Investigation of high resolution stem water potential in mature beech trees and relationships to water supply, demand and storage

Understanding of water transport within soil plant atmosphere continuum (SPAC) is essential for predicting tree functioning under changing environmental conditions. Xylem water potential (Ψ_xylem) reflects the energy state of water in plants, yet continuous monitoring has long been technically challenging, limiting insights into tree response to drivers such as vapor pressure deficit (VPD), soil water availability, and stem water storage. In a throughfall exclusion experiment (KROOF site, Germany) we continuously measured water fluxes and Ψ_xylem in mature beech trees along the entire SPAC. Our objectives were to quantify influence of soil water potential (Ψ_soil), stem water storage and VPD on tree water uptake and Ψ_xylem over diurnal cycles, and to test whether stem water storage predicts the hysteresis relationship between Ψ_xylem and sapflow (J).  We installed soil water potential and water content sensors in four different soil depths and took soil samples for natural abundance δ²H and δ¹⁸O isotopes to assess water uptake depths. Xylem water potential was measured continuously with microtensiometers (FloraPulse) at breast height and at the lower end of the crown, where we also installed sapflow sensors and point-dendrometers. We used 24 XGBoost models, separated by hour and calculated SHAP values to provide information about the importance of soil water potential, stem water storage and VPD on Ψ_xylem generally and over a diurnal cycle. We determined stem water storage using detrended (daily centered) and scaled dendrometer data (SDV) and calculated a mixed model to investigate its relationship of min and max Ψ_xylem values combined with Ψ_soil. Finally, we computed XGBoost models to predict Ψ_xylem hysteresis with J, J + SDV as well as J + SDV + Ψ_soil. Our models show a strong impact of SDV and VPD on Ψ_xylem while the impact of Ψ_soil was marginal.  Water uptake occurred mainly from upper soil layers (0-30 cm depth) but Ψ_soil of depth 30 and 50 showed the largest impact on Ψ_xylem. Diurnally SDV and VPD had the biggest impact, while there was no shift in importance of different soil depths on Ψ_xylem. We observed a linear relationship between min and max Ψ_xylem and SDV. At breast height, we found a significant interaction with Ψ_soil, while this was not observed in the lower crown. Sapflow as a single predictor for Ψ_xylem showed a direct relationship while SDV in addition was able to predict the daily hysteresis of Ψ_xylem. Water uptake was only weakly depended on Ψ_soil, possibly because the observed trees were not limited by water supply. SDV, which can be seen as a proxy for stem water storage, seemed to be a main factor predicting Ψ_xylem. The influence of Ψ_soil on SDV at breast height and its absence in the lower crown could show that storage status may vary within the tree. SDV, in addition to sapflow, is able to provide a second axis of information to also predict hysteresis curves between daily extremes.