DEEP LEARNING APPROACHES FOR PREDICTING DEFORESTATION PATTERNS AND BIODIVERSITY HOTSPOT LOSS IN SUMATRA

Deforestation in Sumatra, Indonesia, represents a critical environmental challenge that has led to the degradation of biodiversity hotspots and poses serious threats to both local ecosystems and global climate stability, driven largely by rapid forest conversion into agricultural land, illegal logging, and extensive land-use changes, making accurate prediction of deforestation patterns essential for effective conservation planning. This study applies deep learning approaches to predict deforestation patterns in Sumatra while simultaneously assessing their impacts on biodiversity hotspots, with the objective of developing a model capable of identifying areas at high risk of deforestation and estimating potential biodiversity losses. The research employs deep learning algorithms, specifically Convolutional Neural Networks and Recurrent Neural Networks, to analyze satellite imagery, historical deforestation data, land-use changes, and biodiversity hotspot maps, enabling the model to capture both spatial and temporal trends in deforestation dynamics. The results demonstrate that the proposed deep learning model achieves a high prediction accuracy of 92 percent in identifying deforestation hotspots and successfully highlights key biodiversity-rich areas that are highly vulnerable to rapid forest loss, with agricultural expansion and infrastructure development emerging as the dominant drivers of deforestation in these regions. Overall, the findings confirm that deep learning provides a powerful and reliable tool for predicting deforestation patterns and assessing biodiversity hotspot degradation, offering valuable evidence-based insights for policymakers and conservation practitioners to prioritize protection efforts and design targeted interventions aimed at mitigating further environmental damage in Sumatra.