A fast radiative transfer model based on the oxygen A-band has been successfully developed. Using the white-sky albedo spectral conversion method and temperature offset parameter optimization, this model enables the unified retrieval of cloud geometric thickness (CGT) over both land and ocean surfaces. It significantly improves computational efficiency while maintaining high accuracy.

A machine learning-driven algorithm for predicting multiple cloud vertical structure parameters has been developed. This algorithm relies solely on oxygen A-band observations to simultaneously retrieve cloud optical depth (COD), cloud top pressure (CTP), and cloud pressure thickness (CPT), eliminating the need for active remote sensing data and significantly enhancing retrieval accuracy.

The traditional limitations of cloud retrieval algorithms have been overcome. The newly developed DIRECT algorithm is capable of handling both land and ocean single-layer liquid clouds and is compatible with various cloud types (such as stratocumulus and nimbostratus) and single- and multi-layer cloud scenarios, achieving precise retrievals on a global scale across different underlying surfaces and cloud types.

A method for retrieving cloud vertical structure by combining the oxygen A-band and infrared channels has been completed. This method has significantly reduced the RMSE for cloud top height (CTH) and cloud geometric thickness (CGT), providing high-quality data support for climate model optimization, cloud-radiation effect studies, and weather forecasting.