Abstract: Heat Field Deformation (HFD) is one of the most important techniques to measure tree sap flow; However, the high accuracy is often required for absolute quantification. Particularly, the potential applications can be expected for scarce water resources in shelterbelts. This study aims to investigate the radial variation in sap flow and their its relationship with wood anatomical structure. The tree species were selected as Populus × euramericana cv. ‘74/76’ and Salix matsudana. HFD sensors were employed to measure sap flux density at different depths (0~50 mm). Whole-tree transpiration was compared using high-precision weighing lysimeters. A systematic analysis was made to determine the measurement error of the HFD and radial pattern of sap flow. Furthermore, the relationship between sap flow and water transport function was constructed to combine with wood anatomical structure. A systematic investigation was made to determined vessel lumen diameter, vessel frequency, and total vessel area per unit xylem area from cores extracted at the same radial positions using light microscopy. The results showed that: 1) The original HFD algorithm systematically underestimated absolute sap flux in a species specific manner. Compared with the weighing lysimeter, the average underestimation was 84.8% for poplar and 56.7% for willow, indicating a stronger bias in tree species with larger vessels and higher hydraulic conductance. Species specific calibration equations were developed using linear regression. Lysimeter measured transpiration was taken as the dependent variable, while HFD derived flux was the independent variable. 2) The accuracy of sap flux density varied significantly among measured depths, compared with the gravimetric whole tree transpiration. Among all radial positions, the 10 mm depth consistently yielded the highest goodness of fit, leading to the preferred single point measurement depth to estimate whole tree transpiration. In contrast, the positions produced increasingly larger deviations, with the percentage errors exceeding 40% for both species. Radial heterogeneity was neglected to rely only on outer layer flux, resulting in substantial overestimation of whole tree transpiration—potentially by several fold depending on the species—indicating the radial profiles after flux calculations. 3) Sap flux density exhibited a radial distribution pattern for both species, progressively decreasing from the outer sapwood toward the inner xylem. Despite both diffuse being diffuse-porous, the attenuation patterns differed markedly between species: Salix matsudana displayed a relatively uniform gradient, with the flux at 50 mm depth, whereas in Populus × euramericana cv. ‘74/76’, sap flow was highly concentrated within the outermost 10 mm zone. The high concentration in poplar was attributed to a steep hydraulic gradient near the cambium. Moreover, the radial profile derived from multi point HFD measurements effectively delineated the boundary between sapwood and heartwood. A non destructive identification was realized in the active water conducting region, suitable for accurate upscaling. 4) Radial variation in sap flow was closely linked to wood anatomical structure. In both species, mean vessel diameter and vessel area per unit xylem area decreased radially from outer to inner sapwood, closely mirroring the decline in sap flux density. Vessel diameter directly governed water transport efficiency, according to the Hagen Poiseuille law. The radial flux distribution was shaped under by a radial gradient of vessel diameter, together with possible pit blockage and tylosis formation. There were the interspecific differences in vessel dimensions and radial variation amplitudes between the two species at an anatomical level. The sap flow profiles highlighted that the species specific hydraulic architecture was considered to interpret HFD data for calibration. In conclusion, the HFD with multi point measurements can be expected for the radial sap flow, indicating bi directional flow identification, high temporal resolution, and sapwood heartwood delineation. Yet its raw estimates can be calibrated for absolute quantification. Radial profiles and calibration equations after multi depth measurements are recommended for the upscaling accuracy from point scale sap flow to whole tree transpiration. These findings can support better estimation of individual tree transpiration and water regulation in arid zone shelterbelts, urban greening, and ecological restoration. Thereby, the precision and reliability of tree water monitoring can also be enhanced under various climatic scenarios.