Natural gas is the primary fuel used in U.S. residences, yet little is known about its consumption patterns and drivers. We use daily county-level gas consumption data to assess the spatial patterns of the relationships and the sensitivities of gas consumption to outdoor air temperature across U.S. households. We fitted linear-plus-plateau functions to daily gas consumption data in 1000 counties, and derived two key coefficients: the heating temperature threshold (
T
crit
) and the gas consumption rate change per 1°C temperature drop (Slope). We identified the main predictors of
T
crit
and Slope (like income, employment rate, and building type) using interpretable machine learning models built on census data. Finally, we estimated a potential 2.47 million MtCO
2
annual emission reduction in U.S. residences by gas savings due to household insulation improvements and hypothetical behavioral change toward reduced consumption by adopting a 1°C lower
T
crit
than the current value.