5 Mingxi Forestry Bureau, Sanming, Fujian 365000
6 Fujian Junzifeng National Nature Reserve Management Bureau, Sanming, Fujian 365000
7 College of Life Sciences, Huaibei Normal University, Huaibei, Anhui 235000
* Author for correspondence. E-mail: 2019117@htu.edu.cn
Abstract Due to extensive poaching and habitat degradation, the Chinese pangolin (Manis pentadactyla ) population had plummeted by approximately 90%, leading the International Union for Conservation of Nature (IUCN) to classify it as a Critically Endangered (CR) species. The scarcity of up-to-date data on the species’ distribution and dynamics presented a significant challenge in developing effective conservation strategies and implementing protective measures within China. Predominantly, China’s national-level nature reserve and administrative departments operated at the county level, thereby limiting the applicability of larger-scale analyses and studies, especially those at the provincial level and above, for these administrative entities. This study employed on eleven widely used modeling techniques created within the BIOMOD2 framework to predict suitable habitats for the pangolin at the county scale, while examining the correlation between environmental variables and pangolin distribution. The results revealed that in Mingxi County, situated in the eastern sector of the Wuyi Mountains, the moderately suitable habitat spanned 260 km², accounting for 15% of the total area, whereas the highly suitable habitat encompassed only 49 km², constituting 3% of the total area. Within the county-managed nature reserve, the proportion of highly suitable habitats reached as high as 52%. However, nearly half of these areas, both moderately and highly suitable, remained inadequately addressed and conserved. The findings underscored the inadequacy of existing protected areas in sustaining the current pangolin population, leading to the identification of nine administrative villages that necessitated prioritized conservation efforts. The study anticipated an overall expansion in suitable habitats over the ensuing two decades, likely associated with an increase in precipitation, with significant growth projected in the eastern regions of Xiayang Township and Hufang Town. This research offered a clear and applicable research paradigm for the specific administrative level at which China operates, particularly pertinent to county-level jurisdictions with established nature reserve. Given the constraints of the existing data and in order to more precisely evaluate the pangolin’s situation at the county scale, the study underscored the paramount importance of conducting field surveys, deemed as the most urgent task at the time.
Keywords: Chinese Pangolin, County-Level Scale, Conservation, Population Dynamics
INTRODUCTION
The Chinese pangolin (Manis pentadactyla ), an endemic scaly mammal unique to Asia, has attracted significant global focus due to its distinct biological attributes and the grave threats to its existence (Wang et al., 2020; Yan et al., 2021; Zhang et al., 2021). Serving as a myrmecophagous organism, it plays an integral role in the regulation of termite and ant populations (Li et al., 2011). Nevertheless, this species confronts substantial survival challenges, primarily attributed to illicit poaching and habitat degradation (Challender et al., 2020; Heinrich et al., 2016; Wu et al., 2002). Factors such as the illegal trade (Gerard et al., 2023; Gu et al., 2023; Nash et al., 2018) and local consumption of pangolin meat (Emogor et al., 2023) are posited as the principal motivators behind its poaching. Presently, the population of the Chinese pangolin has diminished by a staggering 90%, leading to its classification as Critically Endangered (CR) by the International Union for Conservation of Nature (IUCN) (Challender et al., 2019), inclusion in Appendix 1 of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES), and designation as a first-class protected species under the national conservation laws of China (Notice No. 3, 2021, National Forestry and Grassland Administration, Ministry of Agriculture and Rural Affairs,http://www.forestry.gov.cn/ ). The prospects for this species are rather bleak (Bashyal et al., 2021; Yang et al., 2018), and the deficiency of contemporary data regarding its population distribution and dynamics poses a pressing challenge in the formulation and execution of conservation strategies and actions (Hu et al., 2010; Kong et al., 2021; Sharma,Rimal, et al., 2020).
Recent studies have elucidated that the Chinese pangolin (Manis pentadactyla) predominantly inhabits the southeastern territories of China (Ta et al., 2021). The Wuyi Mountain region is identified as the most pivotal habitat for this species in Eastern China (Peng, 2020; Yang et al., 2018; Zhou, 2022). Furthermore, the distribution of the Chinese pangolin is significantly influenced by human activities and variations in precipitation (Ta et al., 2021; Yang et al., 2018), providing crucial support for comprehending its current distributional status. However, the predominance of county-level units in China’s national-level protected areas presents a limitation in conducting analyses at larger scales, notably at the provincial level and above, thus diminishing their practical utility for administrative departments. Consequently, meticulous analyses at the county level are imperative for formulating viable and effective conservation strategies. A grave challenge encountered globally is the dearth of dedicated protected areas with a primary focus on pangolin conservation (Katuwal et al., 2017; Sharma,Sharma, et al., 2020; Wei et al., 2022). The existing sanctuaries lack targeted scope and specificity in policy development (Nash et al., 2016; Sharma,Rimal, et al., 2020). Therefore, a detailed examination of environmental influences such as climatic conditions, geological factors, and anthropogenic disturbances on the Chinese pangolin at finer scales, coupled with predictions of potential suitable habitats, is essential. Such research will not only deepen our understanding of the local population dynamics and distribution of the Chinese pangolin but also furnish administrative entities with direct and efficacious scientific underpinnings.
Sanming City, located in the eastern segment of the Wuyi Mountain Range in southeastern China, with a total area of 1730 km2, is distinguished for its abundant biodiversity and unique natural habitat, historically constituting a critical distribution zone for the Chinese pangolin (Zhou, 2022). Despite its ecological significance, comprehensive scientific studies pertaining to the population distribution and dynamics of the Chinese pangolin in this area are markedly lacking. This investigation aims to forecast the potential distribution zones of the Chinese pangolin in Mingxi County, Sanming City, leveraging field survey data, Geographic Information Systems (GIS), remote sensing technologies, and the Biomod2 model, for both the present and the upcoming two decades. The varied topography and extensive vegetation varieties in Mingxi County provide prospective habitats for the pangolin. The objective of this study is to elucidate the correlation between environmental variables and pangolin distribution, and to predict potential suitable habitats. This is intended to supply actionable scientific recommendations for local policymakers and serve as a paradigm for formulating conservation policies for endangered species, like the Chinese pangolin, in national, provincial, and county-level protected areas throughout China. In light of the global imperative for biodiversity conservation and the practical demands of wildlife protection, this research emphasizes the significance of engaging in detailed, scientific investigations at a granular scale within the field of wildlife conservation.
This research is of paramount importance for county-level administrative and management entities in China, particularly in the context of developing conservation strategies for endangered species like the Chinese pangolin. Our methodology encompassed a series of crucial steps to fulfill the objectives: Initially, detailed location data for the Chinese pangolin were amassed through extensive field expeditions. Subsequently, environmental datasets were meticulously gathered and rigorously corrected to assure their accuracy. Furthermore, an assessment was conducted on the alterations in suitable habitats, both in the current scenario and projected over the next two decades, including an analysis of their potential influencing elements. Lastly, with a consideration of the demarcations of protected zones and the perimeters of administrative villages, tailored conservation proposals were formulated for immediate and long-term implementation. These strategic approaches substantially elevate the study’s reliability and utility, solidifying its vital role in the formulation of local conservation policies.
STUDY AREA
The Wuyi Mountain range, notably its eastern extension, Mingxi County, is recognized as an ecologically significant potential habitat for Manis pentadactyla (Chinese pangolin), underscoring its conservation value (Peng, 2020; Yang et al., 2018; Zhou, 2022). Accordingly, this investigation designates Mingxi County in Fujian Province (depicted in Figure 1A and 1B) as the focal study locale. The county, typified by a subtropical monsoonal ecosystem, averages an annual temperature near 18°C with mean precipitation around 2000 mm (Shi, 2021). An impressive over 80% forest coverage (Zhang & Hunag, 2011) contributes to its biodiverse landscape, previously a stronghold for the pangolin population (Zhou, 1996). Mingxi’s encompassing nine townships and the Junzifeng National Nature Reserve, devoted to the preservation of subtropical evergreen broadleaf biomes and the safeguarding of endemic fauna such as the Cabot’s Tragopan, delineate its ecological significance. This research utilized vector data delineating the townships and administrative village boundaries, sourced from the county’s environmental governance agencies, to frame the geographical scope of the habitat suitability analysis.
METHODS
Species Distribution Data
The elusive and nocturnally active Chinese pangolin (Manis pentadactyla), characterized by its low population density, renders direct observational studies logistically impractical (Macdonald, 2006). Therefore, this research adopts an indirect approach, focusing on the analysis of pangolin burrow systems. This methodology is instrumental in elucidating the species’ habitat preferences and assessing the environmental determinants influencing their burrow distribution (Sharma,Sharma, et al., 2020; Thapa et al., 2014; Wu et al., 2002), thereby providing critical insights for targeted conservation interventions. In Mingxi County, a methodical stratified random sampling framework was applied, deploying 90 transects across diverse ecological niches: 21 in broadleaf forests, 16 in mixed coniferous-broadleaf forests, 16 in coniferous forests, 17 in bamboo-dominated areas, and 20 within agricultural landscapes, each transect extending 1-2 kilometers in length and encompassing a 5 to 10-meter width. Field expeditions were conducted in distinct seasonal windows - November to December 2022, February to March 2023, May to July 2023, and September to October 2023. Geographic coordinates were meticulously documented upon burrow discovery. Given the restricted spatial range of the species, typically confined to less than 1 square kilometer (Sharma,Rimal, et al., 2020), burrows spaced beyond 500 meters were selected for in-depth analysis. The application of infrared camera traps validated the continued occupancy of these burrows by pangolins, with Appendix 1 presenting photographic evidence (Figure S1).
Environmental Data
In this study, an integrative modeling approach was applied to assess a spectrum of environmental determinants, stratified into three primary categories: (i) a suite of 19 bioclimatic variables, encompassing an array of temperature and precipitation metrics for the period 1970-2000; (ii) topographical and anthropogenic factors, including slope, aspect, altitude, hydrological proximity, and infrastructural distance; and (iii) temporal dynamics of vegetation health, quantified through the analysis of Normalized Difference Vegetation Index (NDVI) across 23 temporal intervals in 2020. The compilation of these environmental parameters (refer to Table S1 in Appendix 1) provided a robust foundation for habitat suitability modeling.
The environmental variables utilized in this ecological analysis were processed with a 2.5-minute spatial resolution, standardized to the UTM-WGS1984 coordinate system. The selection criteria for environmental factors involved a two-tiered process: preliminary individual factor analysis using the Maxent model, identifying significant contributors (AUC > 0.9, contribution rate > 10%), followed by a comprehensive collinearity assessment, where one variable from any highly correlated pair (|correlation| > 0.8) was omitted (illustrated in Figure S2 in Appendix 1). This procedure culminated in the identification of nine pivotal environmental factors: Bio03, Bio19, NDVI0321, NDVI0727, NDVI0913, Aspect, Roads, Slope, and Waterway, each thoroughly defined in Table S2 in Appendix 1.
Species Distribution Modeling
(1) Within the ambit of this study, the biomod2 software package (Thuiller et al., 2023)was harnessed, amalgamating a cadre of 11 advanced modeling algorithms for a synergistic prediction of species distribution. Utilizing the algorithms integral to Biomod2, a training subset comprising 75% of extant species distribution data was deployed for model calibration, reserving the remaining 25% for model validation purposes. Each algorithm was iterated thrice to fortify the robustness of the results. The pseudo-absence approach was employed to compensate for the paucity of explicit absence data. Model efficacy was appraised using True Skill Statistics (TSS) and the Area Under the Receiver Operating Characteristic Curve (AUC) as metrics, evaluating the precision of model fit. TSS amalgamates sensitivity and specificity, with a score range from -1 to 1, where values between 0.8 to 1 signify optimal model fidelity (Allouche et al., 2006). AUC values span from 0.5 to 1, with thresholds above 0.7 denoting reasonable predictive accuracy, above 0.8 indicating satisfactory predictions, and values surpassing 0.9 reflecting high precision (Anderson, 2003). Models with TSS exceeding 0.7 were integrated to construct the ensemble model, leveraging the EMwmean method, and AUC values were employed as the definitive standard for prediction appraisal.
(2) A randomized sampling protocol was applied to ascertain Pearson correlations among all predictive and evaluative variables (Guisan et al., 2017; Thuiller et al., 2023), determining the relative import of each variable in species distribution modeling. This non-model-dependent approach allows for streamlined comparisons across different modeling frameworks (Zanardo et al., 2017). Response curves were employed to delineate the gradational changes in species occurrence probability with pivotal predictive variables, elucidating the interplay between species occurrence and environmental drivers, with ecological factors deemed conducive for species survival when the occurrence probability exceeds 0.5.
(3) In this research, ArcGIS software was utilized for visual representation of habitat suitability spatial distribution in TIFF formats. Ensemble model outputs dictated the stratification of habitat suitability into four discrete categories: 0-0.15 as unsuitable, 0.15-0.50 as lowly suitable, 0.50-0.75 as moderately suitable, and 0.75-1.00 as highly suitable. Further, an analysis incorporating the perimeter of Fujian Junzifeng National Nature Reserve and current administrative village delineations was conducted. This analysis was pivotal in identifying key administrative villages on the periphery of the reserve, earmarking them as primary zones for conservation and monitoring initiatives. Distribution extents of diverse suitability levels within all administrative villages were methodically ranked, employing a weighted schema (score = highly suitable area × 0.7 + moderately suitable area × 0.5). A conservation benchmark was set to ensure no less than 75% of suitable habitats outside the reserve are conserved, based on which, administrative villages necessitating immediate conservation actions were identified. We conducted field surveys in the selected administrative villages, establishing at least one transect in each village to verify the presence of pangolin burrows along the survey lines.
(4) For future projections, the BIOMOD_EnsembleForecasting function within Biomod2 was deployed. Predictive variable binary transformation was conducted using ArcGIS’s reclassification tool, setting a critical threshold at 0.5, denoting values ≥0.5 as indicative of species presence, and <0.5 as absence. Subsequent comparative analyses of current and projected distributions under the SSP1-2.6 scenario (similarly for SSP5-8.5) were facilitated using ArcGIS 10.2. Raster layers were initially reclassified based on habitat suitability, attributing new pixel values. Multiplicative raster calculations were then employed, each pixel value acquiring a novel interpretation: ”3” indicating absence, ”4” for expansion, ”6” for contraction, and ”8” for stable regions (He et al., 2018; York et al., 2011). The final phase involved ranking future suitable areas across townships to spotlight regions meriting heightened conservation focus over the next two decades.
All analytical processes were conducted in R software (version 4.3.1, 2023), with spatial analysis executed using ArcGIS (version 10.2; ESRI, Inc., Redlands, CA, USA). Documentation and presentation tasks were facilitated through WPS Office (Kingsoft Office Software,https://www.wps.com/office-free ). The integrated application of these analytical tools endowed the study with robust data processing and analytical prowess, ensuring the accuracy and reliability of the results.
RESULTS
Model Performance Analysis
In this ecological study, a comprehensive dataset of 106 pangolin burrows was collated, with a focus on 23 selected burrows for intensive analysis (illustrated in Figures 1C and D). Within the scope of the biomod2 framework, seven predictive models were meticulously chosen, each surpassing the True Skill Statistics (TSS) benchmark of 0.7 (as outlined in Table 1). Notably, the Random Forest (RF) and XGBOOST models demonstrated superior Receiver Operating Characteristic (ROC) values of 1, eclipsing the ensemble model’s predictive accuracy. Conversely, the Gradient Boosting Machine (GBM), Maximum Entropy (MAXENT), Generalized Linear Model (GLM), Classification Tree Analysis (CTA), and Generalized Additive Model (GAM) yielded ROC values marginally inferior to that of the ensemble model (detailed in Table 1). This delineation of results highlights the ensemble model’s exceptional proficiency in accurately modeling the distribution patterns of Manis pentadactyla (Chinese pangolin) in the near current historical window (1990-2000).