National energy transitions under decarbonization goals. We investigate the environmental impacts (i.e., total air pollution emissions) and equality (i.e., regional distribution of emissions) of different electricity generation investment strategies under eight decarbonization strategies over a 40-year time-period (see Table 1 in Methods). Our decarbonization scenarios include the base case with no additional carbon constraints (Scenario A), two carbon cap scenarios which meet either the US nationally determined contributions (NDC) from the Paris Agreement or a pathway to stay under 1.5°C warming (Scenarios B and C respectively), and five technology specific portfolios which deploy either renewable energy (Scenarios D – F) or low carbon (Scenarios G and H) generation (see Methods Table 1 for scenario descriptions).
The annual generation by technology for the decarbonization scenarios is shown in Figure 1. For Scenario A (the Base Case), that implements no additional carbon constraints or policies, generation in 2010 has the majority of the generation is supplied by coal, and natural gas and nuclear, but by 2050 we see coal generation decrease to 7.5% (0.41 PWh), natural gas generation slightly increase to 20.0% (1.08 PWh), onshore wind increase to 33.8% (1.83 PWh), and solar PV generation increase to 20.9% (1.14 PWh) of total generation. The carbon cap scenarios (B and C), which place a strict limit on CO2 emissions from the electricity sector, achieve their carbon caps primarily through deploying solar PV and onshore wind. In both scenarios, wind and solar represent less than 3% of the generation in 2010, but by 2050 we see solar PV [B: 20.3% (1.11 PWh), C: 27.8% (1.55 PWh)] and onshore wind generation [B: 37.0% (2.0 PWh), C: 50.0% (2.78 PWh)] supplying a large share of total generation in 2050. Scenario C specifically sees complete retirement of coal by 2035 and almost complete retirement of natural gas by 2050 (0.2% of generation). However, contrasting to Scenario C, the carbon cap defined in Scenario B allows an increase of CO2eq. emissions, which are allotted to an increase in coal generation 2040 to 2050 (1.67% of generation in 2040 to 4.67% of generation in 2050).
Onshore wind generation increases the most in technology specific scenarios with an implemented renewable portfolio standard (RPS) (D, E, and F): Scenario D averages a 25.8% annual increase, Scenario F averages a 27.3% annual increase, and Scenario E averages a 28.7% annual increase. These annual increases lead to onshore wind representing approximately 50% of generation by 2050 in all three scenarios. Scenarios D, E, and F also see large solar deployment, again due to the implemented RPS. The highest generation of solar PV, CSP, biopower, and battery storage are deployed to meet the 100% renewable requirement by 2035 in Scenario E, with solar PV technology representing 35.0% of total generation by 2040. In the other seven scenarios, solar PV is still a large contributor to generation, with solar PV supplying 15-20% of total generation in 2040.
Scenarios that require low carbon technology [renewable sources, nuclear, or natural gas carbon capture and storage (CCS)] by 2035 (Scenario G) and 2050 (Scenario H) strategies primarily rely on natural gas generation until their low carbon requirement year when it is completely retired. Upon reaching the technology mandate natural gas is primarily replaced with natural gas CCS. Thus, without a mandate this technology would most likely continue to provide 10-20% of the total generation needs. See Table S-4 in Supplemental Information (SI) for summary of generation profiles by decade for each scenario.
National environmental sustainability. The environmental impacts of the changing power plant profile can be seen in Figure 2, which depicts the national life cycle and operating emissions over the model timeline 2010 – 2050. Operating emissions are classified as emissions produced directly from the power plant creating electricity, while life cycle emissions stem from the production, generation, and retirement of power plants (emissions over their entire lifetime). We estimate life cycle and operational emissions using power plant emissions factors from literature estimates (see Methods). In these results, we discuss life cycle and operating emissions from CO2eq., NOx, SO2, and PM. PM emissions accounted for in life cycle analysis are total PM emissions, whereas PM emissions accounted for in operating emissions analysis is PM2.5. From Figure 2, we find CO2eq. and NOx have similar trends, with emissions decreasing through 2050, but at varying magnitudes across scenarios. SO2 and PM life cycle emissions also see similar trends to each other with emissions rising from 2010 to 2022 due to increased investments in natural gas, but then we see a turning point for SO2 and PM, with emissions decreasing through 2050. Scenario A (base case) is an upper bound for national emissions across all pollutants in our analysis.
In the carbon cap scenario where emissions are limited to hold the climate at or below 1.5℃ (Scenario C), and the technology specific scenarios with the 2035 deadline (Scenarios E and G) life cycle greenhouse gas emissions fall below 500 megatonnes (Mt) of CO2eq. by 2035 and plateau. For operating emissions, these scenarios fall close to zero emissions by 2035.
The lower bound for the emissions is Scenario E (100% renewable by 2035). Life cycle emissions for SO2 reach 1 Mt and PM reach 0.30 Mt emissions by 2035; these emissions then subsequently plateau. Life cycle CO2eq. and NOx emissions in Scenario E in 2050 are at levels 76-93% below their 2010 levels. In SO2 and PM emissions, we see life cycle emissions across all scenarios rise 2010 to 2022 in the interim, driven primarily by increases in natural gas generation.
In the Scenarios A and B, the life cycle PM emissions in 2050 are 9-13% below their 2010 levels, while SO2 emissions in Scenario A and B are just 1-4% lower in 2050 than their 2010 levels. Since scenarios A and B do not require retirements of fossil fuels completely, the PM and SO2 emissions are not mitigated, and instances of associated health implications, like lung and heart disease, will continue if emissions increase or stay the same.
Operating emissions across all pollutants are decreasing over the modeling time horizon, resulting from high deployment of renewable and low carbon technologies. Scenario A again has the highest operating emissions (all pollutants) 2010 – 2050 because it does not have any carbon reduction requirements. The high operating emissions in Scenario A are driven by coal power plants (see SI Figures S-7 – S-10).
By 2035, operating emissions from Scenarios C, E, and G are under 100 Mt CO2eq. emissions. Coal has completely retired by 2035 in these scenarios, so it is not contributing to emissions, and natural gas or natural gas CCS contribute under 10% of generation. By phasing out coal and natural gas plants, emissions from co-pollutants like NOx, SO2, and PM also fall significantly, with NOx levels at or below 0.02 Mt, SO2 levels below 0.003 Mt, and PM levels below 0.002 Mt.
Air pollution distributional regional equality. While national level emissions analyses are important for measuring progress across the energy system as a whole, there are multiple entities that make decisions in the electricity sector (e.g., utilties, states). Regional inequalities resulting from energy transitions can manifest themselves in the unequal distribution of air pollution emissions across regions. There will be differences between who is responsible for emissions (life cycle emissions), and which populations bear the brunt of those emissions due to where they live (operational emissions). Life cycle emissions are important for global CO2eq. emissions, due to greenhouse gases impacting people regardless of where emissions are generated30. However, the operational co-pollutants will largely impact the health of people within the region, so operating emissions reductions of these pollutants will result in regional health benefits31–34. We present an analysis of both life cycle and operating greenhouse gas emissions (CO2eq.) and operating co-pollutant emissions (NOx, SO2, PM2.5) to illuminate how different decarbonization scenarios could impact emissions globally and regionally.
Figure 3 presents subnational (134 regions) regional life cycle emissions (in tonnes/capita) in 2035 and 2050 across all scenarios for CO2eq emissions. These maps indicate how much greenhouse gas emissions each region is responsible for from their power plants. With only 45% of regions beneath the 2.5 metric tons (t) CO2eq. per capita threshold in 2050, Scenario A (no policy implementation) has the largest distribution of emissions across regions. In 2035, we find that Scenarios C, E, and G have 80-90% of regions below 2.5 tonnes CO2eq./capita and does not see much change over the next 15 years where 2050 has the same 80-90% of regions below that threshold. Therefore, once technology or carbon mandates are met in 2035, life cycle CO2eq. emissions in regions do not change. The regions that have more CO2eq. emissions in these scenarios (part of Montana, Arizona, Kansas are a few) will therefore be more responsible for global CO2eq. emissions than other regions.
The distribution of operating emissions for NOx, SO2, and PM is seen in Figure 4 (a, b, c) respectively. Once the technology mandate is met in 2035 (Scenarios E and G) or 2050 (Scenarios F and H), the operating co-pollutant emissions reach their lowest values. However, in Scenario E, PM operating emissions are worse in some regions in 2050 than they were in 2035. The large PM emission increases results from biopower being deployed in the Eastern US to meet the 100% renewable technology requirement. We also find that the low carbon mandate scenarios (G and H) have greater operating NOx and PM emissions in the Eastern US due to emissions from biopower and natural gas CCS investments Thus, this highlights the need to consider co-pollutants in energy transitions due to local health effects (e.g., asthma) that results from increases in these emissions35,36
Emissions distribution across vulnerable groups. Beyond regional analyses that measure the magnitude of air pollution, it is useful to understand the distribution of operating emissions across different demographic and socioeconomic indicators (median income, poverty rate, cost of living, energy burden) across regions. This investigation shows the impact of different energy transitions on vulnerable regions. See Methods for data information about the demographic groups, and methods for the analysis. Here, we focus on operating NOx, SO2, and PM emissions, due to the local health impacts of these co-pollutants. Here we define inequality between regions as the difference in emissions per capita between the worst off (e.g., median income <$50k) and best off (e.g., median income >$70k) demographic groups. While there are multiple avenues for measuring vulnerability, we focus on two measures. The first is an absolute measure, while the second is a relative measure. Absolute measures indicate vulnerability based on a threshold (poverty level or median income), while relative measures indicate vulnerability based on income spent on a given expense (i.e., energy, transportation, rent).
Absolute vulnerability indicators. Our absolute vulnerability analysis focuses on median income and poverty rate across regions. The distribution of life cycle and operating emissions across different median income (Figure 5) illustrates the disparities between high- and low-income regions (see Methods for group definitions and the aggregation of median income census tracts to the ReEDS regions). Operating emissions across median income groups will impact health of vulnerable groups within regions, since pollutants like NOx, SO2, and PM have more local health affects1,37,38. In 2010, regions with the lowest median income see the highest incidence of emissions, and our findings suggest that these regions will continue to bear the largest burden of emissions from these pollutants (and associated health impacts)35. For scenarios that have large penetrations of renewable and low carbon technologies, the distribution of operating NOx and SO2 emissions across median income groups become equal (0 kg/capita) the year the carbon or technology requirement is mandated (2035 or 2050), but PM emissions remain greater in low-income regions even past the mandate year of 2035 because of biopower and natural gas CCS investments. Furthermore, 5 years before the scenario’s mandate, PM operating emissions in these technology mandate scenarios (E, F, G, and H) are 59-71% higher in the lowest income group than the highest income group, so reaching the mandate does help reach equality in NOx and SO2 emissions, but before those are met, the worse-off groups are still the most impacted. However, PM emissions past the mandate year (2035) in Scenarios E and G are still the highest in the lowest income group, indicating that a single-objective approach could leave vulnerable groups worse off even as we reach a low-carbon or 100% renewable energy system. Similar trends are seen across NOx and SO2 emissions in these scenarios five years prior to their mandates: NOx emissions in the lowest income group are 50-74% higher than the highest income group, and SO2 emissions are 18-74%. Low-income communities have been disproportionately affected by NOx emissions39, and our analysis shows that across different decarbonization scenarios, the lowest income regions continue to have the highest NOx operating emissions, unless a national mandate requires complete retirement of fossil fuels (as seen in Scenarios E and F).
In addition to median income, percent of poverty within a region can identify regional vulnerability as another absolute measurement. Poverty groups were identified by the percent of the population in a region that is experiencing poverty, which the US Office of Management and Budget classifies as a family who is under a certain threshold given their family makeup40. The group of regions with the highest percentage of residents in poverty (>15%) consistently had the highest life cycle and operating emissions per capita across 2010 – 2050 for all co-pollutants (SI Figure S-16).
Relative vulnerability indicators. Or relative vulnerability analysis focuses on energy burden or cost-of-living. Energy burden is defined as the percent of household income spent on satisfying energy needs. As nations deploy more technologies, there is a chance that grid costs, and subsequently energy bills, could rise41, which would affect the people who currently have trouble paying for their energy bills. Thus, we highlight how emissions will change across groups that will likely experience difficulty paying for electricity should costs rise. For the aggregation of energy burden and cost-of-living metrics to the ReEDS regions, see Methods. Operating emissions (Figure 6) across different energy burden groups show the highest energy burden group having the highest emissions per capita either across the entire timeline (Scenarios A, B, D) or until the year the carbon cap or technology mandate requirement is 100% low carbon or renewable energy (Scenarios C, E, F, G, H). Before reaching these years (2035 or 2050) the highest energy burden group is worse off. In 2030 in Scenario E, NOx, SO2, and PM operating emissions are 42%, 18%, and 38% lower respectively in the lowest energy burden (energy burden <2.75%) group than in the highest (energy burden >3.75%) before falling to a 0% difference in 2035. Likewise in 2045 in Scenario F, NOx and PM operating emissions are 28% and 40% respectively lower in the lowest energy burden group compared to the highest. Once 2050 is reached, these emissions fall to 0 kg/capita and a 0% difference between the groups. Therefore, without strict fossil fuel retirement mandates ties to decarbonization strategies the most vulnerable regions will continue to experience the highest emission burden.
Cost-of-living is another relative measurement used to understand disparities of air pollution distribution, which is classified as the percent of income spent on housing and transportation. We recognize that absolute income does not capture the relative cost of achieving a certain standard of living in different parts of the country. When accounting for different costs-of-living across regions we find similar results: where the operating emissions are greater in the highest cost of living group (greater than 66% of income to housing and transportation needs) until the year of the technology mandate. The highest and lowest cost-of-living groups have decreasing inequality across all pollutants in all scenarios, meaning the emissions between the highest cost-of-living group is approaching the emissions in the lowest cost-of-living group. (NOx: difference of 18.3 kg/capita in 2010 to an average difference across scenarios of 2.6 kg/capita, SO2: difference of 55.6 kg/capita in 2010 to an average difference of 7.1 kg/capita, PM: difference of 1.8 kg/capita in 2010 to an average difference of 0.32 kg/capita) (SI Figure S-15).