Remote Sensing-based Detection of Cropland Degradation Signals Using Discrete Wavelet Decomposition Analysis – A Case Study in Argentina

Land degradation is a key threat to the productivity of agroecosystems which are increasingly pressured by climate change and growing global populations. Land degradation can impact arable land through a range of pathways including physical processes such as salinization and erosion, loss or inhibition of biological function in soils through chemical or physical deterioration, and through climatic shifts such as aridification which are projected to worsen in coming decades. Land degradation affects a large but uncertain portion of agricultural land globally and the implication of degradation for food production is highly variable. Despite the widely recognized prevalence of cropland degradation and its potential impacts, tools for measuring and monitoring productivity losses over long time periods and large spatial scales are lacking. Many global maps of land degradation rely on outdated statistics, manual surveys, and overly basic image analysis and computational approaches. Existing land degradation assessments also lack the granularity required for decision-making at regional and local levels. The robust spatial and temporal availability of remote sensing imagery presents a unique opportunity to monitor the long-term trends in productivity of cropland through measurements of vegetation greenness as a proxy for yield. In this work, we present a novel methodology for detecting long-term changes in cropland productivity that is globally scalable and robust to changes in land use and management. Using the entire MODIS imagery record (2000-2024), we use a Discrete Wavelet Transform to decompose EVI signals and isolate the long-term trend in vegetation greenness at 250m resolution for a test region covering Argentinian croplands. We find that large areas of maize and soy cropland in Argentina have a negative trend in long-term greenness, with subtle but important long-term declines in productivity that may be attributable to degradation. These declines appear more pronounced in older croplands than in newer croplands suggesting a potential cause in soil health related changes. The approach presented is globally applicable and advances the use of earth observation technology to measure land degradation and monitor land use change.