Where do the CARMA plant numbers come from?
CARMA’s operating philosophy is to maximize possible transparency, so it’s only fair to shed some light onto how we calculated plant data. For the full CGD working paper, please read Calculating CARMA, Global Estimation of CO2 Emissions from the Power Sector by David Wheeler and Kevin Ummel.
CGD decided to focus on the power sector because it accounts for the most CO2 emissions (26%), and because it’s better-documented any other CO2-emitting sector. With that being said, it was no cakewalk to process emissions data from numerous sources, and we recognize that our findings are still far from perfect.
CARMA began with plant-level CO2 emissions publicly disclosed online by the US, EU, Canada, and India. This initial data source was combined with two others: the World Electric Power Plants Database (WEPP, a subscription service), and country-level power production data from the US Energy Information Agency. A fuzzy-logic algorithm, suggested by CGD colleague David Roodman, was used to stitch together the data sources by matching the (frequently inconsistent) identifier codes. The fuzzy-logic algorithm, along with a painstaking visual analysis, formed the basis for our project.
CARMA ended up with 51,373 total power facilities. Of those plants, 26,034 emit no CO2, and 2,922 have publicly-reported CO2 emissions. The CO2 for the remaining 22,417 plants are estimates based on a regression analysis of 2,469 CO2-emitting facilities in the US. As for power production, the public data sources account for 4,071 facilities (3,869 in the US, 202 in India). The power production for the remaining 47,302 plants are estimates based on the following 5-step process:
1. Estimate capacity factors using a regression analysis based on US facilities
2. Multiply estimated capacity factors by WEPP-reported operational plant capacities
3. Combine estimated and reported power to obtain total power production by energy source for each country
4. For each energy source, divide the total by the corresponding total from the US Energy Information Agency to obtain an adjustment factor
5. For each country and energy source, multiply each estimated power output by the relevant adjustment factor (if the power output isn’t publicly-reported)
A plant’s power production in megawatt-hours (MWh) is the product of its capacity factor (% of potential capacity actually used), its capacity (MW), and its potential operating hours per year (usually 24 * 365 = 8,760). Emissions intensity for each plant is calculated by dividing CO2 emissions by power production.
CARMA uses regression models fitted to plant-level variables in a large US dataset. We estimate emissions and power capacity for plants from very detailed information about their capacity, generator age, combustion technology, and energy sources. CARMA’s plant-level emissions and power reports are long-run average estimates based on facilities under standard operating conditions and with similar variables to plants in the US dataset.
To attach geographic coordinates to plants, Kevin and David sifted through several sources, including the US EPA’s eGRID database, MaxMind’s World Cities database, and various European public disclosure databases. CARMA’s goal was not to provide precise geographic coordinates for local mapping, but to allow for regional- and global-scale mapping to reveal broad patterns in the spatial distribution of plants. Another fuzzy-logic algorithm was used to match city names from the WEPP database to the MaxMind cities database. All said and done, about 60% of all plants in CARMA have coordinates. As it states in the working paper, “the geographic totals in CARMA do not measure emissions related to the consumption of people within a region; they simply measure emissions produced by power plants physically located within that region”.