All‐Sky Assimilation of FY‐4B AGRI Reflectance in the CMA‐MESO Model With ARMS as a Forward Operator

Abstract All‐sky assimilation of visible and near‐infrared observations has attracted widespread attention. In the CMA‐MESO model, this study proposes an optimized reflectance scheme based on the Ross‐Li bidirectional reflectance distribution function (BRDF) model. Over land, the Ross‐Li the BRDF model is coupled with the ARMS model using a method based on Fourier transform and Gaussian quadrature decomposition. Over ocean, a static reflectance lookup table is employed. For cloudy sky simulation, an effective cloud particle radius parameterization scheme consistent with cloud microphysical assumptions is introduced. For FY‐4B AGRI data at visible and near‐infrared bands, all‐sky observation errors are analyzed for a dynamic update in the 3DVAR data assimilation system. Five assimilation experiments are conducted, including three single‐channel experiments assimilating FY‐4B AGRI channels 2, 4, and 5 individually, and two combined‐channel experiments assimilating channels 2 + 4 and 4 + 5. Among the single‐channel cases, channel 4 assimilation shows the greatest improvement in the representation of ice cloud coverage. In contrast, assimilating channel 2 or channel 5 alone led to confusion between liquid and ice cloud increments. The combined‐channel assimilation experiments outperformed the single‐channel ones, yielding a significant reduction in the standard deviation (STD) of the simulated reflectance. In summary, assimilating FY‐4B AGRI observations holds a great potential for improving the accuracy of cloud cover representation.