LAUSR

Abstract Land degradation affects over one-third of the global land area and is projected to become even more widespread due to climate change and land use pressures. Despite being a critical issue for climate change mitigation, biodiversity conservation, and food security, the detection of the onset, duration, and magnitude of land degradation remains challenging, as is early identification of short-term vegetation loss preceding land degradation. Here, we present a new approach for monitoring both short-term vegetation loss and decadal degradation in grasslands using satellite data. Our approach integrates Spectral Mixture Analysis and temporal segmentation, and analyzes dense time-series of satellite observations in three steps. First, we unmix all available satellite observations and aggregate them into monthly composites. Second, we calculate the annual Cumulative Endmember Fractions and examine their piecewise trends among years to determine the onset, duration, and magnitude of short-term vegetation loss and decadal degradation. Third, we attribute a decrease in the green vegetation fraction with a concomitant increase in either open soil, or non-photosynthetic vegetation. We tested our method mapping short-term vegetation loss and decadal degradation in grasslands in the Caucasus Ecoregion using the 2001–2018 time series of MODIS 8-day reflectance data. We found strong patterns of short-term vegetation loss and decadal degradation, mostly in the eastern part of the Caucasus Ecoregion in areas of desert- and semi-desert natural vegetation. Short-term vegetation loss episodes (3–9 years) were more common and had greater magnitude than decadal degradation (≥10 years), especially in steppe regions. On average, 9.3% of grassland area was subjected annually to either decadal, or short-term vegetation loss. Desiccation, i.e., the shift from green vegetation to dry vegetation, was the most prevalent type of change pathway, with green vegetation loss to open soil coming second. Decadal degradation and short-term vegetation loss rates were the highest in dry areas where the potential natural vegetation is sub-shrub deserts, or halophytic, alluvial, and wet lowland forests. Our findings support known general degradation patterns in the Caucasus Ecoregion, but provide better understanding of ongoing processes, by detecting exact location, timing, and magnitude of changes. More broadly, our method advances the monitoring of grasslands by detecting both decadal degradation and short-term vegetation loss. This flexibility supports adaptive degradation monitoring, aids sustainable land management, and provides new information for carbon stock analyses and biodiversity conservation.