Enhancing vegetation monitoring: a proposal for a Sentinel-2 based vegetation health index

Abstract

Vegetation serves as a vital carbon sink, crucial for regulating (Formula presented.) and (Formula presented.) levels in the atmosphere. However, the declining health of vegetation can contribute to a rise in greenhouse gas emissions. Utilizing remote sensing satellite imagery, we can effectively monitor global changes in vegetation health in near-real time. Various vegetation indices have been developed to monitor specific biochemical properties. Yet, many of these indices fall short in detecting health deterioration caused by multiple stressors such as excessive heat, salinity, and water scarcity. Indices that are primarily sensitive to single-leaf parameters may not fully capture the complex stress responses in vegetation. To address this limitation, we introduce a novel vegetation health indicator: the Sentinel-2-based Vegetation Health Index (SVHI). This index is designed to detect stress-induced changes in chlorophyll, water, and protein content. It was validated using global sensitivity analysis (GSA) with physical models and laboratory-based spectroscopy experiments. We have performed a global sensitivity analysis utilizing radiative transfer models to support SVHI’s performance. It indicates strong sensitivity to variations in chlorophyll and water content. Following GSA, a lab-based spectroscopy experiment was conducted to detect the effect of water stress and chlorophyll stress on the vegetation indices. In experiment performed on water stress, SVHI demonstrated five and 1.1 times greater sensitivity than NDVI and NDMI respectively in the early stages of water loss (150%–85% leaf water content), confirmed by Tukey’s HSD test (p < 0.05). It was also observed that NDVI failed to show a statistically significant change during this period (p = 0.63). The experiment performed on the effect of chlorophyll revealed that NDMI could not detect chlorophyll degradation, while SVHI retained sensitivity throughout the chlorophyll decline. Further, we have performed a corn crop phenology analysis using Sentinel-2 data to confirm the effectiveness of SVHI. The analysis revealed that the proposed index successfully distinguishes characteristic changes in vegetation over time. In addition, as compared to NDMI, SVHI differentiates non-vegetated areas, such as water bodies, from vegetated areas. Finally, a temporal analysis of the vegetation indices reveals that SVHI is highly correlated with both NDVI (Formula presented.) and NDMI (Formula presented.), indicating its capability to capture variations in both greenness and moisture content.