Socioeconomic and emissions projections from 2020 to 2300 were recently developed under the Resources for the Future Social Cost of Carbon Initiative (Rennert et al., 2021). These include multi-century probabilistic projections of country-level population; GDP; and global emissions of CO2, CH4, and N2O. While uncertainties in multi-century projections are considerable, as discussed in Rennert et al. (2021), these projections represent the largest set of probabilistic socioeconomic and emissions scenarios based on high-quality data, robust statistical techniques, and expert elicitation. These projections also incorporate coupled uncertainty in the time-dependent relationship between GDP and emissions, while also explicitly accounting for potential future climate policy and its contribution to the economy–emissions relationship (Rennert et al., 2021).

The Finite Amplitude Impulse Response model (FaIRv1.6.2)

https://github.com/OMS-NetZero/FAIR/releases/tag/v1.6.2 (last access: 8 July 2022)

The Framework for Evaluating Damages and Impacts (FrEDI) is a reduced-complexity model that assesses and quantifies future impacts to the US from a changing climate. As described in detail in the technical documentation (EPA, 2021b), FrEDI uses a temperature-binning approach and data from previously published climate impact studies (Sarofim et al., 2021) to develop relationships between climate-driven changes in CONUS temperature or global mean sea level rise and the resulting physical and economic damages across 20 sectors (Table A1) in seven US regions. While FrEDI evaluates both negative and positive impacts of climate change across sectors and regions, climate-driven damages outweigh the positive effects for all sectors at the national level. FrEDI also provides insight into differences in impacts under various adaptation scenarios and contains a module that can be used to quantify impacts to socially vulnerable populations. The underlying studies in FrEDI consist of bottom-up detailed sectoral analyses from the CIRA project (EPA, 2017a) and other studies including those from the Climate Impact Lab (e.g., Hsiang et al., 2017) and the American Thoracic Society (e.g., Cromar et al., 2022). FrEDI was designed to fill the current need of monetizing a broad range of climate-driven impacts in the US across various warming, emission, and socioeconomic trajectories, while doing so in a significantly shorter computational time frame (e.g., seconds) relative to existing impact models.

FrEDI currently includes 20 impact sectors for which damages are modeled as functions of a climate driver (CONUS temperature or sea level rise), US GDP, and regional population. The GDP and population projections from the RFF-SPs are at the country level (i.e., total US population). For the analysis, we disaggregate national populations from the RFF-SPs to populations for each of the seven FrEDI regions based on the percentage of regional to total US population in the years 2010–2090 using projected regional populations derived from ICLUS (EPA, 2017b). Neither population projections, ICLUS or RFF-SP, were generated considering future climate changes such as climate-induced migration. The proportions for each region are held constant after 2090. Figure A1 shows that the mean and 95th confidence intervals for US population and time-averaged US GDP per capita growth rates are USD 390 million (95 % CI: USD 260 million–USD 520 million) and 1.5 % (95 % CI: -0.4 % to 4.0 %), respectively, in 2100

All dollar values in this paper are presented in 2020 US dollars. Any necessary transformations in the inputs (e.g., RFF-SPs are in 2011 USD, FrEDI takes in 2015 USD, and FrEDI results are presented in 2020 USD) are performed using the U.S. Bureau of Economic Analysis (BEA) national data on annual implicit price deflators for US GDP, the top row of BEACE3 Table 1.1.9.

For sectoral impacts driven by temperature change, damages in FrEDI are calculated as functions of CONUS degrees of warming over time relative to a 1986–2005 average temperature baseline. In this analysis, CONUS mean temperature change is estimated for each FaIR-derived temperature projection (calculated from each RFF-SP emissions scenario), as CONUS temperature (∘C) is equal to 1.42 × global temperature (∘C) (EPA, 2021b). This relationship between CONUS and global temperatures is relatively stable across global circulation models (GCMs) and over time, allowing the use of these available data points to develop a generalized relationship between global and CONUS temperature anomalies. Sub-national differences in warming are also explored within FrEDI using results derived from a consistent set of GCMs that were also used within the underlying studies (e.g., Sarofim et al., 2021). For example, unique damage functions for each sector (and variant within each sector) are developed for each region and GCM based on its relationship to CONUS temperature. While FrEDI outputs damages by region and GCM, the main results in this analysis present national and regional damages calculated from the average across the GCM ensemble. For sectoral impacts driven by sea level rise (i.e., coastal properties and transportation impacts from high-tide flooding), global mean sea level is calculated within FrEDI from global mean surface temperature using a semi-empirical method that estimates global sea level change based on a statistical synthesis of a global database of regional sea level reconstructions from Kopp et al. (2016). In FrEDI, sea-level-driven damages are calculated for a given year by interpolating between modeled damages at different sea level heights at that same point in time; this enables FrEDI to account for interactions between adaptation costs, increased coastal property values, and sea level rise over time (EPA, 2021b).

This analysis groups mean damages from each of 20 FrEDI sectors into six topical categories and uses the default FrEDI adaptation assumptions of “reactive”, “reasonably anticipated adaptation”, or “no additional adaptation” (see Table A3) for each sector. As discussed further in Sect. A3, reactive or reasonably anticipated adaptation is where decision makers respond to climate change impacts by repairing damaged infrastructure (e.g., road or rail repair) or reactively responding to current conditions (e.g., building sea walls or beach nourishment) but do not take actions to prevent or mitigate future climate change impacts. No additional adaptation largely incorporates historical or current levels of adaptive mitigation that were in place during the time period of each underlying sectoral study. Example sensitivities to projected climate-driven damages are explored within Sects. 3.1 and A3.

FrEDI also has the capability to investigate adaptation options in select sectors. Available adaptation options reflect the treatment of adaptation in the underlying sectoral studies. For most of these studies, because the implicit or explicit impact response functions are calibrated to historical or current data, historically practiced adaptation or hazard avoidance actions are “baked in”, while enhanced adaptation action or new (currently unknown) technologies are not considered. Exceptions include FrEDI's coastal property and select other infrastructure sectors (e.g., roads, rail), where adaptation options and scenarios from the underlying studies have been incorporated into FrEDI. Total damages in these sectors are sensitive to adaptation assumptions, indicating that adaptation has the capacity to both exacerbate and ameliorate future climate-driven damages, with the latter being more common. These results are further explored below and in Sect. A3.

In addition to quantifying differential climate-driven damages across impact sectors, geographic regions, and adaptation options, FrEDI can also compare climate-driven damages across different populations within the US. This capability is based on a recent EPA Report on Climate Change and Social Vulnerability in the United States (EPA, 2021a), which considers differential climate change risk as a function of exposure to where climate change impacts are projected to occur. These differential impacts are calculated in FrEDI at the census tract level as a function of current population demographic patterns (i.e., percent of each group living in each census tract) (US Census), projections of CONUS population (U.S. EPA, 2017), and projections of where climate-driven damages are projected to occur (from census-tract-level temperature–impact relationships in FrEDI). The relative percent of each group in each census tract is from the 2014–2018 US Census American Community Survey dataset (US Census) and is held constant over time because robust and long-term projections for local changes in demographics out to 2090 and beyond are not readily available. We consider four categories for which there is evidence of differential vulnerability (Table A2), including low income, ethnicity, and race

This analysis uses the term BIPOC to refer to individuals identifying as Black or African American, American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, and/or Hispanic or Latino. It is acknowledged that there is no “one size fits all” language when it comes to talking about race and ethnicity and that no one term is going to be embraced by every member of a population or community. The use of BIPOC is intended to reinforce the fact that not all people of color have the same experience and cultural identity. This report therefore includes, where possible, results for individual racial and ethnic groups.

While FrEDI was initially built to project damages through 2090 for temperature scenarios with a maximum value of 10 ∘C of warming, it was extended in this work to project climate damages out to 2300 to quantify the net present damages in the US resulting from an additional tonne of CO2, CH4, or N2O emissions. As described further in Sect. A4, FrEDI is extended by linearly extrapolating its sector-specific, temperature-binned damage functions to account for the full range of temperature scenarios derived from the RFF-SP emission scenarios run through FaIR (some of which have degrees of warming above 10 ∘C). To quantify the net present damages, all 10 000 RFF-SP-derived temperature and socioeconomic scenarios are then run through FrEDI out to 2300 under two cases: a baseline (emissions = RFF-SP emissions) and a perturbed case, where 1 GtC pulse of CO2 (or CH4 or N2O) is added to each of the RFF-SP emissions scenarios in the year 2020. The emissions are identical between the cases for all other years. The annual marginal climate-driven damages are calculated as the difference between the damages in the baseline and perturbed cases, summed across all sectors and all regions for each year. Lastly, these marginal annual damages are discounted to the year of emissions and then aggregated across the time series into a single net present-damage estimate. The results are normalized by the pulse size and gas chemistry (e.g., C to CO2) and reported in 2020 US dollars.

Future monetary impacts are generally discounted relative to present value. Circular A-4 (White House, 2003) recommends a constant value of 3 % for the “social rate of time preference”, which is considered to be the appropriate discount rate to use for impacts on private consumption (which would include most environmental and health impacts). The discount rate of 3 % was calibrated to the real rate of return for 10-year Treasury notes from 1973 through 2003. However, Office of Management and Budget (OMB) Circular A-4 also noted that for intergenerational impacts (a category in which climate change clearly falls), discount rates lower than 3 % might be appropriate. Moreover, recent real rates of return for Treasury notes have been lower than 3 %, adding support for use of a discount rate smaller than 3 % (CEA, 2017). A number of economists, as well as the National Academies of Sciences (NAS, 2017), have alternatively suggested the use of Ramsey discounting (Eq. 2, ρ is the rate of pure time preference, g is a time-varying measure of per capita consumption or income, and η is the elasticity of the marginal value of consumption with changes in gt) as an appropriate approach to discounting long-term problems such as climate change. The effect of Ramsey discounting is to value damages more highly in futures with less economic growth, e.g., future societies that have fewer resources available for adaptation, and vice versa. A recent study from Rennert et al. (2022b) used a Ramsey approach calibrated to a near-term target discount rate of 2 %, with ρ=0.2 % and η=1.24.

For Ramsey discounting calibrated to near-term target discount rates of 1.5 %, 2.5 %, or 3 %, ρ=0.01 %, 0.5 %, and 0.8 % and η=1.02, 1.42, and 1.57, respectively.

The net present value (NPV) for a constant discount rate (r) is calculated such that 1NPVDt=∑t=2020t=2300D(t)(1+r)t. The net present value for a Ramsey discounting approach is calculated using a time-varying and state-specific discount rate

Consistent with Rennert et al. (2022b), we use a stochastic Ramsey discount factor to discount future climate-driven damages.

FrEDI was developed to quantify the physical and economic damages from climate change over the 21st century within contiguous US borders. Figure 2 shows the net annual economic climate-driven damages across 20 sectors in the US in the years 2050, 2070, and 2090, as calculated by the mean from the 10 000 baseline RFF-SP scenarios (i.e., emission, population, and GDP trajectories). Total annual damages throughout this analysis are shown in 2020 US dollars, converted from FrEDI's base units of 2015 USD using annual GDP implicit price deflators (U.S. Bureau of Economic Analysis, 2023). Figure 2 shows that net national damages increase overtime, with mean climate-driven damages estimated to reach USD 2.9 trillion (95 % CI: USD 510 billion to USD 12 trillion), or ∼3 % of US GDP, annually by 2090 for a subset of total climate impacts. Given that the drop in GDP in 2009 during the Great Recession was 2.2 %

Data from https://fred.stlouisfed.org/series/FYGDP, percentage decline in annual GDP from 2008 to 2009.

Climate-driven damages from FrEDI are largest for the health category. The majority of damages in this category are from the estimated valuation of premature mortality attributable to climate-driven changes in temperature and air quality (O3 and PM2.5) but also include monetized health damages attributable to Valley fever, Southwest dust, wildfire smoke exposure and suppression costs, and crime incidents. Another FrEDI category that includes the monetized value of directly estimated physical impacts (rather than a direct modeled relationship between temperature and monetized damages) is labor, which is the third-largest category in 2090 and represents the damages resulting from lost hours of work when temperatures are too hot for workers to work outdoors or in unconditioned workplaces (e.g., warehouses). Table 2 provides the mean physical impacts from each of the sectors in the health and labor categories in 2090, along with the 95 % CI. As shown in Table 2, climate-driven changes in temperature have the largest impact on premature mortality, resulting in nearly 50 000 additional deaths (95 % CI: 19 000–91 000 deaths) annually by 2090, followed by climate-driven changes in air quality (5100 deaths; 95 % CI: 2100–10 000 deaths) and exposure to wildfire smoke (1100 deaths; 95 % CI: 460–1700 deaths).

To further illustrate the distribution of monetized damages across sectors, Fig. 4 shows the range of 2090 annual climate-driven damages in each of the 20 sectors in FrEDI, across all 10 000 RFF-SP emission, GDP, and population scenarios, in decreasing order of sectoral mean damages. Figure 4 shows that national total damages in 2090 are primarily driven by the valuation of premature mortality attributable to climate-driven changes in temperature (mean: USD 2.3 trillion per year; 95 % CI: USD 0.31–USD 9.9 trillion per year). The next four sectors with the largest monetary climate-driven damages include premature mortality attributable to changes in air quality (mean: USD 240 billion per year; 95 % CI: USD 32–USD 1000 billion per year), transportation impacts associated with changes in high-tide flooding (mean: USD 140 billion per year; 95 % CI: USD 110–USD 200 billion per year), national labor hours lost (mean: USD 51 billion per year; 95 % CI: USD 6.7–USD 210 billion per year), and health damages from wildfire smoke exposure and response costs from wildfire suppression (mean: USD 51 billion per year; 95 % CI: USD 8.1–USD 220 billion per year). Climate-driven damages to coastal properties associated with changes in tropical storm frequency and wind strength (mean: USD 28 billion per year; 95 % CI: USD 12-USD 49 billion per year), damages attributable to changes in rail (mean: USD 19 billion per year; 95 % CI: USD 7.7–USD 45 billion per year) and road systems (mean: USD 17 billion per year; 95 % CI: USD 6.6–USD 35 billion per year), health damages from changes in southwestern dust exposure (mean: USD 18 billion per year ; 95 % CI: USD 2.5–USD 77 billion per year), and the health burden of change in Valley fever incidence (mean: USD 14 billion per year; 95 % CI: USD 2.0–USD 58 billion per year) round out the top 10 sectors with the largest annual damages in 2090. Figure A2 provides the mean and 95 % confidence interval total damages for each sector over the entire 2020–2100 time series. The large distribution of damages in each individual sector is driven by a large range of RFF-SP emissions, population, and GDP projections and the dependence of the valuation approach for each sector on these parameters (as described in EPA, 2021b).

These sectoral damages are sensitive to assumptions in the adaptation scenarios (see Sect. A3 for more detail). For example, the coastal property sector considers three different adaptation options, no adaptation, reactive adaptation, and proactive adaptation. The underlying model within this sector, the National Coastal Property Model, has options for optimal (“proactive”) response to future sea level rise, “reactive” or reasonably anticipated response to current conditions (including sea walls, beach nourishment, house elevation, or managed retreat), or rebuilding in place as often as necessary. Historical data suggest that most of our response to sea level rise thus far is in between reactive adaptation and no adaptation (Lorie et al., 2020). Considering the range of possible adaptation options in this coastal property sector, mean damages range from USD 17 billion under no adaptation to USD 7.5 billion under proactive adaptation. Damages under the default reactive adaptation assumption are USD 9.4 billion. While the inclusion of adaptation options for any sector within FrEDI depends on the consideration and treatment of adaptation in the underlying impact studies, Table A3 further illustrates that projected climate-driven damages are sensitive to adaptation options in each sector where they are considered. Notably, the largest impact sector in this study, temperature-related mortality, does not include assumptions about future adaptation. While the primary underlying study (Cromar et al., 2022) is a well-regarded meta-analysis of existing global temperature-related mortality studies, it does not explicitly consider future adaptative measures. Exploring projected 2090 damages from one alternative damage function that assesses impacts of extreme temperature on mortality in 49 US cities (Mills et al., 2014) suggests that damages will be reduced (Table A4) in the event that US cities can gradually adapt to hotter temperatures, for example, through physical acclimatization, increased air conditioning penetration, and behavioral changes. Several other studies have also observed reductions in temperature-related vulnerability over time (Lay et al., 2021); however, there is little consensus regarding the most appropriate way to consider future adaptation in this sector, even though several methods have been applied (Sarofim et al., 2016; Carleton et al., 2022; Heutel et al., 2021). Therefore, we use the most recently published meta-analysis for the central estimate in this analysis but also present results from alternative assumptions and studies (Tables A3 and A4), further illustrating the unique advantage of the FrEDI framework of enabling direct comparisons across studies.

The sectors assessed in this study are independent and therefore damages are additive across these sectors. One potential exception could be temperature-related mortality and the climate–air quality linkage, as most approaches to estimating temperature-related mortality are statistical rather than mechanistic, which could lead to double counting of some health effects between these two sectors. Specifically, Cromar et al. (2022) note that it will be important to continue exploring potential synergies between the effects of temperature and air pollution to adequately capture the potential risk in compound climate events such as these. Conversely, there can also be compounding effects that the FrEDI analytical approach does not account for, e.g., power outages due to increased summer electricity demand could exacerbate temperature-related mortality. However, few studies produce quantitative, monetized estimates of compounding or interacting effects at the national scale as would be required to build into comprehensive impact tools (Clarke et al., 2018).

Results from FrEDI also show that climate-driven damages across the national population vary by geographical region. Figure 5 shows a map of the damages per capita in each CONUS region in the year 2090, with pie charts showing the per capita damages in each region and the share of the four sectors with the largest damages (Fig A3 shows absolute regional damages). Based on the climate impacts included in FrEDI, Fig. 5 shows that the Southeast will experience the largest annual damages per capita (mean: USD 9300 per person annually; 95 % CI: USD 1800–USD 37 000 per person annually), whereas the smallest damages per capita are expected in the Southwest region (mean: USD 6300 per person annually; 95 % CI: USD 840–USD 27 000 per person annually). In each region, the largest monetary damages in 2090 are expected from premature mortality associated with changes in temperature, ranging from USD 4500 per person in the Southwest to USD 6500 per person in the Southeast. Damages from transportation impacts from high-tide flooding and premature mortality attributable to climate-driven change in air quality are the second and third largest in the coastal Southeast and Northeast regions. In the Northwest and Southwest, the sectors with the second- and third-largest climate-driven monetized damages are air quality and wildfires. In the Southern Plains, high-tide flooding transportation impacts and labor hours lost are the second- and third-largest sectors, while rail and wildfires are the second and third largest in the Northern Plains, and labor and rail are the second and third largest in the Midwest. There are some regions and sectors projected to benefit from warming temperatures, including an expected reduction in air pollution attributable mortality in the Midwest under warmer conditions. Overall, however, the negative impacts of climate change outweigh the positives such that net losses are projected in each region.

Lastly, climate change may also broaden existing societal inequalities (EPA, 2021a), and understanding the comparative risks to different populations is critical for developing effective and equitable strategies for responding to climate change. As described in Sect. 2, FrEDI contains a module to generate and report results of disproportionate exposure and distributional physical effects across four groups of potentially socially vulnerable populations for six sectors. For example, results from this module show that Black or African Americans are more likely to be affected by additional premature mortality from climate-driven changes in air quality, while Hispanic or Latino Americans are more likely to experience lost labor hours (Fig. 6) under a changing climate.

Confidence intervals presented throughout this analysis account for uncertainty associated with the range of future emission and socioeconomic projections across the 10 000 RFF-SP scenarios. These also incorporate climate parameter uncertainty as a Monte Carlo approach was used to sample the calibrated parameter set when running FaIR with the 10 000 RFF-SP emissions scenarios. In addition to these uncertainties and sensitivities to adaptation options, damage estimates within FrEDI are also sensitive to uncertainties in the underlying damage functions themselves. Similar to adaptation, FrEDI can incorporate parametric uncertainty in each damage function when the relevant information is available in the underlying study, as well as structural uncertainty when multiple damages functions are available for a single sector. For example, as further described in Sect. A4, FrEDI incorporates three studies of climate-driven temperature-related mortality, two of which include underlying uncertainty estimates. As shown in Table A4, there is a large range of damage estimates from temperature-related mortality across each study; however, these values all fall within the uncertainty range derived from the RFF-SP scenarios presented in the main text.

To place mean damages in the context of alternative future storylines, Table 3 shows a comparison of annual national climate-driven damages in the US in the year 2090 from a subset of four Shared Socioeconomic Pathways (SSPs), which represent projected socioeconomic global changes up to 2100 (Riahi et al., 2017). Annual damages in Table 3 are calculated following the same approach as outlined in Fig. 1 but using SSP trajectories of emissions, US GDP, and US population from the SSP Public Database (v2.0)

https://tntcat.iiasa.ac.at/SspDb/dsd?Action=htmlpage&page=80, (last access: 5 October 2022)

We extend FrEDI to project climate damages through to 2300 (Sect. A4, Table A5) to quantify the net present damages within the US resulting from an additional tonne of CO2, CH4, or N2O emissions.

Net present damages resulting from an additional tonne of CO2 emissions is sometimes characterized as a “domestic social cost of carbon”.

These results show that even considering only the direct CONUS impacts as estimated by FrEDI, damages per tonne of CO2 are almost 20 % of a recently estimated global value (USD 185 per tonne of CO2 under a 2 % Ramsey discounting, Rennert et al., 2022b). This methodology can also be extended to explore the net present value of future damages resulting from an additional tonne of CH4 (USD 500 per tonne of CH4 under a 2 % Ramsey discounting), N2O (USD 9700 per tonne of N2O under a 2 % Ramsey discounting), or other greenhouse gas emissions.

We recognize that multi-century projections are inherently challenging. This is particularly true for socioeconomic projections of GDP, population, and technologies: even projections to the end of the century have been challenged (Barron, 2018). The climate system is better understood, but FaIR only captures the effects of those feedbacks and tipping points that are apparent in the GCMs and historic record to which FaIR was calibrated.

While the damages estimated within FrEDI are constrained to the 48 contiguous United States, it is important to note that the appropriate climate damages to consider when evaluating policy-induced changes in a global pollutant such as greenhouse gases would be damages that account for impacts around the globe. For example, The National Academies of Sciences advised that “[i]t is important to consider what constitutes a domestic impact in the case of a global pollutant that could have international implications that affect the United States” (NAS, 2017). Impacts that occur outside of US borders (and outside of FrEDI) will impact the welfare of US residents and firms because of the interconnectedness of the global economy, international markets, trade, tourism, national security, political destabilization, additional spillover effects, and many other activities not yet captured in FrEDI. Moreover, the act of international reciprocity has been highlighted as motivation for including damages occurring outside of US borders in a social cost estimate of global pollutants (Carleton and Greenstone, 2022; Revesz et al., 2017; and references within). It has also been shown that accounting for global damages in domestic policymaking can be individually rational (Kotchen, 2018). Therefore, we emphasize the contribution of the damages estimated within FrEDI as providing a useful understanding of the channels through which climate change can affect US citizens and residents and their relative magnitudes beyond what is currently possible in many global models yet remain a partial estimate of the total damages from greenhouse gas emissions.