@article{fdi:010079362,
  title = {{R}econstructing global chlorophyll-a variations using a non-linear statistical approach [+ {C}orrigendum 2020, vol. 7, art. 618249]},
  author = {{M}artinez, {E}lodie and {G}orgues, {T}homas and {L}engaigne, {M}atthieu and {F}ontana, {C}. and {S}auzede, {R}. and {M}enk{\`e}s, {C}hristophe and {U}itz, {J}. and {D}i {L}orenzo, {E}. and {F}ablet, {R}.},
  editor = {},
  language = {{ENG}},
  abstract = {{M}onitoring the spatio-temporal variations of surface chlorophyll-a concentration ({C}hl, a proxy of phytoplankton biomass) greatly benefited from the availability of continuous and global ocean color satellite measurements from 1997 onward. {T}hese two decades of satellite observations are however still too short to provide a comprehensive description of {C}hl variations at decadal to multi-decadal timescales. {T}his paper investigates the ability of a machine learning approach (a non-linear statistical approach based on {S}upport {V}ector {R}egression, hereafter {SVR}) to reconstruct global spatio-temporal {C}hl variations from selected surface oceanic and atmospheric physical parameters. {W}ith a limited training period (13 years), we first demonstrate that {C}hl variability from a 32-years global physical-biogeochemical simulation can generally be skillfully reproduced with a {SVR} using the model surface variables as input parameters. {W}e then apply the {SVR} to reconstruct satellite {C}hl observations using the physical predictors from the above numerical model and show that the {C}hl reconstructed by this {SVR} more accurately reproduces some aspects of observed {C}hl variability and trends compared to the model simulation. {T}his {SVR} is able to reproduce the main modes of interannual {C}hl variations depicted by satellite observations in most regions, including {E}l {N}ino signature in the tropical {P}acific and {I}ndian {O}ceans. {I}n stark contrast with the trends simulated by the biogeochemical model, it also accurately captures spatial patterns of {C}hl trends estimated by satellite data, with a {C}hl increase in most extratropical regions and a {C}hl decrease in the center of the subtropical gyres, although the amplitude of these trends are underestimated by half. {R}esults from our {SVR} reconstruction over the entire period (1979-2010) also suggest that the {I}nterdecadal {P}acific {O}scillation drives a significant part of decadal {C}hl variations in both the tropical {P}acific and {I}ndian {O}ceans. {O}verall, this study demonstrates that non-linear statistical reconstructions can be complementary tools to in situ and satellite observations as well as conventional physical-biogeochemical numerical simulations to reconstruct and investigate {C}hl decadal variability.},
  keywords = {machine learning ; phytoplankton variability ; satellite ocean color ; decadel variability ; global scale ; {MONDE}},
  booktitle = {},
  journal = {{F}rontiers in {M}arine {S}cience},
  volume = {7},
  numero = {},
  pages = {464 [20 + {C}orrigendum, 2 p.]},
  year = {2020},
  DOI = {10.3389/fmars.2020.00464},
  URL = {https://www.documentation.ird.fr/hor/fdi:010079362},
}