Mediterranean agricultural landscapes in south-eastern Spain are increasingly affected by gradual, spatially heterogeneous soil degradation processes, including vegetation decline, water stress, bare soil exposure and surface warming. Early detection is challenging when monitoring relies on occasional field observations alone, yet timely identification of at-risk areas is critical for guiding soil health management and combating desertification. Developed within the NEMESIS project, a Mediterranean Soil Health Living Lab network, this study supports efforts to combat desertification, reduce land degradation and promote sustainable land management through soil health descriptors, Earth Observation (EO) data and place-based monitoring.
This work presents a satellite-based framework for assessing and predicting desertification risk in agricultural plots located in the Region of Murcia (37.930981 N, −2.203217 E), using Sentinel-2 imagery acquired at five-day intervals between 2015 and 2026. Seven spectral and biophysical indicators were derived from the multitemporal archive: NDVI and MSAVI to describe vegetation vigour and canopy development; NDMI to represent moisture-related vegetation response; BSI and BSF for monitoring bare soil exposure; a land surface temperature proxy (LST) to capture thermal stress; and FCover to estimate fractional vegetation cover.
These indicators were integrated into a Desertification Degree Index (DDI), a spatial composite combining all seven components using weighted aggregation and global normalisation across the observation period to ensure temporal comparability. Pixel-wise temporal trend analysis was performed using linear regression, producing maps of change and statistical significance masks to identify areas with consistent degradation or recovery signals.
To explore short-term risk evolution, predictive models were applied to the multitemporal DDI series. These models used the historical behaviour of the indicators to forecast future desertification risk and generate maps of expected surface condition, allowing both plot-level interpretation and pixel-level identification of vulnerable areas.
The results identified zones of elevated desertification risk and captured intra-annual vegetation and soil dynamics that would be overlooked by annual composites. Areas with lower vegetation indices, persistent bare soil exposure, reduced moisture response and higher surface temperature were associated with greater vulnerability, while zones with stable vegetation cover and lower thermal stress showed lower apparent risk. The short-term forecasts provided information on likely surface condition, supporting agronomic intervention. By translating a dense decade-long satellite archive into actionable soil health indicators, this work demonstrates the operational potential of EO for early warning and targeted monitoring in Mediterranean semi-arid agricultural systems.