A posteriori clear-sky identification methods in solar irradiance time series: Review and preliminary validation using sky imagers

This study examines all known methods that have been proposed in the literature to identify clear-sky periods in historical solar irradiance time series. Two different types of clear-sky detection (CSD) methods are discussed: those (16 total) that attempt to isolate periods of 1-min or more cloudless conditions, and those (5 total) that only attempt to detect clear-sun periods. All methods are found to rely on a diversity of inputs and on a variety of tests that typically examine the smoothness of the temporal variation of global and/or direct irradiance. Using samples of a few days with variable cloudiness, it is shown that these methods all have obvious strengths and weaknesses. Although this justifies a detailed validation to determine which method(s) could be best suited in the practice of solar radiation modeling or other applications, the current lack of appropriate equipment at high-quality reference radiometric stations prevents such an endeavor. Only a preliminary study is conducted here at seven stations of the SURFRAD network in the U.S., where 1-min irradiance measurements are available, along with sky data from a Total Sky Imager (TSI). The many limitations of the latter prevent its data to be considered "ground truth" here. Nevertheless, the comparison of the results from all CSD methods and 1.2 million TSI observations from all SURFRAD sites provides important qualitative and quantitative information, using a variety of performance indicators. Overall, two CSD methods appear more robust and are recommended, pending better high-resolution and high-performance cloud observations from modern sky cameras to redo these tests.