Climate change predictions are made using complex computer models. But how accurate can a global model be? How good a record do climate models have?
The quick answer
We cannot expect global models to model the complex global climate precisely. But all models have predicted global warming. Actual temperatures have been within the predicted range for almost all models.
Climate models were initially much simpler and approximate than they now are. Some early models predicted warming to 2020 accurately, others didn’t; some over-estimated, and some underestimated.
More recent models, especially those used in the Intergovernmental Panel on Climate Change (IPCC) reports, have been more reliable and more accurate. The models can and will be improved, but they are quite sufficient to show the need for urgent action.
Critics generally have an agenda and some interest to protect. They generally use fallacious arguments, cherry-picking the data they use and making unsupported allegations. We can safely dismiss their claims.
If you’d like more detail, please read on.
Analysis of the claims
These are basically the same models as are used for 7-day weather forecasts, as shown in this CSIRO video.
Model predictions can be tested. The accuracy of the prediction depends on the accuracy of the model and assumptions about how much the world will, or won’t, take action to reduce greenhouse gas emissions.
Most models have a range of predictions, to account for different levels of emissions as well as model accuracy. In this Fact Check I have focused on temperature predictions.
Several recent papers (e.g. NASA’s Study Confirms Climate Models are Getting Future Warming Projections Right) have shown that most models have been reasonably reliable in their predictions.
10 out of 17 models gave accurate predictions while another 4 gave reliable predictions if actual emissions were substituted for assumed emissions.
Critics (e.g. the Heartland Institute and the Cato Institute) have argued that the models haven’t been as reliable as claimed, and have generally over-stated the degree of warming actually experienced.
Review of the critics
On initial reading, the criticisms I have read don’t inspire confidence.
Poisoning the well
The Cato and Heartland reports resort to denigrating the motives of climate scientists. Heartland suggests they exaggerate warming to get more government funding. No evidence is offered for this claim.
This is the logical fallacy of Poisoning the well, where the integrity of the messenger is attacked instead of refuting the message. It is more than ironic that these institutes are funded by, and support, the fossil fuel industry. It seems much more likely that this connection might lead to bias in their work than that the work of thousands of scientists in scores of countries is somehow a conspiracy only visible to these fossil fuel apologists.
Cherry picking and lack of data
The Heartland report presents very little data and no significant references, to support its criticism of the models. It uses spurious logic, such as arguing that a recent sharper temperature rise can be “explained” and the previous lower temperature rise is the norm. In reality, both form part of the record, and models (and criticisms) have to account for both.
The Cato report alleges, based on this recent report, that current climate modellers “ignore decades of scientific best‐practices that would offer more accurate predictions”. But while the report suggests improvements to how model outcomes are assessed, it doesn’t make the criticism that Cato claims. Cato also ignores that the report strongly supports the consensus on climate change (which Cato doesn’t support).
Cato then mentions favourably work by a scientist who doesn’t represent the scientific consensus while ignoring the large majority view. Cato is thus using the fallacy of cherry picking. The Hoover Institute also references the minority climate sceptics while not giving weight to the massive consensus.
Examining the data
But seeing these problems with the critics doesn’t answer our question. For that, we need to look at the actual models and how they predicted subsequent temperature rises.
I went back to the 5 assessment reports of the IPCC, and found that the actual global temperatures from 2000 to 2019 always fell within the predicted range.
Carbon Brief (Analysis: How well have climate models projected global warming?) examined models in greater detail. It found that three of the five IPCC reports over-predicted temperatures and two under-predicted, but the average was almost exactly spot-on. Two out of three earlier reports also over-predicted.
|Model||Over or under prediction|
|Hansen et al 1981||-20%|
|Hansen et al 1988||+30%|
|IPCC #1 1990||+17%|
|IPCC #2 1995||-28%|
|IPCC #3 2001||-14%|
|IPCC #5 2013||+16% / +9%|
Overall, the model predictions were good, though variable, certainly good enough to confirm the extent of the problem of global warming and climate change.
It is generally admitted that while climate models predict global trends well, they still struggle with local predictions. This is to be expected, and will hopefully be improved with time. But our understanding of the global situation is reliable, sufficient to take necessary action.
Some climate model critics (for instance this article by the Hoover Institute) focus on these uncertainties.
- The models are improving.
- They have always shown the rising trend.
- There is enough information and confirmation to justify action NOW.
Thus it is fair to say that the critics cannot be trusted on this matter.
- Evaluating the Performance of Past Climate Model Projections. Zeke Hausfather, Henri F. Drake, Tristan Abbott, and Gavin A. Schmidt. AGU100, 2019
- Analysis: How well have climate models projected global warming? Carbon Brief, 2017.
- Study Confirms Climate Models are Getting Future Warming Projections Right. NASA, 2020.
- What Are Climate Models and How Accurate Are They? Columbia University.
- Critical assessments by Heartland Institute, Hoover Institute and the Cato Institute.
Graphic: Climate model schematic by NOAA Climate.gov