Endogenously calculate process emissions

[robbieorvis]

Now that the model tracks output of the main manufacturing industries, we should consider to endogenously calculating process emissions, which would free us from relying on updates to the Non-CO2 report that fail to account for updated GDP projections. EPA actually takes this approach with estimating annual GHG emissions from industries: they use historical emissions rate per ton or dollar of output and then scale this based on newer data on output/production. We could easily do this for most process emissions.

For others that are less directly tied to output, we could develop some other methodologies:

1) For natural gas systems, we could use a leakage rate 2) For HFCs, we could build in a simplified consumption-emissions model like I've built for China (most accurate and best representation of policies like Kigali). 3) For waste, could scale by population or develop a landfilling model (EPA has several publicly available versions), etc...

We'd need to find a way to modify PERAC, but we could probably do this by embedding the scaling methodology we use within Vensim instead of doing all the work in Excel. This has some appeal since it would be much easier than the current approach in Excel and all the modifications could happen within Vensim.

[jrissman]

Partners we have worked with in the past (I'm thinking of WRI China here) have sometimes had their own process emissions inventories that they wanted to use. This requires the ability to enter arbitrary process emissions via an input data file.

The EPA non-CO2 report is not so easily discarded, because it underlies PERAC as well as BPE. Even partners that had their own process emissions projections never had their own marginal abatement cost curves. As far as I'm aware, the EPA non-CO2 report is the only source for this type of information on a country-by-country basis with global coverage.

This led to the development of a multiplier system that keeps BPE and PERAC synced up. If you have your own data you want to use for BPE (i.e., because you think the EPA non-CO2 report's GDP projections are outdated), you can use your own data for BPE. In BPE, you calculate multipliers (ratios between your data and the EPA data), and use those ratios in PERAC. It is a refined system that has gone through many iterations and a great deal of work, and has been producing reasonable results for us for years.

Yes, Vensim could take a historical year of process emissions and scale up by output, presumably using data from the existing variable BPCiObIC BAU Percent Change in Output by ISIC Code. But that doesn't simplify things because you'd still need to obtain the multipliers between your data and the EPA non-CO2 report's data, and then use them in the PERAC data files. So you'd be doing the same calculations in Excel anyway, and this would introduce the risk that what you do in Excel doesn't exactly match what Vensim does internally. Rather than reducing complexity or saving time, it just adds another potential point where bugs can come in.

You might think to eliminate the multipliers by also scaling PERAC by output (instead of by the multipliers generated in BPE). Unfortunately, that approach is invalid because the non-CO2 report behind PERAC builds in changing technological availability over time, so scaling up the early year values by output would dramatically under-count available mitigation potential.

The existing BPE/PERAC system already is designed to support what you need. All you need to do is to take a historical year of data for process emissions and scale it forward in Excel based on output projections. Scaling historical data based on output is no great burden in Excel, and we do it in a number of other variables. Then you take the ratios with the non-CO2 report's data to find the multipliers, and you past-as-values those multipliers into PERAC as normal (i.e. just like WRI did). I do not, at present, see a way to improve on the simplicity or polish of the system we've already developed, because of the tight linkage between BPE and PERAC, and the need to continue to use the EPA non-CO2 report for PERAC.

[mkmahajan]

As a somewhat related note on potential edits to BPE and PERAC, I'm running into the need to tweak the current multiplier set-up to use annual multipliers rather than multipliers in 5 year increments. I think this is doable, it will just lead to some lengthy formulas in PERAC where we need to interpolate between years. Here's why I think this would be necessary for the US:

• We realized that since moving to AEO 2021 for the US, which accounts for COVID impacts on service demand, we had not adjusted BPE/PERAC to account for the drop in process emissions in 2020. We continued to simply take EPA's unadjusted non-CO2 data for the year 2020. I have now calculated multipliers for years 2020-2025 based on industrial output from AEO 2021, but our multipliers are only set up in 5 year increments. Using our current multiplier system would lead to a linear, 5-year recovery between 2020-2025 rather than what we actually see (a steep dip in 2020, followed by a pretty rapid recovery over the next 1-2 years).

• Given the EPA's HFC announcement last week, we'd like to account for the AIM Act in our BAU. Rhodium has good data available for emissions under the AIM Act, and their trajectory has some sharp dips in years 2024, 2029, and 2034 that follow the Kigali phase-down schedule. But again, these sudden drops would be lost in our current multiplier system. This is another place where we would benefit from annual multipliers. Note that like we've done for a SNAP adjustment in the past, we will customize the abatement potential for the US after shifting our HFC baseline in BPE (but can keep this as a lever in the Country Selector that is only enabled for the US).

Jeff - I'm happy to take a stab at this and run it by you. We're trying to move pretty fast here on the federal modeling we're doing, so I'm trying to get a file set up for our us-analysis repo now. I may have to circle back after to get the structure set up to work fully with all regions.

[jrissman]

Certainly, it would be possible to switch to annual multipliers. The current approach uses multipliers at 5-year increments because the data in PERAC are given in 5-year increments, and we did not anticipate a need to model rapid year-over-year changes near to the present day.

It should be relatively straightforward to switch to annual multipliers in the BPE and PERAC Excel files. In fact, BPE already uses annual increments in the formulas in its blue tabs (it converts the 5-year multipliers to annual multipliers), so it should be particularly easy to switch this file. Just calculate the multipliers directly in the annual section rather than in the 5-year section (and then you can delete the 5-year section). No updates to the blue tabs should be needed, and I don't think it should interact with or harm the file's customizability to different countries or regions.

PERAC is slightly more bother because, as you mention, you would need to update the interpolation formulas, so they interpolate between the values prior to applying any multipliers, and then multipliers are applied on an annual basis after the interpolation is done. Still, it doesn't sound too bad. Again, I don't see it interacting with the file's customizability to different regions.

One small thing to keep in mind is to keep the format of the multipliers table identical between BPE and PERAC, since the normal procedure is to calculate the multipliers in BPE, then use paste-as-values to transfer those multipliers to PERAC.

Thanks for offering to tackle this data-only edit. Would it make sense to do the update officially in the "develop" branch of the us-eps repo and then copy that file over to your special analysis branch for further customization, if needed? (Robbie has indicated we are ultimately going to publish this work and will want to be able to show the scenarios in the public webtool, so this feature - annual multipliers - will need to be part of an official US-EPS release anyway.)

[jrissman]

Update:

BPE and PERAC have been switched to use annual multipliers rather than 5-year multipliers.

We think it might be worthwhile to calculate process emissions endogenously, as long as there is an override BPE file for partners that already have a process emissions inventory, which they demand be used.

To keep in sync with PERAC, and to eliminate the difficult task of maintaining PERAC as an Excel file, we are interested in potentially doing more of the PERAC adjustment calculations in Vensim and fewer in Excel, but we likely still need to do some PERAC calculations in Excel. For instance, the EPA source doesn't apportion anything into price bins, whereas we both apportion into price bins and allocate technological measures to specific policy levers. Those tasks might need to continue to be done in Excel.

[robbieorvis]

I'm changing this to high priority. It will be necessary once we start to allow users to switch between different production technologies (like in iron and steel) to correctly calculate process emissions reductions.