The heat index (HI) metric is used in many studies linking weather conditions (observed or climate model-projected temperature and humidity) to health impacts. Because climate models often provide only daily data, a common approximation method estimates the daily maximum HI (hismaxest) from daily maximum temperature (tasmax) and minimum relative humidity (hursmin), assuming they coincide with the actual peak HI (hismax24). This study evaluates the accuracy of this approximation using hourly station-based data from NOAA’s Integrated Surface Database (ISD-Lite) and the U.S. Climate Reference Network (USCRN). Though we find that hismaxest either matches or slightly underestimates hismax24 for most summer days and locations, it significantly underestimates the occurrence of extreme heat days in the hottest regions. Specifically, hismaxest misses over 35% of days exceeding the 95th percentile of June–July–August hismax24 at 7 of the 37 stations examined, despite performing well in cooler, drier regions. This underestimation follows an increasing nonlinear relationship with local HI thresholds, indicating that hismaxest may underestimate extreme heat risks in vulnerable regions of the southern U.S. These discrepancies arise because HI is more sensitive to relative humidity in hotter regions, making hismaxest less reliable. Our findings quantify an often overlooked uncertainty arising from daily climate model data resolution that can be comparable to scenario and model sensitivity uncertainties. These results highlight the need for careful interpretation of climate model heat projections, and emphasize the value of archiving hourly climate model data for multivariate index calculations.
