A feasibility study of the cosine analysis constraint method for optimizing initial perturbations of convective-scale ensemble prediction

To improve forecast skill of extreme weather events, it is of primary importance to construct accurate initial conditions for a convective-scale ensemble prediction system (EPS). The traditional initial perturbation schemes, e.g., dynamic downscaling, fail to capture the amplitude or structure of convective-scale forecast errors accurately, especially near steep terrain or meso- and small-scale systems during weather system's rapid change. In this study, we developed a new initial perturbation optimizing technique, namely the cosine analysis constraint method. This method is then used to improve the downscaled initial perturbations by introducing smaller-scale information from the analysis increments, generated from data assimilation. We demonstrate the feasibility of the cosine analysis constraint method in the China Meteorological Administration (CMA) convective-scale EPS. Using the control experiment (CTRL) without any modification to the initial perturbation scheme as a reference, we designed the cosine analysis constraint experiment (CONS) and compared it with CTRL. We selected a case study of convective precipitation and two groups of one-month experiments were initialized to verify the feasibility of the new method and exclude case dependence. The results of the one-month test show that adopting the cosine analysis constraint method to optimize the initial perturbations can effectively enhance the consistency between the ensemble mean and the corresponding reanalysis field (regarded as observations). In the case study, the larger horizontal distribution of precipitation spread in CONS indicated the location of convective precipitation more effectively, which is important for operational weather forecasting. The significant effect of the moisture process was confirmed, especially in CONS. The verification results of the entire study domain were significantly improved after the initial perturbations were rescaled. Overall, the forecast skill of meteorological fields at different pressure levels and the extreme precipitation of smaller-scale convective systems were enhanced, which illustrated the potential of the cosine analysis constraint method to improve the quality of initial perturbations in CMA convective-scale EPS. • The consistency between the forecasts and observations is enhanced by the cosine analysis constraint method significantly. • The forecast performance of CMA convective-scale EPS has been improved. • The new method could alleviate the underdispersion problem of meteorological fields, especially in the lower troposphere.