This dataset compiles the results of a critical meta-analysis based on an exhaustive literature search in the Web of Science database, targeting peer-reviewed primary studies published up to early 2024 and focusing on soil organic carbon (SOC) dynamics in countries of the Global South. The initial literature search identified 3,641 unique articles. Abstracts were screened using keywords such as “soil carbon”, “SOC”, “carbon”, “stocks”, “long-term”, “experiment”, “changes”, “decades”, “pluriannual”, leading to a shortlist of 452 articles. From these, studies were selected based on the following criteria: (i) long-term field experiments lasting more than 5 years, (ii) availability of SOC stock or concentration data at both the initial and final stages of the experiment (diachronic approach), and (iii) exclusion of studies based on remote sensing, modeling, or space-for-time substitutions (synchronic approach). In cases where only soil organic matter (SOM) was reported, SOC was estimated using a conversion factor of 0.58 (Pan et al., 2004). Based on these criteria, the final database includes 155 primary studies conducted across 174 long-term experimental sites, resulting in 1,362 paired observations comparing SOC content or stock at two time points. Data were extracted from text, tables, and figures, including digitization of graphical data using WebPlotDigitizer (Burda et al., 2017; Kpemoua et al., 2022), and organized in a structured Excel file (Supplementary Material 1).

For each study site, extensive metadata were compiled, including: - Geographic location - Ecosystem type (forest, grassland, cropland, fallow, agroforestry, crop-livestock systems) - Soil management practices (conventional tillage, minimum tillage, permanent beds, no-tillage) - Crop management details (mineral and organic fertilization rates and types, crop rotation, diversification, etc.) - Climate variables (climate type, mean temperature, precipitation when available) - Soil physicochemical characteristics: clay, silt, sand, pH, bulk density, total nitrogen, C/N ratio, CaCO₃. Soil texture classes were derived using the USDA textural triangle, then grouped into coarse, medium, and fine texture categories following the Soil Survey Manual . Carbon input data: crop yields, aboveground biomass, and exogenous organic matter additions were recorded when available. Where inputs were reported as dry matter, a conversion factor of 0.45 was applied to estimate carbon input. All exogenous organic inputs were expressed in tons of carbon.

To our knowledge, this is the most comprehensive and methodologically rigorous dataset on long-term SOC dynamics in the Global South. It provides valuable insights into the drivers of SOC sequestration, associated co-benefits, and knowledge gaps in tropical, subtropical, and Mediterranean systems. This resource is expected to support evidence-based climate policy design and contribute to international scientific discussions, including those led by bodies such as the Intergovernmental Panel on Climate Change (IPCC).