Prioritization of the vector factors controlling Emiliania huxleyi blooms in subarctic and arctic seas: A multidimensional statistical approach

Producing very extensive blooms in the world's oceans in both hemispheres, a coccolithophore E. huxleyi is capable of affecting both the marine ecology and carbon fluxes at the atmosphere-ocean interface. At the same time, it is subject to the impact of multiple co-acting environmental forcings, which determine the spatio-temporal dynamics of E. huxleyi blooming phenomenon. To reveal the individual importance of each forcing factor (FF) that is known to significantly control the extent and intensity of E. huxleyi blooms and can be retrieved from remote sensing data, we used long-term spatial time series (1998–2016) of sea surface temperature and salinity, incident photosynthetically active radiation, and Ekman layer depth relevant to the marine environments located in the North Atlantic, Arctic and North Pacific oceans, namely the North, Norwegian, Greenland, Labrador, Barents and Bering seas. The FFs retrieved were subjected to statistical analyses. The descriptive statistical approach has shown that E. huxleyi phytoplankton were highly adaptive to the environmental conditions and capable of arising and developing within wide FFs ranges, which proved to be expressly sea-specific. It was also found that there were FFs optimal ranges (also sea-specific), within which the blooms were particularly extensive. The application of the Random Forest Classifier (RFC) approach to each target sea allowed to reliably rank the FFs considered in terms of their role in the spatio-temporal dynamics of E. huxleyi blooms. With the only exception of the Bering Sea, allegedly due to temporally established untypical hydrological conditions, the prediction ability of RFC modeling characterized in terms of precision, recall, and f1-score generally was in excess of 70 %, thus indicating the adequacy of the developed models for FFs prioritization with regard to E. huxleyi blooms.