Apparent temperature: the feels-like number is a model

The feels-like figure is calculated and the air temperature beside it is measured. Every other property of the number follows from that one distinction. A thermometer inside a ventilated screen reports what the air is doing. The second figure reports what a model says that air would do to a person.

Australia's Bureau of Meteorology sets the arrangement out plainly on its own explainer. The bureau measures ambient temperature with instruments, then estimates what the temperature feels like using a mathematical model. Its definition names the move: the temperature that, at a reference humidity and without wind, would produce the same discomfort as the conditions actually present.

The figure exists because one reading answers too little. Air at a given temperature lands differently on a person depending on how much water it carries and how fast it is moving. Folding those into a single number gives the public one thing to look at instead of three.

That convenience costs something worth knowing about. A model has to assume what it cannot observe, and those assumptions are built into every figure it produces. Read the number without them and it looks like a direct measurement of the reader's own afternoon.

Two apps disagree because they run different models, and both can be reporting honestly. Several apparent-temperature indices are in public use. Each was built by a different agency for a different purpose, and each takes a different set of inputs. Feed one afternoon into all of them and they return different answers by design.

The United States National Weather Service publishes a heat index built from air temperature and relative humidity. Nothing else enters it. The service states the assumption openly: the values were devised for shady, light-wind conditions, so standing in full sunshine pushes the real effect above what the index reports.

Environment Canada publishes humidex. Its glossary calls it an index of how hot or humid the weather feels to the average person, and it too runs on temperature and humidity. The Australian apparent temperature adds wind speed to that pair. Three national services, three scales, and no agreed way to convert between them.

Wind is the input that changes the character of the answer. An index that ignores it returns the same figure for a still afternoon and a breezy one. An index that counts it reads lower when air is moving, because moving air carries heat off skin faster.

No arbiter sits above the set of them. Each was calibrated against a particular climate and a particular population, so none is the reference the others should be corrected towards. A scale built where summers run dry and a scale built where they do not will weight the moisture term differently, and each is defensible on its own ground.

Humidity carries most of the gap here because the other inputs barely move. Air temperature in Singapore works within a narrow band across the year, and wind at ground level in a built-up area is often slight. Water content is the variable doing the work, and every index in use leans on it heavily.

The mechanism behind that is shared by all of them. A body sheds heat partly by evaporating moisture off skin, and air already carrying plenty of water accepts less of it. The indices differ in how they express that effect, not in what they are describing.

How far evaporation can still carry a temperature downward is wet-bulb temperature, which has its own page. The temperature at which condensation starts to appear is dew point, and that has another. This figure sits on top of both. It converts what those readings describe into a scale a person can compare against an ordinary temperature.

One consequence catches people out. A day that reads unremarkable on the thermometer can still produce a high feels-like figure, because the whole difference came from the moisture term. The forecast looked ordinary and the afternoon did not, and both accounts are accurate.

Relative humidity alone is a poor stand-in for any of this. It expresses water content as a percentage of what the air could carry at its current temperature. The same percentage therefore means different things on a hot day and a mild one. What a home here should aim for, and which meter reads it, belongs to the indoor humidity target guide.

Which of the two inputs moved is the part worth noticing. A week where the forecast temperature holds steady while the moisture climbs produces a rising feels-like figure with a flat thermometer behind it. That divergence shows up in the published observations, and it accounts for the complaint better than the derived number does.

No. The Meteorological Service Singapore publishes no apparent temperature and no feels-like figure at all. Its current observations page carries rainfall, temperature, relative humidity, wind and visibility. The forecast on its portal gives a temperature range, a humidity range and a wind description.

The public heat measure here is the heat stress advisory. NEA publishes it with the Meteorological Service, for outdoor exposure, and it is built on a different basis. What goes into it, and who it was written for, has its own page. The levels it sets belong to the agencies behind them.

That absence has a practical consequence. A feels-like figure on a Singapore reader's phone came from whoever built the app, and not from the national service. There is no official number to appeal to, and no agency to ask why it read what it read.

One more consequence lands on anyone tracking the figure over time. An app that changes its weather provider changes its index at the same moment, and the number on the tile shifts while the air does not. A household setting this August against last August through one app is comparing two things it cannot see.

Two apps can therefore draw on the same official observations and still disagree. Temperature, humidity and wind are published for anyone to use. What gets done with them afterwards is the app's own choice, and that choice is the part nobody prints on the tile.

Every index described here models a person outdoors, so none of them describes a cooled room. The Australian model walks outdoors in the shade. The American index assumes an outdoor shade condition as well. A shut bedroom with a running unit falls outside the situation all of them were built for.

Where the sensors sit is only half the reason. A coil sitting below the dew point pulls water out of whatever air crosses it, then sends that water down a drain. The room a household is sitting in has had the dominant input to that outdoor figure removed from it. Applying the outdoor number to that room describes air the room no longer contains.

The outdoor half of the system does have a stake in the weather, and this is not the figure that carries it. A condenser rejects heat into outdoor air, and what governs how easily it manages that is the dry air temperature and the airflow it can pull across its coil. Sweat rates and clothing assumptions play no part in it. The reading that describes what an outdoor unit is working against is ambient temperature.

The figure shares no scale with the number on the remote either. A setpoint is a target for air temperature. An apparent-temperature figure is an equivalent-discomfort statement produced by a model. Setting one against the other compares two different quantities that happen to print in the same units.

Push back when a quote leans on the week rather than on a reading. A punishing stretch of weather explains why a household noticed something, and explains nothing about the machine. Ask which instruments went on the system, and what they showed. Weather is context for a complaint and never a substitute for an instrument.

The opposite trap is quieter and just as common. A high figure gets used to wave away a real loss of capacity, so the household waits out the week and the fault carries on underneath it. What settles that is what the room did before the bad stretch began, and what it does once the stretch has passed.