When should we act to reduce our greenhouse gas emissions and tackle climate change: now, or later when we know more?
One person who thinks we should wait is New York University theoretical physicist, and former US Under Secretary of Energy for Science, Steven Koonin.
In an article published by the Wall Street Journal, and reproduced in The Australian, Koonin claims that climate models are still too uncertain and that everyone should hold their horses, arguing that:
… because the natural climate changes over decades, it will take many years to get the data needed to confidently isolate and quantify the effects of human influences.
That’s not to say that the issue isn’t pressing. But Koonin says we should urgently do science, rather than urgently cut emissions:
The science is urgent, since we could be caught flat-footed if our understanding does not improve more rapidly than the climate itself changes.
Well, yes. But we’ve been doing this “urgent science” for decades.
Bad risk management
To put it in risk-management terms, Koonin seems to be saying that the penalty for acting on climate and being wrong is greater than the penalty for not acting strongly enough.
He also claims that “there isn’t a useful consensus at the level of detail relevant to assessing human influences (on the climate)”. But this line of argument confuses the idea of consensus, and also mistakes uncertainty (in the scientific sense) for a lack of confidence in the evidence.
In fact, climate models demonstrate the opposite of what Koonin says. We are confident, despite uncertainties, that the science is saying that it’s past time to act.
Doubt, uncertainty, and confidence
When a scientist talks about uncertainty, they mean something different to what the public thinks. The public understands uncertainty as “being unsure”. But to a scientist, uncertainty refers to the spread of results gained from a set of experiments.
For example, the models used by the Intergovernmental Panel on Climate Change (IPCC) for its recent landmark Fifth Assessment Report project future temperature changes with a spread of results from 0.3C to 4.8C of warming across four different possible scenarios. But this range, wide though it seems, does not imply doubt because we are confident that the models capture the principal characteristics of how the climate will change.
A model’s output will have a certain level of uncertainty attached to it, but the level of trust that can be placed in it is called confidence. Confidence relates to how well the theory and observations stack up, and to the ability of models to represent the main real-life processes. One key test is to predict unanticipated but confirmable outcomes.
For example, in the real world we see a great deal of climate variability due to natural processes, such as El Niño/La Niña events and large volcanic eruptions. The fact that models can reproduce this variability increases our confidence in them.
But it doesn’t make the future any more certain, because climate variability is, well, variable.
Long-term forecasts of warming over the century involve a trade-off between confidence and uncertainty. To have very high confidence, it may be necessary to provide a wide band of uncertainty, but that’s not very useful for decision-making. A narrower band of uncertainty may be more useful, but will generally have lower confidence attached to it. Factoring in uncertainty about future greenhouse gas emissions makes predicting the future especially difficult.
That’s why the uncertainty guidance produced by the IPCC contains advice for managing both uncertainty and confidence.
How much more certain can we get?
If we want to know how much the world will have warmed by 2100, how much more certain can we be if we wait 10 or 20 years? Is there a benefit in waiting?
That question has two main uncertainties:
- socio-economic uncertainties associated with future levels of greenhouse gas emissions, which depend on population growth, economic activity and technological change;
- scientific uncertainties within and between climate models about how the climate system works.
We have a simple climate model setup that can be used to explore these questions. It takes the major scientific uncertainties from the complex climate models, including carbon cycle uncertainties, and estimates the future spread of mean global temperature. It then uses observations since 1900 to narrow these down.
If we wait, and add another decade or two worth of observations, uncertainty comes down slightly.
For example, the following graphic shows current uncertainty in green, what we might know in 10 years in blue, and what we might know in 20 years in brown. That would be after 2030, and given that we are still currently on a high-emission pathway, this would commit the world to more than 4C of warming and its accompanying risks, all for the sake of a limited gain in knowledge.
The graphs show the expected warming in response to two commonly used greenhouse emissions scenarios: on the left, a moderate scenario in which emissions peak by mid-century and then decline; on the right, the worst-case scenario, which sees emissions continue to grow for much of the century.
All these estimates have the same level of confidence. And the best estimate of future warming (the black line) hardly changes for this particular set of experiments (if it did, the graphs would feature diverging black lines representing differing predictions made now and in the future).
Why don’t the extra decades of observations matter? It turns out that uncertainties due to climate sensitivity, the carbon cycle and changes to energy technology, are interchangeable. They cannot be reduced in the way that Koonin is insisting upon, by building better models with lower uncertainty.
The only way to really narrow the uncertainty is to reduce greenhouse gas emissions – that is, to try and ensure that our real-life future is more like the graph on the left than the one on the right. This would narrow the spread of uncertainty much more effectively than sitting tight and doing little for a couple of decades.
The models shows that acting now and learning as we go is a better way to manage uncertainty than waiting and learning. This is the opposite conclusion to Koonin’s.
The case for climate policy now
We need climate models to get a better idea of how the climate may change at the local scale, but our ability to improve predictions of future global warming is limited.
For example, we might design a climate policy to avoid 2C of warming, and that policy might have a 25% chance of failing. Improved climate models might cause that chance to fall to 20% or rise to 35%. But that doesn’t mean the policy should be thrown out, rather than simply tweaked.
Climate policy is so far-reaching that it needs to interact with a host of other policy areas. Therefore policy uncertainty, not scientific uncertainty, becomes the most important factor.
It’s not uncertainty that’s important. It’s confidence, and we are confident that the science tells us it’s past time to act.