Data Assimilation for Rainfall-Runoff Prediction Based on Coupled Atmospheric-Hydrologic Systems with Variable Complexity

摘要

The data assimilation technique is an effective method for reducing initial condition errors in numerical weather prediction (NWP) models. This paper evaluated the potential of the weather research and forecasting (WRF) model and its three-dimensional data assimilation (3DVar) module in improving the accuracy of rainfall-runoff prediction through coupled atmospheric-hydrologic systems. The WRF model with the assimilation of radar reflectivity and conventional surface and upper-air observations provided the improved initial and boundary conditions for the hydrological process; subsequently, three atmospheric-hydrological systems with variable complexity were established by coupling WRF with a lumped, a grid-based Hebei model, and the WRF-Hydro modeling system. Four storm events with different spatial and temporal rainfall distribution from mountainous catchments of northern China were chosen as the study objects. The assimilation results showed a general improvement in the accuracy of rainfall accumulation, with low root mean square error and high correlation coefficients compared to the results without assimilation. The coupled atmospheric-hydrologic systems also provide more accurate flood forecasts, which depend upon the complexity of the coupled hydrological models. The grid-based Hebei system provided the most stable forecasts regardless of whether homogeneous or inhomogeneous rainfall was considered. Flood peaks before assimilation were underestimated more in the lumped Hebei model relative to the other coupling systems considered, and the model seems more applicable for homogeneous temporal and spatial events. WRF-Hydro did not exhibit desirable predictions of rapid flood process recession. This may reflect increasing infiltration due to the interaction of atmospheric and land surface hydrology at each integration, resulting in mismatched solutions for local runoff generation and confluence.