Coupling deep learning and physically-based hydrological models for monthly streamflow predictions

<p>Revision in journal Water Resources Research, Manuscript number: <strong><span>2023WR035618R</span></strong></p> <p><strong>Abstract:</strong><strong>&nbsp;</strong>This study proposes a new hybrid model for monthly streamflow predictions by coupling a physically-based distributed hydrological model with a deep learning (DL) model. Specifically, a simplified hydrological model is first developed by optimally selecting grid cells from a distributed hydrological model according to their soil moisture characteristics.&nbsp;<span>It</span>&nbsp;is then driven by bias corrected general circulation model (GCM) <span>prediction</span>s to generate soil moistures for the forecasting months. Finally, model-simulated soil moisture along with other predictors from multiple sources are used as inputs of the DL model to predict future <span>monthly </span>streamflows. The proposed hybrid model, using the simplified Variable Infiltration Capacity (VIC) as the hydrological model and the combination of Convolutional Neural Network and Gated Recurrent Unit (CNN-GRU) as the DL model, is applied to predict 1-, 3-, and 6-month ahead&nbsp;<span>reservoir </span>inflows&nbsp;<span>for the Danjiangkou Reservoir in China. </span>The results show that the hybrid model consistently performs better than VIC and CNN-GRU models with great improvement in Kling‐Gupta efficiency (KGE) values for lead times up to 6 months. <span>Additional tests indicate that hybrid</span>&nbsp;model<span>s based on CNN-GRU </span>outperform&nbsp;<span>those based on</span>&nbsp;<span>LASSO, XGBoost, CNN, and GRU models. Moreover, compared with the distributed hydrological model, the hybrid model</span>&nbsp;greatly reduce<span>s</span>&nbsp;the <span>computation </span>burden of rolling prediction<span>. It also </span>saves decision-makers the time and effort of trying different combinations of predictors<span>, which is indispensable when building DL models. Overall</span>, the new hybrid model <span>demonstrates great potential</span>&nbsp;for monthly streamflow prediction&nbsp;<span>where</span>&nbsp;training data are limited.</p> <p><strong><span>Keywords:</span></strong>&nbsp;<span>monthly streamflow prediction; deep learning; </span><span>physically-based distributed hydrological model; </span><span>VIC model; soil moisture; hybrid model </span></p>