Enhancing Deep Learning Soil Moisture Forecasting Models by Integrating Physics-based Models

• Published in: Advances in atmospheric sciences Vol. 41; no. 7; pp. 1326 - 1341
• Main Authors: Li, Lu, Dai, Yongjiu, Wei, Zhongwang, Shangguan, Wei, Wei, Nan, Zhang, Yonggen, Li, Qingliang, Li, Xian-Xiang
• Format: Journal Article
• Language: English
• Published: Heidelberg Science Press 01.07.2024; Springer Nature B.V; School of Atmospheric Sciences,Sun Yat-sen University,Guangzhou 510275,China; Southern Marine Science and Engineering Guangdong Laboratory(Zhuhai),Guangzhou 510275,China; Guangdong Province Key Laboratory for Climate Change and Natural Disaster Studies,Guangzhou 510275,China%Institute of Surface-Earth System Science,School of Earth System Science,Tianjin University,Tianjin 300072,China%College of Computer Science and Technology,Changchun Normal University,Changchun 130123,China
• Subjects: AI Applications in Atmospheric and Oceanic Science: Pioneering the Future; Atmospheric Sciences; Deep learning; Drought; Earth and Environmental Science; Earth Sciences; Forecasting models; Geophysics/Geodesy; Hydrologic processes; Long short-term memory; Meteorology; Moisture content; Original Paper; Physics; Predictions; Soil; Soil conditions; Soil moisture; soil moisture forecasting; hybrid model; deep learning; 注意力机制; 卷积长短期记忆模型; attention mechanism; 深度学习; 土壤湿度预报; ConvLSTM; 混合模型
• ISSN: 0256-1530; 1861-9533
• DOI: 10.1007/s00376-023-3181-8

Accurate soil moisture (SM) prediction is critical for understanding hydrological processes. Physics-based (PB) models exhibit large uncertainties in SM predictions arising from uncertain parameterizations and insufficient representation of land-surface processes. In addition to PB models, deep learning (DL) models have been widely used in SM predictions recently. However, few pure DL models have notably high success rates due to lacking physical information. Thus, we developed hybrid models to effectively integrate the outputs of PB models into DL models to improve SM predictions. To this end, we first developed a hybrid model based on the attention mechanism to take advantage of PB models at each forecast time scale ( attention model). We further built an ensemble model that combined the advantages of different hybrid schemes ( ensemble model). We utilized SM forecasts from the Global Forecast System to enhance the convolutional long short-term memory (ConvLSTM) model for 1–16 days of SM predictions. The performances of the proposed hybrid models were investigated and compared with two existing hybrid models. The results showed that the attention model could leverage benefits of PB models and achieved the best predictability of drought events among the different hybrid models. Moreover, the ensemble model performed best among all hybrid models at all forecast time scales and different soil conditions. It is highlighted that the ensemble model outperformed the pure DL model over 79.5% of in situ stations for 16-day predictions. These findings suggest that our proposed hybrid models can adequately exploit the benefits of PB model outputs to aid DL models in making SM predictions.