What CarbonScore measures
CarbonScore answers a direct question: how much carbon dioxide did this resource prevent from being emitted?
Specifically, CarbonScore is the kilograms of CO2 avoided per megawatt-hour of energy saved or generated by a distributed energy resource. It reflects both the reduction in onsite emissions (for resources that displace fossil fuel combustion) and the carbon impact on the electric grid.
A CarbonScore of 400 kg CO2/MWh means that for every megawatt-hour this resource saves or generates, 400 kilograms of carbon dioxide are kept out of the atmosphere. A CarbonScore of 50 kg CO2/MWh means the resource operates in a context where the net carbon benefit per unit of energy is smaller.
Two sources of carbon impact
Distributed energy resources affect carbon emissions in two ways, and CarbonScore accounts for both.
Onsite emissions. Some resources directly eliminate fossil fuel combustion. A heat pump replacing a gas furnace stops burning natural gas at the building. A heat pump water heater replacing a propane unit stops burning propane. These onsite emissions reductions are straightforward to calculate using standard combustion factors for the fuel being displaced.
Grid impacts. Any resource that changes electricity consumption also changes the carbon footprint of what the grid produces. An efficiency measure that reduces electricity demand means less generation is needed. A heat pump that electrifies heating adds electricity demand. A solar panel displaces grid generation. The carbon impact of these grid changes depends on how carbon-intensive the grid’s electricity is at the time.
CarbonScore combines both: the onsite emissions avoided minus the grid emissions associated with any change in electricity consumption.
How grid carbon intensity is calculated
This is where methodology matters. There are two established approaches to measuring grid carbon intensity, and they capture different things.
Average emissions rate reflects the carbon intensity of the grid as a whole for a given hour. It divides total emissions from all generators by total generation. This is an attributional measure: every kilowatt-hour consumed bears its proportional share of the grid’s total emissions. It captures the long-term, system-wide impact of changes in electricity consumption.
Marginal emissions rate reflects the carbon intensity of the generation that responds to changes in demand. When consumption increases or decreases, certain generators ramp up or down in response. The marginal rate captures the emissions intensity of those responsive generators. It reflects the short-term uncertainty that an intervention introduces into the system.
Each approach has strengths. Average emissions rates are stable, well-understood, and grounded in actual generation data. Marginal emissions rates capture the immediate dispatch consequences of a change in demand, but they are harder to pin down precisely because identifying the true marginal resource requires assumptions about grid operations that different providers resolve differently.
CarbonScore uses both. The grid carbon intensity applied in the CarbonScore calculation is the average of the marginal emissions rate and the average emissions rate for each hour. This blended approach reflects both the short-term and long-term impacts of the intervention, without over-indexing on either framework.
The full calculation
For a fuel-switching resource like a heat pump:
CarbonScore = (avoided fuel emissions - grid emissions from added electricity) / net energy changeWhere grid emissions are calculated using the blended hourly grid carbon intensity (average of marginal and average rates).
For an efficiency resource that reduces electricity consumption:
CarbonScore = Σ(hourly savings × blended hourly grid carbon intensity) / total savingsFor a generation resource like rooftop solar:
CarbonScore = Σ(hourly generation × blended hourly grid carbon intensity) / total generationBecause the calculation uses hourly data, it captures the fact that grid carbon intensity varies significantly throughout the day and across seasons. A kilowatt-hour saved at 2 PM when gas peakers are running has a different carbon value than one saved at 2 AM when the grid may be running cleaner baseload generation.
What drives high and low scores
CarbonScore is driven by two factors: the carbon intensity of any onsite fuel being displaced and the grid carbon intensity in the resource’s region.
Typically high CarbonScore:
- Heat pumps replacing oil or propane heating (displacing carbon-intensive fuels, even after accounting for added grid electricity)
- Efficiency improvements in regions with coal-heavy or gas-heavy grids
- Demand response that reduces consumption during high-carbon hours
Typically moderate CarbonScore:
- Solar in regions with mixed generation
- Heat pumps replacing natural gas in regions with moderate grid carbon intensity
Typically lower CarbonScore:
- Any resource in a region with very clean grid electricity (hydro-dominated Pacific Northwest, nuclear-heavy regions)
- Heat pumps replacing gas in regions where the grid itself is heavily gas-fired (the net carbon benefit is smaller when you are replacing gas combustion with gas-generated electricity)
These patterns shift over time as the grid evolves. A region that adds significant renewable generation will see grid carbon intensity decline, which changes CarbonScores for all resources in that area. The hourly calculation captures this automatically through updated emissions data.
Modeled vs. verified scores
Like GridScore, CarbonScore has two tiers.
Verified CarbonScores are calculated from actual meter data through Aristotle. They use real hourly savings or generation figures combined with hourly grid carbon intensity data for the resource’s specific location. These update continuously as new data arrives.
Modeled CarbonScores are estimated from the resource type, location, and typical performance characteristics. They provide a reasonable approximation but cannot account for the specific operating patterns of an individual asset.
On the DER Registry, verified scores carry a “Verified” badge and modeled scores carry a “Modeled” badge. Enrolling an asset in Aristotle upgrades its CarbonScore from modeled to verified.
Open methodology
CarbonScore’s calculation methodology is published as open methodology by the OpenEAC Alliance. The emissions data sources, combustion factors, blending approach, and calculation procedures are documented and subject to peer review.
This openness is important because carbon accounting is consequential. Organizations use these numbers for regulatory filings, investor disclosures, and public commitments. The methodology needs to withstand scrutiny, and anyone should be able to verify how a score was derived.