Inter and Intra-Annual Spatio-Temporal Variability of Habitat Suitability for Asian Elephants in India: A Random Forest Model-based Analysis

Purathekandy Anjali, Deepak N. Subramani2021 · DOI · OpenReview

Abstract

We develop a Random Forest model to estimate the species distribution of Asian elephants in India and study the inter and intra-annual spatiotemporal variability of habitats suitable for them. Climatic, topographic variables and satellite-derived Land Use/Land Cover (LULC), Net Primary Productivity (NPP), Leaf Area Index (LAI), and Normalized Difference Vegetation Index (NDVI) are used as predictors, and the species sighting data of Asian elephants from Global Biodiversity Information Reserve is used to develop the Random Forest model. A careful hyper-parameter tuning and training-validation-testing cycle are completed to identify the significant predictors and develop a final model that gives precision and recall of 0.78 and 0.77. The model is applied to estimate the spatial and temporal variability of suitable habitats. We observe that seasonal reduction in the suitable habitat may explain the migration patterns of Asian elephants and the increasing human-elephant conflict. Further, the total available suitable habitat area is observed to have reduced, which exacerbates the problem. This machine learning model is intended to serve as an input to the Agent-Based Model that we are building as part of our Artificial Intelligence-driven decision support tool to reduce human-wildlife conflict.

Introduction

Asian elephants (Elephas maximus) are currently facing a crisis of habitat fragmentation. In India, this has led to a surge in human-elephant conflicts, resulting in hundreds of elephant deaths due to electrocution, poaching, and accidents. Traditional ecological models often rely on static "presence-only" data, which fails to account for the migratory nature of elephants and how their habitats change over time.

Researchers from the Indian Institute of Science (IISc) have addressed this by developing a Random Forest-based machine learning model to map how habitat suitability shifts across seasons and years. By identifying key environmental drivers, the study provides a data-driven explanation for increasing human-elephant conflicts caused by shrinking and fragmented land.

Methodology

The study employed a Random Forest (RF) Classifier to capture complex, non-linear relationships between environmental variables. The model integrated high-resolution datasets, including 231 species sightings from the Global Biodiversity Information Reserve (GBIF) spanning 2000–2016, climatic data from WorldClim, and satellite vegetation data (NPP, LAI, NDVI) from NASA’s NEO.

To refine the model, the researchers sampled "pseudo-presence" locations within the uncertainty range of actual sightings and "pseudo-absence" points from regions known to be unsuitable. Through hyperparameter tuning, the authors identified the most significant predictors, prioritizing variables that directly impact elephant survival and movement.

Key Insights: The Drivers of Habitat Suitability

The analysis revealed that vegetation density and primary productivity are the most critical predictors for elephant habitat. Specifically, Net Primary Productivity (NPP) and Leaf Area Index (LAI) emerged as dominant factors. NPP measures available biomass (food), while LAI relates to canopy thickness, providing essential cover for these large mammals.

Importance of environmental predictors in the Random Forest model, highlighting NPP and Elevation as top factors.

Beyond vegetation, the model showed that habitat suitability is highly seasonal. Suitability typically peaks post-monsoon (September to November) and hits a low point between December and August. This seasonal "squeeze" likely forces elephants to migrate into human-dominated landscapes in search of resources.

The final Random Forest model demonstrated robust predictive power. It achieved a precision of 0.78 and a recall of 0.77, indicating high reliability in identifying suitable habitats while minimizing false positives.

ROC curve demonstrating the classification performance of the Random Forest model with an AUC of 0.87.

Long-term analysis of the 2001–2016 period showed a general downward trend in the total area of suitable habitat across India. This decline was particularly pronounced in the southern states of Kerala, Karnataka, and Tamil Nadu. Furthermore, years with the lowest habitat suitability (such as 2011 and 2013) coincided with peaks in reported elephant mortality, suggesting that environmental stress is a direct driver of conflict-related deaths.

Conclusion and Future Directions

While the model is limited by the sparsity of historical observation data, it provides a critical foundation for dynamic conservation strategies. This Random Forest model is designed to serve as a sub-component for a larger Agent-Based Model (ABM). The ultimate goal is to create an AI-driven decision support tool that simulates elephant movement in real-time, allowing for proactive measures to mitigate human-wildlife conflict and protect India's elephant populations.

Summary updated

Abstract

We develop a Random Forest model to estimate the species distribution of Asian elephants in India and study the inter and intra-annual spatiotemporal variability of habitats suitable for them. Climatic, topographic variables and satellite-derived Land Use/Land Cover (LULC), Net Primary Productivity (NPP), Leaf Area Index (LAI), and Normalized Difference Vegetation Index (NDVI) are used as predictors, and the species sighting data of Asian elephants from Global Biodiversity Information Reserve is used to develop the Random Forest model. A careful hyper-parameter tuning and training-validation-testing cycle are completed to identify the significant predictors and develop a final model that gives precision and recall of 0.78 and 0.77. The model is applied to estimate the spatial and temporal variability of suitable habitats. We observe that seasonal reduction in the suitable habitat may explain the migration patterns of Asian elephants and the increasing human-elephant conflict. Further, the total available suitable habitat area is observed to have reduced, which exacerbates the problem. This machine learning model is intended to serve as an input to the Agent-Based Model that we are building as part of our Artificial Intelligence-driven decision support tool to reduce human-wildlife conflict.
BibTeX
@inproceedings{anjali2021inter,
  author = {Purathekandy Anjali and Deepak N. Subramani},
  title = {{Inter and Intra-Annual Spatio-Temporal Variability of Habitat Suitability for Asian Elephants in India: A Random Forest Model-based Analysis}},
  booktitle = {2021},
  year = {2021},
  doi = {10.1109/ingarss51564.2021.9792132},
}
Figure