Estimating Water Reflectance from EPIC/DSCOVR Data

A spectral-matching algorithm is developed to estimate spectral water reflectance from hourly EPIC observations at 340, 388, 443, 551, 680, and 780 nm. The perturbing signal from the atmosphere and surface, after correction for gaseous absorption and Rayleigh scattering effects, is represented by its major principal components and associated coefficients. The eigenvectors are determined from a simulated data ensemble that includes expected geometric and geophysical situations. Water reflectance is modeled as a function of chlorophyll-a concentration and a parameter defining the contribution of suspended particles to the backscattering coefficient. The parameters of the perturbing and water signals are retrieved in a 2-step process by minimizing the difference between observed and modeled top-of-atmosphere reflectance. Performance is assessed for typical conditions. System vicarious calibration is performed at the MOBY site, but using MODIS-Aqua water reflectance estimates as reference. In practice, water reflectance, normalized to zero Sun and view zenith angles, is obtained on source grid for each EPIC observation during the day, then remapped to 18.4 km resolution (Equator) Plate Carrée grid, and averaged. EPIC-derived water reflectance is in agreement with MODIS-derived reflectance. Spatial coverage, however, is much improved on a daily time scale using EPIC data.