FN Archimer Export Format
PT J
TI Water Mass and Biogeochemical Variability in the Kerguelen Sector of the Southern Ocean: A Machine Learning Approach for a Mixing Hotspot
BT 
AF Rosso, Isabella
   Mazloff, Matthew R.
   Talley, Lynne D.
   Purkey, Sarah G.
   Freeman, Natalie M.
   Maze, Guillaume
AS 1:1;2:1;3:1;4:1;5:1;6:2;
FF 1:;2:;3:;4:;5:;6:PDG-ODE-LOPS-OH;
C1 Scripps Institution of OceanographyUniversity of California San Diego La Jolla CA, USA
   Ifremer, University of Brest, CNRS, IRD, Laboratoire d'Océanographie Physique et Spatiale, IUEM Plouzané ,France
C2 UNIV CALIF SAN DIEGO, USA
   IFREMER, FRANCE
SI BREST
SE PDG-ODE-LOPS-OH
UM LOPS
IN WOS Ifremer UMR
   copubli-int-hors-europe
IF 3.405
TC 20
UR https://archimer.ifremer.fr/doc/00613/72471/71438.pdf
   https://archimer.ifremer.fr/doc/00613/72471/71439.pdf
LA English
DT Article
DE ;Southern Ocean;Kerguelen Plateau;Argo;unsupervised clustering;machine learning
AB The Southern Ocean (SO) is one of the most energetic regions in the world, where strong air‐sea fluxes, oceanic instabilities, and flow‐topography interactions yield complex dynamics. The Kerguelen Plateau (KP) region in the Indian sector of the SO is a hotspot for these energetic dynamics, which result in large spatio‐temporal variability of physical and biogeochemical (BGC) properties throughout the water column. Data from Argo floats (including biogeochemical) are used to investigate the spatial variability of intermediate and deep water physical and BGC properties. An unsupervised machine learning classification approach is used to organize the float profiles into five SO frontal zones based on their temperature and salinity structure between 300 and 900 m, revealing not only the location of frontal zones and their boundaries, but also the variability of water mass properties relative to the zonal mean state. We find that the variability is property‐dependent and can be more than twice as large as the mean zonal variability in intense eddy fields. In particular, we observe this intense variability in the intermediate and deep waters of the Subtropical Zone; in the Subantarctic Zone just west of and at KP; east of KP in the Polar Frontal Zone, associated with intense eddy variability that enhances deep waters convergence and mixing; and, as the deep waters upwell to the upper 500 m and mix with the surface waters in the southernmost regimes, each property shows a large variability. Plain Language Summary The Southern Ocean strongly influences the global climate system, by absorbing, storing and redistributing heat and carbon across the different ocean basins. Thanks to an increasing number of observations from autonomous instruments, called Argo floats, our understanding of this harsh environment has deepened in the last two decades. Here we use a machine learning technique to automatically classify the float measurements and sort them in regimes with similar properties based on their temperature and salinity vertical structure. The classification results are consistent with previous studies, but are here used to reveal regions where mixing between different types of waters is likely to be occurring. By sorting the float profiles into regimes, we can diagnose regions with larger variation of properties and highlight the transition of the properties across regimes. Given the increasing volume of observations that instruments like the Argo floats are building, a method such as the technique implemented in this study represents a valuable tool that can help to automatically reveal similarities in dynamical regimes. Key points An unsupervised classification technique, applied to temperature and salinity float data, is used to sort the profiles into frontal zones In eddy fields the variability of physical and biogeochemical properties is more than twice as large as the mean zonal variability The intense eddy variability drives lateral physical processes that cause the large property variance 
PY 2020
PD MAR
SO Journal Of Geophysical Research-oceans
SN 2169-9275
PU American Geophysical Union (AGU)
VL 125
IS 3
UT 000534229400026
DI 10.1029/2019JC015877
ID 72471
ER

EF