Assimilating visible (VIS) and near-infrared (NIR) satellite observations into numerical weather prediction (NWP) systems demands radiative transfer (RT) models that combine high accuracy and structural flexibility. This study introduces an enhanced Optical Depth in Pressure Space (ODPS) transmittance model within the Advanced Radiative Transfer Modeling System (ARMS), focusing on the optimization of water vapor predictors to address critical limitations in VIS/NIR applications. The improved algorithm, referred to as $\mathrm{O}\mathrm{D}\mathrm{P}{\mathrm{S}}_{\mathrm{n}\mathrm{e}\mathrm{w}}$, demonstrates substantial accuracy gains across Fengyun (FY)-4B Advanced Geostationary Radiation Imager (AGRI) channels, reducing mean bias and root mean square error by up to 97% and 98%, respectively. By accounting for height-dependent zenith angles, $\mathrm{O}\mathrm{D}\mathrm{P}{\mathrm{S}}_{\mathrm{n}\mathrm{e}\mathrm{w}}$ effectively mitigates angular-dependent errors caused by Earth’s curvature, outperforming the default Optical Depth in Absorber Space (ODAS) algorithm at large viewing angles while maintaining comparable accuracy under nadir conditions. Relative to ODAS, $\mathrm{O}\mathrm{D}\mathrm{P}{\mathrm{S}}_{\mathrm{n}\mathrm{e}\mathrm{w}}$ lowers RMSEs by 63%–96% in line-by-line validations, achieving parity or better accuracy in five channels. Validations using UMBC (University of Maryland at Baltimore County) profiles and real atmosphere profiles confirm enhanced robustness across diverse atmospheric conditions and spatial coherence. Updated water vapor Jacobians exhibit ODAS-level smoothness in 5 of 6 channels and elimi-nate large layer-to-layer jumps, leaving only a minor residual oscillation near 1 hPa. The smoother Jacobians strengthen suitability for variational data-assimilation framework. Overall, $\mathrm{O}\mathrm{D}\mathrm{P}{\mathrm{S}}_{\mathrm{n}\mathrm{e}\mathrm{w}}$ offers a balanced solution that unifies regression accuracy, geometric adaptability, and physical interpretability, providing a promising pathway for assimilating next-generation satellite VIS/NIR radiance.

With the launch of the first civilian early-morning orbit satellite Fengyun-3E (FY-3E), higher demands are placed on the accuracy of radiative transfer simulations for hyperspectral infrared data. Therefore, several key issues are investigated in the paper. First, the accuracy of the fast atmospheric transmittance model implemented in the Advanced Research and Modeling System (ARMS) has been evaluated with both the line-by-line radiative transfer model (LBLRTM) and the actual satellite observations. The results indicate that the biases are generally less than 0.25 K when compared to the LBLRTM, while below 1.0 K for the majority of the channels when compared to the observations. However, during both comparisons, significant biases are observed in certain channels. The accuracy of Hyperspectral Infrared Atmospheric Sounder-II (HIRAS-II) onboard FY-3E is comparable to, and even superior to that of the Cross-track Infrared Sounder (CrIS) onboard NOAA-20. Furthermore, apodization is a crucial step in the processing of hyperspectral data in that the apodization function is utilized as the instrument channel spectral response function to produce the satellite channel-averaged transmittance. To further explore the difference between the apodized and unapodized simulations, Sinc function is adopted in the fast transmittance model. It is found that the use of Sinc function can make the simulations fit the original satellite observations better. When simulating with apodized observations, the use of Sinc function exhibits larger deviations compared to the Hamming function. Moreover, a correction module is applied to minimize the impact of Non-Local Thermodynamic Equilibrium (NLTE) in the shortwave infrared band. It is verified that the implementation of the NLTE correction model leads to a significant reduction in the bias between the simulation and observation for this band.