Habitat suitability analysis in a natural peat swamp forest on Sumatran elephants using remote sensing and GIS

Abstract Context It is essential to assess the suitable habitat for elephants in order to mitigate the effects of forest fragmentation on conservation. Aims This study aimed to estimate the potential suitable habitats for Sumatran elephants in the fragmented forest in the Padang Sugihan wildlife reserve area in southern Sumatra, Indonesia. Methods The habitat suitability was analyzed using some environmental factors such as slope, elevation, land cover, distance to rivers, and distance to agricultural areas. The remote sensing, geographic information system (GIS), and MaxEnt model were used to determine the potential habitat suitability for Sumatran elephants. Results This study revealed areas of suitable habitat were evenly distributed throughout the study area, with the composition being suitable (45%), highly suitable (5%), and less suitable (50%). This study revealed the most suitable habitats were found in dense forest areas (gelam or Melaleuca cajuputi forest), which were highly affected by river adjacency, whereas agricultural areas resulted in constrained suitability and fragmentation of forested areas. Conclusions As a whole, the estimation of elephant habitat using remote sensing and GIS may guide the development of conservation strategies for elephant conservation in this region. KEY MESSAGE The Sumatran elephant (Elephas maximus sumatranus) is on the Red List of the International Union for Conservation of Nature as a species that is “Critically Endangered.” The population of elephants has steadily declined in the Sumatra region, which is characterized by peat swamp forests and a tropical rainforest climate and is vulnerable to habitat degradation.


Introduction
The Sumatran elephant (Elephas maximus sumatranus) is the biggest land mammal in the Sumatra region (Monik et al. 2021). The Sumatran elephant is listed as a critically endangered species on the International Union for Conservation of Nature Red List because of human-elephant conflict and habitat deterioration (Williams et al. 2020). The total number of Sumatran elephants in the Sumatra region was estimated to be less than 2,000 individuals (Khafifi 2020) and would continue to reduce because of urban expansion and land use conversion. According to the authority report from 2022, the population of the Sumatran elephant in the Padang Sugihan wildlife reserve is around 127 individuals. The population is consisted of wild elephants and treated-elephants in that area (Mahanani et al. 2013). This wildlife reserve area has experienced fragmentation due to human activities since several years ago, such as local farms that were planted with oil palm, and rubber (Rendana et al. 2021). Elephant usually lives in tropical primary rainforest with a dense canopy (Lim and Campos-Arceiz 2022) that provides food and shelter (Rahman 2019), and in particular, in a low-lying forest area with enough water availability (Husain et al. 2022). In southern Sumatra, the largest elephant population was found in the natural peat swamp forest area, while others occur in the lowland dipterocarp forest.
The Sumatran elephant distribution across southern Sumatra has exhibited a reducing pattern of occurrence, with no occurrence in some areas (Ermawati et al. 2022). Thus, it is important to know the basic habitat requirements of elephants to prevent the continual reduction of their population. The impact of fragmentation because of habitat degradation has greatly contributed to cases of Sumatran elephant kills (Kuswanda 2018). So, the forestry authority has tried to protect and improve the fragmented forest areas in the South Sumatra region, where planting took place in an area of 600 ha and rehabilitation took place in an area of 40 ha to make up for lost habitat (Kunarso et al. 2019).
Elephants' movements and where they live can be affected by things like water and food sources, the type of forest and how it is used, the size of the herd, and human interference (Mohd Taher et al. 2021). Elephants usually have a wide home range, but sometimes the fragmented lands affect their movement and distribution. There are some techniques for modeling the suitability of a habitat for elephants, such as the Bayesian model (Cheah and Yoganand 2022), maximum entropy modeling (Huang et al. 2019), and a geographic information system (GIS)-based analytic hierarchy process (Mandal and Das Chatterjee 2021). Among these, the analytic hierarchy process has been widely applied to estimate habitat suitability using the greatest contribution of environmental factors based on their suitability index (Talukdar et al. 2020). This technique is beneficial for a study of habitat suitability monitoring in real-time with poor sampling data (Shrestha et al. 2004). It may use environmental factors in both categorial and continuous data types (Rendana et al. 2016;Jayasekara and Mahaulpatha 2022).
Previous studies, that investigated habitat suitability for some elephant species such as Ntukey et al. (2022) has investigated the African elephant habitat suitability in the Wami Mbiki-Saadani Wildlife Corridor, Tanzania using remote sensing and GIS techniques. Mohd Taher et al. (2021) has also characterized the habitat suitability for the Asian elephant in Ulu Jerai Reserve Forest, Malaysia using Maxent model. Other recent studies also analyzed Asian elephant habitat suitability in various regions, China (Chen et al. 2022), Thailand (Pla-Ard et al. 2019, Myanmar (Thant et al. 2023), Nepal (Sharma et al. 2020). In Indonesian region, Sumatran elephants are an unique mammal which are only found in the Indonesian region, specifically in Sumatra Island. The habitat suitability for the Sumatran elephant have been already carried out in Lampung province (Southern Sumatra; Monik et al. 2021), in Aceh province (Northern Sumatra; Wilson et al. 2021), in Riau province (Northern Sumatra; Sukmantoro et al. 2018), and in the central part of the Sumatra region (Nugraha 2019). Compared to those studies, most of them were done in dipterocarp forest areas. In our study, we tried to look at a unique environment, a natural peat swamp forest, which is only found in the South Sumatra region. This gave us new information about how elephant habitat suitability studies should be done. If we compare this study to the previous one by Monik et al. (2021), they studied the elephant's behavior, such as its home range, and identified the best timing for elephants to be observed by tourists. This study focused on the assessment of the Sumatran elephant to increase its potential as a tourist attraction. Therefore, in our study, we aimed to estimate the Sumatran elephant habitat suitability in the fragmented Padang Sugihan wildlife reserve area using the remote sensing and GIS techniques. Our study assumes that the forest fragmentation within the wildlife reserve area would constrain the suitable habitat for elephants in that region since this area has been disturbed by human activities. By identifying suitable areas for elephants in the study site, habitats for elephants can be protected from human activities.

Study area
The study site was in the Indonesian province of South Sumatra. It was in the Padang Sugihan wildlife reserve area, which is in the province's east and is a natural peat swamp forest ( Figure 1). This wildlife reserve area was located between 105 0 0 0 00 E and 105 20 0 0 00 E, and 2 0 0 0 00 S and 3 5 0 0 00 S. Since 2014, this area has been set aside for Sumatran elephants (Elephas maximus sumatranus) to live in their natural environment. But, in recent years, the fragmentation of forest areas due to human activities has had a significant impact on elephant distribution in that area.
The wildlife reserve area had a total area of approximately 88,148.05 ha, was categorized as secondary forest, and consisted of flat land. The soil type is mainly loam, with a pH value ranging from 4.2 to 5.4. There were some conservation strategies in the study area, such as a poaching ban and land use regulations near the wildlife region. The wildlife reserve area majorly consisted of various unique local plants such as terentang (Campnosperma spp.), meranti (Shorea spp.), kempas (Koompassia malaccensis), jelutung rawa (Dyera polyphylla), gelam (Melaleuca cajuputi), laban (Vitex pubescens), pelangas (Aporosa sp), and kelat (Eugenia sp). Most of the study area was covered by shrubs and bushes such as paku udang (Stenochlaena palustris), Nephrolepis spp., Pteridium spp., and seduduk (Melastoma malabathricum) (Kunarso et al. 2019;Waheed et al. 2022). while the remaining area consisted of gelam forest and low, dense secondary forest. The Padang Sugihan Wildlife Reserve area is flanked by large rivers, namely the Air Sugihan River in the east and the Air Padang River in the west, which are the boundaries of the area. This area has more than fifty waterways, including rivers and tributaries, which are important as water sources for elephant habitat. (Figure 1). Water quality is a prominent factor for elephants and other species, and it is frequently affected by land use change (Rendana et al. 2022).

Data sources
In this study, we used several sets of secondary data from the Nature Conservation Agency of South Sumatra and remote sensing data. The Shuttle Radar Topography Mission (SRTM) data were used to generate a digital elevation model and slope map. Furthermore, land use was produced using a supervised classification technique using Landsat 8 satellite images in 2022. The Landsat 8 satellite (Path 124, Row 062) and SRTM data for the study area were obtained from an open USGS portal. Rivers and land use vector shapefile data were obtained from Indonesian geospatial portal data (https://tanahair.indonesia.go.id/). For species occurrences, data were obtained from the Nature Conservation Agency of South Sumatra. The data were taken based on elephant sightings along ranger routes and reports within the study area.

Analytical hierarchy process analysis
In this study, the analytical hierarchy process (AHP) method was used to give each habitat variable a certain amount of weight (Table 1). The AHP is a suitable technique to calculate weights for habitat variables according to a scale of importance. It consists of some key processes, such as constructing the AHP hierarchy, making pairwise comparisons of elements of the hierarchical structure, and assigning an overall priority rating (Sahroni andAriff 2016 Abdo et al. 2022,;Asfaw et al. 2022). Furthermore, during this process, the normalized matrix is produced to assign a priority rating for habitat variables. Rating of environmental variables was shown in Table 2. While, in order to analyze the rationality with pairwise comparison of AHP. It was needed to calculate the consistency level which was useful to validate our decision. The consistency ratio (CR) was used to represent the probability (Equation (1)).
RI represented the mean of the consistency index relying on matrix values obtained by Saaty (1994) and consistency index (CI) was calculated using the specified formula below (Equation (2)): k max was a basic Eigen value, n was the number of input criteria used. All parameters were processed using ArcGIS software by assigning GIS scores for indicating suitability level.

Habitat suitability analysis using the MaxEnt model
In this study, the MaxEnt software version 3 was used to create a model that could estimate the distribution of species in the Padang Sugihan wildlife reserve area based on a number of inputs, such as where the species were found and how the environment was changing. A total of seven elephant occurrences were obtained using a camera trap, which was carried out by the Nature Conservation Agency of South Sumatra, while other additional data were obtained from elephant sightings along ranger routes and reports. The MaxEnt performed a suitability estimation by assessing the probability distribution closest to uniform from the link between environmental parameters and species occurrences. In this study, the default setting used the auto features tool, and the convergence threshold was set ranging from 10 to 5500 maximum iterations. A regularization variable value is set to 1.0 in order to avoid overfitting. Furthermore, classes were produced by examining ordinary thresholds, which matched the MaxEnt result. The threshold of 0.2 indicated the presence of 10th percentile training, and the threshold of 0.4 indicated areas with the most training.
By applying these thresholds, the probability greater than 0.2 was classified as suitable habitat, and the probability greater than 0.4 was classified as highly suitable habitat, while the probability of less than 0.2 was classified as less suitable habitat. Moreover, all environmental variable data was analyzed using ArcGIS software version 10. These data were transformed into ASCII-formatted files. The elevation map was produced from the digital elevation model data and was then used to produce a slope map.
The land use data was produced using one of the most popular used image classification method, namely, maximum likelihood classification was applied for classifying land use classes. The empirical analysis of satellite image, google earth image, and topographical data was examined. For predominant of classes with a minimun total of training samples was 30. The land use data then were validated using the land use map obtained from the Nature Conservation Agency of South Sumatra. The validation process was carried out using the Kappa statistic, which produced an overall accuracy of 90% with Kappa coefficient of 0.85 (Smits et al. 1999). There were four major classes of land cover identified in the study area: water bodies, fallow, grass, and agroforestry. Land cover and river classes were used to assess the distance of a contiguous area from elephant habitat. All variables were of the continuous data type, except land cover. All maps were set at the same spatial resolution of 10 m and projection of Universal Transverse Mercator 48 S using the ArcGIS software. The correlation test was assessed to identify the relationship between the variables. In this study, all variables did not show a significant correlation. Thus, all variables could be used for the model.

Habitat suitability zones for Sumatran elephant
The model indicated a good accuracy in estimating habitat suitability for the Sumatran elephant over the study site. AUC scores of test data were greater than training data, but, these data showed a satisfactory AUC scores of 0.86 and 0.89. The MaxEnt estimation of suitable habitat revealed areas of high suitability that were mostly distributed in the entire study area ( Figure 2). According to the categorization of habitat suitability, a total of 45% of the area was estimated as suitable habitat, 5% was highly suitable, and 50% was moderately suitable (Table 3). The suitability index showed a steady increase with a reduced distance to the river. In contrast, the suitability index decreased toward agricultural areas or away from rivers. Areas of suitable habitat differed based on the total area of the fragmented forest. A study by Mohd Taher et al. (2021) found that a large forest area yielded a smaller area of suitable elephant habitat because of the effect of habitat variables. In our study, we assumed that the agricultural areas hinder the formation of large, suitable habitat areas. Suitable habitat (0.4-0.8) was largely found in forest areas, while agricultural land and builtup areas obtained low suitability (0.2-0.4). Larger areas of suitable elephant habitat were found close to rivers situated in forest areas, but smaller areas for rivers were found in non-forest areas. As a whole, all areas in the Padang Sugihan wildlife reserve area were still classified as suitable for elephant habitat. We also thought that the area of good habitat changed based on how big the wildlife reserve region was as a whole. The Padang Sugihan Wildlife Reserve had a smaller area of settlement (113.9 ha), and it had larger areas of fallow or wasteland (1,218.8 ha). But, the study area was still at a suitable level for elephants due to suitable environmental variables within this region. Menon and Tiwari (2019) explained that roads and railways could hinder the establishment of larger, more prolonged species in a certain habitat area. Lower elevation areas (20 m) had the most suitable habitat, whereas higher elevation areas were dominated by agriculture and settlement areas had a lower suitability.
In this study, we looked at the vegetations that were already there by making a map of the land cover in the study area. The results showed that there was still a lot of vegetation in the area of study (Figure 3). Because there is a lot of food, water, and plants in this area, it was chosen as a good place for elephants to live. The food chosen by the elephants consisted of grass, shrubs, tree leaves, bark, aquatic plants, and fruit (Ndi et al. 2022). Kuswanda et al. (2022) have revealed the highest percentage of plant parts eaten were leaves, twigs, and stems. Specifically based on that study, the part of the plant that elephants consumed the most Figure 2. Habitat suitability map for Sumatran elephant in the study area. Greener color represents higher suitability. was in the grassland area in as many as four parts (leaves, stems, roots, and fruit), while in swamps and secondary forest there were only two parts each, namely leaves and stems in swamps and leaf and bark in secondary forest. This notion was consistent with another study by Riba'i et al. (2013). Table 2 shows how the different levels of suitability were affected by the environment. Based on our study, we found that land cover had the most impact at about 32%, while elevation had the least impact at 5%. It showed that the estimation of suitable habitat greatly relied on the land cover factor. Other studies that investigated habitat suitability in highland areas, such as Northwest Namibia (Wenborn et al. 2022) and Rombo, Tanzania (Mmbaga et al. 2017), have reported this notion. That is why the high suitability class is mostly found in dense forest areas. Furthermore, because we discovered more agricultural lands than urban lands, thus we evaluated the distance from agricultural land factor for GIS model input. This factor had a 23% influence, and the distance from the river factor was about 18%. The effects of those distance factors were verified by the lowest suitability near roads, whereas areas near rivers had the highest suitability. The factors had a 8% influence on slope factor. This was consistent with other studies (Sharma et al. 2019). Several factors, including slope, distance from a river, and elevation, revealed the same pattern of decreasing likelihood of elephant occurrence with increasing factor value, while factor distance from an agricultural area revealed the opposite pattern. Mohd Taher et al. (2021) assumed the elephant habitat was mainly distinguished by areas below 100 meters. The likelihood of occurrence decreased as elevation increased. Our research showed that elephants were more likely to be seen in low-lying areas with slopes of less than 3. This result was similar to another study by Husain et al. (2022), who found that elephant occurrence increased as slope value decreased. There wasn't a clear link between elephant variation and slope, but this factor has been used a lot in studies of elephant habitat (Puyravaud et al. 2022). The study area had a slightly steep topographic feature on the eastern side. This might constrain elephants' movement range because they would need a lot of power to cross the land (Beirne et al. 2021). Therefore, an area with a flat and light slope was preferred by elephants.
Areas near rivers (distance 100 m) had greater habitat suitability for elephants ( Figure 3). Elephants can be limited by the air temperature and the distance to the nearest water source because they are not heatresistant during the day. Elephant habitat preference is expected to rise in the presence of water sources. Elephants bathe and drink every day because water helps to cool their bodies. Elephants regularly coat their bodies with mud, dust, soil, or sand to regulate body temperature (Syafrianti et al. 2021). Two major Rivers (Air Padang and Air Sugihan) and a few primary channels are among the numerous water sources that may be found in the Padang Sugihan region. Based on the suitable location of the elephant distribution (Figure 2), the distance from water sources to elephant activities is quite variable. A statistical analysis showed that the greatest distance (28 m) from water sources. According to Buchholtz et al. (2019), a water body may indicate whether elephants were present in a certain environment. Furthermore, especially during the dry season, the elephant population was typically concentrated near the main river systems in the central area of habitats.
In contrast, the species' occurrences were low in areas close to agricultural areas, with a distance of less than 1000 m. This was caused by the availability of human activities, which could hinder habitat suitability from extending outside of the wildlife reserve areas. Elephant sightings were most likely in areas with a variety of land cover types, such as forests, while agricultural areas had the lowest likelihood. The study area's land cover is dominated by shrubs and bushes, which are less likely to occur. This result was in line with other studies (Sukmantoro et al. 2018). Even though soil type did not give a high contribution to the estimation in this study, because the study area only had one soil type (loam soil type). For instance, areas originating from acidic intrusion, clay, and silt were more suitable as compared to limestone and alluvium. The soil contained more clay and provided a high amount of nutrients to form salt licks (Rendana et al. 2018).
Aside from the land cover factor, another prominent factor in the habitat suitability analysis was the distance to rivers. Figure 2 depicted an elephant-favored area with a low elevation and a flat landscape. Alfred et al. (2012) found that elephants liked living in areas that were up to 300 m high because they had access to the right kind of terrain, a vegetation canopy, and a water source. Elephants would be less likely to show up if the elevation and slope values went up. South Sumatra's forests could be put into different types based on where they are. In the study area, the forest was categorized as a gelam forest, which was a unique plant in the wetland area. Even though the wildlife reserve area has been broken up, high suitability was still found in that area. This shows that elephants and the productivity of forests are linked in a good way. Other areas in the southern part had lower suitability, attributable to the existence of agricultural and settlement areas. This notion showed a close relationship between land cover type and elephant in low-lying regions.
Previous study by He et al. (2023) found the alteration in suitable habitat for elephants was affected by lower anthropogenic interference and undisturbed circumstance. Sharma et al. (2020), who studied elephant habitat suitability in Nepal, they assumed that habitat fragmentation could be decreased by evading the human activities within suitable habitat areas. This way could assist to increase the wildlife corridors management and habitat patches. So, the land cover was one of the most important factors, since habitat loss was the main cause of elephant migration (Sintayehu and Kassaw 2019). This happened when forests were turned into cities and farms. Furthermore, plant types varied according to land cover type, affecting the availability of food for elephants.
Previous study by Anuradha et al. (2019) has shown that the growth of agriculture has hurt elephant habitat and caused habitat loss. Our research discovered that habitat suitability increased with distance from an agricultural area. The best distance for habitat from agricultural areas was more than 1 km. In this study, we did not include the distance to the road because the study area only had a small road or farm track. According to Ahmed et al. (2022), major roads and railways have a greater impact on habitat suitability. Even though the study area had a moderate to high suitability index at this time, the existence of agricultural activities within the area could deter a sustainable elephant habitat due to the role these activities play in impeding elephant movement range. A moderately suitable zone, consisting of agricultural areas and a small portion of settlement, was located on the south side. As we predicted, an area with a suitable habitat tended to be far away from human interference. Therefore, the development of human activities toward wildlife reserve forests ought to be avoided to diminish the likelihood of elephant and human conflict.
Because of the growing number of incidents between elephants and people (Kuswanda et al. 2022) it was important to figure out if the Sumatran elephant's habitat was good for it. Due to habitat loss and forest fragmentation, there are more elephants in places where people grow crops or raise animals. So, better supervision of wildlife reserves in forest areas would help keep people from getting in the way of elephants. Because our study had finite data on the occurrence of elephants and a small coverage of the study area, we think this was enough to provide a model and also some direction toward elephant conservation efforts. Even though the Sumatran elephant could live in a fragmented area, efforts should be made to keep an eye on habitat suitability to keep a balance between the growth of human activity and the elephant's habitat.
The study about waterways was prominent to know water source distribution around the study area. It was useful on understanding variation and movement of elephants (Buchholtz et al. 2019), because they were known as water-dependent animal. The elephants were necessary to hydrate and to minimize thermal stress and evaporative water loss. Sach et al. (2020) explained the elephants consumed a daily water more than 200l. Therefore, the waterways factor described the elephants home range prefer to close to water source. Our results supported several studies of Bastille-Rousseau et al.

Wildfire damage in the study area
The Padang Sugihan Wildlife Reserve area formerly was considered as susceptible for forest fire events. Areas with high susceptible on fire were shown in Figure 4. Those areas dominantly located in the western and northern part of the study area. Previous regional research in peatlands has described how fire negatively affected the biodiversity of plants and animals (Hidayat and Hanif 2020;Syaufina and Abi Hamzah 2021). The biodiversity of Padang Sugihan is consequently seriously threatened by fire-induced habitat degradation and isolation within its bounds, as evidenced by our observations of fire activity there both during wet and primarily dry years. The dry season, which ran from July to November, saw the most fires in the Padang Sugihan area and its adjacent buffer zone. There are frequently fires both inside the reserve and outside it among the wetlands and bushes that make up the natural peat-swamp forest vegetation. While there are more fires in the reserve's meadow, which corresponds to grassland, there are more fires in the buffer area's forest plantations than in any other kind of land use between 2001 and 2020 (Imron et al. 2022). Mammal responses to before and after fire incidents provide information about how well the reserve is maintaining this area's biodiversity. An important insight into the impact of fire on endangered species conservation in the reserve and more broadly across peatlands and other lowland habitats of Sumatra would come from studying the behavioral response of Sumatran elephants, a prominent species in the Padang Sugihan, using GPS collaring and population projection studies (Abram et al. 2022). Additionally, a modeling-based population viability analysis for the elephant population in the reserve and its surroundings will offer suggestions for management strategies (He et al. 2020).

Conclusions
Remote sensing and GIS techniques were successfully used to estimate the habitat of Sumatran elephants in the fragmented wildlife reserve area. This gave a good idea of how elephant habitat changes over time and space. This was obtained by determining that the appropriate habitat for Sumatran elephants was highly dependent on the land cover factor and the distance to water sources or rivers. Because of the fragmentation process, the total number of areas with suitable habitat was lower as compared with the forest areas. As a result, connecting forest zones via wildlife passages could greatly benefit elephant habitat and movement. The output of this study will be beneficial to the management and conservation of elephant habitats in the Sumatran region.

Author contributions
Conceptualization and supervision: MR; writing review and editing: MR, WMRI, SAR and HGA; data curation and formal analysis: MR, WMRI and SAR; evidence collection, review, and editing: SD, SN, HGA, HA and AAA.

Disclosure statement
No potential conflict of interest was reported by the author(s). On behalf of all authors, the corresponding author states that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability statement
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.