Advanced thermal comfort: how Infrared Core computes the heat people feel

One hot afternoon in a Vienna courtyard shows what Infrared Core computes for every square metre: sun and shade, hot walls, trees, water and wind, combined into the heat people actually feel.

A hot afternoon in a Vienna courtyard

It is 15 July, 15:00, in the MuseumsQuartier courtyard in Vienna. The air is 31 °C, yet people on the open square feel strong to very strong heat stress. The thermometer does not explain this. Radiation does: direct sun from above, sunlight that bounces off bright walls and paving, and heat that the stone has stored since the morning.

We tested eight cooling measures on this 4,150 m² square, one at a time, with Infrared Core. For every 1 m cell of a 500 m × 500 m area, the engine computes the felt temperature, the Universal Thermal Climate Index (UTCI).

The open square of the MuseumsQuartier courtyard in Vienna under strong heat stress, before any measure — The open square today: at 15:00 the paving is in strong to very strong heat stress. Schematic pattern of the felt heat.

Which measures cool people, and which do not

Shade does most of the work. Eight street trees cool the ground below them by 1.4 K on average and by up to 4 K in their deepest shade, and they give the best cooling per euro. A shade sail and a PV canopy cool the space below them by 2.3 K and 2.7 K, and the canopy makes electricity at the same time. A green facade cools the street in front of it by 1.3 K.

The MuseumsQuartier courtyard in Vienna with the eight cooling measures we tested — The eight measures, each tested alone: (1) street trees, (2) shade sail, (3) PV canopy, (4) green facade, (5) watered grass, (6) light facade paint, (7) cool pavement and (8) green roof.

Measure	At the measure	Whole square	Construction cost
Street trees (8 trees, 10 m tall)	−1.4 K, up to −4 K in the shade	−0.55 K	€24k to €80k
Shade sail (300 m², 6 m high)	−2.3 K below the sail	−0.27 K	€46k to €108k
PV canopy (384 m² of panels)	−2.7 K below the canopy	−0.41 K	€115k to €192k
Watered grass (300 m², de-sealed)	−1.7 K on the patch	−0.13 K	€31k to €93k
Green facade (climbers, 109 m²)	−1.3 K in front of it	−0.06 K	€11k to €54k
Light facade paint (2,973 m²)	+0.4 K in front of it, at 15:00	Warmer for people	€74k to €178k
Cool pavement (albedo 0.45, 576 m²)	+1.1 K on it, at 15:00	Warmer for people	€6k to €29k
Green roof (extensive)	About 0 K at street level	About 0 K	Not priced
10 trees and 2 shade sails	Found by the AI agent	−1.03 K	€84k to €226k
PV canopy only, larger	For comparison	−1.16 K	€376k to €627k

Green cools people only when it has water. Watered grass cools its patch by 1.7 K, while dry grass or bare soil gives only 0.3 to 0.6 K.

Two results surprise most people. Light facade paint and cool pavement make the square warmer for people at 15:00, because bright surfaces reflect the sun back onto the body. They still help the buildings and the night-time air, but not a person who crosses the square in the afternoon. The green roof cools the roof and does nothing for the street. A rule of thumb misses all three effects. A model that follows every ray of sunlight finds them.

An AI agent finds the cheapest mix

Next, we gave an AI design agent a clear goal: cool the whole square by 1 K at the lowest cost. The agent combined measures, sent each variant to the engine and read the result.

Its best answer was 10 street trees and 2 shade sails with 360 m² of fabric: −1.03 K on the whole square for €84k to €226k. A larger PV canopy alone reaches −1.16 K, but costs €376k to €627k, about three times as much. Measures that stand apart add up almost linearly, so the agent can plan with the single results and then check the mix in one run: it estimated −1.00 K, and the engine measured −1.03 K. Because every run takes seconds, the agent tests more than 30 variants in a few minutes.

Cooling on the whole square against construction cost for each measure and for the cheapest mix — Cooling on the whole square against construction cost (log scale). The best value line climbs from the green facade over the trees to the cheapest mix at the −1 K goal.

What Infrared Core computes for every 1 m cell

The courtyard result comes from a full radiation budget of the 3D scene, hour by hour. For every 1 m cell, Infrared Core computes:

Sun and shadeDirect sun and a Perez all-weather sky that splits the sky into 145 patches.
Sky viewHow much open sky each spot sees, which controls how fast it cools at night.
MaterialsFacades, roofs and ground, from brick and glass to asphalt, soil and water.
Reflected sunlightFrom walls and from the ground. It reaches people even in shade.
Heat stored in walls and groundEach facade solves its own heat balance, and heavy walls keep a memory, so a wall that baked all day still radiates in the evening.
Trees with light through the crownCrowns let some sun pass, and deciduous trees let winter sun through when their leaves are off.
InstallationsShade sails, canopies and PV panels cast shade and have their own surface temperature.
Irrigation and evaporationWatered surfaces cool more on dry days, as evaporation follows the dryness of the air.
Local windLocal wind for each cell from our AI wind model, so a sheltered courtyard and a windy street each get their own wind.

From all of this, the engine computes the UTCI for every hour, with that hour's weather.

Choose the physics for each run

You choose how much physics goes into each thermal comfort run. There are two levels. Standard costs 10 tokens per 512 m tile; Advanced with local wind costs 50 tokens per tile, with the wind runs included. The pricing page lists every simulation.

StandardSun, sky, shade and materials. The default, and the right start for most design questions.Default

Advanced with local windAdds local wind for each cell from our AI wind model. For courtyards, street canyons and windy corners.Adds local wind

Local wind changes the felt heat a lot. In a Vienna calculation for July, a street canyon came out 4.5 K hotter and a closed courtyard 6.5 K hotter than with one wind speed for the whole area, while a street that channels the wind came out 1.2 K cooler. This is a calculation, not a measurement. A cell in the open gets exactly the wind of the weather file, so the two modes agree where nothing blocks the wind.

UTCI: one number for the heat people feel

UTCI is the international index for outdoor thermal comfort. It turns air temperature, humidity, wind and radiation into one temperature: the air temperature that would feel the same to a person who walks at 4 km/h. Infrared Core uses the standard polynomial of the Fiala human heat-balance model and matches an independent reference implementation to within 0.003 K.

UTCI in °C over a city district on the infrared.city platform, with street trees — UTCI in °C on the platform, for April afternoons from 14:00 to 18:00. The tree-lined streets stand out.

UTCI (°C)	Stress class
above 46	Extreme heat stress
38 to 46	Very strong heat stress
32 to 38	Strong heat stress
26 to 32	Moderate heat stress
9 to 26	No thermal stress (the comfort target)
0 to 9	Slight cold stress
−13 to 0	Moderate cold stress
−27 to −13	Strong cold stress
−40 to −27	Very strong cold stress
below −40	Extreme cold stress

Radiant heat matters most. The engine sums all of it into the mean radiant temperature (Tmrt), and in a sunny street Tmrt can sit 20 to 30 °C above the air temperature. That is why shade changes the felt heat so much more than the air temperature, and why the engine spends most of its effort on radiation.

From one hour to a whole season

One hot hour shows the peak. Planning decisions need the whole summer, or the whole year. Thermal Comfort Statistics runs the same engine over any period, also across the year end, such as December to February. For every location it reports the share of hours in comfort (9 to 26 °C), in heat stress and in cold stress.

For a season or a year, the engine picks representative hours that stand for the whole period, and it always keeps the hottest and the coldest hours. Heat waves and cold snaps stay in the result instead of being averaged away.

Thermal Comfort Statistics for a summer on the infrared.city platform — Thermal Comfort Statistics on the platform: the share of time in comfort for each location, June to September, 10:00 to 14:00.

Tested in the street

We checked the radiant heat (Tmrt) of Infrared Core against field instruments in four cities, and we ran SOLWEIG, the established research model, with the same input (same points, weather and 3D scene). The errors side by side:

City	Infrared Core	SOLWEIG
Gothenburg, Sweden	4.5 K	5.0 K
Phoenix, USA	11.1 K	14.3 K
Melbourne, Australia	11.8 K	12.5 K
Guelph, Canada	9.7 K	12.3 K

Error (RMSE) of the mean radiant temperature against measurements. Melbourne and Guelph were held out: we did not tune the engine on them.

The felt heat itself holds up too. At Guelph, a site we never tuned for, the felt heat (UTCI) stays within about 4 K of the street measurements (3.9 K error). In Phoenix, a mobile instrument cart measured 438 points across a desert park: the radiant-heat error of Infrared Core was 11.1 K, against 14.1 to 15.7 K for SOLWEIG in our head-to-head report.

The building blocks pass their own tests. The facade surface temperature stays within 2.95 K of EnergyPlus, the reference tool for building energy, with almost no bias (−0.33 K). Solar radiation agrees with Radiance, the reference for lighting simulation, within about 2.6 % on a Vienna tile and 4.5 % on a Hong Kong tile, and the median difference on facades is 0.1 to 4.3 %. As with any simulation, results depend on the input geometry and the weather file. All sites, data and methods are on the Validation page.

Fast enough to design with

A hot day on 500 m × 500 m takes about 1 s of engine time, and a whole month takes about 2 s. That speed changes how you work. A designer can try an idea while the meeting is still running, and an AI agent can test 30 or more variants in minutes, as it did in the courtyard.

Perceived heat (UTCI) on a 1 m grid over a district in Singapore — Felt heat (UTCI) on a 1 m grid over a whole district in Singapore, from the platform.

What comes next

Custom ground materials are coming, with their own physical properties and desert presets for gravel, decomposed granite and dry sand, for plazas in hot and dry climates. A same-input comparison with ENVI-met is also coming, so you can see both tools side by side on one site.

Run it where you work

The same engine runs on the platform, in Grasshopper, QGIS and ArcGIS and through the API and SDK, with the same results everywhere. Pick a block that gets too hot in summer, add trees, shade and water, and see in seconds what people will feel. To learn more about the engine, read about Infrared Core.

How does this compare with ENVI-met, SOLWEIG or PALM-4U? See when to use which tool.