Climate-Driven or Not? Explaining Seasonal NDVI Variability in the Sahel

Vegetation change in the Sahel has long been framed through competing narratives on desertification and greening, but existing literature often overlooks sub-annual patterns. Therefore, this research investigates the timing, duration and magnitude of vegetation growth in the West African Sahel, which is important for climate-sensitive livelihood practices such as pastoralism and semi-permanent agriculture. Our research shows that studying the range between the seasonal maximum and minimum vegetation conditions (further defined as ‘seasonal envelope’) is more informative than looking at the annual means: it reveals which part of the seasonal cycle is changing and why that matters functionally. Additionally, linear regression modelling indicates that, while climate variables explain most of the seasonal variability, consistent prediction errors at the driest and hottest extremes point to non-linear vegetation responses.

Existing studies often assess vegetation dynamics using long-term satellite-derived Normalized Difference Vegetation Index (NDVI) records. Following this approach, our study uses combined Landsat and MODIS datasets to examine vegetation dynamics over the last 45 years at 500 x 500m resolution. Rather than focusing on annual-average trends, wet-season peak productivity and dry-season minimum conditions are analysed as seasonal indicators that directly guide land use, resource access, and mobility decision of local inhabitants, including pastoralists and agriculturalists in Ghana, Mali and Nigeria.

Over recent decades, dry-season NDVI minima have remained relatively stable, while wet-season NDVI maxima have flattened or declined across large parts of the study regions. This divergence suggests a contraction of the seasonal NDVI envelope driven primarily by reduced peak productivity rather than declining baseline vegetation conditions, as often suggested in literature. Time-series decomposition of month-to-month NDVI variability, combined with analysis of long-term seasonal means and seasonal peak values reveals functional asymmetries in vegetation response over the last 45 years.

Climate variables explain most of the interannual variability and long-term trends in linear predictive NDVI models, yet systematic modelling errors at seasonal extremes indicate the influence of non-climatic factors. To distinguish where vegetation dynamics are climate-driven or not, the second part of the study moves from historical analysis to NDVI modelling using rainfall, temperature, soil moisture, and lagged NDVI as predictor. A linear regression framework is applied, not to maximise predictive accuracy, but to diagnose where standard climate-NDVI relationships succeed or fail in capturing month-to-month vegetation responses. The largest deviations are found under high near-surface temperatures and prolonged low-precipitation conditions, proving that vegetation responses in very hot, dry circumstances are not fully captured by linear climate-NDVI relationships. These nonlinear responses are particularly noticeable during the dry-season extremes.

These results establish a baseline for interpreting vegetation change that is directly relevant to land-use monitoring, early-warning systems, food security planning, and climate adaptation policy. In semi-arid regions, where livelihoods depend on narrow windows of resource availability, small shifts in wet-season productivity or dry-season duration can result in large socio-ecological consequences.