Introduction

Photosynthetic carbon gain and biomass growth in forest ecosystems are key factors in the carbon cycle of terrestrial ecosystems, and will become increasingly important with the predicted changes in the global climate (Canadell et al. 2000; Canadell and Raupach 2008; Ito 2008; Beer et al. 2010; Reichstein et al. 2013). In the past decade, many studies have reported on the carbon budget of the world’s terrestrial ecosystems and described the associated ecological and micrometeorological processes. The methods used in such studies include surveys of ecological processes responsible for biomass growth and soil carbon sequestration (e.g., Ohtsuka et al. 2005; Fang et al. 2014), micrometeorological measurements of CO2 flux (Yamamoto et al. 1999; Baldocchi et al. 2001; Saigusa et al. 2005; Owen et al. 2007), satellite remote sensing (Running et al. 2004; Piao et al. 2008), and simulation models (Bonan 1996; Ito and Oikawa 2002; Piao et al. 2012). Integrated analyses combining the findings from these studies have revealed the spatial and temporal dynamics of carbon cycles and budgets (Canadell et al. 2000; Sasai et al. 2005; Piao et al. 2008; Ohtsuka et al. 2009). To date, however, researchers have not been able to estimate accurately the current status of these budgets, nor have they predicted their future changes from an ecophysiological viewpoint. Such estimates and predictions would provide an improved mechanistic understanding of the responses of plants and ecosystems to environmental changes (Tenhunen and Kabat 1999; Muraoka and Koizumi 2009).

The dynamics of forest-canopy photosynthesis can be predicted from the leaf-level photosynthetic responses to changes in micrometeorological conditions such as photosynthetic photon flux density (PPFD), air temperature, and CO2 concentration (e.g., Farquhar et al. 1980; Baldocchi and Meyers 1998; Bernacchi et al. 2013), which fluctuate over short time scales (e.g., daily). In addition to short-term changes, the seasonal changes (i.e., phenology) in leaf/canopy morphology and physiology reflect the responses to environmental changes (e.g., Wilson et al. 2001; Ito et al. 2006; Muraoka et al. 2010). Recent in situ remote sensing observations of temperate forests have emphasized the functional importance of phenology in determining canopy photosynthesis (e.g., Richardson et al. 2006; Nakaji et al. 2007; Saitoh et al. 2012; Nagai et al. 2013). However, it is important to increase our understanding of leaf-level phenology with respect to photosynthesis and related components such as chlorophylls and nitrogen, because photosynthesis is the fundamental process controlling tree growth and the responses of the forest carbon cycle to environmental change. Understanding leaf-level phenology is also important from the perspective of using remote sensing to observe forests, as such observations rely on our ability to understand the mechanisms responsible for the spectral changes in the forest canopy.

The leaf phenology of deciduous trees can be characterized by the timings of leaf emergence in spring and leaf senescence and fall in autumn, which together determine the length of the growing season. These timings also strongly affect the growth and survival of individual plants (Augspurger 2008) through their effects on the carbon balance between photosynthesis and respiration at an individual plant level. Hence, they also determine the net primary production of forests (Goulden et al. 1996; Keeling et al. 1996; White et al. 1999). In addition to the length of the growing season, the phenology of single leaves (i.e., temporal changes in leaf morphological and physiological characteristics such as leaf area, leaf mass per unit area (LMA), leaf chlorophyll and nitrogen contents, and photosynthetic capacity) is responsible for the canopy carbon gain during the growing season. Our understanding of leaf ecophysiological characteristics of tall trees has progressed remarkably with technological improvements in photosynthetic measurement systems (Reich et al. 1991; Miyazawa et al. 1998; Wilson et al. 2000; Niinemets et al. 2004; Muraoka and Koizumi 2005; Hikosaka et al. 2007; Wang et al. 2008; Yasumura and Ishida 2011). These studies have demonstrated that single-leaf photosynthetic characteristics show substantial seasonal- or leaf age-dependent changes in deciduous and evergreen species. In general, the photosynthetic capacity increases from leaf expansion to leaf maturity and decreases before leaf fall. Previous studies have shown that there are temporal changes in the relationships between photosynthesis and nitrogen or LMA, and in the temperature responses of biochemical processes related to photosynthesis. Those studies have also demonstrated that the single-leaf phenology of ecophysiological parameters varies among species even within a single functional type (e.g., deciduous broadleaf trees) and even when they are growing in the same forest. The phenology of such ecophysiological characteristics is the fundamental process determining plant growth and forest canopy function. Therefore, we must expand on the current knowledge by studying the species-specific characteristics of leaf phenology and the generality of leaf phenology over multiple years. To do so, we must consider micrometeorological conditions rather than just calendar dates, since leaf biochemical and anatomical development depend so strongly on the leaf environment (e.g., Hikosaka et al. 2007).

The phenology of ecophysiological characteristics in a single leaf is driven by environmental conditions such as solar radiation, photoperiod, and air temperature (e.g., Cleland et al. 2007; Hänninen and Tanino 2011). In boreal and temperate regions, air temperature is often the main factor determining the timings of phenological events (Kramer et al. 2000; Peñuelas et al. 2009 ; Chmielewski and Rötzer 2002). Several studies have documented the effects of recent global warming on the timing of phenological events such as leaf emergence and leaf fall in various tree species (Walther et al. 2002) and in terrestrial vegetation at a continental scale (Myneni et al. 1997; Menzel et al. 2006; Piao et al. 2006). Therefore, a more comprehensive understanding of leaf phenological responses to changing temperatures will improve the accuracy of predictions of forest carbon gain under current and future climate change scenarios (Morin et al. 2010; Richardson et al. 2010; Chung et al. 2013). Considering the interannual variations in phenology, such studies should include long-term monitoring over multiple years.

The forest at the Takayama site in Japan is a cool-temperate deciduous broadleaf forest. Long-term multidisciplinary research at this site has quantified the micrometeorological and ecological processes governing the carbon budget of the forest, as well as the degree of interannual variation in the ecological processes of forest carbon assimilation and release (Yamamoto et al. 1999; Saigusa et al. 2005; Yamamoto and Koizumi 2005; Muraoka and Koizumi 2009; Ohtsuka et al. 2009). A model analysis suggested that interannual variations in leaf phenology and summer meteorological conditions are responsible for variations in the forest carbon budget as a result of interannual changes in canopy photosynthesis (Muraoka et al. 2010).

In this study, we focused on the major canopy trees Quercus crispula Blume and Betula ermanii Cham. at the Takayama site. The aims of this study were as follows: (1) to clarify the phenological patterns of leaf photosynthetic capacity and related physiological characteristics, and (2) to identify patterns in these characteristics of deciduous broadleaf trees by examining the phenological dependency on temperature. The results of this ecophysiological study will improve our understanding of ecosystem-scale responses to environmental changes, and help to improve the accuracy of predictions about future responses to climate change.

Materials and methods

Study site and plant species

Field measurements were conducted in a cool-temperate deciduous broadleaf forest (36°08′N, 137°25′E, 1420 m a.s.l.) on the northwestern slope of Mt. Norikura, in central Japan. The Takayama study site belongs to the AsiaFlux network (http://asiaflux.net/) and the JaLTER network (http://www.jalter.org/). Researchers have investigated the carbon cycle and the CO2 exchanges between this forest and the atmosphere using micrometeorological measurements (Yamamoto et al. 1999; Saigusa et al. 2005; Murayama et al. 2010), ecological process research (Mo et al. 2005; Muraoka and Koizumi 2005; Ohtsuka et al. 2007, 2009; Nasahara et al. 2008), process-based modeling (Ito et al. 2006; Muraoka et al. 2010), and remote sensing (Muraoka and Koizumi 2009; Nagai et al. 2010). Table 1 shows air temperature and precipitation during the growing season (April to October) at the forest. The site is covered by snow from December until April, and its peak depth ranges from 100 to 180 cm. The forest canopy is dominated by Q. crispula, Betula platyphylla Sukatchev var. japonica Hara, and B. ermanii. The subcanopy and shrub layers are dominated by Acer rufinerve Sieb. et Zucc., Acer distylum Sieb. et Zucc., Hydrangea paniculata Sieb., and Viburnum furcatum Blume ex Maxim. The understory is dominated by an evergreen dwarf bamboo, Sasa senanensis (Fr. et Sav.) Rehder, with a height of 1.0–1.5 m. The height of the dominant forest canopy ranges from 13 to 20 m. All of the deciduous tree species flush their leaves after snowmelt in mid-May and lose their leaves by late October or early November.

Table 1 Climatic conditions at the Takayama study site from 2003 to 2010
Full size table

In this study, we focused on the sun leaves of B. ermanii, which has a thin crown, and the sun and shade leaves of Q. crispula, which has a multi-layered crown. The sun leaves of B. ermanii and Q. crispula were obtained from the top of the crown or from a south-facing branch. The shade leaves of Q. crispula were obtained from the middle of the crown (ca. 7 m above the ground). In 2006, the monthly average of relative PPFD (measured using IKS-27 PPFD sensors; Koito, Yokohama, Japan), which is the ratio of daily PPFD measured at the height of the shade leaves to the open sky PPFD, was 32.9 % in May, 6.7 % in June, 3.5 % in July, 3.3 % in August, and 4.3 % in September.

Measurements of leaf ecophysiological traits

We conducted our measurements of canopy leaves during the growing season, from budburst in mid-May to senescence at the end of October. Measurements were conducted from 2003 to 2010 (except for 2008) for B. ermanii and Q. crispula sun leaves and from 2003 to 2006 and in 2010 for Q. crispula shade leaves. Quercus crispula has a single leaf flush, which occurs within a single short period in the early growing season, whereas B. ermanii shows indeterminate leaf production and can produce several consecutive leaf flushes (Koike 1988). In this forest, B. ermanii produces two leaf flushes, the first in mid-May and the second in early June. Typically, more than half of the leaves from the second flush are lost to herbivores (data not shown); therefore, we focused on the leaves from the first flush. We measured leaf CO2 gas-exchange traits, chlorophyll content, nitrogen content, and LMA for the same sample leaves on each measurement day. The measurements were conducted once every 2–4 weeks. However, we were unable to obtain some measurements because of the rapid change in leaf size during the expansion period or unsuitable weather conditions at the study site.

Leaf photosynthetic CO2 gas exchange measurements were conducted from a canopy-access tower (18-m height) using a portable photosynthesis measuring system (LI-6400, Li-Cor, Inc., Lincoln, NE, USA). We measured three to five representative leaves of each species and their light environments in the morning (from 0800 to 1200 h) to avoid the influences of midday stomatal closure and photoinhibition. The dark respiration rate (R) and the relationship between the net photosynthetic rate (A) and leaf intercellular CO2 partial pressure (C i) were determined to obtain the leaf photosynthetic characteristics and to quantify net photosynthesis. The relationships between A and C i were obtained under saturating PPFD (provided by the red–blue LED of the LI-6400) and a leaf-to-air vapor-pressure deficit of <1.2 kPa. The temperature conditions during measurements were adjusted using the fan and Peltier device of the LI-6400 to match the ambient conditions in each season (9–19 °C in May and October; 14–24 °C in June, July, and September; and 21–27 °C in August). For the sun leaves of both species, at least one entire curve for the relationship between A and C i was obtained during each field measurement period by adjusting the concentration of CO2 entering the leaf chamber (C a) from 0 to 1500 µmol CO2 mol−1 air. Since the measurement of A/C i curves was time consuming, we measured the photosynthetic rates only at ambient and high CO2 concentrations and dark respiration to increase the number of samples.

We used the one-point method (Wilson et al. 2000; Kosugi et al. 2003; Wang et al. 2008) to calculate the photosynthetic parameters. The maximum carboxylation rate (V cmax) was calculated using data from the range in which ribulose bisphosphate is saturated (C a < 370 μmol mol−1). The photosynthetic rate (A c) was calculated as a function of C i and V cmax using the following equation:

$$A_{\text{c}} = \frac{{V_{\text{cmax}} (C_{\text{i}} - \Gamma^{*} )}}{{C_{\text{i}} + K_{\text{c}} \left( {1 + O/K_{\text{o}} } \right)}} - R_{\text{d}} ,$$
(1)

where K c and K o are the Michaelis–Menten constants of Rubisco for CO2 and O2, respectively; O is the O2 concentration; Γ* is the CO2 compensation point in the absence of photorespiration; and R d is the photorespiration rate, which is assumed to be 40 % of R. We used the generic temperature responses of Γ*, K o, and K c from Bernacchi et al. (2001). The potential rate of electron transport (J max) was calculated from the photosynthetic rate (A j) at which ribulose bisphosphate regeneration is the limiting factor (C a > 1000 μmol mol−1), as follows:

$$A_{\text{j}} = \frac{{J(C_{\text{i}} - \Gamma^{*} )}}{{4 C_{\text{i}} + 8\Gamma^{*} }} - R_{\text{d}} ,$$
(2)

where J is the actual rate of electron transport. J is related to the incident PPFD (Q), and J max was estimated by solving the following equation:

$$\theta J^{2} - \left( {\alpha Q + J_{ \hbox{max} } } \right) J + \alpha QJ_{ \hbox{max} } = 0 ,$$
(3)

where θ is the curvature of the light-response curve and α is the quantum yield of electron transport. The values of α and θ were taken to be 0.3 mol electrons mol−1 photon and 0.90, respectively, according to Medlyn et al. (2002). The R, V cmax, and J max values were standardized to the reference leaf temperature of 20 °C (R 20, V cmax20, and J max20) using the temperature-dependence equations of Harley and Baldocchi (1995).

After measuring photosynthetic CO2 gas exchange, we collected the leaves in plastic bags and transported them to the laboratory (approx. 10 min). Leaf discs (1-cm diameter) were punched out in the laboratory and kept at −80 °C until use. Chlorophylls a and b were extracted from the leaf discs with N,N′-dimethylformamide and their concentrations were calculated from absorbance values determined with a spectrophotometer (UV-1200, Shimadzu, Kyoto, Japan) following the method of Porra et al. (1989). The residual parts of the leaves were scanned with a digital scanner and the area was estimated using image-analysis software “LIA for Win32” (http://www.agr.nagoya-u.ac.jp/~shinkan/LIA32/index-e.html). Then, the leaves were dried at 70 °C for 72 h to measure dry weight. We calculated LMA (g m−2) as the ratio of leaf dry weight to leaf area. The dried leaves were ground into a powder and analyzed with an NC-analyzer (Sumigraph NC-22, Sumika Chemical Analysis Service, Ltd., Tokyo, Japan, or Flash EA 1112, Thermo Electron, Waltham, MA, USA).

Some of the leaf photosynthetic and respiratory data from 2003 have been reported previously by Muraoka and Koizumi (2005), and some of the photosynthetic data from 2003 to 2009 have been used in a modeling analysis of forest canopy photosynthesis (Muraoka et al. 2010, 2013). However, the present study is the first to analyze and report the original leaf characteristics data for the entire study period.

Estimation of leaf nitrogen allocation

Using the abovementioned data, we examined the allocation of leaf nitrogen to the photosynthetic apparatus by applying the model proposed by Niinemets and Tenhunen (1997). The photosynthetic apparatus was divided into three categories: (1) Rubisco, (2) the bioenergetics pool, and (3) the light harvesting components. In the present study, we calculated the nitrogen allocation to these three components to estimate the total nitrogen in the photosynthetic apparatus, which we used in our phenology analysis. Rubisco nitrogen (N r) was calculated as follows:

$$N_{\text{r}} {\,\text{per unit leaf area}} = \frac{{V_{\text{cmax}} }}{{6.25 V_{\text{cr}} }} ,$$
(4)

where 6.25 is the ratio of the weight of Rubisco to the weight of nitrogen in Rubisco and V cr is the specific activity of Rubisco (mmol CO2 g−1 Rubisco s−1). Since V cr is assumed to be a function of temperature, N r per unit leaf area was estimated using the V cmax and V cr values at the leaf temperature during the gas exchange measurement. V cr at that temperature was calculated according to Niinemets and Tenhunen (1997). The amount of nitrogen in the bioenergetics pool (N b) was calculated as follows:

$$N_{\text{b}} {\,\text{per unit leaf area}} = \frac{{J_{ \text{max} } }}{{8.06 J_{\text{mc}} }} ,$$
(5)

where 8.06 is the amount of nitrogen in the bioenergetics pool per unit cytochrome f (mmol cyt f (g N in bioenergetics)−1) and J mc is the photosynthetic electron transport per unit cytochrome f (mmol electrons (mmol cyt f)−1 s−1). Because J mc is also a function of temperature, we used J max and J mc at the leaf temperature during the photosynthetic measurements to calculate N b. J mc at that temperature was also calculated by the equation of Niinemets and Tenhunen (1997). The nitrogen cost for light harvesting (N h) depends on the chlorophyll content associated with photosystems I (PSI) and II (PSII), and with the light-harvesting chlorophyll protein complex of PSII (LHCII). Thus, N h per unit leaf area was calculated as follows:

$$N_{\text{h}} {\,\text{per unit leaf area}} = C_{\text{c}} \left[ {0.46 P_{{{\text{C}}\left( {\text{PSI}} \right)}} + 1.17 P_{{{\text{C}}\left( {\text{PSII}} \right)}} + 0.36 P_{{{\text{C}}({\text{LHCII}})}} } \right] ,$$
(6)

where C c is the leaf chlorophyll content (mmol m−2); P C(PSI), P C(PSII), and P C(LHCII) are the fractions of chlorophyll associated with PSI, PSII, and LHCII, respectively; and 0.46, 1.17, and 0.36 are the weights of nitrogen per weight of chlorophyll bound to PSI, PSII, and LHCII, respectively (Hikosaka and Terashima 1995). PSI content (mmol m−2) was assumed to be 1.7 × C c (Niinemets and Tenhunen 1997), and P C(PSI) was calculated as follows: 1.7 × 0.184; where 0.184 is the amount of chlorophyll (mmol) bound to one PSI complex (Hikosaka and Terashima 1995). Because PSII content (PSII c) per C c is strongly correlated with cytochrome f content per C c, PSII content was estimated from the C c and cytochrome f content, which was obtained as J max/J mc according to Niinemets and Tenhunen (1997). P C(PSII) was calculated as the product of PSII c and the amount of chlorophyll bound to one PSII complex (0.06 mmol; Hikosaka and Terashima 1995). The amount of chlorophyll associated with LHCII was assumed to be the difference between C c and the chlorophyll associated with PSI and PSII. Thus, P C(LHII) was calculated as follows: 1−P C(PSI)−P C(PSII). The amount of nitrogen in the photosynthetic apparatus (N ph) equaled the sum of N r, N b, and N h.

Modeling phenology of leaf morphological and physiological traits

We examined the seasonal changes in leaf characteristics as a function of air temperature. To model aspects of leaf development (chlorophyll content, A max, V cmax20, J max20, and LMA) against air temperature during the growing season (from spring to the end of summer) we calculated the growing-degree days (GDD) index using the following equation:

$${\text{GDD}}_{k} \left( t \right) = \mathop \sum \limits_{i = 1}^{t} \hbox{max} (T_{i} - k, 0) ,$$
(7)

where t is time (day of year, DOY), T i is daily mean air temperature on DOY i, and k is a threshold temperature. To test the threshold temperature for leaf development, GDD was calculated with three temperatures; 0, 2 and 5 °C. The changes in chlorophyll content, A max, V cmax20, and J max20 during the senescence period from late summer to autumn were examined by calculating the chilling-degree days (CDD) index. The CDD index from DOY 213 (31 July or 1 August) to DOY 365 (30 or 31 December) was calculated using the following equation:

$${\text{CDD}}_{k} \left( t \right) = \mathop \sum \limits_{i = 213}^{t} \hbox{min} (T_{i} - k, 0).$$
(8)

The values of k for CDD were set to 15, 18, and 20 °C. The changes in GDD0 and CDD18 as a function of DOY are shown in Fig. 1.

Fig. 1
figure 1

a Growing-degree days above 0 °C (GDD0) and b chilling-degree days below 18 °C (CDD18) as a function of day of year (DOY) from 2003 to 2010

Full size image

In previous studies, canopy phenology was characterized as a simple sigmoid-shaped logistic function based on DOY (Zhang et al. 2003; Richardson et al. 2007; Ide and Oguma 2010). In the present study, we examined a logistic function based on DOY and GDD or CDD to characterize the seasonal changes in ecophysiological parameters:

$$f\left( x \right) = a + \frac{b}{{1 + { \exp }(c - dx)}}.$$
(9)

The data during the development period (DOY ≤ 240) were fitted to the models with x = DOY, GDD0, GDD2, and GDD5. For the data during the senescence period (DOY > 240), we tested models based on DOY, CDD15, CDD18 and CDD20. To apply this function to the senescence period, in which the calendar date is the independent variable, we used the opposite values of DOY [i.e., DOY × (−1)] since all of the CDD values are negative by definition. The parameters a and b control the lower limit (a) and the upper limit (a + b) of the function. The c parameter causes parallel shifts in the response to the driving variable x (here, either DOY, GDD, the opposite of DOY, or CDD), and changes in the parameter d affect the overall steepness of the response to the driving variable. The c parameter is positive for the leaf development model and negative for the leaf senescence model. In the leaf senescence models for chlorophyll content, A max, and V cmax, the values for parameter b were the same as those used in the corresponding leaf development models.

Statistical analysis

To examine the interannual variations in LMA, chlorophyll content, R 20, A max, V cmax20 and J max20, one-way ANOVA was conducted using the free statistical software GNU R (R Core Team 2014, http://www.R-project.org/).

Results

Seasonal and interannual variations in leaf characteristics

Figure 2 shows the seasonal and interannual variations in LMA and chlorophyll contents of B. ermanii and Q. crispula leaves, and Table 2 summarizes the peak values of each parameter and its DOY. In all leaves, LMA increased remarkably from leaf emergence until mid-summer (late July; ca. DOY 200), increased more slowly until late September, and then declined until leaf fall in October (Fig. 2a, b, c; Table 2). The high LMA values in sun and shade leaves of Q. crispula at the beginning of leaf expansion were because the very young leaves were folded and shrunken; thus, the measurements underestimated the actual leaf area. Throughout the growing season, the LMA of Q. crispula sun leaves varied among years (p < 0.05), while the LMA of B. ermanii leaves and Q. crispula shade leaves did not show interannual variations (p > 0.05; Fig. 2a, b, c). Chlorophyll content increased rapidly from leaf expansion until late July/early August (Fig. 2d, e, f; Table 2) and then decreased to almost zero in late October. Although the seasonal patterns were similar in all years, the absolute values of chlorophyll content in B. ermanii leaves varied among the years (p < 0.001); the chlorophyll content was lower in 2006 than in other years (maximum value in 2006 was 35 % lower than that in 2005).

Fig. 2
figure 2

Seasonal changes in a, b, c leaf mass per unit area (LMA) and d, e, f chlorophyll content. Asterisks or “ns” indicate significance of interannual variations (ns: *p > 0.05; **p < 0.05; ***p < 0.01; **** < 0.001; ANOVA). Values are mean ± SD (n = 3–5 leaves)

Full size image
Table 2 Peak values of LMA, chlorophyll content, Amax,V cmax20, and J max20 for two tree species from 2003 to 2010. Values shown are mean ± SD (n = 3–5 leaves) and day of year (DOY) of the peak value
Full size table

Leaf photosynthetic and respiratory characteristics also showed remarkable seasonal changes (Fig. 3). The R 20 was highest immediately after leaf emergence, then it decreased rapidly until mid-June (DOY 165), and then stabilized until leaf yellowing began in the autumn (Fig. 3a, b, c). There were almost parallel changes in A max, V cmax20, and J max20 throughout the growing season, but the peak in J max20 tended to be later than the peaks in A max and V cmax20 in all leaves (Table 2). The values of A max, V cmax20, and J max20 increased from leaf emergence until late August in Q. crispula sun leaves, but increased rapidly until early July or mid-August in B. ermanii leaves (Table 2). In all leaves, A max and V cmax20 started to decrease in late August. The J max20 started to decrease in the sun leaves of Q. crispula in late August, but we could not find a clear trend in J max20 in B. ermanii sun leaves because of insufficient data (measurements of the photosynthetic rate at a high CO2 concentration to determine J max were difficult in that month because of rapid stomatal closure). The A max and V cmax20 of Q. crispula leaves showed yearly variations (p < 0.001); A max and V cmax20 were lower in 2009 than in other years (maximum values of A max and V cmax in 2009 were 26 and 24 % lower than their respective values in 2005).

Fig. 3
figure 3

Seasonal changes in a, b, c dark respiration rate at 20 °C (R 20), d, e, f light-saturated photosynthetic rate (A max), g, h, i maximum carboxylation rate at 20 °C (V cmax20), and j, k, l potential electron transport rate at 20 °C (J max). Asterisks or “ns” indicate significance of interannual variations (ns: * p > 0.05; **p < 0.05; ***p < 0.01;  < ****0.001; ANOVA). Values are mean ± SD (n = 3–5 leaves)

Full size image

The dates of the peak values also differed among years. In 2003, the peak value of V cmax20 was on DOY 173 in Q. crispula shade leaves, but was later in B. ermanii leaves (DOY 260) and Q. crispula sun leaves (DOY 246). In 2009, the peaks of A max and V cmax20 were on DOY 192 in B. ermanii leaves, but approximately two months later (on DOY 256) in Q. crispula sun leaves (Table 2).

Figure 4 shows the seasonal variations in total leaf nitrogen content (N total) and N ph. N total per unit leaf biomass was high at the start of leaf emergence (ca. 50 mg g−1) and decreased rapidly until mid-June, reaching ca. 20–30 mg g−1 (Fig. 4a, b, c). During the summer, N total per unit leaf biomass decreased slowly, and then decreased more rapidly from the end of September. Overall, the N total per unit biomass was similar in the leaves of B. ermanii and Q. crispula (both sun and shade leaves). The N total per unit leaf area was high in very young leaves (ca. 3–4 g m−2; Fig. 4d, e, f), but showed little variation after early June. In Q. crispula, the N total per unit area was lower in shade leaves than in sun leaves.

Fig. 4
figure 4

Seasonal changes in (a, b, c) total leaf nitrogen (N total) per unit leaf biomass, (d, e, f) N total per unit leaf area, (g, h, i) nitrogen in photosynthetic apparatus (N ph) per unit leaf area, (j, k, l) N ph per unit leaf biomass, and (m, n, o) ratio of N ph to N total. Values are mean ± SD (n = 3–5 leaves)

Full size image

Both on a leaf biomass and leaf area basis (Fig. 4g, h, i, j, k, l), N ph changed remarkably throughout the season. Both showed a rapid initial increase, a stable period during most of the summer, and a rapid decline in autumn, but the patterns of N ph differed from those of N total. The ratio of N ph to N total was generally low (ca. 10–15 %) at the start of leaf emergence, but increased rapidly to between 40 and 50 % in mid-June (Fig. 4m, n, o). N ph/N total in the sun leaves of Q. crispula started to decrease in mid-summer, and the decrease continued through the senescence period, but N ph/N total in B. ermanii leaves did not change until leaf fall.

Modeling phenology of leaf characteristics

The seasonal increases in chlorophyll content, A max, V cmax20, J max20, and LMA during the leaf development period fitted well with the development model based on GDD0. The decline in chlorophyll content, A max, and V cmax20 during the autumn senescence period fitted well with the senescence model based on CDD18 (Table 3). The senescence model based on CDD15 and CDD20 showed a poor fit (extremely low r 2 values) with chlorophyll content and V cmax20 of B. ermanii. In Q. crispula sun leaves, A max and V cmax20 fitted the DOY-based model almost as well as the GDD-based model. For all leaves, chlorophyll content and LMA showed a better fit (higher r 2 values) with the GDD2 model than with the GDD0 model. For the senescence model, the model based on the opposite value of DOY showed high r 2 values for chlorophyll content of B. ermanii and A max, V cmax20 and J max20 of Q. crispula sun leaves. Because LMA showed little change during the senescence period (Fig. 2a, b, c), we did not fit LMA to the senescence model. We did not fit J max20 of B. ermanii and of Q. crispula shade leaves to the senescence model because of insufficient data in October.

Table 3 Coefficients of determination (r 2) for logistic models of phenological changes in ecophysiological traits during leaf development and leaf senescence periods
Full size table

We used the developmental model based on GDD0 and the senescence model based on CDD18 (Table 4 and Fig. 5) to clarify the variations in increases and decreases in the ecophysiological parameters of the various species. The maximum values of A max, V cmax20, and J max20 were similar in B. ermanii and Q. crispula sun leaves, and were higher in B. ermanii and Q. crispula sun leaves than in Q. crispula shade leaves. However, the temporal changes in A max, V cmax20, and J max20 differed between B. ermanii and Q. crispula. During the development period, the values of these parameters increased rapidly with increasing GDD0 in B. ermanii sun leaves, but increased more gradually in Q. crispula sun leaves (d values; Table 4). During the senescence period, A max and V cmax decreased more rapidly in B. ermanii leaves than in Q. crispula sun leaves. Chlorophyll content also decreased rapidly in B. ermanii leaves. These analyses revealed that B. ermanii is characterized by rapid growth and senescence, whereas Q. crispula is characterized by gradual growth and senescence. In Q. crispula, the maximum values of A max and V cmax20 were lower, and peaked earlier, in the shade leaves than in the sun leaves (Fig. 5c, e, g).

Table 4 Fitting parameters (a, b, c and d) for logistic models of phenological changes in ecophysiological traits during leaf development period based on growing-degree days above 0 °C (GDD0; development model) and during senescence period based on chilling-degree days below 18 °C (CDD18; senescence model)
Full size table
Fig. 5
figure 5

Logistic models during leaf development period for a chlorophyll content, c light-saturated photosynthesis (A max), e maximum carboxylation rate at 20 °C (V cmax20), g potential electron transport rate at 20 °C (J max20), and (i) leaf mass per unit area (LMA) based on growing-degree days above 0 °C (GDD0) and corresponding values during senescence period for b chlorophyll content, d A max, f V cmax20, and h J max20 based on chilling-degree days below 18 °C (CDD18). Fitting parameters are provided in Table 4

Full size image

Relationships among leaf characteristics

To examine the relationships among the leaf morphological and physiological characteristics throughout the growing season (i.e., as a function of leaf age), we further analyzed the data by dividing the leaves into three age classes according to DOY: young leaves (DOY ≤ 160), mature leaves (160 < DOY ≤ 240), and senescing leaves (240 < DOY). This classification relates to apparent changes in physiological activities: DOY 160 corresponds to the timing of the sharp change in the seasonal curve for dark respiration (R 20; Fig. 3a, b, c), and DOY 240 corresponds to the point when A max begins to decrease from its maximum level as a result of senescence (Fig. 3d, e, f).

In all of the leaves, there were strong and statistically significant positive linear correlations between J max and V cmax (both at measured temperatures) throughout the season, indicating that these parameters changed in the same manner from leaf expansion to maturity and from maturity to senescence (Table 3; Fig. 6). The slopes of the relationships between J max and V cmax (i.e., the ratio of J max to V cmax) were steeper for the young and senescing leaves than for the mature leaves (Fig. 6d, e, f), indicating that J max/V cmax changed in the middle of the growing season.

Fig. 6
figure 6

Relationships between maximum carboxylation rate (V cmax) and potential electron transport rate (J max). Leaves were divided into three leaf-age groups (Y young, M mature, S senescing) according to day of year (DOY) when the trend for certain parameters changed (see text for details). Linear regressions between J max and V cmax for each leaf age group are shown in lower panels (d, e, f) along with slope of each line. Results for linear regressions for the entire growing season are shown in Table 5

Full size image

Chlorophyll content and V cmax20 were also significantly positively correlated throughout the growing season (Table 5; Fig. 7a, b, c). V cmax20 and LMA were significantly positively correlated for all leaf-age groups of B. ermanii leaves and Q. crispula sun leaves, but there was no significant correlation between V cmax20 and LMA for Q. crispula shade leaves (Table 3; Fig. 7d, e, f). In the young and mature leaves, V cmax20 was correlated less strongly with LMA in B. ermanii sun leaves (r 2 = 0.38, P < 0.001) than in Q. crispula sun leaves (r 2 = 0.63, P < 0.001). Throughout the growing season, V cmax20 was only significantly correlated with N total per unit leaf area in B. ermanii leaves and Q. crispula sun leaves (Table 3, Fig. 7g, h). In mature and senescing leaves, the correlations between V cmax20 and N total per unit leaf area were weaker in B. ermanii leaves (r 2 = 0.44, P < 0.001) than in Q. crispula sun leaves (r 2 = 0.60, P < 0.001), but both correlations were stronger than that in Q. crispula shade leaves (r 2 = 0.38, P < 0.001).

Table 5 Linear regression coefficients between primary photosynthetic parameters and leaf characteristics
Full size table
Fig. 7
figure 7

Relationships between (a, b, c) maximum carboxylation rate at 20 °C (V cmax20) and chlorophyll content, d, e, f V cmax20 and leaf mass per unit area (LMA), g, h, i V cmax20 and total nitrogen (N total) per unit leaf area, and j, k, l chlorophyll content and N total per unit leaf area. Leaves were divided into three leaf-age groups (Y young, M mature, S senescing) according to day of year (DOY) when the trend for certain parameters changed (see text for details). Regression lines are shown for statistically significant relationships (see Table 5)

Full size image

Chlorophyll content was significantly positively correlated with N total per unit leaf area in all leaves throughout the growing season (Table 3). The strongest correlation (highest r 2 value) was for Q. crispula sun leaves (Table 3; Fig. 7k), which might have resulted from a lack of chlorophyll data for some of the sampled leaves (i.e., values with a high N total per unit leaf area and low V cmax20 in Fig. 7h). We could not obtain chlorophyll data for these very young leaves because they were too small to obtain enough leaf material for chlorophyll analyses after the photosynthetic measurements. If we only analyze the data for the mature and senescing leaves, the chlorophyll content was strongly positively correlated with N total per unit leaf area in B. ermanii leaves (r 2 = 0.46, P < 0.001) and in Q. crispula sun leaves (r 2 = 0.61, P < 0.001) and shade leaves (r 2 = 0.31, P < 0.001).

Discussion

Seasonal and interannual variations in ecophysiological characteristics

Our goal in this long-term ecophysiological research on tree leaf morphology and physiology is to obtain essential information on the leaf-level characteristics that are responsible for the spatial and temporal dynamics of CO2 fluxes in a deciduous broadleaf forest canopy. In a previous study (Muraoka and Koizumi 2005), we found that B. ermanii and Q. crispula showed remarkable seasonal changes in their photosynthetic capacity and stomatal conductance, and we recognized that leaf phenology must be a fundamental ecological process affecting ecosystem-scale photosynthetic productivity in deciduous forests (see also Wilson et al. 2001). In previous studies on the canopy leaf area index, we revealed the phenology of leaf and canopy structure and functions by investigating the seasonal changes in single-leaf area and canopy-level leaf biomass (Nasahara et al. 2008). We have also estimated ecosystem-level carbon budgets using simulation models (Ito et al. 2006; Muraoka et al. 2010).

The present study was based on 5–7 years of data, although the measurements are continuing. Our analyses show that there were interannual variations in the timings of leaf maturation and senescence (Figs. 2, 3). The A max values varied by approximately 25 % among years (Fig. 3), although this forest is not subject to drought, unlike many temperate North American forests (Abrams and Mostoller 1995; Wilson et al. 2001; Xu and Baldocchi 2003). The Takayama forest site is in the Asian monsoon region, so it receives sufficient precipitation throughout the growing season (Table 1). Wang et al. (2008) investigated the leaf photosynthetic capacity, nitrogen content, and LMA of Fagus crenata Blume along an elevation gradient on a mountain in a cool-temperate region of Japan, and found that its leaf characteristics showed interannual variations from 2000 to 2005.

The multi-year measurements in this study revealed clear interspecific differences in the phenological patterns of ecophysiological properties between B. ermanii and Q. crispula, even though both species belong to the same functional type (here, deciduous broadleaf trees). The seasonal increases in A max, V cmax, and J max and decreases in A max and V cmax occurred more rapidly in B. ermanii leaves than in Q. crispula sun leaves (Figs. 3, 5, 6). For both species, these characteristics were successfully described by our phenology model based on a logistic equation that included cumulative air temperature (GDD and CDD) as the independent variable (Fig. 5). In B. ermanii, the increases in photosynthetic parameters during the leaf development period fitted better with the logistic model based on GDD than with that based on calendar date (DOY). This was because the timings of leaf budbreak differed among the years due to differences in air temperatures during the spring (Table 1). In Q. crispula sun leaves, the development of A max and V cmax were explained equally well by the DOY-based model and the GDD0-based model. In another study at the same forest site, the yearly variations in GDD2 and CDD18 strongly affected the timings of the start of leaf-expansion and the end of leaf-fall, as estimated from digital camera images (Nagai et al. 2013). The findings of that study are consistent with our findings that the development of chlorophyll content and LMA showed the best fit with the GDD2-based model, and the decrease in chlorophyll content during senescence showed the best fit with the CDD18-based model (Table 3). Several other studies have reported interspecific variations in the seasonal patterns of leaf ecophysiological variables (Reich et al. 1991; Wilson et al. 2000; Niinemets et al. 2004). To better understand the phenology of these physiological traits, it may be insufficient to consider only the functional type of the tree.

Another interesting finding was the interannual variations in leaf phenology. The irregular patterns of photosynthetic capacity in the summers of 2003 and 2009 (Fig. 3; Table 2) might be caused by the rainy season (“baiu”). In 2003, the meteorological conditions on the main island of Japan were characterized by remarkable decrease in incident radiation due to thick cloud cover during the baiu (Japan Meteorological Agency 2005), resulting in a lower mean air temperature in July than in the other years (Table 1). In 2009, the end date of the baiu was the latest recorded from 1951 to 2014 (Japan Meteorological Agency 2014), and the number of sunshine hours during July was about half that in normal years (Japan Meteorological Agency 2009). Since leaf photosynthetic development responds to PPFD (Niinemets et al. 2004) and temperature, the sunlight shortage and lower temperatures might have delayed the photosynthetic development of B. ermanii and Q. crispula leaves at the Takayama site. Saigusa et al. (2008) discussed the effects of the sunlight shortage in 2003 on forest-level CO2 uptake in Asia, and found that this sunlight anomaly affected gross primary production at the Takayama site.

It is unclear why the chlorophyll content of B. ermanii leaves was markedly lower in 2006 than in other years (Fig. 2d). This phenomenon was not apparent for the sun leaves of Q. crispula (Fig. 2e). Various factors that we did not examine (e.g., frost in the early spring, herbivory, pathogens) could have affected the biochemical properties of B. ermanii leaves.

Comparison of the sun and shade leaves of Q. crispula provided additional ecologically important information about the multi-layered canopy at the study site. The timings of leaf budbreak of Q. crispula in sunlit and shaded environments were similar in all years. The photosynthetic capacity of the sun and shade leaves increased in a similar manner from leaf budbreak (around DOY 140) to DOY 180. Then, the photosynthetic capacity reached its maximum in the shade leaves but continued to increase in the sun leaves (Fig. 3e, f), resulting in higher maximum capacity in the sun leaves and faster maturation of the shade leaves (Fig. 5). The light availability (relative PPFD) to the shade leaves in the crown of Q. crispula was almost at the seasonal minimum at around DOY 180, and this could have affected the development of shade leaves as they acclimated to the low-light environment (see also Niinemets et al. 2004).

Seasonal changes in leaf morphological, photosynthetic, and nitrogen-use characteristics

Our measurements during the 7 years revealed details of the seasonal changes in chlorophyll and nitrogen contents, photosynthetic parameters, dark respiration, and leaf morphology (LMA), as well as their interactions. In general, leaf photosynthetic capacity is proportional to leaf nitrogen content (Hikosaka 2004), and their relationship has been widely examined for leaves from various species in a range of biomes around the world (e.g., Wright et al. 2004). The strength of the relationship between photosynthetic capacity and leaf nitrogen content has been examined for a wide range of plant species including herbaceous and trees (e.g., Evans 1989; Reich et al. 1994, 1995). The results of such studies have provided a deeper understanding of nitrogen use during photosynthesis (Hikosaka 2004).

Most studies have focused on mature leaves, but analyses of developing and senescing leaves can increase our understanding of leaf phenology. In developing or senescing leaves, there are changes in leaf structural and biochemical components that are tightly linked. These changes reflect the anatomical and physiological responses to a fluctuating environment, as well as the resource acquisition and use strategies of a species throughout the growing season (cf. Marshall and Roberts 2000). Reich et al. (1991) and Wilson et al. (2000) have shown in maple (Acer spp.) and oak (Quercus spp.) that the leaf nitrogen content and photosynthetic capacity, and their relationship, change during the growing season (i.e., as the leaves age). Miyazawa et al. (1998) and Yasumura and Ishida (2011) have shown that for evergreen broadleaf tree leaves, the relationship between leaf nitrogen and photosynthetic capacity (and, in turn, nitrogen use efficiency) changes as the leaves age. In B. ermanii and Q. crispula trees at the Takayama site, the leaf morphological and biochemical parameters showed rapid changes during the development of young leaves, relatively stable values in mature leaves, and decreasing values in senescing leaves (Figs. 2, 3, 4). However, the phenology of LMA, nitrogen content, and photosynthetic capacity did not always change in the same way, and so the relationships among these characteristics changed as the leaves aged.

Figure 8a illustrates the relationship between LMA and photosynthetic capacity in B. ermanii and Q. crispula. From leaf expansion to maturity (i.e., during the leaf development process), LMA and photosynthetic capacity increased proportionally. However, by late summer or autumn, the photosynthetic capacity decreased rapidly (i.e., during senescence) while LMA remained relatively constant. Thus, the relationship between V cmax and LMA showed a pattern of hysteresis during these two parts of the growing season (Fig. 7). Similar hysteresis was reported by Reich et al. (1991) in oak (Quercus ellipsoidalis) and maple (Acer rubrum and A. saccharum).

Fig. 8
figure 8

Conceptual diagrams showing seasonal variations in relationships between a maximum carboxylation rate at 20 °C (V cmax20) and leaf mass per unit area (LMA) and b V cmax20 or chlorophyll content and total nitrogen (N total) per unit leaf area

Full size image

The seasonal changes in R 20 also reflected the leaf developmental stage (Fig. 3a, b, c). In young leaves, R 20 was very high and then decreased rapidly. The respiration rate consists of two components: maintenance respiration, which is associated with energy generation and metabolic processes and is proportional to the amount of biomass; and growth (construction) respiration, which is associated with the construction of plant tissues (Amthor 1989). The high R 20 during the very early growing season probably reflects the high growth component of respiration during leaf development.

Figure 8b illustrates the relationship between the nitrogen content per unit leaf area and photosynthetic capacity. In the early growing season, there was a high nitrogen content (Fig. 4a, b, c, d, e, f) but a low chlorophyll content (Fig. 2d, e, f). At this stage, the photosynthetic capacity had not yet fully developed (Fig. 3d, e, f, g, h, i, j, k, l), leading to a poor correlation between N total per unit area and photosynthetic capacity (young leaves in Fig. 8b). As the chlorophyll content and photosynthetic capacity increased, the relationship became more proportional (mature and senescing leaves in Fig. 8b). Reich et al. (1991) reported a similar relationship between leaf nitrogen content and photosynthetic capacity, and Wilson et al. (2000) and Wang et al. (2008) reported that the relationship between V cmax and N total per unit area was strongest in spring and early summer, and became weaker thereafter.

Implications for ecosystem-scale research in time and space

In large-scale ecosystem research such as carbon cycle models or remote sensing-based models, seasonal changes in leaf biochemical and photosynthetic characteristics are generally ignored or are regarded as the same as seasonal changes in leaf area (e.g., Bonan 1995; but see Ito et al. 2006). However, our analyses indicate that the phenology of leaf characteristics can affect the analyses and predictions of forest-scale photosynthetic productivity and the resulting carbon balance in deciduous forests in temperate regions, as reported in previous studies (Wilson et al. 2001; Ito et al. 2006; Muraoka et al. 2010, 2013).

In this study, we successfully described the seasonal changes of leaf ecophysiological characteristics as a function of cumulative air temperature, during both leaf development (GDD) and leaf senescence (CDD). Many studies on leaf phenology have attempted to explain interannual variations and geographical differences in phenological events based on the calendar date. This is logical, because leaf development is considered to be triggered by certain temperature conditions or by thermal accumulation (Chuine et al. 2003; Hänninen and Tanino 2011). As mentioned above, the long rainy season in the summer of 2003 may have delayed leaf photosynthetic development as a result of the lower temperatures during July of that year. Although PPFD and soil water availability must also be considered, at least for deciduous trees in temperate regions, the results of the present study show that it should be possible to predict the development of leaf ecophysiological characteristics as a function of GDD.

The environmental factors that influence leaf senescence have not yet been unequivocally defined (Schaber and Badeck 2003). More detailed analyses are required to develop a leaf senescence model based on physiological mechanisms. In this study, senescence was successfully fitted to a simple logistic model based on CDD as the independent variable. Our empirical phenological models for leaf ecophysiological characteristics could be incorporated in plant growth, ecosystem carbon cycle, and hydrological models, thereby improving future analyses of the functional roles of canopy processes in growth, carbon cycling, and hydrology. This will also improve predictions of forest ecosystem function under future climate change scenarios. Currently, we are conducting an open-field experiment in which we are warming the upper branches of Q. crispula and B. ermanii to examine the possible effects of global warming on leaf morphology, photosynthesis, and phenology (Chung et al. 2013).

Another contribution of ecophysiological research on forest leaf phenology is to improve the ecophysiological uses of remote sensing data (Muraoka and Koizumi 2009). Recently, researchers have attempted to observe the seasonal changes in the ecophysiological status of a forest canopy using indirect methods such as in situ and satellite remote sensing (Nishida 2007; Muraoka et al. 2013) and digital camera images (Richardson et al. 2007; Ahrends et al. 2009; Nagai et al. 2013). The spectral reflectance of the canopies measured by spectroradiometers and RGB signals recorded by a digital camera can reflect the leaf anatomical, biochemical, and photosynthetic properties of the forest canopy. However, to improve the accuracy of monitoring the spatial and temporal dynamics of forest canopies via remote sensing, a more comprehensive understanding of the relationships among plant morphological and biochemical structures and optical signals (reflectance, RGB signals) is required. In the present study, we found that the chlorophyll content was strongly correlated with V cmax. This is an important finding, because the leaf chlorophyll content strongly affects the optical characteristics of the leaves (Porra et al. 1989; Gitelson et al. 1996; Sims and Gamon 2002). Our findings suggest that it may be possible to estimate the spatial distribution and temporal changes of ecosystem-level photosynthetic capacity from optical remote sensing data. Indeed, an optical index for canopy chlorophyll content obtained by in situ remote sensing was found to be strongly correlated with the daily maximum gross primary production of the canopy from the leaf expansion period to maturity and senescence in multiple years (Muraoka et al. 2013). Further integrated studies of leaf- and canopy-level ecophysiology, combined with process-based models and optical remote sensing at monitoring sites, promise to support and improve analyses of the environmental response of terrestrial ecosystems to climate change.