ncx-gitbot
commented
2 years ago NCX response: We appreciate comments noting that the structure and performance of the baseline model used within this methodology is strongly influential on the predicted and realized climate impact of projects. Our revised methodology increases transparency rather than following an expert review process. This includes both detailed documentation of particular models used, as well as sharing benchmarking and performance information for baseline models. Finally, the revised approach to uncertainty explicitly accounts for imprecision in the baseline model in calculating the final number of credits generated from projects developed under this methodology.
Commenter Organization: The Nature Conservancy
Commenter: Ethan Belair
2021 Deferred Harvest Methodology Section: No Section Indicated
Comment: This methodology is a distinctly new approach and does not bear resemblance to previous methodologies. As such, I think a significan burden of proof lies with the authors to ensure that, not only is a methodology reliable, but it is described in sufficient detail to allow readers, users, and members of the market place to understand the basic assumptions and methods well enough to have an intuitive sense of the quality and reliability of the overall methodology. This is espcially important given the recent criticism of carbon markets generally.
I think the approach here of describing a general model approach as apposed to a specific model with parameters and covariates specified keeps the entire methodolgoy ambiguous and opens Verra and their market up to significant risk. The general model described seems highly refined and has the possibility of producing highly accurate estimates of harvest intensity and timing. However, the information provided is insufficient to allow a true review of the model. Rather, we must review the general structure and then trust to the expertise of NCX staff to produce models which claim not to be just the best in the business, but which claim to predict with precision things that others struggle to predict at all. I think in order for that claim to pass the "sniff test", we need to see a more detailed description and explaantion of this model.
Proposed Change: I propose that NCX provide AT LEAST a subregion example where the detail specific model structure, a full list of predictors, approximate coefficients and forms. Further, there should be a proof of concept example that uses this model to predict harvest timing and intensity of previous timesteps and evaluates model performance.