Soil organic carbon (SOC) sequestration projects require unbiased, precise and cost-effective Monitoring, Reporting, and Verification (MRV) systems that balance sampling costs against uncertainty deductions imposed by regulatory frameworks. Design-based estimators guarantee unbiasedness but cannot exploit auxiliary data. Model-based approaches (VCS Methodology VT0014 v1.0 (2025)) can improve precision but require independent validation for each project. Model-assisted estimation offers a robust compromise, combining model predictions with probability sampling to retain design-based guarantees while improving precision. We evaluate the scientific integrity and efficiency of the simple regression estimator (SRE), a well-known model-assisted estimator, via an extensive simulation study. Our simulations span diverse SOC stock variances, sample sizes, and model performances. We assess three core properties: empirical bias, empirical confidence interval coverage, and precision gain relative to the design-based Horvitz-Thompson estimator (HTE). Results show negligible bias and valid coverage probabilities for n > 40, regardless of SOC stock variance. Below this threshold, variance approximations and normality assumptions yield unreliable uncertainty estimates. With correlated ancillary variables (r^2 = 0.3), SRE achieves 30% precision gains over HTE. With uncorrelated variables, no gains are observed, but performance converges to HTE for n >= 40. Model-assisted estimation can enhance project economics without compromising scientific rigor. Regulators should permit such estimators while mandating minimum sample size thresholds. Project proponents should routinely employ such estimators when correlated ancillary variables exist. The industry should prioritize the retrieval of high-quality, project-specific covariates to maximize precision gains and thereby the project economics.
翻译:暂无翻译