Assessing forest carbon offset additionality with dynamic baselines and uncertainty quantification
Keywords: climate modeling, nearest neighbors, uncertainty quantification, counterfactuals
TL;DR: We present a dynamic baseline model improving accuracy in carbon offset predictions, revealing inaccurate emissions reduction estimates in Brazilian REDD+ projects.
Abstract: Forest carbon projects, like those reducing emissions from deforestation and degradation (REDD+), can help mitigate climate change by sequestering carbon. Their effectiveness is measured against a 'baseline' scenario, which predicts emissions if no project intervention occurred. To maintain the integrity of carbon credits, it's crucial to have accurate baseline emission reduction models with well-characterized uncertainty. However, recent scrutiny has raised concerns about the accuracy of emission reduction claims made by REDD+ projects. These projects rely on ex-ante predictions of future deforestation risk with no standard approach to quantify uncertainty. Ex-post (``dynamic'') baselines could reduce this model uncertainty, but they also lack a standardized framework for uncertainty. We introduce a dynamic baseline model based on remote sensing data and nearest neighbors matching and apply a novel uncertainty quantification framework to assess the accuracy of this model. Applying our approach to seven REDD+ case study projects in Brazil, we found several instances of consistent over/under-estimation of emissions reductions, suggesting potential inaccuracies in current carbon offset measurements. Our findings highlight the importance of our dynamic baseline and uncertainty quantification in enhancing the effectiveness of REDD+ and similar forest carbon projects.
Submission Number: 4
Loading