Abstract Profiles of the sensible heat flux are key to understanding atmospheric-boundary-layer (ABL) structure and development. Based on temperature profiling by a remotely-piloted aircraft system (RPAS), the Small Unmanned Meteorological Observer (SUMO) platform, during the Boundary Layer Late Afternoon and Sunset Turbulence (BLLAST) field campaign, 108 heat-flux profiles are estimated using a simplified version of the prognostic equation for potential temperature $$\theta $$ θ that relates the tendency in $$\theta $$ θ to the flux divergence over the time span between two consecutive flights. We validate for the first time RPAS-based heat-flux profiles against a network of 12 ground-based eddy-covariance stations (2–60 m above ground), in addition to a comparison with fluxes from a manned aircraft and a tethered balloon, enabling the detailed investigation of the potential and limitations related to this technique for obtaining fluxes from RPAS platforms. We find that appropriate treatment of horizontal advection is crucial for obtaining realistic flux values, and present correction methods specific to the state of the ABL. Advection from a mesoscale model is also tested as another correction method. The SUMO heat-flux estimates with appropriate corrections compare well with the reference measurements, with differences in the performance depending on the time of day, since the evening period shows the best results (94 $$\%$$ % within the spread of ground stations), and the afternoon period shows the poorest results (63 $$\%$$ % within the spread). The diurnal cycle of the heat flux is captured by the SUMO platform for several days, with the flux values from the manned aircraft and tethered balloon coinciding well with those from the SUMO platform.