Enhanced deep learning approach for improved maximum temperature forecasting: a case study in the Sahara region, Morocco

The pursuit of insights into the effects of historical climate shifts, both globally and regionally, has intensified. Specifically, air temperature fluctuations are crucial for analyzing climate change impacts across agriculture, ecology, environment, and industry. Precise temperature forecasting is vital for protecting lives and assets, significantly aiding strategic planning efforts. This study introduces ’TempFusionNet’, a novel hybrid algorithm developed to enhance temperature forecasting by integrating multiple deep learning models. Specifically designed to address the unique climate challenges of the Moroccan Sahara, the model focuses on four key cities: Dakhla, Laayoune, Boujdour, and Tantan. By incorporating advanced models such as Temporal Convolutional Networks (TCN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM) networks, TempFusionNet significantly improves the accuracy of one-day-ahead temperature predictions. Evaluation using metrics such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared demonstrates superior performance over traditional techniques, highlighting the potential of hybrid deep learning models in regional climate prediction. This work contributes a valuable tool for informed planning and decision-making in sectors sensitive to climatic variations.