Combination of satellite and ground measurements of hourly surface solar radiation data in Germany

Abstract

<p>As part of a restructuring of the solar radiation measurement network, the German Meteorological Service (DWD) is pursuing the goal of expanding its high-quality surface measurements using pyranometers (to 42 stations) while discontinuing measurements with other instruments. At the same time, surface solar radiation products from satellite data are progressively improving in quality and can be used to compensate for the reduction of ground measurements and to increase the spatial coverage of solar radiation information over Germany. For this purpose, the project DUETT aims at a merging of solar radiation data from the 42 pyranometer stations and near real-time data based on measurements from METEOSAT-SEVIRI (with a spatial resolution of about 5 km). As products, hourly data of the global horizontal irradiance (GHI) and the sunshine duration (SDU) will be provided on a 2 x 2 km grid for Germany with a time delay of 15 minutes after each full hour.</p><p>Merging is performed in three main steps: 1. Time aggregation to derive hourly data from instantaneous 15-minute measurements of the satellite; 2. Determination of deviations between hourly satellite and ground measurements; 3. Interpolation of the deviations to target grid of 2 km resolution using Ordinary Kriging. These three basic steps involve an 'optical flow' technique to generate intermediate images of the satellite data, a correction algorithm for the clear-sky satellite data to consider the impact of varying water vapour columns on surface radiation and a regression approach to reduce systematic deviations between satellite and station data. We present the latest version of the merging procedure for both radiation parameters GHI and SDU. The steps of the algorithms are illustrated, strategies for correcting systematic errors in the satellite data are shown and open issues are discussed. Furthermore, validation results based on independent validation data and based on cross validation will be presented.</p>