Poster

IAMAS

M15 - Advances in the Remote Sensing of Aerosols, Clouds, Precipitation and Radiation

Aerosol Vertical Profile Estimation over China Using Orbiting Carbon Observatory-2 O2 A-Band Data and the Random Forest Model

Aerosol research plays a critical role in enhancing air quality, protecting ecosystems, and mitigating climate change. This study presents a random forest (RF) model to estimate aerosol optical depth (AOD) and vertical profiles of aerosol extinction coefficients, utilizing the Oxygen-A (O2 A-band) observations from the Orbiting Carbon Observatory-2 (OCO-2) over China and its neighboring regions. The model incorporates geographical data (latitude, longitude, elevation) and viewing angle information, with principal component analysis (PCA) employed to reduce the dimensionality of the high-volume OCO-2 O2 A-band data. The model was applied to predict aerosol extinction coefficients, and its performance was validated by comparing the predictions to the Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO) Level 2 extinction data. The AOD model achieved a correlation coefficient (R) of 0.676. Predictions for aerosol extinction coefficients demonstrated a reasonable agreement with the actual values, yielding a R of 0.535 and a root mean square error (RMSE) of 0.107 1/km. Seasonal analysis revealed optimal performance in autumn (R = 0.557), with lower performance in summer (R = 0.442). The model's accuracy showed a clear height dependency, with R and RMSE decreasing as altitude increasing. Additionally, the inversion accuracy improved when AOD was below 0.3, with an RMSE of less than 0.06 1/km.