Uncertainty

We deal with uncertainty all the time. In the simplest sense, “uncertainty” means we don’t have 100% of the information about how a system works. We almost never have all the information about anything, and so we constantly make decisions in the face of uncertainty.

One result of uncertainty is risk. Risk, defined very simply, means that things might sometimes go wrong, or that there might be negative consequences to a given decision or situation. How much risk we face depends in part on how uncertain we are.

There are many different types of risk, from disaster risk to economic risk to health risk. Like uncertainty, we live with risk all the time: everyday activities like driving or walking or swimming involve risk. In other words, risk and uncertainty are a normal parts of everyday life.

How do we, as society, live with the constant threat of risk? Simple: we take proactive steps that either reduce risk or increase the capacity to cope with and recover from impacts when they occur. We take action in the face of uncertainty to reduce risk.

The best way to manage the risk of ongoing climate change is to prevent climate change from occurring in the first place. As we will discover here, we may never know for certain exactly how much climate change will occur in the end. But by taking action in the face of this uncertainty we can reduce risk, rather than doing the opposite: ignoring the science and allowing the risk to grow.

The Scientific Process and Climate Modeling

Science is a process in which ideas about ‘how things work’ are presented, tested with currently available data, and then adjusted or discarded if the data do not support the initial idea. It is a process that never stops assessing and reassessing previously accepted ideas, as new data is collected and ideas are refined. In other words, scientists are continually trying to prove themselves wrong and are always on the lookout for better ideas.

Uncertainty is a normal part of science, including the science of climate modeling, and climate scientists fully acknowledge that their models are not perfect, and never will be. No model—scientific, economic, or otherwise—ever is. Science can never really be 100 percent certain about any predictions it makes, especially when dealing with systems as complex as the global climate.

In the climate modelling community, teams of scientists from around the world build their models independently so that they are free to explore different ways to model the climate. The fact that models agree as much as they do about how the climate is likely to change in the coming decades is very a good sign; the level of agreement across the models, no matter how they simulate the real world, implies that the models are doing a good job of simulating the most important climate processes. Another line of evidence that the models are working properly is that when they are used to simulate the climate of the past, the results are very similar to the climate that we know actually happened.

Uncertainty and Climate Models

When it comes to understanding and modeling future climate changes, climate scientists have to contend with three main sources of uncertainty.

Natural Climate Variability

The climate system has many components, some more important than others, and each operates over different time scales. These internal components affect each other in a variety of ways; a change in one component affects another, which in turn affects another, and so on. Consequently, the average temperature of the planet is always rising and falling from year to year, and even decade by decade, by natural processes. Clearly, this presents a challenge for climate models; reliable models of the climate system should be able to simulate how the system varies by natural processes.

Climate Model Uncertainty

Climate models are very complicated computer programs that simulate atmospheric processes and their interactions with the surface of the Earth. They are most often used to assess how climates around the world are likely to change in the coming years if we continue to increase the concentration of greenhouse gases in the atmosphere.

Although these models include very large numbers of variables and use sophisticated equations to simulate how the climate changes over time, they remain simplifications of the real world. Furthermore, even though they are run on the most powerful super-computers, they cannot simulate conditions everywhere within the climate system; they can only produce output representing conditions at a relatively low resolution (that is, they produce data points that can be quite far apart).

Models are getting better and better as modelers continue to improve how they simulate climate processes, and as computers get more powerful, but the models are, and always will be, simulations that are not perfect. Furthermore, each climate model is somewhat different, depending on how the modelers have chosen to represent processes and their interactions. Consequently, even when the models are given essentially the same inputs, the results will vary.

The level of uncertainty in climate model projections resulting from the models themselves is much larger than that associated with natural variability.

Future Emissions Uncertainty

By far the largest amount of uncertainty in climate model projections arises from us not knowing what the concentration of greenhouse gases in the atmosphere will be in the future. When the climate models are asked to figure out how the climate is likely to change if we do not curtail our greenhouse gas emissions, they all indicate that the climate will be much warmer and very different overall. If they are asked to figure out how it will change if we dramatically reduce our emissions, the warming and changes are much reduced.

So, when the models simulate climate changes for very different emissions scenarios (or Representative Concentration Pathways, as they are called), the range of outcomes is very large. The diagram below clearly shows that the uncertainty related to emissions scenarios is much larger than that from climate model (response) uncertainty and natural variability (which here is assumed to be constant over time).

Representation of the relative importance of the main sources of uncertainty in the modelling of global average temperature change. The grey zone depicts uncertainty in estimates of the global average temperature during an historical period; the black line shows the observed global average temperature that was observed. [1]

Acknowledging Uncertainty

As noted above, uncertainty is a normal characteristic of the modeling process, especially when the models are simulating something as complex as the global climate. Consequently, it is normal and expected that the models used to produce climate change projections for the future are somewhat different. Therefore, it is also normal and expected that reports such as those produced by the Intergovernmental Panel on Climate Change (IPCC) report a range of projected changes.

For example, the IPCC assessment published in 2014 reported that the global average surface temperature for the period 2081-2100 is projected to likely be 2.6°C to 4.8°C warmer than that observed during the period 1986-2005 (assuming that global carbon emissions continue to be very large) [2].

The stated range of temperature change is the result of a statistical analysis of many climate model results (all using the same assumptions about how humans will affect the atmosphere), and represents the 5 to 95% range of outcomes; that is, the conclusion is that there is a 5% chance that the warming would be less than 2.6°C, and there is a 95% chance that the warming would be less than 4.8°C. The word likely in the statement about the range of warming is an assessment, by experts, of the likelihood of the stated outcome being ‘true’, given the level of uncertainty associated with the models. In the 2014 IPCC assessment, likely means a 66 to 100% likelihood of the stated outcome.

Furthermore, another set of words (e.g., very high, high, medium) is used to identify the level of confidence that experts have in the validity of the reported results. The confidence level is affected by the amount, quality and consistency of evidence, and the level of agreement across the models. The confidence in the range of warming noted above was rated as high.

Yes, all of this gets is complicated, but it is important that climate scientists report that there is uncertainty associated with their projections and make clear how much uncertainty there is, and why. That is the goal of this very careful and precise, if technical, use of confidence and likelihood.

Can we rely on model projections?

Models are not perfect, and never will be. But the climate models used to estimate how the climate will change in the coming decades do a very good job of depicting the current climate, and climates of the past, which is a very good indication that we should pay attention to what they say about future climates. Each of the models approximates the climate system and its many moving parts in slightly different ways and, not surprisingly, produces slightly different results.

However, all of the models tell essentially the same story: the climate system is going to continue to change in important ways, especially if we continue to emit large amounts of greenhouse gases into the atmosphere. That is, the climate models are all warning us that severe climate changes are likely to happen if we do not reduce our emissions soon. It would be foolish for us to not pay attention to what the models are projecting; there is great risk in ignoring what they tell us about how we are changing the climate.

References

1. Intergovernmental Panel on Climate Change. Climate Change 2013: The Physical Science Basis.
2. Cubasch, U. et al. “Chapter 1: Introduction.” In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.