Evaluation of cloud-type impact on performance of the GL model version 1.2 for estimation of solar irradiance

Abstract

ABSTRACT This work evaluates the quality of the satellite-based GLobal radiation model version 1.2 (GL1.2) estimates for four cloud classes. The GL1.2 calculates the global solar irradiance at ground level using images from a single visible band channel (VIS) of the Geostationary Operational Environmental Satellite – GOES (GOES-East). The model´s performance was assessed by comparing hourly mean GL1.2 values with ground-based hourly measurements from the Brazilian National Institute of Meteorology (INMET, 354 automatic weather stations), for the entire year 2017. A satellite-based cloud classifier was adopted to discriminate the datasets according to prevailing cloud conditions (cumulus, stratus, cirrus and multilayer), but the clear sky behaviour was presented. The results were analysed for the five Brazilian regions. In the first analysis, we selected days with a predominance for five stations. It was found that the diurnal cycle was well reproduced. Then the regional investigation for the cloud types reveals that the best results are found for the Center West region over multilayer cloud (mean bias error, MBEannual = -2 ± 91 Wm−2 and root mean squared error, RMSE = 91 Wm−2), while the worst ones are in the North over cumulus fields (MBEannual = 101.5 ± 136.4 Wm−2). When considering all cloud types, the MBEannual is lower than 5 Wm−2 for the Northeast, Southeast and South regions, but it reaches 101 Wm−2 in the North. It is noteworthy that winter has the highest MBE in all classes analysed in the North, as well as cirrus situations in other regions. Although the inhomogeneities of cumulus and the semitransparent cirrus clouds tend to propagate errors to the model, the quality of GL1.2 data has a high degree of agreement with the observations. Improvements including an updated monthly minimum reflectance (R min) and water vapour column (H2Ovapour), and a better spatial resolution of the Advanced Baseline Imager (ABI) of GOES-16 (ABI/GOES-16) VIS imagery will allow refinement, especially for cumulus clouds.