Water source contribution
From the natural abundance of δ2H and δ18O in plant xylem and soil water, we used a Bayesian stable isotope mixing model to quantify the contribution of potential tree water sources in each species in the monospecific and four-species mixtures for each sampling date. As the rainwater isotope ratio differs throughout the season (i.e., isotopically more depleted rainwater in winter compared to summer) and the water evaporation decreases with soil depth, each water source has a significantly different stable isotopic composition (Fig. S3). These distinct soil isotopic profiles allow us to determine the contribution of each water source to the tree xylem water under the assumption that there is no isotopic fractionation during water uptake by the roots (Dawson & Ehleringer, 1991). Therefore, the natural isotopic abundance of xylem sap should reflect the water sources used by the plant. We used the package simmr in R (Parnell, 2019), where the isotopic composition (δ 18O and δ2H) for each potential source (i.e., 0-10 cm, 10-20 cm, 20-30 cm, 30-40 cm, rainwater) and each target tree were assigned into the model. We set the TEF (trophic enrichment factor) and the concentration dependence to 0 due to the absence of isotopic fractionation by the roots. We ran the model where 3600 iterations out of 10000 runs were produced over 4 Markov chain Monte Carlo (MCMC) for the isotopic values from each plant with the isotopic values from the soil water source of the corresponding plot in each date (Sun et al., 2022). To increase the clarity of presentation, the contributions from the water sources were grouped a posteriori into three layers: shallow (i.e., 0–20 cm), deep (i.e., 20–40 cm), and water stored in the fractured bedrock (i.e., rainwater). Indeed, the winter precipitation that penetrates deep soil layers and bedrock cracks could be a substantial water source for trees in Mediterranean forests during summer droughts (Eliades et al., 2018). As the bedrock water is not subjected to evaporation (Ehleringer & Dawson, 1992), we used the precipitation collected at our site as a proxy, similar to Grossiord et al., (2017). The cumulative rainwater collected during the winter of 2021 (i.e., January-April) and 2022 (i.e., October-April) was considered bedrock water source for the May campaigns of 2021 and 2022, respectively. During the remaining growing season, the stable isotope composition of the precipitation until the sampling date was added to the winter precipitation (e.g., winter, May, and June for the campaigns in July).