Coupled Atmospheric-hydrological Modeling For Hydrological Forecast Based On Multi-source Data Assimilation Containing Radar Data

Hydrological forecasting can provide the decision taking base for flood control and calamity reducing,reservoir operation,water supply and power generation.Project planning and construction are also need the information of hydrological forecasting.Generally,in order to improve the forecast lead time,the forecast accuracy will decline,while improving the forecast accuracy may reduce the forecast lead time.Coupled atmospheric-hydrological modelling with the data assimilation method can prolong the forecast lead time and improve the forecast accuracy at the same time.In this paper,the weather forecasting model are used to forecast the rainfall,which drives the hydrological model to obtain the runoff process.Data assimilation are used to improve the rainfall forecasting and the real-time correction method are used to further improve the accuracy of runoff process.Firstly,the WRF model was applied with 12 designed parameterisation schemes with different combinations of physical parameterisations,including microphysics,radiation,planetary boundary layer(PBL),land-surface model(LSM)and cumulus parameterisations.The selected study areas are two semi-humid and semi-arid catchments(Fuping and Zijingguan catchments)located in the Daqinghe River basin,northern China.The performance of WRF with different parameterisation schemes is tested for simulating eight typical 24-h storm events with different evenness in space and time.In addition to the cumulative rainfall amount,the spatial and temporal patterns of the simulated rainfall are evaluated based on a two-dimensional composed verification statistic.Regarding the individual parameterisations,Single-Moment 6(WSM6),Yonsei University(YSU),Kain-Fritsch(KF)and Grell-Devenyi(GD)are better choices for microphysics,planetary boundary layers(PBL)and cumulus parameterisations,respectively,in the study area.Among the 12 parameterisation schemes,Scheme 1,4,6,7 and 8 outperforms the other schemes with the best average performance in rainfall simulation.So the WRF multi-physics ensemble is composed by Scheme 1,4,6,7 and 8.Secondly,the radar data and conventional meteorological observations are assimilated using the three-dimensional variational(3D-Var)data assimilation method based on WRF-3DVar.Eleven data assimilation modes are designed for assimilating different combinations of observations in the two nested domains of the WRF model.Results show that the assimilation can largely improve the WRF rainfall products especially the accumulative process of rainfall,which is of great importance for hydrologic applications through the rainfall-runoff transformation process.Both radar reflectivity and conventional meteorological observations are good choices for assimilation in improving the rainfall products,whereas special attentions should be paid for assimilating radial velocity where unsatisfactory results are always found.Simultaneously assimilating conventional meteorological observations and radar data always perform better than assimilating radar data alone.The inclusion of conventional meteorological observations in the nested domains when radar reflectivity and radial velocity are assimilated in the innermost domain show the best results among all the 11 assimilation modes.The assimilation efficiency of the conventional meteorological observations is higher than both radar reflectivity and radial velocity considering the number of data assimilated and its effect.Furthermore,Each type of radar data is divided into seven data sets according to the height layers:(1)<500 m,(2)<1000 m,(3)<2000 m,(4)500?1000 m,(5)1000?2000 m,(6)>2000 m,and(7)all layers.The results show that radar reflectivity assimilation leads to better results than radial velocity assimilation.The accuracy of the forecasted rainfall deteriorates with the rise of the height of the assimilated radar reflectivity.The same results can be found when assimilating radar reflectivity and radial velocity at the same time.Finally,the best data assimilation mode is comfirmed:assimilating radar reflectivity(<500 m)in inner domain and conventional meteorological observations in outer domain at the same time.Finally,coupled atmospheric-hydrological modelling is structured by WRF model and Hebei model.The WRF model is used to forecast the rainfall,which drives the Hebei model to obtain the runoff process.The results show that the forecasted rainfall with data assimilation is much better than the forecasted rainfall without data assimilation for hydrological use as the input of Hebei model.Autoregressive moving average model(ARMA)can also help to improve the runoff forecast.For static forecasting,the error of flood volume and peak flow can reduce 1.03%?66.48%,1.55%?95.54%respectively.The Nash-Sutcliffe efficiency coefficient(NSE)for individual event can increase to 0.958.For dynamic forecasting,the forecast results become worse with the extension of the leading time for the four flood events.For event ?,?,?,the forecast results are still good enough for the leading time 6h,9h,9h,respectively.The forecast results of event VII are worse than the other three events,whereas the results are basically reliable.It means that coupled atmospheric-hydrological modeling based on multi-source data assimilation can obtain not only sufficient forecast accuracy but also plenty of leading time.The results of the study can provide references for hydrological forecast and technical support for flood control decision in semi-humid and semi-arid catchments in northern China.