← The record · All reference data

Network NPDES inventories (EPA ECHO)

Verified inventories of CWA-permitted facilities per watershed basin, pulled from EPA’s ECHO Clean Water Act REST services (cwa_rest_services). Every facility, NPDES ID, and value here was returned by the ECHO API — nothing is fabricated or inferred. The one exception is a curated receiving water, which is never invented either: it is a document-cited correction declared in the curated overlay and always marked as such on the row. Five basins are committed today, each with its own <basin>-wwtp.* fileset: the Maumee (watermark npdes, the default), the Great Miami (watermark npdes --basin great-miami, the Miami-basin sites — Urbana, Springfield, WPAFB, Troy-Piqua, Hamilton-Middletown), the Little Miami (watermark npdes --basin little-miami, the Scenic-River sites Xenia and Wilmington / Todd Fork, a single HUC-8 05090202), Ohio Brush Creek (watermark npdes --basin ohio-brush-creek, the direct-to-Ohio-River branch at West Union / Adams County — a single HUC-8 05090201; #1120), the Portage (watermark npdes --basin portage, the direct-to-Lake-Erie branch at Bowling Green / Middleton Township — a single HUC-8 04100010; #1433), the Scioto (--basin scioto, the Columbus / New Albany data-center cluster — three HUC-8s), the Muskingum (--basin muskingum, Mansfield and Coshocton — all six HUC-8s of subregion 0504, the largest river basin lying wholly in Ohio), and the Sandusky (--basin sandusky, the Sandusky / Fremont / Tiffin / Bucyrus points — a single HUC-8 04100011).

Every basin a registered site sits on now has a committed inventory, and two guards in tests/test_shared_registries.py keep it that way. Until they were added, three basins screened an empty set: muskingum and sandusky were in no registry at all, and scioto was registered here and in basin._BASIN_POTW_INVENTORY — with curated mainstem gages already waiting for it — while its pull sat deferred on an ECHO throttle (the note that used to stand in this paragraph). Seven registered sites therefore reported total=0, which reads exactly like a basin with no dischargers. basin._load_dischargers returns [] for a missing file on purpose — an absent inventory must never fall back to another basin’s rows — so that silence is a safety property and stays; what the guards remove is the chance of a registered site sitting behind it unnoticed. Add a basin by registering it in watermark.hydrology.connectors.echo, then pull it; never hardcode one into the connector.

What the watershed is

Seven USGS HUC-8 subbasins (subregion 0410, Western Lake Erie), queried one at a time via p_huc:

Adjacent WLE subbasins 04100001 (Ottawa-Stony), 04100002 (Raisin), and 04100010 (Cedar-Portage) are not Maumee drainage and are excluded.

Files

Structured YAML, each with a meta: provenance block. null is a genuine ECHO null (never an estimate); true/false flags are booleans.

• maumee-wwtp.all-npdes.yaml — meta: + facilities: list of all active CWA-permitted facilities ECHO returns for the seven HUC-8s, deduplicated to one record per facility by FRS Registry ID (POTW + non-POTW + federal + private/package).
• maumee-wwtp.potw.yaml — same shape, restricted to the subset flagged POTW by ECHO’s CWPFacilityTypeIndicator (municipal plants).
• maumee-wwtp.huc-counts.yaml — huc_counts: per-HUC manifest (ECHO’s reported count vs. rows actually pulled — they match, no pagination loss) plus totals: (raw / deduped / potw).

The three files above are regenerated output — never hand-edit them, the next pull overwrites the edit. Reviewed corrections go in the one hand-authored file here:

• curation/maumee-wwtp.receiving-water.yaml — the curated receiving-water overlay re-applied on every pull (see below).

Method

Per HUC-8: get_facilities (p_huc=<HUC8>, p_act=Y) returns a QID + summary count; get_qid pages the rows as JSON, columns selected by name (mapped to ECHO ColumnIDs). Deduplication keys on FRS Registry ID; a facility holding multiple permits keeps its primary NPDES ID with the rest in npdes_ids_secondary. Two distinct FRS IDs sharing a name are never collapsed.

Verified against cwa_rest_services metadata CWA v2017-10-13 1325 (260 result columns). Numbers come from the API’s structured fields, not any text layer.

Curated receiving water (the refresh path)

ECHO’s CWPStateWaterBodyName is null for most of the Ohio rows (66 of 832 carry one), including plants whose receiving water an Ohio EPA permit names outright. Those corrections used to be typed straight into the files above — so the next watermark npdes --basin maumee silently reverted them and regressed every downstream screen (Lima WWTP and Van Wert WWTP fell back to no_receiving_water, taking the basin’s two starkest effluent-dominance findings with them). That is why the basin inventory sat un-refreshed: a re-pull cost reviewed data.

Corrections now live in curation/maumee-wwtp.receiving-water.yaml and the connector re-applies them on every pull, so a refresh is non-destructive. Each entry carries the FRS Registry ID it pins, the document that names the receiving water, and the ECHO value observed when it was reviewed. Two modes:

• mode: field — the curated value is written into receiving_water (so derived flags like ottawa_discharge and every downstream screen see it), with ECHO’s verbatim value preserved alongside as receiving_water_echo and the row marked receiving_water_source: curated.
• mode: caveat — receiving_water keeps mirroring ECHO verbatim and the correction rides beside it on receiving_water_documented. This is the reviewed posture for OH0135569 (#379), whose correction comes from ECHO’s own facility report rather than an independent regulatory document, and which nothing downstream screens on.

The overlay never overrides live ECHO silently. If ECHO has moved off the reviewed echo_value, the pull names what changed — and whether it can still write depends on what moved:

• conflict (ECHO now supplies a different water) and stale (the facility is gone from a pull that covered its subbasin — a terminated or re-keyed permit) refuse the write. Nothing is emitted until a human reconciles the document against ECHO.
• superseded (ECHO now supplies the same water itself) does not block: the refresh completes normally, ECHO’s own value stands in the field, the row carries no curation provenance, and the run reports that the overlay entry is now redundant and can be retired.

Each pull prints an outcome table and records the applied set in every emitted file’s meta.receiving_water_curation.

Adding a basin’s overlay is a new curation/<basin>-wwtp.receiving-water.yaml plus a catalog entry — no code change. A basin with no overlay pulls exactly as before.

Headline counts (last pull)

2026-07-28 (whole-basin refresh, #1698): 1,662 active-permit rows across the 7 HUC-8s → 1,579 facilities after FRS dedup: 130 POTW, 1,447 non-POTW, 2 federal. POTW design flow present for 113/130 (the 17 blanks are mostly Michigan general-permit stabilization lagoons that don’t report a design-flow number).

Up from 1,006 facilities / 129 POTW at the previous pull. Nearly all of the growth (591 of 593 new rows) is General Permit Covered Facility coverage ECHO’s WATERS geocoding now resolves into these HUC-8s — construction/industrial stormwater, concentrated in the Lower Maumee — not new wastewater outfalls; 20 expired Indiana construction permits dropped out. Movements worth knowing about on the POTW side:

• Toledo Bay View Park WWTP (OH0027740, → Maumee River) is new to the inventory. ECHO carries no design flow for it, so it is counted but not screened.
• Shawnee II WWTP (OH0022675) design flow moved 2.0 → 3.0 MGD, converging on the figure the routed model already used from Ohio EPA fact sheet 2PK00002.
• Harrison Lake State Park was re-keyed from an individual permit (OH0036170) to a general permit (OHGC10322) and lost its receiving water; Miller City HS gained a design flow (0.008 MGD) and lost its receiving water. Both move to unscreened.

Great Miami River basin (great-miami-wwtp.*, #446/#455)

The Great Miami (subregion 0508, an Ohio River tributary) is two Ohio HUC-8 subbasins — same p_huc method and field shape as the Maumee, but the Lima-specific in_lima_subbasin / ottawa_discharge flags are omitted (a Maumee-only concept):

Whitewater (05080003) is predominantly Indiana drainage and is excluded (mirroring the Maumee’s excluded WLE neighbors). Last pull: 289 active-permit rows across the 2 HUC-8s → 286 facilities after FRS dedup, 81 POTW. The City of Springfield WWTP (OH0027481, 25 MGD, → Mad River) is present, but ECHO carries no receiving-water value for it, so the basin-screen reports it unscreened rather than guess — the same gap the Maumee’s curated overlay closes for Lima WWTP, and a candidate for a Great Miami overlay of its own once its receiving water is document-cited. The Hamilton WRF (OH0025445, 32 MGD, the basin’s largest POTW) is a second such candidate for a different reason: ECHO gives it GREAT MIAMI RIVER, TWO MILE CREEK, two different waters, and since #1120 a compound receiving water is refused rather than resolved to the larger of the two — so it reports no_7q10. The Little Miami’s Lower East Fork Regional WWTP (OH0049379) and Yellow Springs WWTP (OH0028215) are in the same position. A document-cited overlay entry naming the outfall’s actual water is what moves any of them back into the screened set.

Files: great-miami-wwtp.all-npdes.yaml, great-miami-wwtp.potw.yaml, great-miami-wwtp.huc-counts.yaml. Those counts are from an earlier pull; the Maumee refresh (#1698) did not re-pull this basin.

Ohio Brush Creek basin (ohio-brush-creek-wwtp.*, #1120)

The network’s first direct-to-Ohio-River branch — West Union / Adams County (#1117), which drains straight to the Ohio with no Scioto or Miami loop. One HUC-8, same p_huc method and field shape as the others, no Lima-specific flags:

Read the scope before reading the file. Unlike every other basin here, this one’s slug is a part naming a whole. maumee covers seven subbasins because they are all one river’s tree; 05090201 is a WBD cataloging unit spanning both banks of the Ohio River, and the pull says so without any outside dataset: its rows land in Kentucky counties (Campbell, Pendleton, Bracken, Mason, Lewis) and Ohio ones (Clermont, Brown, Adams, Highland, Scioto), which face each other across the river. Several separate creeks reach the Ohio inside the unit — West Union’s own receiver, Beasley Fork, drains to Ohio Brush Creek, while Whiteoak, Twelve Mile, Four Mile and the rest reach the river independently. Each is therefore a sibling of Ohio Brush Creek, not one of its headwaters: a discharger on one is neither upstream nor downstream of a discharger on another.

No HUC-12 count or area is stated here. The unit’s WBD composition (how many HUC-12s it holds, how many of them are Ohio Brush Creek, its area) is not in the corpus — no WBD extract for 05090201 is committed, and watermark wbd has only ever been run for Lima’s campus HUCs. Those figures were asserted in an earlier draft of this file and of the connector’s caveats without a source behind them; they have been removed rather than given a citation they never had. Pulling the WBD HU12 sublayer for 05090201 would make them citable and is open work.

Last pull (2026-08-05): 273 active-permit rows in the one HUC-8 → 261 facilities after FRS dedup, 23 POTW (17.87 MGD of design flow, present for 22 of 23). The two-bank geography makes it majority-Kentucky: 168 of the 261 carry KY permits (Campbell 111, Lewis 22, Mason 22, Bracken 11, Pendleton 2) against 93 Ohio ones (Clermont 20, Brown 15, Adams 11, Scioto 3, Highland 1, and 43 for which ECHO returned no county). Completeness therefore needs a cross-check against both Ohio EPA and Kentucky DOW; only what ECHO federalizes is reflected here.

What screens, and what does not

receiving_water is null for 118 of the 261 rows in that pull, and for 13 of the 23 POTWs. The assimilative screen (watermark --site west-union basin-screen) reports:

The five no 7Q10 rows are three different failures, and conflating them would misread all three. Three are ungaged tributaries: Felicity WWTP → Bear Creek, Georgetown WWTP → Town Run, Lewis County SD #1 → Grog Branch. One is not a water body at all: Western Mason County Sanitation District reads DOWNING DRIVE, MAYSVILLE STP — a street address and the name of the plant it sends flow to, i.e. a sewer connection, which is why it also has no design flow of its own. One names two different waters: New Richmond WWTP (OH0021156, 1.1 MGD) reads OHIO RIVER, TWELVE MILE CREEK, and ECHO’s field is a permit-level aggregate over every outfall, so nothing in it says which water carries the design flow. The screen refuses that row rather than crediting it with the larger of the two — at the Ohio’s 9,464 cfs it would have published a 5,560:1 ok for a plant that may discharge to the creek.

Two consequences worth stating plainly rather than leaving to be inferred:

1. No facility in this inventory names Ohio Brush Creek as its receiving water. The creek’s committed 7Q10 (0.50 cfs, USGS 03237500) is a denominator standing ready, not one in use.
2. The West Union WWTP (OH0028088, 0.7 MGD) does not screen, and neither do Peebles, Seaman or Winchester — the four Adams County POTWs off the Ohio River mainstem are exactly the four ECHO leaves blank. The SiteProfile attributes West Union’s discharge to Beasley Fork on the Ohio EPA NPDES service, but that is a live-service reading, not a committed document, so it cannot support a curated-overlay entry (each entry carries the instrument that names the water). The 1993 consent order in the corpus is explicit that it never names a receiving water. Closing this needs the Ohio EPA permit for 0PC00019 ingested — and even then Beasley Fork is ungaged, so the plant would move from no_receiving_water to no_7q10, not into the screened set. Both halves are open work; neither is closed by inference.

The screened five exist because #1120 added the Ohio River mainstem to data/reference/hydrology/mainstem-gages.yaml — USGS 03216600 at Greenup Dam, chosen as the upstream-end gage: it sits east of every facility in the pull, on a river that only gains drainage area downstream, so its 62,000 mi² understates what each facility’s own reach carries and the screen cannot overstate dilution. That entry’s note — copied into low-flow-7q10.derived.yaml, the file the screen actually reads — records what a reader is owed about it: the mainstem is navigation-regulated in fact and unannotated on the record, and one gage standing for the unit’s whole river frontage makes a dilution ratio a magnitude check rather than a reach-specific finding. The entry is scoped to this basin (basins: [ohio-brush-creek]), because that conservatism argument is geographic: the Muskingum and Mahoning basins meet the Ohio hundreds of river miles above Greenup, where the river carries a fraction of the drainage area gaged here, so serving this denominator on the bare name OHIO RIVER network-wide would commit exactly the overstatement it avoids.

Files: ohio-brush-creek-wwtp.all-npdes.yaml, ohio-brush-creek-wwtp.potw.yaml, ohio-brush-creek-wwtp.huc-counts.yaml. There is no curated receiving-water overlay for this basin yet.

Portage River basin (portage-wwtp.*, #1433)

The network’s first direct-to-Lake-Erie branch that does not run through the Maumee — Bowling Green / Middleton Township (#1433), whose Water Pollution Control plant discharges to Poe Ditch and thence to the North Branch Portage River. One HUC-8, the same p_huc method and field shape as the others:

Read the scope before reading the file. Like Ohio Brush Creek above, this slug is a part naming a whole: 04100010 “Cedar-Portage” is a WBD cataloging unit covering the Portage River and the frontal Lake Erie drainage beside it, and the pull says so without any outside dataset. Its rows include the Lake Erie islands — Put-in-Bay, Kelleys Island State Park, ODNR’s South Bass Island campground — plus Catawba Island and Oregon on Maumee Bay, none of which are Portage River drainage at all. Each is a sibling of the Portage, not one of its headwaters: neither upstream nor downstream of a Portage discharger, and never entitled to borrow its low flow.

The split is live at the network’s own watershed point. Bowling Green’s data-center campus sits in HUC-12 041000100703 (Cedar Creek-Frontal Lake Erie) while the city’s WPC plant discharges in 041000100301 (N Br Portage/Poe Ditch), so campus construction runoff does not reach the plant’s receiving water — a fact established from the permits in #1439 and reproduced by this unit’s own composition.

Last pull (2026-08-14): 310 active-permit rows in the one HUC-8 → 296 facilities after FRS dedup, 26 POTW (34.858 MGD of design flow, present for 23 of 26). Ohio-only, spread across seven counties — for the POTWs, Ottawa 10, Wood 9, Seneca 2, Sandusky 2, Erie 1, Hancock 1, Lucas 1. Bowling Green (OH0024139, 10.0 MGD) is the largest POTW in the basin, ahead of Fostoria’s 8.25 and Oregon’s 8.0.

ECHO’s county spelling is not normalized in this pull. Both LUCAS and LUCAS COUNTY come back, likewise Ottawa and Wood, and 154 of the 296 rows carry no county at all. A county tally taken off this file under-counts rather than resolving to zero, and must fold the two spellings together before it means anything.

What screens, and what does not

receiving_water is null for 261 of the 296 rows in that pull, and for 21 of the 26 POTWs. The assimilative screen (watermark --site bowling-green basin-screen) reports:

The one screened row is Bowling Green itself, and it does not screen against a basin proxy: poe ditch (bowling green wpc outfall, rm 2.5) in low-flow-7q10.yaml is permits:-bound to OH0024139 / 2PD00009, so screen_facility takes the permit match ahead of any name match. The binding is safe here in the way the Defiance entry’s note requires — the fact sheet’s 10 MGD average design flow and the 10.0 MGD ECHO carries for this permit agree exactly, so a permit-cited numerator is paired with a denominator that permit corroborates. The bare name poe ditch deliberately resolves to nothing, so a future second discharger on the ditch cannot inherit a low flow computed at this outfall alone.

The four remaining named receivers are Algire Creek (McComb), Wolf Creek (Evergreen Poplar), Turtle Creek (Fenwick Marina) and PORTAGE RIVER, URIE DITCH (Bloomdale). Three are ungaged tributaries with no NWIS daily-discharge record meeting the 20-climatic-year floor. The fourth names two different waters, and ECHO’s field is a permit-level aggregate over every outfall, so nothing in it says which one carries the design flow — the screen refuses that row rather than crediting it with the larger.

No Portage mainstem gage is registered in mainstem-gages.yaml, and that is deliberate. Registering one would screen exactly zero rows today — the only mainstem-naming row is Bloomdale’s compound, which is refused on its own terms — while creating a live hazard: a mainstem gage integrates a drainage area far larger than the North Branch or Poe Ditch above it, so any later alias reaching Bowling Green’s outfall would overstate its dilution by orders of magnitude. That is the same defect class as the 17x overstatements found at Sidney (#1992) and in #1995’s routing table, and the cheapest place to not commit it is here.

Bowling Green’s own outfall is screened anyway, by the better number. Ohio EPA’s fact sheet for 2PD00009 publishes a drainage-area-adjusted 7Q10 of 0.364 cfs at the outfall (Table 12, from USGS 04195500 over 1951-97) — the regulator’s own denominator for this reach, carried in reference/hydrology/bowling-green/routing.yaml with source: document. A cited permit-scoped low flow beats a basin-wide screening proxy wherever one exists, which is why this basin’s zero screened rows are a statement about the other twenty-five plants and not about the site.

One more property of the basin worth stating plainly: it drains a heavily tile-drained lakebed plain (the Great Black Swamp), so summer low flows are small and a large share of the receiving water below a POTW outfall is that plant’s own effluent. Bowling Green’s 15.47 cfs design discharge against a 0.364 cfs 7Q10 is roughly 42x the river it enters. Read an assimilative ratio here as a statement about an effluent-dominated stream, not a diluted one.

Files: portage-wwtp.all-npdes.yaml, portage-wwtp.potw.yaml, portage-wwtp.huc-counts.yaml. There is no curated receiving-water overlay for this basin yet — closing the 21 null receivers means reading 21 permits’ fact sheets, and each correction needs its own citation.

Scioto River basin (scioto-wwtp.*)

The Columbus / New Albany data-center cluster, plus Piketon (PORTS) and Portsmouth on the lower river. Three HUC-8s, the same p_huc method and field shape as the others:

2026-08-21 pull: 377 rows → 362 facilities after FRS dedup (123 POTW, 239 non-POTW). 86 of the 123 POTWs carry no receiving_water and 12 no design flow.

This basin is where the coverage gap took its worst shape. It was already registered in both echo.BASINS and basin._BASIN_POTW_INVENTORY, and the curated Scioto gages — the mainstem at Higby, the Olentangy near Worthington, Big Walnut Creek, Big Darby Creek — had been sitting in mainstem-gages.yaml waiting for it. Only the pull was missing, so the code advertised support while four registered sites screened nothing.

What screens, and what does not

Unlike the Muskingum and Sandusky sections below, some rows here do screen, because this basin’s denominators are partly registered. Read them with the derivation’s own label: each is the 7Q10 at the gage, a screening proxy for a discharge reach that differs by drainage-area ratio. A gage downstream of an outfall integrates a larger drainage area and therefore overstates dilution — which is conservative for a violation finding and the wrong direction for an ok one. Treat a comfortable ratio on a downstream-integrator denominator as unproven rather than clean.

Muskingum River basin (muskingum-wwtp.*)

Mansfield (Rocky Fork of the Mohican) and Coshocton (the Tuscarawas–Walhonding confluence), with Newark and Zanesville tracked above and below them. All six HUC-8s of subregion 0504 — the largest river basin lying wholly within Ohio:

The set was enumerated from the USGS Watershed Boundary Dataset itself (hydro.nationalmap.gov/arcgis/rest/services/wbd/MapServer layer 4), not typed from memory, and every network point was located against the same service: mansfield → 05040002, coshocton → 05040004, newark → 05040006, zanesville → 05040004.

2026-08-21 pull: 659 rows → 630 facilities after FRS dedup (149 POTW, 481 non-POTW) — the network’s largest POTW inventory, ahead of the Maumee’s 130. 97 of the 149 carry no receiving_water and 17 no design flow.

Unlike 04100010 and 05090201, this subregion IS one river’s drainage tree. Both of those carry a “scope is wider than the name” caveat because a WBD cataloging unit is not a watershed; that warning does not transfer here. The Tuscarawas and the Walhonding join at Coshocton to form the Muskingum mainstem, and Wills Creek and the Licking enter below, so every discharger in 0504 is genuinely upstream of the basin mouth and no unit is a frontal-drainage sibling of another.

What screens, and what does not

8 of 149, on two gages registered in mainstem-gages.yaml: the Tuscarawas at Newcomerstown (03129000, DA 2,443 mi²) and the Muskingum at McConnelsville (03150000, DA 7,422 mi²). Both are mouth-ward on their own mainstem, both carry a cited regulation block quoting Straub 2001’s REMARKS verbatim — 17 flood-control reservoirs above McConnelsville, eight plus an out-of-basin diversion above Newcomerstown — and both are therefore confidence: low denominators integrating thousands of square miles. Read a ratio here as a magnitude check, not a reach-specific finding. All eight screened rows currently read ok.

Two things registered here later must still be true, and are why the Ohio River gage at Greenup Dam is scoped basins: [ohio-brush-creek]: the Muskingum meets the Ohio hundreds of river miles above it, and an unannotated gage means “not yet read against the published record”, never “unregulated”.

⚠️ Neither gage screens Mansfield, the network’s own Muskingum point. Its WWTP discharges to Rocky Fork of the Mohican, and HUC-8 05040002 has no usable gage at all: every long record ended between 1975 and 1993 (Mohican at Greer to 1982, Black Fork at Loudonville to 1991, Clear Fork below Pleasant Hill to 1991, Touby Run to 1978) and every live gage is far under the 20-climatic-year floor (Black Fork at Shelby ~14 yr, Rocky Fork at Lucas ~3 yr, Touby Run at W 6th ~2 yr). rocky fork creek and mohican river are deliberately unregistered, so those rows stay no_7q10. Mansfield’s denominator has to come from its permit’s fact sheet, which this corpus does not hold (#1429).

⚠️ “Little Muskingum River” is a different river — a direct Ohio River tributary in Washington County (DA 210 mi²) that the 0504 pull picks up, and one POTW in the inventory names it. It is not an alias of the Muskingum, so that row stays unmatched rather than borrowing a 35x-larger denominator.

Sandusky River basin (sandusky-wwtp.*)

The Sandusky / Perkins Township point, with Fremont, Tiffin and Bucyrus tracked upstream. One HUC-8 — wholly in Ohio, and the largest single subbasin in subregion 0410, bigger than the Auglaize:

All four network points sit inside this one unit, each located against the USGS WBD service.

2026-08-21 pull: 125 rows → 124 facilities after FRS dedup (31 POTW, 93 non-POTW). 22 of the 31 carry no receiving_water.

It is a sibling of the Maumee and the Portage, not part of either — all three reach Western Lake Erie independently — which is why 04100011 is absent from MAUMEE_HUC8S and PORTAGE_HUC8S, and why no discharger here may borrow either basin’s denominator.

What screens, and what does not

1 of 31 — and it is a violation. The inventory contains exactly one discharger naming “Sandusky River”: BUCYRUS WWTP, 3.4 MGD, at the basin’s headwaters.

That single row is why this basin’s entry in mainstem-gages.yaml deliberately does not use the mouth-ward gage. Registered is 04196000 Sandusky River near Bucyrus (DA 88.8 mi²), the gage 1.5 mi west of the plant — not 04198000 near Fremont (DA 1,251 mi²), which the “lowest gage on the mainstem” convention would have selected. Fremont integrates 14x the facility’s drainage area, and watermark.hydrology.basin’s own rule forbids precisely that: a discharger “is never screened against a downstream river’s larger 7Q10 (that would overstate dilution into a false ‘ok’).”

The difference is the whole verdict. On the Fremont gage this plant screens at ~4:1 and reads tight; on its own reach it is a violation — a derived 7Q10 of 2.38 cfs (Straub’s published figure is 1.1) against a 5.26 cfs design flow, so the discharge runs roughly two to five times the river it enters at low flow. The mouth-ward convention would have published “tight” over an effluent-dominated stream. Both gages are carried in the entry — Fremont as a cross_check_gage, labelled as the worked example.

Note what the regulation is here: the reservoirs Straub names 5.3–6.0 mi above the gage are Bucyrus’s own municipal supply. The city withdraws its drinking water above the gage and discharges its effluent below it, and both sides of that loop sit inside the 88.8 mi² this denominator describes.

⚠️ A bay is not a river reach, and a 7Q10 does not describe one. The network’s own point discharges to Sandusky Bay — the inventory carries the surface form lake erie/sandusky bay — an embayment whose mixing volume is set by lake level, seiche and wind rather than streamflow. It is deliberately not an alias, so that row stays no_7q10. A river 7Q10 served to a bay outfall is not a conservative approximation, it is a category error.

⚠️ If a downstream Sandusky River discharger appears in a later pull, do not leave it on this gage. A headwaters 7Q10 understates dilution for anyone below it — the safe direction, but still wrong. The three downstream gages are carried as cross-checks for exactly that re-screening.

Known gaps & caveats (read before using)

1. No CWNS ID. The ECHO CWA facility service has no CWNS column, so the requested CWNS cross-check for POTWs is not available from this API. POTW classification here rests solely on CWPFacilityTypeIndicator. (CWNS IDs would have to come from the Clean Watersheds Needs Survey, separately.)

2. HUC geocoding (WATERS). ECHO links facilities to HUCs via WATERS; not every NPDES ID geocodes, so a pure watershed query can miss facilities whose coordinates didn’t resolve. RadWBDHu8 is frequently null in the raw data — we use FacDerivedHuc (which reliably reflects the queried HUC-8) instead.

3. Cross-state completeness. Four subbasins (St. Joseph, St. Marys, Upper Maumee, Tiffin) extend into Indiana and/or Michigan. ECHO’s p_huc did return IN/MI facilities (e.g. Auburn IN, Amboy Twp MI), but for a complete inventory this pull should still be cross-checked against Ohio EPA, Indiana IDEM, and Michigan EGLE NPDES permit lists; any facility in a state list but absent from ECHO should be flagged. (Not yet performed — those state datasets are a follow-up; this is an ECHO-only inventory.)

4. “All” is the active CWA universe, not just process wastewater. p_act=Y includes some non-NPDES Industrial-User/pretreatment permits and stormwater general permits alongside true wastewater dischargers. The permit_type (CWPPermitTypeDesc) and facility_type fields make the distinction visible — filter on them rather than assuming every record is a wastewater outfall.

5. ottawa_discharge undercounts. This optional flag is keyed on ECHO’s CWPStateWaterBodyName string, which is null for most of the Ohio rows. Lima WWTP (OH0026069, 18.5 MGD), the largest Lima-area POTW, is flagged only because its permit-cited receiving water is supplied by the curated overlay; every other Ottawa-River discharger ECHO leaves blank is still missing from the flag, and per the “no inference” rule none of them is guessed at. Use in_lima_subbasin (every Auglaize + Blanchard record) plus county for the broad Lima/Allen screen, and treat ottawa_discharge: true as a floor, not a complete list.

Field reference

Each entry under facilities: carries these keys (null = ECHO returned nothing):

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

echo-great-miami-wwtp — Great Miami-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:great-miami · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin great-miami

echo-little-miami-wwtp — Little Miami-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:little-miami · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin little-miami

echo-maumee-npdes — Maumee-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:maumee · Refresh: quarterly (ttl 180d), last 2026-07-28

Regenerate: watermark npdes --basin maumee

echo-maumee-receiving-water — Curated receiving-water overlay for the Maumee ECHO NPDES inventory

Source: Hand-authored, document-cited corrections to ECHO’s CWPStateWaterBodyName — Ohio EPA NPDES permit 2PE00000*OD (Lima WWTP), Ohio EPA NPDES fact sheet 2PD00006 Table 12 (Van Wert WWTP), and the EPA ECHO detailed facility report for OH0135569 · License: U.S. Government work (public domain) — the cited permits and fact sheets are Ohio EPA records · Access: public · Site scope: basin:maumee · Refresh: on-demand

echo-muskingum-wwtp — Muskingum-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:muskingum · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin muskingum

echo-ohio-brush-creek-wwtp — Ohio Brush Creek-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:ohio-brush-creek · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin ohio-brush-creek

echo-portage-wwtp — Portage-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:portage · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin portage

echo-sandusky-wwtp — Sandusky-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:sandusky · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin sandusky

echo-scioto-wwtp — Scioto-basin NPDES discharger inventory (EPA ECHO)

Source: EPA ECHO — cwa_rest_services (CWA v2017-10-13) · License: U.S. Government work (public domain) · Access: throttled · Site scope: basin:scioto · Refresh: quarterly (ttl 180d)

Regenerate: watermark npdes --basin scioto