@article{fdi:010095766,
  title = {{O}bservation-only deep learning for gappy satellite-derived ocean color data using 4{DV}ar{N}et},
  author = {{D}orffer, {C}. and {J}ourdin, {F}. and {N}guyen, {T}.{T}.{N}. and {D}evillers, {R}odolphe and {M}ouillot, {D}. and {F}ablet, {R}.},
  editor = {},
  language = {{ENG}},
  abstract = {{M}onitoring optical properties of coastal and open ocean waters is crucial to assessing the health of marine ecosystems. {D}eep learning offers a promising approach to address these ecosystem dynamics, especially in scenarios where gapfree ground-truth data is lacking, which poses a challenge for designing effective training frameworks. {U}sing an advanced neural variational data assimilation scheme (called 4{DV}ar{N}et), we introduce a comprehensive training framework designed to effectively train directly on gappy data sets. {U}sing the {M}editerranean {S}ea as a case study, our experiments not only highlight the high performance of the chosen neural network in reconstructing gap-free images from gappy datasets but also demonstrate its superior performance over state-of-the-art algorithms such as {DI}n{EOF} and end-to-end neural mapping schemes based {CNN} or {UN}et architectures.},
  keywords = {{MEDITERRANEE}},
  booktitle = {},
  journal = {{IEEE} {T}ransactions on {G}eoscience and {R}emote {S}ensing},
  volume = {63},
  numero = {},
  pages = {4212512 [12 ]},
  ISSN = {0196-2892},
  year = {2025},
  DOI = {10.1109/tgrs.2025.3624465},
  URL = {https://www.documentation.ird.fr/hor/fdi:010095766},
}