EPA ECHO Data Downloads (compliance/enforcement + RCRAInfo hazardous waste)

It is the widest single portrait of environmental compliance in the United States: one integrated row per regulated facility, over 1.5 million facilities wide and 130-plus fields deep, built by the agency that holds the enforcement record. Each row stacks three layers. Identity and geography first - facility name, street address through county, latitude and longitude on the NAD83 datum, and a persistent registry ID that stitches together every program database that ever touched the site. Then program membership - flags and identifiers showing which statutes the facility answers to, including its hazardous waste handler ID. Then the enforcement payload itself: inspections, informal and formal actions, penalty totals and dates, quarterly compliance status and the significant-noncomplier flag, all inside a five-year rolling window.

Around that spine sit the program tables. Six hazardous waste files - facilities, evaluations, violations, enforcements, industry classification and month-by-month status history - share one handler ID as their join key, so a single generator's decade of conduct reassembles from flat CSV shapes. Beside them: water discharge permits and effluent violations, discharge monitoring reports reaching back before FY2009, drinking water, air emissions, biosolids, combined sewer overflows, federal enforcement case files, facility demographics, tribal spatial linkages and national PFAS datasets. Where sibling EPA properties narrow their lens, this one keeps every program on one shelf. Get a sample of this dataset cut to the programs, states and handlers you care about before anything else.

Flat, wide and readable without a decoder ring. One integrated facility row, then its handler-side companions:

# echo exporter facility row - each value is the documented example for its field
REGISTRY_ID   : 110000490166
FAC_NAME      : ACME LANDFILL LLC
FAC_ADDRESS   : 1200 WEST CITY RANCH ROAD, PALMDALE, CA 93550  (LOS ANGELES county)
FAC_LAT/LONG  : 34.589, -118.100   (NAD83 decimal degrees)
RCRA_FLAG     : Y          RCRA_IDS: CAD982042076     PERMIT_TYPES: TSDF

# five-year compliance window riding on the same row
INSPECTIONS   : 4          INFORMAL ACTIONS: 1        FORMAL ACTIONS: 2
PENALTIES     : 25000      LAST PENALTY: 2021-06-15
COMPL STATUS  : V          SNC FLAG: N                3YR QUARTERS: VVV_

# handler-side rcrainfo tables, joined on ID_NUMBER
ID_NUMBER              : VTD000509174
FED_WASTE_GENERATOR    : 1            (large quantity generator)
OPERATING_TSDF         : LIT          (land disposal, incinerator, treatment)
VIOLATION_TYPE         : 265          (TSDF standards, 40 CFR part 265)
VIOLATION DETERMINED   : 2019-03-04

Read the top half as a sentence: a California landfill holding a hazardous waste permit took four inspections in five years, drew one informal response, two formal actions and a $25,000 penalty dated June 2021 - and its last three quarters ran compliant after a violation quarter, which is exactly the story the twelve-character quarter string tells at a glance. Read the bottom half as a join: the handler ID pulls generator class and operating sub-universes from the facilities table, and the violation table attaches dated determinations keyed to the CFR part violated. Every value above is the documented example for its field, so the shape is exact even before real handlers fill it.

Twenty-five fields define the core this page documents - the Exporter's facility spine and the hazardous waste columns that carry the enforcement payload - each definition checked against the agency's own column documentation during the August 2026 research pass. Three design choices make the schema unusually joinable: every facility carries a persistent registry ID that links program databases together, the handler ID keys across all six hazardous waste tables, and coordinates arrive in one datum (NAD83) so mapping layers stack without reprojection surprises.

Beyond the verified core, the wider program families fold under additional fields on request - their column sets are program-specific and quoted individually rather than guessed: permit-limit and effluent-violation columns dimensioned by outfall and parameter, discharge monitoring tables spanning fiscal years before FY2009 through FY2026, national PFAS analyte lists, stormwater compliance measures, biosolids parameters, sewer overflow event fields, drinking water columns, and facility demographics variables. Name the programs when you request a sample and the dictionary ships extended to exactly that scope.

Geography - the United States nationwide, tribal lands included, with proximity flags marking facilities near the US-Mexico border. Every row resolves to street address, city, county, state and postal code, and carries NAD83 coordinates precise enough to drop onto any mapping layer.

Temporal - the Exporter carries a five-year rolling compliance window per facility. Discharge monitoring tables reach further, spanning fiscal years before FY2009 through FY2026, and hazardous waste status history runs quarter by quarter with violation records month by month. One boundary is stated outright by the agency: data before November 2000 carries unassessed quality and should be treated as unknown.

Granularity - two grains that never mix. The Exporter holds one integrated row per facility. The program tables hold event-level rows: one per evaluation, violation, enforcement action or monthly status record. Aggregate the events onto the facility row and you have a panel; keep them separate and you have a case file.

On documentation depth, this record sits at the top of the catalog: 10/10 on the field-documentation rubric against an average of 7.81 across the 1,744 datasets tracked - one of the few to reach the ceiling, because every field definition here was verified against published column documentation rather than inferred from headers.

API, files, or your warehouse. Daily, weekly, or hourly.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; extraction, decoding and schema stability are ours. The facility spine arrives typed - statuses decoded beside their raw codes, coordinates in one datum, handler IDs normalized for joining - and scheduled captures accumulate into a panel a one-time pull cannot give you. A sample goes out first, sized to test in your own pipelines the same week.

• Facility-level risk panels - five-year windows joined onto handler IDs and coordinates score industrial assets on inspections, escalation and penalties for ESG screens and counterparty review.
• Enforcement-trend screening - informal-versus-formal separation turns a static flag into a sequence, exposing operators drifting from letters to orders to penalties.
• Territory planning - 1.5 million-plus facilities segmented by program, status and geography rank outreach territories for consulting, testing and remediation vendors.
• Due diligence - violations, penalty dates and determinations attach to named target assets ahead of transactions.
• Water-side compliance - permit, effluent-violation and overflow event tables extend the same join discipline to dischargers and collection systems.
• Accountability journalism - named operators with dated penalties and quarter-by-quarter status give investigators a citable federal record.

Data scientists and ML engineers get a facility-year panel with enums coded and coordinates attached; see data scientists use cases. Developers and data-product builders power vendor screens and location intelligence off stable registry keys; see developers builders use cases. Sales and growth teams aim compliance-services outreach at facilities actually accumulating actions; see sales growth teams use cases. Journalists, academics and students investigate named operators against the federal record; see journalists academics use cases. Market researchers take facility counts as market denominators, investors and quants read worsening quarters as operating risk, and competitive intel teams check a rival's plants before signing - see competitive intel product teams use cases.

Within this slice, every neighbor owns a different job. EPA Envirofacts Web Services is the query counterpart - nine program families answered row by row on demand - better for narrow lookups, lighter for whole-program depth. EPA GHGRP data holds annual greenhouse gas emissions for roughly 8,000 reporters: what facilities emit, where this set records how they behave. EPA Superfund data (SEMS) covers the cleanup ledger - contaminated sites, remedies and boundaries - rather than the operating-permit universe. And when the question drifts toward buildings rather than facilities, the head-to-head on scope and cadence sits in the vs EIA CBECS comparison.

This record owns breadth-with-depth: every regulated program, every state, event-level history, one facility spine. Nothing else in the shelf answers 'how has this operator behaved under regulation for the last five years' without stitching three sources together.

Four things, stated plainly.

First, old records carry a stated limit: anything before November 2000 has unassessed quality by the agency's own caution. Treat those cells as unknown rather than zero, especially in trend work.

Second, statuses arrive coded. Quarterly compliance letters, generator classes 1/2/3/N and treatment sub-universe letters ship exactly as published; the decoding tables ride with your delivery, so analysis never rests on a guessed abbreviation.

Third, verification depth varies by family. The facility spine and the hazardous waste columns were verified field-by-field during the August 2026 pass; narrower program dictionaries - PFAS analytes, biosolids parameters, stormwater measures - fold under additional fields on request and are confirmed against live values when your sample is cut rather than asserted blind.

Fourth, scale changes tooling. At full depth the largest program extracts outrun spreadsheet capacity outright; deliveries land shaped for database or warehouse loading, which is our problem, not yours. Say which programs, states and handlers matter and the sample arrives cut to exactly that form.