Background
Verra is a standards organization for climate action and sustainable development. One of its core activities is support for voluntary-sector REDD (Reducing Emissions from Deforestation and forest Degradation) projects through the Voluntary Carbon Standard (VCS) and associated registry. Verra has played a pivotal role in certifying initiatives that yield measurable, high-integrity outcomes.
Aligned with the United Nations’ Sustainable Development Goal 14, Climate Action, Clark Labs has been actively engaged in the development of support tools for REDD since 2009. Initially, Clark Labs worked with Conservation International to add a REDD analysis facility to the Land Change Modeler, based on the World Bank’s BioCarbon Fund (BioCF) Methodology for Estimating Reductions of GHG Emissions from Mosaic Deforestation. Building on this, Conservational International and Clark Labs further collaborated to develop GEOSIRIS – a national level REDD planning tool, implemented in TerrSet. This commitment to the goals of REDD in mitigating climate change now continues with a direct engagement with Verra.
A core element of REDD is the ability to predict, before the project crediting period begins, the expected deforestation over the project’s validity period. This prediction is then compared with actual deforestation to establish carbon credits and the transfer of funds from project investors. It is therefore critical that the predicted density of deforestation be as accurate as possible. Given Clark Labs’ extensive history of empirical modeling and land change prediction associated with the Land Change Modeler and GEOMOD (an earlier tool also found in TerrSet), Verra contracted with Clark Labs to review existing prototypes and then to develop a new spatially and quantitatively explicit approach. The resulting VT0007 Unplanned Deforestation Allocation Tool(UDef-A) is the product of 15 months of analysis, experimentation and testing, conducted in cooperation with TerraCarbon.
The goals of the methodology are to:
- Set a minimum standard of prediction quality by defining a simple but effective benchmark approach. It is based on the definition of modeling regions defined by the intersection of vulnerability zones and administrative divisions. These modeling regions are then used to tabulate the relative frequency of deforestation during a historical reference period.
- Establish a standard method for the incorporation of more sophisticated empirical models and a methodology for their comparative testing against the benchmark. These might include, for example, neural network models incorporating multiple explanatory variables. In cases where an alternative exceeds the quality of the benchmark for both a calibration period data fit and a test prediction, the alternative qualifies for use in a jurisdiction.
- Output a spatially explicit mapping of expected deforestation expressed in hectares/pixel/year. The output can then be summed over project extents and forest strata to estimate the emissions avoided by the project.
Principal Investigators: J. Ron Eastman (Clark Labs/Clark CGA), Gil Pontius Jr (Clark Graduate School of Geography) and Rebecca Dickson (TerraCarbon).
Resources