Improving heavy precipitation forecasts through phase‐aware nonlinear bias correction of all‐sky infrared radiances using cloud top temperature differences

Abstract Infrared (IR) radiances are highly sensitive to cloud top properties, making all‐sky data assimilation (DA) challenging due to scene‐dependent biases arising from uncertainties in cloud representation and radiative transfer modeling. This study introduces a phase‐aware nonlinear bias correction (BC) framework for all‐sky IR radiances. The method employs the cloud top temperature difference (ΔCTT) between model and observations as a physically based predictor to account for cloud‐phase‐dependent biases, using 273.15 K as the phase threshold. All‐sky IR radiance DA experiments were performed for three heavy rainfall events over Korea using brightness temperature (BT) data from the Advanced Himawari Imager (AHI) onboard the Himawari‐8 geostationary satellite. The proposed CTT‐BC scheme was implemented within a convection‐permitting Weather Research and Forecasting Data Assimilation (WRFDA) three‐dimensional variational (3D‐Var) system and evaluated against the conventional variational BC (VARBC) and a cloud‐amount‐based BC (CA‐BC). Results show that CTT‐BC effectively reduces systematic biases and leads to more Gaussian observation‐minus‐background (OMB) distributions by mitigating extreme outliers in cloud‐affected scenes while preserving sharper OMB gradients. The CTT‐BC analyses produce stronger, better‐located hydrometeor signals and locally enhanced moisture gradients along convective bands, leading to improved precipitation forecasts that more accurately capture the spatial organization of heavy rainfall. Forecast verification indicates consistent reductions in root‐mean‐squared error (RMSE), particularly for humidity at the analysis time. These results demonstrate that phase‐aware nonlinear predictors can substantially enhance all‐sky IR radiance assimilation and improve short‐range forecasts of high‐impact convective systems.