Satellite imagery for monitoring badland evolution and vegetation dynamics in a global change context

Satellite remote sensing provides consistent data collection over large areas. Programs such as Landsat offer the longest continuous space-based record, freely accessible, and have revolutionized the understanding of Earth’s surface dynamics, particularly in regions hard to reach and with limited historical in-situ data. One of the most widely used indices derived from specific satellite bands is the Normalized Difference Vegetation Index (NDVI), a powerful tool for assessing changes in vegetation cover and linking these transformations to broader environmental and anthropogenic factors. A multidecadal approach based on satellite imagery data was applied to an Italian badland site (located in southern Tuscany), with the aim of contributing to the comprehension of its recent morphodynamics. Although this temperate landscape could naturally sustain vegetation due to seasonal soil moisture patterns, it has been heavily impacted by human activities, such as several historical deforestation phases, which removed the protective vegetation cover, leaving the soil exposed to erosion and leading to the formation of distinctive erosive landforms. Nowadays, with the widespread abandonment of rural areas, these landforms appear to be gradually decreasing, allowing the landscape to slowly return to its natural state. Moreover, this process is further triggered by climate change, which alters precipitation patterns and temperatures arising, significantly affecting vegetation dynamics. The specific objectives of this study are to address the following research questions: (i) Is the area undergoing a revegetation process that may lead to the gradual reduction of badland surfaces? (ii) Is there a seasonal trend characterizing the long-lasting vegetation changes? (iii) What is the role of global change, encompassing shifts in precipitation and temperature patterns as well as land use changes, in driving this dynamic? To address these questions, satellite imagery from the Landsat program (Landsat 5 TM, Landsat 8 OLI, and Landsat 9 OLI) is analyzed through Google Earth Engine to generate the 30-meter NDVI dataset for the period 1984–2023. Cloud-covered pixels are excluded based on the Quality Assessment band, and a monthly pixel-wise analysis was conducted using the Mann-Kendall test combined with the Sen's slope estimator, a widely used non-parametric statistical method for detecting monotonic trends in climate data. Only pixels with a confidence level greater than 95% are considered in the analysis. Additionally, in situ data on precipitation (monthly averages and intensity) and temperature (monthly min, max, and mean) are compared with monthly satellite-derived temperature data from the Landsat Surface Temperature product and precipitation data from the Global Precipitation Measurement Mission, using the same analysis period and statistical approach. Finally, land use changes are evaluated using the CORINE Land Cover datasets for 1990, 2000, 2006, 2012, and 2018, alongside a 1978 land use map from the Tuscany Region. The analysis reveals a steady increase in vegetation cover, suggesting a possible prolongation of the growing season. This trend is accompanied by relatively stable precipitation patterns and a general warming trend. These climatic drivers, along with expanding forested and agricultural areas and a decline in badland landforms, highlight the impact of both climate and land use changes on landscape dynamics.