Cooling Tower Wet Bulb Temperature: How It Sets Data Center Heat Rejection

Ask an operator what the cooling tower controls to, and the answer is always a water temperature. Ask what limits it, and the honest answer is the weather — specifically, the wet-bulb temperature of the air entering the tower. Every other number on a cooling tower selection sheet is derived from that one number, and it is the only one the site cannot control.

This matters more in a data center than in any other building type, because a data center tower is asked to do two jobs that pull in opposite directions. In summer it has to reject the entire heat load of the halls at a water temperature the chillers can live with. In winter it has to make water cold enough that the chillers can shut off entirely. Both jobs are bounded by wet bulb, and the same as-selected tower behaves very differently in the two seasons.

This article covers what wet bulb is and why it is the right reference (not dry bulb), how approach and range turn physics into a selection number, the ASHRAE design frequencies and which one a data center should use, how the tower's own approach degrades as the wet bulb falls, how to read a wet-bulb-driven free-cooling budget, the water and power cost that rides along, and the commissioning and diagnostics traps that make a correctly selected tower underperform. It ends with a complete wet-bulb-driven heat rejection review for a 5 MW liquid-cooled hall.

Start with the limit, not the equipment

A cooling tower rejects heat by evaporating a small fraction of the water it circulates. The remainder is cooled by that evaporation. Because the mechanism is evaporation, the theoretical floor for the leaving water temperature is not the air temperature you read on a thermometer — it is the temperature the air can reach if it is fully saturated by evaporation. That is the wet-bulb temperature [1][2].

The distinction is not academic. On an Arizona afternoon the dry bulb might be 110°F while the wet bulb is 70°F. On a Miami afternoon the dry bulb might be 91°F while the wet bulb is 80°F. The Phoenix day is 19°F hotter in the way a person feels it and 10°F cooler in the way a cooling tower experiences it. A tower sitting in Phoenix on that day can make colder water than a tower sitting in Miami [3][4].

Table 1 — Dry bulb is the wrong reference for evaporative heat rejection. Two sites can differ by 19°F in air temperature and rank in the opposite order by cooling potential. ASHRAE design values shown are representative 1% summer conditions for illustration [3][4].

The physical reason the tower cares about wet bulb and not dry bulb is enthalpy. Heat and mass transfer in a cooling tower fill is driven by the difference in enthalpy between the air and the saturated air film at the water surface. Lines of constant enthalpy on a psychrometric chart run nearly parallel to lines of constant wet-bulb temperature. That is why a tower selection curve can be drawn against entering wet bulb alone, and why two days with the same wet bulb but different dry bulbs produce the same leaving water temperature from the same tower at the same flow and load [1][5].

Turn the physics into a selection number

Three temperatures fully define a cooling tower's operating point: the hot water entering the tower (HWT), the cold water leaving it (CWT), and the entering air wet bulb (WBT). From those, everything else follows [2][5][6].

The five expressions every tower conversation reduces to. Approach is the performance number; range is a load number. Range tells you how much heat is moving. Approach tells you how good the tower is at moving it [2][6].

Work the standard rating point through and the numbers line up:

The 15,000 BTU/hr per ton figure is the reason tower capacity is not the same number as chiller capacity. A 1,000-ton chiller needs a tower sized for roughly 1,250 tons of heat rejection [2].

Range is set by the load and the flow, and it moves all day. Approach is set by the tower's physical size, the fill surface area, the airflow, and the wet bulb. Approach is therefore the parameter you can design, buy, and later degrade — and it is the one to put in the specification.

The two variables you trade against each other at selection are flow and range, and the trade is not free. Reselecting the same tower from 3 gpm/ton down to 2 gpm/ton pushes entering water from 94.3°F to 96°F while leaving water drops from 85°F to 82°F — same tower, same fan power, but a range of 14°F instead of 9.3°F, and an approach that tightens from 7°F to 4°F. Lower flow per ton buys a tighter approach, which is why warm-water data center loops are often specified at 2–2.5 gpm/ton rather than the traditional 3 [17].

Pick a design wet bulb and defend it

ASHRAE publishes design weather conditions at several annual exceedance frequencies. The number is the percentage of hours in a typical year the value is expected to be exceeded [3][4][7].

Table 2 — The design frequency is a risk decision, not a climate fact. For a data center the relevant question is what happens during the 35 worst hours of the year, because that is when the halls are running flat out and the chillers are already at their highest condensing pressure [7].

The gap between 0.4% and 1% looks trivial on paper and is not. In South Florida the 0.4% design wet bulb runs about 79–80°F while the 1% value sits roughly a degree lower. Size the tower to the 1% number and on the afternoon the site actually reaches 0.4% conditions, the tower cannot make design leaving water temperature. Condenser water runs hot, compressor head pressure climbs, kW/ton rises on the worst day of the year, and in extreme cases the chiller trips on high pressure [8].

Two more selection rules follow from the standard's own baseline language. ASHRAE 90.1's performance rating method derives a baseline approach of roughly 25.72 − (0.24 × WB) for a 10°F range in the 55–90°F wet-bulb band — about 7°F at a 78°F wet bulb and about 12.5°F at 55°F. That is a compliance floor, not a design target. Second, the standard requires fluid economizers to deliver up to 100% of cooling load at 50°F dry bulb / 45°F wet bulb and below, which is the condition that fixes the winter end of the tower's duty [5][9][10].

Economizing is not a marginal proposition in the climates where it applies. Energy Star's threshold is a wet bulb below 55°F for 3,000 or more hours a year, which covers all but the extreme Southwest and the lower Southeast. The consequence for plant design is that chillers and economizers can no longer be piped in parallel — they must be able to run simultaneously, and because a chiller has a minimum condensing temperature, the plant may need two different tower cold-water supply temperatures rather than one [20].

Accept that the tower's approach is not a constant

This is the counterintuitive part, and the one that causes the most design errors. Approach is not a fixed property of a tower. It is a function of the wet bulb it is operating against. The same as-selected tower has a tight approach on a hot day and a wide approach on a cold day, because cold air holds less moisture and therefore cannot absorb as much evaporated water. When the wet bulb falls, the tower is handicapped, not helped [9][10][11].

Table 3 — Approach at a constant 100% heat rejection load and 10°F range, for one tower selected at a 78°F / 7°F design point. Between 45°F and 78°F of wet bulb the approach worsens by a factor of about 2.3; extending the range down to 30°F wet bulb the change reaches a factor of roughly six [9][11]. Values derived from published tower performance curves.

Read the last two columns together and the operational trap appears. At a 45°F wet bulb the coldest water this tower can make at full load is 61°F — which is excellent in absolute terms. But the tower only got there because the outside air was cold. Its approach, the number you bought, has degraded from 7°F to 16°F. An operator comparing 61°F against a nameplate designed around 85°F might conclude the tower is performing superbly. It is not; it is performing at about two-thirds of its rated effectiveness, and it will be the reason a free-cooling target is missed in shoulder season.

The practical design response is not to buy a tower for the coldest winter day — that is uneconomic and the ASHRAE curves show why. It is to size the tower for the summer design point and then operate it at part load in cold weather, typically by piping one or more towers in series with the economizer, so that the tower stays in the region of its curve where approach is reasonable while still making cold enough water to keep the chillers off [10][11]. Operating practice on the plant side matters as much as the tower itself: trend the plate-exchanger approach and the free-cooling hours season over season, and keep the changeover control point — the wet-bulb threshold with its offset and deadband at which the plant enables or drops free cooling — under review, because a drifted setpoint looks exactly like a fouled exchanger in the data [18].

Convert wet bulb into free-cooling hours

A waterside economizer makes chilled water from the tower alone, with the chiller compressor off or unloaded. Whether it can do so on any given hour is decided by the wet bulb, and secondarily by the approach of every component in the chain between that wet bulb and the load [10][11][12].

The chain is worth writing down, because each link adds to the total:

Every degree you can relax on the facility water supply temperature buys wet-bulb margin, and therefore free-cooling hours. This is why the industry has moved toward warm-water cooling.

Warm-water operation is the single highest-leverage decision for a wet-bulb-driven plant. A 32°C facility water supply is now standard practice at scale, and NVIDIA's DSX reference documentation describes a 45°C technology cooling supply at the rack as a design point that “can widen the opportunity for dry cooling and reduce cooling overhead when site and ambient conditions permit.” The higher the supply temperature the CDU can work with, the more of the year the tower can carry the load without mechanical cooling [12][13]. What that supply temperature does to the rest of the loop — flow rate, pipe diameter, pump head — is covered in our guide to CDU pipe sizing and coolant loop diameter.

Table 4 — Raising the facility water supply temperature moves the entire free-cooling calculation. The tower does not change; the amount of the year it can carry the load does. Uptime Institute reports liquid-cooled systems running on free cooling 50% to 80% or more of the time, and the spread between those two numbers is mostly this table [12].

Know when wet bulb stops being the constraint

Wet bulb is the limit for anything that evaporates. It is not the only heat rejection option, and at the extreme dry end it stops being the right reference entirely [7][12].

Table 5 — The right-hand column is the design constraint, and it is set by climate, not by preference. A hybrid sits deliberately in the middle: full evaporative benefit during the worst hours, dry operation the rest of the year.

LBNL's warm-water analysis puts the climate question in design terms. Across the national laboratory sites assessed, the 0.4% (99.6%) wet bulb was at or below 26.5°C (79.7°F), while the 0.4% dry bulb was at or below 37.5°C (99.5°F). The report therefore uses 26.5°C wet bulb to assess designs where a cooling tower is the primary cooling source, and 37.5°C dry bulb where dry coolers are. Those 35 exceedance hours are not spread evenly through the year; they concentrate in hot daytime hours, which is precisely when a data center cannot afford to be short of capacity [7].

This is the arithmetic behind every “can we go waterless?” conversation. Going dry converts a wet-bulb problem into a dry-bulb problem, and in most US climates that trade spends fan energy, compressor energy on design days, and plot space instead of water [7][12].

Read the water and power bill that rides along

Wet bulb does not just set temperatures; it sets the size of the water and energy bill. Both can be computed from the same three temperatures [14][15][16].

The 1.8 gal/ton-hour evaporation figure is worth memorizing; it is the basis of most published data center withdrawal estimates, which typically fall in the 1.5–2.5 gal/ton-hour range depending on cycles of concentration and drift [14][15].

Run that against a 5 MW liquid-cooled hall. At a 1.20 PUE the total heat rejection is 6,000 kW, and a 10°F range puts the circulating condenser water flow at about 4,100 gpm:

Table 6 — 5 MW IT load, 1.20 PUE, 10°F range, 4,095 gpm circulation, 0.005% drift, computed at 100% load and 8,760 h/yr. Evaporation alone is 21.5 M gal/yr and is irreducible; the leverage is entirely in blowdown, which is set by cycles of concentration and therefore by water treatment [14][15].

Wet bulb also shapes the annual picture, because in cold weather a greater fraction of tower heat rejection is sensible rather than latent, so less water evaporates for the same heat load. The effect works in the operator's favor at northern sites, where the annual make-up total for the same hall is materially lower than at a hot–humid site even though the tower must work harder in the shoulder seasons [9][14].

Commission and diagnose against wet bulb

A tower selected correctly against the right design wet bulb can still fail to deliver, because approach degrades for reasons that have nothing to do with the tower's rated capacity. The first diagnostic step is always to measure the wet bulb at the tower intake at the same moment as the leaving water temperature, and to compare only under representative load [5][6].

Table 7 — The first five rows are physical faults. The sixth is the most common of all, and it is a measurement error, not an equipment problem.

Two commissioning habits follow. First, log the wet bulb and the leaving water temperature together, so the ratio can be reconstructed later; a leaving water temperature alone cannot be judged. Second, record fan blade angle, fill condition and eliminator condition at commissioning, because those are the four things that silently drift.

Work the whole review for a 5 MW hall

Bring the pieces together for a 5 MW liquid-cooled hall in a hot–humid inland climate, with a warm-water loop and a waterside economizer.

A complete wet-bulb-driven review for a 5 MW hall. Every number traces back to the 0.4% design wet bulb in step 2.

Step 4 is where the real decision sits. A required approach of 3.3°C (5.9°F) is achievable but tight, and it is close to a commercial boundary: CTI's guidance is that it is not customary in the cooling tower industry to guarantee any approach below about 5°F (2.8°C) [12]. Selection software typically works from an ASHRAE design wet bulb entered by the engineer, with typical US values in the 72–80°F range, a typical approach of 5–10°F and a typical 10°F range at 3 gpm/ton — and induced-draft counterflow as the default configuration for large data center duty [19]. There are two honest ways to buy margin. Raise the facility water supply temperature, which lowers the required CWT and relaxes the approach — every degree of supply temperature is a degree of relief. Or accept a larger tower, since approach tightens non-linearly with size.

Bought physics is expensive physics. Industry guidance puts the move from a 6–7°F approach to 4–5°F at roughly 25–40% added tower cost, which is justified for data centers because it converts directly into free-cooling hours rather than comfort [5].

Summary

Wet bulb is the only boundary condition a cooling tower cannot negotiate with. Dry bulb describes the day; wet bulb describes the tower's ceiling. Every selection number on the heat rejection sheet traces back to it, and every operational promise — the design heat rejection capacity, the free-cooling hour count, the annual water withdrawal — is a function of it.

Four rules carry most of the weight. Select against the 0.4% design wet bulb, not the 1% number, and treat the tower as the cheapest capacity on the sheet. Put approach, not capacity, in the specification, and understand that its cost is non-linear. Expect approach to degrade as the wet bulb falls, and stage or pipe the towers accordingly instead of comparing every measurement back to a summer design point. Then buy free-cooling hours the cheap way, by raising the facility water supply temperature, before buying them the expensive way, by tightening the tower approach.

If you are specifying heat rejection for a liquid-cooled hall and need tower selection data or an installed reference at your design wet bulb, request a quote and our engineers will work the numbers against your site's ASHRAE design conditions.

Frequently asked questions

Why does a cooling tower follow wet bulb instead of dry bulb?

Because it rejects heat by evaporation. The driving force is the enthalpy difference between the air and the saturated air film at the water surface, and lines of constant enthalpy on a psychrometric chart run nearly parallel to lines of constant wet-bulb temperature. Two days with the same wet bulb but different dry bulbs produce the same leaving water temperature from the same tower at the same flow and load [1][5].

What wet-bulb temperature should a data center use for tower selection?

The 0.4% (99.6%) ASHRAE design wet bulb — the value exceeded roughly 35 hours per year. LBNL's warm-water analysis uses 26.5°C (79.7°F) as the 0.4% wet bulb across the national laboratory sites when a cooling tower is the primary cooling source, and 37.5°C (99.5°F) dry bulb when dry coolers are [7]. Commercial work commonly uses 1%, which is about 88 hours per year and typically one to two degrees warmer — a shortcut that costs capacity on the worst afternoon of the year [8].

Can a cooling tower make water colder than the air temperature?

Yes, and it routinely does. The tower can approach but never reach the wet bulb, which in a dry climate can be 20–40°F below the dry bulb. In a saturated climate the two converge and the tower loses that advantage entirely — which is why the same tower specification performs differently in Phoenix and Miami [1][2].

Why did approach get worse in winter when the water got colder?

Because cold air cannot hold as much moisture, so the evaporative driving force collapses even as the absolute water temperature falls. For a tower selected at a 78°F wet bulb and 7°F approach, the approach at a 45°F wet bulb is about 16°F, and the deterioration continues below that — a factor of roughly six by the time the wet bulb reaches 30°F [9][11]. The water is still colder in absolute terms; the tower is simply further from its ceiling.

When does dry bulb become the right reference instead of wet bulb?

When the heat rejection equipment is dry. A dry cooler approaches the ambient dry bulb, an adiabatic or hybrid cooler sits between the two, and an open or closed-circuit tower approaches the wet bulb. A site evaluating waterless cooling has to re-run its design day against the 0.4% dry bulb, which is a materially harder number than the 0.4% wet bulb in every US climate [7][12].

How much water does a wet-bulb-based plant actually consume?

About 1.8 gallons evaporate per ton-hour of refrigeration, which is irreducible for evaporative cooling. Total withdrawal including blowdown at 5 cycles of concentration is about 2.25 gal/ton-hour, and typical data center estimates land between 1.5 and 2.5 gal/ton-hour. For a 5 MW hall at 1.20 PUE and a 70% annual load factor, that computes to roughly 18 million gallons per year, or a water usage effectiveness near 1.7 L/kWh [14][15].

Does a tighter approach always pay back?

No. The tower cost curve is non-linear — moving from a 7°F approach to 5°F is roughly a 1.55× tower, and to 4°F roughly 1.9×. It pays where the plant runs long hours at high load and where the tighter approach converts into free-cooling hours or lower chiller condensing temperature. For a data center running 8,000+ hours a year, it usually does. For a low-utilization plant, it usually does not [5].

1. CED Engineering, “Cooling Towers” — wet bulb as the hard floor on leaving water temperature (a tower cannot cool below the entering air wet bulb), the enthalpy-driven heat and mass transfer basis, CTI's definition of ambient wet bulb measured 50–100 ft upwind at 5 ft above the tower base, the 1–2°F entering-wet-bulb adder for recirculation, and the note that approach below 5°F is neither economical nor guaranteed. cedengineering.com
2. CalcEngineer, “Cooling Tower Calculator” — the fixed performance model: range = hot − cold, approach = cold − wet bulb, efficiency = range/(range + approach), the physically inconsistent condition when approach is negative, and the industry standard rating point of 95°F hot water / 85°F cold water / 78°F wet bulb equal to a 10°F range and 7°F approach. calcengineer.com
3. University of Wisconsin IRC, “Design weather conditions” — ASHRAE design weather table with 0.4%, 1% and 2% dry-bulb/coincident-mean-wet-bulb and design-wet-bulb/coincident-mean-dry-bulb values for Davenport, Madison, Miami, Phoenix, Portland, Tampa and Sioux City, plus the 10% condenser capacity shortfall that follows from selecting against coincident wet bulb instead of design wet bulb. irc.wisc.edu
4. CalcEngineer, “Makeup Air Unit Sizing per IMC Section 508 + ASHRAE 90.1” — ASHRAE 169-2021 climate-zone design table giving summer 1% dry bulb and 1% wet bulb per representative city (Miami 91/79°F, Houston 96/78°F, Phoenix 110/70°F, Chicago 88/75°F, Minneapolis 88/73°F, Duluth 84/70°F). calcengineer.com
5. CalcEngineer, “How to Calculate Cooling Tower Approach Temperature” — the 6–7°F comfort-cooling approach sweet spot versus 4–5°F for data center waterside economizers, the 25–40% tower cost increase for 3–4°F approaches, why VFD fan staging legitimately raises approach at part load, the rule that measured approach is only comparable to design above 80% of design heat rejection, and the non-fouling causes of degraded approach (drift eliminator damage, fan blade pitch slip, casing or louver air bypass, cracked distribution decks). calcengineer.com
6. IngeNER, “Cooling Tower Performance for HVAC Engineers” — range, approach and effectiveness definitions with typical values (effectiveness 0.65–0.75, approach 7–10°F, range 10°F), heat rejection Q = GPM × 500 × ΔT, the evaporation expression, cycles of concentration 3–6, and the standard 95°F entering / 85°F leaving / 78°F wet bulb rating point. ingener.by
7. Lawrence Berkeley National Laboratory, “‘Hot’ for Warm Water Cooling” — the LCWG decision to use ASHRAE 99.6% (0.4%) design temperatures, the finding that across all sites the 99.6% wet bulb is at or below 26.5°C (79.7°F) while the 99.6% dry bulb is at or below 37.5°C (99.5°F), the resulting convention of 26.5°C wet bulb for cooling-tower-based designs and 37.5°C dry bulb for dry-cooler designs, and the note that the 35 exceedance hours concentrate in hot daytime periods. datacenters.lbl.gov
8. JASCKO Corp. technical note, “ASHRAE 0.4% vs 1% Design Wet Bulb in South Florida Cooling Towers” — the three ASHRAE design frequencies expressed as hours per year exceeded (0.4% ≈ 35 h, 1% ≈ 88 h), Miami's 0.4% design wet bulb near 79–80°F with the 1% value about a degree lower, and the failure chain when a tower is sized to 1%: design leaving water temperature missed on the 0.4% day, rising condenser water temperature, climbing chiller head pressure, higher kW/ton, and high-pressure trip in extreme cases. linkedin.com
9. Trane (M. Schwedler, P.E.), “Effect of Heat Rejection Load and Wet Bulb on Cooling Tower Performance” — the ASHRAE Handbook figure showing approach varying with wet bulb and load, the finding that between 30°F and 85°F wet bulb the approach changes by a factor of six, that air holds less moisture at low temperature so a greater proportion of cold-weather heat rejection is sensible and less water evaporates, and the reference tower design at 78°F design wet bulb with a 4.5°F approach and 14°F range. trane.com
10. R.L. Deppmann, “Water-Side Economizers Part 2: State Energy Codes” (citing ASHRAE 90.1-2013 Sections 6.5.1.2.1 and 6.5.1.3) — the requirement that a water economizer provide up to 100% of expected system cooling load at 50°F dry bulb / 45°F wet bulb and below, the integrated-economizer mandate for simultaneous chiller and free-cooling operation, the 15 ft water-side pressure-drop limit or secondary-loop alternative for pre-cooling coils and water-to-water exchangers, and the computer-room exceptions. deppmann.com
11. JMP Equipment Company, “How To Size A Waterside Economizer Part 1” — Energy Star's guidance that waterside economizers suit climates where the wet bulb is below 55°F for 3,000 or more hours, the ASHRAE 90.1-2010/2013 general application guidelines, why chillers and economizers can no longer be piped in parallel, and the resulting need for two different cooling tower cold water supply temperatures because chillers have a minimum condensing temperature. jmpcoblog.com
12. JMP Equipment Company, “How To Size A Waterside Economizer Part 2: A Cooling Tower In Winter” — the worked illustration that a crossflow tower selected at 79°F wet bulb for 95°F to 85°F (10°F range, 6°F approach) would need a 31.85°F wet bulb to hold the same 10°F range on 60°F entering water, giving an 18.18°F approach, and the resulting practice of sizing for peak summer and operating at part load in cold weather with towers piped in series with the economizer. jmpcoblog.com
13. Harris Thermal, “AI data center liquid cooling: where heat exchangers fit in the cooling infrastructure” — the heat rejection comparison table (open tower approaches wet bulb, closed-circuit fluid cooler wet bulb in wet mode, adiabatic/hybrid between dry and wet bulb, dry cooler dry bulb, water-cooled chiller mechanical), Uptime Institute's 32°C (89.6°F) facility water convergence and the 50% to 80%-or-more free-cooling range, SPX Cooling Tower Fundamentals' definition of approach with the industry convention of not guaranteeing any approach below 5°F, the 2025 CTI paper reporting up to 90% water reduction for adiabatic coolers with scale on pre-cooling media as the design-against failure mode, and ASHRAE TC 9.9 FWS/TCS water quality separation. harristhermal.com
14. NVIDIA DSX reference documentation, as cited in Harris Thermal — the 45°C technology cooling supply design point at the rack and the statement that it “can widen the opportunity for dry cooling and reduce cooling overhead when site and ambient conditions permit,” with the caveat that a hot–dry site such as Arizona may still need trim support for much of the year and mechanical cooling during the hottest conditions. harristhermal.com
15. U.S. Department of Energy, Better Buildings Solution Center, “Water System Info Card” — the cooling tower make-up balance M = evaporation + blowdown + drift, blowdown = evaporation/(cycles of concentration − 1), cycles of concentration = make-up/blowdown, the 1% evaporation per 10°F range rule of thumb, and the annual water usage table for chillers from 100 to 1,000 tons nameplate across 3 to 8 cycles of concentration at 8,760 operating hours. betterbuildingssolutioncenter.energy.gov
16. ICS Cooling Towers, “Cooling Tower Makeup Water Calculation Guide” — M = E + B + D with E = R × ΔT × 0.00085 for Fahrenheit, B = E/(CoC − 1), modern drift eliminators reducing drift to about 0.0005%, cycles of concentration as the primary water-efficiency lever (moving from 2 to 4 cycles halves blowdown), and the comparative make-up reference table for 1,000 to 5,000 gpm recirculation. h2ocooling.com
17. Chemical Engineering (quoted by Quimicas Mundiales), “Quick Calculation of Cooling Tower Blowdown and Makeup” — the mass balance M = E + B + W, evaporation losses of 0.85–1.25% of circulation per 10°F drop with 1.0% as the standard assumption, windage losses of 0.1–0.3% for forced-draft towers and 0.1% for modern formed-draft towers, the relation M = (B + W)·C for cycles of concentration, and the derived blowdown expression B = ΔT/(10(C − 1)). quimicasmundiales.com
18. Trane, “Cooling Towers and Condenser Water Systems Design and Operation” — the worked tower performance comparisons at standard rating conditions (1,500 gpm at 78°F design wet bulb with a 7°F approach giving 94.3°F hot water and 85°F cold water, and the same tower at 1,000 gpm and 2 gpm/ton reselected to a 4°F approach with 96°F hot water and 82°F cold water), the tabulated approach-versus-wet-bulb curve (16.0°F at 45°F wet bulb down to 6.0°F at 80°F wet bulb), and CTI certification limits with an approach above 5°F in the 60–90°F wet-bulb band. trane.com
19. Anvilfield, “Waterside economizer and free cooling field guide” — the definition of a waterside economizer as making chilled water from the tower alone with the compressor off or unloaded, the wet bulb as the driver of free-cooling hours and the tower as the element approaching it, the ASHRAE 90.1 fluid-economizer threshold and integration requirement, the changeover control point with offset and deadband, and the guidance to trend plate-exchanger approach and free-cooling hours season over season to catch fouling and drifted setpoints. anvilfield.com
20. EngDatabase, “Cooling Tower Selection & Sizing” — the selection workflow using ASHRAE 0.4% or 1% design wet bulb (typical US range 72–80°F, default 78°F), typical approach 5–10°F, typical range 10°F at 3 gpm/ton, induced-draft counterflow as the most common configuration for large HVAC and data center duty, and the water balance calculation covering evaporation, blowdown, drift and make-up with cycles of concentration typically 3–7. engdatabase.com

Related: Cooling data centers in the desert — four architectures and the water they drink (WUE/PUE ledger for 1 MW IT)