Comparing the performances of WRF QPF and PERSIANN-CCS QPEs in karst flood simulations and forecasting with a new Karst-Liuxihe model

Abstract. Long-term, available rainfall data are very important for karst flood simulations and forecasting. However, in karst areas, there is often a lack of effective precipitation available to build distributed hydrological models. Forecasting karst floods is highly challenging. Quantitative precipitation forecasts (QPF) and estimates (QPEs) could provide rational methods to acquire the available precipitation results for karst areas. Furthermore, coupling a physically-based hydrological model with the QPF and QPEs felicitously could largely enhance the performance and extend the lead time of floods forecasting in karst areas, the performance of coupling the Weather Research and Forecasting Quantitative Precipitation Forecast (WRF QPF) and Precipitation Estimations through Remotely Sensed Information based on the Artificial Neural Network-Cloud Classification System (PERSIANN-CCS QPEs) with a new fully distributed and physical hydrological model, the Karst-Liuxihe model in flood simulations and forecasting in karst area. This study served 2 main purposes: one purpose is to compare the performances of WRF QPF and PERSIANN-CCS QPEs for rainfall forecasting in karst river basins. The other purpose is to test the effective feasibility and application of the karst flood simulation and forecasting by coupling the 2 weather models with a new Karst-Liuxihe model. The new Karst-Liuxihe model improved the structure of the model by adding the karst mechanism based on the Liuxihe model as follows: (1) Refine the model structure and put forward the concept of karst hydrological response units (KHRUs) in the model. The KHRU, as the smallest unit of the Karst-Liuxihe model, is defined in this paper to be suitable for karst basins; (2) Increase the calculations of water movement rules in the epikarst zone and underground river, such as the division of slow flow and rapid flow in the epikarst zone and the exchange of water flow between the karst fissures and conduit systems; thus, the convergence of the underground runoff calculation method is improved to be suitable for karst water-bearing media; and (3) Add some necessary hydrogeological parameters in the coupled model to reflect the true conditions of rainfall-runoff in the karst underlying surface. Moreover, the flood detention and peak clipping effects due to the upstream karst depressions during flooding were considered and reasonably calculated in the coupled model. The flood detention effect can affect the peak flow time error simulated in the model and make the true peak flow appear later; the flood peak clipping effect can affect the flood peak flow relative errors and the simulation errors of floods volume. The consideration of these 2 factors in the model makes the flood simulations and forecasting effects more credible. The rainfall forecasting result show that the precipitation distribution of the 2 weather models was very similar compared with the observed rainfall result. However, the precipitation amounts forecasted by WRF QPF were larger than that measured by the rain gauges, while the quantities were smaller by the PERSIANN-CCS QPEs. A postprocessing algorithm was adopted in this paper to correct the rainfall results by the 2 weather models. The karst flood simulation and forecasting results showed that the flood peak flow simulations were better by coupling the Karst-Liuxihe model with the PERSIANN-CCS QPEs, and coupling the Karst-Liuxihe model with WRF QPF could extend the lead time of flood forecasting largely, as a maximum lead time of 96 hours can provide an adequate amount of time for flood warnings and emergency responses. The satisfying and rational karst flood simulation evaluation indices proved that coupling the 2 weather models with the new Karst-Liuxihe model could be effectively used for karst river basins, which provides great practical application prospects for karst flood simulations and forecasting. In addition, the postprocessing method used to revise the 2 weather models in this paper is feasible and effective, and this method can largely improve the coupled model application effectiveness and prospect in karst river basins.