What it does to the water
A data center cools itself with water, and the water it gives back is not the water it takes — evaporation is a permanent loss to the basin. Before asking what the campus removes, read what the receiving streams can spare. Ohio EPA already wrote it down, in the low-flow number it uses to set the limits on every discharge.
NPDES fact sheet — the 7Q10 screen
Open the in-browser reader
Rendering oepa-2PH00006-american-ii-fact-sheet.pdf…
Ohio EPA sizes every discharge against the stream's design low flow — and the Ottawa this project discharges into runs at just 0.2 cfs, dropping to zero in the driest weeks. The tributaries are worse: American II's own fact sheet states a dilution of barely 1.3 to 1. The receiving water is near-undiluted before this project adds a drop.
The river is already effluent
Before the campus takes a drop, read what the river already carries. At design low flow the three county WWTPs discharge 37.44 cfs of treated effluent into streams whose natural low flow totals just 1.21 cfs — and the campus adds its own routed 3.87 cfs FM-2 discharge. The Ottawa leaving Lima runs 97% treated effluent, before counting a drop of what the campus evaporates.
| Discharger | Receiving stream | Discharge | 7Q10 |
|---|---|---|---|
| Shawnee II WWTP | Ottawa River | 4.64 cfs | 0.20 cfs |
| American Bath WWTP | Pike Run | 2.32 cfs | 0.03 cfs |
| American II WWTP | Dug Run | 1.86 cfs | 0.78 cfs |
| Lima WWTP | Ottawa River | 28.62 cfs | 0.20 cfs |
⌖ feed assimilative (NPDES fact sheets) + the campus's cited routed discharge (hydrology-scenarios)
That is what the campus discharges into. It also takes water out of the basin:
a data center cools by evaporation, and the water it gives back is not the water it takes. But
read the next comparison carefully — it is a worst-case bound, not the operating
reality. Lima doesn’t pump from the river at low flow; its supply is five off-stream reservoirs
holding ~14.4 billion gallons — roughly 70% of it fed from the Auglaize (Bresler and
Williams), the balance from the Ottawa — filled by pumping at high flow and drawn down through
the dry months. So setting the campus’s net loss against the Ottawa’s low flow measures the
scale of the basin stress, not a withdrawal from the river — the honest tag is
[inference], not [verified]. The reservoir sourcing is [reference], from the City’s
utility record.
Read the operating constraint the right way and it moves from the river to the reservoirs. The
campus draws treated municipal water, so its makeup demand — modeled at about 3.9 MGD, a
fifth of Lima’s ~19 MGD production — draws down that shared storage, while its returns leave
downstream through the public WWTPs, not back into the reservoirs. Run the zero-refill drought
buffer and the campus shortens it from roughly 961 to 762 days (−199). That reservoir
drawdown — not the instantaneous river depletion the screen below bounds — is the binding
low-flow constraint, and the pump stations have to recapture the added draw (~1.4 billion
gallons a year) from the Auglaize and Ottawa when the rivers run high. It regenerates from the
live refill gauges via watermark supply. [inference], a modeled draw on [reference]
storage.
There is a second reason the draw is modeled rather than read. The one number that would
settle it — the cooling system’s design flowrate — was claimed by the developer as a trade
secret, and Ohio EPA granted the claim on 2025-10-08 (eDoc 3859883). The
justification is unusually candid about the stakes: the air permit, it argues, is “the only
public document that requests the size” of the equipment, so “by preventing this information
from becoming public via the air permitting process, the facility has taken a significant
step to ensure the confidentiality of this information.” The flowrate is not missing by
accident — the single public path to it was identified and closed. So we model the draw, tag
it [inference], and can say precisely why the [verified] figure
does not exist.
And because the screen is built from live gauges, you don’t have to take a static number on faith — the hydrology dashboard re-runs it against the Ottawa’s current flow from USGS on every build.
What the campus takes out of the basin
| Season | Ottawa low flow | Loss ÷ low flow | Source |
|---|---|---|---|
| Annual 7Q10 | 0.20 cfs | 24× | Ohio EPA NPDES fact sheet 2IG00001 (Lima Refining Co.), Stream Flows table — Ottawa River at Lima, USGS gage 04187100, 1989-2021 |
| Summer 30Q10 | 1.60 cfs | 3.0× | summer 30Q10 1.6 cfs — Ohio EPA NPDES fact sheet 2IG00001 (Lima Refining Co.), Stream Flows table — Ottawa River at Lima, USGS gage 04187100, 1989-2021 (Ottawa mainstem at Lima nearly dries at design low flow (1Q10 = 0 cfs) — heavily abstracted upstream for Lima's water supply) |
| Driest week 1Q10 | 0.00 cfs | ∞ (dry) | driest-day 1Q10 0 cfs — Ohio EPA NPDES fact sheet 2IG00001 (Lima Refining Co.), Stream Flows table — Ottawa River at Lima, USGS gage 04187100, 1989-2021 (Ottawa mainstem at Lima nearly dries at design low flow (1Q10 = 0 cfs) — heavily abstracted upstream for Lima's water supply) |
Now weigh that against what the proponents say. AEDG’s own FAQ assures that
data centers “increasingly use closed-loop cooling systems that recirculate
water” — but the air permit you read in the last chapter fixes
36 evaporative cooling towers, and evaporative towers lose
water to the sky by design. Google’s “120% water replenishment” pledge is
real, but it’s global and offsite — it says nothing about the
Ottawa. The tower count is [verified]; the consumptive draw above
is [inference]; the claim that this “reduces water use” here is
neither.
Where the water goes when it's done cooling
The water an evaporative tower doesn’t lose to the sky comes back dirtier than it left.
The towers concentrate whatever is in the make-up water — dissolved salts, treatment
chemicals — into a waste stream called blowdown, and that stream has to go
somewhere. For Project BOSC, the somewhere is on the record. In August 2024 — more than a
year before the public knew whose data center this was — the County Commissioners approved
Resolution #679-24: a $47,600 task order to MS
Consultants for a “WWTP Data Center Flows Treatment Evaluation,” studying the campus
blowdown’s demands, its flows over a typical year, and its water-quality parameters for
storage and treatment at the American Bath WWTP — the County to gather
the figures from “the developer and data center manufacturer.” It is the first
primary-source document tying the cooling system to the public sewer.
[verified], from the produced record.
What that study scopes, the build-out commits to: a dedicated BOSC pump station and dual forcemains — a 10-inch line to the American Bath plant and a 16-inch line to Lima’s existing 78-inch interceptor, sized for 2.5 MGD peak (Res #137-26). The campus’s waste heat leaves the site as water, and the public sewer system is being enlarged to receive it — the cost of that enlargement is the next chapter’s business.
The water also comes back warmer
“The campus's waste heat leaves the site as water” is not a figure of speech. Read the same sentence as a physics problem and it becomes the third question this chapter has to ask. Volume is what the campus takes. Chemistry is what the blowdown carries. Heat is the one nobody had a number for — and Ohio writes a numeric standard for it.
For the Ottawa River in this zone and season, that standard is a daily maximum of 29.4 °C (OAC 3745-1-35 Table 35-11 (G)). The design ambient — the river's own reported temperature, measured at a permit-required in-stream station on this reach — runs 24 °C. The whole heat budget is the difference: 5.4 °C of headroom, which at the drought flow Ohio permits at works out to about 0.128 MW of heat the river can absorb before it breaches the standard. A campus rejecting on the order of 316 MW of condenser heat does not fit inside that budget under any of the three ways the heat could be partitioned — not even the smallest, where only the tower blowdown reaches the water. [inference], from the disclosed IT load.
But the sharper finding is the one that needs no model at all. Of the permits already discharging to this reach, 2 report effluent temperatures at or over Ohio's criterion — their own submissions, in their own filings. And they report them from outfalls their permits require them to measure but never cap: not one of the 3 permits reporting a temperature on this reach (OH0002623, OH0002615, OH0026069) carries a numeric thermal limit on its discharging outfall. That is a cited absence, not a clean bill of health. [verified], from the reported record.
None of this is a violation, and the screen does not say it is. It says the river's thermal budget is already spent, that the record shows no one holding a limit against it, and that a new heat load of this size is the kind of question the Clean Water Act answers with a §316(a) demonstration — which nobody on this reach has been asked to make.
It starts in a soybean field
All of it lands on ground that drains the wrong way for it. Of the ~340 acres assembled into the campus (the Brenneman → Bistrozzi LLC deed — the number the landing headlines), roughly 309 acres of former CAUV farmland are the graded stormwater footprint — the construction-stormwater coverage (facility 2GC08468, Turner Construction) was modified to 309.2 acres in June 2026 — graded to discharge into Pike Run, a headwater tributary of the same Ottawa system the 7Q10 above describes. Two on-site isolated wetlands, 0.33 acres of Category-1 (0.29 forested, 0.04 not), were authorized for fill on 2025-08-12, the loss offset by credits bought at the Pearson Metropark mitigation bank a county away (eDoc 3788677).
A broader fill across the site’s 358-acre wetland delineation went differently. The developer applied for it on 2025-12-09 — the day after the first erosion-control inspection recorded clearing and mass grading already underway — and Ohio EPA returned the application incomplete two weeks later, for missing an analysis of practicable on-site alternatives (eDoc 3949585). The corpus doesn’t show how that one resolved. What it shows is the order: the earth was moving before the wider wetland question was answered. Analysis of the produced record, not a legal conclusion.
Explore this in the library
The finding this chapter teaches toward → The impact study, §II·4 Water supply & consumptive use