Soil moisture is a critical variable in environmental monitoring, water resource management and agricultural production. However, accurate soil moisture forecasting remains challenging due to complex spatial-temporal interactions and the large volume of remote sensing data. Traditional prediction methods often struggle to effectively capture these nonlinear relationships. To address these limitations, this study proposes an Attention-Based CNN-LSTM model optimised using the Tabu Search Algorithm for enhanced soil moisture forecasting. The model integrates Convolutional Neural Networks (CNN) to extract spatial features and Long Short-Term Memory (LSTM) networks to model temporal dependencies. An attention mechanism is incorporated to emphasise the most relevant spatial and temporal information, thereby improving predictive performance. Furthermore, the Tabu Search Algorithm is employed to optimise model hyperparameters, reducing forecasting errors and improving efficiency. The proposed approach is evaluated against conventional methods, including standard LSTM and XGBR-GA models, using performance metrics, such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and the coefficient of determination (R2). Experimental results demonstrate that the attention-based CNN-LSTM model achieves superior accuracy, characterised by lower error values and higher R2 scores. These findings highlight the effectiveness and scalability of the proposed framework for large-scale soil moisture forecasting using remote sensing data.
R et al. (Sun,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: