Who owns the wires, who serves the load, and where the plants sit.

Emission intensity: method and data notes

Source

EPA's Emissions & Generation Resource Integrated Database (eGRID). eGRID combines EIA-923 generation and fuel use, EIA-860 plant data and the emissions that power plants monitor and report to EPA under 40 CFR Part 75, and estimates the rest from fuel use and emission factors. The latest edition is eGRID2023 (revision 2, June 2025), built on 2023 EIA data. EPA had not published eGRID2024 by 23 October 2026, so 2024 comes from Cornerstone's run of EPA's own code (below). History comes from EPA's eGRID2014, 2016 and 2018 to 2022 files (eGRID was published every two years before 2018).

What a rate means

The total output emission rate is an area's annual emissions divided by its annual net generation, in pounds (or kilograms) per MWh. Biomass CO2 counts as zero and combined heat and power plants keep only the share of emissions that makes electricity, as in eGRID. Rates are at the point of generation: they count the plants located in an area, not the power its customers import, and leave out transmission and distribution losses and upstream fuel emissions. A subregion card also shows the rate per MWh delivered, using EPA's grid gross loss for the interconnection, as EPA's Power Profiler does. The non-baseload rate weights plants by how rarely they run (all output of plants under a 20% capacity factor, none above 80%). EPA offers it for estimating emissions avoided by efficiency or renewables, not for carbon footprints.

How the numbers were checked

Every rate was recomputed from eGRID's mass emissions and net generation for the US, each subregion and each balancing authority in all nine years, and each matches EPA's published rate within rounding. Subregion and balancing-authority generation each add up to the US total. Two series were made comparable: eGRID2014 published CH4 and N2O rates in lb/GWh (converted to lb/MWh here), and EPA's CO2e uses different warming potentials by edition (SAR through eGRID2016, AR4 for 2018 to 2022, AR5 for 2023), so CO2e here is recomputed with the AR5 values (CH4 28, N2O 265) in every year.

2024 from Cornerstone

Cornerstone ran the development branch of EPA's open-source eGRID code (github.com/USEPA/egrid, commit 8932a07) on EIA and EPA data retrieved on 11 March 2026 and published the output on Zenodo (record 18968658, version 1.0.0, 17 March 2026, CC BY 4.0). It is not an EPA release and EPA has not reviewed it. Cornerstone says it checked that the code reproduces eGRID2023 but did no quality assurance of its own. The file has the same 27 subregions and 66 balancing authorities as eGRID2023, its rates follow from its masses, and its generation adds up to the US total. It is in metric units, converted here. 2024 carries an asterisk on the year slider, hollow dots in the charts and a note in every card. US CO2 is 747.9 lb/MWh in 2024 against 767.2 in 2023 (2.5% lower). Three things differ from EPA's years:

• Non-baseload rates are recomputed from the plant sheet, with emissions and generation each weighted by the plant's nonbaseload factor, because the file's own are inconsistent (below). The same method reproduces EPA's 2023 non-baseload rates within 0.04%.
• SO2 is not shown for 2024 (below).
• NOx, CH4 and N2O include landfill gas, which EPA's 2023 file sets to zero. US NOx is 0.464 lb/MWh in 2024 against 0.452 in 2023; without the 47,589 tons from plants that mainly burn landfill gas it would be 0.442, a fall.

Balancing authorities on the map

eGRID assigns each plant to a balancing authority from EIA-860 but publishes no boundaries, so balancing authorities are drawn on the HIFLD planning-area polygons whose code matches. Of the 95 polygons, 53 can be matched to one of the 66 balancing authorities in eGRID2023: 49 by their own code and 4 by hand. Sierra Pacific's polygon is shown with NV Energy's single balancing authority (NEVP); the Alabama Power, Georgia Power and Mississippi Power polygons with the Southern Company balancing authority (SOCO); and HIFLD's combined PacifiCorp polygon with PACE and PACW added together. That covers 50 of the 66. The other 16 balancing authorities, such as Avangrid Renewables, Hawaiian Electric and the Southeastern Power Administration, appear only in the ranked list. Where polygons overlap, the smaller one is drawn on top.

Draft PM2.5, ammonia and VOC

EPA published draft estimates of primary PM2.5, NH3 and VOC for 2018 to 2021 from the National Emissions Inventory but has not added them to eGRID. Here the subregion totals are rebuilt from the drafts' plant sheets and divided by eGRID's generation for each year, because the drafts' own subregion sheets contain the errors listed below. National PM2.5 roughly halves in 2020 (0.089 lb/MWh in 2019, 0.050 in 2020, 0.060 in 2021) and VOC in 2018 is about 2.4 times later years; the drafts do not explain these steps, so compare those years with care.

Errors found in the source files

• NH3 and VOC drafts, 2019: the subregion and state sheets reuse 2018 generation (AKGD shows 4,641,060 MWh in both years; eGRID2019 has 4,513,906), which shifts their 2019 rates by -3% to +18%.
• NH3 draft, 2020: the plant, state and subregion columns are labeled as PM2.5 (PLPM25AN and similar) although they hold ammonia.
• NH3 and VOC drafts, 2021: the state sheets hold 11,395 rows of plant data instead of 52 state rows.
• PM2.5 draft, 2019: generation for MROW, SPSO and SRMW differs from eGRID2019 by up to 0.8%.
• PM2.5 white papers (2020 and 2021): both say the most recent National Emissions Inventory is for 2017, and both match units with the 2016v1 modeling platform crosswalk although the data are for 2020 and 2021.
• eGRID2023 Technical Guide, Table 3-2: the biomass CO2 is 65,289 tons in the table but 62,289 in step 1 of the worked example (75,817 minus 65,289 gives the 10,528 shown). Step 3 says the CHP adjustment is found by subtracting the final emissions from the biomass emissions; it is subtracted from the emissions after the biomass is removed (10,528 minus 2,094 gives 8,434).
• Power Profiler terms note (revised 6 January 2025): its national averages of 823.1 lb CO2/MWh and 0.473 lb SO2/MWh are eGRID2022 values (eGRID2023: 767.2 and 0.359), its 5.0% loss factor compares with 4.2% in eGRID2023, and it gives home use as 12.194 MWh in one place and derives 11.874 MWh in another.
• HIFLD planning areas: 13 of the 95 polygons have no code, two codes are each used twice (MPC for Minnkota Power and Mississippi Power, GPC for Gulf Power and Georgia Power), Black Hills Corporation carries BPA, and several names predate mergers (Gulf Power, SCE&G, Dayton Power & Light).
• Cornerstone eGRID2024, non-baseload: plant non-baseload generation (PLNGENNB) sums to 1,020 TWh, but net generation times each plant's own nonbaseload factor gives 1,215 TWh, while non-baseload emissions use the factor. The file's non-baseload rates are therefore too high: 1,622 lb CO2/MWh for the US against 1,368 recomputed, and up to 48% off for some areas.
• Cornerstone eGRID2024, SO2: for units whose SO2 is estimated from emission factors, SO2 per unit of fuel is 182 times the 2023 value for gas engines, 130 times for other biogas engines, 26 times for landfill-gas engines and 0.18 times for gas turbines. Port Westward Unit 2 (Oregon, gas engines) goes from 16 to 2,671 short tons on almost the same generation, and SO2 rates for CAISO, Portland General and Idaho Power rise 6 to 22 times.
• Cornerstone eGRID2024, CO2e: the Zenodo description says AR5 (CH4 28, N2O 265), but the comparison workbook published with it computes CO2e with 27 and 273, and the file's own CO2e totals are 0.02% above CO2 plus 28 CH4 plus 265 N2O. CO2e here is recomputed.
• eGRID2022 and eGRID2023, Curtis H. Stanton Energy Center (Florida): its units report 4,230 (2022) and 3,522 (2023) short tons of NOx, but the plant record has 0, and so does every total built from it. Florida Municipal Power Pool NOx shows 0.068 lb/MWh in 2023 instead of about 0.59, and FRCC 0.242 instead of about 0.270. MAS ASB Cogen (Southern Company) drops 589 tons the same way. The layer shows EPA's values as published.
• eGRID2023 Technical Guide, landfill gas: section 2.2 says eGRID2023 no longer adjusts landfill-gas NOx, SO2, CH4 and N2O, but Table C-4 still lists NOx, CH4 and N2O as adjusted, and the 2023 data follow the table (302 plants that mainly burn landfill gas report zero NOx).
• eGRID2023 metric workbook: the US sheet's annual non-baseload generation is 4,347,119,112 MWh, more than all US generation. It is the GJ figure; the standard workbook has 1,207,533,087 MWh.
• This site: it described the 95 planning areas as 95 balancing authorities. Corrected on 23 October 2026.

ERCOT prices: method and data notes

Source

ERCOT's Public Data API: real-time settlement point prices (NP6-905-CD, every 15 minutes) and day-ahead settlement point prices (NP4-190-CD, hourly) for the trading hubs and load zones, from January 2022. Where ERCOT posted an interval more than once, the latest posting is used. The data were pulled for Virya's battery-bidding benchmark and are summarized here, not redistributed as raw files.

Measures

• Real-time price: the mean of each hour's four 15-minute real-time prices, in $/MWh.
• Real-time minus day-ahead: that hour's real-time price minus its day-ahead price. Above zero means real time came in dearer than the day-ahead market expected.
• Daily battery spread (TB2): the mean of the day's two highest hourly real-time prices minus the mean of its two lowest. It is a common yardstick for what a 2-hour battery could earn per MW in a day with perfect foresight. It leaves out losses and ignores that a battery must charge before it discharges, so it overstates what a real battery earns.

Hours are hour-ending Central Standard Time all year (UTC-6), so every day has 24 hours and the daylight-saving changes neither drop nor repeat an hour. For price and battery spread, the map's classes are quantiles of every zone's yearly value across all years, so colors compare from year to year. For real-time minus day-ahead they are fixed bands centered on zero, blue where real time came in cheaper. The heat maps use fixed price bands. The latest year is a part year, marked with an asterisk.

Load zones on the map

ERCOT publishes its load-zone map only as an image. The outlines here were traced from that image (ERCOT media kit, June 2023) after lining it up with the Census outline of Texas, so they are approximate, especially where a zone boundary runs between counties. The North, South, West and Houston zones cover the competitive areas. The Austin Energy, CPS Energy, Lower Colorado River Authority and Rayburn Electric Cooperative zones are the loads those utilities serve, so the LCRA and Rayburn zones are scattered patches inside the others. Parts of Texas outside ERCOT (the Panhandle, El Paso and East Texas) are not colored. Trading hubs are averages of ERCOT's 345 kV buses in each region and have no outline; they appear in the panel list.