SUPPLEMENTING SOLAR-RADIATION NETWORK DATA BY CO-KRIGING WITH SATELLITE IMAGES

Observations made in 1982 from the Swiss meteorological network are used to assess the precision of geostatistically predicted daily global horizontal insolation. Results of such predictions are compared with insolation estimates obtained from the imagery of the meteorological satellite METEOSAT. Application of Kriging to the network data yields cross-validation results that are very close to the validation results of the satellite estimates, namely unbiased prediction with about 3 MJ m-2 root mean square error (RMSE) on a daily total measured mean (MM) of 18.9 MJ m-2. Co-Kriging applied to both data sets helps to reduce the RMSE to about 2.5 MJ m-2. As the application occurs under simplifying assumptions to allow use of a PC, and that there is, therefore, much room for improvements, we conclude that co-Kriging appears promising for detailed global insolation mapping.