Abstract Green roofs are a component of energy-saving architecture. Building energy savings due to green roofs are a function of both vegetation and substrate properties. Direct empirical measurements of heat flux through green roof layers represent a method for comparing green roof vegetation or substrate types, but these methods are limited by the expense of heat flux sensors. This paper proposes to use an inexpensive thermo-module for heat flux measurements. The thermo-module heat flux sensor had a big advantage for both expense and measuring sensitivity compared to a commercial heat flux meter: two orders of magnitude less cost and exceeding three times higher sensitivity. Then the thermo-module heat flux sensors were installed in a replicated extensive green roof, comparing heat flux measurements during winter conditions on a roof in western Japan among seven different vegetation type treatments. Vegetation had strong effects on both temporal mean and range of heat flux values. The strongest performing plant type was Luzula capitata, a low-growing graminoid with dense leaf cover even in winter, showing up to 50% less heat loss than the poorest performing species. This kind of sensor is recommended for further replicated empirical comparisons of green roofs or other energy-saving architectural technologies.