I started this morning the way a lot of Dubai did, reading about the fog. Gulf News had a piece up early — Is fog back in UAE? Why September humidity is triggering foggy mornings in Dubai, Abu Dhabi — with the National Centre of Meteorology's alert running from around 1 am to 8:30 am, the usual advice to slow down and keep your distance, and a short paragraph explaining why it happens. Humidity near 90% at night. A desert surface that cools fast once the sun is gone. Moist air rolling in off the Gulf, touching that cold ground and turning to droplets.
I read that paragraph and stopped. Then I read it again, because I have spent years explaining exactly that mechanism to people, just not about roads. About ceilings. The article describes fog forming on the desert at 4 am. My team spends every September dealing with the same thing forming on pipework at 4 pm, above a false ceiling, in a villa where nobody will notice until the smell arrives.
The parallel is so close that I have decided to borrow the article's explanation and walk it indoors. If you understand why it was foggy this morning, you already understand most of what goes wrong in air-conditioned buildings this month. There is one idea to add, and it is not complicated.
Think of air as a sponge. Warm air is a big sponge and holds a lot of water. Cool the same air and the sponge shrinks; it can hold less. Keep cooling and at some point it is full. One degree more and water has to leave. The temperature at which the sponge is full has a name, the dew point, and it is the only number in this article you really need.
You have seen dew point at work a thousand times without calling it that. Take a cold bottle out of the fridge in September and inside a minute it is wet. The same thing happens on the outside of a glass, on the bathroom mirror after a shower, and on the outside of your windows on a September morning, when the glass, cooled all night by the air conditioning behind it is streaming with water before you have opened the curtains. And it happened on the desert floor at 4 am this morning, which is what the NCM was warning drivers about.
Two things follow from this, and they explain everything else I am going to say. The first is that the number that matters is not humidity. Humidity tells you how full the sponge is right now. Dew point tells you the temperature at which trouble starts, and that is a number you can hold up against a surface. If a surface is colder than the dew point of the air touching it, that surface gets wet. Not probably. Always. The second is that cooling the air does not change its dew point. The amount of water in the air is what sets it. Cooling just brings the air closer to the point where it has to let go.
Not probably. Always.
Fog needs a cold surface, and outdoors that surface is the open ground, which is only cold enough on a minority of nights. Across 42 years of readings at Dubai International, only about 6% of September night-time observations show the air within 2 °C of its dew point. That is why a foggy morning makes the news: it is uncommon.
Inside a building the cold surface is not a weather event. It is designed in, and it never warms up. The cooling coils, the chilled pipework, the diffusers and the ceiling void around them typically sit somewhere between the low teens and the low twenties in °C, every hour the system runs. Meanwhile the air outside is carrying a great deal of water. At Dubai International the typical September dew point is 23.6 °C. For 45% of the month's hours it is at 24 °C or above, for 27% it is at 26 °C or above, and one hour in twenty it passes 28 °C. In October that 24 °C figure drops to 18% of hours. By November it is almost nothing.
Put the two together. Air with a dew point of 26 °C finds a surface at 18 °C, which is 8 °C colder than the point at which that air must give up its water. It is the cold bottle, permanently, in a space nobody looks at. This is the reason I treat September as a season of its own. The air is carrying as much water as it did in July, the September typical dew point of 23.6 °C is level with July's 23.9 °C — but the nights are starting to cool, people are coming back from summer, and buildings that ran on holiday settings for two months are being switched on and off again. That is the combination that produces the calls.
Almost never in the middle of a room. Water appears wherever humid air can reach a cold surface you cannot see: in the ceiling void above a diffuser, around chilled pipework where the insulation has a gap or the vapour barrier was cut through during some earlier repair, behind wardrobes and skirtings on external walls, and inside the AC unit itself, on drain pans and insulation faces where dust has settled. Dust matters more than people expect. It soaks up water at a lower humidity than the clean material underneath it would, so a dusty coil starts getting wet before a clean one.
What decides whether anything grows is sustained wetting, not a momentary one. Field-validated work puts the critical relative humidity for growth on wood-based boards at roughly 75–85% at 22 °C, with cement board, glass-fibre and EPS effectively resistant and it is the cumulative time above that limit, not any single reading, that settles it.
The mistake I see most often is a natural one. The room feels sticky, so you set it colder. Go back to the sponge and you can see why that backfires. A thermostat only changes the temperature of the air. It does nothing to the amount of water in it, so the dew point stays exactly where it was. Take air at 26 °C and 60% humidity; its dew point is 17.6 °C. Cool that air to 22 °C and the dew point is still 17.6 °C. The only thing that moved is the humidity reading, which climbs to 76%. Nothing has dried. But every surface in the system is now colder, so the gap between those surfaces and the dew point has narrowed. You have made the cold bottle colder.
Water leaves a building by two routes and only two: a cooling coil that runs long enough and cold enough to wring the water out and drain it away, or a dehumidifier doing the same job deliberately. A system that is too big for the space cools the air quickly, satisfies the thermostat and shuts off before it has removed much water at all. That is how a room can read a comfortable 23 °C while it quietly deposits water in the ceiling above it.
You have made the cold bottle colder.
What I would do this month? I would start by measuring the dew point rather than temperature and humidity separately. Most decent loggers will calculate it for you; a week of readings, with attention to the overnight hours, tells you more than any spot check. Then I would compare that dew point to the surfaces. An infrared thermometer or a thermal camera on the diffusers, the pipework and the wall corners will show you which ones are colder than the air's dew point, and those are your problem spots, whether or not they look wet today.
Before touching the controls I would fix the air leaks and the insulation gaps, because air that never reaches a cold surface cannot condense on it. Then, when I did adjust the system, I would run it longer rather than colder: longer operation at a lower fan speed removes more water than short cold bursts ever will. And I would deal with the dust. Clean coils, drain pans and ducts remove the layer that holds water where bare metal would not.
One honest note on the figures. They come from the public Dubai International record, which is an airport station. A building's microclimate is not an airport's; a shaded, irrigated coastal plot and an exposed inland one can differ materially. Treat the numbers as the regional envelope your system is working inside, and measure the building itself before concluding anything about it.
The NCM will have lifted this morning's alert by breakfast. The one over your ceiling does not lift until November.
| Claim | Source |
|---|---|
| Median Sept dew point 23.6 °C, p95 28.0 °C | Meteostat stn 41194, 1983–2025, n = 28,781 [computed] |
| Dew point ≥ 24 °C for 45% of Sept hours, ≥ 26 °C for 27% | same |
| Oct 18% / Nov <1% | same |
| Only 6% of Sept nights are fog-capable (gap ≤ 2 °C) | same |
| Sept median level with July's 23.9 °C | same |
| 26 °C/60% RH → DP 17.6 °C; cool to 22 °C → 76% RH, DP unchanged | emc.py [computed] |