Assessing the performance of NDVI, 2-band EVI and MSAVI vegetation indices for land degradation monitoring across variable biomass cover at global scale.

Diminished overall land productivity and reduced resilience in the face of climate and environmental change have made addressing land degradation a global priority, formalized by the United Nations Convention to Combat Desertification and the Sustainable Development Goals, in particular Target SDG 15.3 on Land Degradation Neutrality. Remote sensing offers the most cost-effective approach to assess large scale Earth surface change and Trends.Earth was developed to facilitate land degradation assessments using Earth observation data. Vegetation indices (VI) have been broadly used as proxies to estimate land productivity, given that they can be readily derived from imagery covering large extents and over long time-series, serving as one of the indicators to map and monitor land degradation. The Normalized Difference Vegetation Index (NDVI) is the most studied and accepted vegetation index for monitoring changes in land productivity. Nevertheless, several studies have found that NDVI tends to asymptotically reach a plateau over areas with high-biomass, and that it also has limited capacity to cope with influences from background soil in sparsely vegetated areas. Other vegetation indices could provide improved sensitivity when measuring land productivity for locations with biomass on the two extremes of the spectrum and as such help on the assessments of changes in land condition and achieving land degradation neutrality by 2030. The 2-band Enhanced Vegetation Index (EVI2) has been demonstrated to not saturate over highly vegetated areas, and the Modified Soil Adjusted Vegetation Index (MSAVI) has been shown to be a robust VI for sparsely vegetated lands. We used global MODIS data to evaluate the performance of NDVI compared to these two alternative vegetation indices. Performance of each VI was assessed against a global dataset representing biomass from sparsely vegetated grasslands to dense tropical forests. Our results support robust recommendations on what indicator is best suited for assessing land degradation in different regions of the globe for monitoring progress towards a land degradation neutral world.