heatwave-prediction

This paper presents a GPU-accelerated deep learning framework using ConvLSTM models to predict next-day urban thermal conditions and generate heat risk maps, achieving strong performance (R²=0.8877) on Sarajevo data.