Annex B: Model Evaluation

Value Chain Interventions/Programs that are using models (e.g., process-based) to calculate the baseline(s) or intervention(s) scenario must fill out this Annex.

The information provided will be used to assess the modeling approach. Since no two models and no two contexts are the same, the assessment will be undertaken by an expert, recognizing that while presented stepwise, the process may require a non-linear, iterative review approach. Expert assessors may also come from diverse backgrounds, experiences, and qualifications.

In this section:

The Certificate Holder (or an Intervention Representative, IR, acting on its behalf) shall fill out the Tables below
The Certificate Holder or its IR shall submit the calibration and validation data set(s) as links (URLs) or attachments (submitted documents)

Table 4: General Information of Scope and Quality of the Modeling Approach

Criteria		Recommended Criteria	Justification/Response
Scope
G1	Model version	A uniquely traceable record of all files needed to reproduce a given model output from its calibrated parameter set and model-driving input data. These may include source code, internal parameters that are not adjusted during calibration, default values for parameters or input data, or generally any other information that may change model behavior. A model version must change any time there is a change in any of the model files. For a given parameter set and set of model-driving inputs, any copy of the model reporting the same model version must always produce the same output.
G2	Is the modeling approach suitable for the context of the intervention, notably geography and activity?	Some modeling approaches are global / generic but most are targeted at specific activity types and geographies. Companies should not try to apply modeling approaches that do not properly account for their activity or that use approaches intended for other geographies that are different from the context of the intervention.
G3	Is the approach suitable for the reporting purpose sought?	Companies may report for many reasons, most notably for Greenhouse Gas Protocol or other reporting protocols such as GRI and CDP. The company should satisfy themselves that the method chosen conforms with any criteria stated in the reporting protocols it is intended for. Likewise, if any part of the intervention is seeking carbon credits, a methodology that is approved by the credit issuer is likely necessary.

Quality
G4	Is the scientific framework of the model third-party authored or reviewed?	It is generally not recommended that companies create their own models. Relying on well-chosen third-party models may avoid accusations of poor quality or conflicts of interest. Where self-authoring of models are unavoidable, these should be based on peer-reviewed science and reviewed and approved by a suitable third party.
G5	Is the model recent and based on up-to-date science?	It is recommended that the latest version of a model is applied. Ideally, the version would have been published within the two to three years prior to application though longer may be necessary in circumstances where this is unavoidable. It is recommended in such cases that the company highlights this transparently in reporting and explores ways to bring the method up to date.
G6	Is the modeling approach scientifically peer-reviewed?	Modeling approaches that are peer-reviewed or include scientifically peer-reviewed approaches should be chosen over those that do not.
G7	Has the model been approved by a Standards body, such as Gold Standard, CAR, VERRA?	For enhanced confidence in the model chosen, or where it is desirable to use for the purposes of certification, then the selection of a model that has been approved by a body such as Gold Standard may be beneficial.
G8	Does the company have the technical capacity and resources to properly use the model?	Companies should choose a modeling approach commensurate with their skill and capacity to implement. While more rigorous approaches are often desirable, it is better to choose an approach that can be realistically implemented than one that is unlikely to be properly maintained.
G9	Are all assumptions and the level of uncertainty/confidence in the model output generated clear and transparent?	Per the Interventions Guidance, companies should transparently report any assumptions inherent in the modeling approach chosen and the level of uncertainty in the results generated. This helps to avoid accusations of over-reporting or over-stating results where a more practical but less rigorous approach is being applied. Different sources of uncertainty should be distinguished (e.g., sample uncertainty vs. heterogeneity). Choice and form of distributions need to be set for all parameter values to reflect second-order uncertainty.