AUTHOR=Phillips Lachlan R. , Carroll Gemma , Jonsen Ian , Harcourt Robert , Roughan Moninya TITLE=A Water Mass Classification Approach to Tracking Variability in the East Australian Current JOURNAL=Frontiers in Marine Science VOLUME=7 YEAR=2020 URL=https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2020.00365 DOI=10.3389/fmars.2020.00365 ISSN=2296-7745 ABSTRACT=

The East Australian Current (EAC) is a southward flowing western boundary current that transports relatively warm and nutrient-depleted subtropical water along Australia's east coast. The EAC is a highly variable system that is formed by temporally-varying mixtures of water in the Coral Sea that do not form a linear density gradient or conform to a set range of temperature and salinity values. It can therefore be difficult to track EAC dynamics across both space and time using traditional analytical approaches. In order to more accurately quantify variability and trends in penetration of the EAC we develop a novel machine-learning classification approach to quantify variability in coastal EAC dynamics along a latitudinal gradient within the EAC extension zone in southeastern Australia. Applying our method to data from a 22-year free running regional hydrodynamic model revealed significant decadal-scale changes to EAC dynamics in the region. The annual period (generally in the austral summer) when the EAC is the dominant water mass in the region increased by approximately 2 months over the model time series. The encroachment of the EAC's traditional period of summer dominance into winter may have significant ecological implications through the acceleration of poleward range extensions by vagrant tropical species, facilitation of community phase shifts from temperate to tropical assemblages, and a phenological shift in the timing of major phytoplankton blooms. These results highlight the need to further understand the rapid changes occurring within western boundary current systems, and illustrates how classification approaches may assist in uncovering patterns in these highly variable systems.