LAUSR

Abstract Ocean-colour radiometry is recognised as an Essential Climate Variable (ECV) according to the Global Climate Observing System (GCOS), because of its capability to observe significant properties of the marine ecosystem at synoptic to global scales. Yet the value of ocean colour for climate-change studies depends to a large extent not only on the decidedly important quality of the data per se, but also on the qualities of the algorithms used to convert the multi-spectral radiance values detected by the ocean-colour satellite into relevant ecological, bio-optical and biogeochemical variables or properties of the ocean. The algorithms selected from the pool of available algorithms have to be fit for purpose: detection of marine ecosystem responses to climate change. Marine ecosystems might respond in a variety of ways to changing climate, including perturbations to regional distributions in the quantity and in the type of phytoplankton present, their locations and in their seasonal dynamics. The ideal algorithms would be capable of distinguishing between abundance and type, and would not mistake one for the other. They would be robust to changes in climate, and would not rely on assumptions that might be valid only under current climatic conditions. Based on such considerations, we identify a series of ideal qualitative traits that algorithms for climate-change studies would possess. Necessarily, such traits would have to complement the quantitative requirements for precision, accuracy and stability in the data over long time scales. We examine the extent to which available algorithms meet the criteria, according to the work carried out in the Ocean Colour Climate Change Initiative, and where improvements are still needed.