Analysis of uncertainties in SEVIRI cloud property retrievals for climate monitoring

METEOSAT-8 is the first geostationary satellite with an imager (Spinning Enhanced Visible and Infrared Radiometer Instrument (SEVIRI)) designed to provide accurate spatial distributions of cloud properties with a 15 minutes temporal resolution. Within the Climate Monitoring Satellite Application Facility (CM-SAF) KNMI has developed a cloud properties retrieval algorithm, using narrowband visible and near-infrared radiances from either NOAA/AVHRR or METEOSAT-8/SEVERI. This algorithm provides Cloud Optical Thickness (COT), droplet effective radius and Cloud Liquid Water Path (CLWP) over Europe. In order to use these cloud properties for climatological and process studies it is important to access their accuracy and how these values vary daily and annually. Although there has been progress in quantifying the accuracy of cloud property retrievals from MODIS and NOAA/AVHRR, little research has been done on the validation of these retrievals from METEOSAT-8.

This study examines if SEVIRI retrieved cloud properties have sufficient accuracy for climate research and process studies over Europe. We analyse a one-year dataset of COT, CLWP and droplet effective radius retrievals at a 15 minutes sampling frequency. The SEVIRI cloud property retrievals are compared to microwave radiometer observed LWP and pyranometer derived COT for the three CloudNET stations. The CloudNET research project, supported by the European Commission, provides quasi continuously measurements of ground-based cloud properties for the development and implementation of cloud remote sensing algorithms. The accuracy of the ground-based measurements is superior to current satellite remote sensing techniques, which makes these observations the appropriate data source for assessing the accuracy of SEVIRI retrievals. The accuracy of the cloud property retrievals is evaluated for daily and annual variations in illumination geometry. Radiative transfer simulations are used to quantify and understand these differences. Uncertainties caused by the different footprints of the ground based and satellite retrieved cloud properties are investigated. Based on a large set of MODIS cloud liquid water path fields the magnitude of these uncertainties is quantified and an optimal sampling strategy is proposed.

The results from both the validation with CloudNET observations and the sensitivity study indicate that cloud property retrievals over northern Europe at low solar zenith angles from METEOSAT-8/SEVIRI have a higher sensitivity to errors than from NOAA/AVHRR. This indicates that at high latitudes present geostationary satellites, such as METEOSAT-8, are less suited than polar satellites for the analysis of seasonal variations in cloud properties. It will be shown that a portion of the observed differences between ground and space based observations is explained by the different footprints of the ground and space based instruments. These biases can be reduced by selecting the most representative pixels for validation, taking into account wind speed and direction and the altitude of the clouds. A method for optimizing the collocation of ground and satellite observations is proposed.