A spatially explicit dynamic biomass baseline for forest-management carbon projects in China

Technical Summary: A Spatially Explicit Dynamic Biomass Baseline for Forest-Management Carbon Projects in China

Problem Statement
Forest carbon projects, particularly Improved Forest Management (IFM) initiatives, rely on counterfactual baselines to determine "additionality"—the carbon gains achieved beyond what would have occurred without the project. Current baseline methodologies often suffer from significant biases. Static reference levels and coarse regional averages fail to account for the dynamic nature of forest biomass, which is influenced by stand age, historical management rotations, and natural or anthropogenic disturbances. These oversimplifications can lead to two critical errors: baselines set too low overstate mitigation gains (over-crediting), while baselines set too high obscure genuine project achievements. Existing post-hoc evaluation methods (e.g., matching or synthetic controls) are valuable for review but cannot provide standardized, spatially explicit ex-ante baselines required at the project registration phase. There is a need for a framework that integrates age-dependent growth, spatial biomass potential, historical management continuity, and observed loss risks into a single, auditable projection system.

Methodology
The authors developed a dynamic, spatially explicit counterfactual baseline framework for China, utilizing a coupled growth-curve approach. The methodology proceeds through the following steps:

1. Data Stratification and Growth Curve Fitting:

   • A national dataset of 24,237 forest plot records was stratified into 155 ecological strata based on vegetation zone, forest type, and origin (natural vs. planted).
   • For each stratum, age–biomass accumulation curves were fitted using Richards, Logistic, and Weibull functions. A library of 67 final growth curves was established.
   • Uncertainty analysis indicated that the asymptotic biomass term (the maximum potential biomass) was the primary driver of uncertainty in these fitted curves.

2. Spatial Biomass Ceiling Estimation ($B_{eco}$):

   • To address the uncertainty in the asymptotic term, the authors replaced the static asymptote in the growth equations with a spatially explicit estimate of attainable near-natural biomass ($B_{eco}$).
   • $B_{eco}$ was modeled using an ensemble of 50 Random Forest models trained on mature, low-disturbance forest pixels (selected from protected areas and nature reserves).
   • Predictors included climate variables (e.g., monthly temperature range), soil properties, topography, distance to protected areas, and tree cover. The model achieved an $R^2$ of 0.39 in spatial block cross-validation.

3. Coupled Growth-Curve Framework:

   • The framework couples the stratum-specific growth curve shape with the pixel-level $B_{eco}$ ceiling.
   • A sensitivity analysis compared a standard model ($M_0$, where only the asymptote varies with $B_{eco}$) against an alternative model ($M_1$, where growth rates also covary with $B_{eco}$). The study found that $M_1$ offered negligible improvement (<5% RMSE reduction) for most strata, justifying the use of the more parsimonious $M_0$ for national application.

4. Counterfactual Scenario Projection:

   • Annual counterfactual trajectories were projected from 2000 to 2020 (with 20-year horizons) under three scenarios:
     • BAU (Business as Usual): Unconstrained continued growth without harvest or biomass reset.
     • S1 (Management-Constrained): Continuation of historical management rotations (e.g., 15-year short-rotation, 60-year long-rotation) and biomass resets.
     • S2 (Observed Loss): S1 plus empirically estimated probabilities and severities of biomass-loss events derived from the China Forest Aboveground Biomass Dynamic (CFATD) time series. Note: These losses are identified by biomass change, not attributed to specific drivers (e.g., fire vs. harvest).

Key Results

• Baseline Divergence: The unconstrained BAU scenario consistently produced higher baselines than the management-constrained scenarios (S1 and S2). The national mean difference between BAU and S1 increased from 9.98 Mg ha⁻¹ at 10 years to 14.40 Mg ha⁻¹ at 20 years (for a 2000 start). For a 2020 start, this gap reached 29.30 Mg ha⁻¹ at 10 years and 36.99 Mg ha⁻¹ at 20 years.
• Drivers of Bias: The divergence was most pronounced in short-rotation plantations, oil palm, and agroforestry systems. Planted forests showed significantly larger BAU–S1 gaps than natural forests. Spatially, the largest discrepancies occurred in southern provinces with high growth potential and intensive management (e.g., Hainan, Guangdong, Guangxi, Fujian).
• Ecological Strata Contribution: A small number of strata dominated the national difference. The top six ecological strata (including natural and planted Masson pine, Chinese fir, and typical evergreen broad-leaved forests) contributed over 70% of the total national BAU–S1 difference.
• Impact of Loss Events: Incorporating observed biomass-loss events (S2) had a minimal effect on national averages compared to S1 (differences < 0.22 Mg ha⁻¹), though it altered local trajectories and timing of declines.
• Uncertainty Propagation: Uncertainty in the baseline projections was dominated by the attainable biomass ceiling ($B_{eco}$), which accounted for the vast majority of the variance. Uncertainty related to loss-event parameters (probability and retention) was negligible at the national scale (<0.04% contribution to interval width), though it was slightly higher in specific sensitive strata.

Significance and Claims
The paper claims to provide a "national reference for baseline setting" that enables accurate and credible quantification of climate mitigation outcomes for IFM projects in China. The framework's significance lies in its ability to:

1. Limit Developer Discretion: By using external, independent data (national forest inventories and remote sensing) rather than project-specific reporting, the framework reduces the risk of systematic over-crediting.
2. Ensure Auditability and Updateability: The database is built on consistent rules and can be updated as new inventory data or biomass products become available.
3. Represent Historical Continuity: Unlike static baselines, this dynamic framework explicitly incorporates historical management rotations and observed loss regimes, preventing the misclassification of natural recovery or ongoing carbon accumulation as project additionality.
4. Support Policy Development: The authors posit that this framework can serve as a foundational module for developing IFM methodologies in China, addressing current issues of fragmented rules and insufficient risk prevention.

The authors maintain a modest stance, noting that while the framework provides a robust counterfactual biomass module, it does not resolve all carbon crediting challenges such as leakage, permanence, harvested wood products, or soil carbon. They emphasize that the framework is a tool for transparent assessment of additionality, not a complete solution for all carbon pool accounting.