The Eumetsat Multi Sensor Precipitation Estimate (mpe): Concept and Validation

Precipitation is the meteorological parameter affecting people in the most direct way. Forecasting the spatial and temporal distribution of rain and snow is therefore one of the major challenges for the meteorological services. Satellite remote sensing of precipitation could help to improve these forecasts. The most direct method to retrieve precipitation from spaceborne measurements is based on data from passive or active microwave sensors. These sensors are only available on low orbiting satellites up to now. In addition the spatial resolution of microwave sensors is hardly sufficient to resolve small scale precipitation events. Infrared brightness temperatures from geostationary satellites are only related indirectly to the precipitation at the ground but have a high spatial and temporal resolution. Therefore methods to combine both retrieval systems have been developed. EUMETSAT implemented such a ‘blending’ technique in the re-processing branch of its Meteorological Product Extraction Facility (MPEF). Rain rates derived from measurements of the Special Sensor Microwave / Imager (SSM/I) onboard of the US-DMSP satellites are combined with brightness temperatures from the Infrared channel of the MTP-METEOSAT satellites. The basic assumptions of the Multisensor Precipitation Estimate (MPE) method is that colder clouds are more likely to produce precipitation than warmer clouds. The relation between the cloud top temperature and the surface rain rate is non-linear and depends strongly on the current weather situation. Temporally and spatially coregistrated SSM/I and METEOSAT measurements are used to derive look-up tables (LUTs) which describe the rain rate as a function of the METEOSAT IR brightness temperature. These LUTs are applied to METEOSAT images in order to derive rain rates in the full spatial and temporal resolution. Co-registered data for the LUT derivation are accumulated in geographical windows for a specified time. The size of these temporal and spatial accumulation windows must be sufficient to have enough data sets for the LUT creation but should be not too large to represent the current weather situation. We selected temporal windows of 6-12 hours and spatial windows of 5° longitude and 5° latitude. Validation was performed with historical data sets comparing the MPE-results with rain rates derived only from SSM/I and with ground based radar data. The method performs as expected. Due to the basic assumption that cold clouds produce the most rain it is very efficient to estimate the spatial distribution and the strength of convective precipitation. This is valid for large scale tropical convection as well as for small scale convective processes and cold fronts. The precipitation at warm fronts and orographically induced precipitation is usually detected but miss-located by up to 100km.