Comparison of cloud-filling algorithms for ocean color data
Andy Stock
Ocean color remote sensing provides timely, temporally and spatially exhaustive descriptions of chlorophyll concentrations in surface waters, allowing for the monitoring and study of phytoplankton biomass and primary productivity at broad spatial scales. However, cloud cover is a major challenge in marine satellite monitoring, causing ocean color sensors to see only small parts of the ocean surface in a given satellite image. Researchers have thus proposed various algorithms to fill data gaps “below the clouds”, but a comprehensive and quantitative comparison of the several algorithms’ performance has not yet been conducted. Based on analyses in three study areas, this poster thus aims to answer three related research questions. First, how accurately can the existing and various new, geostatistical and statistical-learning based cloud-filling algorithms predict chlorophyll-a concentrations “under the clouds”? Second, which of these algorithms have the best ratios of performance, computational cost and effort of implementation? Third, for researchers interested in developing new methods, what are the shared properties of the best-performing algorithms?