Using Sub-Millimeter Channels for Cloud Correction around 183 GHz

Abstract:

With increasing focus towards using sub-millimeter wavelengths for meteorological applications, future satellite missions like EPS-SG (EUMETSAT Polar System-Second Generation) ICI (Ice Cloud Imager) will include sensors at higher frequencies. In this study, we examine the potential of all sub-millimeter channels available on ICI to provide a cloud filtering/correction for data measured around 183 GHz. We also compare their usefulness against what can be achieved with 229 GHz for cloud filtering. The channel 229 GHz will be available on MWS (MicroWave Sounder) for similar purposes. We use a machine learning based approach for cloud correction and a QRNN (Quantile Regression Neural Network) based correction algorithm is developed. The algorithm uses combinations of data from 183 GHz and sub-mm channels to predict the clear sky brightness temperatures for 183 GHz channels. We also look in the special case of using just 325 GHz channel. Both 183 and 325 GHz channels have similar characteristics, but 325 GHz has a higher sensitivity to the cloud impact. Thus with a combination of two, even the data with cloud impact of the order of noise and modeling uncertainties can be accounted for. It is shown that the correction algorithm using 325 GHz channel is successful in making clear-sky predictions with a high accuracy and only minimal data is rejected. Another benefit of using QRNN is the estimation of case-specific uncertainties which makes it ideal for use in assimilation systems.