Tree decline in arid and semi-arid forest ecosystems causes severe socioeconomic and ecological problems and thus needs to be thoroughly quantified and monitored across space and time. This study investigates tree and forest decline in Iran’s Zagros forests, considering environmental factors (e.g. topographic, soil, and climatic variables). We used field data from Chaharmahal-and-Bakhtiari (a study area covering 165 km2) and environmental data derived from freely available databases. Relationships between tree, forest decline, and environmental data were analyzed using generalized additive models. Our findings reveal that slope and the BioClim-16 variable (precipitation of the wettest quarter) significantly influence tree decline across various decline classes (P-values: slope = .009, BioClim-16 = .02). The best multivariate model for forest decline incorporated soil organic carbon and silt as predictive variables, with soil organic carbon emerging as the key factor (P-value = .04). Additionally, a spectral analysis of bare soil in declining and non-declining areas consistently demonstrated reduced reflectance values in declining regions across 10 Sentinel-2 bands, with VNIR-3, SWIR-2, red, green, and blue bands consistently showing significant differences as unveiled by the Wilcoxon test in all seasons except winter. These reduced reflectance values may indicate that forests stocked on soils with larger grain size (a higher fraction of sand) and/or higher organic carbon content may be more vulnerable to decline. This study contributes to our hitherto understanding of the main drivers of tree and forest decline in semi-arid forests, among others underscoring the potential utility of the spectral properties of bare soil in sparse semi-arid forests to predict the likelihood of tree decline.