Abstract
The Rayleigh scattering approach to absolute radiometric calibration of satellite optical sensors is investigated for SNPP VIIRS. The vicarious calibration coefficients are obtained by comparing measured and predicted top-ofatmosphere (TOA) reflectance sampled under “favorable” scene conditions, i.e., oceanic areas with stable water optical properties, where situations of clear sky, low wind speed, and low aerosol content are frequent. In such situations, and under suitable geometry, the dominant contribution to the satellite signal in the visible is due to molecular (Rayleigh) scattering, which can be computed accurately. A sensitivity study performed for suitable calibration sites indicates that, using thresholds of 7 m/s for wind speed, 0.1 for aerosol optical thickness, 0.0005 for Sun glint reflectance, and 60° for Sun and view zenith angles, the error on the TOA reflectance is 2.3% at 410.5 nm (M1), increasing to 4.5% at 671.4 nm (M5). The error budget is dominated by the effect of uncertainty on aerosol optical thickness at the longer wavelengths and on marine reflectance at the shorter wavelengths. Two methods are developed to compute the VIIRS reflectance. The first method (Method 1) consists in using for the region of interest environmental conditions obtained from ancillary data or climatology. The second method (Method 2) addresses the uncertainty in aerosol model and optical thickness by accounting for the aerosol content of the scene. Application of Method 1 to VIIRS data yields calibration coefficients that are site and air mass dependent, which is not satisfactory. This results from using climatology data to specify aerosol parameters. Method 2, which utilizes the measured VIIRS reflectance in the near infrared to estimate the aerosol reflectance, reduces substantially the site and angular geometry dependence of the calibration coefficients. Accuracy can be improved by better specifying the marine reflectance, estimating the aerosol model, and further stratifying the calibration data. The large number of realizations allows for absolute calibration of individual detectors, studies of angular dependence, linearity, and polarization sensitivity, which is not possible using vicarious techniques based on (too few) in situ measurements.
