Remote Sensing-Based Assessment of Interannual Change of the Ecological Sustainability of Agricultural Landscapes in Northern Benin

Abstract

Assessing ecological sustainability in agricultural landscapes requires approaches that integrate land-cover change, its ecological effects, and their spatial determinants. This study analysed changes between 2023 and 2024 across six agricultural landscapes in northern Benin using the Landscape Ecological Sustainability Index (LESI), calculated for 1 km2 landscape cells from Human Disturbance Coefficients (HDCs) assigned to satellite-derived land-cover classes. Interannual changes were assessed using ΔLESI, the Wilcoxon signed-rank test, Global Moran’s I, and the Local Indicators of Spatial Association (LISA). The contribution of land-cover transitions to HDC change was quantified and a Monte Carlo sensitivity analysis based on classification accuracy was additionally used to assess the robustness of the observed interannual changes to classification uncertainty, and complementary univariate and bivariate regression models examined the relationships between LESI variations and 11 biophysical and geographical variables, including their pairwise interactions. The interannual comparison between 2023 and 2024 showed a decrease in ecological sustainability in four sites, a slight increase in one, and relative stability in another. Significant spatial autocorrelation was detected in all sites (Moran’s I = 0.44–0.65; p < 0.001). Some spatially limited transitions exerted important effects on HDC change. Sensitivity analysis confirmed that the direction of ΔHDC was robust to classification uncertainty in five of the six landscapes. The low explanatory power of the univariate models (R2 ≤ 0.081) indicates that no single variable independently explains the observed changes. The bivariate interaction analyses further showed that some associations were context-dependent, although their explanatory power remained limited, with the best-performing model accounting for only 10.9% of the variation in ΔLESI. This integrated framework provides a reproducible approach for assessing, mapping, and prioritising interannual change of ecological sustainability from remote sensing data to support evidence-based agricultural landscape planning and sustainable land management.