There is no single answer to this question. The datasets include different values due to the use of different clear sky algorithms and aerosol and water vapor inputs. The combination that works best in one climate region may not work best in another. This is why multiple datasets are provided.

Not sure how to choose a model? Follow the steps below and use the provided Google Earth files to view the validation results at public stations around the world.

1. Review the regional and local validation results in the downloadable kml files and identify which model performs best in your region of interest.

2. If you have a ground station within 50-100 km of the project location, compare the different data options for the concurrent period of time and evaluate which ones match the ground data the best.

3. If you have no ground data to serve as a reference, we don’t recommend using the dataset with the highest or the lowest values. It's better to choose a version that is in the middle of the results with good validation statistics in the region.

4. The final consideration would be the technology to be employed. If a tracking photovoltaic or concentrating solar plant are under consideration, all other statistics being equal, we would suggest one of the REST2 based models (2.0 or 2.1) because of the increase in accuracy of the DNI irradiance component.