Application of Observation Perturbation-based EDA Method to CMA Ensemble Forecasting

To facilitate the development of convective-allowing ensemble forecasting technology based on the China Meteorological Administration's (CMA) Mesoscale Model (CMA-MESO), an observation perturbation scheme is designed. This scheme further enhances the ensemble data assimilation (EDA) method for generating initial conditions for CMA-MESO convective allowing ensemble forecasting system. The design and distinctive characteristics of the observation perturbations are studied, and several severe convective events are analyzed. It can be concluded that the observation perturbation scheme developed for CMA-MESO aligns with actual observation error characteristics, it can address uncertainties in the model initial analysis field stemming from observations, and multiple sets of observations generated can effectively represent uncertainties in observations. Observation sensitivity experiments are conducted to explore the impact characteristics of observation perturbations, and a typical convective weather event in Beijing is analyzed, results indicate that observation perturbations primarily affect the short-range forecast performance of CMA-MESO model, causing relatively small forecast perturbations. The growth of perturbations reaches saturation within a 12-24 h forecast range, while the energy of observational perturbations gradually dissipates as the forecast range extends. Observational uncertainties significantly influence the local convective characteristics and the spatiotemporal distribution of convective-related elements in short-range forecasts. Based on observation perturbations, an EDA initial value perturbation scheme is constructed, and a convective-scale ensemble forecasting experiment with a 3 km resolution is conducted over the North China. Results indicate that EDA scheme can effectively generate initial perturbations for convective-scale ensemble forecasting. Compared to traditional dynamic downscaling methods, EDA scheme minimizes uncertainties arising from large-scale background fields in convective-scale ensemble forecasting, while emphasizing uncertainties that originate from observations. Ensemble forecast verification results indicate that EDA scheme can effectively enhance the reliability of element forecasts. Case studies of severe convective precipitation demonstrate that EDA scheme can improve the forecast accuracy of precipitation location and significantly enhance the effectiveness of precipitation probability forecasts. Results demonstrate the feasibility of constructing observational perturbations and EDA scheme in the development of CMA-MESO convective-allowing ensemble forecasting. Although ensemble spread may be slightly compromised due to data assimilation, there is a significant improvement in the quality of initial values for ensemble members and the accuracy of short-range forecasts, highlighting the practical application value of this method. Given that data assimilation only significantly impacts short-range forecasts, it remains essential to improve the associated model perturbation techniques to enhance the forecast performance of CMA-MESO convective-allowing ensemble forecasting for longer ranges.