Study On Multi-scale Initial Condition Perturbation Schemes In Convection-allowing Ensemble Forecasting

The focus of this reaseach is on the initial condition perturbation(ICP) and lateral boundary perturbation(LBP) methods, both of which are affected by the muti-scale interactions in an convection-allowing ensemble forecast system. There are four components to this research. First, a WRF-ETKF update and forecast system is constructed with one-way double nested(resolutions are 4 and 24 km). The original perturbation for each domain is downscaled from NCEP global ensemble forecast system(GEFs) using cluster method, then perturbation for inner domain get updated with ETKF method with large scale LBP from GEFS during each updating cycle. Second, the impacts of inconsistencies between ICP and LBP are evaluated in a case study as well as the chosen technology of Blending zone. Third, the multi-scale evaluation of two initial perturbation methods(ETKF and dynamical downscaling, DOWN) and their interaction with large scale LBP under two different synoptic-scale forcing heavy rain cases(case one is Beijing "7.21" extreme event and case two is 2014 Anhui squall line event) are examined. Fourth, blended LBP also with improved Blending ICP are proposed and examined in case two which has weaker synoptic-forcing than case one. The conclusions are written as follows,1.The ICP obtained from ETKF cycle with large scale LBP from GEFs contains multi-scale uncertainty, but it can not satisfy the consistency between LBP which leading to spurious gravity waves at the lateral boundary; The multi-scale evaluation characteristics of ETKF and DOWN are also studied during Beijing "7.21" heavy rain event using Haar wavelet scale decomposition in order to get the most reasonable Blending wavelength. Basing on that, the fast-growing components of ETKF are extracted and blended with the large scale component of the perturbation from GEFS. The optimum perturbation(BLEND) cured the inconsistencies between ETKF and LBP, at the same time the forecast skill for front precipitation period of "7.21" case has been improved compared to ETKF.2.While under strong large scale forcing, ETKF has more total perturbation energy in early hours and DOWN in later hours when both ICP and LBP were considered; while the large scale forcing is weak, ETKF has better perturbation energy in whole forecast period, indicating the importance of consistency between initial and lateral perturbation depending on cases, so does the multi-scale evaluation of perturbation(ETKF and DOWN). When it comes to the interaction with LBP, initial perturbation always dominating the first several hours and LBP the following hours. The dominating hours of initial perturbations is more under strong large scale forcing than weak large scale forcing. Also the synoptic-scale forcing has significant impact on quantitative precipitation forecast and probability forecast skill.3.Based on the results above, blended LBP was designed which combined large scale perturbation from GEFs and mesoscale perturbation from ETKF cycle in 24 km domain. Then the blended LBP was blended with ETKF perturbation in 4 km domain. The double-Blending method was named SSETKF here which has best performance in case 2.Up to now the optimal design of convection-allowing ensemble forecast systems remais largely unknown in the whole world, this study has important scientific significance not only for initial and lateral boundary perturbation schemes in the convection-allowing ensemble forecast systems, but also providing the basis for operational application.