Taozhong Huang - February 16

Weather and Climate Systems

A Correlation-based Inversion Method for Aerosol Property (CIMAP) Retrieval: A Case Study Using AERONET and POLDER Measurements

Taozhong Huang

Wednesday, February 16

03:00 PM

Online

Aerosol remote sensing often involves a large volume of spatial and temporal data to be processed. Both retrieval efficiency and accuracy are essential. To this end, we developed a correlation-based inversion method for aerosol property (CIMAP) inversion. CIMAP utilizes that the aerosols over a targeted area are typically contributed by a few prevailing types, and their spatial and temporal variations are correlated and can be captured by a few principal components (PCs). By retrieving the PC quantities, the retrieval efficiency and stability are improved. For the preliminary test, instruments with ground-based measurements and satellite-borne measurements are chosen to test the robustness of the algorithm. CIMAP is implemented to retrieve 60 sets of ground-based AERONET observations of aerosol optical depths (AODs) and sky radiances in three selected areas, including southern California, southern Africa, and north China around Beijing, where anthropogenic aerosols, carbon aerosols, air pollutants prevail, respectively. Capitalizing on the correlation in aerosol properties, CIMAP reduces the aerosol retrieval parameter space and saves retrieval time by more than 80% from adopting 7-8 PCs in the retrieval, compared to retrievals in regular parameter space. The retrieval results are compared to the AERONET measurement and operational retrieval products. The mean absolute differences are found to be 0.004, 0.019, 0.032, 0.039, 0.003, and 0.044mm for aerosol optical depth, single scattering albedo, Ångström exponent, real and imaginary parts of refractive index, and effective radii of whole size distribution, respectively. These differences are within or close to AERONET measurement or retrieval uncertainties. CIMAP is also implemented for the retrieval of POLDER measurements from the PARASOL satellite. With the increased parameter space, the neural network is applied to CIMAP for further improvement of computational efficiency. The preliminary results show that the retrieval results compared to the collocated AERONET measurements have a high level of agreement. Based on the improvement of retrieval efficiency and stability, CIMAP is under further application to retrieve past and future satellite observations of high spatial and temporal resolutions and broader spatial coverage.