Variation of woody plants diversity and land use along a bioclimatic gradient of agroforestry landscapes in Senegalese Sahel

In the Sahel region, landscape configuration is closely linked to factors such as climate, ecology, soil composition, agronomy, livestock, and biology. Over the past decades, significant changes in these factors have been observed, including shorter rainy seasons, irregular precipitation, a decrease in biomass productivity, and rapid population growth, negatively impacting local agricultural and pastoral systems. In response to this pressure, mitigation strategies have been implemented to contribute to the improvement of local food, nutritional, and economic security. Agroforestry systems, involving a combination of trees, shrubs, crops, and animals in the same plot, represent one of these strategies. Therefore, characterizing these systems in the current context of climate change is crucial for sustainable natural resource management.

In this study, three agroforestry landscapes of the Senegalese Sahel were described, spanning a bioclimatic gradient from the Louga region (Ouarkhokh) in the north to the Fatick region (Niakhar) in the center, and the Tambacounda region (Koussanar) in the south. The data utilized included satellite imagery synthesis (Sentinel-2 and Spot), landscape variables (rainfall, evapotranspiration, biomass, and vegetation), spectral indices (NDVI, NDRE, GNDVI), and field data on land use and woody cover. The methodology consisted of three main approaches: (i) landscape stratification involving Sentinel image segmentation in 2021, selection of relevant landscape variables, and mixed discriminant factor analysis to establish landscape heterogeneity; (ii) land use and land cover mapping through supervised pixel-based classification using a Random Forest (RF) machine learning classifier with 500 trees; (iii) floristic diversity analysis by assessing floristic composition and calculating diversity indices (i.e., Shannon, Pielou, and Simpson indices).

Landscape stratification identified seven classes with distinct landscape characteristics. Classes (1, 2, and 4) in the Ouarkhokh site had lower average biomass, rainfall, and actual evapotranspiration values than classes (3 and 4) in the Niakhar site. Similarly, classes (5, 6, and 7) in the Koussanar area had higher average biomass, rainfall, and actual evapotranspiration values than the first two sites. Land use mapping showed vegetation predominance in the Ouarkhokh site, significance in the Koussanar site, and low presence in the Niakhar area. Other identified units (cultivated areas, built-up areas, water, and bare land) were dominant in the Niakhar area, present in the Koussanar site, and low in the Ouarkhokh area. Likewise, vegetation dominated in classes 1, 5, 6, and 7. Class 1 was exclusively found in Ouarkhokh, while classes 5, 6, and 7 were located in the Koussanar site. The majority of cultivated surfaces were in class 3, exclusively located in the Niakhar area. Species richness was higher in the Niakhar area (60 species, 21 families) and lower in the Koussanar area (56 species, 16 families) and Ouarkhokh area (31 species, 13 families). This landscape distribution of land use, landscape classes, and identified species highlights the influence of anthropogenic, soil-related, and climatic factors specific to each site.