To estimate whole-tree or stand transpiration based on sap flow measurements, sap flux density (Fd) needs to be scaled up over the entire sapwood area. We investigated radial variations in Fd in the dominant oak (Quercus liaotungensis) and a commonly occurring oriental arborvitae (Platycladus orientalis) species using Granier-type thermal dissipation probes in a semi-arid forest in northern China. Our results show that Fd in oak trees reached a maximum just below the cambium and decreased towards the inner sapwood. On the other hand, radial sap flow patterns in oriental arborvitae trees differed among individuals. Fd values at different depths within an individual were highly correlated with each other in both species. We assessed the error of single-depth measurements when upscaling to whole-tree water use using multi-depth measurements as a reference. Omitting radial variations resulted in an overestimation of daily water use by 38% and 78% in oak trees with sapwood depths of approximately 2 and 3 cm, respectively. The outer sapwood annuli (0–1 cm) contributed 74% and 65% to the total water use in the two classes of sapwood depth. These contribution coefficients may be used as correction indices for estimating whole-tree water use. Errors of omitting radial variations in oriental arborvitae trees ranged from 42% to 21%. Using relatively long sensors (e.g. 2 cm) may improve the estimate accuracy in this species. The results would be applicable to the estimation of tree or stand transpiration based on routine depth measurements using one set of sensors. It is suggested that radial variations should be considered for accurately upscaling locally measured sap flow data to whole-tree water use in trees with wide sapwood and that different approaches may be applied to species with different patterns of radial variations.