A Multi-Task CNN-LSTM Model with Attention Mechanism for Climate Temperature Forecasting over Iraq

Accurate forecasting of climate temperature at the regional scale is crucial to agriculture planning, water resource management, and disaster preparedness. Still, due to the nonlinear and complicated temporal patterns in regions such as Iraq, conventional climate models fail to meet this imperative. This paper presents a novel deep learning model that not only predicts monthly average temperature values but also predicts the temperature trend, hence addressing the regional climate prediction problem holistically. The proposed model is designed as a multi-task deep learning framework integrating Convolutional Neural Networks, Long Short-Term Memory networks, and an attention mechanism to process historical monthly temperature data obtained from the Berkeley Earth dataset using a sliding window method. The model is learned a shared representation for both regression and classification, therefore optimized together. Classification accuracy of 94.81%, ROC AUC of 0.9839, and R² score 0.9773 for regression were obtained, thus proving the capability of model to efficiently capture local and long-range temporal dependencies and give importance to influential time steps. As such, the proposed CNN-LSTM-Attention model in multi-task learning configuration presents a promisingly accurate, interpretable, and computationally-effective temperature prediction and trend classification, applicable to regional climate modeling and decision support for climate-sensitive areas.