A Primer on Climate Projections and Future Climate Scenarios

This primer is excerpted from Mid-Atlantic Regional Climate Impacts Summary and Outlook: Winter 2023–2024.

Introduction

Climate models are physics-based computer simulations that represent Earth systems, such as the ocean and atmosphere, and help us see how different potential future states of the world may result in different climate-related outcomes and impacts.¹ There are dozens of groups of scientists that have developed global climate models all over the world, leading to the use of tens of different models to study global climate change.² Through the Coupled Model Intercomparison Project (CMIP) and in support of the Intergovernmental Panel on Climate Change (IPCC) assessments, every five years or so, global climate modeling groups come together under the leadership of the World Climate Research Programme (WCRP) and, among other modeling goals, agree to run their models with the same set of potential future scenarios as their inputs with the goal of compiling and comparing model results.^3,4 Comparing many models run with the same inputs leads to a broader understanding of the range of future climate changes and makes it possible to provide information about their levels of confidence and uncertainty, which is important for users to be able to interpret and act on the climate projections.⁵

Projections of global climate change from CMIP efforts are utilized by a range of scientific, practitioner and policy stakeholders from academic papers examining future tropical cyclone activity and intensity to designing infrastructure to withstand future climate conditions. But doing so requires an understanding of the terminology, datasets and limitations of these projections. Because each new phase of CMIP produces new projections under updated future climate scenarios and leads to a new round of downscaled climate projections, keeping up to date with modeling, scenarios and other changes between CMIP phases can be time consuming. This primer is intended to provide a broad overview of the two most recent phases of CMIP - the fifth phase of CMIP (CMIP5) and the sixth (CMIP6), as well as their relevant future climate scenarios and related downscaled climate model datasets. We also describe how MARISA products utilize these datasets and our planned changes to shift to the most recent CMIP6 projections.

Coupled Model Intercomparison Project 5 (CMIP5)

The fifth phase of CMIP, CMIP5, took place from roughly 2008 to 2013. Similar to other phases, it was managed by WCRP and included over 20 different global climate modeling groups and 50 different earth systems models.⁶ CMIP5 had a number of scientific goals, chief among them to evaluate and compare global climate models, as well as to produce a standardized set of projections of global climate change to 2100.⁷

To meet both scientific and practitioner needs, the CMIP5 effort produced a publicly-available set of global climate model outputs in a consistent format and under a consistent set of scenarios of future greenhouse gas emissions for 60 atmospheric variables including maximum daily temperature, humidity, windspeed and daily precipitation.⁸ These scenarios are termed Representative Concentration Pathways (RCPs) and are described in the next section.

These data served as the scientific basis for climate projections in the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (AR5), as well as in the Fourth National Climate Assessment (NCA4). ^9,10 Outside of these national and global consensus efforts, CMIP5 data have been used at state to local levels for mitigation and adaptation planning throughout the United States. CMIP5 has now been updated to the sixth phase of CMIP, CMIP6, which is described below.

Coupled Model Intercomparison Project 6 (CMIP6)

In 2021, the sixth phase of CMIP, CMIP6, released a new suite of global climate model outputs from 100 distinct models and 49 different modeling groups that went through a development process similar to CMIP5.¹¹ The number of modeling groups and models included in CMIP6 are double those from CMIP5, representing both the growing body of scientific work in global climate change as well as an overall expansion of the CMIP effort. CMIP6 also included a number of key technical updates. Additional or augmented model components and greater spatial and temporal resolutions enhanced the ability of models to capture key dynamics and smaller-scale processes or in key topographies, such as mountains.

The most notable technical difference between the suite of CMIP5 and CMIP6 models is their sensitivity to atmospheric greenhouse gas concentrations. Specifically, CMIP6 models have a higher climate sensitivity, meaning they project a greater change in global temperature for the same amount of atmospheric greenhouse gas concentrations.¹² For many of the CMIP6 models, this is true for both the short-term climate response to emissions, and in the long-term.¹³ This is important as it helps characterize the scope and timing of potential warming to different trajectories of emissions. Because CMIP6 models also have a greater range in their climate sensitivity than those from CMIP5, scientific groups are able to better understand the role of climate sensitivity in producing changes in our global climate.

From a practitioner perspective, some literature has recommended excluding those CMIP6 models with extremely high climate sensitivities (termed "hot" models), but studies largely suggest changes in atmospheric variables have been relatively consistent across CMIP5 and CMIP6.^14,15,16 Some studies suggest that CMIP6 has a potentially greater capacity to capture extremes, most notably for temperature, but a narrower range of projections (and thus smaller amount of uncertainty) for hydrologic indicators, such as precipitation variables. ^17,18

Representative Concentration Pathways (RCPs)

Representative Concentration Pathways (RCPs) were developed as quantitative descriptions of future atmospheric greenhouse gas concentrations due to potential emissions trajectories and factors such as land use and population growth.¹⁹ RCPs were initially used as inputs to CMIP5 models in order to examine how future emissions would produce changes in our global climate.²⁰ For CMIP5, four RCPs were developed, as shown in Table 7. RCPs are named for their level of radiative forcing, or the amount of additional energy entering Earth's atmosphere due to human activities. Of the RCPs, the most commonly used from CMIP5 are RCP4.5 and RCP8.5 which were generally considered as a pathway following current policies and emissions (RCP4.5) and a pathway with higher emissions (RCP8.5). CMIP6 introduced additional RCPs, which are detailed below with Shared Socioeconomic Pathways (SSPs).

	Greenhouse Gas Emissions	Agricultural Area	Air Pollution
RCP2.621	Very low	Medium for cropland and pasture	Medium-low
RCP4.522	Medium-low	Very low for cropland and pasture	Medium
RCP623	Medium	Medium for cropland and very low for pasture	Medium
RCP8.524	High	Medium for cropland and pasture	Medium-high

Source: Table reproduced from VanVurren et al., 2012 https://link.springer.com/article/10.1007/s10584-011-0148-z#Tab2

Shared Socioeconomic Pathways (SSPs)

The other notable difference between CMIP5 and CMIP6 was how it conceptualized future climate scenarios. In CMIP5, global climate models were run under RCPs as their future scenarios. As described above, RCPs characterize different future pathways for atmospheric greenhouse gas concentrations. At the time scientists were developing RCPs, another group began modeling how socioeconomic factors may change in the future. This parallel effort was designed to complement the RCPs and when used together provide a more holistic set of trajectories of how human activities could influence global climate change. These socioeconomic pathways were termed Shared Socioeconomic Pathways (SSPs).

SSPs use narrative and quantitative descriptions of different potential pathways for how societies may evolve and change over time based on factors such as population growth, ecological and environmental conditions, inequality, and technological innovation.²⁵ There are five different SSPs, each with a different story about the trajectory of the world:²⁶

• SSP1: Sustainability—Taking the Green Road. Low challenges to mitigation and adaptation.
• SSP2: Middle of the Road. Medium challenges to mitigation and adaptation.
• SSP3: Regional Rivalry—A Rocky Road.High challenges to mitigation and adaptation.
• SSP4: Inequality—A Road Divided. Low challenges to mitigation, high challenges to adaptation.
• SSP5: Fossil-fueled Development—Taking the Highway. High challenges to mitigation, low changes to adaptation.²⁷

Because SSPs were fully published in 2017,²⁸ they did not feature prominently in CMIP5 and instead form the basis of future climate scenarios for CMIP6. For CMIP6, the SSP futures were combined with an updated set of future atmospheric greenhouse gas concentrations (formerly RCPs) to provide the future climate scenarios, which carry the name SSPs, that served as inputs to CMIP6 models. Under CMIP6, RCPs were expanded from the four used in CMIP5 (2.6, 4.5, 6.0, 8.5) to also include futures with different levels of atmospheric greenhouse gas concentrations 1.9, 3.4 and 7.0. As an extension of RCPs, these are also named after their radiative forcing in 2100. Figure 7 illustrates how these were combined with the socioeconomic scenarios to form comprehensive SSPs. In data and literature, it is common to see SSPs described by their combination of SSP and their level of radiative forcing, such as SSP2-4.5, the combination of SSP2 and RCP4.5.

SOURCE: Reproduced from Chen et al., “Framing, Context, and Methods,” in Masson-Delmotte et al., eds., Climate Change 2021: The Physical Science Basis, p. 232, Cross-Chapter Box 1.4, CC BY-NC-ND 4.0.

To learn more about RCPs and SSPs, read the GLISA Practitioner’s Guide To Climate Model Scenarios.

Downscaled Climate Projections

The resolution of global climate models can be too large for some applications at regional and local scales. Fortunately, there are methods called downscaling that can take the global climate models and increase their spatial or temporal resolution. This increase in resolution does not necessarily make the downscaled models more accurate than the global models but develops datasets with resolutions more appropriate for local needs.

Because each phase of CMIP produces new climate projections, related downscaled climate model projections are developed in close succession. For CMIP5, the most commonly used downscaled products in the United States included Locally Constructed Analogs (LOCA), Multivariate Adaptive Constructed Analogs (MACA), and North America Coordinated Regional Downscaling Experiment (NA-CORDEX). Each dataset offers tradeoffs in terms of its downscaling approach, LOCA and MACA are statistically downscaled while NA-CORDEX is dynamically downscaled,²⁹ as well as their temporal and spatial resolution. LOCA data is perhaps the most widely used and featured in work such as the National Climate Assessment. Technical details for these datasets are included in Table 8.

Dataset	Number of Models	RCPs	Spatial Resolution	Temporal Resolution
MACA	20	4.5, 8.5	2.5 miles	Daily
LOCA	32	4.5, 8.5	3.7 miles	Daily
NA-CORDEX	5	4.5, 8.5	15.5 to 31 miles	Sub-daily to daily

SOURCE: Table reproduced from Miro et. al, 2021.³⁰

NOTE: Spatial resolutions across all of the datasets are approximate and are based on conversions from degrees to miles at mid-latitudes.

For CMIP6, downscaled data products are still being produced, but a number have already been developed, including NASA Earth Exchange Global Daily Downscaled Projections

(NEX-GDDP-CMIP6), LOCA2 and Seasonal Trends and Analysis of Residuals, Empirical-Statistical Downscaling Model (STAR-ESDM). These are all statistically downscaled datasets. While use cases are still underway for these datasets, both LOCA2 and STAR-ESDM are featured in the Fifth National Climate Assessment (NCA5).

Dataset	Number of Models	SSPs	Spatial Resolution	Temporal Resolution
NEX-GDDP-CMIP631	35	SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5	15.5 miles	Daily
LOCA232	27	SSP2-4.5, SSP3-7.0, SSP5-8.5	3.7 miles	Daily
STAR-ESDM33	16	SSP2-4.5, SSP3-7.0, SSP5-8.5	15.5 miles	Daily

All of our previous climate summaries have used the LOCA downscaled versions of the CMIP5 suite of climate models. Our future climate summaries will use the LOCA2 downscaled versions of the CMIP6 climate models. LOCA and LOCA2 both use statistical downscaling methods, but the historical training dataset and bias correction methods used in LOCA2 were designed to help do a better job at preserving any extremes in daily precipitation from global climate models.³⁴ For a detailed comparison between the two, refer to the LOCA website: https://loca.ucsd.edu/loca-version-1-vs-loca-version-2/. We will be exploring the differences between the LOCA2 and LOCA climate projections in the next two climate summaries. For the upcoming Spring edition, we will take a look at the precipitation projections and for the Summer edition, we will look at temperature.

The Northeast Regional Climate Center recently held a webinar that included an initial examination of the differences between the LOCA and LOCA2 climate model projections for the Northeast region, which is available to view online.