We quantified the growth dynamics and climatic responses of three tree species that have dominated Changbai Mountain: Korean pine (Pinus koraiensis), Yeddo spruce (Picea jezoensis), and Erman's birch (Betula ermanii). Standardization curves and moving correlations were used to assess growth rate trends and analyze changes in growth-climate relationships of trees at their upper forest boundaries and individual species elevation limits, respectively. Contrasting growth patterns were observed between trees at each upper forest boundary and species-specific upper elevation limits. Korean pines and Yeddo spruces grew faster at their upper forest boundaries than at their individual species limits. A higher growth rate of Erman's birches at their forest upper boundary only occurred before 1960. Relative to the strong effect of temperature on tree growth at individual upper elevation limits, the stable effect of precipitation and changing effect of temperature on tree growth were observed at the upper forest boundaries. Temperature increases have had a significantly negative effect on Korean pine and Erman's birch since 1980, whereas temperature increases were associated with Yeddo spruce growth. This study elucidated the differential growth patterns and temporal changes in climate–growth relationships of these species between their upper forest boundaries and elevation limits.
Introduction
Changbai Mountain, located in northeastern China, has experienced climate warming over the past 100 years but without any accompanying increase in precipitation (Wang et al. 2002). Owing to its varied topography and wide variation in climatic conditions, Changbai Mountain is characterized by a vertical zonation of three forest ecosystems along its altitudinal gradient. The first consists of a mixed broad-leaved and Korean pine forest extending from ca. 740 m to 1100 m a.s.l. This relatively low-altitude forest type is dominated by Korean pine (Pinus koraiensis), Amur linden (Tilia amurensis), Manchurian ash (Fraxinus mandshurica), and other deciduous species. From 1100 to 1700 m a.s.l., a spruce–fir forest, also known as a dark coniferous forest occurs, which is dominated by Yeddo spruce (Picea jezoensis var. komarovii), Korean pine, Manchurian fir (Abies nephrolepis), and Erman's birch (Betula ermanii). Lastly, there is an Erman's birch forest occurring from 1700 m to 2000 m a.s.l., which corresponds with the upper limit of all forests (Wang et al. 1980; Xu et al. 2004). Previously-published models have predicted that climate warming will push the distributions of the mixed broad-leaved and Korean pine forest toward higher elevations (Yan et al. 2000; Cheng and Yan 2008), whereas the average growth of Korean pine is predicted to decrease in these areas (Wei et al. 1995). Conversely, projected warming is predicted to benefit the growth of Yeddo spruce, thereby expanding the distribution of the dark coniferous forests (Hao et al. 2001) while further enhancing the dominance of Erman's birch at the tree line (Yu et al. 2011). However, many dendroclimatology and dendroecology studies conducted in this region suggest increases in temperature and precipitation during the growing season enhance both the growth of Korean pine (Li et al. 2011) and Erman's birch (Yu et al. 2007) at their upper limits, but are not expected to benefit radial growth of Yeddo spruce (Yu et al. 2006). Notably, the trees in these dendroclimatology studies were all sampled from their individual upper elevation limits, which is the elevation limit of the species (Wang et al. 2013), whereas trees in the models were from their respective forest upper boundary, which is still the main distribution area of the species, but the area experiencing stress for the forest type (He et al. 2005). The differences in condition and species composition may explain the discrepancy between these predictions.
The relationship between dominant species and climate was used by Crimmins et al. (2011) as basic assumptions and preconditions in many models to simulate future growth patterns and the potential distribution area under a range of different climate change scenarios. However, during the second half of the Twentieth Century, unstable relationships between tree growth and climate factors were observed in the European Alps by Marco and Carlo (2006), in Alaska by Lloyd and Fastie (2002), in Siberia by Jacoby et al. (2000), and in the northeastern Tibet Plateau by both Shi et al. (2010) and Zhang et al. (2009, 2010, 2011). These studies have shown an anomalous reduction in forest growth indices and temperature sensitivity (Tessier et al. 1997). Trees in temperature-limited areas have a greater tendency to display divergent responses (Marco and Carlo 2006; Zhang et al. 2010). Temperature-induced drought stress has been suggested as the main cause for a reduction in sensitivity to temperature (D'Arrigo et al. 2008). Water availability during the relatively short vegetative growth period becomes a key factor for tree growth when temperatures are higher than has historically occurred at a particular thermal boundary (Marco et al. 1998). Divergent responses of individual trees and populations in recent decades therefore seem to not only occur in areas of high elevation (Cook et al. 2004), but also to tree-line species in the mid-latitudes of Asia (Zhang et al. 2011). Accordingly, temporal instability of climate–growth relationships should be considered in order to provide better estimates of future forest distributions, especially in regions where temperatures are expected to increase (Wilmking et al. 2004). However, only a few investigations on the temporal stability of climate–growth relationships have been conducted at Changbai Mountain. Furthermore, these studies were restricted to one of the dominant species of the mixed broad-leaved Korean pine forest (Wang et al. 2011; Yu et al. 2013). Hence, an important question remains: does the climate sensitivity of dominant tree species change with climate or remain constant? The answer to this question may further explain the observed divergent growth pattern.
The main objective of this study was to provide qualitative empirical data for understanding the growth dynamics of Korean pine, Yeddo spruce, and Erman's birch at the upper boundary of their respective forest type. To accomplish this task, the response patterns to climate variability over the Twentieth Century (1900–2006) and 95 years of growth for three dominant temperate forest tree species, Korean pine, Yeddo spruce, and Erman's birch, were analyzed to identify shifts in radial growth rates and growth sensitivity to climate drivers. This study tested the following two hypotheses. (1) The trees distributed at the upper boundary of each forest are under less stress and therefore grow faster than those located at the individual upper elevation limits. (2) Compared with the trees at the individual upper elevation limits, there is differential response of the three species to climatic drivers at the upper boundary of each forest. A comprehensive analysis of climate–growth relationships of these three dominant trees at the upper boundary of each forest type will facilitate accurate estimates of their future growth patterns and provide valuable insight for forest management.
Materials and Methods
Study Site
The study area is located along the north slope of Changbai Mountain within the territory of the Changbai Mountain Nature Reserve (128°07′49″E, 42°19′10″N, ca. 1085--1950 m a.s.l). Three dominant tree species, Korean pine (Pinus koraiensis) from a mixed broad-leaved and Korean pine forest, Yeddo spruce (Picea jezoensis) from a spruce–fir forest, and Erman's birch (Betula ermanii) from an Erman's birch forest were selected for detailed analysis. The area has a continental and temperate monsoon climate (Guo et al. 2005). Consistent with observations of other mountains (Schweingruber 1996), the temperature decreases while precipitation increases with elevation at Changbai Mountain. The mixed broad-leaved and Korean pine forest has longer and colder winters followed by short and warm summers. The average annual temperature is 3.6°C. Monthly mean temperatures range from −15.4°C in January to 19.6°C in July. Average annual rainfall amounts to 707 mm, with 74% falling from May to July. The spruce–fir forest has cold, windy winters and rainy, wet summers. Ca. 80% of the total annual precipitation at the site occurs between June and September. The average annual temperature is −1.8°C. In the Erman's birch forest, the annual mean temperature ranges from −8.6°C in January to −5.2°C in July. Annual precipitation ranges from 700 to 1500 mm (meteorological scientific data, http://data.cma.cn/).
Tree-Ring Collection
Our study focused on sites that have been little impacted by human disturbance at the upper boundaries of each forest type for each of the three species. Tree-ring samples of Korean pine came from two sites (at elevations of 1085 and 1100 m a.s.l., respectively), whereas samples from Yeddo spruce came from three sites (at an elevation of 1700 m a.s.l.). Ring-width cores of Erman's birch were obtained from a single site (at an elevation of 1950 m a.s.l.). At each site, adult trees that dominated in the canopy layer were chosen. During the summers of 2009 and 2013, increment cores were obtained from healthy trees using a 5.5-mm increment borer at breast height (1.3 m). One or two cores were taken from each tree, depending on their quality. The data corresponding to individual upper elevation limits comes from previous studies. In these data, Korean pine samples were from two sites at an elevation of 1300 m (Gao et al. 2011), Yeddo spruce samples were from one site at an elevation of 1800 m (Gao et al. 2014), and Erman's birch samples were from one site at an elevation of 2010 m (Wang et al. 2013). These data were used in the current study to compare the growth pattern between the forest upper boundary and individual upper elevation limits. In total, 148 trees were sampled: 57 Korean pines, 60 Yeddo spruce trees, and 31 Erman's birches. All samples were dried, mounted, and sanded using sandpaper of up to 600-mm grit until ring boundaries became clearly visible. Cores were visually crossdated by using the point year technique (Stokes and Smiley 1968). The ring widths were measured to the nearest 0.01 mm on the LINTAB5 measuring stage and statistically verified using COFECHA software (Holmes 1999).
Climate Data
Because most meteorological stations in China were established after 1950, gridded data from the high-resolution 0.5° × 0.5° gridded climate dataset CRUTS3.0 (128°25′E, 42°25′N) was used (Climatic Research Unit, University of East Anglia, https://crudata.uea.ac.uk). In order to evaluate the quality of gridded data over time, a 10-year moving standard deviation (SD) was employed. The moving SDs of temperature and precipitation data were stable during the period from 1930 to 2007. Therefore, gridded CRU climate data for monthly mean temperatures and precipitation levels from AD 1930 to 2007 were used. The homogeneity of the climate data was tested using a Kendall test (Kendall 1970) and double-mass analysis (Kohler 1949).
Tree-Ring Data Processing
Radial Growth Trend Analysis
As individual trees within a site have shown different growth trends or divergent climate–growth associations throughout the Twentieth Century (Wilmking et al. 2004; Zhang et al. 2009), growth trends for each individual tree-ring series were analyzed by dividing trees into groups (i.e. positive and negative). Temporal trends in the growth of our three species were assessed by regression analyses in R (R Development Core Team 2010). Growth trends over the last nine decades were summarized by the growth rate per decade of each tree as estimated from a regression analysis in order to compare differences between the upper boundary of each forest and individual upper elevation limits in the growth patterns of each of the three tree species. Tree-ring chronologies based on these groups were established using a negative exponential curve or a simple linear regression with a negative or zero slope in order to remove variance related to age fluctuations using the program ARSTAN.
Climate–Growth Relationship Analyses
Response function and moving correlation analyses were carried out using the program DENDROCLIM2002 (Biondi and Waikul 2004). Correlations between tree-ring standard chronologies and climatic variables, i.e. monthly mean temperature and total monthly precipitation over 13-month windows from September of the previous year (i.e. the previous year's ring formation) to September of the current year (i.e. the year of ring formation) were examined. Temporal stability of the growth-climate relationships obtained was evaluated with moving correlation functions (MCFs; Biondi and Waikul 2004). The statistical significance and stability of MCFs were tested with a bootstrap procedure using 1000 replicates. We chose 35 years as the moving interval for each calculation within the analysis. The analysis started with the window corresponding to the 1930–1965 period and ended with the 1972–2007 period window, for a total of 42 analyses of the 35-year relationship between growth and climate. The statistical significance of the coefficients was determined by calculating 95% quintile limits based on the 1000 bootstrap re-samplings of the data.
Results
Radial Growth Trend Analysis
Differential growth patterns of the three species were observed between the trees at the upper boundary of each forest and those at the individual upper elevation limits (Figure 1). Korean pine and Yeddo spruce at their respective forest upper boundaries grew faster than those at the individual upper elevation limits. Additionally, Korean pines had undergone a marked decline in growth since 1960, whereas the growth of Yeddo spruces appeared to have remained relatively stable and even increased as early as the 1930s. Moreover, the growth patterns of Erman's birch trees at their forest upper boundary contrasted with those Erman's birch trees at their individual upper elevation limits. Erman's birch trees at the forest upper boundary grew faster than those at the individual upper elevation limits over the long-term (i.e. 1870–1960), but slower more recently (i.e. 1960–2007; Figure 1).
Figure 1.
Rates of growth per decade for three species at forest upper boundary (triangles) and at individual upper elevation limits (circles). PK = Pinus koraiensis; PJ = Picea jeroensis; EB = Betula ermanis; x = decade years; y = rate of growth per decade. Solid (dashed) lines indicate the growth trend of species located at forest upper boundary (individual upper elevation limits).
As similar growth patterns were observed among individual trees within each site, six standardized tree-ring chronologies were developed (Figure 2). The expressed population signal (EPS) values of all chronologies exceeded the accepted cut-off of 0.85 (Wigley et al. 1984). Erman's birch trees had higher mean sensitivity, showing variation consistent with the effect of climate. Except for Erman's birch trees at the forest upper boundary, all chronologies had a higher signal-to-noise ratio (SNR). The smaller tree sizes could account for the low SNR, which is related to the number of trees and the correlation among the trees.
Stability of the Tree-Ring Growth–Climate Relationships
Growth responses to monthly climate variables were species specific (Figure 3) and varied across temporal scales (Figure 4). Korean pine growth was significantly and negatively correlated with late spring temperatures (current April and May), but positively correlated with precipitation at the end of the growing season (previous September and current August). The previous winters’ temperatures (previous October) and precipitation (previous December) were the major factors limiting the growth of Yeddo spruce trees. Erman's birch trees were significantly correlated with the monthly mean temperature and precipitation during the growing season (June and July) and significantly and negatively correlated with the early spring temperatures (February, March, and May), but positively correlated with the previous winters’ precipitation total.
Figure 3.
Correlations between each tree-ring residual chronology and mean climatic variables derived from grid data (1930–2007) from September of previous year to September of the current year. Dashed lines indicate significant correlations at p < 0.05. PK = Pinus koraiensis; PJ = Picea jeroensis; EB = Betula ermanis.
Figure 4.
Moving correlations between standard chronologies of three dominant species and limited climatic factor. Black horizontal lines indicate 95% significance levels.
The moving 35-year correlations between tree-ring chronologies and major climatic variables confirmed a stable relation between the three dominant species and precipitation, but there was a shift over time in the relationship between the growth of trees and temperature (Figure 4). The negative relationship between Korean pine growth and late spring temperature has been significant since the year 1980. Over the same time interval, the positive relationship of Yeddo spruce growth and the previous winter temperature became insignificant. In contrast, Erman's birch trees were consistently and negatively correlated with temperature in the early spring. The negative effect of the previous winter temperature on Erman's birch growth was weakening, while there was an increasingly negative relationship of its growth with the cumulative precipitation of the previous winter.
Discussion
Growth Patterns among Species
The radial growth rates of trees located at their species-specific individual upper elevation limits differed from those growing at the upper boundary of each corresponding forest type. Korean pine and Yeddo spruce grew faster at the upper boundary of each forest than at the individual upper elevation limits. This indicated that the most suitable conditions for Korean pine and Yeddo spruce growth occurs in the forest edge, in support of hypothesis (1). Moreover, at the forest upper boundary, the growth rate of Korean pine has decreased sharply since 1960, while the increase in the growth rate of Yeddo spruce was stable as early as the 1930s. This confirms the model simulation that indicated the average growth of Korean pine would decrease (Wei et al. 1995) and the biomass of Yeddo spruce would dramatically increase (Hao et al. 2001). However, at individual upper elevation limits, the growth of Korean pine increased whereas the growth of Yeddo spruce remained unchanged. These findings were consistent with earlier reports focused on individuals at their upper elevation limits (Yu et al. 2006; Gao et al. 2011; Wang et al. 2013; Yu et al. 2013; Gao et al. 2014; Wang et al. 2016).
These discrepancies may be explained by the site occurring along ecological gradients. The Korean pine and Yeddo spruce cores from the forest upper boundary collected at elevations of 1100 m a.s.l. and 1700 m a.s.l., i.e. the upper boundary of the mixed broad-leaved and Korean pine forest and dark forest, respectively. Samples of Korean pines and Yeddo spruce at the upper limit of individual tree species were obtained at elevations of 1400 m a.s.l. and 1800 m a.s.l., respectively. The ring width of trees within a forest depends on a variety of factors, such as climate (Campelo et al. 2013) and competition (Martín-Benito et al. 2008). Conversely, the trees growing at their upper limits appear to be mostly impacted by climate (Briffa et al. 2002). Given that trees that have grown to the height of the forest canopy are less affected by competition, greater variation among individuals of the same species across different environmental conditions should be impacted by physiological adaptation to climate (Wang et al. 2011; Galván et al. 2014).
Growth trends among Erman's birches changed abruptly at 1960. The Erman's birches at the forest upper boundary grew faster than those at individual upper elevation limits during the 1900–1960 period, supporting hypothesis (1), after which, growth rates decreased, in conflict with hypothesis (1). This pattern was not reported by earlier studies on Erman's birches from Changbai Mountain (Yu et al. 2007; Wang et al. 2013). However, similar growth patterns have been observed in spruce in north central China (Fang et al. 2012) and along the tree line in Alaska (Wilmking et al. 2004). Brienen et al. (2012) and Bowman et al. (2013) have shown that forest stand structures that capture growth responses to heterogeneous environmental conditions will produce biased growth estimates. The structure of birch forests and physiological features of birch trees differed between the forest upper boundary and individual upper elevation limits. Mono-dominant Erman's birch forests have crown densities of 0.4–0.6 at their forest upper boundary (an elevation of 1700–2000 m), whereas patches of Erman's birches are located in alpine tundra at the individual upper elevation limits (>2000 m a.s.l.; Shi et al. 2000). The nonstructural carbohydrate (including starch) concentrations in Erman's birch branches increased significantly with elevation, while the ratio of starch to sugar decreased with elevation in branches and leaves (Zhou et al. 2009). Consequently, it appears that growth models based on trees from individual upper elevation limits are less informative of growth at the forest upper boundary (Marco et al. 2011). These results highlight the need for investigations of tree growth at a given type of forest site, prior to applying growth models.
Temporal Stability of Climate–Growth Relationships among Species
As expected, a differential growth response to climate was observed between the trees at the forest upper boundary and trees at the individual upper elevation limits, supporting hypothesis (2). We confirmed the positive effect of precipitation at the end of the previous growing season (in September) on Korean pine (Gao et al. 2011; Yu et al. 2011; Wang et al. 2013). However, the negative effect of temperature during late spring was inconsistent with an earlier report suggesting that the temperature in late spring and the growing season was positively correlated with Korean pine growth at the upper limit of the species (Yu et al. 2011; Yu et al. 2013). Average annual precipitation increases with elevation, from approximately 750 mm at 900 m a.s.l. to 1100 mm at 1800 m a.s.l. The average annual temperature also decreases with elevation from about 1.8°C at 900 m a.s.l. to −2.8°C at 1800 m a.s.l. The temperature in the canopy is moderated by transpiration from leaves. This effect prevents the daytime temperature from rising rapidly, making the sub-canopy space cooler than a clearing during the day. At night, the canopy prevents rapid heat loss via radiation into the atmosphere from the understory. Hence, the air temperature remains higher than outside the forest. High temperatures in late spring nights may have increased evapotranspiration and/or respiration, which would have resulted in water stress during the growth of Korean pine early in the season (Wu and Shao 1996).
Our results support the suggestion that temperature and precipitation in the previous winter positively affect the radial growth of Yeddo spruce (Yu et al. 2006, 2007). Schweingruber (1996) reported that most spruce ring width occurs during the early growing season. Warmer winters may initiate early cambium growth and thus increase ring widths (Wu 1990). Unexpectedly, previous reports of significant and negative effects of precipitation in the current May (Gao et al. 2011) and August (Gao et al. 2014) and the positive effect of current growing season temperatures (Yu et al. 2011) on growth were not supported by our observations. Instead, only precipitation in the current May and temperature in the current growing season had a slight effect on the radial growth of Yeddo spruce. The stand structures of our sites are very complex because of the presence of other species, i.e. spruce (Picea koyamae var. koraiensis), fir (Abies nephrolepis), and a few broad-leaved tree species. Accordingly, the observed differential growth response to climate may be partly explained by complex competitive stress. However, further investigations would be required to test this hypothesis.
Significantly positive correlations of Erman's birch growth with monthly mean temperature and precipitation during the growing season suggest warm summers favored tree growth at the upper forest boundary (Wang et al. 2013). However, the factors limiting tree-ring growth of Erman's birch at the individual upper elevation limit do not include the temperature of the current growth season but that of the previous winter and the current March (Yu et al. 2005; Yu et al. 2006). Some Qinghai spruce trees (Picea crassifolia) have shown increasing and stable negative responses to temperatures during the growing seasons throughout the most recent decades (Zhang et al. 2010). Erman's birch forests are situated below the tree line, where windstorms and heavy snowfall occur and are known to substantially affect the population dynamics and short-term growth trends of trees at the tree line (D'Arrigo and Jacoby 1993). The climate near the forest edge also reflects the transition from the forest to clearings (Karl 1984). Changes in climate from the forest edge into openings depend on the size of the opening and the geometry of the edge. The climate in a clearing progresses from that of a protected forest near the edge to that of a totally exposed opening near the center. Near the edge, the high back-radiation from the tree crowns slows the rate of surface cooling. This effect decreases with distance from the forest edge, which appears to be partly responsible for the differential growth pattern. It suggests that trees growing at the tree line are more sensitive to climate warming, but those small extensions of habitat heterogeneity or disturbance events may have masked the response (Yu et al. 2016).
The forests at Changbai Mountain are sensitive to climate change and have experienced climate warming since 1960 (Ren et al. 2005; Figure 5). As temperatures have increased, a stable relationship between precipitation and tree growth and variability of temperature–growth relationships have been observed. Previous studies have suggested that precipitation is a critical factor limiting radial growth of trees on Changbai Mountain and that trees growing at the upper limit of their distribution range are more affected by precipitation than by temperature (Yu et al. 2007). Our results indicate that the effect of precipitation on tree growth is still strong for trees within forests, although forests increase in both abundance and frequency with local precipitation. The divergence between tree growth and mean temperatures in late spring and the previous winter became clearly recognizable and has continued to increase up until the end of the common record, at ca. 1980. This phenomenon has also occurred for trees at the individual upper elevation limits at Changbai Mountain (Yu et al. 2013) and extensively in mid-to-high latitudes of the northern hemisphere (Driscoll et al. 2005) in studies focusing on the late Twentieth Century (Briffa 1998; Briffa et al. 1998; Barber et al. 2000; D'Arrigo et al. 2008). More investigations are necessary to improve our understanding of growth responses to climate change within forests and at elevational limits.
Acknowledgments
Authors thank the anonymous Referees and Associate Editor Qi-Bin Zhang for the detailed revision and constructive comments that greatly improved manuscript. This research is supported by the Program of Key Project of National Key Research and Development Plan (2017YFC0504003), the Program of National Natural Science Foundation of China (31600509; 31670643), and Beijing Excellent Talents Training Project (2013D009046000001).
