In this study, I built species distribution models (SDMs) for three mammal species in Kon Plong District, Kon Tum Province, Vietnam, to predict potential habitat suitability and identify key environmental variables influencing their distributions. The district of Kon Plong is located in the Central Highlands of Vietnam, at the heart of the Kon Tum Plateau. With approximately 80% of the district still covered in relatively intact natural forests, Kon Plong represents one of the largest remaining areas of contiguous forest outside of Vietnam’s protected area network (Wearn et al., 2021). Recent biodiversity surveys have recorded 955 species in the district, including 29 listed as ‘threatened’ on the IUCN Red List (IUCN, 2024). Kon Plong is home to the world’s largest population of the Critically Endangered grey-shanked douc langur (Pygathrix cinerea) and Vietnam’s largest population of the Endangered northern yellow-cheeked gibbon (Nomascus annamensis) (Wearn et al., 2021). Through species distribution modelling, I aimed to identify current and potential habitat areas to inform proactive conservation and restoration efforts.

I selected three “umbrella” species that differ in their dependence on forest structure: the northern yellow-cheeked gibbon (N. annamensis), Owston’s palm civet (Chrotogale owstoni), and grey-shanked douc langur (P. cinerea). N. annamensis is a highly arboreal primate dependent on continuous canopy cover for locomotion. Roads and narrow canopy gaps can act as barriers, with crossings generally occurring only where canopy connectivity remains high (Asensio et al., 2021). This species, therefore, represents other arboreal taxa with similar ecological requirements. In contrast, C. owstoni is a terrestrial quadruped tolerant of habitat degradation and capable of inhabiting fragmented or disturbed areas (Nguyen et al., 2022). It was used as a representative of more generalist terrestrial species. P. cinerea occupies an intermediate ecological niche, employing both arboreal and terrestrial movement patterns (Bailey et al., 2017). Together, these species capture a range of habitat dependencies, allowing for comparative modelling of ecological niches within the same landscape.

I built an SDM for each species using Maxent v. 3.4.1. Maxent was selected for its high predictive performance and capacity to model presence-only data (Su et al., 2021). The software applies the principle of maximum entropy to estimate the probability distribution of suitable habitat based on known occurrence records and environmental predictors (Phillips & Dudík, 2008).

I used nine ecological variables relating to vegetation, topography, hydrology, and anthropogenic factors at a resolution of 30 m to create each MaxEnt model. Tree height and canopy cover data from Potapov et al. (2021) were used as the vegetation variables. These are global layers based on Landsat satellite data and, for tree height, Lidar data from the Global Ecosystem Dynamics Investigation instrument (Potapov et al. 2021). For the topographic variables, I used a digital elevation model (DEM) from the Shuttle Radar Topographic Mission (Farr et al., 2007). I calculated the slope and aspect of the study area with ArcGIS 10.8 using the DEM. The Euclidean distance from all main rivers and large bodies of water was used as the hydrological variable. The anthropogenic variables were Euclidean distance from human populated areas using datasets from WorldPop (Tatem, 2017), land-use datasets from the Earth Observation Research Center (Phan, et al., 2021), and remoteness. The remoteness variable was a proxy for hunting pressure, calculated in the ‘movecost’ R package using Tobler's off-path hiking function (Alberti, 2019). The focal points input to movecost, which act as the points of entry into the forest for hypothetical hunters, were created by dividing roads in the study area into a series of points separated by 1 km.

I assessed multicollinearity among predictor variables using the Band Collection Statistics tool in ArcGIS 10.8, calculating Pearson’s correlation coefficient (r) between all variable pairs. No strong correlations (r > 0.7) were detected, so all nine variables were retained for modelling.

Each species’ model was run using 20 cross-validated replicates, ensuring all occurrence data contributed to both training and testing to improve predictive reliability (Hijmans, 2012). I generated multiple candidate models for each species by varying the regularisation multiplier (0.5, 1, 1.5, 2) to evaluate model performance and avoid overfitting (Radosavljevic & Anderson, 2013). All other Maxent parameters were left as default.

Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUC) metric, a standard measure of predictive accuracy for SDMs. Models with an average AUC value below 0.7 were considered insufficient for reliable habitat prediction (Phillips et al., 2006). The best-performing model for each species was selected for further spatial interpretation of suitable habitat distribution within Kon Plong and its surrounding landscape. Areas of low, medium, and high suitability were defined using thresholds of 0.2, 0.5, and 0.7, respectively, following SDM methods from Thapa et al. (2018).  

Owston’s civet was predicted to have the most expansive distribution of the three species, while the two primate species were predicted to have a fragmented distribution with the largest areas of suitable habitat occurring in the core of Kon Plong District, centred on Mount Ngoc Boc, and directly south-east of Ngoc Linh – Kon Tum Nature Reserve, in Tu Mo Rong District.

References

Alberti, G. (2019) movecost: An R package for calculating accumulated slope-dependent anisotropic cost-surfaces and least-cost paths. SoftwareX, 10, 100331.

Asensio, N., Kachanan, J., Saralamba, C. & José-Domínguez, J.M. (2021) The impact of roads on the movement of arboreal fauna in protected areas: the case of lar and pileated gibbons in Khao Yai National Park, Thailand. Journal of Tropical Ecology, 37, 276–285. Cambridge University Press.

Bailey, K.E., Lad, S.E. & Pampush, J.D. (2017) Functional morphology of the douc langur (Pygathrix spp.) scapula. American Journal of Primatology, 79, e22646.

Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R., Hensley, S., et al. (2007) The Shuttle Radar Topography Mission. Reviews of Geophysics, 45.

Hijmans, R.J. (2012) Cross‐validation of species distribution models: removing spatial sorting bias and calibration with a null model. Ecology, 93, 679–688.

IUCN (2024) The IUCN Red List of Threatened Species. Version 2024-1 https://www.iucnredlist.org/species/39827/17941672 [accessed 28 December 2024].

Nguyen, A. T., Nhung, C. T. H., Galante, P. J., & Lê, M. Đ. (2022). Rapid decline and fragmentation of the distribution of an enigmatic small carnivore, the Owston’s Civet, in response to future climate change. Frontiers of Biogeography, 14(1). https://doi.org/10.21425/f5fbg53201

Phan, D.C., Trung, T.H., Truong, V.T., Sasagawa, T., Vu, T.P.T., Bui, D.T., et al. (2021) First comprehensive quantification of annual land use/cover from 1990 to 2020 across mainland Vietnam. Scientific reports, 11, 9979. Nature Publishing Group UK London.

Phillips, S.J. & Dudík, M. (2008) Modeling of species distributions with MaxEnt: new extensions and a comprehensive evaluation. Ecography, 31, 161–175.

Phillips, S.J., Anderson, R.P. & Schapire, R.E. (2006) Maximum entropy modeling of species geographic distributions. Ecological modelling, 190, 231–259. Elsevier.

Potapov, P., Li, X., Hernandez-Serna, A., Tyukavina, A., Hansen, M.C., Kommareddy, A., et al. (2021) Mapping global forest canopy height through integration of GEDI and Landsat data. Remote Sensing of Environment, 253, 112165. Elsevier.

Radosavljevic, A. & Anderson, R.P. (2014) Making better MaxEnt models of species distributions: complexity, overfitting and evaluation. Journal of Biogeography, 41, 629–643.

Su, H., Bista, M. & Li, M. (2021) Mapping habitat suitability for Asiatic black bear and red panda in Makalu Barun National Park of Nepal from Maxent and GARP models. Scientific Reports, 11, 14135. Nature Publishing Group.

Tatem, A.J. (2017) WorldPop, open data for spatial demography. Scientific data, 4, 1–4. Nature Publishing Group.

Thapa, A., Wu, R., Hu, Y., Nie, Y., Singh, P.B., Khatiwada, J.R., et al. (2018) Predicting the potential distribution of the endangered red panda across its entire range using MaxEnt modeling. Ecology and Evolution, 8, 10542–10554.

Wearn, O.R., Trinh Dinh, H., Nguyen Quyet, T., Dao Cong, A., Nguyen Van, P., Nguyen Minh, P., et al. (2021). Myth to reality in the forests of Kon Plong: The exceptional biodiversity value of Kon Plong District, Kon Tum Province. Fauna & Flora International – Vietnam Programme. Hanoi, Vietnam