When the Heat Rises

Section One

What Is an Outdoor Design Temperature?

Ask a facility manager how their cooling system was sized and the honest answer is usually: someone decided how many tons the building needed. Ask how that decision was made, and the trail leads back to a single climatic number.

Engineers design against statistics, not weather

ASHRAE publishes climatic design data for more than 9,000 weather stations in Chapter 14 of the Handbook—Fundamentals. For each station it gives cooling design dry-bulb temperatures at three probability levels: 0.4%, 1%, and 2%. These are annual cumulative frequencies of occurrence. ASHRAE is explicit:

“The 0.4, 1.0, 2.0, and 5.0% values are exceeded on average 35, 88, 175, and 438 h per year, respectively, for the period of record.”

Two points of precision matter, and most trade literature gets them wrong. These are exceedance values — the temperature is beaten about 35 hours a year at the 0.4% level, not “reached.” And ASHRAE says on average, for the period of record. Any individual summer can exceed it considerably more. The design condition is a long-run statistic, not a promise about this year.

ASHRAE publishes all three percentiles and deliberately declines to say which to use: “The designer, engineer, or other user must decide which set(s) of conditions and probability of occurrence apply to the design situation under consideration.”

Commercial practice commonly sizes to the 0.4% condition. So the working number for a Cleveland commercial building is roughly 90°F. On a 95°F afternoon your building is operating five degrees past the condition it was engineered for. On a 97°F afternoon, seven.

Your cooling system was not designed to defeat the hottest day of the decade. It was designed to hold the line on a temperature that gets exceeded about thirty-five hours a year — and to run efficiently for the other eight thousand seven hundred.

What “correctly sized” actually means

Equipment is selected to meet the design load — with a modest margin, not a generous one. In practice that means installed capacity somewhere near the load itself: enough to hold the line at the design condition, and not so much that the system runs badly for the rest of the year.

We are deliberately not going to hand you a standard-mandated percentage here. The relevant commercial standards are not freely available, we have not read them, and we are not going to cite a standard we have not read. What we can do is show you exactly how much the sizing decision matters — which we do, at length, in Section 3 and the Appendix.

The essential point survives without any appeal to authority: the margin at the design condition is small on purpose. A system with generous headroom at 90°F is a system that short-cycles, fails to dehumidify, and wears itself out for the other 8,725 hours of the year.

Why not simply install a bigger unit?

This is the first question every building owner asks, and it deserves a real answer.

This is not hypothetical. When PG&E instrumented commercial rooftop units in the field, it found that although “25% over sizing is generally considered by the HVAC design and construction professions to be a safe and acceptable practice,” in fact “over 40% of the units are greater than 25% over sized” and about 10% were oversized by more than half. In at least 40% of cases, the unit could have been half the size.

The industry already tried buying bigger. It is a large part of why buildings are uncomfortable — and, as Section 3 will show, it is also why so many buildings that should be coping in a heat wave are not.

And your building is probably not carrying the load you think

There is a widespread belief that modern offices have overwhelmed their 1990s cooling systems with computers and equipment. The measured evidence says the opposite.

NREL metered plug loads across commercial buildings and found that designers in the late 1980s and 1990s sized systems for plug load densities of 3 to 5 watts per square foot, based on nameplate ratings. Actual measured office plug loads today average about 0.28 W/ft², with peaks around 0.50 — lower by a factor of five to ten than what design and lease documents assume. Laptops replaced desktops. LCDs replaced CRTs. Efficiency standards did their work.

Your 1995 building was very likely sized for an internal load it never actually carried. That is another reason legacy commercial buildings tend to be oversized rather than starved — and another reason adding tonnage is rarely the answer to a comfort complaint.

Section Two

How Commercial Cooling Load Is Built

“Cooling load” is the total rate at which heat must be removed from a building to hold a setpoint. It arrives through several independent pathways — and only some of them care what the outdoor temperature is.

The roof is the villain, and it is not close

Most people assume the roof matters because hot air sits on it. The truth is more aggressive.

A sunlit surface does not exchange heat with the air at the air’s temperature. It exchanges heat at what ASHRAE calls the sol-air temperature — the temperature the outdoor air would have to be, with no sun at all, to drive the same heat into the surface.

ASHRAE publishes sol-air values for July at 40°N latitude. Cleveland sits at 41.4°N — this table is not an analogy, it applies directly. At solar noon, with the outdoor air at 90°F and the building held at 75°F:

A dark commercial roof under a July sun behaves as though the outdoor air were 162°F. The driving temperature difference across it is nearly six times what the thermometer would suggest.

And in a single-story building — the warehouse, the strip center, the suburban office, the school wing — the roof area equals the floor area. It is the largest surface the building has, it is aimed at the sun, and it is running at that multiplier.

On a 90°F afternoon, the thermometer reads 90. Your dark roof is behaving as though it were 162.

One honest caveat: that 5.8× figure is a ratio, and ratios shrink as their denominator grows. At 97°F outdoor the same dark roof runs at about 4.3× the air-to-room difference — still enormous, but the multiplier itself does not grow with the heat. The absolute roof load does.

Outdoor air: the load you are not allowed to turn off

Every commercial building is required by code to bring in outdoor air. ASHRAE Standard 62.1 sets the minimum — for a typical office, 5 cfm per person plus 0.06 cfm per square foot of floor area.

You cannot opt out. It is a code obligation, and it means that on the hottest afternoon of the year your building is required to draw in thousands of cubic feet per minute of the hottest air available and cool it down.

There is a timing point worth knowing, too. Outdoor-air load hits the cooling coil the moment the thermometer moves. Roof heat arrives late — the roof soaks up sun at noon and hands it to the building around four o’clock. When the temperature spikes, you feel the outdoor-air load immediately, and the roof load from that same spike is still on its way.

The insight that carries the rest of the paper

In the conceptual 50,000 ft² Cleveland office specified in the Appendix, the temperature-driven components add up to a slope of about 11,500 Btu/h for every degree of outdoor temperature.

But the temperature-independent components — solar gain, lighting, people, plug load, fan heat, and dehumidification — total roughly 975,000 Btu/h and do not move at all when the outdoor temperature changes.

At the 89.7°F design condition, that building’s total load is about 1,145,000 Btu/h (roughly 95 tons) — and about 85% of it is completely indifferent to the outdoor temperature.

Cooling load rises steeply in absolute terms — but modestly in percentage terms. Because such a large share of the load is fixed, the total load curve has a high intercept, not a shallow slope. Total load climbs only about 1% per degree. On its own, that is manageable. It is not on its own — and Section 3 is about the other half.

Section Three

The Scissors: Where Load and Capacity Cross

It is not a switch. It is a margin.

The common mental model of air conditioning is binary: the AC works, or the AC is broken. Nearly every frustrated call to a service contractor during a heat wave begins from that assumption.

The engineering reality is a margin that narrows — continuously, predictably, and by an amount you can calculate.

Correct sizing does not mean comfortable headroom. It means enough capacity to hold the line at the design condition — and not so much that the system runs badly for the rest of the year.

The scissors

Two things happen simultaneously above the design temperature, and the reason buildings fall behind is that they compound.

The load rises — about 1.0% of the design load per degree, in our conceptual example.

The capacity falls — about 0.7% of the design load per degree, applying the derate slope that published manufacturer tables actually show. (Section 4 explains the physics.)

Together, the deficit opens at roughly 1.7% of the design load for every degree — about 60% of it from load rising, 40% from capacity falling. Both blades of the scissors are real. Neither alone explains what building owners experience.

Nothing broke between the 89.7°F row and the 97°F row. No filter clogged. No compressor failed. Seven degrees happened.

The crossover is a range, not a number — and this matters more than the number does

It would be easy to read “about 93°F” off that table and treat it as a fact about buildings. It is not. It is a fact about this building, and the largest single thing it depends on is how the equipment was sized — a choice a designer made years ago, not a property of the weather.

Which means the honest answer to “at what temperature will my building fall behind?” is: it depends on your building, and specifically on a decision someone made when it was built. Anyone who hands you a single universal number — including anyone quoting this paper — is overselling.

And now the part that matters most: the oversized building

Section 1 established that more than 40% of commercial rooftop units are oversized by more than 25%. That is not the exception. In the field, it is the most common building there is.

So run the model on one.

An oversized system does not fall behind at 97°F. It has capacity to spare. So if an oversized building cannot cool on a 97°F afternoon — and most commercial buildings are oversized — the weather is not the explanation. Something is broken.

This is the most important paragraph in the paper, and it is the one a contractor is least likely to tell you.

Everything else here explains why a correctly sized system runs out of margin in extreme heat. That explanation is true, well-supported, and should change how you interpret a hot afternoon. But it applies to a building that was sized correctly — and the field data says most were not.

If your building was generously sized (and the odds favor it), the physics in this paper buys you no excuse whatsoever. Section 6 is how you find out what is actually wrong.

Why a small deficit feels like a catastrophe

When capacity falls short of load, a building does not simply “stay a bit warm.” The indoor temperature rises until the load drops enough to match the capacity available. The building finds a new equilibrium — at a higher temperature.

But recall from Section 2: only about 15% of this building’s cooling load is temperature-sensitive at all. The solar gain, the lights, the people, the computers, the fan heat, the dehumidification — none of it gets smaller because the room got warmer.

So to shed a deficit, the temperature-sensitive part of the load must fall by enough to cover the whole gap. And since that part is small, the indoor temperature has to move a lot.

A 7% capacity shortfall does not make a building 7% less comfortable. The physics is not gentle.

That is why heat waves feel less like a gradual decline and more like a building giving up. Because most of a commercial cooling load does not care how warm the room gets, a modest capacity shortfall forces a large temperature drift.

Section Four

The Second Curve: Why Capacity Falls as It Gets Hotter

Chart 1 has two lines. Section 3 was about the one going up — the one everyone expects. This section is about the one coming down, which almost nobody does.

A cooling system does not make cold

An air conditioner is not a cold-generator. It is a heat pump — a machine that picks heat up inside your building and puts it down outside.

Which raises an obvious question: what happens when outside is hot?

Rejecting heat into 85°F air is straightforward. Rejecting that same heat into 105°F air is much harder — you are trying to push heat into a place that is already hot. The machine still works. It just cannot move as much.

That is the entire concept. What follows is the mechanism, for readers who want it.

What the manufacturers actually publish

These are not estimates. These are performance tables for the equipment on Northeast Ohio roofs.

Across three published models from two manufacturers, commercial rooftop capacity falls roughly 0.5% to 0.8% per degree above the 95°F rating point — and the derate steepens as it gets hotter. Our model uses the mean of the three, about 0.63%/°F, and the Appendix shows what happens at either end of the range.

The nameplate is a rating, not a promise

The cooling capacity stamped on your rooftop unit is certified under AHRI Standard 340/360. That standard rates capacity at a specific, defined condition: 95°F outdoor air, with 80°F dry-bulb / 67°F wet-bulb entering the evaporator.

Above 95°F, the standard specifies something different. It defines a “Maximum Operating Condition” of 115°F — a requirement that the unit keep running. Not a statement of how much it will deliver.

The part your CFO should read

At 95°F the unit produces 193,100 Btu/h for 11.78 kW. At 115°F it produces 161,900 Btu/h for 15.16 kW.

Less cooling. More power. Every hour.

Efficiency falls by about a third precisely when the building needs cooling most — and precisely when the utility is setting your demand charge, which is billed on your peak. The hottest afternoon of the year is simultaneously your worst performance and your most expensive hour.

Section Five

Why Systems Run Continuously During Heat Waves

Cycling is what surplus looks like

A cooling system cycles off when it has caught up — when it has removed heat faster than the building gained it, driven the space to setpoint, and has capacity left over. Cycling is the visible evidence of surplus capacity.

At the design condition, a correctly sized system has very little surplus. So it barely cycles. It runs. And past the design condition, when load exceeds capacity, it cannot catch up — so it cannot shut off. A system running continuously in a heat wave is not failing to shut off. It has nothing left to shut off with.

The field data

Pacific Gas & Electric instrumented commercial rooftop units with one-minute interval monitoring across more than 18,000 hours of operation, during a period that in many cases exceeded ASHRAE 1% design conditions:

“About 20% of the monitored units ran continuously under peak load conditions, as one would expect for a properly sized unit.”

And the mirror image:

“Frequent cycling, especially under peak load conditions, is a reliable indicator of roof top unit over sizing.”

Nearly 60% of the monitored units were cycling three or more times per hour at peak conditions — typically six minutes on, sixteen minutes off.

Scope, honestly stated: this is 1998 data from Northern California, drawn from 145 usable datasets of 250 monitored units, most of them around 5 tons. The mechanism generalizes. The specific percentages are theirs, not ours.

Section Six

When Continuous Operation Does Indicate a Real Problem

Everything to this point has explained why a healthy system struggles in extreme heat. This section exists so that explanation is never used as an excuse — including by us.

The second question, and it is nearly as important

Was your system generously sized?

If you have more installed tonnage than a current load calculation would call for — and the field data says more than 40% of buildings do — then the physics in this paper does not apply to you at 97°F. An oversized system has surplus at 97°F. It should be coping. If it is not, something is wrong.

Signals consistent with normal full-load operation

• The unit runs continuously only during genuine extreme heat
• Supply air is still measurably cold, for the current indoor humidity
• The building holds within a few degrees of setpoint and recovers overnight
• Performance returns to normal when outdoor temperatures do
• Other buildings nearby are experiencing the same thing at the same time

Signals of genuine mechanical failure

Warm or lukewarm supply air. The clearest signal. If the temperature drop across the cooling coil has collapsed, the system is not removing heat.

But have it measured correctly. The “18–22°F temperature split” that every technician quotes is not a universal. The correct target depends on indoor humidity, and it collapses as humidity rises — at 75°F return air with a muggy 71°F wet bulb, the correct target split is about 11.7°F, not 20°F.

If a technician tells you the split “looks fine” — or “looks bad” — ask what the return wet bulb was. A tech who measured only dry bulb has told you very little. And on a rooftop unit with an economizer, what they measured may have been mixed air rather than return air, which shifts the target again. This is a screening tool, not a substitute for direct airflow measurement.

A frozen evaporator coil. Ice on a cooling coil in July is always a fault — never “the system working too hard.” It points to low airflow or low refrigerant charge. Note the trap: those underlying faults are most likely to surface precisely when the unit runs long and hard. A coil that freezes during a heat wave is a genuine failure that the heat wave exposed, not caused.

Short cycling during a heat wave. Read Section 5 again. At peak load, a healthy right-sized unit should be running. If yours is switching off and on, it is either badly oversized or tripping on a fault.

Low refrigerant charge. A collapsed temperature split across the evaporator “indicates that the compressor is not extracting load from the air stream, which indicates a high possibility of a low refrigerant charge.” And an important framing point: refrigerant is not consumed. A system low on charge has almost always sprung a leak — and a leak is a failure, not a maintenance item.

A dirty condenser coil. This interacts directly with everything in Section 4. A fouled condenser cannot reject heat efficiently, which drives condensing temperature and pressure up — the same mechanism that high outdoor temperature imposes. A dirty coil is therefore a self-inflicted heat wave, stacked on top of the real one. In high ambient, the combination is what trips the high-pressure switch.

This is often the honest answer to “why did my unit die on the hottest day of the year?” The coil was already dirty. The 97°F afternoon simply revealed it.

A stuck economizer. This one hides for years and almost nobody checks it. Economizers fail at a remarkable rate — across eleven field studies, failure rates average about 60%. Most commonly they fail closed, quietly costing you free cooling every spring and fall.

But when one fails open, it becomes actively destructive in hot weather. In a simulated building in Bakersfield, California, an economizer stuck fully open added 84% to the summer peak load. The magnitude would be smaller in Cleveland’s milder climate — but the mechanism is identical, and the unit will appear hopelessly undersized on hot days, because it is being force-fed outdoor air while trying to cool the building.

Section Seven

What Building Owners Should Do

Know your two numbers

Most building owners cannot state their building’s design temperature or its installed cooling capacity. These determine what your building can and cannot do in July, and both are knowable.

Set expectations before the heat wave, not during it

The most damaging thing about a heat wave in a commercial building is often not the temperature — it is the surprise. Tenants and staff who have been told in advance that the building may run a few degrees warm during genuine extreme heat respond very differently from those who conclude, at 3 p.m. on a Thursday, that the building is broken.

A one-paragraph email in May is worth more than a service call in July.

Protect the capacity you already own

Everything in Section 4 says capacity is scarce on hot days. Every item below is capacity you have already paid for and are at risk of throwing away:

• Condenser coils. A dirty condenser imposes the same physics as a heat wave. Cleaning it is the most direct way to protect hot-weather capacity.
• Filters and airflow. DOE: 5–15% of energy consumption, and low airflow is a leading cause of frozen coils.
• Refrigerant charge. Low charge almost always means a leak. Find it.
• Economizers. Field failure rates average around 60%. If nobody has checked yours in the last few years, the statistics say it is more likely broken than not.

Understand where load creep is real

Section 1 showed that the “modern offices overwhelmed old systems” story is a myth — measured plug loads today are far below what 1990s designers assumed.

Where load creep is real is at the zone level. The storage closet that became a server room. The open floor that got densified. The tenant who installed a commercial kitchen. NREL’s data shows offices with data centers running three to four times the plug load of a plain office.

That is a zoning and air-distribution problem, not a tonnage problem. The instinct to solve it by adding tonnage will make the humidity and cycling penalties in Section 1 worse, not better. The right fix is usually to get the cooling to the right place, not to buy more of it.

Reduce the load before you buy more capacity

The cheapest ton of cooling is the one you never have to produce. The highest-leverage moves are usually on the load side:

• Roof reflectivity. Section 2’s table is the argument: a light-colored roof runs at roughly a 3.1× effective temperature multiple instead of a dark roof’s 5.8×. On a single-story building this is the largest surface you own. (In our conceptual example, switching from dark to light cuts the design load by about 9 tons — and it is why we show that sensitivity in the Appendix rather than quietly leaving it out.)
• Glazing film or shading. Solar gain through glass is a fixed load that does not care about the outdoor temperature — which means you carry it at every condition, not just the hot ones.
• Controls and scheduling. Pre-cooling the building’s thermal mass ahead of a forecast heat event shifts load out of the worst hours.
• A planned setpoint strategy. A communicated 2°F relaxation during a genuine extreme-heat event is vastly better than an unplanned 8°F drift.

Conclusion

What This All Comes Down To

Outdoor design temperature is not an excuse. It is an engineering boundary condition — and a knowable one.

Commercial HVAC systems in Northeast Ohio are sized against a statistical condition: roughly 90°F in Cleveland, a temperature exceeded on average about 35 hours a year. Below that line, a correctly sized system has margin. At that line it has little — because having little there is what correct sizing means.

Above that line, two curves move against each other. Load rises with every degree. Capacity falls with every degree, because a cooling system must reject heat into outdoor air that is itself getting hotter. In our conceptual example the deficit opens at roughly 1.7% of the design load per degree.

Seven degrees past the design point, a correctly sized building can be short several percent of the cooling it needs — with nothing broken and nothing to repair.

Continuous operation during extreme heat is therefore, very often, the sign of a system working. When commercial rooftop units were monitored at peak conditions, about 20% ran continuously — which, in the words of the engineers who ran the study, is “as one would expect for a properly sized unit.”

But honesty cuts in two directions

More than 40% of commercial rooftop units are oversized by more than 25%. Those systems do not run out of margin at 97°F. They have capacity to spare. If a generously sized building cannot cool on a hot afternoon, the physics in this paper explains nothing about it — and a contractor who offers that explanation is not being straight with you.

Likewise, a building that cannot cool on an ordinary 82°F afternoon is not a victim of the weather. It has a problem, and that problem has a name: a fouled condenser, a leaking circuit, a failed economizer, a collapsed airflow path.

The right question is not “why isn’t my air conditioning keeping up?”

It is: is my building’s cooling capacity actually matched to the load it carries today — and do I know what happens to both when the temperature climbs past ninety?

That is an answerable question. And it is worth answering before the next heat wave, not during it.