Hydrology — Tier-0 municipal water-flow findings

Generated by watermark (watermark.hydrology). Tier-0 SCS screening — auditable and fast, not a substitute for SWMM/HEC-RAS. Every figure is tagged [verified] (read from a record or a live gauge) or [inference] (assumption/derived).

1. The municipal loop and its low-flow squeeze

The Lima system is one closed loop on two rivers:

Auglaize/Ottawa → Lima WTP → municipal + data-center demand → WWTPs → Ottawa River

Low-flow assimilative screen — each discharge against its receiving stream’s cited design low flows, the design flow matched to the criterion type (chronic aquatic-life at the 7Q10, acute at the 1Q10):

• ❌ Shawnee II WWTP → Ottawa River: chronic 7Q10 0.2 cfs vs discharge 4.64 cfs → 0.04:1 dilution (violation). [verified] Ohio EPA NPDES fact sheet 2IG00001 (Lima Refining Co.), Stream Flows table — Ottawa River at Lima, USGS gage 04187100, 1989-2021 Against the acute 1Q10 0 cfs → 0.00:1 (violation). Crediting the 28.62 cfs of permitted effluent already in the reach [inference] → 6.21:1 (tight).
• ❌ American Bath WWTP → Pike Run: chronic 7Q10 0.03 cfs vs discharge 2.32 cfs → 0.01:1 dilution (violation). [verified] Ohio EPA NPDES fact sheet 2PH00007 (American Bath WWTP), Stream Flows table — USGS Gauge 04186500 adjusted for drainage area Against the acute 1Q10 0.016 cfs → 0.01:1 (violation).
• ❌ American II WWTP → Dug Run: chronic 7Q10 0.78 cfs vs discharge 1.86 cfs → 0.42:1 dilution (violation). [verified] Ohio EPA NPDES fact sheet 2PH00006 (American II WWTP), Stream Flows table — USGS Station 04187500 Against the acute 1Q10 0.6 cfs → 0.32:1 (violation).
• ❌ Lima WWTP → Ottawa River: chronic 7Q10 0.2 cfs vs discharge 28.62 cfs → 0.01:1 dilution (violation). [verified] Ohio EPA NPDES fact sheet 2IG00001 (Lima Refining Co.), Stream Flows table — Ottawa River at Lima, USGS gage 04187100, 1989-2021 Against the acute 1Q10 0 cfs → 0.00:1 (violation). Crediting the 4.64 cfs of permitted effluent already in the reach [inference] → 0.17:1 (violation).

At design low flow the receiving streams carry less than the effluent they receive — the discharges are effectively undiluted by natural flow. Where a plant shares the reach with other permitted dischargers, the effluent-credited ratio shows the more honest system picture: it is diluted by standing effluent, not clean water.

A second campus pathway — stormwater to Pike Run. Distinct from the FM-2 process discharge above (routed to Lima’s WWTP), the campus’s stormwater leaves the site via a constructed BOSC Storm Outfall channel that discharges to Pike Run — the loop’s most flow-starved tributary (7Q10 0.03 cfs, already shown undiluted by the American Bath WWTP). Per the roundabout/outfall SWP3 (Ohio EPA eDoc 4091286; operator George J. Igel & Company, Inc., engineer WSP USA Inc.; prepared 2026-04-16) [verified: document], the site drains east-to-west by subsurface tile and the outfall channel terminates at Pike Run. That SWP3 documents construction disturbance (5.71 ac), not a continuous low-flow discharge, so the pathway is recorded as a receiving-water fact, not added to the routed mass balance below.

The cited 7Q10 is independently reproducible. The denominator above is a single number read off a fact sheet. Computing it ourselves from the raw record — the USGS daily-mean discharge at the same gage the fact sheet names (NWIS 04187100, Ottawa River at Lima OH, 1988-09-30..2024-12-31, 24 complete climatic years) — lands on it. Annual n-day minima by climatic year, fit with log-Pearson III and bracketed by the non-parametric Weibull plotting position [inference: derived]:

The computed 7Q10 is 0.2387 cfs against the cited 0.2 cfs — agreement to within rounding, from an independent method on a longer record than the fact sheet used. The 1-day record is dry in 21% of complete years, so the computed 1Q10 is 0 cfs — the mainstem literally stops, matching the cited 1Q10. So the assimilative screen’s denominator is not an Ohio EPA artifact to be argued with; it is what the river actually carries at design low flow, reproducible by anyone with the public gage record. These computed figures are [inference: derived] and corroborate — they do not replace — the cited regulatory statistic.

The design flow is matched to the criterion type. Acute aquatic-life criteria are evaluated at the 1Q10, chronic at the 7Q10, and human-health (carcinogen) criteria at the harmonic-mean flow — the whole-record statistic that weights the low tail the way a lifetime exposure does. Computed from the same daily record (N/Σ(1/Q) over 9671 non-zero days, 63 zero-flow days excluded), the harmonic mean is 4.8232 cfs and lands on the cited 4.8 cfs [inference: derived]. So each criterion in the toxic screen is read against its own design flow, not a single 7Q10 pressed into every service.

The whole loop at design low flow: a routed mass balance

The screen above reads each plant against its own tributary in isolation. Routing the cited headwater 7Q10s, the document-cited WWTP/campus discharges, and the cooling draw through the cited confluence graph (data/reference/hydrology/network.yaml) shows the system picture the per-stream rows miss. At design low flow the loop’s streams carry, in total, only 1.01 cfs of natural low flow (Ottawa 0.2 [verified: document] + Dug Run 0.78 [verified: document] + Pike Run 0.03 [verified: document]). The three small county WWTP discharges alone add 8.82 cfs of treated effluent — 8.7x the streams’ entire natural low flow, with no data center in the picture. The river at design low flow is effluent, not stream. The City of Lima WWTP — the major municipal plant — then adds its own 28.62 cfs design discharge (18.5 MGD, OEPA NPDES 2PE00000 / OH0026069, outfall to the Ottawa at RM 37.6), and the campus adds its documented 3.87 cfs FM-2 industrial discharge (piped to that same plant via Lima’s 78” sewer), together taking the Ottawa leaving Lima to 98% treated effluent — now counting the municipal plant’s own design flow that the corpus previously lacked (#1536; the FM-2 is the campus’s increment on top of the plant’s permitted baseline, not a double count).

Unbuffered bound. If the cooling load of 4.85 cfs were pumped straight from the Ottawa at 7Q10 (4.8x the loop’s entire natural low flow) it consumes the Ottawa mainstem’s entire design low flow — it runs dry at the intake, leaving a 3.84 cfs shortfall the river cannot supply. But that is not how Lima’s supply works — the city draws treated water from ~15 billion gallons of off-stream reservoir storage (see the next section), so this is a worst-case bound, not the operating reality. The routed balance still conserves mass (base + gains - applied loss reconciles to the 43.29 cfs outlet) [inference: derived]. The order-invariant system totals are the robust result; the per-reach values depend on the cited-but-approximate confluence order and are screening-grade.

Downstream: the Auglaize confluence. The outlet is the Ottawa’s actual receiving water, not a bare label. The Auglaize’s own 1.17 cfs 7Q10 [inference: derived] — a drainage-area-ratio transfer, not the raw USGS 04186500 (Fort Jennings) gage value, which sits below the confluence and already carries the Ottawa’s own historical contribution (full method in low-flow-7q10.derived.yaml) — joins here, diluting the 42.12 cfs routed Ottawa flow (mostly treated effluent) arriving from Lima. ❌ At 0.03:1 dilution the confluence screens violation by the same band as the per-plant checks above — the Auglaize’s own flow is dwarfed by what the loop sends it. The final flow leaving toward the Maumee is 43.29 cfs, 95% treated effluent — down from 98% at the Lima gage, so the Auglaize genuinely dilutes the loop’s discharge without resolving the underlying effluent-dominance. Even this transfer is a coarse screening estimate — it nets out the Ottawa’s own drainage area at Lima, not at its actual mouth a few miles further downstream — so it likely still slightly over-states the Auglaize’s own contribution.

The supply side: off-stream storage, not a 7Q10 intake

The screen above reads the campus draw as if it depleted the Ottawa at design low flow. It does not — and the real mechanism is a stronger finding. Lima’s raw water is held in 5 upground (off-stream) reservoirs totalling ~14.4 billion gallons (Auglaize River 10.1 BG, Ottawa River 4.3 BG), filled by pumping from both the Auglaize (west) and the Ottawa (east) through four pump stations at high flow [verified: document]. So Lima never withdraws at the 7Q10 — it lives off stored water, and the binding low-flow constraint is reservoir drawdown, not intake depletion.

The data center draws treated municipal water like any large customer, so its 3.92 MGD makeup is an added draw on this shared storage — 20.7% of the 18.92 MGD the plant would then produce (against ~15 MGD today, 30 MGD rated) [inference: derived]. At that draw the zero-refill drought reserve falls from 960.9 to 761.8 days (-199.1). Its evaporative 3.136 MGD consumptive is a permanent loss to the basin — the returns (FM-2/FM-1) go downstream to the Ottawa via the WWTPs, never back to the reservoirs, so the full makeup draws storage down. This is a far harder number to rebut than the 7Q10 multiple: the campus alone is a fifth of the city’s water production, drawn from a finite reserve that must be refilled by high-flow pumping from two rivers whose yield is lowest in exactly the season the draw is highest. Quantifying that refill against the Auglaize (USGS 04185750) and Ottawa (04187100) flow records is the next increment.

Can the rivers refill the reservoirs — even in drought?

Off-stream storage only helps if high-flow pumping keeps it filled. Two questions, two answers from the gauged record (Auglaize at Fort Jennings + Ottawa at Lima, 1988-09-30—2024-12-31). In a normal year, refill is amply adequate: the two rivers’ combined mean flow at the intakes (324.6 cfs, Auglaize scaled to its intake reach) is ~11.1x the city+campus demand [verified: connector]. The binding case is drought: the Ottawa reaches 0 cfs and the Auglaize sits below the city+campus draw ~25.8% of the time, so the system draws down storage.

The sequent-peak storage requirement — the active storage the worst gauged drawdown calls on at a constant demand — measures the drought margin and the campus’s bite:

The worst gauged drought (the 1999 event, a ~338-day drawdown) is survived with large margin — but the campus raises the storage it calls on from 31.3% to 41.1% of capacity (+1,412 MG, and ~13 more days of drawdown). At the high cooling bound (27.5 MGD) it rises to 62.7%. [inference: derived] So refill is adequate and the system survives the historical record — but the campus measurably erodes the buffer, and a drought longer or deeper than 1988—2024 is the residual exposure this screen cannot bound.

The estimate remains optimistic (the 0.614 drainage-area transfer scaling the downstream Auglaize River gage to the intake is itself a coarse over-estimate; no pump-rate cap), so the real margin is still somewhat tighter than shown. A first-order reservoir-evaporation sink is folded in — FAO-56 ET0 over the 1603-acre reservoir surface, ~5.095 MGD mean (peaking at 8.8479 MGD in JUL) [derived], which tightens the drought bound above.

Industrial toxic dischargers on the same reaches. The municipal screen above covers the three WWTPs; the industrial side is larger. Of the 13 EPA-RSEI facilities that release toxics to water in the county, 3 sit on a near-undiluted reach. Placing each on its receiving stream (ECHO-cited where available, else inferred from the Ottawa River industrial corridor) and reading it against the same cited 7Q10:

* = receiving water inferred from the corridor coordinate cluster, not independently cited. The screen mg/L is a coarse [inference: derived] value (annual reported water pounds, fully mixed at the 7Q10) — an order-of-magnitude screen, not a measured concentration.

The seasonal pinch compounds it: the Ottawa’s 1Q10 is 0 cfs (and summer 30Q10 1.6 cfs [verified: document]) — the mainstem effectively dries at design low flow. That floor falls in the May-Oct window where reference ET exceeds precipitation (§3), so the largest toxic loads meet the smallest assimilative capacity exactly when the river is lowest.

Outfall flood exposure. None of the 3 plant sites sits in the FEMA Special Flood Hazard Area at its ECHO-reported point, but the discharge infrastructure is flood-adjacent on streams already shown to be undiluted at low flow [verified: document]:

ECHO coordinates are the facility location, a proxy for the NPDES outfall; the actual outfall discharges at the receiving stream and is likely closer to the mapped floodplain than the facility centroid. So the mapped exposure understates the outfalls’: the discharge points themselves sit at the receiving water, inside or at the edge of the AE floodplain.

2. The Maumee Nutrient TMDL: the same discharges are capped phosphorus loads

These discharges don’t just strain a local stream. The Ottawa flows to the Auglaize and on to the Maumee — Lake Erie’s largest tributary and the driver of its western-basin harmful algal blooms. The 2023 Maumee Watershed Nutrient TMDL (Ohio EPA, US-EPA-approved) assigns each individually permitted discharger a total-phosphorus wasteload allocation: a spring-season (March-July) cap, also stated as a daily equivalent. The plants the low-flow screen flags as effectively undiluted are the same permits carrying these caps [verified: document]:

Across the whole grouped category of individually permitted dischargers the cap totals 64.1 metric tons (418.8 kg/day) of spring phosphorus. So the local dilution failure compounds a basin-scale constraint: at design low flow these effluents are near-undiluted, and every pound of phosphorus is metered against a Lake Erie nutrient budget.

3. Stormwater: paving the corridor

Climate baseline (NASA POWER). The Lima point averages ~996 mm/yr of precipitation (corrected), peaking in May, at a mean annual temperature of 10.7 °C [reference: NASA POWER climatology]. The satellite climate normal sets the long-run water budget; the design storm below is the NOAA Atlas-14 extreme the corridor must detain — the two are complementary.

Reference ET (FAO-56 Penman-Monteith). Atmospheric water demand runs ~1,085 mm/yr of reference ET0, computed from the same POWER normals (temperature, humidity, wind, solar) [derived: FAO-56 Penman-Monteith]. Net of precipitation that is -89 mm/yr — and ET0 exceeds rainfall across the May-Oct growing season, so summer soil moisture, pond evaporation, and any consumptive cooling draw compete for water in the months the Ottawa is already near its low-flow floor (§4).

A 25-yr 24-hr design storm (4.25 in [inference: assumption]) over the 340-ac footprint [verified]:

• 25-yr 24-hr storm (4.25 in): peak 373 -> 431 cfs (+58, CN 85 -> 87, Tc 1 -> 0.78 hr)

• runoff volume 77 -> 81 ac-ft (+4 ac-ft to detain for pre-development control)

The footprint sits just outside the FEMA floodplain — but only just. The recorded campus parcels intersect no FEMA Special Flood Hazard Area, yet Zone AE and AE (FLOODWAY) (1%-annual-chance floodplain and regulatory floodway) reach within ~50 m of them (FEMA DFIRM 39003C_FIS5) [verified: document]. The post-development runoff increase routes toward that corridor; a regulatory floodway tolerates no rise, so added peak discharge there is a permitting constraint, not only a detention-sizing question.

Drainage scope vs the design storm

The roundabout program budgets $1,068,530 of drainage across 6 OPC sub-estimates [verified: document], but the engineering basis is thin. Auditing what the estimates actually quantify against the corridor design rainfall:

Atlas-14 corridor design storm (24-hr) [verified: connector]: 2-yr 2.52 in, 10-yr 3.58 in, 25-yr 4.25 in, 50-yr 4.81 in, 100-yr 5.39 in.

• $1,068,530 of drainage across 6 sub-estimates (7.5% of the $14,233,081 program), but only 1 of 6 carry an extracted line-item breakdown — the rest is a bare section subtotal.

• $100,000 of $120,440 (83%) is lump-sum ‘Drainage improvements’. The only sized conveyance is: 6” shallow pipe underdrains with geotextile fabric, as per plan.

• No estimate cites a design storm or return period. The corridor design rainfall (NOAA Atlas-14): 25-yr 24-hr 4.25 in, 100-yr 24-hr 5.39 in [verified: connector] — the basis the unsized storm-sewer / detention scope must meet.

• Neither the OPC drainage scope nor the 95% SPS grading & storm plan itemizes detention/retention storage (detention_shown=false), echoing the corpus’s own open question on the lump-sum DRAINAGE items.

This is a design-basis / scope-completeness reading, not a sizing of the roundabouts’ hydraulics — the corpus carries no per-roundabout footprint area, so runoff/detention volumes are deliberately not computed.

4. Scenario: data-center cooling vs the Ottawa River’s low flow

The cooling demand is sourced, derived from disclosed campus data by two methods:

• top-down: IT load 275.00 MW [inference: derived] x WUE 1.80 L/kWh [inference: assumption] → 3.14 MGD consumptive
• bottom-up: FM-2 blowdown x 5 cycles, capped at the physical evaporative-WUE ceiling (FM-2 is not purely cooling blowdown, so the raw blowdown figure is unreachable for cooling alone) → 3.84 MGD consumptive (upper bound)

They bracket the consumptive demand at 3.14-3.84 MGD. The conclusion is robust to the range.

Buildout adds 4.85 cfs of net consumptive draw — 24.3x the Ottawa River’s cited 7Q10 (0.2 cfs). At design low flow the Ottawa River nearly dries (1Q10 = 0 cfs); a data center’s cooling draw competes for water the river does not have — even the low estimate is tens of times the 7Q10.

The seasonal pinch: the draw lands when the river is lowest

The annual-7Q10 multiple understates the constraint. The growing season (MAY-OCT, where reference ET exceeds precipitation — §3) is exactly when the Ottawa sits at its summer design low flow, with no rainfall buffer. Reading the same consumptive draw against the cited seasonal floor:

In the MAY-OCT window the draw is 3x the cited summer 30Q10 (1.6 cfs) — vs 24.3x the annual 7Q10. And the summer 30Q10 is the generous floor: the Ottawa’s absolute design low flow is 1Q10 = 0 cfs [verified: document], so in the driest growing-season weeks there is no flow to draw against at all. The cooling draw peaks against supply precisely when the atmosphere is also taking the most.

5. Tier-1 escalation (EPA SWMM)

watermark tier1 runs the real EPA SWMM5 engine on the footprint under the design storm for two questions Tier-0 only approximates: the detention volume that holds the post-development peak to the pre-development rate, and the sanitary wet-weather surcharge (dry-weather base + RDII) against each plant’s documented wet-weather headroom. Hydraulic routing parameters (RDII, basin geometry) are assumptions; the footprint, its declared impervious acreage — which drives the as-permitted case’s imperviousness — the storm, and the plant design flows stay document/connector-sourced.

The committed run (pyswmm 2.1.0, 25-yr 4.25-in storm; mass-balance continuity error 0.00%) [inference: derived] sizes the detention the corridor needs. Paving the footprint as permitted takes the design-storm peak from 169 cfs (cropland) to 436 cfs — 33.9% impervious, the declared permanently-impervious acreage; holding the release back to the pre-development rate (168 cfs) takes a 20 ac-ft basin (13.6 ac, 9.28-ft bottom orifice). Every input deck is committed under data/reference/hydrology/swmm/ so anyone can re-run them in EPA SWMM.

Built out — the whole parcel at the blanket 90% impervious assumption, the Tier-0 screen’s full_buildout bound rather than the permitted project — the same storm peaks at 657 cfs and the basin holding it to the pre-development rate grows to 45 ac-ft (4.76-ft orifice). The two are separate readings of separate projects [inference: derived]; the as-permitted one is what the SW1225 application describes.

The campus’s storm-driven sanitary peak does not stay on site — it rides the forcemains to the treatment plants. It is judged only against the plants that actually receive it:

Campus sanitary routing: FM-1 → American Bath WWTP + American II WWTP; FM-2 → City of Lima WWTP. Receives campus flow but peak hydraulic capacity not cited (campus share not quantified): American Bath WWTP, City of Lima WWTP. Excluded — no campus routing: Shawnee II.

That 18.6 MGD is the campus’s total wet-weather sanitary peak; it splits across FM-1 (the small American Bath / American II plants) and FM-2 (the City of Lima sewer). The corpus does not quantify the split, so it is not apportioned — but the total alone is several times even American II’s whole wet-weather headroom (2.4 MGD), so the small FM-1 plants cannot absorb their share. The RDII rate is an uncalibrated screening assumption — but the direction is robust, and it lands on the regulatory fact below.

The surcharge lands on a system with no headroom to give. Permitted average / peak design flows are document-cited [verified]: American II 1.2/3.6 MGD (headroom 2.4); Shawnee II 3/12.6 MGD (headroom 9.6). The decisive fact is regulatory: the collection system is already under a 2005 OEPA mandate to eliminate all SSO bypassing by 2015, with $11.8M of storm-water I/I remediation and a 21-inch trunk replaced by 48-inch purely to equalize wet-weather I/I. So each plant’s nominal wet-weather headroom (peak minus permitted average) is already documented as effectively spent before the campus adds load. The campus’s documented dry-weather contribution is the 2.5 MGD FM-2 industrial discharge; the storm RDII multiplier on top remains an assumption.

Detention is the absent control, not a modeled redesign. The campus grading & stormwater plan (1A-C-3104, 95% SPS Design, [verified]) routes runoff via catch basins -> inlets -> storm sewer to headwall outfalls (with rock check dams and overland flood routing) and shows no detention, retention, or infiltration storage across its 207 storm-structure rims (820-829 ft). So the SWMM-sized basin is the on-site control the as-drawn 95% design omits. Pipe connectivity/inverts are drawn as vector geometry with no schedule table, so a routable network is deliberately not transcribed (omission over invention).

Sources: USGS NWIS (streamflow), NOAA Atlas-14 (design rainfall), NASA POWER (climate normals), Ohio EPA NPDES fact sheets 2PH00006 / 2PH00007 / 2IG00001 (receiving-stream 7Q10), Maumee Watershed Nutrient TMDL Appendix 4 (phosphorus WLAs), recorded Bistrozzi parcels (footprint). Regenerate with watermark hydro-report --write.