Approximate Bayesian inference for calibrating the IPCC tier-2 steady-state soil organic carbon model for Canadian croplands using long-term experimental data

We conducted a Bayesian calibration of the IPCC tier-2 Steady-State (IPCCT2) model using long-term experimental (LTE) data from Canadian croplands. A global sensitivity analysis identified key parameters influencing the prediction of soil organic carbon (SOC) stocks, including those governing the temperature response curve, optimal decay rate in the passive pool, and stabilization efficiencies for decay products in different pools. We used Sampling-Importance-Resampling to obtain posterior parameter and hyperparameter distributions for the sensitive parameters and the tillage disturbance modifiers. The calibration significantly narrowed parameter ranges compared to the original parameter range provided by the IPCC guidelines, reducing relative uncertainty in SOC point estimates from 27-33 % to 3.5–4 % - an 85 % reduction in model uncertainties. However, calibration was much less efficient in reducing model uncertainties if the correlation structure in the posterior samples was unaccounted for. Calibrated parameters effectively minimized Root Mean Squared Error and bias in SOC predictions in a validation dataset. The default IPCC tier-2 steady-state model parameters performed comparably to those obtained from maximum a priori distributions. Our findings highlighted the broad nature of original IPCC guideline boundaries, leading to uncertain SOC stock predictions and limiting model informativeness and emphasizing the need for parties to adapt parameters to their country-specific conditions. Simulation results suggested that the calibrated model parameter ranges are essential for accurate predictions. When simulating the impact of reducing tillage or adding inorganic nitrogen to annual crops without manure amendments, model calibration substantially reduced uncertainties in long-term impact predictions—by ∼15 % for tillage and ∼75 % for nitrogen addition. This study underscores the accuracy of default IPCCT2 parameters in simulating SOC dynamics in Canadian LTE studies. However, it emphasizes the need for calibrated model parameters in conducting uncertainty analyses. The Bayesian calibration improved uncertainty assessments of cropland management practices leading to reliable carbon accounting. This work supports informed decision-making towards sustainable agriculture, guiding management strategies that optimize carbon storage while aligning with national and international carbon reporting frameworks. • ·Bayesian calibration of IPCC Tier 2 steady state SOC model was performed with Canadian long-term experimental data. • Calibration of Model parameters reduced uncertainty of the SOC predictions. • Calibration reduced uncertainty in evaluating impacts of tillage and N application on soil organic carbon.