Cooling towers reject the largest share of a data center’s heat, and they do it by design — through evaporation. When that rejection degrades, the loss rarely announces itself. Approach temperatures creep, fans work harder, and the first clean signal often shows up where nobody is looking: on the roof of the plant itself.
An aerial thermal survey is one instrument for locating cooling-tower problems while they are still patterns rather than failures. Here is what it can show, what it cannot, and how a finding becomes a verified work order.
How a Cooling Tower Rejects Heat — and Where It Loses It
Almost all of a tower’s work is evaporation. The U.S. Department of Energy’s Federal Energy Management Program (FEMP) describes the four ways water leaves a cooling tower system: evaporation — the heat-rejection process itself; drift — the small share carried out as mist, limited by baffles and drift eliminators; blowdown — the deliberate removal of concentrated water, which flushes out the dissolved solids (calcium, magnesium, chloride, silica) that evaporation leaves behind; and basin leaks or overflows — which, as FEMP puts it, properly operated towers should not have.
The balance is simple: Make-Up = Evaporation + Blowdown + Drift. A tower consuming unusual make-up water is telling you water is leaving through a path that is not evaporation — or that control equipment is letting it go.
Heat loss in this context has two faces. One is lost capacity: fill that no longer contacts water and air efficiently, distribution that no longer spreads water evenly. The other is lost water and chemistry: leaks, overflows, excess drift. An aerial thermal survey is built to find the first, and to flag the second.
What Heat Loss Looks Like From Above
A cooling tower is, thermally, a machine designed to stand out. A radiometric thermal sensor measures radiated surface temperature across every pixel in the frame, and from an aerial vantage the patterns are legible in ways a ground walk rarely is:
- Plume behavior. Which cells carry visible vapor, how evenly, and where flow appears weak or intermittent.
- Fan-deck discharge. Working cells discharge warm air through the fans; uneven or missing discharge between neighboring cells can point to staging, dampers, or airflow problems.
- Wet and dry deck zones. Equal water distribution produces uniform evaporative cooling across the fill. Zones that read differently from their neighbors — hotter or cooler, under the same conditions — are the first suspicion.
- Basin and casing surfaces. Wet streaks, overflow tracks, and standing water near the tower’s base.
One discipline up front: a thermal image shows surface radiance, not efficiency. A hot patch is a pattern to verify, not a measured ton of lost capacity. Keeping that distinction in the report is what makes the finding usable.
The Signature Catalog
Across cooling plants, the same failure modes keep producing visible thermal patterns:
- Plugged spray nozzles and uneven distribution. Fill that should be uniformly wetted runs partially dry — losing evaporative cooling — or floods in the areas of least resistance. Both show up against neighboring cells.
- Scale-loaded fill. Mineral deposits reduce the contact area between water and air; the affected cell rejects less heat and shows weaker cooling against its neighbors under the same load.
- Air bypass. Air that slips around the fill — through damaged sections, missing splash bars, or open gaps — fails to pick up heat. The cell runs hot relative to its peers.
- Fan and gearbox anomalies. Bearing or motor heat is visible on the fan deck; a gearbox running warmer than its neighbors under similar duty is a maintenance flag.
- Basin leaks and overflows. Saturated roof areas near the tower after dry weather, overflow paths, and wet-insulation signatures around the basin.
- Risers and valves. Hot spots on distribution piping, flanges, and valves — often the first evidence that insulation has been damaged or displaced.
None of these is self-diagnosing. Each one generates a location and a hypothesis to verify — which is the entire purpose of a detection layer.
What Airborne Radiometry Can’t Resolve
Honest limits, stated before any sales conversation:
- Water vapor attenuates infrared. The plume itself is a shield; the hottest, wettest zone can be the least readable.
- Wet surfaces complicate emissivity. A water film changes how a surface emits and reflects; wet areas can read cooler or noisier than they are.
- Casing and structure block the view. Thermal imaging reads external surfaces; it does not see inside a closed tower or through steel.
- Angle and distance matter. From altitude, small features can blur into patches; resolution sets a floor on what can be called a discrete finding.
- Load is unknown at scan time. Without the heat load and wet-bulb conditions, a temperature difference is a pattern — not a performance number.
That last point is why a temperature anomaly should never be published as a capacity number. A survey’s value sits in locating suspects reliably and handing off cleanly — not in inventing precision it does not have.
Timing, Weather, and False Positives
Thermal contrast is driven by change, not by absolute temperature. Survey windows are chosen for conditions that separate wet from dry and working from struggling: low sun and post-sunset cooling, when stored heat releases at different rates across a plant; low wind, because airflow equalizes surfaces and hides patterns; and dry weather, because after rain everything is evaporatively cool, and wet-from-rain mimics wet-from-leak.
Three false positives are routine: solar loading that heats one face of a tower beyond its neighbors; residual heat after shutdown; and high load that makes a healthy tower look alarming. A pattern means more when the same asset is captured in the same window and compared against itself — which is why recurring surveys, not one-off flights, are how towers are actually tracked from the air.
From Finding to Verified Work Order
Aerial thermal is the detection layer. Confirmation belongs to the plant: visual inspection of fill and nozzles, water-flow measurement, conductivity and blowdown logs, approach-temperature trends from the BMS, and invasive checks where the physics warrant it.
Handoff language matters. “Suspect uneven water distribution, cell 3 north quadrant — verify spray nozzles” is a work order candidate. “Hot spot” is a riddle.
The strongest signal a tower program produces is not any single image — it is change over time: the same assets, the same windows, compared against their own baseline. Drift is the finding; transients are noise.
For the wider plant picture — towers, liquid cooling, and where each risk concentrates — the field note on data center cooling and liquid cooling is the map. The data centers vertical covers how aerial thermal programs run over heat-rejection plants, the intelligence scan covers the assessment itself, and a sample report shows how findings are written.