Assessing cloud representation in two microphysics schemes of MPAS–Atmosphere model (version 8.2.2) over the Maritime Continent with Himawari-9 AHI observations

Tropical cloud prediction over the Maritime Continent remains challenging for numerical weather prediction models, largely due to uncertainties in cloud microphysics parameterizations under complex topography and multiscale convective conditions. Accurate cloud forecasts are essential for assimilating cloud-affected observations, particularly high-resolution radiances from geostationary satellites, which also provide a valuable basis for diagnosing model performance of clouds. In this study, tropical cloud characteristics simulated by the Model for Prediction Across Scales–Atmosphere (MPAS–A) with two cloud microphysics schemes – the single-moment WSM6 and the double-moment NSSL, are comprehensively evaluated using multi-band radiances from the Himawari-9 Advanced Himawari Imager (AHI) and associated Level-2 cloud products. The proposed infrared radiance–based framework enables more effective use of high-resolution satellite data for model assessment. Results indicate that NSSL better captures the spatial distribution of cloud amount and exhibits smaller warm forecast-minus-observation ( F − O ) brightness-temperature (BT) biases in cloud-sensitive window channels. In contrast, WSM6 shows smaller biases in water-vapor-sensitive channels but shows less skill in distinguishing cloudy and clear-sky conditions during the spin-up period. Cloud-type diagnostics further reveal opposite characteristics in the upper troposphere, with WSM6 underpredicting and NSSL overpredicting the occurrence of cold ice clouds. Nevertheless, both schemes show higher forecast skills for liquid-water and thin ice clouds than for other cloud types, whereas the largest forecast errors occur in multilayer ice-cloud regimes. An overprediction of cloud-top heights for high-level clouds is also evident in both schemes, with diurnal variations in F − O being more pronounced over land in NSSL.