Enhancing Convective Precipitation Forecasting through Cloud-Aware Bias Correction and Hybrid IR–MW All-Sky Data Assimilation

Abstract

This dissertation addresses key limitations in all-sky infrared (IR) radiance data assimilation and multi-sensor integration for improving the prediction of high-impact precipitation events. In conventional all-sky data assimilation (DA) systems, cloud-induced biases in IR radiances, the limited vertical sensitivity of IR observations, and the dominance of IR-driven analysis increments often lead to substantial model–observation mismatches and suppress the effective use of complementary microwave (MW) information. To overcome these challenges, this study develops and evaluates a set of physically grounded strategies targeting both bias correction (BC) and multi-sensor integration. First, a cloud-aware all-sky IR radiance bias correction framework is proposed based on cloud-top temperature (CTT) and case-based cloud classification. The nonlinear, phase-aware correction explicitly accounts for cloud-induced radiance biases and reduces imbalances in all-sky IR assimilation. By improving consistency between observed and simulated radiances, the proposed method enhances the assimilation of localized deep-convective signals and leads to improved forecasts of heavy rainfall events over the Korean Peninsula. Second, this dissertation demonstrates the value of integrating temporally asynchronous IR and MW radiance observations. While MW radiances provide cloud-penetrating information that is unavailable from IR measurements alone, their influence is often diminished in conventional synchronized assimilation systems due to the overwhelming temporal density of IR observations. An asynchronous IR–MW integration framework is introduced to amplify the contribution of MW radiances within joint assimilation, revealing strong synergistic effects between the two sensor types. The complementary sensitivities of IR and MW radiances result in more realistic representations of cloud structure and convective evolution than can be achieved using either sensor independently. Finally, a unified all-sky assimilation framework is developed by combining cloud-aware IR BC with asynchronous IR–MW integration. Whereas IR assimilation primarily provides localized corrections near cloud tops, enhanced MW influence enables broader thermodynamic and dynamical adjustments throughout the troposphere. The unified framework effectively leverages the strengths of both observation types, yielding additional improvements in precipitation forecasts and producing a more balanced and physically consistent representation of deep convective systems. Overall, this research demonstrates that advancing all-sky radiance assimilation requires not only improved cloud-aware BC but also optimized multi-sensor integration strategies. The proposed approaches contribute to enhancing predictive skill for convective rainfall events and provide a practical foundation for next-generation all-sky data assimilation systems. Key words: all-sky radiance data assimilation, bias correction, IR-MW integration, heavy rainfall