The US National Renewable Energy Laboratory (NREL) is responding to a growing demand for high-accuracy solar resource data with uncertainties significantly lower than those of existing solar resource datasets, such as the National Solar Radiation Database (NSRDB). Measurements for long-term solar resource characterizations require years to complete, which is an unacceptable timeline for the rapidly emerging needs of renewable energy applications. This contribution seeks methods of reducing the uncertainty of existing long-term solar resource datasets by incorporating lower-uncertainty site-specific ground measurements of a limited period of record. In particular, various techniques are being explored to make full use of the existing high-resolution radiation data available in the NSRDB and other sources, and extrapolate them over time using locally measured data and other supportive information. The interannual variability in global and direct radiation is studied here using long-term data at various sites. NSRDB's modeled data for the 1998-2005 period are compared to quality-controlled measurements to assess the performance of the model, which is found to vary greatly depending on climatic condition. The reported results are encouraging for applications involving concentrators at very sunny sites. Large seasonal biases are found at some cloudy sites. Various improvements are proposed to enhance the quality of the existing model and modeled data. The measurement of solar radiation to characterize the solar climate for renewable energy and other applications is a time consuming and expensive operation. Full climate characterization may require several decades of measurements-a prospect that is not practical for an industry intent on rapid deployment of solar technologies. This study demonstrates that the consistency of the solar resource in both time and space varies widely across the United States. The mapped results here illustrate regions with high and low variability and provide readers with quick visual information to help them decide where and how long measurements should be taken for a particular application. The underlying data that form these maps are also available from NREL to provide users the opportunity for more detailed analysis. (C) 2011 Published by Elsevier Ltd.