Evaluating NDVI-Based Inference of Vegetation States in a Semi-Arid Rangeland Restoration Site

Abstract

Rangelands are critical land systems that pose unique management and monitoring challenges due to their size and complexity. The Normalized Difference Vegetation Index (NDVI) is commonly used to monitor rangeland condition and guide restoration, both as an indicator of vegetation condition and as a basis for classifying land-cover types. However, these applications rely on assumptions that may not be accurate in semi-arid systems. Here, we directly evaluate the strengths and limitations of Landsat-based NDVI data in rangeland monitoring with field vegetation surveys and soil sampling at a rangeland restoration site in northern Jordan. Our results showed that NDVI was positively correlated with aboveground indicators, including vegetation cover (Spearman’s ρ = 0.70, p < 0.001) and aboveground biomass (Spearman’s ρ = 0.45, p = 0.034), supporting its utility as a measure of vegetation condition. In contrast, NDVI showed limited evidence of association with soil organic carbon, captured only a weak portion of variation in plant species composition, and did not clearly distinguish vegetation types defined by dominant species in the field. These results suggest that Landsat-based NDVI is useful for tracking continuous vegetation attributes but should be interpreted cautiously when used to infer vegetation types or belowground recovery in semi-arid systems. We highlight the importance of aligning the use of NDVI with ecological properties it can reliably represent and emphasize the continued need for field data to accurately assess rangeland structure, composition, and restoration outcomes.