Speaker
Description
The data record of the Aura Microwave Limb Sounder (MLS) currently covers more than 18 years, far exceeding its original 5-year design life. As part of the A-train constellation, the long MLS data record and the comparatively high spatio-temporal resolution of its retrieval products provide a unique opportunity to apply machine learning techniques to enhance its observational capabilities.
This talk presents four particular examples where artificial neural networks (ANNs) were trained on MLS observations to reliably predict a variety of different atmospheric properties: (i) cloudiness and cloud top pressure in the vicinity of individual MLS profiles, (ii) more accurate near-real-time profiles of atmospheric properties and constituents, (iii) methane and chlorine nitrate profiles at the MLS spatial and temporal resolution, and (iv) the bias in MLS 100-hPa geopotential height with respect to MERRA2.