Monte Carlo-based uncertainty analysis of UV array spectroradiometers

Recent efforts (including effective ex-post stray light corrections) have allowed significant improvement in the performance (tested by intercomparison) of low-cost array spectroradiometers. However, the problem of uncertainty evaluation of UV measurements by these instruments has not been comprehensibly addressed mainly because of the difficulties involved in the uncertainty propagation. We have applied a Monte Carlo-based approach that allowed us to overcome those difficulties and to evaluate the uncertainty associated with spectral measurements carried out using a 12-bit CCD array of 2048 pixels, model SD2000, manufactured by Ocean Optics (R). In the UV-A part of the spectrum the standard uncertainty was found to be roughly constant (about 5%). However, mostly due to the noise affecting the readings, a sharp increment was observed at wavelengths shorter than 315 nm; indeed, the driving factor determining the UV-B uncertainty (and in turn the UV index uncertainty) was the signal-to-noise ratio (SNR). Low SNRs lead to standard uncertainties greater than 20% for irradiances at 300 nm and greater than 10% for UV index values. Although each instrument has an individual character, comparable figures are expected in the case of instruments whose measurements undergo similar quality assurance activities (i.e. similar stray light correction methods). Our uncertainty figures provide a frame to assess the significance of differences when intercomparing, and in turn to check the quality of the measurements. The applied uncertainty evaluation methodology is general and agrees with recommendations of the ISO Guide to the Expression of Uncertainty in Measurement.