Revealing Bias of Cloud Radiative Effect in WRF Simulation: Bias Quantification and Source Attribution

Accurate prediction of cloud radiative effect (CRE) is important to weather forecast and climate projection, and solar energy production?a major renewable energy source toward decarbonization. Here, we evaluate the capability of the Weather Research and Forecast (WRF) model to simulate solar irradiance on a short-term timescale (days) against observations in a remote region in north China. Results illustrate that our WRF simulation systematically underestimates the CRE and three error sources are identified: (a) incorrectly predicted cloud occurrence (i.e., missed clouds and false clouds), (b) underestimated cloud condensate mass, and (c) simplified parameterization of solar irradiance extinction. The incorrect cloud occurrence is the leading bias source, because it occurred most frequently and results in a substantial magnitude of errors. The cloud occurrence bias is subject to simulations of large-scale air ascends and planetary boundary layer turbulence. Even when cloud occurrence is correctly simulated, our WRF simulation still underestimates CRE. This is because (a) the shallow convection scheme and cloud microphysics scheme underestimate cloud condensate mass and (b) cloud water path that feeds in the radiation scheme neglects precipitating cloud condensates (i.e., raindrops and graupels). Furthermore, an evaluation of cases with small bias in cloud condensate mass and effective radius demonstrates the parameterization of solar irradiance extinction for clouds induces a mean root mean square deviation of 110 W/m2. A possible reason is the simplified calculation of cloud extinction efficiency by applying Monte Carlo integration. The gained knowledge is important for understanding CRE simulation and solar irradiance forecast.