Sap Flow and Environmental Responses Across Salix psammophila Stand Ages

Water use and its environmental responses in artificially planted sand–fixing shrubs in arid and semi–arid regions are important for evaluating the sustainability of ecological restoration. This study investigated artificially planted Salix psammophila stands aged 3, 8, and 15 years in the Mu Us Sandy Land. From May to September 2025, branch sap flow rates were continuously monitored using the tissue heat balance method, while environmental factors, including solar radiation (Rs), air temperature (Ta), relative humidity (Rh), soil temperature (Ts), soil volumetric moisture content (Sh), wind speed (Ws), precipitation (Ra), and vapor pressure deficit (VPD), were simultaneously observed. Pearson correlation analysis, partial correlation analysis controlling for day of year (DOY), segmented regression, Random Forest (RF), and SHAP were used to compare the statistical associations and response patterns between sap flow and environmental factors among stands of different ages. The results showed that: (1) The daily mean branch sap flow rate during the growing season generally followed the order 8a > 3a > 15a, although the ranking varied among seasonal stages. Because the diameters of the monitored branches differed among stand ages, this result primarily reflects differences in branch–scale absolute sap flow rates among the sampled stands and cannot be directly equated with differences in hydraulic capacity per unit conducting tissue. (2) Sap flow in the 3a and 8a stands was strongly associated with Sh. In the 15a stand, sap flow was significantly negatively correlated with Ta, Ts, and VPD at the whole–growing–season scale; however, these relationships were no longer significant after controlling for DOY, indicating that they were largely influenced by common seasonal trends. (3) RF analysis showed that the relative predictive contributions of environmental variables differed among the stand–age–specific models. OOB–R2 ranged from 0.303 to 0.526, whereas CV–R2 values from 5–fold forward–chaining time–series cross–validation were below 0 for all models, indicating that the models could describe sap flow–environment relationships within the observation period to some extent but had limited temporal generalization to subsequent unseen periods. Overall, the sampled S. psammophila stands of different ages exhibited distinct sap flow–environment association patterns, with the 15a stand showing more pronounced seasonal dependence.