Prediction of risk zones and wildlife exposure assessment in Periyar Tiger Reserve using machine learning approach in Kerala, India

Forests are extremely risky and these risks frequently cause animals to go extinct. Similar to other hilly regions, the Periyar Tiger Reserve is habitat to a variety of natural hazards and animal-human interactions, although it is unclear how the risk areas are spread or how the exposed species are distributed. The high-resolution trans boundary models illustrating risk to floods, landslides, wildfires, and human-wildlife interactions is proposed in order to assess wildlife distribution vulnerability to high-risk zones across the Periyar Tiger Reserve. An inventory map of the first Using field surveys and various official data, four different types of risks flood, landslides, forest fires, and Human wildlife conflict were created. Using the Max Ent (Maximum Entropy) machine learning technique, a total of 13 geo-environmental parameters were chosen as predictors to create the risk maps. Generating receiver operating characteristic (ROC) curves and computing the area under the ROC curve (AUCROC) allowed us to assess the predictive models' accuracy. The Max Ent model not only performed exceptionally in terms of degree of fitting but also produced significant results in terms of predictive performance. Indicators of the relative relevance of the four categories of risks under study revealed that elevation and distance from streams were, two most crucial determinants for flooding. For detecting landslides, soil, topographic roughness index, and forest cover were important factors. The closest roads, the amount of forest cover, and livestock were each ranked as the three most crucial determinants for human-wildlife conflict. The research area's annual mean temperature, elevation and distance from water bodies, as well as the presence of livestock, were key factors in the mapping of forest fires.An integrated multi-hazard map was finally produced by merging the high-risk zones for flood, landslides, wildfire, and Human wildlife conflict risk. The results demonstrated that 55 % of the area is subjected to risks, reaching a proportion of landslides up to 31%, human wildlife conflict up to 9%, flood up to 5% and fire up to 10 % in the whole territory. Human settlements in the Periyar Tiger Reserve are disproportionately concentrated in areas of high risk. In contrast, low-risk areas are disproportionately unpopulated. Nearly half of Tiger population in the region lives in areas that are highly susceptible to landslide. Few percentages of elephant population live in areas that are highly suspect able to human wildlife conflict zones. Fire and flood risk areas suspect able to the wildlife is comparatively less. Using this type of multi-hazard map may be a useful tool for local administrators to identify areas susceptible to hazards at large scales as we demonstrated in this research.