The heat index is the US National Weather Service's estimate of how hot humid air feels, worked out in two stages: a simple formula first, then Rothfusz's regression if that first estimate is warm enough. On 6 November 2026 the procedure put 272 of 365 paired Singapore station-hours from data.gov.sg at Caution or above in NWS Amarillo's table. Seven reached Danger, all between 12:00 and 14:00. The handover depends on whose wording is followed. The equation page of NWS's Weather Prediction Center (WPC) says 80 °F or higher; its calculator's code tests above 79 °F. The page's rule alone would have changed the class of 9 of those station-hours. In the 10 cells of NOAA's larger heat index chart where the two rules disagree, the chart sides with the code every time.
A first estimate from the simple formula flips the procedure to Rothfusz's regression
Lans P. Rothfusz fitted his regression in 1990, in NWS technical attachment SR 90-23, to Steadman's heat index table. He writes that "No true equation for the Heat Index exists" and gives the fitted value an error of ±1.3 °F, which describes the fit rather than a measured accuracy. His model person is 5 ft 7 in and 147 lb, walking at 3.1 mph in a 5-knot wind.
WPC's Heat Index Equation page adds a gate and two patches. It prints the simple formula as HI = 0.5 * {T + 61.0 + [(T-68.0)*1.2] + (RH*0.094)}. That leading half of a sum is already the average with the air temperature. One patch subtracts a term below 13% humidity at 80–112 °F; the other adds one above 85% at 80–87 °F.
WPC's calculator code throws the switch just above 79 °F
The page's prose says the simple formula "is computed first and the result averaged with the temperature", and the regression applies if that reaches 80 °F. The Heat Index Calculator averages 61 + (T − 68) × 1.2 + RH × 0.094 once with T and runs the regression only if the result is above 79. At 79 °F air, the code hands over from 51.1% humidity.
The NWS Amarillo heat index page links a NOAA chart of 996 printed cells, covering 76–126 °F air at 4–100% humidity and reaching 183 °F. To the whole degree, the calculator code reproduces 992 of them, each miss 1 °F off at 48% humidity; the page's 80 °F rule reproduces 982. The chart's edges are not a stated limit of Steadman's data.
Crossed wires in this calculator's old kernel kept humid hours on the simple branch
Until a revision dated 7 November 2026, the kernel behind this wind chill and heat index calculator departed from WPC's code twice. It averaged the printed formula with the temperature a second time, and it gated on the page's 80 °F instead of the code's 79. At 79 °F air it reached the regression only from 93.7% humidity. Alone, the gate changed the class of 9 of the 365 station-hours on 6 November and the extra average 10. Together they changed 72, every one from Caution to no class, all between 25.5 and 26.1 °C, 45 of them between 19:00 and 06:59.
Enter 26.1 °C and 94% humidity and the first estimate is 81.0 °F. The old kernel averaged it again to 79.99 and gave no class. The revised kernel averages once, passes the gate and reports 83.2 °F (28.4 °C), Caution, and it matches our transcription of WPC's code in all 89,991 cells we compared, 41–130 °F in 0.1 °F steps at every whole humidity. It now refuses humidity above 100%, grades on the one-decimal value it shows, and past 183 °F, the chart's top value, shows "Above 183.0°F" instead of a figure. The revision covers this calculator's kernel only.
An always-on regression overloads mild air by up to 11 °F
Across 20–26.6 °C at every humidity, the regression without its gate ran up to 10.99 °F above the full procedure on our arithmetic: 77.4 °F at 20 °C and 41% humidity, where the full procedure gives 66.4 °F. Over 26–40 °C at 40–100% humidity, in 8,601 cells, the bare regression and the full procedure differed in 580, almost all through the high-humidity patch, by at most 2.08 °F (26.7 °C, 100%); 38 cells changed class.
Anderson, Bell and Peng compared it with Steadman in Environmental Health Perspectives in 2013. Of the 21 heat index algorithms they found in environmental research, the NWS one performed best, "on average, within 0.4°C/0.7°F" of Steadman's values. At the hot end nothing stops the curve: at 100% humidity it passes 150 °F from 34.1 °C air. WPC says the regression "is not valid" beyond the range of data Steadman considered, and gives no number.
Most of Singapore's station-hours on 6 November powered up to Caution or above
We paired hourly air temperature and humidity readings from 17 stations over 23 hours of 6 November 2026. NWS Amarillo's table starts Caution at 80 °F, Extreme caution at 90, Danger at 103 and Extreme danger at 125.
| NWS Amarillo class | Station-hours | Share of 365 |
|---|---|---|
| No class | 93 | 25.5% |
| Caution | 182 | 49.9% |
| Extreme caution | 83 | 22.7% |
| Danger | 7 | 1.9% |
The peak was 105.9 °F (41.1 °C) at station S06 at noon, from 32.2 °C and 70.2% humidity. The counts use unrounded values; graded on one decimal, one hour at 89.97 °F moves up to Extreme caution. NOAA's larger chart grades differently, from Very Warm at 80 °F to Extremely Hot at 130.
A second day, 15 May 2026, put 437 of 443 station-hours at Caution or above, but its series looks odd, warmest at 01:00, and we did not check it against Meteorological Service Singapore records. Two days make a snapshot and say nothing about climate. Contains information from Air Temperature across Singapore and Relative Humidity across Singapore accessed on 7 November 2026 from data.gov.sg (National Environment Agency) which is made available under the terms of the Singapore Open Data Licence version 1.0.
Wind chill stays disconnected above 50 °F and in near-calm air
NWS's wind chill page defines the index only at or below 50 °F with wind above 3 mph. In its example, 0 °F in a 15 mph wind gives −19 °F, at which exposed skin can freeze in 30 minutes; frostbite needs air below freezing. This calculator keeps both edges: at 10 °C with a humidity entered, 4.8 km/h of wind leaves it on the heat index's simple estimate and 5 km/h switches it to wind chill, with no frostbite grade above freezing.
Singapore's heat-stress readings run on a separate circuit
Singapore's own heat-stress figure is WBGT. Meteorological Service Singapore's heat stress page, read on 7 November 2026, calls it "a composite measure that takes into account air temperature, humidity, wind and solar radiation", averaged over the past 15 minutes. Rothfusz's regression takes only temperature and humidity, and NWS Amarillo notes that direct sunlight can add up to 15 °F to a heat index. How WBGT compares with shortcut formulas fed by station weather has its own guide, as does why a humidity reading needs its temperature.
View author profile → · Editorial policy
