Training and Validation of the Fast PCRTM_Solar Model

Abstract:

Fast and accurate radiative transfer model is the key for satellite data assimilation for remote sensing application. The simulation of the satellite remote sensing radiances is very complicated since many physical processes, such as absorption, emission, and scattering, are involved due to the interactions between electromagnetic radiation and earth surface, water vapor, clouds, aerosols, and gas molecules in the sky. The principal component-based radiative transfer model (PCRTM) has been developed for various passive IR and MW instruments.

In this work, we extended PCRTM to including the contribution from solar radiation. The cloud/aerosol bidirectional reflectances have been carefully calculated using the well-known Discrete-Ordinate-Method Radiative Transfer (DISORT) model under over 10 millions of diverse conditions with varying cloud particle size, wavelength, satellite viewing direction, and solar angles. The obtained results were compressed significantly using principal component analysis and used in the mono domain radiance calculation.

We used 1352 different atmosphere profiles, each of them has different surface skin temperatures and surface pressures in our training. Different surface emissivity spectra were derived from ASTER database and emissivity models. Some artificially generated emissivity spectra were also used to account for diverse surface types of the earth. Concentrations of sixteen trace gases were varied systematically in the training and the remaining trace gas contributions were accounted for as a fixed gas. Training was done in both clear and cloudy skies conditions. Finally the nonlocal thermal equilibrium (NLTE) induced radiance change was included for daytime conditions.

We have updated the PCRTM model for instruments such as IASI, NASTI, CrIS, AIRS, and SHIS. The training results show that the PCRTM model can calculate thousands of channel radiances by computing only a few hundreds of mono radiances. This greatly increased the computation efficiency since we do not need to calculate the millions of mono radiances and do the convolution process.

The results from fast PCRTM_Solar simulation were compared to the instrument observed data. The simulated results were excellently agreed with the observations.