On the improvement of cloud radiation in GRAPES global model and its forecast impact

The global numerical weather prediction model GRAPES at the National Meteorological Center of the China Meteorological Administration is identified with substantial systematic discrepancies from satellite-retrieved cloud covers, cloud water contents, and radiative fluxes. In particular, it produces insufficient total cloud cover and liquid water amounts, and consequently greatly underestimates cloud radiative forcings and causes substantial radiation budget errors. New parameterization schemes are then incorporated to more realistically represent cloud-radiation interactions. These include predictions for cloud cover, liquid water, and effective radius, as well as radiative effects of partial clouds and in-cloud inhomogeneity. As a result, radiation fluxes and cloud radiative forcings at both the surface and top of the atmosphere are made much closer to the best available satellite data, with global mean model biases reduced from 20-30 to a few W m-2. These improvements enhance model weather forecast skills for key surface variables including precipitation and 2m temperature as well as for height and temperature in the lower troposphere. Although non-trivial biases still exist, this study nonetheless represents the first essential step to correct the radiation imbalance before tackling other formulation deficiencies, such that significantly enhanced GRAPES weather forecast skills can be eventually achieved.