Integrating multivariate ordination and machine learning to disentangle the environmental drivers of xylem sap redox metabolism in trees

Abstract

Xylem sap is increasingly recognized as a dynamic biological matrix reflecting whole-plant physiological status, but its biochemical variation under field conditions remains insufficiently characterized. We investigated oxidative stress markers, osmolytes, antioxidant enzymes, and redox-related enzymes in xylem sap from three focal tree individuals representing Fraxinus excelsior, Populus nigra, and Pinus sylvestris. Sap was collected by passive stem tapping using a custom-built apparatus, and biochemical patterns were evaluated using multivariate statistical and machine-learning approaches. The three focal trees showed distinct biochemical profiles within the present dataset. The focal P. nigra individual was associated with relatively higher antioxidant enzyme activities, whereas the focal F. excelsior and P. sylvestris individuals were more closely associated with oxidative-damage and metabolic-adjustment traits. Precipitation and wind direction were retained as the main meteorological variables associated with biochemical variation, with wind direction interpreted as an atmospheric correlate rather than a direct physiological driver. Exploratory machine-learning analyses highlighted catalase and selected meteorological variables as influential predictors. Overall, the findings support the potential of xylem sap for integrative ecophysiological monitoring while emphasizing the exploratory nature of patterns derived from repeated measurements of three focal trees.