Indicator: Soil Carbon Stock

Indicator	Soil Carbon Stock
Description	Plot-level estimates of soil organic carbon stored in the top 30 cm soil from 2018 to the present. Provides a baseline for the current year and changes over the past years.
Unit	Metric tons (total and per hectare)
Temporal Resolution	Annual Assessment
Spatial Resolution	field-level model, 20m and 250m input data
Input Data	IRSIC SoilGrids v2: organic carbon stock, bulk density, texture; satellite observations of cover crops
Method	Soil carbon stock assessment is conducted by coupling soil baseline data with monitoring of land-use practices (such as land use change and cover cropping) and then modelling annual carbon changes using a carbon cycle model (CoolFarmTool) that is based on IPCC tier 1 coefficients. Soil carbon baseline and texture information is extracted from a global soil properties map that was produced in 2018 (IRSIC soilgrids v2 (Poggio, 2020)) and that is the result of a predictive machine learning model that was trained using a large database of publicly available soil samples and environmental covariates, such as geography, terrain, climate, vegetation type, geology etc. Input information, which drives changes in soil carbon stocks encompasses the detection of cover crops history and historic land-use changes since 2018. Information about farming practices on cover crop application is derived from time-series of satellite observations of vegetation and soil dynamics. TLG assesses only the presence of cover crops and land use changes derived via satellite-based crop monitoring, and assumes conventional carbon inputs otherwise (full tillage, no manure application, no land-use change in the past 20 years).
Confidence interval	Confidence intervals are derived from two main sources of uncertainty. First, SoilGrid data, which provides quantiles (05, 50, and 95), is not symmetrically distributed around the median or mean. To prevent negative standard deviation values, a log-normal distribution is assumed, leading to confidence intervals that are not perfectly symmetrical around the median. Second, uncertainty arises from land use, tillage, and management practices. The impact of these factors on soil carbon varies due to micro-environmental differences. For example, tillage may increase soil carbon by a factor of 0.02 in a specific climate and soil type, but the actual effect can range between 0.01 and 0.03.
Benchmarking	Benchmarks are created for each combination of climate zones and soil types, treating each combination as a separate group. Within each group, soil carbon levels are analyzed separately for croplands and grasslands. The poor benchmark is set at the lower 25th percentile of SoilGrid estimates, representing low carbon input conditions. The good benchmark is estimated using the CFT model under optimal management practices. The average benchmark is the midpoint between the poor and good values.