Key Technologies of CMA-MESO and Application to Operational Forecast

Abstract

To meet the requirement of numerical weather prediction for local severe convective weather, especially disastrous weather and extreme weather events, based on GRAPES-MESO 10 km system, many works have been completed, which include improving the calculation accuracy and stability of the model dynamic framework, selecting and testing the physical parameterization schemes suitable for high-resolution model, establishing a national radar quality control preprocess system, applying the national (SA/SB/CB) three-dimensional network mosaic data through the cloud analysis system, establishing a convective resolvable assimilation system and land surface data assimilation system for small and medium-scale systems, implementing the assimilation and application of unconventional local dense data such as radar radial wind, wind profile radar, FY-4A imager emissivity, satellite cloud motion wind, satellite GNSSRO, surface precipitation and the near surface data, and developing the rapid cycle technology. By integrating all the jobs mentioned above, the nationwide rapid analysis and forecast system CMA-MESO (GRAPES-MESO 3 km)has been established and put into operational run since June 2020 with 3 km horizontal resolution and 3 h time interval. The operational verification results in flood season from June to September of 2020 show that the forecasts of near surface elements (precipitation, 2 m temperature and 10 m wind) of CMA-MESO forecast surpass the results of GRAPES-MESO 10 km system, and the threat score for 3 h accumulated precipitation forecast is outstanding. The threat score for 24 h accumulated precipitation of CMA-MESO is slightly worse than the result of ECMWF, but the threat score for 3 h accumulated precipitation forecast is significantly better. For the precipitation exceeding 10.0 mm, CMA-MESO performs better than ECMWF within all the lead times, and the advantages are more obvious with the increase of precipitation threshold. Compared to ECMWF, CMA-MESO shows more obvious advantages on daytime forecast. For 25 mm precipitation threshold, the improvement rate exceeds 50% in most of the daytime and reaches about 100% in the later stage of forecast. The spatial distribution of mean 24 h accumulated precipitation predicted by CMA-MESO and ECMWF models is close to the observation, but the amount predicted by CMA-MESO is slightly larger. The frequency and intensity of precipitation simulated by CMA-MESO, which can characterize the ability of model to predict the spatial-temporal fine characteristics of precipitation, are consistent with observation in terms of both horizontal distribution and magnitude. The comprehensive performance of CMA-MESO in flood season in China exceeds that of ECMWF fine grid model.

Keywords

• rapid analysis and forecast system
• local dense data application
• kilometer scale variational framework
• high-resolution model