Neural network estimation of atmospheric profiles usingAIRS/IASI/AMSU data in the presence of clouds

摘要

As the forthcoming launch of the NPOESS Preparatory Project (NPP) nears, pre-launch predictions of onorbit performance are of critical importance to illuminate possible emphasis areas for the intensive calibration/ validation (cal/val) period to follow launch. During this period of intensive cal/val (ICV), quick-look performance assessment tools that can analyze global data over a variety of observing conditions will also play an important role in verifying and potentially improving environmental data record (EDR) quality. In this paper, we present recent work on a fast and accurate sounding algorithm based on neural networks for use with the Cross-track Infrared Sounder (CrIS) and the Advanced Technology Microwave Sounder (ATMS) to be flown on the NPP satellite. The algorithm is being used to assess pre-launch sounding performance using proxy data (where observations from current satellite sensors are transformed spectrally and spatially to resemble CrIS and ATMS) from the Atmospheric InfraRed Sounder (AIRS) and the Advanced Microwave Sounding Unit (AMSU) on the NASA Aqua satellite and the Infrared Atmospheric Sounding Interferometer (IASI) and AMSU/MHS (Microwave Humidity Sounder) on the EUMETSAT MetOp-A satellite. The algorithm is also being developed to provide a highly-accurate quick-look capability during the NPP ICV period. The present work focuses on the cloud impact on the infrared (AIRS/IASI/CrIS) radiances and explores the use of stochastic cloud clearing (SCC) mechanisms together with neural network (NN) estimation. A stand-alone statistical algorithm will be presented that operates directly on cloud-impacted AIRS/AMSU, IASI/AMSU, and CrIS/ATMS (collectively CrIMSS) data, with no need for a physical cloud clearing process. The algorithm is implemented in three stages. First, the infrared radiance perturbations due to clouds are estimated and corrected by combined processing of the infrared and microwave data using the SCC approach. The cloud clearing of the infrared radiances was performed using principal components analysis of infrared brightness temperature contrasts in adjacent fields of view and microwave-derived estimates of the infrared clear-column radiances to estimate and correct the radiance contamination introduced by clouds. Second, a Projected Principal Components (PPC) transform is used to reduce the dimensionality of and optimally extract geophysical profile information from the cloud-cleared infrared radiance data. Third, an articial feedforward neural network (NN) is used to estimate the desired geophysical parameters from the projected principal components. The performance of the method was evaluated using global (ascending and descending) EOS-Aqua and MetOp-A orbits co-located with ECMWF forecasts (generated every three hours on a 0.5-degree lat/lon grid) for a variety of days throughout 2003, 2004, 2005, and 2007. Over 1,000,000 fields of regard (3 x 3/2 x 2 arrays of footprints) over ocean and land were used in the study. The performance of the SCC/NN algorithm exceeded that of the AIRS Level 2 (Version 5) algorithm throughout most of the troposphere while achieving approximately 25-50 percent greater yield. Furthermore, the SCC/NN performance in the lowest 1 km of the atmosphere greatly exceeds that of the AIRS Level 2 algorithm as the level of cloudiness increases. The SCC/NN algorithm requires signicantly less computation than traditional variational retrieval methods while achieving comparable performance, thus the algorithm is particularly suitable for quick-look retrieval generation for post-launch CrIMSS performance validation.