A practical demonstration on AMSU retrieval precision for upper tropospheric humidity by a non-linear multi-channel regression method

[摘要] A neural network algorithm inverting selected channels from theAdvance Microwave Sounding Unit instruments AMSU-A and AMSU-B wasapplied to retrieve layer averaged relative humidity. The neuralnetwork was trained with a global synthetic dataset representingclear-sky conditions. A precision of around 6% was obtained whenretrieving global simulated radiances, the precision deterioratedless than 1% when real mid-latitude AMSU radiances were invertedand compared with co-located data from a radiosonde station. The 6%precision outperforms by 1% the reported precision estimatefrom a linear single-channel regression between radiance andweighting function averaged relative humidity, the more traditionalapproach to exploit AMSU data. Added advantages are not only abetter precision; the AMSU-B humidity information is more optimallyexploited by including temperature information from AMSU-A channels;and the layer averaged humidity is a more physical quantity than theweighted humidity, for comparison with other datasets.The trainingdataset proved adequate for inverting real radiances from amid-latitude site, but it is limited by not considering the impactof clouds or surface emissivity changes, and further work is neededin this direction for further validation of the precision estimates.

[学科分类] 大气科学