Fig. 4

Linear regression (a, c, d) of total daily photosynthetically active radiation (PAR, μmol∙m^− 2∙d^− 1) and average sap flow density (J_s, g∙m^− 2∙s^− 1) in the daytime, and linear regression (b, d, f) of residuals plotted against soil water content (SWC, %) in old (a, b), mature (c, d), young (e, f) in ambient (A) and throughfall reduction treatment (TRT). The significance level (p < 0.001) is marked by Asterisks (*) for these significant fitting model. Different letters behind r-square value denote the significant difference in the slopes of the regression lines between ambient and drought (ANCOVA, (a) F (1, 22) = 59.339, p < 0.001; (c) F (1, 22) = 10.343, p = 0.001; (e) F (1, 34) = 73.831, p < 0.001). Data only in sunny days during growing seasons in 2017 and 2018 were included (n = 211)