← The record · All reference data

RSEI toxic-release inventory (EPA RSEI Public Data Set)

Per-facility Risk-Screening Environmental Indicators (RSEI) results for the reading site’s own county, reduced from EPA’s RSEI Public Data Set. Every figure in a committed inventory was summed from RSEI rows — nothing is fabricated, inferred, or estimated by BOSC. Regenerate with watermark rsei.

What RSEI is

RSEI is EPA’s screening model over Toxics Release Inventory (TRI) reports. For each reported release it combines the pounds released, the chemical’s toxicity weight, and (for the Score) fate-and-transport + the surrounding population into two comparative measures:

• RSEI Score — modeled, population-weighted, unitless. Comparative only: it is not a risk estimate, a dose, or a concentration. Used to rank/triage.
• RSEI Hazard — pounds × toxicity weight, with no exposure/population term.
• Cancer / Non-cancer Score — the Score split by health endpoint (CScore/NCScore).

Source

The prior v234 release was distributed through the AWS Open Data bucket s3://epa-rsei-pds; that bucket is frozen at v234/2016 and does not carry v2.3.12, so the connector was re-pointed to EPA’s gaftp Public Release Data (#1148). The legacy per-table-gzip layout is still reachable via rsei_distribution="s3_gz".

One vintage network-wide (#436). Every registered site’s committed inventory — this one and each reference/rsei/<slug>/inventory.yaml — is pulled from this same v2.3.12 archive. The prior v234 ceiling capped all of them at last_year: 2014, which is a modeling vintage, not a per-county truncation; mixing vintages across sites would silently corrupt the cross-site comparison the network is for. So the inventories are re-pulled as a set, never one site at a time. A site whose last_year is still below 2022 after this pull is reporting its county’s real record (the West Union site stops at 2018, when Killen Station closed) — not a gap.

How the inventory is built

RSEI is a relational dump. watermark rsei joins five tables and keeps only the rows that roll up to a facility in the site’s own county (its profile’s rsei_fips):

elements   (ReleaseNumber)   -> Score, CScore, NCScore, Hazard, Population
  via release    (ReleaseNumber -> SubmissionNumber, Media, PoundsReleased)
  via submission (SubmissionNumber -> FacilityNumber, ChemicalNumber, SubmissionYear)
  via facility   (FacilityNumber -> name, coords, parent, NAICS1)
  via chemical   (ChemicalNumber -> name, CAS, ToxicityCategory)

• Pounds are summed from the reported release rows (PoundsReleased), bucketed by media via media.csv (MediaCode: 1 air, 3 water, 4 underground, 5 land, 6 POTW, 7 offsite).
• Score / Cancer / Non-cancer / Hazard are summed from the modeled elements rows. elements carries Hazard directly — BOSC does not recompute it.
• Codes use the primary reported NAICS1 field. v2.3.12 is NAICS-only — it dropped the SIC1 and NPDESPermit facility columns v234 carried, so sic / npdes_permit are now always null. Text is Latin-1.

Files

• inventory.yaml — provenance meta block + the county’s facilities, ranked by Score, each with cumulative Score/Cancer/Non-cancer/Hazard/pounds, a per-year series, a by-media pounds breakdown, and the top contributing chemicals. The reference build keeps the un-slugged path; every other site’s copy is reference/rsei/<slug>/inventory.yaml.
• toxic-discharge-screen.yaml — the toxic-load × assimilative-capacity screen (watermark toxics): the subset of those facilities that release toxics to water, placed on their receiving stream and read against the cited 7Q10 (see below).

Toxic-discharge screen (toxic-discharge-screen.yaml)

watermark toxics extends the hydrology low-flow assimilative screen from the three municipal WWTPs to the industrial dischargers — the RSEI facilities with water-media releases — using only committed artifacts (RSEI × ECHO × the cited 7Q10), no network.

• Receiving water is resolved on a ladder, never invented: ① a coordinate match to an EPA ECHO facility carrying a cited receiving water (source: connector); ② else membership in the site’s own industrial receiving-water corridor — the active profile’s toxic_corridor_bbox / receiving_water_name, a coordinate-cluster inference (source: assumption, flagged * in the CLI); ③ else left null and reported uncharacterized.
• Screening concentration is a coarse derived order-of-magnitude value — annual reported water pounds, fully mixed at the receiving stream’s 7Q10, no decay/mixing zone. It is a screen, not a permit determination or a measured concentration.
• Flag bands key on that concentration (the water pathway), not the total RSEI Score (which can be air-driven): critical ≥ 1 mg/L, elevated ≥ 0.01 mg/L.

What the screen finds is per-site, and is reported in that site’s instance note — a receiving stream with no assimilative capacity is a finding about one river, not about RSEI.

Caveats / gaps

• A facility with reported pounds but a zero Score released only non-modeled media/chemicals in the modeled years — that is faithful to the data, not a gap, and every county inventory carries some.
• RSEI covers TRI reporters only. Small/unpermitted sources and non-TRI chemicals are out of scope by construction.
• The Score reflects the modeling vintage and population layer of v2.3.12; absolute values are comparable within this version, not across RSEI versions.
• The bulk archive (RSEIv2312_Public_Release_Data.zip ~447 MB; elements ~1.2 GB unzipped) is not committed — it caches under the git-ignored data/cache/rsei/ and tables are streamed straight out of the zip. Only this curated YAML is committed.
• v2.3.12 no longer carries a facility NPDESPermit, so receiving-water joins to an ECHO NPDES inventory go through coordinate matches rather than a permit key.

Per-site instances

The findings — which facilities dominate a county’s Score, what the water bucket screens at, which corridor entities the parents resolve to — belong to the site whose county was reduced, and live beside this file under instances/:

• Lima’s instance

Cataloged datasets — generated from data/catalog/reference/; run watermark catalog render --apply after editing an entry.

A {…} placeholder in a title resolves to the site reading it — a slug-scoped dataset holds each site’s own copy, so its title names each site’s own county.

rsei — RSEI Toxic-Discharge Water Screen

Source: EPA RSEI (per-chemical water-media releases) × EPA ECHO (receiving water) × Ohio EPA cited design low flows (1Q10/7Q10/harmonic mean) × Ohio WQS criteria (OAC 3745-1) — a multi-source derivation · License: U.S. Government work (public domain) · Access: public · Site scope: basin-shared · Refresh: on-demand

Regenerate: watermark rsei

rsei-inventory — {county_state} RSEI Toxic-Release Inventory (EPA RSEI Public Data Set v2.3.12)

Source: EPA RSEI Public Data Set v2.3.12 (EPA gaftp Public Release Data), version v2312 · License: U.S. Government work (public domain) · Access: public · Site scope: slug-scoped · Refresh: on-demand

Regenerate: watermark rsei