Forecast experiments of revised cloud and convection schemes in NMC T213L31 global model

In order to solve the obvious underestimation of moderate to extraordinary heavy rainfall in the T213L31 operational global model of China, this paper diagnoses and analyzes the precipitation forecast performance of the model in 2005. The results show that the predicted total precipitation is largely composed of grid scale rainfall which can reach and even exceed the strength of observed total precipitation in regions where overestimation is the most remarkable. This implies that the main reason for inaccurate precipitation forecast is the overestimation/underestimation of resolvable/unresolvable scale rainfall. Causes for overestimation of resolvable scale rainfall include flaws in the initialization of soil moisture and cloud related variables, and in the cloud and convective parameterizations which directly produce precipitation. This paper is devoted to analyzing and improving the cloud and convective parameterizations. Specific measures include: adding one more calling of the cloud scheme before the convection scheme is called; changing the convective closure from “dynamical” to “Convective Available Potential Energy (CAPE) adjustment”; using more complex convection triggering mechanisms; and improving calculations of ice sedimentation and precipitation flux. The main purpose is to make the convective scheme more active, thus reducing the grid scale convection. With these improvements, sensitivity experiments and continuous assimilation and forecast experiments for the summer of 2005 are carried out, and the results are compared with rainfall observations at 400 standard stations in China and GPCP global observations. Comparative results show that revised schemes have a better capability than the original operational schemes in forecasting both China and global summer rainfall, except that the overestimation of rainfall still exists in the Sichuan Province and the equatorial eastern Pacific. In addition, statistical verifications of precipitation forecasts also indicate that the threat scores of precipitation forecasts for most rainfall categories except light rainfall, are increased, and biases are reduced, for most of the forecast leading times.