Multi-decadal satellite monitoring of soil carbon and its role in farm carbon footprint: a case study for the Czech Republic

This paper presents a study on the retrieval of long-term trends in soil organic carbon (SOC) using advanced time-series analysis applied to multi-temporal Landsat and Sentinel-2 satellite data processed in the form of 5-year bare soil composites. The research is focused on the Czech Republic, a region that is characterized by a diverse range of soil types and a variety of agricultural practices. The study integrates a support vector regression algorithm with soil type mapping to construct a model of soil organic carbon (SOC) and their long-term trends at different spatial scales, including laboratory and field spectroscopy and Earth observation (EO) scales. The highest accuracy of SOC retrieval was obtained for soil optical properties measured using a field spectroradiometer (LAB2LAB case) for a scenario involving wet soil (r² = 0.73), followed by satellite retrievals (EO2EO scenario) with r² = 0.63. When applied retrospectively to Landsat satellite observations, long-term SOC trends from Landsat observations between 1985 and 2023 were retrieved across different soil types, thereby highlighting the influence of soil characteristics on these trends. Furthermore, the methodology proposed offers a novel approach to evaluating the carbon footprint at the farm level, integrating soil organic carbon dynamics into carbon accounting systems.