Meteorological–hydrological coupling flood forecast and error propagation characteristics based on radar data assimilation in small- to medium sized River Basin: a case study of Zhanghe River Basin in China

In small- to medium sized river basins, flood forecast accuracy and adequate lead times are especially important for the scheduling of catchment management decisions, involving flood prevention measures and disaster mitigation. For this study, the Zhanghe River Basin in China was selected as the study area. A meteorological–hydrological coupled model, which linked the Weather Research and Forecasting (WRF) model to the WRF-Hydro model was used with Doppler radar data, including radial velocity and reflectivity in the S-band, to explore the influence of data assimilation frequency on rainfall and runoff forecasts, as well as the differences in error propagation characteristics between meteorological and hydrological models. The results were as follows: (1) Doppler radar data assimilation has the ability to improve the temporal and spatial variability of rainfall forecasts. Appropriate data assimilation show positive effect on improving the rainfall forecast. 3 h assimilation intervals data assimilation may result in over-estimating under the influence of complex topography in Zhanghe River Basin. The rainfall forecast results based on 6 and 12 h assimilation intervals were more accurate than those derived from a 3 h interval, with the average cumulative rainfall errors being reduced by 44.86% and 53.26%, respectively. (2) Rainfall forecasts have a significant impact on the accuracy of subsequent runoff forecasts. The runoff results showed that the assimilation of radar data at higher frequencies does not guarantee the further improvement of the runoff simulations due to the overestimation of forecast rainfall. The average flood peak error under the 6 and 12 h assimilation intervals was 27.52% and 20.0%, respectively, less than that using the 3 h interval. Therefore, the effective information contained in the assimilation data is more important than the amount of data. (3) Error propagation between models differs with the changing assimilation frequency of the radar data and the consequent effect of the rainfall forecast. With the increase in assimilation frequency of the radar observations, the error range increases. Compared with the rainfall errors, the runoff errors show greater variability. Through quantitative analysis, it was found that there is no well-defined linear relationship between the rainfall and runoff errors. At the same time, the potential of radar data assimilation is discussed, and some suggestions for improvement are put forward.