Water, Land, and Noise:
The Non-Electric Constraints
That Kill Projects

Power projects rarely die on economics. They die on a nighttime noise limit written into a municipal code decades ago, a gas main that turns out smaller than the proposal assumed, a fire separation that does not fit the yard, or a water connection the local purveyor declines to enlarge.

Section 01The veto class

Every serious energy evaluation produces two kinds of findings. The first kind is continuous: a price is higher or lower, an output is larger or smaller, a payback is longer or shorter. Continuous findings move a decision. The second kind is binary: the ordinance is met or it is not, the setback fits or it does not, the water is available or it is not. Binary findings end a decision, and no spreadsheet resurrects a technology that a site fact has vetoed.

The binary constraints share three properties that make them dangerous. They are local: noise limits, fire-code amendments, and water policies are written jurisdiction by jurisdiction, so nothing learned on the last project can be assumed on this one. They are physical: no financing structure shortens a lot line, and no incentive enlarges a gas main. And they are discovered late by default, because they live in documents that projects tend not to read until permitting begins: municipal codes, general-plan noise elements, fire-code local amendments, utility service records, and water-supplier policies. Late discovery is the expensive part. The screen itself is disciplined reading, and disciplined reading is cheap.

Air permitting is the famous member of this family and is treated at length in companion papers in this series. This paper covers five constraints that get less attention and kill just as reliably: water, land and setbacks, noise, fire code, and fuel availability. Each section states what the constraint is, how the major technology classes fare against it in both directions, and which fact settles the screen. The paper closes with the screening matrix we start from and a six-question site screen an owner can run in the first week.

Section 02Water: the constraint that arrives with the cooling choice

Start at grid scale to calibrate the stakes. In the federal government's most recent national census of water use, thermoelectric power plants withdrew 133 billion gallons per day in 2015, 41 percent of all water withdrawn in the United States for any purpose.¹ Withdrawal is not consumption: most of that water passed through once-through cooling systems and returned to its source, and consumptive use ran near three percent of thermoelectric withdrawals.¹ The pair of numbers establishes the central fact anyway. Generating electricity with heat is, by default, a water business.

The peer-reviewed literature is unusually clear about what drives the number: cooling architecture, not fuel. The most widely used compilation of operational water factors reports that a gas combined-cycle plant on wet cooling towers consumes a median of roughly 205 gallons per megawatt-hour, with a range of about 130 to 300, while the same plant dry-cooled consumes a median of about 2. Once-through systems consume little but withdraw enormously, a median near 11,380 gallons per megawatt-hour.² Choose the cooling system and you have largely chosen the water bill.

At distributed, behind-the-meter scale the defaults are friendlier, which is exactly why the screen gets skipped. Packaged reciprocating engines, microturbines, and small gas turbines are ordinarily radiator- or air-cooled, so generation water is close to zero. The water re-enters through side doors. Add a steam bottoming cycle, evaporative inlet cooling for hot-day output, or a combined-heat-and-power absorption chiller with its own cooling tower, and the site acquires tower duty, make-up water demand, chemical treatment, and Legionella-management obligations it did not have on paper.

Electrochemical platforms vary by design: some consume modest volumes of treated water for fuel reforming, others run close to water-neutral in steady operation. The screening fact is not a slogan about the class but the data sheet: the water connection, the required water quality, and the reject stream from on-site treatment, since deionization systems discard a meaningful fraction of their feed. Solar photovoltaics consume essentially no operating water beyond periodic panel washing. Battery storage consumes none in operation, yet can still acquire a water requirement through the fire code, where some officials require documented fire-protection water supply near installations. Water is not always an operating input; sometimes it is a code artifact.

Supply is only half the screen. Cooling-tower blowdown and treatment reject must legally go somewhere: a sewer connection under an industrial discharge permit with limits on dissolved solids, or trucking, or on-site treatment. Sites without sewer access discover this constraint expensively. And in California the supply itself is a commitment, not an assumption. The settling fact is a written statement from the water purveyor that the incremental demand can be served. A technology whose water balance exceeds what the supplier will put in writing is disqualified at any price.

The trade cuts both ways, which is why this is a screen and not a verdict. Wet cooling buys thermal efficiency and hot-day output at the cost of consumption, plume, and treatment. Dry cooling buys water independence at the cost of parasitic fan load and output derates on hot afternoons, precisely when many California loads peak. Neither is right in general. One of them is usually right at a specific site, and the site's water facts decide which.

Section 03Land: the footprint is not the equipment

The land question is really three questions, and proposals routinely answer only the first. Equipment footprint is what the brochure shows. Effective footprint adds what codes require around the equipment: separation distances, fire-apparatus access, working clearances, ventilation and exhaust offsets, and the acoustic buffer that Section 04 will impose. Construction footprint adds laydown area, crane pads, and delivery routes that must survive for months. Sites that hold the first number comfortably fail the second and third with regularity.

Power density separates the technology classes more starkly than almost any other property. Combustion and electrochemical plants are compact: continuous output from equipment yards measured in parking-stall counts. Solar's fuel is area. The most complete federal survey of built plants found utility-scale solar requiring a capacity-weighted average of 8.9 total acres per megawatt of alternating-current capacity across the projects studied.³ Because photovoltaic capacity produces at a fraction of nameplate around the clock, the area per unit of continuous power is several times larger still. Rooftops and canopies change the currency rather than the price: acreage becomes structural dead load, seismic anchorage, roof-warranty terms, and attachment engineering, all of which are answerable questions that must actually be asked.

Setbacks and encumbrances are the quiet killers in this class. Zoning setbacks at lot lines; recorded utility and pipeline easements that read as open land on a site walk but cannot be built over; flood zones; airport overlay height limits near runways. The screening documents are unglamorous: the title report, the survey, the zoning district table. Reading them costs an afternoon. Not reading them costs a redesign.

Stated honestly in both directions: compact classes concentrate their externalities, so one corner of the yard now holds the noise, the exhaust, and the hazardous materials, and mitigation spends money in that corner. Area-hungry classes escape those concentrations and spend the site itself, and at industrial facilities the land they consume is often the expansion room the project was meant to serve. There is no dominant row here. There is a fit.

Section 04Noise: regulated town by town, enforced at night

Noise regulation in California has a structure worth understanding, because the structure explains the surprises. State planning law requires every city and county general plan to contain a noise element, with community noise analyzed in metrics, community noise equivalent level or day-night average level, that add penalties to evening and nighttime sound.⁴ Municipal ordinances then set the enforceable numbers: exterior limits measured at the property line, commonly stepping down after evening hours. Two consequences follow. First, the binding limit is local, town by town and sometimes overlay by overlay. Second, the metrics are built to penalize exactly what continuous generation does, which is run at two in the morning.

As an illustrative range only, nighttime exterior limits in the mid-40s to mid-50s of decibels are common in California municipal codes, and compliance is measured at the nearest property line, not at the site average. Character matters as much as level. The low-frequency exhaust note of reciprocating engines carries far, penetrates buildings, and draws complaints out of proportion to its measured level; some ordinances add explicit penalties for tonal or impulsive sound. Turbine noise sits higher in frequency and yields more readily to silencers. Electrochemical plants, photovoltaic inverters, and battery thermal-management systems are markedly quieter, but quiet is relative to the limit, not to daytime traffic: a fan bank or transformer hum a short distance from a bedroom window at night can still fail.

Every mitigation that works spends one of two currencies. Enclosures, silencers, and barrier walls spend capital. Distance spends land, and Section 03 already priced land. The two-way reading: reciprocating engines fail this screen more often than any other class, and they also carry the most mature, most accurately priced mitigation industry in the field. The quiet classes pass the screen more often but are not entitled to pass it unexamined. And a compliant plant can still generate complaints; complaint-driven permit conditions, with restrictions on testing hours as the classic example, carry real operating cost.

A site constraint discovered in month nine costs what it would have cost to learn in week one, plus the nine months.

Section 05Fire code: the newest screen moves the fastest

The fire code is the newest member of the veto family and the fastest-moving. California regulates stationary battery installations under the energy-systems chapter of the California Fire Code, which draws on the national standard for stationary energy storage. The load-bearing requirements are consistent: systems must be listed to the applicable product-safety standard; where thresholds are crossed, large-scale fire testing must demonstrate that a fire in one unit does not propagate to the next; and installations carry separation distances from lot lines, buildings, and each other, per-unit and aggregate energy limits, monitoring, and fire-service access.⁵ The specific distances and limits depend on the listing, the test results, and the local authority, which is the practical point. The fire official is a design input, and the correct time for the first conversation is before technology selection, not after.

This body of rules has been rewritten repeatedly within a single equipment generation, and local amendments frequently exceed the state minimums. A storage layout that screened clean against last cycle's code deserves a fresh check at permit time; assuming otherwise has cost real projects real quarters.

Fueled classes carry an older, more settled version of the same screen. Gaseous fuel brings gas-train requirements, seismic shutoff, and ventilation rules for indoor equipment, and hydrogen-capable installations bring their own separation tables under the hydrogen technologies code. Liquid fuel brings tank rules, secondary containment, and hazardous-materials inventories that are bureaucratically routine but not free. Rooftop photovoltaics bring access-pathway and rapid-shutdown provisions. None of these is ordinarily fatal; all of them consume schedule, and several consume yard space, which loops back to Section 03. The class-by-class contrast is fair to state plainly: combustion's fire-code burden is mature and predictable, storage's is heavier and still moving, and solar's is light and largely standardized.

Section 06Gas availability: the assumption that fails silently

Every gas-fueled proposal contains an assumption so basic it is rarely written down: that the gas system at the street can serve the machine. The screening facts are physical. The diameter and operating pressure of the main; the distance from an adequate main to the meter location; whether the pressure the equipment needs exceeds what the local distribution system delivers. Turbines and microturbines generally want higher inlet pressures than engines, and where the street cannot supply it, a fuel-gas compressor joins the project, bringing capital, parasitic load, maintenance, and one more noise source for Section 04.

The regulatory ground shifted recently enough to deserve a date. In September 2022 the California Public Utilities Commission eliminated ratepayer-funded gas line extension allowances for requests submitted on or after July 1, 2023, the first state to do so; new gas main and service extensions are now built at the applicant's cost.⁶ The decision was aimed at building decarbonization, but its screening consequence for on-site generation is direct. If the site needs an extension, that cost belongs in the model at full freight, with the utility's written estimate and schedule attached, not a legacy assumption that the utility funds the pipe.

Service class matters as well. Large gas users typically take noncore or transport-level service whose curtailment provisions are worth reading before a firm-power commitment is built on top of them, not after. Contracted renewable gas is an accounting instrument layered on the same physical molecule and does not change the pipe. On-site biogas, where a site has it, changes the screen entirely and brings its own gas-quality questions, since some equipment classes tolerate sulfur and siloxanes far less than others and cleanup equipment is its own line item.

Stated both ways: the gas-fueled classes offer the most compact firm power an owner can site, and their fuel path is now a project in its own right, with a real cost and a real calendar. The electric-only portfolio escapes this section entirely and re-enters Sections 02 through 05 carrying the largest land case and the heaviest fire-code case on the table. No path skips the screen. Different rows fail different columns.

Section 07The screening matrix

The discipline that catches vetoes early is not sophisticated. It is a matrix: technology classes down the side, constraints across the top, and in every cell a pass, a priced fix, or a documented kill. The table below is the qualitative version we start from, with rows for reciprocating engines, gas turbines and microturbines, electrochemical plants, solar photovoltaics, and battery storage. A real study replaces every cell with the site's own facts and sources, and newer classes, such as linear generators, run through the same columns with an added diligence line on fleet history.

Two features of the matrix do the work. First, no row sweeps its columns. The class that wins water tends to lose noise; the class that wins noise tends to lose the fire-code column; the class that wins every site column still answers for its fuel. Second, every kill is written down with its source: the ordinance section, the code edition, the utility letter, the purveyor's policy. A documented kill is knowledge. An undocumented one is an opinion that will be relitigated every quarter until someone finally documents it.

Section 08Run the screen before the model

Six questions settle most of this in the first week, before any economic model earns the right to be built.

1. Get the water balance in writing.The candidate technology's water and drain connections from the data sheet, the treatment reject stream, and a written statement from the water purveyor that the incremental demand can be served.
2. Draw the effective footprint, not the equipment footprint.Equipment plus code separations, working clearances, apparatus access, and construction laydown, drawn on the survey with easements and setbacks shown.
3. Pull the nighttime number at the nearest lot line.The municipal ordinance section itself and the general-plan noise element. Model the quietest hour against the closest receptor, not the daily average.
4. Meet the fire official before selecting the technology.A pre-application conversation, the local amendment list, and the current code edition. For storage, ask what testing documentation the office expects to see.
5. Get the gas answer from the utility, dated.Main size and pressure at the street, extension cost and schedule at applicant expense, and the curtailment terms of the service class the site would actually take.
6. Name the discharge path.Where blowdown, reject, and washwater legally go, under whose permit, at what limits. No sewer access is a finding, not a footnote.

None of these six questions requires engineering. They require reading documents that exist today, and the willingness to let the answers eliminate options before anyone has fallen in love with a machine. In our study work the matrix runs before the economics: every path that dies earns its elimination in writing, and the paths that survive get modeled with their constraints priced in rather than discovered later.

The pattern behind every section of this paper is the same. A seller screens against its own machine's constraints last, if at all, because a veto found early ends a sale. An owner's screen runs in the opposite order: vetoes first, economics second. That ordering is most of what independence means in practice, and it is the least expensive risk management available in this field. Six questions, one week, and the documents are free.