Estimating uncertainties in bio-optical products derived from satellite ocean color imagery using an ensemble approach

Abstract

We propose a methodology to quantify errors and produce uncertainty maps for satellite-derived ocean color bio-optical products using ensemble simulations. Ensemble techniques have been used by the environmental numerical modeling community to propagate initialization, forcing, and algorithm error sources through-out the full simulation process, but similar approaches have not yet been applied to satellite optical properties. Uncertainties in retrievals of bio-optical properties from satellite ocean color imagery are related to a variety of factors, including sensor calibration, atmospheric correction, and the bio-optical inversion algorithms. Errors propagate, amplify, and intertwine along the processing path, so it is important to understand how the errors cascade through each step of the analysis, to assess their impact and identify the main factors contributing to the uncertainties in the final products. Also, we are interested in producing short-term forecasts of the bio-optical property distributions, by coupling the satellite imagery with physical circulation models. So, in addition to the uncertainties in the satellite-derived bio-optical properties due to the above-mentioned factors, the uncertainties in the model currents used to advect the bio-optical properties add another layer of complexity to the problem. We outline these processes and present preliminary results for this approach.