the EMPATHI framework
EMPATHI (Energy Metrics for Project Assessment of Trade-offs & Human Impacts) translates a user defined electricity sector intervention (generation addition, demand reduction, or fossil reduction) into projected avoided emissions and downstream climate, public health, and ecological outcomes through 2100. It covers thirteen technology options across all fifty states, the District of Columbia, Puerto Rico, and a national default, with up to twelve grid emissions projection schemes and five Shared Socioeconomic Pathways. It gathers scattered impact metrics into one defensible workflow by integrating EPA AVERT, NREL Cambium, EPA eGRID, and EPA COBRA derived health relationships. Its impact engine involves a reconstruction of the Bressler (2021) DICE-EMR mortality cost of carbon, decoupled from original assumptions and rebuilt to accept user provided inputs, an extension that makes the mortality cost of carbon a scenario dependent quantity at the project scale.
What is a metric tonne of CO₂?What is a Gtonne of CO₂?!
Every decision to build, or retire, a power plant is at once an economic, climate, and public health decision; and because the burdens and benefits land on different people, it's also a justice decision. Yet the metrics we usually offer (tonnes of CO₂, levelized cost) don't answer the questions communities are asking:
How many fewer asthma attacks in this school district?
How much sea level rise will it have averted by the time I can finally afford to buy a house?
What changes here, for people, if this plant is built, extended, retired, or replaced?
These questions have answers. The data, models, and assumptions required to produce those answers live in different scientific literatures, are maintained by different agencies, use different geographies, and have not been wired together in a form usable for project level screening. Up until now.
Technologies: generation, demand & fossil reductions ▸
- Onshore Wind
- Utility-scale land-based wind feeding the regional grid. Default capacity factor ~37%.
- Offshore Wind
- Utility-scale turbines in coastal or Great Lakes waters, cabled to shore. Default ~42%.
- Utility PV
- Large ground-mounted solar selling directly to the bulk grid. Default ~25%.
- Utility PV + Storage
- Utility solar paired with batteries that shift output across hours. Storage shifts timing, not annual energy. AVERT publishes no rate of change for storage-paired profiles, so the AVERT schemes are not offered.
- Distributed PV
- Customer-sited rooftop or carport solar on homes, businesses, or campuses. Default ~20%.
- Distributed PV + Storage
- Rooftop solar with behind-the-meter batteries for on-site use and limited export.
- Portfolio EE
- A mix of energy-efficiency measures (lighting, HVAC, industrial) that permanently lower electricity use.
- Uniform EE
- Idealized efficiency that cuts demand by the same amount in every hour of the year.
- Uniform Load Growth
- New round-the-clock demand: data centres, electrification, new industrial load. Uniform EE with the sign reversed, so results are negative because emissions are added, not avoided.
- Nuclear
- Firm, low-carbon baseload generation. Default ~91%.
- Gas Reduction
- A reduction in fossil gas-fired generation (retiring or curtailing a gas plant). Credits the plant's own emission rate, net of the chosen replacement. Illustrative default 100 MW; offered only where a gas fleet exists.
- Coal Reduction
- A reduction in coal-fired generation. Credits the plant's own (higher) emission rate, net of the chosen replacement. Illustrative default 300 MW; offered only where a coal fleet exists.
- Oil Reduction
- Retiring or curtailing oil-fired generation. Rare on the mainland but material in Hawaii and Puerto Rico, where oil still carries real load. Like the other fossil reductions it needs a replacement resource, and the plant’s own rate is held fixed while the backfill grid decarbonizes.
Emissions-projection schemes ▸
How the grid your project displaces is assumed to change over the project's life. Not every scheme applies to every technology: the AVERT schemes need a published rate of change, and a specific gas or coal plant's own rate is held fixed regardless. Where a scheme cannot be applied the result reads N/A rather than a substituted number. For a gas, oil, or coal reduction it instead sets how fast the replacement (backfill) grid decarbonizes; the reduced plant's own rate stays fixed.
- Constant
- Today's grid emission rates held fixed for all future years. No policy or technology change assumed.
- Regional Goals
- Rates decline in a straight line from today to the region's stated clean-energy or decarbonization target years.
- AVERT 3 Year Linear
- Takes AVERT's 3-year rate of change and removes the same absolute amount every year. Bounded, and reaches zero on a datable year.
- AVERT 7 Year Linear
- The same arithmetic over a longer 7-year window.
- AVERT 3 Year Regression
- Fits a log-linear trend instead of taking the two endpoints, and compounds it, so the rate approaches zero without reaching it.
- AVERT 7 Year Regression
- The same fit over seven observations. The better-estimated slope, since it is not hostage to one noisy endpoint year.
- Cambium Mid-case
- NREL's central scenario for technology costs, fuel prices, demand, and policy (as of Aug 2024). Not available for National, AK, HI, PR.
- Cambium Low RE Cost
- Mid-case with cheaper renewables and batteries, so faster clean deployment and steeper emissions declines.
- Cambium High RE Cost
- Mid-case with costlier renewables, so slower deployment and flatter declines.
- Cambium High Demand
- Mid-case with higher electricity demand (e.g., faster electrification), which shifts the pace and shape of grid change.
- User Linear
- You set the annual decline yourself. The rate falls each year by a fixed percentage of the 2023 rate, so it is a straight line and reaches zero after 100 divided by your rate, in years. Available everywhere, including nationally, where Cambium is not published.
- User Geometric
- The same idea, compounding: the rate falls each year by a fixed percentage of the previous year, approaching zero without reaching it. One rate is applied to all six pollutants, which is a simplification: the measured trends differ by species, and sulfur dioxide has been falling several times faster than carbon dioxide.
Replacement resource (gas, coal & oil reductions only) ▸
When fossil generation is removed, something serves that load instead. This sets what, which determines whether the result is net or gross. The reduced plant's own emission rate is held fixed over its life; the projection scheme decarbonizes only the backfill grid.
- Grid's current marginal mix (net, default)
- Subtracts what the grid actually dispatches on the margin to backfill the lost output. The honest net effect. Can be negative if the backfill is dirtier than the plant reduced, which is common for gas and rare for coal or oil.
- Efficiency or zero-emission (gross)
- Assumes the lost output is met by efficiency or zero-carbon resources, so the plant's full emission rate is credited. The optimistic bound.
- Grid average intensity (benchmark)
- Nets against the fleet-average intensity. A benchmark, not a dispatchable backfill; the average includes zero-carbon generation that would not actually ramp to replace the plant, so it overstates the benefit.
Shared Socioeconomic Pathways (SSP) ▸
The global development and climate future used to scale population and climate-related impacts.
- SSP1-1.9
- Sustainable, low-inequality world with very aggressive mitigation (about a 1.5°C class outcome).
- SSP1-2.6
- Same sustainable path, slightly weaker mitigation (about a 2°C class outcome).
- SSP2-4.5
- Middle-of-the-road development and policy with intermediate emissions. A common default.
- SSP3-7.0
- Fragmented, regionally focused world with weak climate policy and high emissions.
- SSP5-8.5
- Fossil-fuel-intensive growth with very high emissions.
Capacity, capacity factor, start year & lifetime ▸
- Installed Capacity (MW)
- The project's nameplate size.
- Capacity Factor (0 to 1)
- The fraction of the time the project runs at full output over a year. A 0.91 nuclear plant generates far more energy per MW than a 0.25 solar plant.
- Start Year
- When the project comes online (2015 to 2100). Sets where on the projection and SSP timelines the analysis begins.
- Expected Lifetime (years)
- How long the project operates. Avoided impacts accumulate across this whole period.
Advanced: custom scenario ▸
Off by default. Switching it on reveals four things, for checking this tool against another, for a project whose emission rates you already know, or for a grid outside the United States.
- Pick a specific region
- Normally a state that spans more than one AVERT region or Cambium area uses a blend of them. Here you can choose one, to model a project sited within it.
- Enter your own emission rates
- Three different rates can be replaced, in tonnes per MWh, and they are not the same number. The displaced rate is the grid your project pushes off the system. The plant rate is the fossil unit being retired, if that is what you are modelling. The fleet average rate is the regional average, used only when the replacement is the grid average. Leave a pollutant blank to keep the published value.
- Custom grid
- A grid this tool does not cover. You supply all six displaced rates and the tool does the rest. Emissions, climate and the equivalencies are calculated normally. Health is not available on a custom grid, and this is a limit rather than an oversight: the health numbers depend on where the emissions happen and who lives downwind, which no emission rate can supply.
- Avoided emissions
- Greenhouse gases (CO₂e) and criteria air pollutants the project keeps out of the air, per year and over its lifetime.
- Equivalencies
- The same avoided CO₂e expressed in everyday terms: tree seedlings grown, homes' annual energy use, gasoline gallons, miles driven.
- Health outcomes
- Avoided pollution-related deaths, asthma symptoms, and hay-fever cases in the affected region, plus the monetized value of those avoided damages.
- Climate & heat
- The project's small contribution to slowing temperature rise, the avoided heat-related deaths worldwide, and people spared from unprecedented heat exposure.
- Sea level
- Avoided land-ice mass loss tied to the project's avoided warming.
- Share a link
- Copies a link that reopens the exact scenario, including anything entered under Advanced, so a result can be cited or checked by someone else.
The tool links five main components:
- Electricity system baselines and marginal emissions (EPA AVERT, EPA eGRID, NREL Cambium)
- Policy and scenario pathways (simple projection schemes, regional goals, and Cambium scenarios)
- Demographic and climate futures (SSPs, UN WPP / IIASA WiC population)
- Reduced-form air-quality and health damages (EPA COBRA, resolved by source state)
- Health coefficients are specific to the state the emissions come from, and the tool reports both the nationwide total and the share of it falling inside that state. Geography matters a great deal here: the same project can deliver several times more or less health benefit depending on how many people live downwind. EPA's COBRA does not model Alaska, Hawaii, or the territories, so those jurisdictions read N/A for health rather than borrowing a national average.
- Avoided deaths are reported as a range rather than a single number. The lower figure follows Wu et al. (2020) and the higher follows Pope et al. (2019); EPA reports these two studies separately and does not average them.
- Climate-mortality and physical-climate responses (DICE-EMR-style MCC, temperature, ocean, and sea-level modules)
Each run follows one path. Your project becomes an annual change in generation or load. That change is multiplied by marginal emission rates that evolve along the projection scheme you pick, giving avoided emissions year by year. For a gas, oil, or coal reduction the logic flips: the tool credits the plant's own (fixed) emission rate and subtracts a backfill rate that decarbonizes along the chosen scheme, so the result is net of replacement and may be negative. Those avoided emissions then feed the health, climate, temperature, ocean, and sea-level modules to produce the downstream results. The method is reduced-form: deterministic formulas chained together for fast, transparent, screening-level estimates rather than a full process-based simulation.
Sources: EPA AVERT (2024), EPA eGRID (2023), NREL Cambium 2024, EPA COBRA (2023 baseline, state-resolved, 3% discount rate), IPCC AR6 SSPs, UN WPP 2024, with the mortality cost of carbon after Bressler (2021) and human-niche framing after Lenton et al. (2023).
Technical reference: McNevin, J. (2026). The EMPATHI Framework for Clean Energy Impact Assessment: Technical Reference Zenodo. doi.org/10.5281/zenodo.21347959 CC BY 4.0. The methodology of record: data sources, governing equations, parameters, validation and limitations for every module.
what are the benefits of clean energy on your grid?
interested in learning more about it?
interested in a consultation?
interested in professional access?
