Abstract: Tracking the sap flux of woody plants in savannas is essential for understanding their response to climate change and human management. Solar-induced fluorescence (SIF) has potential to predict transpiration yet its applicability for estimating savanna sap flux is unclear. Using three years of tower-based far-red SIF observations and ground-based sap flow monitoring in a temperate savanna of Otindag Sandy Land, China, we investigated the relationship between far-red SIF and sap flux density and developed linear and random forest models for estimating. The results show a variable correlation between SIF and sap flux density for Ulmus pumila var. sabulosa (J.H. Xin) G.H. Zhu & D.H. Bian (U. pumila.) at an hourly scale. The strongest correlations were during the mid- growth period July and August when considering the time lag between SIF and sap flux (0–0.5 h). Photosynthetically active radiation was the primary factor driving the SIF and sap flux density relationship. Soil moisture, vapor pressure deficit, and air temperature also influenced this relationship on daily and monthly scales. Compared to SIF-based linear regression models, the SIF-based random forest model performed better in tracking the seasonal sap flux density. The results suggest the feasibility of accurately monitoring vegetation sap flux using SIF, woody fractional vegetation cover, and environmental factors in a temperate savanna. This method could also be used in modeling land surface processes in savanna-type ecosystems.

Key words: Temperate savanna, Ulmus pumila., Far-red SIF, Sap flux relationships, Environmental factors