The science behind the credit
Terrabit's verification model integrates direct soil measurement with continuous satellite monitoring — producing a statistically robust SOC estimate that meets international MRV standards.
Request White Paper
Soil organic carbon measurement
Soil organic carbon (SOC) quantification begins with physical soil cores extracted at standardized depth intervals: 0–30 cm, 30–60 cm, and 60–90 cm. This depth standardization aligns with the requirements of Verra VCS VM0042 and enables comparability across monitoring periods.
Laboratory analysis uses Loss-on-Ignition (LOI) as the primary method: dried soil samples are combusted at 550°C, and weight loss is attributed to organic matter oxidation. LOI results are converted to SOC using the van Bemmelen factor (0.58) with field-specific calibration. Walkley-Black wet oxidation analysis is performed on 10% of samples as a quality control cross-check, with a reconciliation procedure applied when LOI and Walkley-Black results diverge by more than 0.3 percentage points.
Sampling density follows a stratified random design: minimum 1 composite sample (5 cores pooled) per 40 acres, with composite formation by soil mapping unit. GPS coordinates and timestamp are recorded for every core location. Bulk density measurements from undisturbed cores enable conversion from SOC concentration (%) to carbon stock (tCO₂e/ha).
The carbon stock equation follows IPCC Tier 2 methodology: C stock (tCO₂e/ha) = SOC% × BD × depth × 3.67, where BD is bulk density in g/cm³ and 3.67 is the molecular weight ratio of CO₂ to C.
Analysis methods
Primary Method
Loss-on-Ignition (LOI)
550°C combustion, van Bemmelen conversion
QC Cross-Check
Walkley-Black
Wet oxidation, 10% of samples
Precision
±0.3% SOC
Per certified laboratory protocol
Satellite spectroscopy
Satellite spectroscopy provides the continuous monitoring layer between sampling campaigns. Sentinel-2 MultiSpectral Instrument (MSI) Level-2A atmospherically corrected imagery is processed for SWIR1 (1610nm) and SWIR2 (2190nm) band reflectance. These two shortwave infrared bands are selected because organic matter in soil absorbs more infrared radiation at higher concentrations — producing a measurable inverse correlation between SOC% and SWIR reflectance.
Field calibration is performed by co-registering satellite observations with soil sample locations from the most recent sampling campaign. A field-specific calibration regression is fit to the reflectance-SOC pairs. The calibrated model is applied to the full satellite time series to produce monthly SOC distribution maps across enrolled acreage.
Landsat 9 TIRS-2 thermal data is used for soil moisture correction, as moisture significantly affects SWIR reflectance independent of SOC. Uncorrected moisture effects are a common source of error in satellite-only SOC estimation — Terrabit's dual-method approach allows direct quantification and correction of moisture-induced bias.
Temporal monitoring cadence: Sentinel-2's 5-day revisit frequency (at mid-latitude Midwest sites, with cloud masking) produces approximately 30–50 usable observations per year per field. This enables detection of practice changes and confirms consistency of SOC trajectories between annual sampling campaigns.
Machine learning integration
The Terrabit verification model integrates direct soil measurements and satellite observations into a final SOC estimate using an ensemble of gradient-boosted trees and random forests. The ensemble is trained on the USDA NRCS SSURGO digital soil survey dataset for Iowa and Illinois — approximately 120,000 soil horizon records representing a comprehensive characterization of Midwest agricultural soils.
Model features include: laboratory SOC measurements from current and prior sampling campaigns, satellite SWIR reflectance time series statistics (mean, variance, trend), bulk density, soil texture class, elevation, topographic wetness index, and land management practice indicators. The ensemble combines the predictions of 500 trees with balanced variance-bias tradeoffs.
Cross-validation against 847 USDA NRCS direct measurements (not used in training) yields a root mean square error (RMSE) of 0.28% SOC and a Pearson r of 0.91. This performance is reported in the Terrabit methodology white paper and is available for third-party review.
Uncertainty quantification uses conformal prediction to produce field-level 95% confidence intervals on each SOC estimate. The net sequestration calculation subtracts a permanence buffer pool deduction (5–10% depending on protocol) from gross sequestration before credit issuance.
Model performance summary
Methodology alignment with voluntary carbon standards
Compatibility statements reflect methodology alignment, not pre-certification or endorsement by the named standards bodies.
| Terrabit Methodology Step | Verra VCS VM0042 | Gold Standard Soil Carbon | CAR Soil Enrichment |
|---|---|---|---|
| Stratified random sampling design | Section 8.1 | Annex I | §3.3 |
| 0–30, 30–60, 60–90 cm depth intervals | Required | Required | Required |
| LOI primary / Walkley-Black QC | Accepted | Accepted | Accepted |
| Bulk density undisturbed core method | Required | Required | Required |
| Satellite supplementary monitoring | Optional Tier 3 | Remote sensing allowed | Approved supplemental |
| Additionality assessment (financial + common practice) | VCS combined tool | GS additionality tests | CAR regulatory surplus |
| Permanence buffer pool deduction | AFOLU buffer pool | Soil carbon reversal buffer | CAR buffer account |
| 95% confidence interval reporting | Required for Tier 3 | Uncertainty disclosure required | Statistical uncertainty required |
Download the methodology white paper
The complete Terrabit methodology document — suitable for ESG disclosures, CDP responses, and due diligence review. Contact us with your name and organization to receive the current version.