Assessing Uncertainties in Satellite Ocean Color Bio - Optical Properties

Uncertainties in retrievals of bio-optical properties from satellite ocean color imagery are related to a variety of factors, including errors associated with sensor calibration and degradation, atmospheric correction, and the bio-optical inversion algorithms. Here we examine the impact of uncertainties in the top-of-atmosphere (TOA) radiances and coefficients in the quasi-analytical algorithm (QAA) on the satellite estimation of the normalized water-leaving radiances (nLw) and downstream bio-optical properties, such as the absorption and backscattering coefficients and chlorophyll. We use a Moderate Resolution Imaging Spectroradiometer (MODIS) image, free from clouds, covering the Gulf of Mexico, as our data set. We apply ensembles at each pixel in the image to simulate noise on the satellite TOA radiance values (randomly-distributed in the range of ±2.0%) at MODIS visible wavelengths, as well as the two near-infrared (NIR) wavelengths used for atmospheric correction. For each set of ensemble runs, we assess the effect of uncertainty on nLw and bio-optical retrievals by analyzing the results spectrally for a coastal and open ocean region within our test image. In general, we found that as little as 2% uncertainty (noise) in the TOA radiances can significantly change „Lw retrievals, absorption and backscattering coefficients across the visible wavelengths, as well as derived chlorophyll values. We also use ensembles to analyze sensitivity of coefficients in the QAA algorithm. By varying coefficients in the algorithm, we can quantify the impact of uncertainties on the backscattering coefficient of suspended particles (bbP). Within the QAA algorithm, we vary coefficients used in calculating remote sensing reflectance (Äre) and the exponent used in calculating