Machine Learning Models for Predicting Species Distribution Under Climate Change and Urban Expansion Scenarios

Abstract

Abstract: Conserving biodiversity, managing ecosystems and sustainable urban planning requires predicting the distribution of species in response to predicted changes in the environment. This study introduces a machine learning-based approach to predict the distribution of species under combined land-use change and climate change scenarios by incorporating species occurrence data, bioclimatic data, topographic data, urban expansion data, and land-use data. The proposed framework is a combination of environmental variable processing, feature engineering, climate scenario modelling and urban growth simulation to build a strong habitat suitability prediction. Several machine learning algorithms are tested to find the best model to describe the complicated nonlinear relationships between the environment and the occurrence of species. The most important ecological factors affecting habitat suitability and shifts in habitat distribution are identified by feature importance analysis. The framework also outlines areas of potential habitat loss, expansion and fragmentation in future climate and urbanization scenarios and conservation priority zones and ecological corridors. Comparative performance analysis shows high predictive accuracy, reliability and generalization ability in various environmental conditions. Distribution maps created from the data help to gain insight into the spatial distribution of biodiversity, which is important in monitoring, conservation planning, habitat restoration, and for evidence-based decision making. In general, the suggested approach provides an effective and scalable approach to providing support for proactive conservation and sustainable landscape management in the context of a rapidly changing environment and growing pressures from urbanization.