Differential vulnerability of key threatened mammals to climate and land cover changes in the Central Himalayas

Abstract

Anthropogenic climate change affects biological diversity by altering their suitable habitat ranges. The Himalayan region is one of the world's most sensitive biodiversity hotspots to global climate change. The Chitwan Annapurna Landscape (CHAL) in the central Himalayas serves as a vital north–south linkage among the protected areas in central Nepal and provides suitable habitats for threatened mammals in different ecological zones, such as snow leopards (in the alpine zone), Himalayan red panda (in the temperate zone), and one‐horned rhinoceros (in the lowland tropical zone). The biodiversity of CHAL is threatened by climate change and land use alterations. This study assessed the potential impacts of climate and land cover changes on the above three key threatened mammals in CHAL by employing maximum entropy (MaxEnt) modeling to predict the current potential habitat and project it for future climate change scenarios under different greenhouse gas concentrations. Further, we used the cellular automata and Markov Chain models to simulate and predict the temporal and spatial changes in land cover of CHAL. Our results indicate that the snow leopard and Himalayan red panda will experience more significant vulnerability than the one‐horned rhinoceros in all future climate scenarios. Approximately 36.3% and 41.8% of the suitable habitat of the snow leopard and 32.5% and 56% of the Himalayan red panda in CHAL are projected to be lost in 2050 and 2070, respectively, under representative concentration pathway (RCP6.0). Climate refugia, representing areas of suitable habitat for 2070 (under the RCP6.0) in CHAL, are projected to cover 958 km² (80.37% of the current range), 1052 km² (43.73% of the current range), and 2375 km² (58.21% of the current range) for one‐horned rhinoceros, Himalayan red panda, and snow leopard, respectively. Among the land cover attributes in CHAL, snow cover is predicted to decrease by 24% in 2070. Our findings indicate that species inhabiting alpine and temperate environments are more susceptible to human‐induced climate change than those inhabiting lowland tropical areas. These findings will help to implement the adaptation actions that are crucial to addressing future conservation challenges arising from climate and land cover change.

INTRODUCTION

Global climate change poses a threat to biological diversity (Gonzalez-Orozco et al., 2016; Sharma et al., 2008) and is a growing concern for biodiversity conservation. The effects of climate change on species distribution, diversity, and abundance vary across different ecosystems (Hansen et al., 2001; Harris et al., 2006; Nogués-Bravo et al., 2007). However, such impacts on wildlife and their habitats are not yet fully understood. Though it is not fully clear why, the Himalayan highlands are experiencing more pronounced temperature increases than the global average (IPCC, 2022; Shrestha et al., 2012), resulting in more extreme weather events in South Asia (Sharma et al., 2008). The central Himalayan region is highly vulnerable to climate-induced hazards such as droughts, floods, the invasion of invasive species, fires, and heat stress (IPCC, 2022; Lamsal et al., 2017). Nepal, located in the southern lap of the Himalayas, has a remarkable diversity of flora and fauna because of its location at the crossroads of two major biogeographic regions (the Indo-Malayan to the south and the Palearctic to the north) and extreme elevational gradients resulting in unique geography and climates (Bhattarai & Vetaas, 2006; Grytnes & Vetaas, 2002; NBS, 2002). The relatively high rate of warming in mountainous and mid-hill areas of Nepal Himalaya could have significant impacts on essential natural resources such as forestry and wildlife (MoFE, 2021).

Landscape approaches are initiated in Nepal to preserve biodiversity in heterogeneous areas that face challenges due to climate and land use alterations (WWF, 2013b). The Chitwan Annapurna Landscape (CHAL) is situated in the central Himalayas and extends across different physiographic zones, from the tropical lowlands of Terai (~200 m above sea level [asl]) to the high mountains of the Annapurna range (~8000 m asl), connecting the protected areas of the lowlands of Nepal with those of the higher Himalayas (Luitel et al., 2020; Thapa et al., 2015; WWF, 2013a). The landscape harbors habitats for several endemic faunas, such as the Himalayan field mouse (Apodemus gurkha Thomas 1924), Csorba's mouse-eared bat (Myotis csorbai Topál 1997), the Nepalese mountain vole (Neodon nepalensis), and the spiny babbler (Turdoides nepalensis) (BPP, 1995). In addition, CHAL is home to more than 104 mammals (BPP, 1995; WWF, 2013b), 500 birds (Baral & Inskipp, 2005), and 5930 plant species (WWF, 2013b) which account for approximately 50% of the mammals, 56% of the birds, and 91% of the plants recorded from Nepal. CHAL provides suitable habitats for key threatened mammals, such as snow leopards (Panthera uncia Schreber, 1775), the Himalayan red panda (Ailurus fulgens Cuvier, 1825), the one-horned rhinoceros (Rhinoceros unicornis Linnaeus, 1758), and the Bengal tiger (Panthera tigris Linnaeus, 1758), which are distributed across different eco-physiographic zones. Climate change can have different effects on the habitats of species based on their climatic and ecological requirements. For example, lowland Terai and mid-hill inhabitants of the Himalayan range might undergo elevational and latitudinal range shifts toward higher elevations under anthropogenic climate change, whereas the high elevation inhabiting species might suffer from range size contraction and run out of room (Freeman et al., 2018). Therefore, long-term and periodic research is essential to develop a more detailed understanding of the impacts of climate change on threatened species and their habitats.

Species distribution modeling (SDM) is an effective statistical tool for detecting the environmental factors that account for species distributions, mapping the potential distributions of species (overcoming the Wallacean shortfall) and estimating suitable areas for a species in the past or future (Araújo et al., 2011; Elith et al., 2005, 2006; Guisan & Thuiller, 2005; Pearson et al., 2004). SDMs are used to predict the distribution of a species across time and geographic space using two types of data, viz. species occurrence data and environmental variables (Elith & Leathwick, 2009). The models assume that the species is in equilibrium with its environment within its native range and its niche is conserved over time (Phillips et al., 2006). These models play crucial roles in quantifying species responses, particularly range shifts, in response to climate change (Araújo et al., 2011; Newbold et al., 2020). As the availability of extensive species occurrence data and environmental data layers increases, SDMs can model how species ranges have been and will be affected by ongoing environmental changes (Liu et al., 2020; Mi et al., 2016). These models have become the most well-known advanced methods for predicting distributions via species occurrence records (Elith & Leathwick, 2009; Sambrook & Russell, 2001).

Among various SDMs, maximum entropy (MaxEnt) is the most common because of its unique features, such as its ability to make robust predictions from presence-only records, flexibility with sparse or noisy input information, good predictive performance even with limited occurrence data, and easy implementation of algorithms in the public domain (Figueirido et al., 2012; Fourcade et al., 2014; Lande, 1988; Phillips et al., 2006; Phillips & Dudik, 2008). Earlier studies have used SDMs to model potential distributions and predict the impact of future climate change on mammals, birds, and reptiles in Nepal (Adhikari et al., 2023; Aryal et al., 2016; Baral, Kunwar, et al., 2023; Katuwal et al., 2023; Khanal et al., 2023; Sharma et al., 2020; Singh et al., 2020). The extent and specific impacts of climate change and land cover change on wildlife are still unclear; however, these impacts are expected to shift in distribution range, increase species invasion, and increase the probability of local extinction.

In this study, we selected three charismatic and globally threatened mammals—the snow leopard, Himalayan red panda, and one-horned rhinoceros—to examine the potential effects of climate and land use change on the fauna of the central Himalayan region. Each of these species is specialized to thrive in a distinct ecological zone within CHAL: snow leopards in alpine areas (3000–5500 m asl), Himalayan red pandas in temperate areas (2100–3800 m asl), and one-horned rhinoceros in tropical areas (110–850 m asl) (Jnawali et al., 2011). These three keystone species (Caro, 2010) represent the community in the respective ecological zone as model species for understanding the potential impacts of climate change in the CHAL. The snow leopard is the apex predator in alpine ecosystems and plays a key role in trophic niche functioning in alpine ecosystems, which are also facing the impact of climate change (Forrest et al., 2012). One-horned rhinoceros and Himalayan red panda are habitat and dietary specialists in tropical and temperate areas, respectively (Thapa, Hu, & Wei, 2018). Limited scientific evidence is available on the effects of environmental change on wildlife habitat in the central Himalayas, and long-term species-specific conservation plans are essential to address effective actions to cope with climate change-induced effects such as range shifts. The central Himalayan region, with an extreme altitudinal gradient (200–8000 m) within a narrow (<200 km) north–south span, provides ideal conditions for testing the differential effects of anthropogenic climate change on the elevational range shifts of the species inhabiting different elevational zones. The three threatened mammals selected for this study are important wildlife fauna that represent the faunal diversity of the different ecological zones of the CHAL in the central Himalayas. However, the potential impacts of climate and land cover changes on the habitats of these threatened mammals are poorly understood. It is hypothesized that the species inhabiting higher elevations are more vulnerable to anthropogenic climate change than those inhabiting lower elevations. Thus, we aimed to predict the impacts of climate and land cover changes on the potential habitats of the three key threatened mammals and identify climate refugia in the foreseeable future. The findings of this study will contribute to the formulation and implementation of species-specific action plans for the threatened mammals in the Himalayas.

METHODS

Study area

The CHAL (27°35′–29°33′ N, 82°88′–85°80′ E) serves as a north–south vertical linkage, connecting the Annapurna Conservation Area (CA), Manasalu CA, Langtang National Park (NP), and Shivapuri Nagarjun NP in the northern region with the Chitwan NP and Parsa NP in the southern region (Thapa et al., 2015; WWF, 2013a, 2013b) (Figure 1). CHAL covers an area of 32,068 km², which includes Siwalik (11.4%), mid-hills (37%), and mountains (50.8%), representing 22% of the land area of Nepal (WWF, 2013a). It covers the entire Gandaki River basin and includes major rivers such as Kali Gandaki, Seti, Marshyangdi, Daraundi, Budhi Gandaki, Trishuli, and Narayani/East Rapti (WWF, 2013a). It extends over 19 districts (Arghakhanchi, Gulmi, Palpa, Baglung, Parbat, Myagdi, Mustang, Syangja, Kaski, Tanahun, Lamjung, Gorkha, Manang, Rasuwa, Nuwakot, Dhading, Nawalparasi, Chitwan, and Makwanpur). The landscape ranges from lowland Terai (~200 m) to alpine high mountains and the cold and dry Trans-Himalayan region (above 4000 m) (Thapa et al., 2015). It consists of different bioclimatic zones, including lower tropical, upper tropical, lower subtropical, upper subtropical, temperate, lower subalpine, and upper subalpine bioclimatic zones. These bioclimatic zones are under the influence of climate change, with average temperature increases ranging between 0.022 and 0.051°C/year between 1970 and 2019 (Luitel et al., 2020). The landscape provides important habitats for Asian elephants (Elephas maximus Linnaeus, 1758), leopards (Panthera pardus Linnaeus, 1758), Himalayan red pandas, Himalayan black bears (Ursus thibetanus Cuvier, 1823), one-horned rhinoceros, snow leopards, sloth bears (Melursus ursinus Shaw, 1791), Bengal tigers, etc. (Jnawali et al., 2011; WWF, 2013a).

[IMAGE OMITTED. SEE PDF]

Species occurrence records and environmental variables

Occurrence records of the Himalayan red panda, snow leopard, and one-horned rhinoceros were compiled primarily from our own field surveys and from the published literature, gray literature, and national survey database (December 2019) (Ale et al., 2014; Bista et al., 2016, 2017; Chetri et al., 2019; DNPWC, 2009; Thapa & Puri, 2019; Thapa, Wu, et al., 2018). In total, we collected 98, 120 and 284 occurrence records for the Himalayan red panda, snow leopard, and one-horned rhinoceros, respectively (Figure 1). These records include both direct sightings and indirect evidence of feces, dung, scats, and pugmarks that were observed in the field. We deployed spatial filtering via the “SpThin” package (Aiello-Lammens et al., 2015) in R software (R Core Team, 2022) to minimize the spatial autocorrelation of sampling occurrence. We retained only one record per 1-km² grid, which included 66, 84, and 109 occurrence locations to model suitable habitats for Himalayan red panda, snow leopard and rhinoceros, respectively (Aryal et al., 2016; Thapa, Wu, et al., 2018).

Bioclimatic variables (19 variables; 11 temperature and 8 precipitation) with a cell size of 30 arcseconds (~1-km resolution) for the current (average for 1950–2000) and future (2050 and 2070) climatic conditions were downloaded from WorldClim () (Hijmans et al., 2005). These climatic layers include annual trends (mean annual temperature and precipitation), seasonality (annual range in temperatures and precipitation), and limiting environmental factors (temperature and precipitation of quarters) (Hijmans et al., 2005). Additionally, a digital elevation model (DEM) of the same resolution as the climate variables was downloaded from WorldClim, and the aspect and slope were derived from the DEM via ArcGIS 10.1 (ESRI, 2011). These bioclimatic variables have been used in several studies to predict the distributions of the Himalayan red panda (Panthi et al., 2019; Shrestha et al., 2021; Thapa, Wu, et al., 2018), one-horned rhinoceros (e.g., Pant et al., 2021) and snow leopards (Aryal et al., 2016; Li et al., 2016, 2021) in the Himalaya.

Modeling and evaluation

We performed variance inflation factor (VIF) analysis for the three species to minimize multicollinearity among the predictor variables, utilizing the “vif” function in “car” package (Fox & Weisberg, 2019) in R (R Core Team, 2022) following the methods of previous similar studies (Aryal et al., 2016; Lauria et al., 2015; Ranjitkar, Kindt, et al., 2014; Ranjitkar, Xu, et al., 2014; Thapa, Wu, et al., 2018). The VIF values were examined, with a threshold of 10 to determine strong collinearity, as VIF values greater than 10 can affect model performance (Quinn & Keough, 2002). Out of the 21 variables considered (19 bioclimatic and two topographic), we identified and used least correlated variables (VIF < 10) for the snow leopard (bio3, bio11, bio13, bio14, and slope) and Himalayan red panda (bio3, bio15, bio16, bio19, and slope) each and six variables (bio1, bio3, bio4, bio9, bio12, and bio17) for the rhinoceros for modeling (Appendix S1: Tables S1–S3). We used the MaxEnt algorithm, a robust and superior bioclimatic modeling approach that is best fit for presence-only data (Elith et al., 2006) to predict the current potential habitat and project it for the future under climate change. We predicted the impact of climate change in 2050 and 2070 via the Community Climate System Model (CCSM) to forecast future distributions via the same variables examined for the current distribution under the three greenhouse gas emission scenarios, referred to as representative concentration pathways (RCP2.6, RCP6.0, and RCP8.5). We chose the three RCPs representing the lowest emission (RCP2.6), most plausible emission (RCP6.0) (IPCC, 2013; Pielke et al., 2022) and the highest emission (RCP8.5) of greenhouse gases. Impacts of climate change on the species were studied through individual-species climate models with respect to the RCPs. Field-validated species presence and ecologically important variables were used to build a correlative bioclimatic model in MaxEnt v.3.4.4 (Phillips et al., 2024) and predict the spatial vulnerability of each species to climate change under three RCPs for 2050 and 2070. In the modeling process, 75% of the presence data of the species were used to build the model, and 25% were used for model validation. The predictive ability was tested with five replicates and validated based on the area under the curve (AUC) of the receiver operating characteristics. The Jackknife test was done to evaluate the contribution of environmental variables (Shcheglovitova & Anderson, 2013). The average logistic threshold value of the maximum training sensitivity plus specificity (MTSPS), an inbuilt functionality in MaxEnt, was used to distinguish between suitable and unsuitable regions (Liu et al., 2013). An AUC < 0.7 indicates poor model performance, 0.7–0.9 indicates moderate performance, and >0.9 indicates good performance (Pearce & Ferrier, 2000). Although AUC is a commonly used model evaluation parameter, it is influenced by the geographic extent of the models (Lobo et al., 2008). Therefore, we also used the thresholds-dependent method, that is, true skill statistics (TSS) to evaluate the accuracy of the model predictions (Allouche et al., 2006; Merow et al., 2013). TSS was calculated for all model outputs (0–4 replications) and the final TSS averaged from all five replicates.

Projected land use

Anthropogenic activities and climate change can lead to land use and land cover changes. Mapping land use changes can also reveal how the potential habitat may improve or degrade, indicating where the species could thrive (existing microrefugia/stable habitat) or disperse (new refugia/new habitat). A land cover database from the International Centre for Integrated Mountain Development (ICIMOD) () (Uddin et al., 2015) for 2002 and 2010, which is more precise than other large-scale global land cover databases, was downloaded. The future land cover changes for 2050 and 2070 were projected based on the land cover database of 2002 and 2010 as input in a Cellular Automata and Markov Chain model within TerraSet 18.21. Several studies have used Cellular Automata and Markov Chain models to predict land cover change scenarios (Huang et al., 2015; Kumar et al., 2014; Yuan et al., 2015; Zhao et al., 2017) because of their high predictive accuracy. Here, the Cellular Automata and Markov Chain model equation is $$$ S\left(t,t+1\right)=f\left(S(t),N\right) $$$ (Mondal et al., 2016), where S denotes the set of limited and discrete cellular states, N is the cellular field, t and t + 1 denote different times, and f is the transformation rule of the cellular states in the local space.

Vulnerability assessment

Changes in potential suitable habitat under the current and future climate scenarios were assessed by identifying vulnerable habitats, increased suitable habitats, and climate refugia under the following criteria:

• Vulnerable habitat is an area of habitat that is currently suitable and is predicted to be unsuitable under future climate scenarios.
• Increased suitable habitat is an area of habitat that is currently unsuitable and is predicted to be suitable under future climate scenarios.
• Climate refugia are areas of habitat that are currently suitable and are predicted to be suitable under future climate scenarios.

These three indicators were used to demonstrate the impacts of climate change on currently suitable habitats: (1) AC: suitable habitat change percentage; (2) SH_c: current suitable habitat loss percentage; and (3) SH_f: increased suitable habitat percentage under future climate scenarios (Duan et al., 2016; Li et al., 2017). Each indicator is expressed via the following relationship: $$$ \mathrm{AC}=\frac{A_{\mathrm{f}}-{A}_{\mathrm{c}}}{A_{\mathrm{c}}}\times 100\%, $$$ $$$ {\mathrm{SH}}_{\mathrm{c}}=\frac{A_{\mathrm{c}}-{A}_{\mathrm{c}\mathrm{f}}}{A_{\mathrm{c}}}\times 100\%, $$$ $$$ {\mathrm{SH}}_{\mathrm{f}}=\frac{A_{\mathrm{f}}-{A}_{\mathrm{cf}}}{A_{\mathrm{f}}}\times 100\%, $$$ where A_c is the projected area of current suitable habitat, A_f is the projected area of future suitable habitat, and A_cf is the area of climate refugia.

RESULTS

Predictors of threatened mammal distribution in the CHAL

The performance of the models for all three species was robust, with high values for AUC (all >0.92), as well as TSS (all >0.82) for all five replicates (Appendix S1: Figures S1–S3). These findings indicate that the models performed better than a random when the potential habitat suitability of these species was considered. The annual mean temperature, isothermality, and mean temperature of the coldest quarter were the dominant variables for predicting the suitable habitat of one-horned rhinoceros, Himalayan red panda, and snow leopard, respectively (Appendix S1: Figure S4). The mean temperature of the coldest quarter contributed the most (97.7%) to the model performance in determining the distribution of snow leopards in the alpine habitat range, followed by the precipitation of the wettest month (3.94%) (Figure 2a). Isothermality contributed the most (71.8%), followed by precipitation in the wettest quarter (23%), to the prediction of potential habitats for the Himalayan red panda in the Lesser Himalaya (Figure 2b). The annual mean temperature contributed the most (90%) to the prediction of suitable habitat for one-horned rhinoceros in lowland Terai, followed by temperature seasonality (5.4%) and isothermality (2.1%) (Figure 2c).

[IMAGE OMITTED. SEE PDF]

Projected distributions under current and future climates

The predicted current potential habitats for one-horned rhinoceros, Himalayan red panda, and snow leopard within the CHAL region are estimated to be 1192, 2417, and 4080 km², respectively (Table 1). The suitable areas for one-horned rhinoceros, Himalayan red panda, and snow leopard represent approximately 4%, 8%, and 13% of the total area of the CHAL. Currently, the predicted suitable habitat of one-horned rhinoceros is predominantly (75%) encompassed within the existing protected areas in lowlands (Chitwan NP and Parsa NP), and that of snow leopards (38.8%) is within the protected areas of higher mountains (Annapurna CA, Manaslu CA, and Langtang NP). However, a large portion (77%) of the predicted habitat of the Himalayan red panda exists outside the protected areas in the landscape.

TABLE 1 Predicted suitable habitats of three threatened mammals under current and future climate change scenarios in Chitwan Annapurna Landscape.

Species	Current suitable area (km2)	Future suitable area (km2)
RCP2.6	RCP6.0	RCP8.5
2050	2070	2050	2070	2050	2070
Snow leopard	4080	3823	3788	3483	3934	3204	3126
Himalayan red panda	2417	1221	1168	1674	1108	1089	984
One-horned rhinoceros	1192	2277	2294	2857	2684	3159	3080

We predicted the potential distributions of the three threatened mammals under RCP2.6, RCP6.0, and RCP8.5 and observed a consistent decrease in the future suitable habitats for the Himalayan red panda and snow leopard; however, the same would increase for the one-horned rhinoceros (Table 1). We presented detailed results with a focus on the most plausible greenhouse emission scenario (RCP6.0). Under RCP6.0, the potential habitat of snow leopards in CHAL is projected to decrease by 15% and 4% in 2050 and 2070, respectively, resulting in estimated areas of 3483 and 3934 km² (Table 1, Figure 3a–c). This reduction in habitat primarily affects the Langtang, Manang, Mustang, and Manaslu regions. Currently, the predicted habitat spans elevations ranging from 3600 to 5100 m, with an average elevation of 4400 m. However, in the future climate change prediction scenario (2070, RCP6.0), the elevation range is projected to shift between 4100 and 5600 m, with an average elevation of 4900 m. Most of the predicted habitats are concentrated at approximately 4800 m, both in the current and future projection scenarios. This suggests a contraction of the suitable habitat that remains relatively stable at this elevation range (Appendix S1: Figure S4).

[IMAGE OMITTED. SEE PDF]

The predicted habitat of the Himalayan red panda in CHAL will be reduced by 7% and 13% under RCP 6.0 in 2050 and 2070, respectively (Table 1, Figure 3d–f). The current predicted habitat is distributed between 2000 and 4100 m, with an average elevation of 3000 m, but the prediction increased in elevation to 5300 m, with an average of 3200 m in the future (2070). Habitat contraction is observed in the eastern (Rasuwa) and western (Mygadi) parts of the CHAL. A large part of the predicted habitat is concentrated within the elevational range of 3000–3300 m under the current future scenario, indicating that most of the suitable habitat is stable. Furthermore, the predicted habitat suitability peaks shifted toward higher elevation from the current values (Appendix S1: Figure S5).

The potential habitat of one-horned rhinoceros in CHAL is projected to increase under RCP6.0 in 2050 and 2070 (Table 1, Figure 3g–i). The habitat increases are expected to be primarily in the northern and northwestern regions of the current habitat in the lowlands toward the inner river valleys. The current predicted habitat range of rhinoceros is between 120 and 320 m in elevation, which is projected to increase up to 700 m in the future. This indicates an increase in the elevation of suitable habitat for rhinoceros in the future. The predicted habitat pixels of the density plot revealed that the maximum suitable habitat area peaked below 200 m in the current and future scenarios, indicating stable suitable habitat (Appendix S1: Figure S6).

Geographic features of the climate refugia

The impacts of climate change on the snow leopard and Himalayan red panda are expected to increase their vulnerability. In contrast, rhinoceros are less vulnerable. The current potential habitat of snow leopards is projected to decrease by 36.3% and 41.8% under RCP 6.0 in 2050 and 2070, respectively (Table 2). Approximately 60% of the predicted habitat between elevations of 3755 and 5630 m acts as climate refugia for snow leopards in the CHAL. The climate refugia for snow leopards are in Lomanthang, Neshyang, Narphu, Noshong, Chum Nurbi, Rubi Valley, Darche, and Langtang (Appendix S1: Figure S4). Approximately 32.5% and 56% of the Himalayan red panda habitat will be lost under RCP6.0 in 2050 and 2070, respectively. Approximately 1632 and 1052 km² will remain as climate refugia under RCP6.0 in 2050 and 2070, respectively (Table 2). The climate refugia of the Himalayan red panda are located between elevations of 2005 and 4700 m in the areas of Dhaulagiri range, Rubi Valley, Gosaikunda, and Helambu of Langtang NP (Appendix S1: Figure S5). The current suitable habitat of one-horned rhinoceros is projected to be less vulnerable in the future, indicating the stability of the majority of the habitat and the ability to act as climate refugia in the future (Table 2). Mostly, the climate refugia of the one-horned rhinoceros range in altitude from 120 to 337 m inside Chitwan NP (Appendix S1: Figure S6).

TABLE 2 Predicted changes in potentially suitable habitats for the study species in Chitwan Annapurna Landscape under the RCP6.0 for 2050 and 2070.

Species	Scenario (RCP6.0)	Ac (km2)	Af (km2)	Acf (km2)	Ac (%)	SHc (%)	SHf (%)
Snow leopard	2050	4080	3483	2600	−14.6	36.3	25.4
2070	4080	3934	2375	−3.6	41.8	39.6
Himalayan red panda	2050	2417	1674	1632	−30.7	32.5	2.5
2070	2417	1108	1052	−54.2	56.5	5.0
One-horned rhinoceros	2050	1192	4157	1190	248.8	0.10	71.4
2070	1192	2884	958	142.0	19.6	66.8

Predicted land cover changes in CHAL

Most of the land cover attributes, with the exceptions of snow cover and water and barren areas, are expected to increase in the future compared with the initial land cover inputs (2002 and 2010). Urban areas are expected to increase by 20%, followed by shrubs (8.8%), agriculture (3.6%), and forests (3.5%) in CHAL. Notably, other land cover attributes, including snow cover (24%), water (4.7%), and barren areas (4.3%), are likely to decrease in 2050 (Figure 4). By 2070, urban areas are projected to increase by 37%, followed by shrubland (15%), grassland (8.9%), forest (8.3%), barren land (7.5%), and water bodies (4.7%), whereas snow cover is expected to significantly decrease (Appendix S1: Figure S7).

[IMAGE OMITTED. SEE PDF]

DISCUSSION

Animals exhibit differential responses to climate change due to specific climate requirements and tolerances (Graham et al., 1996; Thuiller et al., 2011). SDM tools are valuable for predicting the current ecological niche of a species and projecting its future distribution under various scenarios, based on the principle of niche conservatism (Wiens et al., 2010). Projecting species niches in the context of future climate change provides essential insights for conservation, ecosystem management, policy planning, and adaptation strategies and ultimately helps to minimize risks and enhance the resilience of biodiversity in a changing climate (Schwartz, 2012). This study tested whether the threatened keystone mammals (Himalayan red panda, one-horned rhinoceros, and snow leopard) in the CHAL in the central Himalayas would exhibit varied responses to future climate and land use change due to their unique climatic space in different bioclimatic zones. We observed a prominent effect of climate change on the future distribution of threatened mammals in the landscape, but the effect was not uniform across the species considered. The results revealed that high-altitude species such as snow leopards and the Himalayan red panda are more vulnerable to climate change than one-horned rhinoceros dwelling at lower altitudes. The greater vulnerability of mammals residing at higher elevations might be attributed to the potential rapid increase in temperature at higher altitudes than at the lowlands of the Himalayas (Luitel et al., 2020).

Climatic variables associated with threatened mammals in CHAL

The Himalayan red panda, one-horned rhinoceros, and snow leopard are the keystone threatened species that are suffering from habitat loss and hunting (WWF, 2013a). We employed MaxEnt SDM to predict the potential distributions of the three mammals in CHAL. Our findings suggest that the mean temperature of the coldest quarter plays an important role in determining the potential habitat of snow leopards in CHAL. The optimum mean temperature of the coldest quarter ranging between −15 and 5°C defines the suitable habitat for snow leopards (Appendix S1: Figure S8C). Previous studies have shown that the annual mean temperature, mean temperature of the warmest quarter, mean temperature of the coldest quarter, annual precipitation, elevation, and landcover attributes are important variables for predicting the distribution of snow leopards (Aryal et al., 2016; Li et al., 2016; Shrestha & Kindlmann, 2022). High-altitude environments in the Himalayas are characterized by extremely harsh climates consisting of low temperatures and low oxygen pressures (Shrestha et al., 2012). High elevations with rugged mountain terrains, low precipitation in warm annual quarters, and low annual average temperatures have created large alpine meadows as suitable habitats for snow leopards (Farrington & Li, 2016); however, topographic complexity and the microclimate may limit their distribution locally.

Precipitation-associated variables were dominant in predicting the Himalayan red panda distribution. Annual precipitation, precipitation of the wettest quarter, precipitation of the coldest quarter, precipitation seasonality, and precipitation in the driest month play vital roles in the distribution of Himalayan red pandas (Hu et al., 2020; Panthi et al., 2019; Shrestha et al., 2021; Thapa, Wu, et al., 2018; Thapa et al., 2020). Our results revealed that isothermality defines the suitable habitat for Himalayan red pandas (Appendix S1: Figure S8A). The Himalayan red panda inhabits the high-altitude temperate forest, where they are adapted to cool and moist conditions (Pradhan et al., 2001; Thapa, Hu, & Wei, 2018; Wei & Zhang, 2011; Yonzon & Hunter, 1991). The isothermality in Himalayan red panda habitat may influence their ability to maintain consistent body temperatures and metabolic rates. A high-isothermality environment, characterized by minimal temperature fluctuations, may be especially beneficial for the Himalayan red panda, as it alleviates the stress caused by extreme temperature changes. The central and eastern regions of Nepal experience heavy rainfall during the monsoon season, resulting in moist and humid conditions that support the growth of bamboo and bamboo shoots, which serve as a staple food for red pandas (Bista et al., 2019; Thapa, Wu, et al., 2018). The Himalayan red panda is recognized as a habitat and dietary specialist that primarily depends on a bamboo diet and dwells in the bamboo understories of Himalayan broadleaf and coniferous forests (Thapa, Hu, & Wei, 2018; Yonzon & Hunter, 1989). Owing to specific environmental constraints or changes, habitat and diet specialist species may face an increased probability of extinction at both local and global scales.

Our models emphasized that the annual mean temperature was the major variable with the greatest contribution to the prediction of suitable habitats for one-horned rhinoceros. The mean annual temperature ranging between 23 and 26°C predicted the most suitable habitats for the one-horned rhinoceros (Appendix S1: Figure S8B). Temperature and precipitation-related variables were shown to contribute significantly to predicting the habitat suitability of the rhinoceros (Pant et al., 2021). The mean temperature of the driest quarter, mean annual temperature, and annual precipitation have been identified as the critical factors associated with predicting the distribution of one-horned rhinoceros (Pant et al., 2021). In addition to climatic variables, environmental correlates such as distance from grasslands, distance from wetlands, and slope play crucial roles in determining the potential distribution of grasslands (Adhikari & Shah, 2020; Pant et al., 2020a, 2021, 2022; Subedi, 2012). Currently, the known distribution of one-horned rhinoceros in Nepal extends between the elevation range of 100 and 500 m (DNPWC, 2009, 2017a; Pant et al., 2021), and those areas are hot and humid, creating suitable habitats for the megaherbivore of the lowland in CHAL (WWF, 2013a).

Distribution of threatened mammals under climate change

We projected the current climatic niches of snow leopards, Himalayan red panda, and one-horned rhinoceros in CHAL for the future (2050 and 2070) under three different greenhouse gas emission scenarios (RCP2.6, RCP6.0, and RCP8.5). Our results revealed that the potential habitat of snow leopards in CHAL represents 19.5% of the earlier estimates within the country (approximately 12,815 km² [DNPWC, 2017b]; 19,945 km² [Li et al., 2016]; 20,000 km² [Forrest et al., 2012]; and 22,625 km² [Aryal et al., 2016]) and 13% of the CHAL. These potential habitats of snow leopards in CHAL are distributed in Annapurna CA, Manaslu CA, and Langtang NP. Under all the future climate change scenarios, there will be a reduction in the habitat of snow leopards in CHAL in 2050 and 2070. Aryal et al. (2016) predicted that the potential habitat of snow leopards in Nepal would decrease by 14.57% in 2030 and by 21.57% in 2050, whereas Li et al. (2016) anticipated a 47% loss of habitat by 2070. The habitat of snow leopards is predicted to decline by 30% in the southern distribution range due to an upward shift in the tree line and subsequent shrinkage of the alpine zone (Forrest et al., 2012). However, on a global scale, the suitable snow leopard habitat will increase by 20% in 2080 (Farrington & Li, 2016). Owing to variations in spatial coverage and methodological approaches, there are discrepancies in potential habitat estimation among different studies.

The Himalayan red panda habitat is predicted to decrease remarkably in the future due to climate change. Our results revealed that the potential habitat of the Himalayan red panda in CHAL was greater than that reported by Bista et al. (2017), possibly because we used spatial occurrence records of the species that were not included in that study. Prior studies estimated an area of approximately 21,680 km² (Thapa, Wu, et al., 2018), 22,400 km² (Kandel et al., 2015), and 23,977 km² (Bista et al., 2016) as the habitat of the Himalayan red panda in Nepal, approximately 11% of which lies in the CHAL. A large proportion of the suitable habitat for the Himalayan red panda is located in the eastern part of the country, where it gradually decreases from east to west (Thapa et al., 2020). Our results indicated that approximately 32.5% and 56% of the habitat in CHAL will be lost due to climate change in the future. Panthi et al. (2019) predicted an increase of 6.5% in suitable habitat for the Himalayan red panda in Nepal by 2070 based on climate data alone; however, when land use and land cover changes were combined, a potential loss of 0.5% was estimated. In future projections, the distribution of the Himalayan red panda population is likely limited by the continuous expansion of regional agriculture and the ecological or physiological limitations of bamboo (Lyon et al., 2022). The future projection still reveals a suitable climate within much of the 2200–4803-m range, which is congruent with our results (Lyon et al., 2022).

We observed that the potential habitat of the one-horned rhinoceros (currently 1192 km²) will increase in 2050, reaching a peak of 4175 km² and then starting to shrink to reach 2884 km² in 2070 under RCP6.0. However, Pant et al. (2020b) estimated that 35% of the currently suitable habitat of the one-horned rhinoceros will be lost by 2070 under the SSP8.5 climate change scenario, and Adhikari and Shah (2020) also predicted a decrease of more than 50% by 2070 under RCP8.5. Both previous studies used Nepal as the background area for modeling, although the occurrence points were mainly concentrated in Chitwan NP. Additionally, those predictions were based on the highest greenhouse emission scenarios, which are considered unlikely. Our predictions using the most plausible scenario of greenhouse gas emission and only the CHAL as the background area might have provided a more reliable estimate.

Climate refugia for threatened mammals and land cover attributes

The diverse climate and topographic complexity in the Himalayas have created different microclimatic spaces, contributing to the formation of climate refugia for the species. Thus, the identification of past and future refugia is now considered necessary for wildlife management in light of anthropogenic climate change (Lenoir et al., 2008; Noss, 2001), and developing methodologies for their identification and description is a high research priority (Steffen, 2009). Species distribution models (Elith & Leathwick, 2009; Guisan & Thuiller, 2005) are widely used in biogeography and phylogeography to predict species distributions and have been applied to identify climate refugia (Fløjgaard et al., 2009; Huang et al., 2015). In the alpine areas of CHAL, 60% of the predicted habitat ranges between 3755 and 5630 m, with an average elevation of 4705 m, and acts as a climate refugia for snow leopards. Previous large-scale studies have indicated the existence of refugia in high Asia due to the unique mountain environment, which maintains a relatively constant arid or semiarid climate (Li et al., 2016). These arid climates prevent forest and glacier formation, keeping stable alpine steppes functioning as climate refugia for other alpine animals (Li et al., 2016). Similarly, the forest understory of bamboo between 2000 and 4700 m, with an average elevation of 3100 m, is distributed in the area, particularly in the Dhaulagiri range. Rubi Valley, Gosaikunda, and Helambu in Langtang NP are climate refugia for the Himalayan red panda. At low altitudes, most of the suitable habitat of the rhinoceros remains stable in the future, acting as climatic refugia. However, potential habitats of the rhinoceros will expand along the river valleys of the western part of Chitwan NP in the future. The river valley consists of water sources and grasslands, which create favorable conditions for rhinoceros in the future, as predicted by (Pant et al., 2021).

Among the land cover attributes in CHAL, snow cover is likely to decrease remarkably. How species' responses to climate change are correlated with their response to land use change is poorly understood in the Himalayas. Conversely, land use and land cover changes can also influence climate change, resulting in a loss of forest cover from increased climate-related disturbances, the expansion of woody vegetation into grasslands, and changes in surface albedo (Wickham et al., 2018). Land use change and climate change are driving a global biodiversity crisis. One apparent effect of climate warming in mountain environments is the upward shift in species distributions (Parmesan & Yohe, 2003). With climate change, higher elevation areas of Nepal are becoming suitable habitats for snow leopard, which is expected to push large cats into new areas and create more problems (Baral, Adhikari, et al., 2023). For snow leopards, changes in the distribution and abundance of primary prey resources are crucial (Aryal et al., 2016). Climate change enables forests to move to higher elevations, occupying present-day grasslands, the main habitats of snow leopards. Therefore, the snow leopard habitat has shrunk and has experienced more significant fragmentation. Climate change also increases the vulnerability of grasslands to degradation, thus threatening the prey availability of snow leopards (Shen, 2020).

The distribution of red pandas in Nepal is influenced by landscape variables such as elevation, distance to water sources, and bamboo cover (Thapa et al., 2020). An assessment of human impacts on endangered species living in high-altitude regions in Nepal revealed that human activities such as livestock grazing, fuelwood collection, and forest clearing have a negative impact on Himalayan red panda habitats (Panthi et al., 2019). The northern aspect of the Himalayas has more moisture with greater canopy cover than the southern aspect does, and these two factors are important habitat factors associated with red pandas (Thapa et al., 2020). One-horned rhinoceros largely depend on the floodplain grass Saccharum spontaneum, particularly during the monsoon season (Pradhan et al., 2008). Grasslands in alluvial floodplains, which are the prime habitats of one-horned rhinoceros, have been shrinking due to invasive plants (Subedi, 2012). Studies on one-horned rhinoceros are limited, and more focus needs to be provided to understand the impact of climate change.

This study has some limitations which should be considered while interpreting the results. We employed bioclimatic and topographic variables only to model the current and future potential distribution of key threatened species in CHAL by MaxEnt SDM. SDM can enhance our understanding of how climate change affects different species, but it has its limitations (Lissovsky et al., 2021). Some of these shortcomings arise from the data constraints, underlying assumptions, and the static nature of the model (Elith & Leathwick, 2009). Accurate species occurrence data are vital for the reliability of the model output. We employed occurrence data of the species from our field work in some parts of the CHAL and supplemented it with data from secondary sources. Projections of future climates are subject to significant uncertainty due to variability in emission scenarios and climate feedback introducing additional uncertainty on future projections. To improve estimates of climate change impacts on suitable habitats and climate refuge of the studied species, we combined species distribution models with land use land cover changes. Future studies with comprehensive field surveys across the species range and use of high-resolution data on climate and environmental variables, habitat and land use data, and information on species interactions, dispersal, and genetic traits are essential.

CONCLUSION

Our study highlights that climate and land use changes can have severe impacts on species distributions in the future and that these impacts can vary according to the species and their current ecological niches. Species from alpine and temperate ecological zones are more vulnerable to the consequences of climate change than those from lowland tropical zones. Current management regimes may require special attention to understand and incorporate these habitat dynamics for more climate-resilient planning and conservation. The current potentially suitable habitat covered 4%, 8%, and 13% of the CHAL for the one-horned rhinoceros, red panda, and snow leopard, respectively. Almost the entire habitat of snow leopards and rhinoceros encompasses protected areas (Chitwan NP, Annapurna CA and Manaslu CA), but the habitat of the Himalayan red panda lies outside the protected area system of the country. Both snow leopards and Himalayan red pandas are vulnerable to future climate change scenarios due to decreased habitat, in contrast to rhinoceros, which will be relatively less vulnerable and will experience increased habitat. Under future climate change, 1190, 2375, and 1052 km² in CHAL will act as climate refugia for the rhinoceros, snow leopard, and Himalayan red panda, respectively. Most land cover attributes are likely to increase in the future; however, snow cover is likely to decrease. Our findings can help local or provincial government bodies develop new conservation strategies to combat the future challenges of climate change. Climate change adaptation actions should be implemented in areas that have been identified as climate change refugia. This study recommends that conservation-related institutions pay attention to land use planning in the future to mitigate the potential negative impacts on these key species. Future studies should focus on areas where microrefugia are aggregated and isolated.

AUTHOR CONTRIBUTIONS

Arjun Thapa, Shanta Raj Jnawali, and Kanchan Thapa conceived and supervised the project. Arjun Thapa, Rabin Bahadur K. C., Hari Basnet, Rima G. C., Kapil Khanal, and Rajan Prasad Paudel conducted field work, data collection, and preparation. Arjun Thapa, Laxman Khanal, Suraj Baral, and Gokarn Jung Thapa performed the analysis. Arjun Thapa, Rabin Bahadur K. C., Rajan Prasad Paudel, and Laxman Khanal wrote the manuscript with input from all coauthors.

ACKNOWLEDGMENTS

We would like to thank the Department of National Parks and Wildlife Conservation (DNPWC) for granting permission to conduct activities in CNP, LNP, and ACA. Arjun Thapa is thankful to the President's International Fellowship Initiative (PIFI) program of the Chinese Academy of Sciences, China. WWF-Nepal Hariyoban Program-II supported the field work.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

DATA AVAILABILITY STATEMENT

Data (Thapa et al., 2025) are available from Dryad: .

References

Adhikari, A., and D. N. Shah. 2020. “Potential Impact of Climate Change on One‐Horned Rhinoceros (Rhinoceros unicornis) in Nepal.” bioRxiv. [DOI: https://dx.doi.org/10.1101/2020.05.04.076562].