Uncertainty & Conservativeness

This section describes how ERS accounts for uncertainty and the rules enforced to ensure conservative carbon estimations.

UNCERTAINTY

1. Woody AGB Estimation

   To minimise and account for uncertainty related to AGB estimation, ERS implements best practices outlined in the Aboveground Woody Biomass Product Validation Good Practices Protocol. This implies that:

   1. AGB error estimation must be considered in the entire process, from field measurements to modelling errors, including those associated with allometric equations.
   2. The propagation of uncertainty through these various stages must be effectively managed. ERS's AGB benchmark (Appendix 1) demonstrates different methods of AGB uncertainty propagation.

2. AGB Model Uncertainty

   1. Pixel-level uncertainty. ERS uses Chloris’ model to generate AGB maps, including a pixel-level standard error for AGB density (AGBD) change estimates at the 95% confidence level. This uncertainty is derived from a Map of Standard Error, based on error propagation analysis across all layers in the time series.
      1. The standard error is calculated by considering geolocation, allometric, and model-based errors for AGBD predictions at each time point. The confidence interval (C.I.) for each pixel trajectory is then used to determine the standard error of the AGBD change. Reported AGBD change statistics are based on the sum of significant pixel-level changes (p-value ≤ 0.05).
   2. Site-level uncertainty. To estimate AGB uncertainty at the site level, ERS applies Monte Carlo simulations. This approach accounts for variability in pixel-level uncertainties, ensuring robust estimates for large datasets and when spatial correlations are present.

3. Quantification of Project Uncertainty

   1. The Monte Carlo approach involves randomly sampling AGB values at the pixel level from their respective probability density functions. These sampled values are then aggregated to calculate the overall AGB for the designated plot. Through iterative sampling, the method constructs a comprehensive probability density function, capturing site-level uncertainty with precision. The key steps are outlined below:
      1. For each pixel, a single AGB value is randomly selected from its predefined probability density function and its associated standard error, reflecting the variability inherent at the pixel level;
      2. AGB values are expanded to include BGB estimates. Both AGB and BGB are transformed into their CO2e values;
      3. The determined pixel-level GHG removals obtained are aggregated to estimate the total net GHG removals for the plot in the specific iteration. Once aggregated, deductions are made for leakage and baseline emissions from the verification cycle to derive the net GHG removals achieved during the cycle. This process ensures an accurate and conservative estimation of the project's actual contribution to GHG removal;
      4. These steps are iterated to build a comprehensive probability distribution of net GHG removal at the plot level. During the iterations, the mean net GHG removal estimate stabilises as the simulation progresses. A minimum of 500 iterations is performed to ensure robust and reliable results. More iterations may be conducted based on empirical observations.
      5. The resulting distribution represents the range of potential net GHG removal values.

CONSERVATIVENESS

The conservative approach applied by ERS consistently and systematically selects the uncertainty boundary to keep the most conservative estimates. This prevents any potential overestimation of GHG removals. In addition, uncertainty parameters calculated in the Uncertainty section are factored in. 

The following section provides details about the conservative approach taken at each step.

1. GHG Removal Capacity. The GHG Quantification methodology ensures reliability through a conservative approach using Monte Carlo simulations to model uncertainties distribution in biomass estimates. To enhance precision, ERS applies an uncertainty threshold that discounts the average biomass estimate. This adjustment yields a conservatively lower final total carbon stock value compared to using the unadjusted average from AGB data, providing a robust basis for GHG calculations.
   1. PRU Accounting. For the quantification of PRUs, the lower bound of the 95% confidence interval is chosen from the distribution generated by Monte Carlo simulations. 
   2. VRU Accounting.  For the quantification of VRUs, the lower band of the 95% confidence interval is chosen from the distribution generated by Monte Carlo simulations.
2. Leakage. The uncertainty assessment approach incorporates potential leakage within the Monte Carlo simulations. Leakage values are accounted for indirectly through the modelling of AGB growth uncertainty. During each iteration, leakage is sampled alongside other GHG components, allowing for a comprehensive calculation of the Project's overall uncertainty.
3. Biennial Quantification. The same conservative approach is applied to measure the carbon stock of the Restoration Site before every Verification. 
   1. The lower band of the 95% confidence interval is selected for Woody AGB values.
   2. The woody or non-woody biomass uncertainty, derived from equation (26), is retrieved from the biomass stock. 
4. Loss Events. An inherent challenge in assessing the impacts of loss events is determining the BGB loss through satellite imagery. ERS conservatively considers a complete loss of BGB and consequently deducts both AGB and BGB from the carbon stock quantification.