Sap flow probes are routinely used in forest and horticulture hydrology for estimating tree water use. This requires unbiased measurements when upscaling from tree to stand level, but accuracy and comparability of different thermometric methods have been questioned. Three sap flow measuring techniques were compared against gravimetric flow measurement in cut stem segments: ‘Granier-type’ thermal dissipation probes (TDP; three different sensor types), the heat field deformation method (HFD), and the heat ratio method (HRM). For the empirical methods (TDP and HFD), new calibration parameters were estimated using a nonlinear hierarchical modelling approach. 66 stem segments from five temperate, diffuse-porous tree species (9–16 cm stem diameter, 100 cm stem length) were exposed to a wide range of flux densities by applying subatmospheric pressure (−50 to −650 hPa) analogous to natural flow conditions in the field. All TDP probes underestimated flux density by 23–45% when calculated with Granier's original calibration parameters, with the deviation increasing with flux rate. The accuracy was significantly improved by estimating new calibration parameters, especially for probes differing from Granier's original sensor design. Species-specific parameters further improved accuracy, although the species differences might partially be explained by variation in the observed ranges of sap flux. The HFD sensor overestimated gravimetric flow by ∼11%; empirical calibration did not improve its accuracy compared to the manufacturer's equation. At low to medium flow rates, the HRM system achieved higher accuracy than the other probes (0.8% underestimation), while performing poorly at high flux rates under our measurement settings (energy input of 25 J). Both for TDP and HFD sensors, we observed a surprisingly large variability in calibration parameters between different stems of the same species. We conclude that (i) TDP and HFD sensors require species-specific calibration to measure sap flux with high accuracy, (ii) the original Granier equation cannot be used for TDP probes with deviating design, and (iii), at low to medium flow rates, the highest accuracy can be achieved with HRM sensors. Our results help to increase the accuracy of tree sap flow measurements with thermal dissipation probes, and to assess various levels of errors related to the different thermometric methods. This is important when synthesizing forest transpiration data on regional and global scales.