Abstract A statistical regression method is presented for estimating hourly anthropogenic heat flux (AHF) using an anthropogenic pollutant emission inventory for use in mesoscale meteorological and air-quality modeling. Based on bottom-up AHF estimated from detailed energy consumption data and anthropogenic pollutant emissions of carbon monoxide (CO) and nitrogen oxides (NO x ) in the US National Emission Inventory year 2005 (NEI-2005), a robust regression relation between the AHF and the pollutant emissions is obtained for Houston. This relation is a combination of two power functions ( Y = aX b ) relating CO and NO x emissions to AHF, giving a determinant coefficient ( R 2 ) of 0.72. The AHF for Houston derived from the regression relation has high temporal ( R = 0.91) and spatial ( R = 0.83) correlations with the bottom-up AHF. Hourly AHF for the whole US in summer is estimated by applying the regression relation to the NEI-2005 summer pollutant emissions with a high spatial resolution of 4-km. The summer daily mean AHF range 10–40 W m −2 on a 4 × 4 km 2 grid scale with maximum heat fluxes of 50–140 W m −2 for major US cities. The AHFs derived from the regression relations between the bottom-up AHF and either CO or NO x emissions show a small difference of less than 5% (4.7 W m −2 ) in city-scale daily mean AHF, and similar R 2 statistics, compared to results from their combination. Thus, emissions of either species can be used to estimate AHF in the US cities. An hourly AHF inventory at 4 × 4 km 2 resolution over the entire US based on the combined regression is derived and made publicly available for use in mesoscale numerical modeling.