Divergent Biodiversity Change Within Ecosystems SEE COMMENTARY

Divergent biodiversity change within ecosystems SEE COMMENTARY

Anne E. Magurrana,1,2, Amy E. Deacona,b,1, Faye Moyesa, Hideyasu Shimadzua,c, Maria Dornelasa, Dawn A. T. Phillipb,3, and Indar W. Ramnarineb

aCentre for Biological Diversity, School of Biology, University of St Andrews, St Andrews KY16 9TH, Scotland, United Kingdom; bDepartment of Life Sciences, The University of the West Indies, St. Augustine, Trinidad and Tobago; and cDepartment of Mathematical Sciences, Loughborough University, Loughborough LE11 3TU, United Kingdom

Edited by Nils Chr. Stenseth, University of Oslo, Oslo, Norway, and approved December 20, 2017 (received for review July 14, 2017) The Earth’s ecosystems are under unprecedented pressure, yet the levels anticipated by two different null models, even though nature of contemporary biodiversity change is not well under- temporal α-diversity metrics detected no systematic change (7). stood. Growing evidence that community size is regulated high- The contrasting conclusions emerging from the growing lights the need for improved understanding of community number of analyses of biodiversity change could, at least in part, be dynamics. As stability in community size could be underpinned rooted in the different methods used to evaluate diversity. One by marked temporal turnover, a key question is the extent to possibility is that temporal β-diversity metrics can identify com- which changes in both biodiversity dimensions (temporal α- and munity reorganization not well captured by species richness and temporal β-diversity) covary within and among the assemblages other α-diversity indices (16–18). Another possibility is that change that comprise natural communities. Here, we draw on a multias- in temporal α- and temporal β-diversity emerges on different semblage dataset (encompassing vertebrates, invertebrates, and timescales (19). A third option is that temporal stationarity in unicellular plants) from a tropical freshwater ecosystem and em- α-diversity, together with substantial compositional change, is a ploy a cyclic shift randomization to assess whether any directional general, but largely unrecognized, pattern in community ecology change in temporal α-diversity and temporal β-diversity exceeds (13). These explanations are not mutually exclusive and jointly lead baseline levels. In the majority of cases, α-diversity remains stable to the prediction that there will be a stronger signal of temporal over the 5-y time frame of our analysis, with little evidence for β α systematic change at the community level. In contrast, temporal -diversity than temporal -diversity in contemporary communities. β-diversity changes are more prevalent, and the two diversity di- Biodiversity is usually assessed at the assemblage level (14). mensions are decoupled at both the within- and among- Assemblages are groups of taxonomically related species that assemblage level. Consequently, a pressing research challenge is coexist and are often sampled together (14, 20). A community, to establish how turnover supports regulation and when elevated by contrast, is a broader concept encompassing the organisms temporal β-diversity jeopardizes community integrity. that co-occur but are not restricted by phylogeny or resource use (20). Single taxon (i.e., assemblage) assessments can provide only biodiversity change | tropical ecology | freshwater | temporal turnover | partial insight into overall community change as even well- community-level regulation studied groups, for example, birds and vascular plants, are not necessarily informative surrogates of biodiversity patterns in rave concern about the fate of the world’s biodiversity (1) Ghighlights the need for a better understanding of bio- Significance diversity change. The wholesale transformation of ecosystems, for example, from primary forest to pastureland, markedly re- The world’s biodiversity is under unprecedented threat due to duces the types and numbers of species present (2–4). However, human activities, yet we have an incomplete understanding of some localities and assemblages gain species from migrations ecosystem change in response to these pressures. Here we and introductions, while others support remarkably constant present data from a new 5-y study of a tropical freshwater numbers of species. The overall effect is that we do not detect ecosystem showing that change in the two dimensions of systematic biodiversity loss over the time period (typically years biodiversity—assemblage diversity (number and abundance of or decades) during which assemblages have been rigorously species) and assemblage composition—is decoupled from and monitored (1, 5). This raises questions about the nature of bio- uncorrelated among taxa. Assemblage diversity is typically diversity change and the extent to which ecological communities stable over time. However, in line with Darwin’s expectation are regulated (6). It also poses challenges for policy makers that community composition is constantly changing, this sta- charged with quantifying variation in biodiversity over space and bility can be accompanied by marked turnover in species time and with protecting ecosystem integrity. identities. Our paper thus identifies an important question for Biodiversity change is made up of temporal α-diversity— future research: at what point does compositional turnover change in the numbers and/or relative abundances of species— threaten ecosystem resilience? and temporal β-diversity—change in composition (7). Temporal α-diversity is often quantified using species richness (e.g., refs. 6– Author contributions: A.E.M. and I.W.R. designed research; A.E.M., A.E.D., and D.A.T.P. 8) but can be assessed by a variety of metrics such as the Shannon performed research; A.E.M., A.E.D., F.M., H.S., and M.D. analyzed data; and A.E.M., A.E.D., and Simpson indices (5). Accelerating anthropogenic pressures F.M., H.S., M.D., D.A.T.P., and I.W.R. wrote the paper. on the natural world, including climate change (9), habitat The authors declare no conflict of interest. transformations, and species introductions (10), may lead to This article is a PNAS Direct Submission. species loss (11) and hence be picked up by temporal α-diversity Published under the PNAS license. metrics (1, 12). However, as noted above, it is becoming clear Data deposition: The data reported in this article are available at dx.doi.org/10.17630/ that species loss (or gain) is not necessarily apparent in assess- ede726cd-3ab0-41a7-a6ef-5063732af297. ments of contemporary communities (5–8, 13). See Commentary on page 1681. Temporal β-diversity, the other dimension of biodiversity, tracks 1A.E.M. and A.E.D. contributed equally to this work. compositional change and is typically measured using Jaccard, 2To whom correspondence should be addressed. Email: [email protected]. – 3 Bray Curtis, and related (dis)similarity metrics (7, 14, 15). A re- Deceased October 28, 2017. ECOLOGY cent analysis has documented widespread compositional change, This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. with contemporary rates of temporal turnover often exceeding the 1073/pnas.1712594115/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1712594115 PNAS | February 20, 2018 | vol. 115 | no. 8 | 1843–1847 Downloaded by guest on October 2, 2021 other taxa (3). The need to understand biodiversity change at over 20 time points (four times per year: 2× in the dry season and 2× in the this broader “community” level is sharpened by the appreciation wet season for 5 y). that ecosystem function is linked to taxonomic composition and species richness and depends on the contributions of species in Sites. Sites in Trinidad’s Northern Range (SI Appendix) were sampled four different trophic groups and taxonomic assemblages (21, 22). times per year (twice in the wet season and twice in the dry season) for 5 y – Protecting function thus means preserving diversity across a (2010 2015). On each sampling occasion the diversity of fish, benthic in- wide range of taxa (23). However, different taxa within, as vertebrates, and diatoms at a site was quantified. Sampling methodology was consistent throughout. Eight of our sites were exposed to human well as among, assemblages may respond to different envi- “ – pressure in the form of recreational use (31). These recreationally dis- ronmental drivers (e.g., refs. 24 29) so there is no a priori turbed” (d) sites were matched with “recreationally undisturbed” (u) sites in reason why biodiversity change, as expressed by either temporal α β the same river system. Trinidad is a continental island that became separated -diversity or temporal -diversity should be correlated across from South America within the last 10,000 y (32). As result, the diversity of its taxonomic groups. fauna and flora is reduced relative to the mainland but still greater, relative Here we evaluate biodiversity change in the tropical fresh- to its size, than in other Caribbean islands. The γ-diversity of fish in Trinidad’s water community (vertebrates, invertebrates, and unicellular Northern Range, for example, is in the region of 30 species (33). The em- plants) in Trinidad’s Northern Range. These river systems are phasis in this paper on biodiversity change at the site level reflects the fact exposed to a range of global (e.g., climate change) and local that the rivers have distinctive fish faunas. As yet, the diversity and distri- (e.g., recreational use) stressors, but have not been subjected to bution of benthic invertebrates and diatoms in this region is poorly un- wholesale transformation in recent decades. If community reg- derstood. Nonetheless, these assemblages are included here as a better ulation is a general pattern (as predicted by ref. 13), we would representation of understudied groups in biodiversity assessments and, in expect to see little systematic change in temporal α- diversity the quantification of biodiversity change, are a priority (34). in any assemblage, but marked shifts in temporal β-diversity. We further expect these compositional shifts to be uncorre- Sampling. The same 50-m stretches of stream were revisited each session. Fish lated across assemblages. Ecological assemblages are not static were exhaustively sampled (35, 36) using a seine net followed by electro- entities, but instead experience continual flux in species com- fishing (37). The species of all individuals were identified and counted. Fish were returned unharmed to the capture site at the end of a session. A surber position and abundance (22). Quantifying biodiversity change sampler (36) was deployed to assess benthic invertebrates. Samples were thus means evaluating both temporal α-diversity and tempo- β returned to The University of the West Indies, and individuals identified to ral -diversity relative to a meaningful baseline. One way of family—the best taxonomic resolution achievable for this system. Diatoms doing this is to employ a cyclic shift randomization (30) that were sampled by collecting rocks from the stream bed. No taxonomic key is preserves within-species temporal autocorrelation and hence available for Trinidadian diatoms; specimens were therefore identified to captures realistic population dynamics, but breaks species cross- morphospecies using a photographic catalog (SI Appendix). We dropped the correlation. first sample in each assemblage from the analysis as there were a few in- consistences in sampling effort during the first session. Since stream sections Materials and Methods had comparable morphology, and sampling methods and effort were con- We quantify the temporal α-diversity and temporal β-diversity of each as- stant throughout the remaining n = 19 sessions, we did not correct for semblage [(i) fish, (ii) benthic invertebrates, and (iii) diatoms] at 16 river sites detectability. (See SI Appendix, B: Detailed sampling methodology,for

A C Chao1 PIE Dom ExpH MH BC ChaoJ Jturn Jnest Jtot PIE Dom Chao1 ExpH MH BC ChaoJ Jturn Jnest Jtot 1 1 S S 0.8 0.8 Chao1 Chao1 0.6 0.6 PIE PIE 0.4 0.4 Dom Dom 0.2 0.2 ExpH ExpH 0 0 MH MH 0.2 0.2 BC BC 0.4 0.4 ChaoJ ChaoJ 0.6 0.6 Jturn Jturn 0.8 0.8 Jnest Jnest Fig. 1. Correlations between metrics. These plots 1 1 show the pairwise correlations (Spearman) of bio- B D diversity change (linear regression slope, n = 16 sites) for a range of α- and β-diversity metrics for each as- Dom ExpH Jturn Jnest Jtot Chao1 PIE MH BC ChaoJ PIE ExpH MH BC Jturn Jnest Jtot 1 Chao1 Dom ChaoJ 1 semblage: (A) fish species; (B) fish families; (C)in- S S vertebrates; and (D) diatoms. Strength and direction 0.8 0.8 Chao1 Chao1 of the correlations are denoted by circle size and 0.6 0.6 color (as per scale bar). The scale bar extends from PIE PIE = 0.4 0.4 perfect correlation (dark blue, rs 1) to perfect Dom Dom anticorrelation (dark red, rs = −1). Larger circles also 0.2 0.2 ExpH ExpH denote a stronger correlation. Metric abbreviations: 0 0 S, species richness; Chao1, estimated S; PIE, proba- MH MH bility of interspecific encounter; Dom, McNaughton 0.2 0.2 BC BC Dominance; expH, exponential Shannon; MH, Morisita– 0.4 0.4 Horn; BC, Bray–Curtis; ChaoJ, Chao Jaccard dissimilarity; ChaoJ ChaoJ 0.6 0.6 Jturn, Jaccard turnover; Jnest, component of Jaccard Jturn Jturn dissimilarity due to richness change; Jtot, total Jaccard 0.8 0.8 Jnest Jnest dissimilarity. See SI Appendix, Table S7, for a description 1 1 of metrics.

1844 | www.pnas.org/cgi/doi/10.1073/pnas.1712594115 Magurran et al. Downloaded by guest on October 2, 2021 methodological details and SI Appendix, E: Detectability and repeatability, The analysis presented in Fig. 1 shows that the α-andβ-diversity for discussion of detectability issues.) metrics (PIE and Jaccard dissimilarity) that we used to quantify SEE COMMENTARY biodiversity change are representative of their class and, with some Analysis. We conducted a comprehensive analysis of biodiversity change by exceptions, behave consistently across assemblages. Our analysis first evaluating the temporal diversity trend of each assemblage at each site α also provides reassurance that taxonomic resolution (species ver- using a suite of metrics [ -diversity: species richness; Chao1; probability of sus families in fish) does not obscure general patterns. However, it interspecific encounter (PIE); McNaughton Dominance; exponential Shan- non; β-diversity: Morisita–Horn dissimilarity; Bray–Curtis dissimilarity; Chao does not tell us when and where change exceeds the expected Jaccard dissimilarity; Jaccard turnover component; Jaccard nestedness (rich- baseline. Our null model uncovered evidence for both increases α ness) component; Jaccard dissimilarity (7, 14, 38–40)]. The rationale for the and decreases in temporal -diversity (Figs. 2 and 3), suggesting choice of these measures is provided in SI Appendix, D: Detailed statistical that these changes are approximately balanced as anticipated by methodology, and Table S7; it also reflects Anderson et al.’s (41) point about refs. 7 and 13. For example, fish α-diversity declined substantially the advantages of taking a multifaceted approach when evaluating at Acono U while benthic invertebrate α-diversity increased at β-diversity. Metrics were computed in R (42) using vegan (43) and betapart Maracas D. Change exceeding the baseline expectation for tem- (44). To ensure a fair comparison with the benthic invertebrate and diatom poral β-diversity was more prevalent than change in temporal assemblages, we repeated the analyses of the fish assemblage at two tax- α-diversity (Figs. 2 and 3), and predominantly in the direction of onomic levels (species and family). increased dissimilarity over time (Fig. 3). Within sites, biodiversity Next, to measure the strength of directional change in diversity, we fitted a change was disassociated across the two components of diversity linear model (ordinary least squares) of each metric against time, within each and among assemblages (Fig. 3). assemblage (fish, invertebrates, diatoms), and calculated the slope of the The results produced by the two methods of handling dis- trend, as in ref. 7. In the case of β-diversity metrics, we adopted current similarity trends were almost identical with baseline changes in accepted practice in the field in setting the initial dissimilarity at time 1 to > zero, plotted shifts in the composition of the assemblage at each successive dissimilarity (as measured using the criterion in Fig. 2, i.e., Z 2) time point relative to composition at time 1, and fitted a linear model to as follows: fish species—two sites in each case; invertebrates— these data (SI Appendix, Figs. S5 and S6). We then computed the pairwise two sites in each case; and diatoms—four sites (constrained) correlation of slopes (Spearman index, n = 16 sites) to examine the re- versus five sites (not constrained). For diatoms, there were also a lationship between the metrics and visualized the results using the corrplot few instances where dissimilarity was reduced relative to expec- R package (45). tation (Z < −2): one site (constrained) versus two sites To test whether biodiversity change exceeded baseline expectations, we (not constrained). focused on two representative metrics, as informed by our correlation Seven localities (three with, and four without, recreational analysis - PIE (α-diversity) and Jaccard dissimilarity (β-diversity). We also chose disturbance: Fig. 3) exhibited no biodiversity change in any as- α PIE because it is a robust and informative -diversity measure (38); like the semblage or in either diversity dimension relative to the baseline Jaccard index, it is widely used and well-understood. However, we recognize that all measures have limitations and that other metrics may uncover different patterns (15, 39, 41, 46). We compared the observed slope of biodiversity change at each site (for each assemblage) against slopes (1,000 draws), computed using a cyclic shift permutation [cyclic_shift func- AB tion in the R package codyn (30)]. A cyclic shift permutation randomly selects the start time for each taxon in an assemblage. As such, species abundances vary independently, but within-species temporal autocorrelation is pre- served. This helps prevent bias that could occur with a free permutation in the presence of the moderate or strong autocorrelation associated with density dependence and seasonality (47). In each case, we asked whether the observed biodiversity change exceeded the expectation of a two-tailed test. We also computed Z-scores and the actual quantile of the observed slope, relative to the null distribution.

Results Over 670,000 individual organisms were recorded and identified in our investigation (see SI Appendix, Figs. S2 and S3 and Tables S3–S6, for details). As Fig. 1 shows, the metrics of biodiversity C change fell into two clusters representing trends in temporal α- and temporal β-diversity. As such, they revealed that temporal α- and temporal β-diversity metrics elucidate complementary aspects of biodiversity change. This disassociation of the two dimensions of biodiversity change was repeated across assemblages and was apparent at both taxonomic resolutions of the fish assemblage. Some of the measures used to describe temporal α-diversity, for example, slopes of PIE and McNaughton Domi- nance, were strongly anticorrelated. This is because these metrics track opposing aspects of assemblage structure, namely evenness and dominance (e.g., refs. 14 and 39). We note, however, that S and Chao 1 were strongly correlated in all cases, showing that Fig. 2. Concordance of change in temporal α- and temporal β-diversity, observed S and estimated S gave comparable evaluations of relative to baseline expectations, as informed by the cyclic shift randomi- biodiversity change. Similarly, there was generally good agree- zation, in the assemblages comprising the Trinidadian freshwater commu- ment in the patterns uncovered by Chao Jaccard and classic nities [fish species (A), benthic invertebrates (B), and diatoms (C)]. The − ≥ ≥ Jaccard dissimilarity, suggesting that unseen species were not number of sites exhibiting change (using the criterion of 2 Z 2) is in- β β dicated in each cell, with cells in white indicating “no change” in either biasing estimates of temporal -diversity. However, -diversity diversity dimension. For example, in A (fish species), there is one site where

partitioning revealed a predominant contribution of turnover to change in excess of the baseline is detected for both α-(α*) and β-diversity ECOLOGY compositional change in the fish assemblage and nestedness (β*), one with a change in α-diversity only, one with a change in β-diversity (richness) in the invertebrate and diatom assemblages. only, and 13 sites exhibiting no change.

Magurran et al. PNAS | February 20, 2018 | vol. 115 | no. 8 | 1845 Downloaded by guest on October 2, 2021 Fig. 3. Assemblage and geographic differences in temporal change. Each block of three cells reveals the presence of directional change (relative to the expected baseline set by the null model) at a given site for fish (species), invertebrates, and diatoms, respectively. In each case, the top row of cells denotes temporal α-diversity, and the bottom row temporal β-diversity. Cells where diversity has changed in ways that are generally considered unfavorable are shaded red with darker shading indicating a more marked change (Z scores are indicated on the scale bar: ≤−3, ≤−2, ≤−1forα-diversity and ≥3, ≥2, ≥1for β-diversity). These are sites and/or assemblages in which α-diversity has declined or where compositional dissimilarity has increased. Gray cells are those where there is no detectable change in diversity. The blue cells indicate an increase in α-diversity or cases where compositional dissimilarity has changed less than would be expected by chance. The locations of the 16 sites across Trinidad’s Northern Range are identified by black circles with abbreviations as follows: AD, Acono d; AU, Acono u; CD, Caura d; CU, Caura u; LAD, Lower Aripo d; LAU, Lower Aripo u; LD, Lopinot d; LU, Lopinot u; MD, Maracas d; MU, Maracas u; QD, Quare d; QU, Quare u; TD, Turure d; TU, Turure u; UAD, Upper Aripo d; UAU, Upper Aripo u. Arrows indicate the cells associated with a site.

criterion. (See SI Appendix, Fig. S7, for further examination of also reflected in a trend toward spatial biotic homogenization (52, the relationship between disturbance and biodiversity change.) 53) in fish and invertebrates and a trend toward spatial biotic heterogenization (52) in diatoms (SI Appendix,Fig.S12). Whether Discussion this homogenization/heterogenization dichotomy is linked to di- These results provide support for the argument (13) that com- chotomy in the turnover/nestedness reported above remains to be munity regulation—resilience in temporal α-diversity—is a gen- tested. Responses at more fine-scaled levels of organization may eral feature of ecological systems. As such, it points to a set of further contribute to the compositional shifts. Recent theory (54) processes that contribute to the stability of ecological commu- suggests that higher-order interactions among species can support nities (48). We detected little directional change, relative to the coexistence while modularity, with different subsets of taxa null expectation, in α-diversity, at any site or assemblage. In responding to different drivers, has the potential to promote sta- contrast, we found more evidence of directional change in bility (55, 56). β-diversity in all assemblages but few correlated compositional Notwithstanding the above, the noncorrelated pattern of as- shifts within sites. semblage turnover observed here sheds little light on whether The observation that resilience in α-diversity can occur despite community resilience (as shown by nontrending α-diversity) oc- substantial temporal β-diversity raises questions about the role of curs because of temporal β-diversity, or despite it. As noted by turnover in community processes. Darwin took the view that Darwin [and inherent in MacArthur and Wilson (57) and similar temporal turnover is a mechanism linked to the persistence of ecological theory], some degree of turnover is essential to pro- community properties, particularly the community’s overall size. mote community persistence. However, we also know that con- In The Origin of Species (22, p. 69), he states: temporary rates of temporal turnover exceed the theoretical expectations of existing ecological models (7, 13). It is likely that ... .we forget that each species, even where it most abounds, is con- anthropogenic impacts, including climate change and transloca- stantly suffering enormous destruction at some period of its life, from tions of exotic species, are contributing to the temporal turnover enemies or from competitors for the same place and food; and if that we observe in this and other studies. Moreover, while the these enemies or competitors be in the least degree favoured by any slight change of climate, they will increase in numbers; and as each functional consequences of biodiversity change within the eco- area is already fully stocked with inhabitants, the other species system as a whole remain unclear, research suggests that eco- must decrease. system function will be prejudiced if diversity is reduced in any taxonomic or functional groups (21). A pressing research ques- The noncorrelated assemblage pattern of temporal β-diversity tion, then, is, at what point does elevated temporal β-diversity that we observed in Trinidad’s Northern Range supports this jeopardize community integrity? perspective. Taxa may wax and wane in abundance as a result of Our study provides strong support for the quantification of temporal variation in different sets of biotic and abiotic factors, temporal β-diversity alongside temporal α-diversity in assessments but the assemblages to which they belong tend not to show a of biological diversity (16–18). Although α-diversity, particularly directional trend in α-diversity. Compositional changes can be species richness, is an intuitive and widely used metric (5), the use decoupled across assemblages because of varying taxonomic re- of complementary measures will deepen understanding of bio- sponses to environmental change. Diatoms, for example, are re- diversity change (19, 41, 58). A particular challenge will be iden- sponsive to variation in temperature and pH, macroinvertebrates tifying when (and why) temporal β-diversity provides an early to small-scale spatial modifications, and tropical freshwater fish to warning signal of change. The next generation of pattern- and a range of variables including changes to river-channel morphology process-based models will be crucial in meeting this goal (19). (49–51). (See ref. 29 for a modeling framework linking diversity Monitoring schemes are not evenly distributed across the responses to environmental variables.) Assemblage differences are Earth’s surface (28), and data gaps from tropical ecosystems and

1846 | www.pnas.org/cgi/doi/10.1073/pnas.1712594115 Magurran et al. Downloaded by guest on October 2, 2021 the freshwater realm mean that these systems are under- ACKNOWLEDGMENTS. We thank the field team—Rajindra Mahabir, Kharran — represented in appraisals of biodiversity change (7, 8). Compo- Deonarinesingh, and Avinash Deonarinesingh and Khadija Huggins, Sarah SEE COMMENTARY Shageer, and Devan Inderlall for their help in processing the invertebrate sitional shifts are particularly difficult to generalize as rates of samples; and Mary Alkins-Koo and Julian Duncan for advice on inverte- temporal turnover can vary across localities and groups (34, 59, brate identification and diatom identification, respectively. We are grate- 60), yet it is becoming clear that evaluating temporal β-diversity ful to the reviewers for their helpful comments. This project was funded by is essential not just to safeguard the world’s ecosystems, but also the European Research Council (AdG BioTIME 250189 and PoC BioCHANGE 727440). A.E.M. also acknowledges support from the Royal Society and to answer fundamental questions about the resilience of eco- M.D. from the Scottish Funding Council (Marine Alliance for Science and logical communities and the maintenance of biodiversity. Technology for Scotland Grant HR09011).

1. Newbold T, et al. (2015) Global effects of land use on local terrestrial biodiversity. 33. Phillip DAT, et al. (2013) Annotated list and key to the stream fishes of Trinidad & Nature 520:45–50. Tobago. Zootaxa 3711:1–64. 2. McGill BJ, Dornelas M, Gotelli NJ, Magurran AE (2015) Fifteen forms of biodiversity 34. Lawton JH, et al. (1998) Biodiversity inventories, indicator taxa and effects of habitat trend in the Anthropocene. Trends Ecol Evol 30:104–113. modification in tropical forest. Nature 391:72–76. 3. Godfray HCJ, May RM (2014) Open questions: Are the dynamics of ecological com- 35. Deacon AE, Shimadzu H, Dornelas M, Ramnarine IW, Magurran AE (2015) From munities predictable? BMC Biol 12:22. species to communities: The signature of recreational use on a tropical river ecosys- 4. Nimmo DG, Mac Nally R, Cunningham SC, Haslem A, Bennett AF (2015) Vive la tem. Ecol Evol 5:5561–5572. résistance: Reviving resistance for 21st century conservation. Trends Ecol Evol 30: 36. Southwood R, Henderson PA (2000) Ecological Methods (Blackwell Science, Oxford), – 516 523. 3rd Ed, 592 pp. 5. Isbell F, et al. (2017) Linking the influence and dependence of people on biodiversity 37. Deacon AE, Mahabir R, Inderlall D, Ramnarine IW, Magurran AE (2017) Evaluating – across scales. Nature 546:65 72. detectability of freshwater fish assemblages in tropical streams: Is hand-seining suf- 6. Vellend M, et al. (2017) Estimates of local biodiversity change over time stand up to ficient? Environ Biol Fishes 100:839–849. scrutiny. Ecology 98:583–590. 38. Gotelli NJ (2008) A Primer of Ecology (Sinauer, Sunderland, MA). 7. Dornelas M, et al. (2014) Assemblage time series reveal biodiversity change but not 39. Maurer BA, McGill BJ (2011) Measurement of species diversity. Biological Diversity: systematic loss. Science 344:296–299. Frontiers in Measurement and Assessment, eds Magurran AE, McGill BJ (Oxford Univ 8. Vellend M, et al. (2013) Global meta-analysis reveals no net change in local-scale plant Press, Oxford), pp 55–64. biodiversity over time. Proc Natl Acad Sci USA 110:19456–19459. 40. Baselga A (2010) Partitioning the turnover and nestedness components of beta di- 9. Stuart-Smith RD, Edgar GJ, Barrett NS, Kininmonth SJ, Bates AE (2015) Thermal biases versity. Glob Ecol Biogeogr 19:134–143. and vulnerability to warming in the world’s marine fauna. Nature 528:88–92. 41. Anderson MJ, et al. (2011) Navigating the multiple meanings of β diversity: A road- 10. Chown SL (2010) Temporal biodiversity change in transformed landscapes: A southern map for the practicing ecologist. Ecol Lett 14:19–28. African perspective. Philos Trans R Soc Lond B Biol Sci 365:3729–3742. 42. RCoreTeam (2014) R: A Language and Environment for Statistical Computing (R 11. Vilà M, et al. (2011) Ecological impacts of invasive alien plants: A meta-analysis of their effects on species, communities and ecosystems. Ecol Lett 14:702–708. Foundation for Statistical Computing, Vienna). Version 3.3.3. Available at www.r- 12. Elahi R, et al. (2015) Recent trends in local-scale marine biodiversity reflect community project.org/. Accessed October 1, 2017. structure and human impacts. Curr Biol 25:1938–1943. 43. Oksanen J, et al. (2015) vegan: Community Ecology Package. R package Version 2.2-1. = 13. Gotelli NJ, et al. (2017) Community-level regulation of temporal trends in biodiversity. Available at cran.r-project.org/package vegan. Accessed October 19, 2017. Sci Adv 3:e1700315. 44. Baselga A, Orme D, Villeger S, De Bortoli J, Leprieur F (2013) betapart: Partitioning 14. Magurran AE (2004) Measuring Biological Diversity (Blackwell Science, Oxford). Beta Diversity into Turnover and Nestedness Components. R package Version 1.3. 15. Jost L, Chao A, Chazdon RL (2011) Compositional similarity and beta diversity. Available at cran.r-project.org/package=betapart. Accessed October 19, 2017. Biological Diversity: Frontiers in Measurement and Assessment, eds Magurran AE, 45. Wei T, Simko V (2016) corrplot: Visualization of a Correlation Matrix. R package McGill BJ (Oxford University Press, Oxford), pp 66–84. Version 0.77. Available at https://cran.r-project.org/package=corrplot. Accessed October 16. Magurran AE, Henderson PA (2010) Temporal turnover and the maintenance of di- 19, 2017. versity in ecological assemblages. Philos Trans R Soc Lond B Biol Sci 365:3611–3620. 46. Greenacre M (2017) ‘Size’ and ‘shape’ in the measurement of multivariate proximity. 17. Santini L, et al. (2017) Assessing the suitability of diversity metrics to detect bio- Methods Ecol Evol 8:1415–1424. diversity change. Biol Conserv 213:341–350. 47. Henderson PA, Magurran AE (2014) Direct evidence that density-dependent regula- 18. Hillebrand H, et al. (2017) Biodiversity change is uncoupled from species richness tion underpins the temporal stability of abundant species in a diverse animal com- trends: Consequences for conservation and monitoring. J Appl Ecol 55:169–184. munity. Proc Biol Sci 281:20141336. 19. Hill SLL, et al. (2016) Reconciling biodiversity indicators to guide understanding and 48. May RM (1973) Stability and Complexity in Model Ecosystems (Princeton Univ Press, – action. Conserv Lett 9:405 412. Princeton, NJ). 20. Fauth JE, et al. (1996) Simplifying the jargon of community ecology: A conceptual 49. Soininen J, Jamoneau A, Rosebery J, Passy SI (2016) Global patterns and drivers of – approach. Am Nat 147:282 286. species and trait composition in diatoms. Glob Ecol Biogeogr 25:940–950. 21. Oliver TH, et al. (2015) Biodiversity and resilience of ecosystem functions. Trends Ecol 50. Vilmi A, Karjalainen SM, Nokela T, Tolonen K, Heino J (2016) Unravelling the drivers – Evol 30:673 684. of aquatic communities using disparate organismal groups and different taxonomic 22. Darwin C (1859) On the Origin of Species by Means of Natural Selection, or the levels. Ecol Indic 60:108–118. Preservation of Favoured Races in the Struggle for Life (John Murray, London). 51. Tejerina-Garro FL, et al. (2005) Effects of natural and anthropogenic environmental 23. Soliveres S, et al. (2016) Biodiversity at multiple trophic levels is needed for ecosystem changes on riverine fish assemblages: A framework for ecological assessment of riv- multifunctionality. Nature 536:456–459. ers. Braz Arch Biol Technol 48:91–108. 24. Peters MK, et al. (2016) Predictors of elevational biodiversity gradients change from 52. Socolar JB, Gilroy JJ, Kunin WE, Edwards DP (2016) How should beta-diversity inform single taxa to the multi-taxa community level. Nat Commun 7:13736. biodiversity conservation? Trends Ecol Evol 31:67–80. 25. Thom D, et al. (2017) The impacts of climate change and disturbance on spatio- 53. Magurran AE, Dornelas M, Moyes F, Gotelli NJ, McGill B (2015) Rapid biotic homog- temporal trajectories of biodiversity in a temperate forest landscape. J Appl Ecol enization of marine fish assemblages. Nat Commun 6:8405. 54:28–38. 54. Grilli J, Barabás G, Michalska-Smith MJ, Allesina S (2017) Higher-order interactions 26. Bowler DE, et al. (2015) A cross-taxon analysis of the impact of climate change on – abundance trends in central Europe. Biol Conserv 187:41–50. stabilize dynamics in competitive network models. Nature 548:210 213. 27. Bowler DE, et al. (2017) Cross-realm assessment of climate change impacts on species’ 55. Haldane AG, May RM (2011) Systemic risk in banking ecosystems. Nature 469: – abundance trends. Nat Ecol Evol 1:67. 351 355. – 28. Korhonen JJ, Soininen J, Hillebrand H (2010) A quantitative analysis of temporal 56. Scheffer M, et al. (2012) Anticipating critical transitions. Science 338:344 348. turnover in aquatic species assemblages across ecosystems. Ecology 91:508–517. 57. MacArthur RH, Wilson EO (1967) The Theory of Island Biogeography (Princeton Univ 29. De’ath G (2012) The multinomial diversity model: Linking Shannon diversity to mul- Press, Princeton, NJ). tiple predictors. Ecology 93:2286–2296. 58. Soininen J, Heino J, Wang J (2017) A meta-analysis of nestedness and turnover 30. Hallett LM, et al. (2016) codyn: An r package of community dynamics metrics. components of beta diversity across organisms and ecosystems. Glob Ecol Biogeogr Methods Ecol Evol 7:1146–1151. 27:96–109. 31. Dudgeon D, et al. (2006) Freshwater biodiversity: Importance, threats, status and 59. Westgate MJ, Barton PS, Lane PW, Lindenmayer DB (2014) Global meta-analysis re- conservation challenges. Biol Rev Camb Philos Soc 81:163–182. veals low consistency of biodiversity congruence relationships. Nat Commun 5:3899. 32. Kenny JS (1995) Views from the Bridge: A Memoir on the Freshwater Fishes of 60. Heino J (2010) Are indicator groups and cross-taxon congruence useful for predicting Trinidad (J. S. Kenny, Maracas, Trinidad and Tobago). biodiversity in aquatic ecosystems? Ecol Indic 10:112–117. ECOLOGY

Magurran et al. PNAS | February 20, 2018 | vol. 115 | no. 8 | 1847 Downloaded by guest on October 2, 2021