Application Of Two Ensemble-based Methods To The Simulations Of Two Convective Cases

The ensemble Square Root Filter(En SRF) and three-dimensional assimilation method based on the forecast-ensemble and singular value decomposition(SVD-En3DVar) are used to assimilate real radar observations to simulate two convective cases occurred on summer over Jiangsu Province of China. All experiments are conducted with the Weather Research and Forecasting(WRF) model to investigate the effectiveness and ability of two ensemble-based assimilation methods. Meanwhile, the identical assimilation experiments are performed with WRF 3DVar as a comparison. Furthermore, a simplified scheme is proposed based on En KFto reduce the computation cost of the En KF. The simplified scheme is estiblished based on En SRF, andits effectiveness and ability is examined with the WRF model and a convective case. The experimental results are shown as follows:(1) Both SVD-En3 DVar and En SRF methods have preferable capability of assimilating radar observations.The meso-micro scaleinformationis introduced effectivelyinto initial fields by adjusting horizontal wind, equivalent potential temperature and water vapour fields. In the area where strong reflectivity exists, the wind convergence becomes stronger, the vapour mixing ratio is increased and the equivalent potential temperature isolines are denser. However, the vapour mixing ratio, wind direction and wind speed have some differences between the two methods in some area.(2) The two ensemble-based methods both improve forecastskill of thecomposite reflectivity. The location and evolution with time of composite reflectivityfrom the SVD-En3 DVar and En SRF methods are consistent with the observed, but the performance of two methods is different for two convective cases. For the squall-line case in June, 2009 on Jiangsu, SVD-En3 DVar outperforms WRF 3DVar and En SRF. For the other convective case happened on 2013, En SRF obtains the best forecast of composite reflectivity compared with SVD-En3 DVar and WRF 3DVar.(3) It is documented from two convective cases that the assimilation skill of the SVD-En3 DVar method is slightly sensitive to horizontal localization parameter for the convective-scale weather system. For the squall line case, the horizontal localization radius of about 50 km can achieve the best prediction of composite reflectivity; but for the severe rainfall case, the SVD-En3 DVar method with horizontal localization radius of about 15 km performs better.(4) Byanalyzing error variance of horizontal wind, water vapour and other variablesthat calculatingfrom forecast ensemble at adjacent assimilation cycle, it is found that the error variance from the forecast ensemble at the start time of assimilation is smoother, the error variances of adjacent forecast ensembles produced by assimilation and forecast process demonstrate small-scale features and they are close both on the value and distribution shape.(5) The traditional En KF has to update the forecast ensembles each assimilation process and it is time assuming and high computing cost. Given that the error variance change little awithin the assimilation process, a simplified scheme is proposed based on En SRF(marked with SS_En SRF) in which the update frequency of forecast ensemble is reduced to save some computational cost.The preliminary experimental results with a severe rain happened on 2013 show that the SS_En SRF performaned well in assimilating radar observation. The assimilation and forecast skill of SS_En SRF can catch up with that of En SRF when the forecast ensemble from the assimilation and forecast process is used. Besides, SS_En SRF performs better than WRF 3DVar in the forecast of composite reflectivity.