**3. Results**

#### *3.1. Interspecific Association among Species*

The overall species association measured by the VR test revealed no significant correlation within the forest community in either 2012 or 2015 (in 2012, VR = 1.10 > 1, W = 569.10, χ2 (0.95, N) = 569.95, χ2 (0.05, N) = 464.32; in 2015, VR = 1.03 > 1, W = 604.23, χ2 (0.95, N) = 646.57, χ2 (0.05, N) = 533.71).

The association measurements between each pair of species revealed generally negative correlations between *P. davidiana* and co-occurring species (Table 1). In 2012, there was no significant correlation between *P. davidiana* and *T. mongolica*, while the relationship between *P. davidiana* and *B. dahurica*, *B. platyphylla* and *Q. mongolica* were all negative. In 2015, the relationship between *P. davidiana* and all the co-occurring species became negatively correlated.


**Table 1.** Yates correlations and Ochiai indices showing interspecific associations among main species in the forest in 2012 and 2015.

The interspecies association among *B. dahurica*, *B. platyphylla* and *Q. mongolica* in both years were strong and significantly positive (V > 0, χ2 > 6.635 for all associations except the one between *B. platyphylla* and *Q. mongolica* in 2015; OI > 0.6), showing the distribution of the three species tend to be convergen<sup>t</sup> (Table 1).

A strong increase of *T. mongolica* could be detected from 2012 to 2015. The total number of *T. mongolica* in plot increased from three in 2012 to 486 in 2015 (Figure 1e), and the percent of quadrats with *T. mongolica* increased from 0.2% in 2012 to 12.0% in 2015. Consequently, the association between *T. mongolica* and co-occurring species had prominently been increased (Table 1). In 2012, none of the association between *T. mongolica* and other species were significant (χ<sup>2</sup> < 3.841 in Yates correlation; OI < 0.1), while in 2015, *T. mongolica* showed extremely significant positive correlation with *B. dahurica* and negative correlation with *P. davidiana* (χ<sup>2</sup> > 6.635). For Ochiai index, though the association between *T. mongolica* and other species were both weak, OI value in 2015 were generally higher than OI in 2012.

Principle component analysis (PCA) integrated the interspecific associations within the forest in these two years (Figure 2). The first two components reflected 52.53% and 52.42% of the total variabilities in 2012 and 2015, respectively. *B. platyphylla* and *B. dahurica* located nearby the first component at the opposite of *P. davidiana* in both years. *Q. mongolica* located close to the first component in 2012, while close to the second component in 2015, indicating *Q. mongolica* being increasing decoupled with *P. davidiana*, *B. platyphylla* and *B. dahurica* over the years. The location of *T. mongolica* in 2015 was inversed to it in 2012 and be closer to the first component than 2012, indicating a closer association emerged over the years. Quadrats at the bottom of the mountain located generally at the top–left in PCA figures, while quadrats at the top of the mountain usually distributed at the bottom–right in figures, indicating *P. davidiana* to occupy more proportion at the bottom of the mountain, while *T. mongolica* and *B. platyphylla* occupying more proportion at the top of the mountain.

**Figure 2.** Principle component analysis (PCA) showing the interspecific associations among main species in the forest in 2012 and 2015. Arrows show the load of each species on each principle components. Dots indicate the PCA scores for the quadrats. Color of the dots represents the location of the quadrats along the altitude.

#### *3.2. Hydraulic Safety and E*ffi*ciency Di*ff*erences among Species*

There were significant differences in Kl between species (Figure 3). The Kl value of *B. dahurica*, *B. platyphylla*, *T. mongolica* and *Q. mongolica* increased significantly in sequence (*p* < 0.05), ranging from 1.1 × 10−<sup>4</sup> to 11.2 × 10−<sup>4</sup> kg m MPa−<sup>1</sup> s<sup>−</sup>1. There were no significant differences between the Kl value of *P. davidiana* and *B. dahurica*, nor between *P. davidiana* and *B. platyphylla*.

**Figure 3.** Leaf-specific hydraulic conductivity (Kl) of the co-occurring species are shown in the bar figure. Error bars show ± 1 SE. Different characters show the significant differences according to a variance analysis (*p* < 0.05).

The stem-vulnerability curves showed generally convergen<sup>t</sup> sensitivity to early stages of drought, but different stem cavitation resistance abilities among co-occurring tree species (Figure 4). All the species except *Q. mongolica* had the same P12. P12 of *Q. mongolica*, −1.24 MPa, was significantly lower than *B. platyphylla*, *B. dahurica* and *T. mongolica*, indicating *Q. mongolica* to be more sensitive to the early stages of drought. Though some species pairs did not show significant differences, the P50 and P88 values indicated the stem cavitation resistance of species to be generally ordered as: *T. mongolica* = *B. platyphylla* > *P. davidiana* > *B. dahurica* > *Q. mongolica*.

**Figure 4.** (**a**) Comparison of the water potential of stems at 12%, 50% and 88% loss of stem conductivity (P12, P50 and P88, respectively) of each species. The different characters indicate the significant differences among the thresholds of each species based on variance analysis; (**b**–**f**) Vulnerability curves are shown as the dynamic of percent loss of stem hydraulic conductivity in response to centrifugal-force xylem tension for each tree species. Gray curves are fitted by each sample, while the black curves are fitted by the mean values. For each species, P12, P50 and P88 are shown as the single-dashed, solid and double-dashed vertical lines, respectively.

The interspecific differences on hydraulic strategies between safety and efficiency can be integrated as Figure 5 when representing xylem hydraulic safety by P50. Generally speaking, the hydraulic strategies of different species were divergent. Hydraulic strategy of *P. davidiana* has much overlap with *B. platyphylla* and *B. dahurica*, though the strategies of the latter two were different. *Q. mongolica* located at the right bottom corner, suggesting the xylem have high hydraulic conductance but poor ability for drought resistance. *T. mongolica* lied in the right top corner of figure, suggesting *T. mongolica* have both high hydraulic conductance and high threshold for cavitation. The same pattern could be found if the xylem hydraulic safety were represented by P88 (Figure S1).

**Figure 5.** Relationship between stem hydraulic transportation efficiency (as measured by leaf-specific hydraulic conductivity, Kl) and safety (as measured by the water potential of stems at 50% loss of stem conductivity, P50) of all species. Error bars show ± 1 SE.

#### *3.3. The Relationship between Interspecific Association and Species Hydraulic Strategy*

The correlation between interspecific association indices and distances between species pairs calculated by hydraulic traits, Kl and P50, were generally positive, but not significant. The one between Canberra distance and V value in Yates correlation in 2015 was significant (Figure S2). By linear regression, V2015 could be predicted by Canberra distance with the model

V2015 = −19,283 + 28,794 × Canberra distance (Figure 6, *p* = 0.05, R<sup>2</sup> = 0.39). About 40% of the variance of interspecific association could be explained by the interspecific differences on hydraulic strategies in the forest we studied (Figure 6).

**Figure 6.** Relationship between hydraulic differences of each pair of species and their interspecific associations, represented by Canberra distance and V2015 in Yates correlation tests, respectively. Blue line shows the linear model result, while the gray region shows the confidence interval.
