CSSP Product 2.1a - the 16% question

The US Climate Change Science Program (CCSP) has initiated a series of Synthesis and Assessment Products that are intended to demonstrate the state of the art in fields relevant to the climate science questions. The recent Stern Review on the economics of climate change received widespread media coverage in stating that the costs of reducing CO2 output sufficiently to prevent major global warming would be perhaps 1% of Gross World Product (GWP), while costs of allowing global warming to continue would be as high as 20% of GWP. In light of this, the CCSP Product 2.1a effort on developing scenarios is of particular interest, as at least one of the scenarios saw abatement/mitigation costs as high as 16% of GWP in the year 2100 - a number featured in a NY Times article this past week by Andy Revkin.  

So where did this 16% figure come from, and why is it so much higher than Stern's 1%?

First off, there's a very simplistic model that explains why 1% should be a reasonable number, perhaps even an overestimate.

In a $40 trillion world economy, $100 billion/year is 0.25%, well within Stern's 1% number.

Now, you can't replace every coal plant with a nuclear plant or wind farm instantly, and CO2 emissions reductions beyond eliminating coal will be harder. But will they really be orders of magnitude harder? That's what those CCSP numbers seem to say.


CCSP Product 2.1 isn't in final form yet - a second public draft was just recently posted on the Department of Energy's website. The report includes analysis of three models of the climate and economy taken to the year 2100, with reference models under no carbon-reduction constraints, and four scenarios of increasingly stringent abatement (the actual criteria for abatement are on total radiative forcing, rather than direclty on CO2 concentrations).

The three simulations were carried out independently by an MIT group (using the Integrated Global Systems Model (IGSM) together with the EPPA-4 economic model), the "Joint Global Change Research Institute" (PNNL and U. Maryland - the "MiniCAM" model), and a Stanford University/Electric Power Research Institute collaboration (the "MERGE" model). Section 3 of the draft describes the reference simulations, while section 4 discusses the constrained simulations with mitigation scenarios.

The bottom-line numbers on GWP loss due to regulation of carbon and other emissions are in table 4.8 and figure 4.26. For some reason, the assumptions of the IGSM model are leading to much higher costs than for the other two (which are between them remarkably similar) - in the "Level 1" (most stringently regulated) case, a factor of 10 higher costs by 2100.

The "carbon prices" in these scenarios, roughly equivalent to the carbon abatement cost of $27/ton above, are to be found in table 4.6, and figure 4.19. All the modeling scenarios show higher prices than our baseline as mitigation efforts get more serious later in the century, but MERGE and MiniCAM stay under about $600/ton (a factor of 20 higher than the naive coal-nuclear estimate) while IGSM/EPPA (always costlier for some reason) skyrockets off to $6000/ton in their 2100 level-1 scenario.

The order of magnitude difference in carbon prices seems to be directly related to the order of magnitude GWP reduction difference - 16% in IGSM/EPPA vs. about 1.5% for the other two. Worse, from the price curve in figure 4.19, it seems highly likely that if the abatement percentage had need of being higher than the roughly 80% IGSM found, the costs could have been an order of magnitude higher, totally devastating the world economy (assuming governments didn't react to pull back the abatement regulations).

An uncertainty analysis would surely show very large error bars on the IGSM/EPPA level-1 numbers in 2100, just based on the slope of that price curve. The question is, why does IGSM/EPPA have this "knee" in the abatement cost curve that seems absent in the other models, and that appears to prevent abatement of more than about 80% of CO2 emissions?

Is there something fundamentally missing in the EPPA analysis? The following aspects of the IGSM/EPPA model (to the extent I understand them) don't seem right:

  1. The historic shift of perhaps 1%/year from non-electric to electric energy demand is missing
  2. Carbon abatement drove down electric production in IGSM/EPPA far more than in the other models (figure 4.11) - it looks like from reference to level 1, IGSM/EPPA showed a drop of about 35% in electric production, while the other two dropped only 6-7%. Even at level 4, IGSM/EPPA had what seems to be over 20% reduction in electric, while the others were just a few percent.
  3. Even in Level 1, IGSM/EPPA shows no additional energy from non-biomass renewables (figure 4.9). I understand the restriction imposed on nuclear power in the model, but why did the non-biomass renewables piece not change, as it did in the other models?
  4. There doesn't seem to be an accounting of economies of scale or learning curves on the supply side; or perhaps it's just that the historical path-dependence of the abatement curve of figure 4.19 is missing? Shouldn't relative costs of alternatives decrease as we use them more? The historical pattern in the solar and wind industries seems to be still seeing 10-20% cost reductions for doubling of production, not fixed prices.
  5. The models show demand shifting between alternatives based on the abatement costs; however, it's not clear to me that the timescale-dependence of elasticity of demand is accounted for here (demand tends to be inelastic on short time-scales due to sunk capital, but much more elastic on longer time-scales as alternative capital investments become more economically viable). Getting to the bottom of that probably requires real familiarity with the EPPA model... but there's got to be some core reason (associated perhaps with the transportation sector modeling we've already discussed a bit) for the high inelasticity to carbon reductions beyond 80% in figure 4.19.

Understanding that core inelasticity seems to be the key to understanding the 16% question. The other differences between the models appear to be relatively small compared to this one. So what's going on here?

Created: 2006-11-04 04:57:12 by Arthur Smith
Modified: 2006-11-04 05:03:13 by Arthur Smith