How different are RCP8.5 and RCP4.5 climate change impacts?
A starting point for thinking about how our understanding of climate change impacts has changed (or not changed) with the retirement of RCP8.5.
Casual observers of the ongoing RCP8.5 debate should resist false dichotomies (which social media—and sometimes traditional media—love to promote).
For example, climate scientists are right to note that RCP8.5 being retired is an example of science updating as it should. But critics are also right to note that retreating from RCP8.5 took too long; there were too many false claims made about it along the way (e.g. mislabeling it “business as usual”); and there has been too much revisionist history after the fact (e.g., the narrative that climate policy is the main reason it’s implausible). For a nuanced discussion, check out the recording below of Andy Revkin’s podcast with Zeke Hausfather, Roger Pielke Jr, Richard Tol, and myself.
The most important false dichotomy to reject is this one:
RCP8.5 being retired changes nothing about what you’ve heard on the news for the past 15 years about climate change impacts.
vs.
RCP8.5 being retired changes everything and climate change is actually no problem anymore.
Both of these claims are false.
To crudely illustrate this, I asked Claude Opus to help me compare projected impacts between RCP8.5 and a more plausible scenario, RCP4.5. As the figure below (which GCR readers have seen before) roughly shows, headline findings from climate impact studies are often produced by RCP8.5 (or SSP5-8.5) projections, whereas RCP4.5 (or SSP2-4.5, for emissions at least) is a more realistic “business as usual” scenario.
I asked Claude to calculate elasticities of various climate change impacts to temperature increase:
I specifically asked Claude to calculate the elasticities to temperature increase of climate change impacts discussed in the latest IPCC Working Group II report (AR6). For example, an elasticity of 2 (0.5) would imply that a 1% increase in the global average surface temperature (above pre-industrial temperatures) results in a 2% (0.5%) increase in the climate change impact. An elasticity of 1 would imply a linear relationship between temperature and the impact.
RCP/SSP2-4.5 projects roughly 40% less warming than RCP/SSP5-8.5 in 2100 (with uncertainty bounds around both scenarios). The difference is smaller (~20%) in 2050. See the top-left panel in the figure below from the latest IPCC Working Group I report (AR6). So, an impact with an elasticity of 1 would be roughly 40% less severe in RCP/SSP2-4.5, compared to RCP/SSP5-8.5, in 2100, and 20% less severe in 2050. With an elasticity of 2, the impact would be ~64% less severe in 2100 and ~36% less severe in 2050. With an elasticity of 0.5, it would be ~23% less severe in 2100 and ~10% less severe in 2050. With an elasticity of 1.5, it would be ~54% less severe in 2100 and ~28% less severe in 2050. (See footnote1 for details.)
The point being: the higher the elasticity, the greater the difference between RCP/SSP5-8.5 and RCP/SSP2-4.5 impacts, and the more we should be revising down our expectations of these impacts today, compared to the RCP/SSP5-8.5-dominated narrative of the recent past. The converse also applies. The lower the elasticity, the less we should revise down our understanding of the likely 21st-century impacts. (And the world also doesn’t end in 2100, so we also keep warming if we haven’t gotten to net-zero emissions by then.)
Claude’s estimate of the elasticities:
While Claude’s numbers look at least in the right ballpark for the areas I know well, I encourage interested readers to check these numbers yourself. Here is what Claude came up with:
Here is Claude’s qualitative summary:
“If you want a single answer: the empirical economic literature centers on elasticity ~1 (linear); the theoretical-IAM literature centers on elasticity ~2 (quadratic); the physical impact literature centers on 0.5–1.5 depending on the metric and time horizon.
The honest answer is that “typical elasticity” depends critically on what you’re computing damages over. For end-of-century snapshot impacts on most metrics, an elasticity of 1.0–1.5 is reasonable. For integrated lifetime damages, multi-century outcomes, or anything involving the upper tails of climate sensitivity, the implicit elasticity is much higher — 2 to 3 — because both the cumulative time spent at high temperatures and the share of tail risks that materialize scale super-linearly with the scenario.”
What does this mean?
Assuming these numbers are approximately right, here are my takeaways:
This doesn’t much change our understanding of what impacts we can expect from now until 2050. (Unfortunately, we also don’t have much control over these near-to-medium-term impacts at this point, so we need to pay more attention to adaptation.)
Our expectations of a few other severe longer-term impacts may be similar, too. For example, coral reefs may be broadly in trouble above 2C, which includes both 4.5 and 8.5 scenarios. (I believe there is some debate about this in the more recent literature, but the previous sentence represents my understanding of the consensus as of the latest IPCC report.)
In areas with lower elasticities (e.g., 0.5 or lower), scenario revisions shouldn’t change our understanding of the projected impacts much, but climate change also may not be the main driver of these impacts (which contributes to the low elasticity). For example, an elasticity of 0.4 for species extinction means that lowering the 2100 temperature projection by ~40% (from RCP/SSP5-8.5 to RCP/SSP2-4.5) should reduce our projection of extinctions by ~19%. However, habitat loss—not climate change—is the main driver of global extinction risk. Similarly, human factors (fuel, buildings in the wildland-urban interface, forest management, etc.) affect wildfire risk (0.2 elasticity for wildfire area) more than climate change, and managing these better can offset RCP/SSP2-4.5 risk increases in many (if not most) places.
In areas with higher elasticities (e.g., 1.5 or greater), we should be revising down our understandings of projected impacts substantially, at least for the end of the century. For example, with an elasticity of 2 (as some economic models assume), the 2100 impacts would be ~64% smaller under RCP/SSP2-4.5 compared to RCP/SSP5-8.5. Impacts would be ~54% smaller with a 1.5 elasticity (e.g., for cropland suitability). These are pretty big differences.
It’s very important to remember that the specific numbers above come with large error bars—in some cases much larger than we might think from reading one or two studies. Finbar Curtin and I have recently shown how this applies to economic damage estimates.
It’s also important to remember that many impact studies whose headline findings are based on RCP8.5 could easily produce—or already did produce—projections with RCP4.5. For example, if someone makes a model of how fish move in response to temperature and projects their model forward with RCP8.5, their model is still useful. They can easily (and should, if they haven’t already) re-run it with RCP4.5.
So, we shouldn’t throw out—or cast a pall over—all climate impact studies that use RCP8.5. Major assessments (and journalism) should summarize them more carefully going forward, though.
All things considered, the retirement of RCP8.5 is a major event in climate change science and it does substantially revise down our projections of some impacts. But it does not change the fact that climate change is real and comes with worrying impacts that we should want to work to avoid and adapt to, as necessary. (It also doesn’t change the fact that renewables and batteries are cheap and getting cheaper.)
The precise way to make this calculation—assuming a constant elasticity—sets impact (I) equal to a constant times T^e, where T is temperature above pre-industrial and e is the elasticity. If we standardize I by setting RCP8.5 impacts equal to 1, and RCP8.5 temperatures (T) equal to 1 in the relevant year—where RCP4.5 temperatures are assumed to be 60% as high in 2100 and 80% as high in 2050—then (0.6)^e gives us our 2100 RCP4.5 impacts, as a fraction of RCP8.5 impacts; and (0.8)^e gives us our 2050 RCP4.5 impacts, as a fraction of RCP8.5 impacts. Again, this is a crude back-of-the-envelope calculation that assumes constant elasticities. So, don’t take my numbers above literally. They are simply meant as a device to help us think a bit more carefully about these issues than typical social-media/op-ed discourse allows.