| |
|
|
|
IntroductionFinanced by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), the study "IASI/AVHRR Visual Scenes Analysis and Cloud Detection" has the aim of improving the IASI cloud detection by optimising the coefficients used for a predefined set of cloud tests performed in the IASI Level 2 processing.For the optimisation, a dataset comprising 25,000 "handpicked" samples is assembled by a visual analysis of AVHRR Level 1B scenes overlaid by collocated IASI footprints. Each sample classifies a single IASI instantaneous field of view (IFOV) as being cloud-free (clear sky), cloudy, clear partly cloudy and cloudy partly cloudy - partly cloudy samples with less than 20% cloudiness and with more than 20% cloudiness. 
A sample comprises the visual cloud classification together with all data relevant for the IASI cloud detection tests and the assessment of their performance after optimisation: the IASI sounder pixel, all AVHRR pixels covered by the footprint of the IFOV and auxiliary data not directly provided by the IASI or AVHRR products at IFOV position.
In order to enable an expert on visual scene analysis to efficiently collect the 25,000 samples a set of interactive tools has been developed. These tools as well as the IASI reader has been developed as plug-ins to the ESA BEAM 4 software, an open-source image processing, visualisation and analysis application developed by BC.
|