Abstract: In this study, we apply the Beijing flood forecast model and both the gauge-measured and radar-monitored rainfall data to reassess the "July 2023" flood event that occurred in the key regions of Beijing. Results reveal that the radar-derived rainfall data closely align with ground observations, offering a more nuanced representation of the rainfall's temporal and spatial variations. Comparative evaluation of the forecasting capabilities based on these rainfall datasets demonstrates their substantial equivalence, affirming the radar data's viability as a credible alternative to ground measurements. Our specialized Beijing flood forecast model, meticulously tailored to the distinctive runoff characteristics of the city’s mountainous areas, consistently exhibits a high accuracy across a wide range of scenarios. The intricate hydrological processes in the city's mountainous terrains are inherently nonlinear; the parameters of its hydrological model, often derived from the historical floods of varying magnitudes, inherently harbor uncertainties. Recognizing the dynamic nature of runoff and flood events, we emphasize the necessity of proactive model parameter optimization. This optimization procedure should integrate real-time conditions and the most current data so as to bolster the reliability of flood predictions.

Key words: flood hindcasting, flood forecasting, "July 2023" flood, radar retrieval rainfall