CL-WP-004 / WORKING PAPER / v1.0

Peak Water, Not Average Water.

A reproducible public-evidence capacity screen for data-center growth in Loudoun County, Northern Virginia.

Abstract. This release tests how published utility and regional water evidence can be translated into a transparent compute-capacity screen without treating annual averages, nominal future system capacity, reclaimed supply and hydrologic consumption as interchangeable quantities. The contribution is a peak-water screening method and evidence workflow—not a utility will-serve determination or a final Critical Load rating.

Download PDF Data & model

All publications Data & reproducibility Related research Methodology

Research claim boundary

Peak-day water demand is treated as a necessary first-order utility service-capacity screen, not as a complete hydrologic carrying-capacity determination.

01 / Published baseline

The infrastructure signal becomes materially stronger at the peak.

The ratios below are Critical Load calculations using Loudoun Water's published baseline values. They should be read as ratios of published values, not as proof that each system and data-center maximum occurred at the same instant.

9.0%2.6 MGD / 29.0 MGD · published average-day values

16.0%8.0 MGD / 50.0 MGD · published maximum-day values

3.08×Data-center potable peak / average ratio

2.1–5.2MGD peak per added GW · official regional scenario range

~1.8×

The data-center peak/average ratio is about 1.8 times the system-wide peak/average ratio using the published baseline figures.

Peak first

An annual-average water metric can materially understate the service-capacity question relevant to large additions of compute.

Convert an explicit peak-water budget into compute without hiding the allocation assumption.

The PWCS starts from a declared maximum-day potable allocation. It deliberately does not call nominal future utility capacity “available headroom.”

Peak Water Capacity ScreenB = a × C_future − D_{DC,peak,current}
P_screen = max(0, B / I_peak)

Illustrative v1.0 values: Cfuture = 90 MGD; a = 16%; current published data-center peak potable demand = 8 MGD. The 16% allocation is an E4 analyst scenario, not utility policy.

3.05 GWLow · 2.1 MGD/GW

2.21 GWMedium · 2.9 MGD/GW

1.23 GWHigh · 5.2 MGD/GW

0.75 GWFacility-high sensitivity · 8.5 MGD/GW

These are not estimates of “remaining Loudoun capacity.” They demonstrate the consequence of one explicit allocation rule and different cooling-intensity assumptions. A verified capacity result requires project-specific utility and hydrologic evidence.

Reclaimed supply can reduce potable burden without removing total consumption.

The method keeps water source and hydrologic consumption separate. This prevents reclaimed water from being treated as “zero water.”

Source and consumptionPotable burden = D_total × (1 − r)
Estimated consumption = D_total × c

v1.0 uses the ICPRB 75% consumptive factor only as a transparent regional-analysis sensitivity. Project cooling cycles, blowdown, discharge and seasonality should replace it when available.

Reclaimed share	Potable peak / +1 GW medium case	Reclaimed peak	Consumption sensitivity
0%	2.90 MGD	0.00 MGD	2.175 MGD
50%	1.45 MGD	1.45 MGD	2.175 MGD
100%	0.00 MGD	2.90 MGD	2.175 MGD

The output is a screen and an evidence-request workflow.

Suitable forEarly-stage infrastructure decisions

Site screening; infrastructure diligence; cooling/source comparison; scenario comparison; utility data requests; framing project-specific evidence gaps.

Not suitable forFinal service or regulatory determination

Utility will-serve; hydraulic design; permit approval; drought allocation; water-right determination; final Critical Load rating.

Compute basis

Declare IT MW and project phasing.

Water boundary

Potable, reclaimed, watershed and discharge.

Peak evidence

Average plus maximum-day demand.

Cooling cases

Project WUP or explicit scenario bounds.

Escalate

Hydraulics, hydrology, drought and review before rating.

Paper → source register → inputs → code → model → scenario outputs.

The release package is designed so the headline arithmetic can be recomputed without trusting the website.

Working PaperPDF · v1.0 Research BriefPDF · 3 pages Capacity ModelXLSX Model InputsCSV Reproduction CodePython Source RegisterCSV MethodMarkdown ManifestSHA-256 Complete PackageZIP

The next improvement is external evidence, not more decorative modelling.

v1.0 intentionally stops before a final water-constrained compute envelope. The unresolved gates are substantive: utility service/hydraulic evidence, project-specific seasonal WUP, low-flow and drought conditions, wastewater/blowdown treatment, project phasing and independent domain review.

Release interpretationUse this as a reproducible public working method for screening and diligence. Do not cite it as a peer-reviewed final carrying-capacity determination.

Cite the release at the strength it actually supports.

Critical Load Research Group (2026). Peak Water, Not Average Water: A reproducible public-evidence capacity screen for data-center growth in Loudoun County, Northern Virginia. CL-WP-004 / R04 / CL-P03 / CL-001. Working Paper v1.0, 19 September 2026. DOI pending.

Citation text R04 programme CL-001 application

Primary and official evidence remains identifiable.

The web summary does not replace the source register. Core inputs are linked below, while the downloadable CSV records evidence class, use, access path and limitations.

Source	Role in v1.0	Open
Loudoun Water · Data Center Water Use	Published average-day, maximum-day and future permitted-capacity baseline	Source ↗
Loudoun Water · Reclaimed Water Program	Local reclaimed-water context and service framing	Source ↗
ICPRB · 2025 WMA Water Supply Study	Regional direct-water scenarios and planning context	Study ↗
ICPRB · Data Centers and Water Use	Peak, seasonality and consumptive-use context	Source ↗
R04 source register	Machine-readable evidence trail and evidence classes	CSV →