Many of the issues discussed above become much more noticeable and complicated for validation of models based on satellite input data. The questions of temporal and spatial consistency are particularly vexing, as satellite data, while uniform, are usually sparse in time compared to surface observations. Spatial concerns are an even greater problem, since surface observations are ‘point’ observations, and satellite observations are spatially extended, even if at very high spatial resolution. Perez et al. (1997, 2001) provide a detailed review of these issues. In particular, as one observes the degradation in correlation between solar radiation measurements as ground site-separation increases, one sees the same sort of degradation in the accuracy of satellite model estimates as one moves away from a “ground truth” site. Magnitudes of this degradation start at about ±15% for sites within a few kilometers of each other to above 40-60% at distances of several hundred kilometers, for both ground stations and pixels removed from validation sites in satellite estimates (Perez et al. 2001).
As the number of satellite platforms and models evolve, there are also the issues of degradation and ‘recalibration’ of space-based sensors. These issues, similar to the calibration, degradation, and uncertainty of ground-based sensors discussed in Chap. 1, should be kept in mind when using models using satellite-based input data.
The NASA Surface and Meteorological and Solar Energy (SSE) website (http:// eosweb. larc. nasa. gov/sse) provides a great deal of helpful information on both accuracy and methodologies.
The European Community’s Helioclim project has links describing the Heliosat model used for the European Solar Radiation Atlas (http://www. helioclim. net/ heliosat/index. html), as well as links describing the calibration of MeteoSat instruments (http://www. helioclim. net/calibration/index. html), and solar radiation data quality control (http:// www. helioclim. net/quality/index. html).
Comparisons between results from models using either satellite or ground-based input data are also possible. An example provided here is between the METSTAT meteorological model (Maxwell 1998) for ground-based input data—but modified for use in the 1991-2005 update to the 1961-1990 United States National Solar Radiation data base (http:// rredc. nrel. gov/solar/old_data/nsrdb)—and the recently developed Perez satellite model (Wilcox et al. 2007). Figure 20.6 shows a comparison of the annual average direct beam estimates as gray-scaled background (satellite model estimates) and circles (station-based estimates). The observed differences (up to several kWhm~2 per day) are due mainly to issues with the quality of the input data for the meteorological model. Most particularly, the move from human observers to automated ceilometer measurements has severely compromised the cloud cover data needed for the modified METSTAT model inputs.