The Hydrometeorological Evaluation Of Multiple Quantitative Precipitation Estimation And Radar_based Nowcasting Precipitation: Ganjiang River Baisn

Hydrological model is the important tool for the hydrometeorological researches, it plays an indispensable role in streamflow simulation, water resource management, and flood forecast. Precipitation is one of the most important forcing data for hydrological model. The precipitation products may be divided into two types: quantitative precipitation estimation(QPE) products and quantitative precipitation forecast(QPF) products. QPE products could include gauge observations interpolation precipitation, ground-based weather radar products and satellite remote sensing retrieval products. QPF products could include the forecast precipitation products based on Numerical Weather Prediction(NWP) models and the products based on radar echo or other initial data ext rapolation approaches. In this study, we select Ganjiang River Basin which is one of the typical rainstorm regions in China as the study area and used the QPE products and QPF products as the driving data for the hydrological model.We develope statistically and hydrologically assess the hydrological utility of the latest Integrated Multi-satellit E Retrievals from Global Precipitation Measurement(IMERG) multi-satellite constellation over the mid-latitude Ganjiang River basin in China. The reference data are multi-radar-based Quantitative Precipitation Estimation(QPE) product(RQPE), and a dense network of 1200 gauges to find the best precipitation. The investigations are conducted at hourly and 0.1° resolution throughout the rainy season from March 12 to September 30, 2014(period 1: March 12 to May 31; period 2: June 1 to September 30). For the implementation of the study, we compare IMERG and RQPE with gauge data, respectively, to evaluate the statisitally performance and hydrologically utility satellite IMERG. The results indicate that both remote sensing products can estimate precipitation fairly well over the basin, but RQPE significantly outperforms IMERG in almost all the studied cases. The correlation coefficients of RQPE(period 1:CC = 0.98 and CC = 0.79,period 2 :CC = 0.97 and CC = 0.59) are greater than those of IMERG(period 1:CC = 0.79 å'Œ CC = 0.34,period 2 :CC = 0.82 and CC = 0.33) at basin and grid scales. The hydrological assessment is conducted with the Coupled Routing and Excess Storage(CREST) model based on multiple parameterization scenarios, calibrated using the gauge reference, RQPE, and IMERG respectively. During the calibration period(from March 12 to May 31), the streamflow simulated from the gauge network exhibits the highest Nash–Sutcliffe coefficient efficiency(NSCE) value(0.92), closely followed by the RQPE(NSCE = 0.84), while IMERG performs just acceptable(NSCE = 0.56). During the validation period(June 1 to September 30), the three rainfall datasets were used to force the CREST model based on their respectively calibrated parameter sets. RQPE outperforms gauge and IMERG in all validation scenarios, possibly due to its advantageous capability in capturing high space-time variability of precipitation systems in the humid climate during the validation period. Overall, RQPE and the reference gauge exhibit similarly better performance compared with the newly available IMERG data, both given their statistical and hydrological analysis in the Ganjiang River basin. Although n ot quite as good as RQPE and gauge data, the only recently available IMERG has good hydrological performance based on gauge- and RQPE- calibrated parameters. IMERG has huge room to improve given the fact that its predecessor, the Tropical Rainfall Measuring Mission(TRMM), has continuously upgraded its algorithms from Version 0 to Version 7 over the satellite’s 17 year life span. Therefore, future studies should promote the hydrological application of RQPE datasets at local scales, and continuously improve the IMERG Day-one algorithm from its Global Precipitation Measurement(GPM) satellite constellation.For the QPN products, we selected the ―data-driven‖ extrapolation approaches based on radar echo. The radar echo tracking algorithm called multi-scale tracking radar echoes(MTREC) and the pixel-based algorithm called Pixel- Based Nowcasting(PBN) for Quantitative Precipitation Forecasting using radar-based rainfall data was used in this study. To analysis the nowcasting ability of MTREC and PBN, and the applicability of CREST hydrological model for the two nowcasting precipitation products, three rainstorms events in Ganjiang River basin during 2014 rainy season were selected. Both of MTREC’s and PBN’s forecasting precipitation were compared with RQPE. The results indicate that:(1) Along with the lead time increasing, the forecasting abilities of the two nowcasting method are getting worse. Compared with the PBN method, the changing of the forecasting abilities for MTREC method is gentle.(2) the precipitation magnitude forecasted by the MTREC method is smaller than RQPE, and the low precipitation forecasted by the MTREC method is closer to RQPE; the precipitation magnitude forecasted by the PBN method is smaller than RQPE, and the high precipitation forecasted by the PBN method is closer to RQPE;(3) the precipitation area forecasted by the MTREC method is smaller than RQPE, the low precipitation area forecasted by the MTREC method is closer to RQPE; the precipitation area forecasted the PBN method is larger than RQPE, and the high precipitation area forecasted by the PBN method is closer to RQPE.(4) As the lead time increases, the changing for probability distribution function(PDF) of precipitation forecasted by the MTREC method is stable and matches well with the PDF of RQPE. While, the PDF of the precipitation(>0.4mm/h) forecasted by the PBN method is larger than that of RQPE. The PDF of the precipitation(<0.4mm/h) forecasted by the PBN method is smaller than that of RQPE.(5)In terms of hydrological simulation results, the hydrological simulation driven by the forecasting products from MTREC method and PBN method during the three events had excellent hydrological performance,especially for PBN method. As the lead time increases, the hydrologic skill of precipitation forecasted by the PBN method is getting worse obviously.(6) The streamflow simulated by precipitation forecasted by the PBN method overestimates the observed streamflow and the streamflow simulated by RQPE. While, The streamflow simulated by precipitation forecasted by the MTREC method underestimates the observed streamflow and the streamflow simulated by RQPE.