If your work involves thinking about future climates, you’ll be familiar with the alphabet soup of climate scenarios. Have you ever wondered how scientists build them? In this technical explainer, JBA Risk Management’s Head of Science, Dr Paul Young unpacks the ingredients that make up the commonly used climate scenarios.
How will society develop? What does that mean for greenhouse gas emissions? What does that mean for our climate?
Scenarios are systematic ways to explore the “future worlds” that emerge in response to these questions. Their widespread use for long-term strategic planning across businesses, governments, militaries, and more evolved from mid-20th century ideas separately developed in the USA and France. Today, as well as being associated with climate science, scenarios are a core approach in the academic field of futures studies, where definitions, classifications, and methodologies are further developed, discussed, and debated.
Throughout the insurance, reinsurance, property, finance, and international development sectors, our clients encounter future climate scenarios through regulator-mandated stress tests, disclosure requirements, or as part of strategic planning processes, to name but a few.
For future climates, we often want to consider several scenarios and evaluate their implications. Individuals could generate many potential futures, depending on their needs, values, and expertise. But using a standard set of scenarios makes it easier to collaborate and compare impacts and implications across different studies and use cases.
The most encountered climate scenarios are those developed for, and analysed in, the IPCC reports. These are chiefly the Representative Concentration Pathways (RCPs) and the Shared Socioeconomic Pathways (SSPs), which underpinned much of the IPCC’s 5th and 6th Assessment Reports. Additionally, the Network for Greening the Financial System (NGFS) scenarios, primarily developed for central banks, offer a framework to investigate both physical and transition risks related to climate change, and are used in several regulatory contexts to guide policy and economic decisions.
For each of these scenario sets, there are several individual scenarios that span a range of projections for society, policy, greenhouse gas emissions, and the resulting climate. The nomenclature and specifications for these scenarios are detailed elsewhere (e.g., here for RCPs, here for SSPs, here for NGFS scenarios).
Here, we remind the reader of the major sets of climate scenarios and their naming conventions:
Scenarios are often related to the amount of global warming they cause, as measured against the global mean surface temperature of pre-industrial times (usually taken to be the 1850–1900 average) For instance, in policy and political circles there is a long history of the 2°C target, which preceded the 1.5°C target adopted as an aspiration in the 2015 Paris Agreement. SSP1-2.6 (or RCP2.6) and SSP1-1.9 are, in IPCC language, unlikely to exceed global warming levels of 2°C and 1.5°C, respectively. Global warming levels are explicitly built into many NGFS scenarios (e.g., “below 2°C”) and, in general, are a useful shorthand for overall physical risk.
While many in the climate risk world take “climate scenario” to mean the projection of physical climate change, this is only one aspect of these scenarios. In fact, we can think of a climate scenario as integrating three connected components: socioeconomic conditions, climate forcing, and climate change.
The socioeconomic component describes the underlying parameters of society, such as population, economic growth, urbanisation, resource availability, and technology developments. The climate forcing component describes how greenhouse gas and other pollutant emissions, as well as land use and land cover, could evolve consistently with the socioeconomic data. The climate forcing is the collection of things that drive (“force”, in the parlance) physical climate change.
The coupling of the socioeconomic and climate forcing components of the climate scenario is complex and the full details are beyond the scope of this blog. For those interested in some detail, Box 1 summarises how this was done for RCPs, SSPs, and the NGFS scenarios. In brief, the socioeconomic data are created, or chosen, to be consistent with an underlying socioeconomic narrative or storyline. These socioeconomic data are then used to generate climate forcing data with integrated assessment models (IAMs). These models represent the connected energy, economy, and environmental systems, and can be configured to simulate their evolution consistent with a target climate forcing. Figure 1 shows some socioeconomic data for the SSP narratives as well as the resulting CO2 emissions when those data are used to target certain climate forcing projections (see also Box 1).
Figure 1: From socioeconomic data to climate forcing for the SSP climate scenarios. Time series of (a) population and (b) GDP for some selected SSP narratives and (c) the resulting CO2 emissions when those narratives are used with IAMs to generate climate forcing data consistent with different mitigation targets. Note that SSP1 is used to generate two different climate forcing datasets. Data from the SSP Database.
This box sets out some of the nuts and bolts of the major scenario families: the RCPs, the SSPs, and the NGFS scenarios.
With the RCPs and SSPs, IAMs were targeted at producing climate forcing data consistent with a limited set of radiative forcing “mitigation targets”: 2.6, 4.5, 6.0, and 8.5 Wm-2 (adding 1.9 and 7.0 Wm-2 for SSPs only). To produce this climate forcing data, the IAMs simulate different pathways for variables such as energy generation and use, negative emissions, and carbon pricing, all consistent with the socioeconomic data. The “results” from an IAM simulation are different stories of policies for the economy, energy, and climate.
For RCPs, the socioeconomic data driving the IAMs come from a range of literature sources, meaning that they do not form a consistent exploration of different societal choices when taken together. However, the RCPs were primarily aimed at producing stylised climate forcing data at their “target” radiative forcing values for rapid use by GCM teams.
The SSPs build on the RCPs with five connected "socioeconomic narratives" (SSP1 to SSP5). These describe different future worlds with a range of opportunities and challenges to address climate change, which are then used to generate consistent socioeconomic data. While we often call these climate scenarios “SSPs”, they are really part of a single SSP-RCP framework, which links the SSP socioeconomic narratives with RCP mitigation targets, or climate forcing outcomes (see also Figure 1). Different SSP narratives can be targeted at a single climate forcing outcome, allowing exploration of how, for instance, ambitious climate targets can be met with different socioeconomic conditions. Additionally, the same SSP narrative can be targeted at multiple climate- forcing outcomes (e.g., see SSP1-1.9 and SSP1-2.6 in the figure below).
For the NGFS scenarios, the socioeconomic data are a mixture of real-world values for the near-term (e.g., IMF GDP projections) and data consistent with the SSP2 “middle of the road” narrative for beyond.
The climate change projection is the physical climate outcome of the emissions and land use change and is derived from the climate forcing data using climate models.
Simple climate models have already entered the scenario process prior to this point as part of the IAMs, where they connect energy use, greenhouse gas emissions, and agricultural productivity, for instance. However, for this final component we are concerned with the more detailed physical climate projections that we get from simulations by global climate models (GCMs) and other similar tools.
As noted above, it is the output from these simulations that usually characterise a “climate scenario” in the world of physical climate change risk. Rather than the evolution of the energy mix and carbon price under SSP2-4.5, for example, one might ask about that scenario’s change in extreme rainfall or temperature and how that changes the physical risks for a portfolio of assets.
While physical climate risk assessments take the climate change output from climate scenarios, they seldom use output directly from the GCMs. This is not least because they have well-documented problems simulating extremes, among other shortcomings. At JBA Risk Management, we use statistical techniques to process climate model output and calculate return period change factors. These are used to adjust present day hazard intensities to be appropriate for future climates, including allowing for different spatial and temporal extents of extreme events.
Typically, the non-physical aspects of climate scenarios are not – and generally cannot – be fully integrated into physical climate risk assessments. For instance, climate scenarios do not provide detail on the flood defences that local authority X commissions under SSP1 versus SSP4. They also do not say whether future building codes in region Y are more stringent under SSP3 or SSP2, nor what rate old building stock is upgraded or replaced. (Of course, one can ask and answer these types of questions using our loss model – see here, for example.)
Given the physical climate focus, and the lack of non-physical risk information to inform asset-level analyses, it is as well to think of a particular scenario and time horizon in terms of its change in global mean surface temperature (GMST). From a climate science point of view, this has solid underpinnings: the change in many hazards scales well with the change in GMST.
If the primary concern is the change in GMST, different scenarios just describe different amounts and rates of climate change: larger and faster for SSP3-7.0, smaller and slower for RCP2.6, for example. Using the change in GMST, different families of scenarios and time horizons can be compared among one another. Scenarios may also be selected that correspond to a particular change in GMST or because they provide a consistent view with other risk products or an earlier analysis.
In a later blog, we shall drill down into the complex world of uncertainty and climate change. For instance, between different GCMs there is no “agreed” climate change for a particular scenario and time horizon (e.g., a 2041–2060 time slice for SSP3-7.0 might equate to GMST changes of between 1.6°C and 3.2°C above their pre-industrial level) and the same GMST change can be consistent with very different regional climate changes, with materially different outcomes of risk.
All this colours any analysis of future physical risk, especially at the asset-level scales that climate risk professionals are interested in. Yet, we can think about using the wealth of climate data in another way through storylines, which will feature in that later blog too.