Update methodology for avoided deforestation emissions based on Brazil team feedback

[robbieorvis]

Current methodology to calculate avoided deforestation emissions is to take the amount of land year affected in a single year, and cumulate this over time based on the policy setting. Then we multiply by the avoided emissions per unit land area of the land area affected. In other words, if our input data suggested that the potential land area in a given year that could experience avoided deforestation is 1,000,000 hectares, a policy value of 1 would add an additional 1,000,000 hectares to the total each year, and after 10 years we would have 10,000,000 hectares of land with avoided deforestation.

Based on our conversations with our Brazil partners, this approach is wrong for avoided deforestation in particular, though it may be correct for other policies liked afforestation/reforestation and improved land management, where land changes are carried forward in future years and continue to sequester.

In conversations with the Brazil team, our understanding is that avoided deforestation is considered a one-time single year emissions event. In other words, if we avoid deforesting 1,000,000 hectares this year, we can't just assume that we are avoiding that in perpetuity. Emissions are modeled as a single year flow of the area that is avoiding deforestation.

So, it seems to reason that for this particular policy, we should not be cumulating land area, but should instead be taking annual values to represent the flow, not the stock, of land area affected.

[jrissman]

The avoided deforestation lever we have in the model represents permanent protection of a unit of land, not just an avoidance of a one-time deforestation event. Because the protection is permanent, the protected forest continues to sequester CO2 year after year, as it grows. None of that sequestration would have happened if the forest had been turned into farmland in year 1. (This is true at least until the forest becomes fully mature, old-growth, carbon-neutral forest; we currently assume that protected forest does not reach this carbon-neutral state within the model run timeframe.)

I do not think modeling a one-time avoided CO2 release makes sense for a permanent conversion of forest to non-forest land (often farmland or pasture), because the effects of that conversion are ongoing and eternal, and these ongoing effects, cumulated across time, are more important than the one-time release of CO2 from the harvested trees. In fact, that CO2 release may be delayed, because some of the harvested trees might go into building products such as structural lumber that won't burn or decay for many years.

Perhaps the new proposed methodology misses the forest for the trees! :-D

Or if the issue is that we're talking about removing old-growth, carbon-neutral forests in Brazil, so there simply is no ongoing CO2 sequestration benefit from protecting them (i.e. the only benefit is the avoided one-time CO2 release), we might need to consider this a different policy. That is not usual among forests in countries we have modeled, and we don't want to ruin our existing policy by changing it to be particular to Brazil's situation.

[jrissman]

We might be able to keep it as one policy lever if we add an input data variable that specifies the fraction of avoided deforestation that is "carbon-neutral old growth" and the fraction that is younger forest that is still net sequestering carbon.

[jrissman]

I think the most thorough approach here, which keeps it at just one policy lever, would be:

• Add an input variable that says the percentage of forest subject to deforestation in the BAU case that is "old growth" (i.e. is carbon-neutral if left alone) and the percentage of forest subject to deforestation in the BAU case that is not old growth (sequesters carbon every year if left alone).

• Add an input variable that specifies the amount of CO2 emitted as a one-time event from deforesting one unit of land area.

• Add an input variable that specifies the share of CO2 in removed wood that has not yet reached the atmosphere after a particular number of years. If deforestation is done by wildfire, then we can assume all the CO2 reaches the atmosphere immediately. If it is done by timber harvesting, some percentage reaches the air within the first year or so (say, leaves and other plant waste, or harvested wood burned as biomass fuel), some lasts a few years (maybe paper or cardboard), and some lasts longer (furniture, building materials). This might look like a time-series input data file that starts at 100% and declines from there, showing the remaining share of CO2 that is still captured in the wood after that many years go by (assuming the deforestation happened in year 1 - we will time-shift the file to account for deforestation in each model year).

• We keep the existing variables that allow us to specify how much carbon is sequestered per year per unit land area from a forest that is protected from deforestation, if it would have been removed in the BAU case. This only applies to the young forest, not the old growth forest.

This way we capture the emissions from the plant materials removed from the forest or burned, as well as the loss of ongoing sequestration benefits, but only for non-old-growth forest that has ongoing sequestration benefits. And the time delay on the release of CO2 from harvested wood will realistically spread out the emissions (or can be set to zero, if you need to model immediate release, as in the case of burning down the forest).