Lowland biotic attrition revisited: body size and variation among climate change ‘winners’ and ‘losers’

Abstract

The responses of lowland tropical communities to climate change will critically influence global biodiversity but remain poorly understood. If species in these systems are unable to tolerate warming, the communities—currently the most diverse on Earth—may become depauperate (‘biotic attrition’). In response to temperature changes, animals can adjust their distribution in space or their activity in time, but these two components of the niche are seldom considered together. We assessed the spatio-temporal niches of rainforest mammal species in Borneo across gradients in elevation and temperature. Most species are not predicted to experience changes in spatio-temporal niche availability, even under pessimistic warming scenarios. Responses to temperature are not predictable by phylogeny but do appear to be trait-based, being much more variable in smaller-bodied taxa. General circulation models and weather station data suggest unprecedentedly high midday temperatures later in the century; predicted responses to this warming among small-bodied species range from 9% losses to 6% gains in spatio-temporal niche availability, while larger species have close to 0% predicted change. Body mass may therefore be a key ecological trait influencing the identity of climate change winners and losers. Mammal species composition will probably change in some areas as temperatures rise, but full-scale biotic attrition this century appears unlikely.

1. Introduction

Climate change presents a major threat to biodiversity. The responses of lowland tropical organisms are particularly critical to understand [1,2]. Species at higher latitudes and elevations are undergoing range shifts, such that communities that lose some species, for example through northwards or upwards range shifts, may gain new ones moving in from lower latitudes or elevations. But this highlights a critical area of uncertainty: there is nowhere ‘lower’ in latitude or elevation than the tropical lowlands. Whether these areas—currently the most biodiverse on Earth and hotbeds for speciation—will remain diverse, reassemble into novel compositions or become depauperate in the face of climate change remains unknown. If temperatures that are rising to levels unprecedented for the last 2 million years do drive substantial species loss in tropical lowlands, this ‘biotic attrition’ [2] will therefore majorly influence the ecology and evolution of life on Earth.

Ongoing warming could push ambient temperatures above the thermal neutral zones of many tropical endotherms [3–5], suggesting that these species could go extinct or abandon the hot lowlands. Elevational range shifts have indeed been observed in mammals at higher latitudes [6,7] and by other taxa in the tropics (e.g. [8]). But in all areas, mammal population responses to climate change are highly idiosyncratic across species [6], rendering it currently impossible to accurately predict climatic impacts on this ecologically and economically important taxon that is disproportionately threated by extinction [9]. Indeed, in all taxonomic groups, determining which species will benefit or suffer from climate change (‘winners’ and ‘losers’, respectively) has become a critical research priority [10–13].

Temperature sensitivity partly determines a species' niche. The abiotic conditions under which a species can persist—its fundamental niche—can be mapped onto both space and time. Spatial niche mapping is commonly used, underlying the use of species distribution models and environmental niche models to forecast climate change impacts on species' ranges and community structure. Assessing how climate change alters temporal niches (species' daily, seasonal and annual activity patterns) is less common, though possibly just as important. Rarer still is the assessment of how changing abiotic conditions might simultaneously affect spatial and temporal niches, despite the fact that these two niche dimensions are clearly interdependent [14]. Understanding how species' temporal niches are influenced by sensitivity to climatic factors would greatly improve predictions of climate change impacts across organisms. For example, organisms might be able to cope with being at or near their upper critical thermal limits via behavioural plasticity [4], shifting their activity progressively away from midday and towards twilight and night-time in order to avoid high temperatures [15,16].

Here, we assess both the spatial and temporal components of the niche across a suite of mammal species in the tropical rainforests of Borneo. We measure species' spatio-temporal niches—their activity at each hour at each location on the landscape. We use weather station data and regional temperature models to predict hourly temperatures across elevation gradients, and assess species' activity as a function of temperature. We then ask whether relative temperature sensitivity varies with body size. It is thought that much of the variation among species in susceptibility to climate change could be trait-based [17,18]; body size, in particular, is a fundamental ecological trait that affects numerous facets of an organism's ecology [19,20]. But the effects of body size on mammal responses to warming are mixed [6,20,21], suggesting that size-based vulnerability of animals to climate change remains incompletely understood.

We conducted standardized camera-trapping in 16 study areas spanning 980 linear kilometres and elevations from 0 to 1600+ m, corresponding to an ambient temperature gradient of approximately 2°—the minimum amount by which global temperatures now seem likely to rise over the next century [22]. Cameras were triggered by motion and active 24 h d⁻¹, thereby providing continuous measurements of activity for a suite of vertebrates at different sites.

2. Methods

Our study region was in tropical rainforest that received more than 2000 mm of rain per year and was dominated by Dipterocarpaceae trees. We deployed motion-triggered camera traps in 16 study areas across Malaysian Borneo, resulting in a total of 303 camera locations. For seven of these study areas, detailed descriptions of the sampling is provided in Brodie et al. [23,24]. We also included data from Danum Valley (A. Granados 2015, unpublished data) and Imbak Canyon [25,26] in Sabah, and seven areas in Sarawak (J. Mohd-Azlan 2015, unpublished data; [27]). Some of our cameras were in selectively logged forests, where many of the trees above a certain size threshold (usually approximately 60 cm dbh) had been removed. In all areas, cameras were attached to trees, usually approximately 30–60 cm above the ground. Study areas, and camera locations within them, spanned gradients in elevation and logging intensity, and were chosen with no prior knowledge of daily animal activity patterns.

Multiple factors besides temperature could potentially affect animal activity patterns. For example, animals may alter their activity to avoid predators or competitors [28]. Perhaps most importantly, mammals could potentially become more nocturnal in areas where human activity is high [16] such as logged or hunted areas; indeed, we noticed anecdotally that large animals in parts of Gunung Mulu National Park (subject to both tourism and legal hunting) were more rarely photographed during the day than they were at other sites with less human activity. We tried to account for this by including camera location as a categorical fixed effect in our analysis.

We analysed data on 25 native mammal species (electronic supplementary material, table S1) that ranged in size from the 800 g moonrat (Echinosorex gymnura) to the 178 kg sambar (Rusa unicolor). We excluded species that were detected only sporadically, such as several squirrel and tree shrew species, and also species with extremely limited geographical ranges due to human influence, such as the banteng (Bos javanicus). We deemed photographic records of the same species at the same location to be independent if they were non-consecutive (i.e. another species was recorded in between) or if they occurred more than 30 min apart; these are increasingly common criteria used in camera trap data analysis [29,30].

We gathered hourly temperature data from all weather stations in Malaysian Borneo and Brunei from the National Oceanic and Atmospheric Administration, USA, Integrated Surface Database (http://www.ncdc.noaa.gov/isd; accessed 10 July 2015). All weather stations were less than or equal to 30 m elevation.

We developed a regional time-lapse model to predict hourly temperatures at locations across Malaysian Borneo. The model used observed temperatures at the full set (S = 10) of weather stations in East Malaysia and Brunei with usable data and adjusted for vertical and horizontal differences between each weather station and each camera trap station; the model gives more weight to measurements from nearby weather stations than to those more distant. Specifically, for each hour of each day at each camera station, we estimated the temperature () as

where T_s,h is the measured temperature at weather station s at hour h, E_cs and D_cs are the differences in elevation and horizontal distance, respectively, between camera c and weather station s, and the lapse rate in temperature with elevation (−0.0055°C m⁻¹) is based on estimates from Gaffen et al. [31]. Hourly varying lapse-rate estimates are not available from our system; Henn et al. [32], however, found that a long-term lapse rate model (as we use here) had a better fit than hourly lapse-rate models to observed temperature data from a North American mountain range. We note that our model-predicted temperatures are useful for relative comparisons across time and space but may be imprecise as absolute estimates because weather stations are in open areas whereas our field sampling was under forest canopies.

We assessed how temperature and elevation affected the probability of detecting each mammal species using logistic regression. Because we had so many more non-detections than detections in this hourly dataset, we used rare events logistic regression models with the package Zelig [33] in the statistical program R [34]. We restricted the analysis to camera stations for which at least one detection occurred for a given species. For each species, we compared intercept-only, temperature-only, elevation-only and temperature × elevation models using Akaike information criteria (AIC)-based model selection.

We used the regional lapse-rate climate model above, parametrized with weather station data, to predict current average hourly temperatures across 1 km² landscape pixels in East Malaysia and Brunei. We then added the predicted temperature changes from a general circulation model (GCM) ensemble analysis to the current temperatures to assess average hourly temperatures at each landscape pixel in 2070 [35]. We used predicted temperatures based on two emissions trajectory scenarios, relative concentration pathway (RCP) 2.6 and RCP 8.5, representing optimistic and pessimistic (respectively) warming scenarios. We averaged the predicted maximum and minimum temperature changes from the ensemble GCM analysis.

We then used the spatio-temporal temperature predictions with the rare events logistic regression coefficients to predict the occurrence probability of each species at each hour of the day across the study region. For a given species, we calculated current and future spatio-temporal niche availability as the sum of predicted hourly occurrence probabilities across the study region.

Finally, we assessed relative temperature sensitivity across species, measured as either model-averaged temperature coefficients or predicted proportional changes in future spatio-temporal niche availability (based on RCP 2.6 and RCP 8.5). We regressed these metrics of relative heat sensitivity against body size (log₁₀ transformed) from the PanTHERIA database [36] in a phylogenetically explicit framework, using the ‘best dates’ mammal supertree from Bininda-Emonds et al. [37] and the package Caper [38] in R [34].

3. Results

Across 10 weather stations (electronic supplementary material, figure S1), we obtained 2 455 486 hourly temperature measurements from 1949 to 2015. Average hourly temperatures were highly consistent across weather stations (electronic supplementary material, figure S2). Based on the ensemble analysis of GCMs, average temperatures across the study region are predicted to rise 1.32°C (RCP 2.6) to 2.78°C (RCP 8.5) by 2070.

Out of 32 605 camera trap days across 303 camera trap stations in 16 study areas, we obtained 20 876 independent detections of 25 species (electronic supplementary material, table S1). Model-averaged temperature coefficients from the rare events logistic regression models ranged from −0.49 to 0.75 across species (mean = −0.09). Model-averaged predicted proportional changes in spatio-temporal niche availability ranged from −0.05 to 0.02 (mean = 0.00) across species for RCP 2.6 and from −0.09 to 0.06 (mean = −0.01) across species for RCP 8.5 (electronic supplementary material, table S1).

Using phylogenetic generalized least-squares regression, maximum-likelihood-estimated λ = 0 for the models estimating both RCP 2.6 and RCP 8.5 proportional niche changes, indicating that niche changes were unrelated to phylogeny. Thereafter, we used non-phylogenetic general linear regressions weighted by the number of detections of each species. Body size did not significantly affect predicted RCP 2.6 niche changes (p = 0.28), predicted RCP 8.5 niche changes (p = 0.41) or model-averaged temperature coefficients (p = 0.58) across species. The coefficient of variation in each of these temperature sensitivity metrics was 2.7- to 12.4-fold higher for species weighing approximately 1–3 kg than for larger species (figure 1).

Figure 1.

Relationships (grey circles) between body size and (a) temperature effects on occurrence, and predicted spatio-temporal niche changes under RCP 2.6 (b) and RCP 8.5 (c). Coefficient of variation (bars) for 0.5 log₁₀(kg) body mass categories shown on the right y-axes. Sizes of the circles are proportional to the sample size for each species.

4. Discussion

How global biodiversity responds to climate change largely hinges on how rising temperatures affect lowland tropical organisms. Many tropical endotherms may soon experience ambient temperatures above their laboratory-measured thermal neutral zones [3]. Species' behavioural plasticity may allow them to persist even in the face of excessive heat. Our combined temporal–spatial niche analysis provides a preliminary test of potential heat sensitivity, and the ability of behavioural plasticity to buffer this sensitivity, across 25 species.

We detected negligible temperature responses in most species. Kaliq et al. [3] estimated that most mammals—across all latitudes—currently experienced ambient temperatures well within their thermal neutral zones, but suggested that ongoing warming could soon generate ambient temperatures above the ‘thermal safety zone’ for numerous species. This is consistent with our results that suggest widespread temperature insensitivity, but we note the possibility that effects on some species could manifest as temperatures continue to rise.

Understanding which species are most susceptible to anthropogenic threats is of growing importance. While mammal responses to logging-induced habitat disruption may be predictable based on phylogeny [39], our results suggest that shared evolutionary history does not influence vulnerability to rising temperatures. Instead, variation in predicted climate change responses appears to be based on an ecological trait, body size. As body size increases, surface-area-to-volume ratios decline, making it relatively harder for large endotherms to shed heat [6]. But by virtue of their larger body mass reserves and slower metabolic rate [20], it might be that large animals can go for a longer time without feeding than smaller species can, and thus have greater behavioural plasticity to ‘sit out’ hot spells. Additional research on body mass and other key ecological traits could improve the accuracy of predictions about species responses to climate change.

Biotic attrition implies upward distributional shifts where possible (and local extinction otherwise). Elevational range shifts in animals, however, are complicated by the fact that upward shifts in temperature regimes do not necessarily coincide with upward shifts in plant distributions, raising the possibility that animals might have to shift their distributions to track the temperatures they require but, in so doing, abandon particular plant resources (as well as conditions such as salt licks, cover, etc.). Evidence along these lines remains very scant, but certainly supports the possibility that plants in the tropics may be relatively resilient to rising temperatures [1]. Plant distributions across elevation gradients could also be more influenced by montane soil zonation than by temperature [40]. Likewise, if animal distributions were to shift upwards to a greater extent (or at a faster rate) than plant distributions, the loss of certain plant-based resources could make thermoregulation more difficult—the upper thermal critical limit is the point at which organisms need to increase metabolic rates in order to survive. Range shifts due to climate change could therefore not only influence individual species, but could also therefore have important effects on the rest of the tropical forest community.

Ethics

Research permissions were granted by the Sabah Biodiversity Centre, the Sarawak Forestry Corporation, Yayasan Sabah, the Forest Department Sarawak, the Malaysian Economic Planning Unit and local communities.

Data accessibility

Camera trap data are archived in the Dryad Digital Repository at: http://dx.doi.org/10.5061/dryad.13vk0 [41].

Authors' contributions

J.F.B. conceived of and led the project; J.F.B., J.M.-A., A.G. and H.B. led the fieldwork; O.E.H. and A.J.G. contributed to fieldwork; J.F.B. and M.S.-M. conducted the analysis; J.F.B. wrote the manuscript and all authors contributed to revisions.

Funding

Funding was provided by the Natural Science and Engineering Research Council of Canada, the Canadian Foundation for Innovation, the Denver Zoological Foundation, and Panthera; J.M.-A. was also supported by Ministry of Higher Education—FRGS990/2013(31) and NRGS/1087/2013(01).

Competing interests

We have no competing interests.