Abstract
The Spinning Enhanced Visible and Infrared Imager (SEVIRI) instrument, onboard the Meteosat Second Generation (MSG) is a radiometer with 8 infrared (IR) spectral bands. IR retrievals of Layer Precipitable Water (LPW) and Lifted Index (LI) allow to identify potential severe weather when the system is still in a preconvective state. Statistical retrieval is computationally fast and it is a requirement for the SAFNWC PGEs. The study presented here, is part of an attempt to improve the algorithm developed in the SAFNWC framework to calculate Layer Precipitable Water and Stability Analysis Imagery (SAI) from SEVIRI radiances. The first codified algorithms (in the SAFNWC version 0.1 package) are a statistical retrieval where neural networks were trained with the available data (simulated radiances using numerical profiles from 60L-SD and RTTOV-7). These statistical retrievals have been evaluated against co-located products obtained from numerical weather analysis and radiosonde profiles, as well as MODIS products obtained in the areas scanned at the same time. The availability of real SEVIRI radiances allows us to compare real SEVIRI radiances with simulated radiances and to detect systematic bias among both datasets. In this study, first the retrieved LPW and LI will be evaluated, and the error sources will be identified. And later, the method for correcting the detected bias, between real and simulated radiances, will be analysed, and the improvements will be compared to calculated ("clear") values from the nearest (in space and time) ECMWF profiles and similar MODIS products.
