Species Environmental Niche Distribution Modeling for Panthera Tigris Tigris ‘Royal Bengal Tiger’ Using Machine Learning

TLDR: In this article, the authors predict and model the distribution of the species Panthera Tigris Tigris by combining various climatic, human influence, and environmental factors so as to predict alternate ecological niche for the already dwindling tiger habitats in India.

Abstract: Biodiversity loss due to habitat degradation, exploitation of natural deposits, rapid change of environment and climate, and various anthropogenic phenomenon throughout the last few decades in the quest of development have led to rise in safeguarding species ecological domain. With natural habitat of the endangered Panthera Tigris Tigris fast declining, coupled with factors such as loss in genetic diversity and disruption of ecological corridors, there is an urgent need to conserve and reintroduce it to newer geographic locations. The study aims to predict and model the distribution of the species Panthera Tigris Tigris by combining various climatic, human influence, and environmental factors so as to predict alternate ecological niche for the already dwindling tiger habitats in India. 19 Bioclimatic variables, Elevation level, 17 Land Cover classes, Population Density, and Human Footprint data were taken. MAXENT, SVM, Random Forest, and Artificial Neural Networks were used for modeling. Sampling bias on the species was removed through spatial thinning. These variables were tested for Pearson correlation and those having coefficient greater than 0.70 were removed. Kappa statistic and AUC were used to study the results of the methodology implemented. Testing data comprises 25% of the presence only points and test AUC value of MAXENT was found to be the highest at 0.963, followed by RF at 0.931, ANN at 0.906, and lastly SVM at 0.898. These indicated a high degree of accuracy for prediction. The most recent datasets were taken into consideration for the above variables increasing accuracy in both time and spatial domain.