Impact of spatial distribution information of rainfall in runoff simulation using deep learning method

Abstract. Rainfall-runoff modeling is of great importance for flood forecast and water management. Hydrological the traditional commonly used approach rainfall-runoff modeling. In recent years, with development artificial intelligence technology, deep learning models, such as long short-term memory (LSTM) model, are increasingly applied to However, current works do not consider effect rainfall spatial distribution information on results. Focusing 10 catchments from Catchment Attributes Meteorology Large-Sample Studies (CAMELS) dataset, this study compared performance LSTM different look-back windows (7, 15, 30, 180, 365 d) future 1 d discharges multi-day simulations 15 d). Secondly, differences between LSTMs individual models trained independently in each catchment regional were also across catchments. All driven by mean data spatially distributed data, respectively. The results demonstrate that regardless whether or improves data. a model did obtain better than an our study. we found using can reduce difference model. summary, (a) adding about another way improve where long-term records absent, (b) understanding utilizing help runoff simulations.