Neural Networks for Atmospheric Temperature Retrieval from AQUA AIRS/AMSU/HSB Measurements

The high spectral resolution sounding suite instruments on EOS Aqua are expected to provide estimated atmospheric profiles with higher accuracy and better vertical resolution. To validate and test it’s performance with it’s orbital measurements, temperature profiles at 101 pressure levels, up to 0.005 hPa, with 1km vertical resolution at troposphere, were retrieved on different types of terrain with different spectral bands in the middle latitude area by using a three-layered feed-forward neural networks with back-propagation algorithm. Results show that the Qinghai-Tibet Plateau has a measurable impact on the retrieval accuracy which is corresponding to the spectral bands used in performing retrievals. A promising approach to the elimination of this effect is to apply additional predictors which are non-satellite observed (e.g. surface altitude)