Discussion
Our analyses of the effects of projected near-future climate change on bats in South Asia show a general pattern of moderate potential loss in all scenarios with relatively high degrees of retention in climatic suitable areas. As expected, smaller losses were projected under the ‘middle-of-the-road’ SSP2-RCP4.5 scenarios compared to the more pessimistic ’fossil-fuelled development’ SSP5-RCP8.5 scenarios, reinforcing the need for changing behaviour and avoiding business as usual (Peters et al., 2013). While loss was moderate, our results also revealed variability among species and scenarios with winners and losers. For example, Myotis csorbai , a species restricted to Nepal, was projected to lose all its current suitable area in some scenarios, while Saccolaimus saccolaimus , a species with a very wide but fragmented distribution in South and Southeast Asia, was projected to double its current suitable area in some scenarios. Previous work has also reported varied responses of bats to climate change including negative impacts (Rebelo et al., 2010; Hughes et al., 2012; Bellard et al., 2013; Thapa et al., 2021), no or limited effects (Bandara et al. 2022), and positive effects (Costa et al., 2018; Hayes & Piaggio et al., 2018; Thapa et al., 2021). This variation likely reflects differences in niche plasticity, robustness to changing conditions, and migratory propensity among species. Additionally, variation can occur due to data limitations and modelling choices. Indeed, our results showed uncertainty in projections for some species: many species involved small (< 30 occurrences) sample sizes, and the minimum sample size of five occurrences has been used before but may not be fully reliable (Pearson et al., 2007). However, most advanced ENM algorithms, especially techniques such as MAXENT, can account for fewer occurrences in species with low prevalence and narrow ranges (Pearson et al., 2007; van Proosdij et al., 2016; Morales et al., 2017), and ensemble modelling can provide higher predictive accuracy by accounting for inter-model variation. Capturing and reporting this variation by considering various scenarios, scales, and algorithms is important to ensure conservation and management recommendations are not misguided due to over- or under-fitting and a lack of reliability in results.
Similar to other studies on bats (in Europe by Rebelo et al., 2010; in Southeast Asia by Hughes et al., 2012; in Nepal by Thapa et al., 2021), we also found a trend for likely shifts towards higher latitudes in climatically suitable areas. In SSP2-45 scenarios, suitable areas for most species were projected to be located north-northeast or northwest from current areas; in SSP5-85 scenarios the shift was more generally northeast. Overall, climate change is expected to lead to latitudinal shifts towards polar regions, but arguably, our projected shifts could also reflect the geographic configuration in our study area, where the Southern region is mostly coastal. However, this is unlikely the reason for the observed trend, as for species currently in northern and central areas we also did not generally find projected southward shifts. Future suitable areas were also usually not located far from current suitable areas, and it is likely bats could track changes in climatic suitable in many cases. However, there are several barriers that could limit shifts, including the Himalayas in the north and northeast, the Thar Desert and the Great Rann of Kutch in central India, and the hill ranges of south and central India. Similarly, moving across large tracts of water may prevent tracking of climatically suitable areas in some species. Importantly, even if there are no strong geographical barriers, movement may be prevented due to lack of other abiotic and biotic resources or variation in dispersal behaviour.
Combining information from individual species we identified current and future suitability hotspots - areas of climatic suitability for >30% of study species. The current hotspots aligned with the four biodiversity hotspots in the region defined by Myers et al. (2000), but representation varied. While a large proportion of the current suitability hotspots were projected to be retained into the future, projected losses outweighed areas of projected gain. In the Himalaya and Indo-Burma hotspots, projected losses would lead to high fragmentation and isolation between hotspots (especially under the SSP2-RCP4.5 scenarios). Movement corridors and continuous protected area networks are likely to be critical to allow movement and occupation of new suitable areas. Interestingly, new suitable areas were consistently identified nearby rivers, around the northern regions of the Indus valley, northwestern Pakistan, and southern Gujarat. The small areas of suitability in southern Myanmar were projected to increase only in the HadGEM3 scenarios, and in Had5-85 there were a few additional isolated fragments of projected gain in suitability in Bangladesh and northeast India, regions with typically moist deciduous and wet evergreen habitats (Champion & Seth, 1968; Olson et al., 2001), quite similarly to southern Myanmar. The Western Ghats suitability hotspot was mostly retained and contiguous across all scenarios, and increases were restricted to small areas on the margins, showing that the Western Ghats are likely to remain stable as a suitability hotspot into the near future. Decreases were consistently seen in the Nilgiri hills and in northern Maharashtra - both regions with unique vegetation and habitat structures, ranging from semi-arid scrublands in the north of the Western Ghats to tropical moist deciduous forests in the Nilgiris. Many regions in peninsular India which are known to be specific in geography and habitat consistently showed patterns of projected loss in suitable areas. Combined with a growing understanding of the effects of climate change in the Himalayas and Western Ghats, and how it impacts biodiversity (Srinivasulu et al., 2021; Thapa et al., 2022), our results showing high degrees of retention combined with severe losses and minimal gains further support the importance of these regions.
We evaluated projected impacts of climate change using ensemble ecological niche modelling, an approach with good predictive accuracy that captures uncertainty from algorithm choice (Hao et al., 2020). However, this method has some limitations. First, ensemble ENM requires considering the balance between selecting many algorithms and the increased computational time and resources. Generally, the suggestion is that if a smaller number of algorithms is used, the ones with higher predictive accuracy and robustness are selected, as done here (Drake, 2014). A second limitation is the challenge to generate suitable pseudoabsences for presence-only occurrence data (Engler et al., 2004; Lütolf et al., 2006; Barbet-Massin et al., 2012). Our method for generating pseudoabsences incorporates random sampling within geographic limits based on spatial resolution and scale of the analysis and estimated foraging distances of study species. We aimed to balance statistical rigour and ecological realism while working with the constraints of lacking accurate species-level data on bat ranging and movement in South Asia. Further analyses for species with available data could incorporate information on ecological distance and environmental profiling as additional limiting factors to pseudoabsence generation (Iturbide et al., 2015). Model certainty is also an issue with ENM that greatly impacts the reliability and interpretability of these results. For instance, our analysis showed high certainty of climatic unsuitability in most of the study area (Supplementary Material 3). However, it is important to note that per-species results are much more informative and broad interpretations must be very cautious. Finally, another limitation of ENM is the need to carefully select variables to balance ecological importance, methodological constraints including the effects that multicollinearity and collinearity shift have on different algorithms, and model transferability (Feng et al., 2019). Here, we also considered this balance selecting robust methods, reducing variables to avoid high collinearity while prioritising ecologically relevant information.
Understanding species responses to rapid climate change is vital in conservation planning, especially in regions with high biodiversity and rates of endemism (Warren et al., 2008; Quintero & Wiens, 2013; Raman et al., 2023). This study is an initial assessment of potential effects of near-future climate change on bat species in South Asia, offering reasons for concern but also some hope. While climatically suitable areas may be reduced, many currently suitable areas are likely to remain, and shifts may be within the dispersal potential of many species. Nonetheless, retention of suitable areas and moderate loss does not necessarily ensure population persistence. Our initial assessment of impacts focuses on abiotic climate effects, but climate change may influence habitat and food resources differently and may result in climatically suitable areas being effectively unable to support healthy bat populations. Analysing climate effects alone ignores the potential combined effects of climatic and ecogeographic impacts on species distributions (Newbold, 2018), but it is a first step to understand potential changes in areas where climate and ecogeographic data may not be of the same quality, interpretability, or accessibility, possibly biasing ecological interpretations. Future climatically suitable areas may also be unsuitable due to geology and topography, factors linked to human activities including human land use, human population densities, and proximity to various infrastructures and developed areas. However, while it is possible to model both types of factors together (e.g. Hughes et al., 2012; Simões & Peterson, 2018; Raman et al., 2023), the lack of accurate data on their interactions when it comes to bats in this region could lead to overfitting, erroneous predictions, and misinterpretations across highly diverse species and functional groups (Araújo et al., 2007; Peterson et al., 2011; Fordham et al., 2012; Newbold, 2018; Simões & Peterson, 2018). It is important to analyse the effects of climate (and future climate change) and ecogeography independently as well as together, and also study their interactions, in order to truly understand the drivers of suitability for each species. Future work that considers these additional factors and the current intactness of future suitable areas would be important to inform conservation actions for climate change mitigation. In addition, it is important to consider that climatic niche modelling based on data from current occupied areas may not reflect the full climatic limits or niche of a species or its capacity to adapt (Hoffmann & Sgrò, 2011). Our results offer a first evaluation that highlights the need to further study climate change impacts in megadiverse regions such as South Asia and to develop robust conservation plans that integrate this information. Effective conservation requires the integrated study of species responses to climate, biotic, geographic, and anthropogenic factors, and effective communication with and outreach to policymakers and stakeholders at all levels.