Clearwater corrections were proposed to deal with the situation that part of the heat dissipation (HD) or Granier probes was in contact with the non‐conducting xylem or heartwood in trees with shallow sapwood depth (<20 mm) (Ds). To perform these corrections requires an accurate determination of the Ds. However, in practice, it is often error‐prone to distinguish the sapwood from heartwood. This error may transmit into and potentially bias the Clearwater corrections. In the present study, we quantitatively analysed the responses of Clearwater corrections to assumed Ds error of up to a 2 mm range for two evergreen tree species: Eucalyptus citriodora and Acacia auriculiformis. We found that the corrected sap flux density (Js) was substantially affected by Ds inaccuracies. Underestimation of Ds led to overestimation of Js, and Js was underestimated more when Ds was overestimated of the equal magnitude. Relative error in Js increased quadratically with Ds error. By contrast, the responses of relative Js error to varying Js and Ds could be fitted by linear models. Decreasing Ds as well as increasing Js jointly increased the relative error in Js up to ~50%. This study provides convenient references for assessing the impact of Ds error on Js corrections, contributing to more reliable water consumption evaluations for trees and forest stands.