Gorenstein I. auteur aut IRD Wainer I. auteur aut IRD Pausata F. S. R. auteur aut IRD Prado L. F. auteur aut IRD Khodri Myriam auteur aut IRD Dias P. L. S. auteur aut IRD text journalArticle eng text/pdf reformatted digital access Northeast Brazil and Western Africa are two regions geographically separated by the Atlantic Ocean, both home to vulnerable populations living in semi-arid areas. Atlantic Ocean modes of variability and their interactions with the atmosphere are the main drivers of decadal precipitation in these Atlantic Ocean coastal areas. How these low-frequency modes of variability evolve and interact with each other is key to understanding and predicting decadal precipitation. Here we use the Self-Organizing Maps neural network with different variables to unravel causality between the Atlantic modes of variability and their interactions with the atmosphere. Our study finds an 82% (p<0.05) anti-correlation between decadal rainfall in Northeast Brazil and Western Africa from 1979 to 2005. We also find three multi-decadal cycles: 1870-1920, 1920-1970, and 1970-2019 (satellite era), pointing to a 50-year periodicity governing the sea surface temperature anomalies of Tropical and South Atlantic. Our results demonstrate how Northeast Brazil and Western Africa rainfall anti-correlation was formed in the satellite era and how it might be part of a 50-year cycle from the Tropical and South Atlantic decadal variability. specialized ATLANTIQUE AFRIQUE DE L'OUEST AMERIQUE DU SUD BRESIL ZONE TROPICALE 021 032 4 1 427 [11 ] 2023 https://www.documentation.ird.fr/hor/fdi:010088712 10.1038/s43247-023-01073-0 [F B010088712] https://www.documentation.ird.fr/hor/fdi:010088712 https://horizon.documentation.ird.fr/exl-doc/pleins_textes/2024-01/010088712.pdf IRD - Base Horizon / Pleins textes 2024-01-12 2025-02-24 fdi:010088712 fre