Research On Flood Forecasting In The Upper Reaches Of The Huaihe River Basin Based On Radar Now Rainfall Forecasting

In recent years,flood events caused by sudden severe convective weather are more and more frequent,and the accuracy requirements of flood forecasting are also improved.Radar extrapolation based rainfall nowcasting can obtain high spatial and temporal resolution precipitation information in the next 6 hours.Applying it to hydrological forecasting by means of meteorological hydrological coupling forecasting will help to solve the uncertainty of Meteorological Elements Forecasting under the condition of climate change,improve the accuracy of flood forecasting and extend the flood forecasting period,which is the main trend in the field of Hydrometeorological forecasting at home and abroad.Based on the historical rainfall data in the upper reaches of the Huaihe River,a fully distributed hydrological model(CREST)based on grid structure was applied to flood forecasting.Firstly,the model parameters were calibrated and verified by rainfall data of rainfall station,and the final mean Nash efficiency coefficient of the model was 0.78,which reached the level of grade B flood forecast.Then,two typical flood processes in 2016 were selected,combined with rainfall observation data of Fuyang radar station,and the radar rain inversion data were evaluated and simulated.Finally,the short term ensemble precision system of British meteorological office was used,STEPS(Short Term Ensemble Precipitation System)is used to forecast the maximum rainfall time of the basin flood process,which is close to 3 hours,2 hours,1 hour and 30 minutes.According to different forecast time,25 members of the rainfall forecast set are generated.Finally,the flow of each forecast process is simulated by CREST hydrological model to analyze the combination effect of different rainfall forecast data,different forecast periods and distributed hydrological model,as well as the impact of rainfall spatial-temporal distribution on the basin runoff process.The main conclusions are as follows:(1)Comparing radar quantitative precipitation estimation with rainfall stations,QPE(Quantitative Precipitation Estimate)can better reflect the temporal and spatial distribution and movement of rainfall.In the analysis of two flood processes,the simulation of flood peak discharge by radar QPE is closer to the measured value than that by rainfall stations,but the overall simulation effect of the flood process has its own advantages and disadvantages according to the Nash efficiency coefficient.This shows that it is necessary to improve the radar rainfall inversion formula and calibrate the parameters to improve the accuracy of radar rainfall inversion.(2)The flood simulation results of radar rainfall nowcast forecast show that the accuracy of radar rainfall nowcast forecast decreases with the increase of forecast time.Within one hour forecast time,the accuracy of radar rainfall near forecast is very close to radar rainfall.In the flood process of July 20,2016,from the 3-hour forecast period to the 30 minute forecast period,the peak error of radar deterministic forecast decreased from-26.7% to-19%,and the Nash efficiency coefficient increased from 0.705 to 0.740;in the flood process of October 28,2016,from the 3-hour forecast period to the 30 minute forecast period,the peak error of radar deterministic forecast decreased from-34.5% to-14.25%,and the Nash efficiency coefficient increased from 0.624 to 0.823?(3)The ensemble forecast results of two flood processes show that the flood process of rainfall ensemble forecast in hydrological simulation can effectively include the measured rainfall flood peak process in the previous flood season,and the longer the forecast period is,the more effective the forecast of flood peak magnitude will be compared with the deterministic forecast.At the same time,the flow hydrograph of ensemble forecast in the whole flood process is consistent with the flow hydrograph simulated by rainfall station and radar QPE.(4)In the radar rainfall forecast within 3 hours,the error range of flood peak forecast can be quantified by rainfall ensemble forecast.With the setting of different flow thresholds,the nearer the arrival time of the maximum flood peak,the smaller the probability of low and medium risk of flood occurrence,and the greater the probability of high risk of flood occurrence.(5)The results show that the simulation effect is good and can meet the requirements of real-time flood forecasting.In the flood process of July 20,2016,from the 3-hour forecast period to the 30 minute forecast period,the Nash efficiency coefficient of the radar deterministic forecast is greater than 0.7,reaching the level B forecast standard;in the flood process of October 28,2016,from the 3-hour forecast period to the 30 minute forecast period,the Nash efficiency coefficient of the radar deterministic forecast is close to or greater than 0.7,also reaching the level B forecast standard.