Time-based shifts in xylem vulnerability curves of angiosperms based on the flow-centrifuge method

Centrifuges provide a fast and standard approach to quantify embolism resistance of xylem in vulnerability curves (VCs). Traditionally, embolism formation in centrifuge experiments is assumingly driven by centrifuge speed, and thus pressure, but unaffected by spin time. Here, we explore to what extent embolism resistance is not only pressure but also spin time dependent, and hypothesise that time-stable hydraulic conductivity (K h ) values could shift VCs. We quantified time-based shifts in flow- centrifuge VCs and their parameter estimations for six angiosperm species by measuring K h at regular intervals over 15 minutes of spinning at a particular speed before a higher speed was applied to the same sample. We compared various VCs per sample based on cumulative spin time, and modelled the relationship between K h , xylem water potential (Ψ), and spin time. Time-based changes of K h showed considerable increases and decreases at low and high centrifuge speeds, respectively, which generally shifted VCs towards more positive Ψ values. Values corresponding to 50% loss of hydraulic conductivity ( P 50 ) increased up to 0.72 MPa in Acer pseudoplatanus , and on average by 8.5% for all six species compared to VCs that did not consider spin time. By employing an asymptotic exponential model, we estimated time-stable K h , which improved the statistical significance of VCs in 5 of the 6 species studied. This model also revealed the instability of VCs at short spin times, and showed that embolism formation in flow-centrifuges followed a saturating exponential growth curve. Although pressure remains the major determinant of embolism formation, spin time should be considered in flow- centrifuge VCs to avoid overestimation of embolism resistance. This spin-time artefact is species- specific, and likely based on relatively slow gas diffusion associated with embolism spreading. It can be minimized by determining time-stable K h values for each centrifuge speed, without considerably extending the experimental time to construct VCs.