Decoding Wildlife Habitat Shifts in a Changing Environment Through Machine Learning

Abstract

Climate change affects wildlife habitats. Data analytics and machine learning have been used in this study, which focused on different ecosystems in the state of Virginia. In this paper, we propose a holistic approach that integrates climate data from the Open-Meteo API and wildlife observations from the iNaturalist API during the years 2020-2023 using real-time data collection, machine learning models, and interactive visualization techniques. Our approach uses ensembling machine learning methods such as XGBoost, Random Forest, and Gradient Boosting classifiers with 85% accuracy using only climate variables as predictors of wildlife presence. Among them, very strong correlations between temperature patterns and observations of wildlife were found (r = 0.72, p < 0.001); temperature range and seasonal timing explained about 65% of the model fit. We developed an interactive web-based dashboard using Dash that visualizes temporal trends and spatial distributions. This study informs conservation planning and habitat management decisions in the face of climate change and demonstrates how the integration of multi-analytical approaches and real-time data collection provides detailed insights into complex ecological relationships.