Air Mass Factor Calculations Based on Support Vector Regression Machine

Nitrogen dioxide (NO₂) plays a key role in the chemistry of the atmosphere. Based on the space borne measurements, the Differential Optical Absorption Spectroscopy (DOAS) method determines the NO₂ slant column density (SCD) along the light path through the atmosphere in a given spectral window between 405 and 465 nm. Based on radiative transfer model, a Look-Up-Table (LUT) of Air Mass Factor (AMF) is derived. The look-up parameters of the LUT are the solar zenith angles (SZA) and the detector viewing zenith angles (VZA), azimuth angles (AA), the surface albedo, the aerosol loading, the cloud parameters and the profiles of pressure and temperature. Conventionally, the MF values are derived using interpolation from LUT, which leads to bias. The accuracy can be improved by the application of Support Vector Machine (SVM). The AMF of the LUT is used to be samples of SVM. By training the samples and predicting, the appropriate AMF values are derived. For comparison, the interpolation AMF values are also calculated. The intercomparison result shows the accuracy of SVM method is comparable to that of interpolation approach.