When evaluating which model of the dataset to choose you'll want to know which version is the most accurate for your location. There are overall global statistics to consider, or you can look at those results specific to your region of interest. Our validation results are always calculated using independent ground stations that have been quality controlled and have the best global coverage available.

Many people look at the Mean Bias Error to understand how closely the satellite data is matching ground observations in the same region or project area. It is a good way to get a quick idea of whether the satellite models may be over or underestimating actual available resource for a particular area.

When we discuss the accuracy of a particular dataset most people are referring to the "uncertainty" or the Standard Deviation of the bias seen over several stations in a particular region. The smaller the uncertainty the more likely that model is representative of conditions in that region and is not biased in a particular direction.

As part of our processing, we regularly update our global validation studies. The version that corresponds to the data currently available in the Solar Time Series Tools is available here.

Of particular importance in validation of satellite derived solar data sets is the quality of the ground station data used. If uncertainty is considered as the difference between the satellite estimate and the ground station – then the uncertainty of the ground station is part of that equation. We have seen that the uncertainty of some models (2.1 for example) is cut in half in the U.S.A. region when considering many very high quality private (customer) stations as opposed to the public stations in the global validation study. Our “professional grade” validation paper is available here.

Also see here for our KMZ validation files that allow for a more detailed exploration of error at specific locations using Google Earth.