Cloud masking using MSG and HIRLAM

Clouds occurrence is the direct consequence of atmospheric processes relevant to the (scientific) community. Their presence is a significant indication about the atmospheric state. Knowledge about cloud patterns is therefore important for forecasters. Remote sensing of clouds from space has become an irreplaceable information source for meteorological and climatological applications. Lacking of the satellite information would lead to a significant degradation of the forecast skill of meteorologists.

Since January 2004 observations from a new type of satellite instrument the Spinning Enhanced Visible and Infra Red Imager (SEVIRI) became available. SEVIRI is located on a new geostationary platform, referred to as METEOSAT 8, operated by the intergovernmental organisation EUMETSAT, located at Darmstadt, Germany. The SEVIRI instrument offers observations in eleven spectral channels, with improved spatial ( in the infrared nadir 3x 3 km²) and temporal resolution (15 minute repetition cycle) , where the predecessor instrument, the Meteosat Visible and Infra Red Imager MVIRI offered three spectral channels, a spatial resolution at nadir in the infrared of 5x5km², and a repetition cycle of 30 minutes. To evaluate the wealth of information provided by SEVIRI the forecaster can no longer rely on visual inspection of all the spectral channels separately. New synergetic interpretation methods using all of the spectral channels are required to profit optimal from the possibilities SEVIRI offers.

This project aims to help forecasters, climatologists, and numerical weather prediction modellers by producing an automated interpretation of the satellite observation. First a cloud mask is produced, differentiating cloud contaminated pixels from cloud free pixels. The cloud mask is the first important part of automated interpretation. The cloud masking is then followed by cloud typing.

To achieve the goal an algorithm is developed. It uses the SEVIRI observations, and the surface temperature from the KNMI numerical weather prediction model (HIRLAM). The method is based on knowledge attained in a BCRS USP funded project, named MetClock (METeosat CLOud Characterisation KNMI). The output is a cloud mask. The results are evaluated by comparison to surface observations (called synops) done at meteorological stations.

During the period of the project approval and the actual project execution EUMETSAT released through the Satellite Application Facility on Now casting and short range forecasting (NWC-SAF) a software package. The software package also provides also a cloud mask. Within the project it was decided to evaluate this cloud mask simultaneous with the KNMI derived cloud mask. The chosen approach allows for a comparison between the NWC-SAF products and the KNMI algorithm. The NWC-SAF software package supplies, next to the cloud mask, a cloud type product and a cloud top height product