@article{fdi:010093410,
  title = {{P}redicting {A}tlantic and {B}enguela {N}iño events with deep learning},
  author = {{B}ach{\`e}lery, {M}. {L}. and {B}rajard, {J}. and {P}atacchiola, {M}. and {I}llig, {S}erena and {K}eenlyside, {N}.},
  editor = {},
  language = {{ENG}},
  abstract = {{A}tlantic and {B}enguela {N}iño events substantially affect the tropical {A}tlantic region, with far-reaching consequences on local marine ecosystems, {A}frican climates, and {E}l {N}iño {S}outhern {O}scillation. {W}hile accurate forecasts of these events are invaluable, state-of-the-art dynamic forecasting systems have shown limited predictive capabilities. {T}hus, the extent to which the tropical {A}tlantic variability is predictable remains an open question. {T}his study explores the potential of deep learning in this context. {U}sing a simple convolutional neural network architecture, we show that {A}tlantic/{B}enguela {N}iños can be predicted up to 3 to 4 months ahead. {O}ur model excels in forecasting peak-season events with remarkable accuracy extending lead time to 5 months. {D}etailed analysis reveals our model's ability to exploit known physical precursors, such as long-wave ocean dynamics, for accurate predictions of these events. {T}his study challenges the perception that the tropical {A}tlantic is unpredictable and highlights deep learning's potential to advance our understanding and forecasting of critical climate events.},
  keywords = {{ATLANTIQUE} ; {BENGUELA}},
  booktitle = {},
  journal = {{S}cience {A}dvances},
  volume = {11},
  numero = {14},
  pages = {eads5185 [12 p.]},
  ISSN = {2375-2548},
  year = {2025},
  DOI = {10.1126/sciadv.ads5185},
  URL = {https://www.documentation.ird.fr/hor/fdi:010093410},
}