A statistical learning approach to Chang'E microwave radiometer data calibration

Abstract

We have proposed a statistical learning model which accounts for the effect due to the Chinese Chang'E (CE) cold space calibration antenna may have been contaminated by thermal microwave radiation from the lunar surface because its field of view overlapped with the lunar surface instead of aiming just at the cold space. This effect can corrupt the simple two-point calibration procedure and lead to the discrepancy between the available microwave radiometer Level 2C data from CE1 & 2. An algorithm to implement the statistical learning model is developed. Recalibration of the Level 2C radiometer data from localized locations (around the Apollo 17 lunar mission landing site on the lunar surface) by regression analysis based on our statistical learning calibration model successfully removes the discrepancy between the two sets of microwave brightness temperatures data from CE1 and 2. Residual analysis on the combined dataset shows that that the residuals are normally distributed with a mean value equals to zero. This sophisticate statistical learning algorithm developed here may be considered as a recalibration effort to remove the systematic inconsistency between the two microwave brightness temperature datasets. The combined and enlarged CE radiometer dataset may potentially increase our understanding of the physical reality hidden behind.