EPA Regulatory Skill (Envirofacts + ECHO + GHGRP + Subpart W)

EPA Regulatory Skill (Envirofacts + ECHO + GHGRP + Subpart W)

One skill, four subsystems. Route by intent to the correct data source.

---

Credential Handling

No API key is required today. All four subsystems return data without authentication. The old build plan referenced an api.data.gov key for Envirofacts; testing in 2026 confirmed it is not enforced.

Contingency credential resolution (if EPA adds key requirements):

1. ~/.config/epa/credentials — parse api_key=<value>
2. EPA_API_KEY env var — fallback
3. Prompt user — "EPA may now require an API key. Get one free at https://api.data.gov/signup/ — store in ~/.config/epa/credentials as api_key=YOUR_KEY (chmod 600)."

Never hardcode or log keys. If activated, pass as ?api_key=<KEY> query parameter.

Reading the credentials file (bash):

KEY=$(grep '^api_key=' ~/.config/epa/credentials 2>/dev/null | cut -d= -f2)
[ -z "$KEY" ] && KEY="${EPA_API_KEY}"

Reading the credentials file (Go):

func resolveAPIKey() (string, error) {
    if home, err := os.UserHomeDir(); err == nil {
        creds := filepath.Join(home, ".config", "epa", "credentials")
        if data, err := os.ReadFile(creds); err == nil {
            for _, line := range strings.Split(string(data), "\n") {
                if strings.HasPrefix(line, "api_key=") {
                    return strings.TrimPrefix(line, "api_key="), nil
                }
            }
        }
    }
    if k := os.Getenv("EPA_API_KEY"); k != "" {
        return k, nil
    }
    return "", fmt.Errorf("no EPA API key found")
}

---

Mode Selection (Intent Routing)

Use this lookup table first. Match the user's question to a mode, then follow that mode's workflow section.

CRITICAL: UIC Class II wells are NOT on Envirofacts

The UIC_WELL, UIC_VIOLATION, and SDWIS_WATER_SYSTEM tables return {"error": "...The table is not available."} on the current Envirofacts endpoint. Do not query them. Instead:

State UIC programs for oilfield-relevant Class II:

Direct users there for UIC well-level data. Use ECHO for compliance flags.

---

Mode 1: Envirofacts

API structure

Base URL: https://data.epa.gov/efservice/

The legacy enviro.epa.gov/enviro/efservice/ domain now 301-redirects here. Always use data.epa.gov.

URL pattern:

GET https://data.epa.gov/efservice/{TABLE}/{COL}/{VAL}/.../rows/{S}:{E}/{FMT}

Comparison operators (between column and value):

Verified working tables

Tables that return 404 on this endpoint

See references/envirofacts_tables.md for complete column listings.

Workflow

1. Match user intent → choose table
2. Build URL: chain COL/VAL pairs, add rows/0:99, add /JSON
3. curl -sL <URL> — response is a flat JSON array
4. Aggregate / format / present

Example — TRI facilities in Monongalia County, WV:

curl -sL "https://data.epa.gov/efservice/TRI_FACILITY/STATE_ABBR/WV/COUNTY_NAME/MONONGALIA/rows/0:50/JSON"

Example — FRS cross-reference for a registry ID:

curl -sL "https://data.epa.gov/efservice/FRS_PROGRAM_FACILITY/REGISTRY_ID/110000344896/rows/0:20/JSON"

Example — TRI barium releases in WV:

curl -sL "https://data.epa.gov/efservice/TRI_REPORTING_FORM/STATE_ABBR/WV/CAS_CHEM_NAME/CONTAINING/BARIUM/rows/0:20/JSON"

Response shape

Flat JSON array at root — no wrapper, no pagination metadata. Detect more pages by comparing returned row count against your requested range.

[
  {"tri_facility_id": "26504PTRTM1090C", "facility_name": "PATRIOT MINING...", ...},
  {"tri_facility_id": "26504PTRTM12MIA", "facility_name": "PATRIOT MINING...", ...}
]

---

Mode 2: ECHO

API structure

Base URL: https://echodata.epa.gov/echo/

ECHO uses a two-step query pattern: search returns a QueryID, then retrieve pages through get_qid.

Common search parameters

Workflow

# Step 1 — search
RESULT=$(curl -sL "https://echodata.epa.gov/echo/cwa_rest_services.get_facilities?output=JSON&p_st=WV&p_co=MONONGALIA")
QID=$(echo "$RESULT" | jq -r '.Results.QueryID')
TOTAL=$(echo "$RESULT" | jq -r '.Results.QueryRows')
echo "QueryID: $QID, Total: $TOTAL"

# Step 2 — retrieve page 1 (20 records)
curl -sL "https://echodata.epa.gov/echo/cwa_rest_services.get_qid?output=JSON&qid=$QID&pageno=1&pagesize=20"

Always:

• Check .Results.Message (Success or Working both indicate valid)
• Check .Results.Error.ErrorMessage if present
• Read .Results.QueryRows for total count
• Page with pageno / pagesize on get_qid

Response shape

{
  "Results": {
    "Message": "Working",
    "QueryRows": "2153",
    "QueryID": "357",
    "PageNo": "1",
    "Facilities": [
      {"CWPName": "FACILITY NAME", "SourceID": "WV0045233",
       "PercentPeopleOfColor": "12.4", "FacPopDen": "45.2", ...}
    ]
  }
}

Useful ECHO patterns

Oil & gas facilities by SIC:

# SIC 1311 = Crude Petroleum and Natural Gas
curl -sL "https://echodata.epa.gov/echo/cwa_rest_services.get_facilities?output=JSON&p_st=WV&p_sic=1311"

Drinking water systems in a county:

curl -sL "https://echodata.epa.gov/echo/sdw_rest_services.get_systems?output=JSON&p_st=WV&p_co=MONONGALIA&p_pswsid_code_type=community"

Facilities in violation for 4+ quarters:

curl -sL "https://echodata.epa.gov/echo/cwa_rest_services.get_facilities?output=JSON&p_st=WV&p_qiv=GT4"

Environmental justice fields (in all ECHO program results)

• PercentPeopleOfColor — percent people of color in nearby area
• FacPopDen — population density around facility
• AcsPopulationDensity — ACS population density
• PercentLowIncome — percent low-income households

Use to flag EJ concerns near disposal sites, gas plants, processing facilities.

---

Mode 3: GHGRP

API structure

Base URL: https://data.epa.gov/efservice/

Same Envirofacts URL pattern as Mode 1. GHGRP uses dedicated views.

Primary table: V_GHG_EMITTER_SUBPART — facility emissions broken down by subpart, gas, and year. This is the correct table — the old PUB_FACTS_SUBPRT_GHG_W table referenced in pre-consolidation skills returns 404 and must not be used.

Secondary table: V_GHG_EMITTER_GAS — facility emissions by gas only (no subpart breakdown). Good for total facility emissions.

Tables that are NOT served:

• V_GHG_EMITTER_FACILITIES — 404
• V_GHG_EMITTER_SECTOR — 404
• PUB_FACTS_SUBPRT_GHG_W — 404 (do not use)
• PUB_FACTS_FACILITY_GHG_EMITTER — 404

See references/ghgrp_api.md for full column reference and all subpart codes.

Key columns (V_GHG_EMITTER_SUBPART)

Use co2e_emission for ranking and totals. Use ghg_quantity for gas-composition analysis (e.g., tons of methane leaked).

Each facility appears as multiple rows (one per gas per subpart). Aggregate by facility_id before ranking.

Key subparts

Workflow

1. Check count first for large queries: .../count/JSON
2. Default to most recent complete year (current year minus 2 is safe)
3. Fetch with appropriate subpart, state, year filters
4. Aggregate by facility_id
5. Rank by total co2e_emission
6. Present table + gas breakdown

Get count first:

curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/SUBPART_NAME/W/YEAR/2022/count/JSON"
# Returns: [{"TOTALQUERYRESULTS": 5782}]

Subpart W in WV, 2022:

curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/SUBPART_NAME/W/STATE/WV/YEAR/2022/JSON"

Methane only, specific state:

curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/SUBPART_NAME/W/STATE/PA/YEAR/2022/GAS_CODE/CH4/rows/0:99/JSON"

Search by operator substring:

curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/FACILITY_NAME/CONTAINING/EQT/SUBPART_NAME/W/YEAR/2022/JSON"

Aggregation (bash + jq):

curl -s "URL" | jq '[group_by(.facility_name)[] |
  {facility: .[0].facility_name, state: .[0].state,
   total_co2e: ([.[].co2e_emission] | add)}] |
  sort_by(-.total_co2e) | .[:10]'

---

Mode 4: Subpart W Deep-Dive

Subpart W is a GHGRP subset — use V_GHG_EMITTER_SUBPART with SUBPART_NAME/W. This mode adds industry context and segment-level interpretation.

Subpart W industry segments (40 CFR Part 98 Subpart W)

Largest Subpart W source nationally: pneumatic devices (~35-40% of reported CH4), followed by gathering compressors and storage tanks.

NOTE on segment-level detail: The Envirofacts API returns subpart-level totals. Segment-level breakdowns (production vs transmission compression within Subpart W) are only available via the EPA FLIGHT bulk download at https://ghgdata.epa.gov/ghgp/ — not this API. If the user needs segment-level detail, direct them to FLIGHT.

CH4 GWP reference

Envirofacts reports in mtCO2e. To convert to physical mass:

CH4 mass (t) = CH4 (tCO2e) / GWP

Example: 100,000 tCO2e CH4 at AR4 = 4,000 tonnes CH4.

National Subpart W benchmarks

Subpart W workflow

Standard GHGRP workflow (Mode 3) with SUBPART_NAME/W locked in. Add these aggregation steps:

Aggregate CH4 by facility:

curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/SUBPART_NAME/W/STATE/WV/YEAR/2022/GAS_CODE/CH4/rows/0:499/JSON" | jq '
  group_by(.facility_id) |
  map({
    facility: .[0].facility_name,
    county: .[0].county,
    total_ch4_co2e: (map(.co2e_emission | tonumber) | add)
  }) |
  sort_by(-.total_ch4_co2e) | .[:15]
'

Year-over-year trend (merge per-year requests):

for YEAR in 2018 2019 2020 2021 2022; do
  curl -s "https://data.epa.gov/efservice/V_GHG_EMITTER_SUBPART/SUBPART_NAME/W/STATE/WV/YEAR/${YEAR}/JSON"
done

The API does not accept year ranges in one call — issue one request per year and merge client-side.

---

Output Format

Match the house style: markdown table + narrative summary. Cap tables at ~20 rows. Always include units, year, and data caveats.

Envirofacts / ECHO example

## TRI Facilities — Monongalia County, WV

| Facility | TRI ID | City | Status | Parent |
|----------|--------|------|--------|--------|
| PATRIOT MINING CO INC OSAGE MINE | 26504PTRTM1090C | Star City | Active | ANKER GROUP |
| ... | ... | ... | ... | ... |

**Summary:** 4 TRI-reporting facilities in Monongalia County (EPA Region
03). Two coal mines (PATRIOT MINING) report zinc and other releases.
Cross-reference via FRS for CWA/RCRA status.

**Caveats:** TRI data is industry self-reported annually; release
quantities may be estimated (E), measured (M), or calculated (C).
Reporting thresholds apply — sub-threshold releases do not appear.

GHGRP / Subpart W example

## Subpart W Emissions — West Virginia (2022)

### Top facilities by total CO2e

| Rank | Facility | County | Total CO2e (MT) |
|------|----------|--------|-----------------|
| 1 | DTM Appalachia Gathering, LLC | Doddridge | 183,921.6 |
| 2 | HG Energy II Appalachia, LLC | Ritchie | 172,815.1 |
| ... | ... | ... | ... |

### Gas breakdown (all WV Subpart W)

| Gas | CO2e (MT) | % of total |
|-----|-----------|------------|
| Carbon Dioxide | 601,604.8 | 53.7% |
| Methane | 518,635.5 | 46.3% |
| Nitrous Oxide | 593.0 | <0.1% |

**Summary:** WV Subpart W facilities reported 1,120,833 MT CO2e in 2022
across 43 facilities. Methane accounts for 46% of CO2e despite being a
small fraction by mass (AR4 GWP=25). Top 3 facilities account for 43%
of state total. Gathering/processing dominates, consistent with WV's
role in Marcellus/Utica production.

**Units:** metric tons CO2-equivalent (MT CO2e), AR4 GWP basis.
**Reporting threshold:** 25,000 MT CO2e/yr facility-wide.
**Data vintage:** 2022 final.

---

Pagination

Envirofacts / GHGRP

all_rows = []
start = 0
page = 500
while True:
    rows = fetch(table, filters, start, start + page - 1)
    all_rows.extend(rows)
    if len(rows) < page:
        break
    start += page

Warn user if total > 5,000 rows; suggest narrowing by state, year, SIC, or subpart.

ECHO

pageno and pagesize on get_qid:

page = 1
while page * pagesize <= total:
    records = echo_retrieve(program, qid, pageno=page, pagesize=20)
    page += 1

---

Error Handling

---

Cross-Reference Strategy

EPA maintains facility IDs across programs. FRS Registry ID is the master cross-reference key:

FRS Registry ID (e.g., 110000344896)
  |-- TRI:   tri_facility_id  (via epa_registry_id in TRI_FACILITY)
  |-- NPDES: permit ID         (via FRS_PROGRAM_FACILITY, pgm_sys_acrnm=ICIS)
  |-- RCRA:  handler ID        (via FRS_PROGRAM_FACILITY, pgm_sys_acrnm=RCRAINFO)
  |-- SDWA:  PWS ID            (via FRS_PROGRAM_FACILITY, pgm_sys_acrnm=SDWIS)
  |-- ECHO:  DfrUrl            (uses registry_id as fid parameter)
  |-- GHGRP: facility_id       (match by name+location; no direct key)

Link a facility across programs:

1. Find facility in any table → get its FRS Registry ID
2. Query FRS_PROGRAM_FACILITY with that ID
3. Use returned program IDs to query specific tables or ECHO endpoints

---

EPA Region Reference

---

SIC / NAICS Codes for PNGE Research

---

PNGE Research Context

Carbon intensity of produced water operations

Subpart W covers produced-water-adjacent emissions: separators, tanks, pneumatics, flares at well sites. For produced water Li/Mg recovery TEA:

1. Query Subpart W for target basin to set a baseline facility-level O&G emissions benchmark
2. Compare against proposed DLE energy inputs
3. Remember: co-producing Li/Mg from existing brine handling may add minimal incremental emissions vs standalone hard-rock mining

Methane as the key concern

In Subpart W, methane typically accounts for 40-60% of CO2e despite being a small mass fraction (AR4 GWP=25). Key sources:

• Pneumatic controllers and pumps
• Equipment leaks (fugitive)
• Associated gas venting and flaring
• Liquids unloading
• Gathering and boosting compressor stations
• Produced water tanks (flash emissions)

Cross-skill patterns

• pnge-core:usgs-produced-waters — match GHGRP facility locations with produced water chemistry (Li/Mg vs emissions profile)
• pnge-state-regulatory:wvges-wells, pnge-state-regulatory:tx-rrc, pnge-state-regulatory:co-ecmc, pnge-state-regulatory:ok-occ, pnge-state-regulatory:nm-ocd, pnge-state-regulatory:nd-dmr, pnge-state-regulatory:calgem — link GHGRP facilities to state well records by operator + county
• pnge-federal-data:fracfocus — cross-reference GHGRP reporters with HF chemical disclosures
• pnge-core:eia-data — compare GHGRP emissions against EIA production volumes for emissions intensity (MT CO2e per barrel or per Mcf)
• pnge-federal-data:ejscreen-cejst-svi — overlay ECHO EJ fields with EJScreen

---

Caveats and Data Limitations

Across all modes

• Self-reported: Envirofacts TRI, GHGRP, and most ECHO program data are submitted by regulated entities. EPA verifies but errors and underreporting occur.
• Publication lag: TRI ~18 months, GHGRP ~9-15 months, ECHO ~1 quarter. As of April 2026, most recent complete GHGRP year is 2024 (released ~Oct 2025); 2025 data expected Oct 2026.
• Data currency: Envirofacts and ECHO refresh asynchronously; FRS is the most current cross-reference index.

Envirofacts specific

• fac_latitude / fac_longitude are DDMMSS integer format (e.g., 393800 = 39d 38m 00s). Use pref_latitude / pref_longitude (decimal degrees) when available.
• PCS tables are legacy — replaced by ICIS-NPDES. Use ECHO CWA for current data.
• UIC_WELL, UIC_VIOLATION, SDWIS_WATER_SYSTEM, RCRA_FACILITY, CERCLIS_SITE return 404 — use ECHO or state sources.

ECHO specific

• Compliance data updates quarterly; new violations may not appear for up to 90 days.
• Major/minor facility distinction (p_maj) is an EPA classification; small facilities may be under-represented in "major" queries.

GHGRP / Subpart W specific

• Reporting threshold 25,000 MT CO2e/yr facility-wide. Thousands of small O&G facilities are not in the dataset.
• Engineering calculations, not measurement. Subpart W allows emission factors + throughput rather than direct measurement. Independent atmospheric studies suggest actual emissions are 1.5-3x higher for some equipment categories.
• GWP basis: EPA uses AR4 (CH4=25). State and corporate inventories may use AR5 or AR6. Be explicit about which GWP is used when comparing values.
• Segment-level detail NOT in API. Use FLIGHT bulk downloads.
• Double counting risk: Facilities may report under multiple subparts (e.g., C combustion + W process). Sum across subparts carefully.
• Supplier vs. emitter: subparts UU, PP, MM are type S — they report quantities entering commerce, not direct atmospheric emissions. Do not add supplier and emitter subpart totals for facility emissions.
• 2024 EPA methane rule (40 CFR Part 98 revisions, effective 2025): Expands monitoring and revises emission factors. Reported 2025+ values may show step-change increases reflecting methodology change rather than real emission increases.
• Offshore: Subpart W covers OCS platforms but they report differently. Use pnge-federal-data:boem-offshore for offshore production context.

---

Implementation Notes

• Prefer bash with curl -sL + jq for all modes
• Go client examples in references/go_example.go
• Python stdlib client in references/python_example.py — covers Envirofacts, ECHO, and GHGRP
• Response types:
  • Envirofacts / GHGRP: flat JSON array at root
  • ECHO: nested JSON with .Results wrapper + .Error check
• co2e_emission and ghg_quantity in GHGRP responses come as numeric but may round-trip as strings; always tonumber in jq
• Cap output tables at 20 rows for readability; note total count
• Always state units (MT CO2e, lbs, count) and data year
• Note when a query returns facility-wide vs subpart-specific results