Abstract: It is a basic approach to estimate the stand transpiration through computing the average of sap flow density (Js) from a number of sample trees and then up-scaling. However, without considering the effect of tree size or spatial position in community on Js, the approach is likely to either under-or over-estimate the stand transpiration, especially in a high-density stand. In order to understand the difference in Js among trees of different size or degree of dominance, and further improve the accuracy of the up-scaling approach, a field experiment was conducted in a Larix principis-rupprechtii stand located at the semi-arid region of Liupan Mountains. In the experiment, the Js of five sample trees with different dominances (relative height) was continuously monitored by using thermal diffusion probes. Then the differences in Js among individual trees under three kinds of soil water conditions were analyzed with the data at mid-growing season when the tree leaf area reached to the maximum and remained relatively stable. It is found that the Js in trees with higher dominance ("higher trees") started earlier in the morning but ended later at night than in trees with lower dominance ("lower trees"). In addition, the maximum of Js during the daytime appeared earlier and was higher in higher trees. Thus the daily average of Js in higher trees was apparently higher as well. Furthermore, the Js in higher trees was more sensitive to the solar radiation (Rs) and the vapor pressure deficit (VPD) on a 5-min time scale, while on daily scale to the soil drought, higher trees showed less sensitive than the lower trees, implying a stronger ability for obtaining light and water. In general, however, no significant difference was found in the pattern of Js response to environmental conditions among sample trees, and the relative difference in Js was relatively stable. Correlation analysis indicated that the most important factors positively affecting the Js were the degree of dominance (relative height) and the tree height (PPP>0.05). In an improved approach, the Js of the forest stand was taken as the average of Js for all individual trees in the stand which was calculated by using the linear relation between Js and degree of dominance. The value calculated by the average of Js for the five sample trees was 16% lower than that calculated by the widely used approach. Therefore it is proposed that the degree of dominance should be taken into account in the up-scaling approach for the stand's Js or transpiration estimate in future.