Study on the Influence of Temporal and Spatial Resolution of Rainfall Data on Watershed Flood Simulation Performance

To investigate the impact of temporal and spatial resolution of rainfall data on watershed flood simulation performance, the rainfall data from meteorological stations and the gridded rainfall data from meteorological forecasts for a rainfall event were adopted in this study. Interpolation methods were applied to generate rainfall processes with different spatial and temporal resolutions. A hydrodynamic model was employed to simulate the flow rates at various sections of the watershed under different rainfall scenarios. The results show that as the spatial and temporal resolutions decreased, the flood variation patterns at various sections remained consistent. Namely, the determination coefficient (R2) decreased, whereas the root means square error (RMSE) and mean absolute error (MAE) increased, and the errors in peak flow rates and the fluctuation amplitudes of the flow rates at the sections increased as well. Moreover, a decrease in temporal resolution led to a delay in the peak flow timing. Significant differences were observed between the simulation results generated from the two different rainfall datasets. The R2 values for the simulated flow rates at each section were all above 0.75 for the observed rainfall data, while 40% of the results based on meteorological forecast data were below 0.5. Overall, the simulation results using observed rainfall data outperformed those using meteorological forecast data. Through the comparative analysis of simulation results including the rainfall characteristic parameters such as the watershed-averaged precipitation (AVP) and the coefficient of variation (CV), it was found that AVP had a strong correlation with the peak flow and its increase or decrease directly affected the peak flow. On the contrary, CV showed a negative correlation with the peak flow.