Study on the influence of ECMWF short-term forecast field assimilation on regional mesoscale model forecast

Abstract

The forecast of the global numerical weather prediction model is often used as the background field to drive the regional mesoscale weather models, and its forecast quality has an important impact on the prediction skill of regional models. The mesoscale operational model in Central China uses the Global Forecast Field (NCEP GFS) of the US Center for Environmental Prediction as the background field, and its forecast accuracy needs to be improved. This paper proposes a method to improve the initial field of the model by using the three-dimensional variational method to assimilate the high-quality European Centre for Medium-range Weather Forecasts (ECMWF) fine-grid forecast field to improve the short-term forecasting ability of the model. At first, the error characters of the 12 h forecast products of NCEP GFS and ECMWF were analyzed by using sounding observations. The RMSE (root mean square error) of temperature, horizontal wind field and relative humidity of ECMWF 12 h forecast are smaller than those of NCEP GFS forecast. Second, sensitivity experiments of ECMWF forecast field assimilation with different resolutions were conducted. Finally, 1°×1° resolution was selected based on the sensitivity experiments and a series of the data assimilation experiments with 1°×1° resolution was performed for August 2021. The results are as follows. (1) The assimilation forecast of heavy rain from 11 to 13 August 2021 shows that after 12 hours of spin-up, the forecast error of the element field has been significantly improved, especially at the bottom of the model. The TS score of 12-36 h, 36-60 h and 60-84 h cumulative precipitation has a certain improvement, especially the forecast of rainstorm magnitude. (2) ECMWF forecast field assimilation sensitivity experiments with different resolutions show that the forecast effect of ECMWF forecast field with 1°×1° resolution is better than 0.5°×0.5° and 0.25°×0.25°. (3) A series of the data assimilation experiments for August 2021 shows that the assimilation of ECMWF 12 h 1°×1° predicted variables has a lower RMSE of the 12 h, 36 h and 60 h forecasted temperature, humidity and wind field in the vertical direction than the control experiments, and the TS score was significantly improved, especially for 50 mm heavy rain precipitation, with an increase of 13.33%, which can effectively improve the forecast skills of the Wuhan Mesoscale Model.

Keywords

• ecmwf
• 3dvar
• heavy rain
• wuhan mesoscale model