Evaluating Default and Domain‐Calibrated Soil Models for Carbon Farming: Evidence From Danish Long‐Term Experiments Using C‐ TOOL and RothC

ABSTRACT

Accurately quantifying soil organic carbon (SOC) dynamics is essential for evaluating carbon farming (CF) practices, which are increasingly integrated into climate mitigation strategies and carbon crediting schemes. This study demonstrates transparent procedures based on Danish long‐term experiments (LTEs) to quantify SOC stock changes (ΔSOC) and test/improve the C‐TOOL and RothC models for CF assessments. Simulations of ΔSOC from straw management (−/+straw) and conversion to grassland with default ‘off‐the‐shelf’ parametrizations were assessed against data from the CENTS and Askov Sandmarken experiments. Stepwise calibration was carried out on the same data, guided by Morris sensitivity analysis (SA) targeting the most influential parameters. The calibrated models were applied to the BRAK experiment to forecast SOC changes under CF scenarios. Morris SA identified initial SOC, grassland C input and the humified‐pool decomposition rate constant as most influential. Targeted calibration improved accuracy and reduced discrepancies between the models. When applied to BRAK scenarios, domain‐calibrated models predicted scenario ΔSOC—the difference between the baseline (−straw) and CF scenarios (+straw or grassland)—to be largest for fertilized winter rye with straw retention (C‐TOOL: 0.67; RothC: 0.69 Mg C ha ⁻¹  year. ⁻¹ ) and for unfertilized spring barley converted to grassland (C‐TOOL: 0.60; RothC: 0.72 Mg C ha ⁻¹  year. ⁻¹ ). Moreover, inter‐model ΔSOC—the difference between RothC and C‐TOOL—decreased after calibration across BRAK scenarios (average: 9.9 → 3.5 Mg C ha ⁻¹ ). Overall, this LTE‐based and SA‐guided calibration framework improves the credibility of SOC model predictions for CF, strengthening their use in monitoring, reporting and verification (MRV) frameworks and carbon trading schemes.