Climate-growth relationships at different altitudes and ages
The climate variation characteristics in Huize climate station from 1953 to 2016 are shown in Fig. 2. Temperature significantly increased at a rate of 0.169 °C/10a (R2 = 0.321, p < 0.001), annual total precipitation showed decreasing trend at a rate of 13.26 mm/10a, although not to a significant level (p > 0.05).
The correlation coefficients between chronologies and seasonal climate factors, and the climate factors selected into the simplified regression model and their explanatory rates for the radial growth of P. yunnanensis varied with altitudes (Fig.3a and Table S1). Trees living at the altitude of 1838 m negatively response to temperature in winter, spring, summer and growing season (p < 0.01), while at the altitude of 2520 m trees showed positive correlations with temperature in winter (p < 0.01), spring (p < 0.05), summer (p < 0.05) and growing season (p < 0.01). The explanatory rate of temperature on radial growth of P. yunnanensis at the altitude of 1838 m and 2520 m was 23.6% and 59.7%, respectively. Trees living at the lower altitude (1838 m and 2010 m) positively response to precipitation in current spring (p < 0.01) and growing season (p < 0.05), the explanatory rate of precipitation on growth was 39.6% and 36.4%, respectively. At the lowest altitude, trees showed significant positive correlations with RH in all seasons, whereas trees showed significant negative correlations with RH at the higher altitude (2369 m and 2520 m). The explanatory rate of RH on radial growth at the altitude of 1838 m, 2010 m, 2369 m, and 2520 m was 12.2%, 23.5%, 40.9%, and 31.7%, respectively. Trees showed positive correlations with PDSI at the lowest altitude (1838m), whereas showed negative correlations with PDSI at the highest altitude (2520m). The explanatory rate of temperature factors on radial growth of P. yunnanensis at the altitude of 1838 m, 2010 m, 2369 m, and 2520 m was 23.6%, 33.3%, 39.0%, and 59.7%, respectively; the explanatory rate of moisture factors on radial growth of P. yunnanensis at the altitude of 1838 m, 2010 m, 2369 m, and 2520 m was 76.4%, 66.7%, 61.0%, and 40.3%, respectively.
The correlation coefficients between chronologies and seasonal climate factors, and the climate factors selected into the simplified regression model and their explanatory rates for the radial growth of P. yunnanensis varied with ages (Fig.3b and Table S2). Radial growth of AC1 and AC4 age trees exhibited positive correlations with precipitation in past summer (p < 0.05), past autumn (p < 0.01), and current spring (p < 0.01). The explanatory rate of precipitation on radial growth of AC1, AC2, AC3, and AC4 age class trees was 8.6%, 14.7%, 34.1%, and 23.5%, respectively. Radial growth of AC1 age class trees showed positive correlations with RH from past summer to current spring (p <0.05), the AC4 age class trees showed positive correlations with RH in past summer (p <0.05), past autumn (p <0.05), and current spring (p <0.01). The explanatory rate of RH on radial growth of AC1, AC2, and AC4 age trees was 31.7%, 63.7%, and 45.8%, respectively. Radial growth of AC4 age trees showed positive correlation with PDSI in past summer (p < 0.05), the explanatory rate of PDSI on growth was up to 21.7%. The explanatory rate of temperature factors on radial growth of P. yunnanensis of AC1, AC2, AC3, and AC4 age trees was 59.7%, 14.6%, 65.9%, and 9.1%, respectively; the explanatory rate of moisture factors on radial growth of P. yunnanensis of AC1, AC2, AC3, and AC4 was 40.3%, 85.4%, 34.1%, and 90.9%, respectively.
The RDA of tree radial growth with climate factors at different altitudes and age classes is shown in Fig. 4. Of the 68 climate variables, 7 variables showed significant effects on radial growth of P. yunnanensis at different altitudes (Fig. 4a), and 8 variables showed significant effects on radial growth of different age classes (Fig. 4b). Trees living at the lower altitudes (1838 m and 2010 m) significantly positively correlated with precipitation in May of the current year, and the trees living at the altitude of higher altitudes (2369 m and 2520 m) significantly positively correlated with temperature in November of the past year. The precipitation in February of the current year (cP2), the PDSI in June of the past (pPDSI6) and current (cPDSI6) year, the RH in October of the past year (pRH10) and in July of the current year (cRH7) showed significantly positive correlations with radial growth of trees at higher altitudes (2369 m and 2520 m), whereas showed significantly negative correlations with radial growth at lowest altitude (1838 m). Similarly, the young (AC1), middle-age (AC2), and mature (AC4) trees showed significantly negative correlation with PDSI in May of the current year (cPDSI5), and the near-mature (AC3) trees showed significant positive correlation with temperature in November of the past year (pT11). The precipitation in August (pP8) of the past year, the RH in October of the past year (pRH10) and in May of the current year (cRH5) showed significantly positive correlations with young (AC1), middle-age (AC2), and mature (AC4) trees, whereas showed significantly negative correlation with near-mature (AC4) trees.
Radial growth response of P. yunnanensis to drought
We defined the drought events based on the PDSI during January to May. The four strongest drought events were identified as occurring in 2010, 2012, 2013, and 2014, and were used in the analysis of tree resilience to droughts. Among these, 2012 was defined as a moderate drought year, whereas 2010, 2013, and 2014 were defined as the severe drought years (Fig. 5).
The Rt, Rc, and Rs of P. yunnanensis to drought varied among different altitudes (Fig. 6a) (Table S3; Table S4) and age classes (Fig. 6b) (Table S3; Table S5). The Rt of trees at lower altitudes was stronger than that at higher altitudes in the 2010 and 2014 drought events, and the Rc of trees at the lower altitudes were stronger than that at the higher altitudes in 2012 and 2013 drought events; the Rs of trees at low altitude (1838 m) was significantly stronger than that at higher altitudes (2010, 2369, and 2520 m), the Rs significantly decreased with altitudes (p<0.001) in the 2013 and 2014 drought events.
The Rt of trees enhanced with age except the AC4 age class in 2012 drought events, whereas it decreased with age except the AC4 age class in the 2014 drought event. The Rc was strongest for AC1 age class trees and weakest for AC3 age class trees. The Rs for the AC3 age class trees was lower than those of other age classes during the 2013 and 2014 drought events.
Temporal change in tree growth resilience to drought
We evaluated the temporal change in resilience of all sampled trees (Fig.7). The results revealed that the Rt, Rc, and Rs of the sampled trees varied significantly in different drought events (p<0.001). The Rt of the trees increased significantly from 2010 to 2013 drought events, whereas decreased significantly in 2014 drought event; the Rc showed decreasing trend from 2010 to 2013 drought event, whereas increased in 2014 drought event, and the variation trend is opposite to that of Rt; the Rs of trees in 2014 drought event significantly enhanced than 2010 and 2012 drought events.
We analyzed the relationship between Rc and Rt of trees in the different altitudes and age classes (Fig. 8). The results revealed that the Rc and Rt of trees showed significant (p < 0.05) negative relationships in each altitude and age class.
Correlations of resistance, recovery and resilience with average pre-drought growth of trees
We tested the relationships of Rt, Rc, and Rs of P. yunnanensis with the average pre-drought growth at different altitudes and age classes (Fig. 9). The Rt of trees was significant negatively correlated with average pre-drought growth (p < 0.01). The relationship between Rt and average pre-drought growth at the higher altitudes was stronger than those at lower altitudes. Similarly, the Rt of the AC3 (Slope = −0.710, p < 0.0001) and AC4 age class trees (Slope = −0.702, p < 0.0001) showed stronger relationships with average pre-drought growth than the AC2 and AC1 age class trees.
The Rc of trees was negatively associated with average pre-drought growth and varied significantly across the four altitudes and age classes. The Rc of trees at the altitudes of 2010 m (Slope = −0.186, p < 0.001) and 2369 m (Slope = −0.548, p < 0.05) showed significant negative correlations with average-pre-drought growth, whereas the trees at 1838 m and 2520 m altitude showed no significant correlations (p > 0.05) with average pre-drought growth. The Rc of the AC1 (p < 0.01), AC2 (p < 0.001), and AC4 (p < 0.01) age class trees showed significant correlations with average pre-drought growth, whereas the AC3 age class trees showed no significant correlation (p > 0.05).
The Rs was negatively significantly correlated with average pre-drought growth for all altitudes and age classes (p < 0.001). The relationship between Rs and average pre-drought growth at the highest altitude 2520 m (Slope = −0.356, p < 0.001) was weaker than at the lower altitudes (1838 m, 2010 m, 2369 m). The Rs of the AC1 age class trees showed the strongest relationship with average pre-drought growth (Slope = −2.410, p < 0.0001).