## Uncertainty and interpretation of aerosol remote sensing due to vertical...

*J. Quant. Spectrosc. Radiat. Transfer, 114*, 91-100, doi:10.1016/j.jqsrt.2012.08.006.

We have built an aerosol retrieval algorithm which combines the Look Up Table (LUT) and least squares fitting methods. The algorithm is based on the multi-angle multiwavelength polarized reflectance at the Top Of the Atmosphere (TOA) measured by the Research Scanning Polarimeter (RSP). The aerosol state parameters are the aerosol particle effective radius, effective variance, complex index of refraction, and aerosol column number density. Monomodal aerosol size distribution is assumed. The Cost Function (CF) of the least squares fitting is designed in consideration of the RSP instrumental characteristics. The aerosol retrieval algorithm inherently assumes one layer of aerosols within the atmosphere. Synthetic polarized radiance data at the TOA have been created assuming either one or two layers of aerosols using the vector radiative transfer code based on successive order of scattering method. Test cases for one-layer aerosol systems show great performance. Around 90% of the total 1200 test cases have CF values smaller than 50. For these cases, the correlation coefficients of the input and retrieved parameters are generally around or larger than 0.98. The effective variance is slightly worse with the correlation coefficient of 0.76938. On the other hand, test cases for two-layer aerosol systems show that only 50% of the total (also 1200) tested cases have final CFs smaller than 50. Among these successful cases (CF r 50), the retrieved optical depth can still be interpreted as the total column optical depth, though the correlation coefficient is decreased in comparison with the one-layer aerosol cases. We propose to interpret other retrieved aerosol parameters as the average of corresponding parameters for each layer weighted by its optical depth at 865 nm. The retrieved effective radius and complex refractive index can be explained by this scheme (correlation coefficient around 0.9). The effective variance, however, shows decreased performance with the correlation coefficient of 0.46421. This may be due to the strong nonlinearity dependence of the scattering properties on the effective variance.

n