%0 Journal Article
%9 ACL : Articles dans des revues avec comité de lecture répertoriées par l'AERES
%A Ouattara, B.
%A Thiel, M.
%A Forkuor, G.
%A Mouillot, Florent
%A Laris, P.
%A Tondoh, E. J.
%A Sponholz, B.
%T Fire Impacts, vegetation Recovery, and environmental drivers in West African savannas (2014-2023) : a High-Resolution remote sensing assessment
%D 2025
%L fdi:010094852
%G ENG
%J International Journal of Applied Earth Observation and Geoinformation
%@ 1569-8432
%K Burned area ; Net primary productivity ; Post-fire recovery ; Machine learning ; Harmonized Landsat-Sentinel-2 ; Savanna fires ; West Africa
%K AFRIQUE DE L'OUEST ; BURKINA FASO ; GHANA ; COTE D'IVOIRE
%M ISI:001547687400001
%P 104783 [22 ]
%R 10.1016/j.jag.2025.104783
%U https://www.documentation.ird.fr/hor/fdi:010094852
%> https://horizon.documentation.ird.fr/exl-doc/pleins_textes/2025-09/010094852.pdf
%V 143
%W Horizon (IRD)
%X Savanna fires are a dominant ecological force in West Africa, shaping land systems, carbon dynamics, and biodiversity. Yet, their impacts on ecosystem productivity and recovery remain poorly quantified at meaningful spatial and temporal scales. This study presents a decadal assessment (2014-2023) of fire activity and post-fire vegetation response across a similar to 229,000 km(2) transboundary region of Burkina Faso, Ghana, and C & ocirc;te d'Ivoire. Using Harmonized Landsat-Sentinel (HLS) imagery and VIIRS active fire detections, we mapped burned areas (BA) at 30 m resolution-capturing extensive small fires often missed by global datasets. Fire-induced Net Primary Productivity (NPP) losses were estimated using downscaled MODIS productivity data, and post-fire recovery times were tracked at monthly and annual scales. Fires were highly seasonal, with > 80 % of BA occurring between November and January, peaking in December. Despite a dip around 2017, interannual BA remained relatively stable (0.29 % yr(-)1 increase, p > 0.05). Immediate NPP losses averaged similar to 11 x 10-2 Mg C ha(-)1 per year, with higher per-hectare losses in forested and high-biomass zones. Roughly 65 % of BA recovered to pre-fire NPP levels within a year, primarily in grasslands and croplands. However, recovery in woody and mesic areas was slower and more variable. We emphasize that recovery was assessed in terms of NPP (carbon uptake), not structural biomass or species composition-functional recovery does not necessarily imply full ecological recovery. Using machine learning, we identified soil moisture (dry-season NDMI) and temperature as dominant predictors of recovery time, with soil fertility (nitrogen content) and water retention capacity emerging as key drivers. Interestingly, fire frequency and land cover type had limited predictive power once climate and soil factors were accounted for, suggesting that environmental factors, more than fire regime characteristics, shape recovery. These findings support the idea that well-timed, low-intensity fires-particularly early-season burns-can promote carbon resilience in fire-adapted landscapes. This underscores the value of high-resolution remote sensing and soil data in guiding fire-smart management and balancing ecological and livelihood goals under climate change.
%$ 082 ; 126