Validating physical and semi-empirical satellite-based irradiance retrievals using high- and low-accuracy radiometric observations in a monsoon-influenced continental climate

Are high-accuracy radiometric observations strictly indispensable for the validation of satellite-based irradiance retrievals, or might low-accuracy observations serve as adequate substitutes? Owing to the scarcity of sites with redundant radiometers, such inquiries have seldom been contemplated, much less subjected to systematic examination; rather, it has been customary to employ all accessible observations during validation, frequently with only minimal quality control. In this investigation, we address this question by validating two distinct sets of satellite-retrieved irradiance - one derived through physical methods, the other through statistical means - against collocated high- and low-accuracy observations. Departing from the majority of validation studies, which rely exclusively upon an array of performance measures, we advocate and implement a rigorous distribution-oriented validation framework, yielding more profound insights and more comprehensive conclusions. Beyond the validation methodology itself, the dataset utilized in this study is noteworthy in its own regard: It incorporates radiometric observations from the newly established and first-ever Baseline Surface Radiation Network (BSRN) station situated within a monsoon-influenced continental climate (specifically, the Dwa Köppen classification), in conjunction with irradiance retrievals from the Fengyun-4B geostationary satellite, which are likewise new to the community. The accumulated evidence strongly suggests that the use of low-accuracy observations as a reference in validating irradiance retrievals may entail significant risks, because the discrepancies they introduce can be of a magnitude comparable to the commonly accepted margins of error or improvement (approximately several W m-2 or a few percent) upon which numerous scientific assertions depend.