Calculate the heat index from temperature and humidity, with heat risk level.
The heat index quantifies how hot it actually feels when relative humidity is factored in alongside air temperature. High humidity impairs the body's primary cooling mechanism — sweat evaporation — so a humid 95°F day feels substantially hotter than a dry 95°F day. The NWS uses the Rothfusz multiple-regression equation, which was fitted against Steadman's 1979 empirical lookup tables. Two boundary adjustments refine the regression at low humidity (where it over-predicts) and high humidity near 80°F (where it under-predicts). The formula is only meaningful at or above 80°F (26.7°C) with at least 40% relative humidity.
Rothfusz regression (NWS)
HI = −42.379 + 2.049×T + 10.143×RH − 0.225×T×RH − 0.00684×T² − 0.0548×RH² + 0.00123×T²×RH + 0.000853×T×RH² − 0.000002×T²×RH²
Below 80°F, the difference between apparent and actual temperature is small enough to be negligible, and the Rothfusz polynomial was not fitted for that range. At lower temperatures, other factors such as sun angle, wind and clothing matter more than humidity for perceived warmth.
Heat index (NWS, US) and humidex (Environment and Climate Change Canada) both express 'apparent temperature', but use different empirical formulas calibrated against different datasets. Humidex often produces higher numbers than heat index for the same conditions. This calculator implements the NWS heat index.
The NWS risk categories (Caution, Extreme Caution, Danger, Extreme Danger) are guidelines for healthy adults at rest in the shade. Direct sun can increase the heat index by 10–15°F. Elderly people, infants, and those with certain medical conditions face higher risk at lower thresholds.
The Rothfusz equation is a polynomial fit to the middle of its valid domain. At very low humidity (< 13%) and temperatures between 80°F and 112°F, it overestimates the apparent temperature, so NWS subtracts a correction. At very high humidity (> 85%) between 80°F and 87°F, it underestimates, so NWS adds a correction. Both bring the regression into agreement with Steadman's original tables.