Hydrograph and recession flows simulations using deep learning: Watershed uniqueness and objective functions

• Globally trained deep learning models do not always perform the best. • Adding global model outputs as input to a locally trained model. • Recession flow simulation may be improved by combining global and local information. • Watershed uniqueness may have significant control over global model performance. This study examines streamflow simulations using deep learning (DL) to understand the information extraction capability of global DL models trained on multiple watersheds. The study examined the entire streamflow time series and recession flow predictions separately. It introduces a global–local (GL) modeling strategy, where global model outputs are fed as input to a locally trained model, with the hypothesis that the local model can leverage watershed-specific information that the global model may miss. GL models demonstrate enhanced accuracy in recession flow prediction for 30% of the watersheds compared to global and local models. However, considering the entire hydrograph, GL models often perform worse than the global model. Further, the DL models were trained on two different objective functions. The performance of the global model in a watershed depended strongly upon the objective function used. These results suggest that watershed uniqueness plays a significant role in the performance of global models, suggesting that even a global DL model should be tailored to individual watersheds for optimal performance.