Spatial representativeness of ground-based solar radiation measurements estimated from high-resolution Meteosat data

The validation of gridded surface solar radiation (SSR) data often relies on the comparison with ground-based in situ measurements. This poses the question on how representative a point measurement is for a larger-scale surrounding. We use the high-resolution (0.03° ) MVIRI data from the Satellite Application Facility on Climate Monitoring (CM SAF) to study the spatial sub-grid variability in all-sky surface solar radiation (SSR) over Europe, Africa, and parts of South America as covered by the Meteosat disk. This is done for the CERES EBAF 1° standard grid and a grid of 3° resolution. Furthermore, we quantify the spatial representativeness of surface sites from the BSRN (22) and the GEBA (1730) for their site-centered larger surroundings varying in size from 0.25° to 3°. These analyses are done on a climatological annual and seasonal mean basis over the period 2001–2005. The spatial variability of the CM SAF dataset itself agrees very well with surface measurements in Europe. Persistent gradients in cloud cover, often orographically induced, appear to be the major source of high subgrid variability. Hence, the subgrid variability over land masses is largest in mountainous and coastal regions and very low over deserts and rainforests. Seasonal variations are mainly due to the shifting location of the ITCZ. Aggregating the Meteosat data (0.03°) onto coarser grids of different resolutions between 0.06° to 0.5°, a linear decay in variability is evident. By contrast, the variability in a field with random spatial structure is inversely proportional to resolution.

The random representation error at the surface sites with respect to surroundings of 1° and 3° are on average 2% (4 Wm¯²) and 3% (5 Wm¯²). The majority of sites is considered representative with errors well below the uncertainty of in-situ measurements. Typically, the surface sites exhibit neither particularly high nor low representation errors as compared to extreme cases studied over Africa, i.e., mountainous Ethiopia and the Saharan desert. As shown over Europe, these representation errors are reduced when site-specific correction factors are applied or when multiple sites are available in the same grid cell, i.e., three more sites reduce the error by 50%.

Hakuba, M.Z., Folini, D., Sanchez-Lorenzo, A., and Wild, M. 2013: Spatial representativeness of ground-based solar radiation measurements, J. Geophys. Res., 118, 8585–8597, doi:10.1002/jgrd.50673.