
https://zenodo.org/record/5109892. (Zenodo, 2021). 
84.  S. Sethi et al., Avifaunal and Herpetofaunal point counts with recorded acoustic data. 
https://zenodo.org/record/3742834. (Zenodo, 2020). 
85.  D. Shapiro, R. M. Ewers, Investigating Temperature Tolerance In Mosquito Disease Vectors 
Across A Land-Use Gradient. https://zenodo.org/record/1237720. (Zenodo, 2018). 
86.  A. Sharp, M. Barclay, A. Chung, R. Ewers, Beetle Diversity. 
https://zenodo.org/record/1323504. (Zenodo, 2018). 
87.  E. M. Slade, E. Bush, D. J. Mann, A. Y. C. Chung, Dung beetle community and dung removal 
data 2011. https://zenodo.org/record/3247492. (Zenodo, 2019). 
88.  E. M. Slade, A. Y. C. Chung, J. Parrett, Dung beetle community data 2018. 
https://zenodo.org/record/3832076. (Zenodo, 2020). 
89.  E. M. Slade, S. Milne, D. J. Mann, A. Y. C. Chung, J. Parrett, Dung beetle community and dung 
removal data 2015. https://zenodo.org/record/3247494. (Zenodo, 2019). 
90.  E. M. Slade, S. Milne, A. Y. C. Chung, J. Williamson, J. Parrett, Dung beetle community and 
dung removal data 2015. https://zenodo.org/record/3906118. (Zenodo, 2020). 
91.  E. M. Slade, J. Williamson, A. Y. C. Chung, J. Parrett, H. Heroin, Dung beetle community 
2017/18. https://zenodo.org/record/3906441. (Zenodo, 2020). 
92.  E. C. Turner et al., Tree census data from the SAFE Project 2011-12. 
https://zenodo.org/record/5729342. (Zenodo, 2021). 
93.  J. Twining, R. M. Ewers, Terrestrial Scavenger Trapping Data. 
https://zenodo.org/record/1237731. (Zenodo, 2018). 
94.  M. Vollans, L. Cator, R. M. Ewers, A. Chung, Investigating the impact of human settlements 
upon the availability of larval habitats and Aedes albopictus population. 
https://zenodo.org/record/3929764. (Zenodo, 2020). 
preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 
The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint 

95. C. Wilkinson et al., All Fish catch data at the SAFE project 2011-2017.

https://zenodo.org/record/3982665. (Zenodo, 2020).

96. J. Williamson, Movement Patterns Of Invertebrates In Tropical Rainforest.

https://zenodo.org/record/1487595. (Zenodo, 2018).

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Table S1. List of data sources compiled for analysis. For each data source, we present: the surname

of the first author and a data citation; a weblink to a data publication or, if that is unavailable, then a

weblink to a published paper presenting the data; the broad taxonomic grouping(s) that were the

focus of the study; the number of taxa that were shared with other surveys; the number of sampling

methods used; the number of sampling periods; and the final number of surveys we extracted from

that data source.

First author

Link

Taxon

type(s)

No.

taxa

No.

sampling

methods

No.

sample

periods

No.

surveys

Bernard (49)

https://zenodo.org/record/3908128

mammal

Bishop (50)

https://zenodo.org/record/1198839

invertebrate

Both (51)

https://zenodo.org/record/3247631

plant

133

Brant (52)

https://zenodo.org/record/1198846

invertebrate

Carpenter (53)

https://zenodo.org/record/5562260

invertebrate

Chapman (54)

https://zenodo.org/record/2579792

mammal

Deere (55)

https://zenodo.org/record/4010757

mammal

Döbert (56)

https://zenodo.org/record/2536270

plant

150

Drinkwater (57)

https://zenodo.org/record/3476542

invertebrate

Ewers (58)

https://zenodo.org/record/3975973

invertebrate

Faruk (59)

https://zenodo.org/record/1303010

amphibian

Fayle (60)

https://zenodo.org/record/3876227

invertebrate

Fraser (61)

https://zenodo.org/record/3973551

amphibian

Fraser (62)

https://zenodo.org/record/3981222

bird

Gray (63)

https://zenodo.org/record/1198302

invertebrate

Gray (64)

https://zenodo.org/record/3475406

invertebrate

Gregory (65)

https://zenodo.org/record/3994260

invertebrate

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Hardwick (66)

https://zenodo.org/record/4275386

invertebrate

Hemprich-Bennett

(67)

https://zenodo.org/record/3247465

mammal

Heon (39)

https://doi.org/10.1890/15-1363

mammal;

bird

Heon (68)

https://zenodo.org/record/3955050

mammal;

bird

Heon (69)

https://zenodo.org/record/1304117

mammal

Jebrail (70)

https://zenodo.org/record/3475408

invertebrate

Kendall (71)

https://zenodo.org/record/1237736

invertebrate

Konopik (72)

https://zenodo.org/record/1995439

amphibian

Lane Shaw (73)

https://zenodo.org/record/1237732

invertebrate

Luke (74)

https://zenodo.org/record/5710509

invertebrate

Luke (75)

https://zenodo.org/record/1198833

invertebrate

Mackintosh (76)

https://zenodo.org/record/4630980

invertebrate

Mitchell (40)

https://doi.org/10.1016/j.ecolind.202

0.106717

bird

132

Mullin (77)

https://zenodo.org/record/3971012

mammal

Noble (78)

https://zenodo.org/record/3485086

amphibian

Pianzin (79)

https://zenodo.org/record/3897377

mammal

Pillay (80)

https://zenodo.org/record/3366104

bird

Qie (81)

https://zenodo.org/record/3901735

mammal;

bird;

invertebrate

Qie (82)

https://zenodo.org/record/1400564

plant

275

Seaman (83)

https://zenodo.org/record/5109892

mammal

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Sethi (84)

https://zenodo.org/record/3997172

bird;

amphibian

227

Shapiro (85)

https://zenodo.org/record/1237720

invertebrate

Sharp (86)

https://zenodo.org/record/1323504

invertebrate

288

Slade (87-89)

https://zenodo.org/record/3247492

https://zenodo.org/record/3247494

https://zenodo.org/record/3832076

invertebrate

Slade (90, 91)

https://zenodo.org/record/3906118

https://zenodo.org/record/3906441

invertebrate

Turner (92)

https://zenodo.org/record/5729342

plant

Twining (93)

https://zenodo.org/record/1237731

mammal

Vollans (94)

https://zenodo.org/record/3929764

invertebrate

Wilkinson (95)

https://zenodo.org/record/4072959

fish

Williamson (96)

https://zenodo.org/record/1487595

invertebrate

The copyright holder for thisthis version posted February 12, 2024. ; https://doi.org/10.1101/2024.02.09.576668doi: bioRxiv preprint

Thresholds for adding degraded tropical forest to the conservation estate

Article

Full-text available

Jul 2024
NATURE

Logged and disturbed forests are often viewed as degraded and depauperate environments compared with primary forest. However, they are dynamic ecosystems¹ that provide refugia for large amounts of biodiversity2,3, so we cannot afford to underestimate their conservation value⁴. Here we present empirically defined thresholds for categorizing the conservation value of logged forests, using one of the most comprehensive assessments of taxon responses to habitat degradation in any tropical forest environment. We analysed the impact of logging intensity on the individual occurrence patterns of 1,681 taxa belonging to 86 taxonomic orders and 126 functional groups in Sabah, Malaysia. Our results demonstrate the existence of two conservation-relevant thresholds. First, lightly logged forests (<29% biomass removal) retain high conservation value and a largely intact functional composition, and are therefore likely to recover their pre-logging values if allowed to undergo natural regeneration. Second, the most extreme impacts occur in heavily degraded forests with more than two-thirds (>68%) of their biomass removed, and these are likely to require more expensive measures to recover their biodiversity value. Overall, our data confirm that primary forests are irreplaceable⁵, but they also reinforce the message that logged forests retain considerable conservation value that should not be overlooked.

Maximum temperatures determine the habitat affiliations of North American mammals

Article

Full-text available

Dec 2023
P NATL ACAD SCI USA

Addressing the ongoing biodiversity crisis requires identifying the winners and losers of global change. Species are often categorized based on how they respond to habitat loss; for example, species restricted to natural environments, those that most often occur in anthropogenic habitats, and generalists that do well in both. However, species might switch habitat affiliations across time and space: an organism may venture into human-modified areas in benign regions but retreat into thermally buffered forested habitats in areas with high temperatures. Here, we apply community occupancy models to a large-scale camera trapping dataset with 29 mammal species distributed over 2,485 sites across the continental United States, to ask three questions. First, are species’ responses to forest and anthropogenic habitats consistent across continental scales? Second, do macroclimatic conditions explain spatial variation in species responses to land use? Third, can species traits elucidate which taxa are most likely to show climate-dependent habitat associations? We found that all species exhibited significant spatial variation in how they respond to land-use, tending to avoid anthropogenic areas and increasingly use forests in hotter regions. In the hottest regions, species occupancy was 50% higher in forested compared to open habitats, whereas in the coldest regions, the trend reversed. Larger species with larger ranges, herbivores, and primary predators were more likely to change their habitat affiliations than top predators, which consistently affiliated with high forest cover. Our findings suggest that climatic conditions influence species’ space-use and that maintaining forest cover can help protect mammals from warming climates.

How to measure response diversity

Article

Full-text available

Mar 2023
Methods Ecol. Evol.

1. The insurance effect of biodiversity—that diversity stabilises aggregate ecosystem properties—is mechanistically underlain by inter- and intraspecific trait variation in organismal responses to the environment. This variation, termed response diversity, is therefore a potentially critical determinant of ecological stability. However, response diversity has yet to be widely quantified, possibly due to difficulties in its measurement. Even when it has been measured, approaches have varied. 2. Here, we review methods for measuring response diversity and from them distil a methodological framework for quantifying response diversity from experimental and/or observational data, which can be practically applied in lab and field settings across a range of taxa. 3. Previous empirical studies on response diversity most commonly invoke response traits as proxies aimed at capturing species’ ecological responses to the environment. Our approach, which is based on environment-dependent ecological responses to any biotic or abiotic environmental variable, is conceptually simple and robust to any form of environmental response, including nonlinear responses. Given its derivation from empirical data on species’ ecological responses, this approach should more directly reflect response diversity than the trait-based approach dominant in the literature. 4. By capturing even subtle inter- or intraspecific variation in environmental responses, and environment-dependencies in response diversity, we hope this framework will motivate tests of the diversity-stability relationship from a new perspective, and provide an approach for mapping, monitoring, and conserving this critical dimension of biodiversity.

Functional traits predict species responses to environmental variation in a California grassland annual plant community

Article

Full-text available

Feb 2022
J ECOL

Turnover in species composition and the dominant functional strategies in plant communities across environmental gradients is a common pattern across biomes, and is often assumed to reflect shifts in trait optima. However, the extent to which community‐wide trait turnover patterns reflect changes in how plant traits affect the vital rates that ultimately determine fitness remain unclear. We tested whether shifts in the community‐weighted means of four key functional traits across an environmental gradient in a southern California grassland reflect variation in how these traits affect species' germination and fecundity across the landscape. We asked whether models that included trait–environment interactions help explain variation in two key vital rates (germination rates and fecundity), as well as an integrative measure of fitness incorporating both vital rates (the product of germination rate and fecundity). To do so, we planted seeds of 17 annual plant species at 16 sites in cleared patches with no competitors, and quantified the lifetime seed production of 1360 individuals. We also measured community composition and a variety of abiotic variables across the same sites. This allowed us to evaluate whether observed shifts in community‐weighted mean traits matched the direction of any trait–environment interactions detected in the plant performance experiment. We found that commonly measured plant functional traits do help explain variation in species responses to the environment—for example, high‐SLA species had a demographic advantage (higher germination rates and fecundity) in sites with high soil Ca:Mg levels, while low‐SLA species had an advantage in low Ca:Mg soils. We also found that shifts in community‐weighted mean traits often reflect the direction of these trait–environment interactions, though not all trait–environment relationships at the community level reflect changes in optimal trait values across these gradients. Synthesis. Our results show how shifts in trait–fitness relationships can give rise to turnover in plant phenotypes across environmental gradients, a fundamental pattern in ecology. We highlight the value of plant functional traits in predicting species responses to environmental variation, and emphasise the need for more widespread study of trait–performance relationships to improve predictions of community responses to global change.

Addressing context dependence in ecology

Article

Full-text available

Oct 2021
TRENDS ECOL EVOL

Context dependence is widely invoked to explain disparate results in ecology. It arises when the magnitude or sign of a relationship varies due to the conditions under which it is observed. Such variation, especially when unexplained, can lead to spurious or seemingly contradictory conclusions, which can limit understanding and our ability to transfer findings across studies, space, and time. Using examples from biological invasions, we identify two types of context dependence resulting from four sources: mechanistic context dependence arises from interaction effects; and apparent context dependence can arise from the presence of confounding factors, problems of statistical inference, and methodological differences among studies. Addressing context dependence is a critical challenge in ecology, essential for increased understanding and prediction.

Vertebrate responses to human land use are influenced by their proximity to climatic tolerance limits

Article

Full-text available

May 2021
DIVERS DISTRIB

Aim: Land-use change leads to local climatic changes, which can induce shifts in community composition. Indeed, human-altered land uses favour species able to tolerate greater temperature and precipitation extremes. However, environmental changes do not impact species uniformly across their distributions, and most research exploring the impacts of climatic changes driven by land use has not considered potential within-range variation. We explored whether a population's climatic position (the difference between species' thermal and precipitation tolerance limits and the environmental conditions a population experiences) influences their relative abundance across land-use types. Location: Global. Methods: Using a global dataset of terrestrial vertebrate species and estimating their realized climatic tolerance limits, we analysed how the abundance of species within human-altered habitats relative to that in natural habitats varied across different climatic positions (controlling for proximity to geographic range edge). Results: A population's thermal position strongly influenced abundance within human-altered land uses (e.g. agriculture). Where temperature extremes were closer to species' thermal limits, population abundances were lower in human-altered land uses (relative to natural habitat) compared to areas further from these limits. These effects were generally stronger at tropical compared to temperate latitudes. In contrast , the influences of precipitation position were more complex and often differed between land uses and geographic zones. Mapping the outcome of models revealed strong spatial variation in the potential severity of decline for vertebrate populations following conversion from natural habitat to cropland or pasture, due to their climatic position. Main conclusions: We highlight within-range variation in species' responses to land use, driven (at least partly), by differences in climatic position. Accounting for spatial variation in responses to environmental changes is critical when predicting population vulnerability, producing successful conservation plans, and exploring how biodiversity may be impacted by future land-use and climate change interactions.

Disturbance type and species life history predict mammal responses to humans

Article

Full-text available

May 2021
GLOBAL CHANGE BIOL

Human activity and land use change impact every landscape on Earth, driving declines in many animal species while benefiting others. Species ecological and life history traits may predict success in human‐dominated landscapes such that only species with “winning” combinations of traits will persist in disturbed environments. However, this link between species traits and successful coexistence with humans remains obscured by the complexity of anthropogenic disturbances and variability among study systems. We compiled detection data for 24 mammal species from 61 populations across North America to quantify the effects of (1) the direct presence of people and (2) the human footprint (landscape modification) on mammal occurrence and activity levels. Thirty‐three percent of mammal species exhibited a net negative response (i.e., reduced occurrence or activity) to increasing human presence and/or footprint across populations, whereas 58% of species were positively associated with increasing disturbance. However, apparent benefits of human presence and footprint tended to decrease or disappear at higher disturbance levels, indicative of thresholds in mammal species’ capacity to tolerate disturbance or exploit human‐dominated landscapes. Species ecological and life history traits were strong predictors of their responses to human footprint, with increasing footprint favoring smaller, less carnivorous, faster‐reproducing species. The positive and negative effects of human presence were distributed more randomly with respect to species trait values, with apparent winners and losers across a range of body sizes and dietary guilds. Differential responses by some species to human presence and human footprint highlight the importance of considering these two forms of human disturbance separately when estimating anthropogenic impacts on wildlife. Our approach provides insights into the complex mechanisms through which human activities shape mammal communities globally, revealing the drivers of the loss of larger predators in human‐modified landscapes.

Recovery of logged forest fragments in a human-modified tropical landscape during the 2015-16 El Niño

Article

Full-text available

Mar 2021

The past 40 years in Southeast Asia have seen about 50% of lowland rainforests converted to oil palm and other plantations, and much of the remaining forest heavily logged. Little is known about how fragmentation influences recovery and whether climate change will hamper restoration. Here, we use repeat airborne LiDAR surveys spanning the hot and dry 2015-16 El Niño Southern Oscillation event to measure canopy height growth across 3,300 ha of regenerating tropical forests spanning a logging intensity gradient in Malaysian Borneo. We show that the drought led to increased leaf shedding and branch fall. Short forest, regenerating after heavy logging, continued to grow despite higher evaporative demand, except when it was located close to oil palm plantations. Edge effects from the plantations extended over 300 metres into the forests. Forest growth on hilltops and slopes was particularly impacted by the combination of fragmentation and drought, but even riparian forests located within 40 m of oil palm plantations lost canopy height during the drought. Our results suggest that small patches of logged forest within plantation landscapes will be slow to recover, particularly as ENSO events are becoming more frequent.

Localised climate change defines ant communities in human‐modified tropical landscapes

Article

Full-text available

Jan 2021
FUNCT ECOL

Logging and habitat conversion create hotter microclimates in tropical forest landscapes, representing a powerful form of localised anthropogenic climate change. It is widely believed that these emergent conditions are responsible for driving changes in communities of organisms found in modified tropical forests, although the empirical evidence base for this is lacking. Here we investigated how interactions between the physiological traits of genera and the environmental temperatures they experience lead to functional and compositional changes in communities of ants, a key organism in tropical forest ecosystems. We found that the abundance and activity of ant genera along a gradient of forest disturbance in Sabah, Malaysian Borneo, was defined by an interaction between their thermal tolerance (CTmax) and environmental temperature. In more disturbed, warmer habitats, genera with high CTmax had increased relative abundance and functional activity, and those with low CTmax had decreased relative abundance and functional activity. This interaction determined abundance changes between primary and logged forest that differed in daily maximum temperature by a modest 1.1°C, and strengthened as the change in microclimate increased with disturbance. Between habitats that differed by 5.6°C (primary forest to oil palm) and 4.5°C (logged forest to oil palm), a 1°C difference in CTmax among genera led to a 23% and 16% change in relative abundance, and a 22% and 17% difference in functional activity. CTmax was negatively correlated with body size and trophic position, with ants becoming significantly smaller and less predatory as microclimate temperatures increased. Our results provide evidence to support the widely held, but never directly tested, assumption that physiological tolerances underpin the influence of disturbance‐induced microclimate change on the abundance and function of invertebrates in tropical landscapes. A free Plain Language Summary can be found within the Supporting Information of this article.

Loss of grazing by large mammalian herbivores can destabilize the soil carbon pool

Article

Oct 2022
P NATL ACAD SCI USA

Grazing by mammalian herbivores can be a climate mitigation strategy as it influences the size and stability of a large soil carbon (soil-C) pool (more than 500 Pg C in the world's grasslands, steppes, and savannas). With continuing declines in the numbers of large mammalian herbivores, the resultant loss in grazer functions can be consequential for this soil-C pool and ultimately for the global carbon cycle. While herbivore effects on the size of the soil-C pool and the conditions under which they lead to gain or loss in soil-C are becoming increasingly clear, their effect on the equally important aspect of stability of soil-C remains unknown. We used a replicated long-term field experiment in the Trans-Himalayan grazing ecosystem to evaluate the consequences of herbivore exclusion on interannual fluctuations in soil-C (2006 to 2021). Interannual fluctuations in soil-C and soil-N were 30 to 40% higher after herbivore exclusion than under grazing. Structural equation modeling suggested that grazing appears to mediate the stabilizing versus destabilizing influences of nitrogen (N) on soil-C. This may explain why N addition stimulates soil-C loss in the absence of herbivores around the world. Herbivore loss, and the consequent decline in grazer functions, can therefore undermine the stability of soil-C. Soil-C is not inert but a very dynamic pool. It can provide nature-based climate solutions by conserving and restoring a functional role of large mammalian herbivores that extends to the stoichiometric coupling between soil-C and soil-N.

The macroecology of landscape ecology

Article

Feb 2022
TRENDS ECOL EVOL

One of landscape ecology’s main goals is to unveil how biodiversity is impacted by habitat transformation. However, the discipline suffers from significant context dependency in observed spatial and temporal trends, hindering progress towards understanding the mechanisms driving species declines and preventing the development of accurate estimates of future biodiversity change. Here, we discuss recent evidence that populations’ and species’ responses to habitat change at the landscape scale are modulated by factors and processes occurring at macroecological scales, such as historical disturbance rates, distance to geographic range edges, and climatic suitability. We suggest that placing landscape ecology studies in a macroecological lens will help to explain seemingly inconsistent results and will ultimately create better predictive models to help mitigate the biodiversity crisis.