Guest post: How not to interpret the emissions scenarios in the IPCC report

On Monday 4 April 2022, Working Group 3 (WG3) of the Intergovernmental Panel on Climate Change (IPCC) is expected to publish its latest assessment of climate change mitigation. This can be considered the definitive guide to halting global warming.

The report will be WG3’s contribution to the sixth assessment report (AR6) of the IPCC. After the assessments of the physical science and climate change impacts, the WG3 report presents the last piece of the puzzle to have a complete overview of the causes, consequences and solutions to Earth’s rising temperatures.

A key means to understanding solutions are long-term emissions scenarios that describe how society could evolve towards low-carbon futures. Such scenarios are typically created by integrated assessment models (IAMs).

Since its fifth assessment report (AR5), published in 2014, WG3 has been compiling emissions scenarios from the literature into a database that supports the assessment. These scenarios are contributed by the wider research community – rather than being chosen by the IPCC – and are made publicly available.

A similar effort took place for the scenarios underlying the IPCC’s special report on 1.5C. And the forthcoming AR6 WG3 report is expected to offer a new version of that database containing both global and regional or national emissions scenarios.

This level of transparency and data availability encourages further research by others. For example, the 1.5C report’s database has been used in research papers, media articles and policy documents alike to provide context and evidence.

However, providing convenient access to these resources and making them available for the world also opens a Pandora’s box of misinterpretation and inappropriate use.

Here, I look at some ways in which these scenarios are often being misunderstood and, as a consequence, inappropriately used or misrepresented.

No crystal ball

Scenarios can be thought of as stories of what could happen in the future. What they are not, it is important to note, are forecasts or predictions for the future.

Therefore, the scenarios in the IPCC database have no inherent predictive power – and no amount of analysis, selection or staring at spaghetti plots will change that. They represent the outcome of “what-if?” thought experiments – some of which may inform questions that others are interested in, but in most cases they do not.

A good example here is the eternal question of where the world is heading in terms of greenhouse gas emissions. A scenario of future emissions could be based on, for example, current policies, climate pledges or historical trends with an assumption of no additional policies at all.

All of these scenarios tell us something interesting about what future emissions could look like, but none of them answers the question of where the world is actually heading. 

They are also imperfect. Assumptions contained in scenarios are often several years old, meaning that the policy and geopolitical context today can be very different from the time when the scenario was created.

Not a complete perspective

It is important to note that the IPCC scenario database does not perfectly encompass all potential futures. Instead, the database should be considered an “ensemble of opportunity” – that is, it was not designed to be a single, coherent collection.

In fact, the variety and number of scenarios that are found in the IPCC scenario database are coincidental. This is because the database was created by inviting modelling teams from around the world to contribute scenario data they wish to share.

This collection of data, therefore, represents in essence an arbitrary collection of pathways that researchers were able to model to answer questions they at some point cared to ask. This is neither an exhaustive set of questions nor a systematic approach to exploring the answers to them. Questions other stakeholders are trying to answer might well not be covered by this.

For example, the AR5 scenario database contained no scenarios that limited warming to 1.5C. This might be easily misinterpreted as it being impossible to devise ways to achieve this most stringent climate goal. This has already been proven wrong. In this case, modelling teams had simply focused their efforts on exploring questions surrounding how we limit warming to 2C. Apart from a few individual papers, the question of 1.5C was not covered by the research community until the UN Framework Convention on Climate Change (UNFCCC) requested that the IPCC produce a special report, which it did in 2018.

It requires careful assessment of the underlying scenario studies to understand what the absence of a specific scenario means. In some cases, the absence of a scenario does communicate information. For example, studies using the structured scenario framework of the Shared Socioeconomic Pathways (SSPs) have shown that when a very unequal and non-cooperative world is assumed, no scenarios can be modelled that keep warming to either 1.5C or 2C.

Equally, scenarios are created with comprehensive models that nevertheless only capture part of our realities. Many of the local challenges, barriers, opportunities, or other things society might care about – such as food security or reducing inequalities – are not part of the problem a scenario tries to solve. Using scenario characteristics – be it carbon prices, shares of renewables or energy efficiency improvements – to inform real-world policy should, therefore, always be accompanied by an understanding of the strengths and limitations of the tools that created them.  

Scenario distribution

In addition to not encompassing the full range of futures, scenario databases also cannot tell us the likelihood of any particular scenario becoming reality. 

In other words, mere availability of scenarios in the database says little about their likelihood. Equally so, the distribution across scenarios is not a measure of probability.  

For example, if 10 modelling teams submit 2C scenarios that reduce emissions without considering behavioural changes, and one team submits a scenario that does, this has no bearing on how likely, possible or preferable one strategy is compared to the other. Similar examples exist for scenarios with or without bioenergy with carbon capture and storage (BECCS) or other mitigation strategies.

Using data to create and correctly interpret scenario ranges requires a careful assessment of the underlying studies.

Equally, also averaging across baselines to determine what the most likely future reference emissions would be is not useful. This would be akin to determining the world’s most common animals by asking 100 of the world’s smartest children to submit their favourite animal and then averaging it. Scenario assumptions do not include the probabilities of how plausible they are.

Informed picks and targeted discussion

Having covered everything that scenarios cannot do, it is necessary to stress that they are incredibly useful – if handled in the correct way.

Using scenarios correctly involves selecting key scenarios carefully and understanding the assumptions behind them – ideally, presenting scenarios that illustrate contrasting implications of the choices available for decision makers.

A good example are the four illustrative pathways of the IPCC’s 1.5C report, which show four different pathways that return warming below 1.5C by 2100, but with very different levels of reliance on carbon dioxide removal.

However, here pitfalls exist as well. For example, the use of nuclear power in all these illustrative pathways has led some to believe that nuclear power is always a necessary part of the mitigation solution. Given that this was not the aim of this specific selection of illustrative pathways, these scenarios cannot answer the question of whether nuclear power is required for 1.5C.

When the 1.5C report scenario database was published in 2018, a group of researchers and I published an accompanying commentary, which picked out five “dos and don’ts” for carrying out analysis of the scenario ensemble. 

Given the overlap with the points I have made here, it seems apt to close with the same list: 

• Don’t interpret the scenario ensemble as a statistical sample or in terms of likelihood/agreement in the literature;
• Don’t focus only on the medians, but consider the full range of the scenario set;
• Don’t cherry-pick individual scenarios to make general conclusions;
• Don’t overinterpret scenario results and don’t venture too far from the original research focus; and
• Don’t conclude that the absence of a particular scenario means that this scenario is not possible.

IPCC scenario databases provide an unparalleled service to the scientific community. And with the latest scenarios collated for AR6, they can become a central resource that catalyses further research and informs climate policy.