Spatial Variation and Temporal Instability in the Growth/Climate Relationship of White Birch (Betula platyphylla Suk) in the Changbai Mountain, China

Tree growth in mountain ecosystems is affected by complex environments, and its relationship with climatic and environmental factors varying with elevation. In order to examine the spatial variation and temporal stability of the growth/climate relationship of Betula platyphylla (BP), the dendrochronological method was used to analyze the radial growth/climate relationship between 1946 and 2016 of the BP trees along an altitudinal gradient in the Changbai Mountain of northeast China. Our results showed that the mean sensitivity of BP was higher than that of other species studied in Changbai Mountain. The growth/climate relationship of BP trees varies with altitude, and this conclusion has reached a consensus from the study of tree growth response to climate change. More specifically, at low altitudes (550–995 m a.s.l.), the radial growth of BP is mainly affected by spring precipitation and temperature in May and October of the current year. However, at high-altitude areas (1210–1425 m a.s.l.), it is mainly affected by the temperature in September of the previous year and May of the current year. Furthermore, the growth/climate relationship of BP trees showed temporal instability. After 1970, the rise in temperature inhibited the growth of BP at low altitudes and promoted the growth of BP trees at high altitudes. In the context of continued warming in the future, the white birch stands in Changbai Mountain will move to higher altitudes.


Introduction
The rising global temperature has a major influence on high-altitude mountain ecosystems of the continental interior of the mid-high latitudes in the Northern Hemisphere [1], causing significant changes in species phenology, biomass, diversity, distribution, and ecosystem processes in this region [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. There is increasing evidence that since the 1970s, the northeast China region has experienced rapid warming, leading to a warming and drying trend in this region [17][18][19][20]. The Northeast Forest is one of the three major state-owned forest areas in China, which is extremely sensitive to climatic and environmental changes [21][22][23][24], and the relationship of its dynamics with climatic factors has been receiving more and more attention.
The increasing temperature is affecting the growth of trees in Northeast China [22,25]. The results from models indicated that the distribution of dominant conifer species in northeast China will tend to move northward and to higher altitudes, and the forest structure and composition will change under the coming climate warming and drying scenarios [25][26][27]. Meanwhile, based on empirical/observational evidences, dendrochronological studies by Jiang et al. (2016) [8] suggested that the radial growth of Larix gmelinii was decreasing under the observed temperature increases in the Greater Khingan Mountains of northeast

Study Area and Sample Collection
The study area is located at the Changbai Mountain (CBM) Natural Reserve in northeast China (41°31′-42°28′ N, 127°9′-128°55′ E) (Figure 1), where the climate is affected by the temperate continental monsoon, which is characterized by moist summers and cold, windy winters [44]. The annual average temperature (Ta) is 2.14 (±0.75) °C, and the annual precipitation (P) is 833 (±114) mm from 1958 to 2016 ( Figure 2). About 88.4% of annual precipitation occurs from April to September. The sampling was conducted in November 2016. Five sampling sites were chosen to represent the elevation range of BP: the lowelevation sites (E550 at 550 m a.s.l.; E760 at 760 m a.s.l.; and E995 at 995 m a.s.l.), and the high elevation sites (E1210 at 1210 m a.s.l.; and E1425 at 1425 m a.s.l.).  The sampling sites are all on the north slope and are covered by similar weather patterns. The P is from 767 mm at E550 to 886 mm at E1425; the Ta is from 3.8 • C at E550 to 0.4 • C at E1425 m. The low-altitude sampling point was located at the lowest edge of the species' distribution, while the high-altitude sampling point was located on the upper tree line. The vertical distance between the lowest edge and upper tree line was about 925 m. All the selected forests were natural secondary forests, which can be regarded as pure birch forests, because there were only some pine seedlings. In order to minimize the impact of non-climatic factors on the growth of trees, the selected BP forests were natural and open forests with no signs of recent fires or human disturbances. According to the prevalence of white birch, two or three plots of 30 m × 30 m were established for each sampling site. All of the healthy, presumably oldest and largest canopy trees (with a height of 20-27 m and a diameter of 19-32 cm) were chosen and then one or two oppositely oriented increment cores were sampled at the trunk 1.3 m above the ground. In some cases, when the core was partially fragmented or rotted, only one core was taken. At each site, 22-33 trees were selected; thus, in total, 272 cores from 142 BP living trees were collected. The slope of the sampling points was from 0 • to 15 • . The sampling sites are all on the north slope and are covered by similar weather patterns. The P is from 767 mm at E550 to 886 mm at E1425; the Ta is from 3.8 °C at E550 to 0.4 °C at E1425 m. The low-altitude sampling point was located at the lowest edge of the species' distribution, while the high-altitude sampling point was located on the upper tree line. The vertical distance between the lowest edge and upper tree line was about 925 m. All the selected forests were natural secondary forests, which can be regarded as pure birch forests, because there were only some pine seedlings. In order to minimize the impact of non-climatic factors on the growth of trees, the selected BP forests were natural and open forests with no signs of recent fires or human disturbances. According to the prevalence of white birch, two or three plots of 30 m × 30 m were established for each sampling site. All of the healthy, presumably oldest and largest canopy trees (with a height of 20-27 m and a diameter of 19-32 cm) were chosen and then one or two oppositely oriented increment cores were sampled at the trunk 1.3 m above the ground. In some cases, when the core was partially fragmented or rotted, only one core was taken. At each site, 22-33 trees were selected; thus, in total, 272 cores from 142 BP living trees were collected. The slope of the sampling points was from 0° to 15°.

Development of Ring-Width Chronologies
In the laboratory, the samples were pretreated, naturally air dried, glued to wooden holders, and then sanded with successively finer grits of sandpaper to highlight the tree rings [45,46]. The tree rings were cross-dated under a binocular microscope, and then the annual ring width was measured using a Velmex measuring system interfaced with the 'Time Series Analysis Program' (TSAP; Frank Rinntech, Heidelberg, Germany) with an accuracy of 0.001 mm. The COFECHA program was used in the quality control of the cross-dating and the measurements [45]. To eliminate the effects of stand dynamics, age, and any other non-climate-related growth variation, the cross-dated tree-ring data were detrended using two different techniques-negative exponential (EXP) and regional curve standardization (RCS), via ARSTAN [46]. Each technique produced three chronologies in the program ARSTAN: autoregressive (ARS), residual (RES), and standard (STD).
The STD chronologies in both techniques contain the most climate signals, while the STD chronology with an EXP detrending was the most suitable chronology because it contained more low frequency signals ( Figure 3). Analyses were restricted to the period with an expressed population signal (EPS) greater than 0.85 [46,47]. Thus, according to Table 1, site E760 has an EPS value greater than 0.85 from 1951, while the other four sites have EPS values greater than 0.85 from 1946. The period from 1946 to 2016 was used for further analysis. In addition, in order to test if the BP trees were sensitive to highfrequencies climate variations, correlations between the RES chronologies and climate data were also carried out.    in Changbai Mountain area. The meteorological data provided by these three weather stations are all after 1958. There is a significant correlation between the grid data and monthly climate variables at the three stations from 1958 and 2016 (r = 0.90, p < 0.001). An upward or downward trend was detected in the climate data ( Figure 3) and was removed through linear regression; after that, the Pearson correlations between the detrended climate data and STD/RES chronologies between 1946 and 2016 were analyzed. An analysis of the climate dynamics from 1946 and 2016 showed that, since 1970, the monthly mean temperatures have risen rapidly, and the severity of drought has gradually increased ( Figure 4). To verify whether warming has an effect on the relationship between tree growth and climatic factors, correlations between the radial growth and climate factors before and after 1970 were conducted.
Correlations between the tree-ring chronologies and monthly climatic variables from the previous year's September (the mean temperature > 5 • C) to the current year's October (the mean temperature > 0 • C) were carried out, and five climate variables were applied for the analyses, including monthly maximum temperature (Tmax), monthly minimum temperature (Tmin), monthly mean temperature (T), monthly total precipitation (Pre), and monthly Palmer Drought Severity Index (PDSI). In addition, in order to explore if the growth of BP could be linked to seasonal climate conditions, the correlations between seasonal climate data and radial growth was also calculated. According to the T and the Tmin of the study region, four seasons were delineated: (a) the winter dormant period (WD) (T < 0 • C), during the previous year's November and current year's March; (b) the Correlations between the tree-ring chronologies and monthly climatic variables from the previous year's September (the mean temperature > 5 °C) to the current year's October (the mean temperature > 0 °C) were carried out, and five climate variables were applied for the analyses, including monthly maximum temperature (Tmax), monthly minimum temperature (Tmin), monthly mean temperature (T), monthly total precipitation (Pre),

Standard Chronologies' Characteristics
The five final standard chronologies with an EPS greater than 0.85 between 550 m a.s.l. and 1425 m a.s.l. is displayed in Figure 4 Table 1). The mean ring width ranged between 1.03 and 1.58 mm, which showed a significant negative correlation with elevation ( Figure 5a). RBAR is the mean inter-series correlation [45], ranging from 0.545 to 0.613. The standard deviation (SD) of the tree-ring widths in the five sites ranged from 0.14 to 0.22, whereas the mean sensitivity (MS), describing the interannual variation in the ring widths, ranged from 0.281 to 0.377 (Table 1). Significant first-order autocorrelations (AC1) were showed in all tree-ring series (Table 1), ranging from 0.482 to 0.756, indicating a significant correlation between the current-year ring width and previous year's ring width [28].
growth of BP could be linked to seasonal climate conditions, the correlations between seasonal climate data and radial growth was also calculated. According to the T and the Tmin of the study region, four seasons were delineated: (a) the winter dormant period (WD) (T < 0 °C), during the previous year's November and current year's March; (b) the early growing season (BG) (Tmin < 5 °C), during April and May; (c) the growing season (GS), from June to August; and (d) the end of the growing season (EG) (Tmin < 5 °C), from the current year September to current year October.

Standard Chronologies' Characteristics
The five final standard chronologies with an EPS greater than 0.85 between 550 m a.s.l. and 1425 m a.s.l. is displayed in Figure 4 Table 1). The mean ring width ranged between 1.03 and 1.58 mm, which showed a significant negative correlation with elevation ( Figure 5a). RBAR is the mean inter-series correlation [45], ranging from 0.545 to 0.613. The standard deviation (SD) of the tree-ring widths in the five sites ranged from 0.14 to 0.22, whereas the mean sensitivity (MS), describing the interannual variation in the ring widths, ranged from 0.281 to 0.377 (Table 1). Significant first-order autocorrelations (AC1) were showed in all treering series (Table 1), ranging from 0.482 to 0.756, indicating a significant correlation between the current-year ring width and previous year's ring width [28].

Climate-Growth Relationship
Correlation analysis showed that the relationships between the radial growth and climatic factors of BP trees were complicated and varied with altitude and time periods ( Figure 6). During the period from 1946 to 2016, the radial growth of BP had a significant positive correlation with the monthly mean temperature (T) in May at all five locations, while the radial growth of BP at the locations of E550, E760, E995/E1210, and E1425 had a significant positive correlation with the monthly mean maximum temperature (Tmax)/the Forests 2021, 12, 589 9 of 16 monthly mean minimum temperature (Tmin) in May (Figure 6a). Meanwhile, significant positive correlations with T and Tmin were found for the previous year's (the previous year was abbreviated as t-1) September at E1210 and E1425, and significant positive correlations with T and Tmax occurred for October at E550 and E760 (Figure 6a). In addition, significant negative correlations with monthly total precipitation (Pre) occurred for March at E550, E760, and E995 (Figure 6a). BP tree growth has no significant correlation with PDSI ( Figure 6a).
The climate/growth relationships between 1946 and 1969 were significantly different from that during 1946-2016 (Figure 6a,b). During 1946-1969, the growth of BP trees was not sensitive to temperature changes at sites E1210 and E1425; significant correlations were only found at site E1425 for Pre and PDSI in September (t-1) and October (t) (Figure 6b). The BP trees' radial growth was significantly impacted by monthly mean temperatures at site E550 and E760, and significant positive correlations were found during March-August (Figure 6b). Meanwhile, the significant positive correlations occurred for Tmax at site E550, for T, Tmin and Tmax at site E760 and for Tmin and T at site E995 in September (t-1) (Figure 6b). Significant positive correlations were also found for monthly mean temperatures in June at site E995 (Figure 6b). In addition, the radial growth at site E550 and E760 was significantly affected by Pre, with significant negative/positive correlations occurring for March(t)/November (t-1) (Figure 6b). Pre and PDSI in September (t-1) and October (t) had a significant positive effect on the tree growth of BP at site E1425 (Figure 6b).

Climate-Growth Relationship
Correlation analysis showed that the relationships between the radial growth and climatic factors of BP trees were complicated and varied with altitude and time periods ( Figure 6). During the period from 1946 to 2016, the radial growth of BP had a significant positive correlation with the monthly mean temperature (T) in May at all five locations, while the radial growth of BP at the locations of E550, E760, E995/E1210, and E1425 had a significant positive correlation with the monthly mean maximum temperature (Tmax)/the monthly mean minimum temperature (Tmin) in May (Figure 6a). Meanwhile, significant positive correlations with T and Tmin were found for the previous year's (the previous year was abbreviated as t-1) September at E1210 and E1425, and significant positive correlations with T and Tmax occurred for October at E550 and E760 (Figure 6a). In addition, significant negative correlations with monthly total precipitation (Pre) occurred for March at E550, E760, and E995 (Figure 6a). BP tree growth has no significant correlation with PDSI ( Figure 6a).  The climate/growth relationships between 1946 and 1969 were significantly different from that during 1946-2016 (Figure 6a,b). During 1946-1969, the growth of BP trees was not sensitive to temperature changes at sites E1210 and E1425; significant correlations were only found at site E1425 for Pre and PDSI in September (t-1) and October (t) ( Figure  6b). The BP trees' radial growth was significantly impacted by monthly mean tempera-  (Figure 6). During the periods of 1970-2016, the temperature had a more pronounced effect on the growth of BP trees of E1210 and E1425 (Figure 6). At site E1210, significant positive correlations were found with monthly mean temperatures in September (t-1), Tmin and T in May, Tmin in June, and Tmin and T during September (t) and October (t), while at site E1425, significant positive correlations occurred for T in September (t-1), monthly mean temperatures in May, Tmin in June, T and Tmax in September (t), and monthly mean temperatures in October (t) (Figure 6c). The growth of BP trees is not sensitive to climate change from September (t-1) to April at sites E550, E760, and E995 (Figure 6c). Meanwhile, compared with 1946Meanwhile, compared with -1969Meanwhile, compared with and 1946Meanwhile, compared with -2016, the sensitivity of the radial growth of BP trees to temperature variations in May, September, and October was increased during the period 1970-2016 in these three sites ( Figure 6). Significant positive correlations were found with T and Tmax in May for all the tree sites, Tmin and T in September (t) for E760, monthly mean temperatures in October(t) for E550, Tmax in October (t) for E760, and T and Tmax in October (t) for E995 (Figure 6c). In addition, contrary to 1946-1969, the temperatures in June to August during 1970-2016 displayed a negative effect on the radial growth. Significant correlations were observed for monthly mean temperatures in June and July at E550 and E760, while at E995, significant correlations were found for Tmin and T in June, monthly mean temperatures in July, and T and Tmax in August (Figure 6b,c). The Pre also had a significant impact on tree growth during 1970 and 2016, and significant positive correlations were observed in April at E1210 and E1425 and negative correlations were observed in May at E550 and E760 (Figure 6c).
During 1946 and 2016, the chronologies had significant positive correlations with T and Tmax for the early growing season (BG) at E550 and E760, and also had significant positive correlations with Tmax at E550 and monthly mean temperatures at E760 for the end of the growing season (EG) (Figure 7a). During 1946 and 1969, there were no significant climate/growth correlations observed at E995-E1425 (Figure 7b). At sites E550 and E760, BP growth had significant positive correlations with Tmax for the winter dormant period (WD) and with temperatures from the early growing season (BG) to the growing season (GS). Meanwhile, the ring-width indices at E550 had a significant positive correlation with Tmax in EG (Figure 7b). After 1970, the climate-growth relationships were obviously different, as the effects of BG and EG temperature on radial growth increased (Figure 7c). The radial growth was significantly positively correlated with temperatures in BG and EG at E550 and E760, with temperatures in EG at E1210 and E1425, with T and Tmax in BG and EG at E990, and also with Tmin in BG at E1210 and E1425 (Figure 7c). In addition, after 1970, the radial growth had a significant negative correlation with monthly mean temperatures in GS at E550-E995 (Figure 7c).

Chronology Evaluation
Statistical analyses indicated a quite high quality for the STD chronologies ( Table 1). The differences between altitudes were well represented in the chronology statistics. The values of the mean inter-series correlation and expressed population signal showed a certain amount of coherence between the series of the chronologies. The values of autocorrelation suggested that the influence of climatic factors in the previous year on the growth of BP trees is gradually strengthened with an increase in altitude. The average tree-ring width of white birch decreases linearly (R 2 = 0.51) with altitude ( Figure 5a). Our findings are consistent with previous studies of other species (e.g., Abies nephrolepis, Pinus koraiensis, Larix olgensis, and Picea koraiensis) in Changbai Mountain [21,33]. The value of MS increased linearly with altitude (R 2 = 0.99) (Figure 5b), indicating that the variation in BP radial growth is greater at higher elevations. Our results showed that the MS (the value: 0.28-0.31) of BP was higher than that of other species studied in Changbai Mountain, such as Betula ermanii (0.23-0.34) [33,35], Picea koraiensis (0.13-0.17) [33], Larix olgensis

Chronology Evaluation
Statistical analyses indicated a quite high quality for the STD chronologies ( Table 1). The differences between altitudes were well represented in the chronology statistics. The values of the mean inter-series correlation and expressed population signal showed a certain amount of coherence between the series of the chronologies. The values of autocorrelation suggested that the influence of climatic factors in the previous year on the growth of BP trees is gradually strengthened with an increase in altitude. The average tree-ring width of white birch decreases linearly (R 2 = 0.51) with altitude ( Figure 5a). Our findings are consistent with previous studies of other species (e.g., Abies nephrolepis, Pinus koraiensis, Larix olgensis, and Picea koraiensis) in Changbai Mountain [21,33]. The value of MS increased linearly with altitude (R 2 = 0.99) (Figure 5b), indicating that the variation in BP radial growth is greater at higher elevations. Our results showed that the MS (the value: 0.28-0.31) of BP was higher than that of other species studied in Changbai Mountain, such as Betula ermanii (0.23-0.34) [33,35], Picea koraiensis (0.13-0.17) [33], Larix olgensis (0.053-0.29) [32,33,37,49], Pinus koraiensis (0.11-0.20) [21,32,50,51], and Fraxinus mandshurica (0.11) [33].

The Relationship of Climate/Growth Along an Altitudinal Gradient from 1946 to 2016
Our results showed that, at low altitudes, BP has a higher correlation with temperature than precipitation (Figure 6a). At high altitudes (E1210-E1425), the growth of birch trees was significantly correlated with temperature ( Figure 6a). The altitude-dependent growth/climate relationships were also reported in the Betula ermanii [35] and conifer species of CBM [21,[32][33][34]. The current summer temperatures had no significant effect on the growth of BP trees; similar results were also found in Mount Norikura, central Japan [41], unlike the BP trees reported in Kamchatka [41]. At the low-elevation sites (E550-E995), the excessive March and May precipitation inhibited growth (Figure 6a). Adequate snowfall in March and May might meet the water requirements of BP in the early growing season, but it may delay the beginning of the growing season of the BP forest in Changbai Mountain. Many researchers have also found that high precipitation in March-May inhibits tree growth in mountain ecosystems [35,52]. It was found that the current May temperatures has a positive effect on the radial growth of BP at all altitudes ( Figure 6a). Warmer spring temperatures may promote the emergence or germination of leaves earlier, resulting in better growth during prolonged growth seasons [3]. In contrast to our results, Alexander Gradel et al. (2017) [39] found that the growth of white birch (Betula platyphylla Suk.) was negatively correlated with the temperatures in May in Altansumber (49 • 29 07.29" N; 105 • 31 30.36" E), Northern Mongolia, and this relationship may be due to a water deficiency in spring [39]. Our results indicated that the temperature in September of the previous year was also a key factor affecting the radial growth of BP in high-altitude areas (E1210-E1425) (Figure 6a). This result was also observed in the study of the relationship between the growth of Pinus koraiensis Sieb. et Zucc. trees and climatic factors in Changbai Mountain [21]. The warmer conditions at the end of the growing season of the previous year helps the accumulation of more carbohydrates, which is beneficial for the growth of trees in the next year [3]. Our results indicated that the amount of carbohydrate storage in the previous year had a great impact on the growth of birch trees in the current year in the colder environments. In addition, BP stands at low altitudes were mainly responding positively to October temperatures in the current year (Figure 6a). In autumn, warmer conditions may prompt BP to continue photosynthesis [53]. Furthermore, the effects of the current spring and growing season temperatures on the radial growth of trees were complex and site-dependent (Figure 7a). At lower elevations, temperature during the early growing season and at the end of the growing season displayed positive correlations with tree growth, and there is no significant correlation at high altitudes ( Figure 7a). Our results confirmed the following general conclusion: The relationship of climate/growth varies with altitude [52,53]. This agreed well with previous findings in this region [21,33].

Temporal Stability of the Climate-Growth Relationship
The results revealed that long-term sensitivity of BP growth to climate factors changed strongly along an altitudinal gradient, and the growth/climate relationship of BP trees showed temporal instability (Figure 7). This phenomenon has been observed in the study of the response of Pinus koraiensis Sieb. et Zucc. [21] and Larix olgensis Henry [37] growth to climate change at different altitudes in Changbai Mountain. Our results showed that, at high-altitude areas, before 1970 (Figures 6b and 7b), growth was not impacted by the current climate (Figure 6b). However, after 1970, the region has gradually warmed up (Figure 3), and the sensitivity of high-altitude BP trees to temperature has generally increased (Figures 6c and 7c). Tree growth displayed significant positive correlations with temperature in the early growing season (BG) and the end of the growing season (EG) (Figure 7c). Therefore, the warming after 1970 would promote the growth of BP trees in these areas.
At low-altitude areas, compared with the past, after 1970, the impact of precipitation on the growth of birch trees in March was gradually reduced, and the impact of precipitation in May was strengthened (Figure 6c). In addition, before 1970, the growth of trees was significantly positively correlated with the temperature from March to August (t) and from September to November (t-1) (Figure 6b). However, after 1970, as the temperature rose, the number of months in which the growth of birch trees was positively correlated with temperature decreased with time ( Figure 6c). Significant positive correlations between tree growth and temperature were found only in May (t), September (t), and October, while the correlation coefficients in the growing season have become negative (Figure 6c). The growth period of trees in CBM is from June to August. Therefore, as the temperature continues to rise, the growth of BP trees at low altitudes will be inhibited. The correlation results for monthly climate variables were also verified by results related to seasonal climate variables ( Figure 7c).
As mentioned above, under continued warming in the future, the BP stands in Changbai Mountain will move to higher altitudes. Similar results were also obtained by Yu et al. (2013) and   [21,34], who suggested that the radial growth of Korean pine increased at higher elevations, while decreased at low elevations under climate change characterized by warming and drought in Changbai Mountain. In addition, climate change can also cause changes in the distribution of tree species at certain latitudes in northeast China [22,27].

Conclusions
Temperature is a key factor affecting the growth of BP trees in Changbai Mountain. Although the growth of BP trees at all sampling points was significantly positively correlated with the temperature in May, in the period 1946-2016, the growth of birch trees distributed at high altitudes and low altitudes had different response patterns to climate change. Higher precipitation in March (t) and May (t) at low altitudes inhibits radial growth and warmer October (t) temperature accelerates radial growth, while the warmer September (t-1) temperature at high altitudes promotes radial growth. Furthermore, the results revealed that long-term sensitivity of BP growth to climate factors changed strongly along an altitudinal gradient, and the growth/climate relationships of BP trees showed temporal instability. Thus, the rise in temperature will inhibit the radial growth of BP at lower elevations and promote the growth of BP trees at higher elevations, dating from 1970. In the context of global climate change, the results will help simulate the radial growth and distribution dynamics of BP under various environmental conditions.