<![CDATA[Accelerating climate change has intensified the frequency and severity of extreme weather events, posing substantial threats to regional biodiversity. Conventional climate-biodiversity models often struggle to capture non-linear interactions between climatic variables and ecological responses, limiting their predictive accuracy under extreme scenarios. This study aims to integrate deep learning techniques with climate modelling to improve the prediction of regional biodiversity loss under extreme weather conditions. We employed a hybrid modelling framework that combines high-resolution regional climate model outputs with deep learning architectures, specifically convolutional and recurrent neural networks. The model was trained using multi-decadal climate data, species distribution records, and ecological indicators across selected regions. Extreme weather scenarios were simulated based on projected temperature anomalies, precipitation extremes, and drought indices. Model performance was evaluated using cross-validation and comparative benchmarks against traditional statistical models. The integrated deep learning–climate model demonstrated significantly higher predictive accuracy and robustness in identifying biodiversity loss hotspots under extreme weather scenarios. Results reveal pronounced spatial heterogeneity, with ecosystems exposed to compound extremes showing disproportionately higher vulnerability. Integrating deep learning with climate modelling offers a powerful approach for anticipating regional biodiversity loss under extreme climate events, providing valuable insights for adaptive conservation planning and climate-resilient biodiversity management.]]>
