Environ Monit Assess. 2026 Jan 23;198(2):169. doi: 10.1007/s10661-026-14993-z.
ABSTRACT
As a crucial freshwater resource, groundwater plays an indispensable role in arid and semi-arid regions characterized by low annual precipitation and frequent droughts. Developing computational frameworks for groundwater level prediction is essential to advance sustainable water resource management. This study proposes a hybrid deep learning model (CNN-GRU-Attention) for groundwater depth forecasting, integrating convolutional neural networks (CNN), gated recurrent units (GRU), and attention mechanisms. The methodological framework commenced with a spatiotemporal analysis of groundwater depth dynamics in Zhengzhou, China. Subsequently, multiple machine learning and deep learning algorithms were systematically evaluated to predict groundwater depth using four input variables: monthly evaporation, precipitation, average temperature, and groundwater extraction. These variables were rigorously selected through the Shannon entropy method. Model performance was quantified using three statistical metrics: MAE, RMSE, and R2. Results indicate that the CNN-GRU-Attention model demonstrates superior performance in groundwater depth forecasting, achieving MAE values of 0.4-0.5, RMSE values of 0.5-0.6, and R2 values of 0.8-0.9. To fully evaluate the performance of the model, we designed two hypothetical scenarios. First, we analyzed changes in the model’s predictive performance under conditions of reduced data, when the data volume is reduced by 10-25%, the CNN-GRU-Attention model still outperforms other models in predictive performance. Second, to maintain stable groundwater depth under drought-induced rainfall reduction conditions, controlled extraction measures should be implemented to balance recharge and withdrawal. Under this special rainfall scenario, a reduction in extraction volume of 42 million m3 is more conducive to maintaining groundwater stability. This model provides an effective predictive framework and offers valuable insights for sustainable groundwater management in arid regions.
PMID:41577970 | DOI:10.1007/s10661-026-14993-z
