Seasonal course of translocation, storage and remobilization of 13C pulse-labeled photoassimilate in naturally growing Larix gmelinii saplings


Author for correspondence: A. Kagawa Tel: +81 29 8733211 Fax: +81 29 8743720 Email:


  • • Autocorrelation – correlation of tree-ring parameters such as ring width, density and isotope ratios to the environmental conditions of the previous year(s) – is associated with the use of previous photoassimilate for current year's tree ring formation.
  • • To clarify the seasonal course of carbon allocation patterns among needles, branches, stem and roots, we pulse-labeled 10 Larix gmelinii growing in a continuous permafrost zone with 13CO2.
  • • Photoassimilate incorporated in June was allocated mainly to above-ground parts, indicating active above-ground growth in spring. Very little was allocated to below-ground parts (2.6–7.9%), probably because root growth is inhibited by low soil temperatures in spring. Conversely, a higher proportion of July and August photoassimilate was allocated to below-ground parts (32–44 and 12–24%, respectively).
  • • About half the carbon in new needles was derived from stored material. The starch pool in non-needle parts, which can be used for xylem formation, drew approx. 43% of its carbon from the previous year's photoassimilate, indicating that carbon storage is a key mechanism behind autocorrelation in (isotope) dendroclimatology.


In order to reconstruct past climates and predict the effects of climatic changes on tree growth, numerous investigators have studied tree-ring parameters such as ring width, ring density (Fritts, 1976; Schweingruber, 1987), and stable isotope ratios (Leavitt, 1993; McCarroll & Loader, 2004). However, despite many dendroclimatological and dendrochronological studies, there is still not enough knowledge on the physiological mechanisms behind tree-ring formation to explain the observed correlation between climate and tree-ring parameters. In particular, autocorrelation – correlation of current ring parameters to environmental conditions in previous year(s) – is often observed in dendroclimatological studies (Lamarche, 1974; Fritts, 1976; Meko et al., 1993), and the presence of autocorrelation hinders climate reconstruction at annual time resolution (Monserud & Marshall, 2001). Proposed causes include the use of stored carbohydrates and the effects of climate on the formation of buds, leaves, roots, fruits and hormones (Fritts, 1976). The order and strength of autocorrelation vary among tree species and growth sites (Meko et al., 1993). On one hand, the tree-ring parameters of evergreen conifers, which retain needles for more than 1 yr, are highly influenced by the effects of environmental conditions on needle formation during previous years (Lamarche, 1974). However, in the case of deciduous trees, stored carbohydrates are probably a major cause for autocorrelation. Therefore an understanding of how much stored photoassimilate from the previous year(s) is remobilized in the formation of the current year's tree ring is needed to improve our interpretation of dendroclimatological data.

Autocorrelation is also observed in carbon isotope ratios of tree rings, in which it has been found to be mostly first order (1-yr lags) or second order (2-yr lags). For example, an evergreen conifer with a long needle-retention time (Pseudotsuga menziesii) displays a second-order autocorrelation (2-yr lag) (Monserud & Marshall, 2001). There is a model (Farquhar et al., 1982) that explains carbon isotope fractionation of plants (Vogel, 1980) relatively successfully at the leaf level, but this model excludes downstream fractionation processes during the storage and remobilization of sugars and starch (Terwilliger, 1997; Jäggi et al., 2002; Damesin & Lelarge, 2003). It cannot explain the intra-annual δ13C variation, for example, of ring-porous oak (Helle & Schleser, 2004) or of tropical mangrove (Verheyden et al., 2004). Labeling experiments are therefore necessary to clarify the relationship between the intra-annual δ13C variation of tree rings and the seasonal dynamics of stored carbon (Jordan & Mariotti, 1998). Furthermore, the downstream processes create a time lag between the uptake of carbon during leaf-level photosynthesis and the incorporation of carbon into stem xylem during tree-ring formation (Schleser et al., 1999). This time lag may differ among photoassimilates incorporated in different seasons, or among trees living in different climatic zones. One example is the fast-growing plantation species of temperate regions, Cryptomeria japonica D. Don, which consumes spring pulse-labeled 13CO2 relatively quickly, in the order of 1 month or less (Kagawa et al., 2005). However, a portion of the photosynthate assimilated near the end of the growing season is sometimes stored and remobilized the following year in the formation of earlywood (Hansen & Beck, 1990), that is, there may exist a time lag of several months between the uptake of 13CO2 and its incorporation into rings. Another example is the English oak (Quercus robur). This deciduous, ring-porous tree forms large vessels before bud-break (Pilcher, 1995), and the earlywood is therefore thought to be formed primarily from stored material (Brown, 1971b). The high correlation of the δ13C of earlywood to that of the previous year's latewood reported by Hill et al. (1995) and Robertson et al. (1997) follows naturally. In general, deciduous trees are expected to use more stored carbohydrates for early season stem xylem production (earlywood formation) than evergreens (Kozlowski, 1992).

Many carbon allocation studies have been conducted with 14C labeling with the aim of understanding seasonal trends of carbon turnover; however, most of these studies have focused on short-term turnover and utilized chase periods (time elapsed since pulse-labeling) of <1 yr (Hansen et al., 1997). In order to apply the physiological knowledge gained from these studies to dendroclimatology work and, in particular, autocorrelation theory, long-term carbon allocation studies are necessary, with chase periods of up to a few years. The stable isotope 13C is an attractive alternative for ecological research as regulatory constraints associated with its use under field conditions are absent.

The first objective of our study was to clarify the carbon-turnover pattern of photosynthate assimilated at different times in the growing season (spring, summer and autumn) over longer chase periods (up to 3 yr) than preceding pulse-labeling studies. This is necessary to examine to what extent the trees rely on stored carbohydrates from previous year(s) for growth. Combined with high-resolution analysis of intra-annual distribution of the pulse-labeled 13C in rings of these trees (Kagawa et al., 2006), we anticipated that this work would enable us to understand the mechanisms behind autocorrelation observed in isotope dendroclimatogical studies (Lipp & Trimborn, 1991; Hill et al., 1995; Switsur et al., 1995; Robertson et al., 1997). The second objective of our study was the investigation of seasonal variations of carbon allocation patterns in Larix gmelinii, also known as Larix cajanderi Mayr. (Abaimov et al., 1998), in order to understand the characteristic allocation pattern of L. gmelinii as found in continuous permafrost regions.

Larix gmelinii forests account for 63% (215 million ha) of the total forested area in eastern Siberia, and play an important role as carbon reservoirs in the global carbon cycle (Tseplyaev, 1965; Schulze et al., 1999; Benkova & Schweingruber, 2004). The Larix on which our study focuses were located in a continuous permafrost zone in this area. Because of the continental climate, the growing season of Larix in this region lasts only approx. 3 months (June to August; Kagawa et al., 2001), and the spring, summer and autumn periods correspond roughly to June, July and August. Natural disturbances, such as frequent forest fires (Osawa et al., 1994), moth attacks on needles (Galkin, 1992; Gninenko & Sidelnik, 2003) and cold waves, make the unreliable photosynthetic production of L. gmelinii even more unreliable. In order to survive the extreme environment, the survival strategy of boreal trees is expected to include a particularly slow rate of carbon turnover. Furthermore, as sunlight is not a strong limiting factor for growth in boreal regions, and there is less need to reach the crown to receive sunlight, they tend to store carbohydrates for longer periods and consume less carbohydrate for above-ground structural growth compared with trees in temperate or tropical areas (Malhi et al., 1999). The extreme continental larch forests in eastern Siberia are often characterized as ‘light taiga’ because of their sparse crown density (Walter, 1985). Earlier investigations found that Larix in Siberia are characterized by higher percentages of carbon allocation to their below-ground parts, compared with the temperate Larix (Kanazawa, 1994; Kajimoto et al., 1999; Kajimoto et al., 2003). The slow carbon turnover of Siberian trees may cause a problem for future climate reconstruction studies in this area. In order to examine the long-term carbon turnover of stored carbohydrates and their use for growth in different tree parts, we set chase periods of up to 3 yr for the trees pulse-labeled with 13CO2 at the end of the growing season. Saplings of different sizes growing under natural conditions were used for a series of pulse-labeling studies on Siberian trees of different age classes.

Materials and Methods

Sample trees and pulse-labeling

We selected 10 saplings of L. gmelinii (Rupr.) Rupr. (Table 1) growing naturally in gaps created by fire disturbances in the Spasskaya Pad experimental forest of the Institute for Biological Problems of the Cryolithozone. The forest is located approx. 20 km north of Yakutsk City (62°15′ N, 129°37′ E) in a continuous permafrost zone. Trees were watered individually during the evening, and the following day were enclosed in individual chambers and pulse-labeled with 13CO2. The method of pulse-labeling is described further in our previous paper (Kagawa et al., 2005). Solutions of 98 at % Ca13CO3 (Isotec Inc., Miamisburg, OH, USA) and phosphoric acid were mixed to generate the 13CO2 gas. The chambers were made of 100 µm-thick polyethylene bags, 42 cm in diameter and 60 cm in height (volume 83 l). Smaller chambers (volume approx. 1 l) were used for pulse-labeling trees D and H. To ensure efficient incorporation of the label, pulse-labeling was conducted strictly on sunny days, and 13CO2 was injected twice in the morning, when CO2 assimilation rate is at its peak (Vygodskaya et al., 1997, A.S., unpublished data), and once in the afternoon. Each exposure lasted for 2 h. To prevent water condensation on needles and subsequent limitation of the CO2-incorporation rate, we used electric fans equipped with silica gel cartridges to circulate the interior air continuously. The amount of silica gel in the cartridge was adjusted to the minimum required to prevent condensation on the inner surface of the chambers, because a superfluity of silica gel has been shown to cause a low vapor pressure deficit and consequential closure of stomata (Vygodskaya et al., 1997). Regulation of the vapor pressure deficit is therefore recommended for future pulse-labeling experiments.

Table 1.  Data for pulse-labeled Larix gmelinii saplings
TreePulse-labeling dateSampling dateChase periodTotal tracer (ml 13CO2)Height (cm)Base diameter (mm)Biomass (g DW)
  1. At the time of sampling, tree C was found to be dead. Judging from microscopic observation and 13C analysis of the tree rings formed, the time of tree death was estimated to be around August 2001 (Kagawa et al., 2006). The chase period for July pulse-labeled trees D–F was calculated as the time elapsed from the middle of the two labeling dates.

A15–16/6/0119/6/01 4 d24036 5.4 × 5.81.431.040.941.620.540.650.35
B15–16/6/0114/8/0160 d24066.5 6.1 × 5.71.452.061.421.730.410.520.41
C15–16/6/0119/8/02 1.2 yr24065 8.4 ×
D13 & 31/7/00 5/9/0045 d 3.810 6.2 × 5.50.691.090.970.700.801.070.17
E13 & 31/7/0019/6/01 0.9 yr 3270 7.1 × 6.62.473.413.073.451.962.260.58
F13 & 31/7/0014/8/01 1.1 yr 3273 9.0 × 9.02.962.482.412.771.471.791.15
G15–16/8/01 1/8/02 1 yr24044 6.8 × 6.91.702.691.922.081.721.670.25
H21–22/8/0019/8/02 2 yr 3.850 9.6 × 9.60.794.482.933.430.820.870.21
I15–16/8/0110/8/04 3 yr24072.510.5 ×
J15–16/8/0112/8/04 3 yr24066 8.3 × 8.52.814.604.363.533.892.530.82

We pulse-labeled three trees (A–C) in June, another three (D–F) in July and four (G–J) in August, then sampled the trees after different chase periods ranging from 4 d to 3 yr. We also studied the intra-ring δ13C of these trees (trees and identification letters correspond to those given by Kagawa et al., 2006). We used total amounts of 3.8–32 ml 13CO2 (2.2–19 mg 13C) for the initial pulse-labeling experiments, but later increased the amount to 240 ml (140 mg 13C) to ensure sufficient incorporation of 13CO2 for detection (Table 1). For example, for trees A, B, C, G, I and J, 240 ml was separated into 40 ml × 3 (injections d−1) × 2 (d), yielding an expected initial concentration of 482 ppm 13CO2. The openings of the polyethylene bags were sealed to the stem bases with oil-based clay and wrapped with string. In this manner the chambers were kept airtight for 2 h after each injection. During recovery and sampling of trees, the roots were gently hand-washed to remove the soil while preserving as much of the root system as possible.

Grinding and starch extraction

After sampling, the trees were oven-dried at 70°C for 24 h and then separated into seven constituent tissue types: needles, branches, stem bark, stem xylem, root bark, root xylem, and fine roots (diameter <0.5 mm). After weighing the tissues, they were ground to 40 mesh (350 µm particle size) in a steel ball mill (Wig-L-Bug Model 30, International Crystal Laboratories, Garfield, NJ, USA) and homogenized (Vibrax mixer, IKA, Staufen, Germany). Starch was extracted from the tissues according to Jäggi et al. (2002). About 250 mg of each tree tissue was mixed with 5 ml 30% ethanol (3 min, Vibrax) and centrifuged at 10 000 g for 10 min to remove polar substances. This process was repeated twice. To remove nonpolar substances, each pellet was then mixed with 5 ml methanol/chloroform/water (12 : 5 : 3) and centrifuged in the same manner. The supernatants were removed, and this process was repeated at least three times until the supernatant was clear. The remaining pellets were each washed in 30% ethanol, then mixed three times with 2 ml, 2 ml and 1 ml of 20% HCl in order to solubilize starch. Supernatants of HCl extractions were collected in a syringe and filtered. Next, 20 ml 100% ethanol was added (the ethanol concentration of the final solutions was >80%) and samples were then refrigerated overnight, causing precipitation of hydrolysed starch. This mixture was centrifuged, the supernatants carefully decanted, and the remaining pellets (gelatinized starch) dissolved in water, frozen and freeze-dried. Each powdered sample was then extracted again. Starch was weighed and expressed as a percentage on a dry weight basis (%) or total amount in each tree part (mg). The two starch samples taken from the same tree and tissue type were mixed and homogenized before carbon isotope analysis.

Carbon isotope analysis

The combined system of an elemental analyser (NC 2500; CE Instruments, Milan, Italy) and an isotope ratio mass spectrometer (MAT252; Thermo Electron, Bremen, Germany) was used for δ13C measurements. Approximately 2 mg prepared tissue or 1 mg prepared starch was weighed out for each analysis, and five standards were used for calibration of the data (C-13 labeled UL-glucose IAEA-309 (A) +535.3, C-13 labeled UL-glucose IAEA-309 (B) +93.9, l-glutamic acid USGS 41 +37.8, International Atomic Energy Agency, Vienna, Austria; l-histidine −9.6, Shoko Co. Ltd, Tokyo, Japan; dl-alanine −23.5, Center for Ecological Research, Otsu, Japan). Two aliquots of each tissue sample were analysed to confirm the homogeneity of the ground samples. During the analysis of several samples with high δ13C, values registered beyond the scale. In these cases, another 0.5 mg tissue was weighed and analysed again. The standard deviation for replicate combustions of the dl-alanine standard was 0.08. Excess δ13C values () were calculated as deviations from the averaged baseline δ13C values of the tree rings formed before pulse-labeling of each individual sapling. For calculation of the baseline values, the intra-annual δ13C values of both the tree-ring part formed before pulse-labeling in the current year (if present), and the previous year's tree ring, were averaged. To accomplish this, a rotary microtome (HM340E; Microm International, Walldorf, Germany) was used to take serial tangential sections from the tree rings and the δ13C of each section was analysed later (Kagawa et al., 2006). Natural fluctuation of tree-ring δ13C of eight L. gmelinii trees at this site was 1.1 (±1σ) during 1996–2000 (Kagawa et al., 2003).

Excess 13C was also calculated according to procedures described by Boutton (1991) and Simard et al. (1997) as follows: each δ13C value was first converted to the absolute isotope ratio of the sample (R):

Rsample = [(δ13C/1000) + 1] × Rstd(Eqn 1)

where Rstd = 0.0112372, the carbon isotope ratio of the international VPDB standard. Then the 13C abundance ratio (A) was calculated as follows:

A = [13C/(13C + 12C)] = (R/R + 1)(Eqn 2)

Total 13C content of the sample was calculated from 13C abundance ratio (A) and total carbon weight of each tree part:

mg 13Csample = A × (13C + 12C)(Eqn 3)

13C in excess of natural abundance (mg 13Cna) was calculated as:

excess mg 13Csample = mg 13Csample − mg 13Cna(Eqn 4)

where mg 13Cna was calculated from the baseline δ13C value of each individual sapling and used for calculating excess mg 13Csample of all tissue types. To account for the different amounts of 13CO2 supplied to the sample trees (Table 1), the results are expressed in percentages of bulk or starch excess mg 13C in each tree part against total excess mg 13C in the whole tree (hereafter referred to as excess 13C).


Allocation of June photoassimilate

A significant amount of photoassimilate is known to be allocated to sugar as well as starch (Schneider & Schmitz, 1989). Boreal L. gmelinii has an especially high needle sugar content compared with, for example, temperate Larix decidua (Hoch et al., 2003), and sugars account for seven times the amount of starch in boreal L. gmelinii needles (Sudachkova et al., 2001). Four days after pulse-labeling, 64% of the total pulse-labeled 13C detected in the tree still remained in needles (Fig. 1a, tree A) and 3.7% was measured in the needle starch fraction. Pulse-labeling efficiency (ratio of excess mg 13C in the whole tree to mg 13C injected into the labeling chamber) was 20% for tree A. A decrease in needle excess 13C was found after a 60-d chase period, and branch and stem proportion (Br + Sb + Sx) increased over the same time period (Fig. 1b, tree B), suggesting export of photoassimilate from needles. At 60-d chase, a total of 37% had been exported to the stem and roots (Sb + Sx + Rb + Rx + Rf, Fig. 1b). Schneider & Schmitz (1989) observed a similar proportion (46%) of 14C-labeled spring photoassimilate exported at an earlier stage (7-d chase) from the labeled branch to the stem and roots of L. decidua growing in a temperate region, suggesting faster carbon turnover of temperate Larix compared with boreal Larix. Tree C died of unknown causes in August 2001, and all the needles had been shed by the time of sampling. This tree showed the lowest starch fraction among the 10 trees sampled, and we believe it died of starvation (Fig. 1c). Apart from needles and branches, a significant proportion of June photoassimilate was also detected in the stem part after a 60-d chase period (Sb + Sx = 33 and 47% for trees B and C, respectively, Fig. 1b,c), signifying export of photoassimilate to the stem, especially to the xylem.

Figure 1.

Allocation of June pulse-labeled 13C, fraction, and amount of starch in each tree part; needles (N), branches (Br), stem bark (Sb), stem xylem (Sx), root bark (Rb), root xylem (Rx), fine roots (Rf) of Larix gmelinii saplings. Allocation of pulse-labeled 13C is expressed as percentage of excess mg 13C in each tree part. Open and closed bars, percentage of bulk and starch excess 13C present in each tree part, respectively; open and closed circles, fraction of starch per tissue dry weight and absolute amount of starch (×10 mg) in each tree part, respectively. June photoassimilate was mainly allocated to above-ground parts. Tree C died of unknown causes and lacked needles when sampled.

In contrast, only a very small portion of June photoassimilate was allocated to below-ground parts (Rb + Rx + Rf = 4.2 and 2.8% in trees A and B, respectively, Fig. 1a,b). Even 1.2 yr after pulse-labeling, the below-ground allocation ratio remained low (Rb + Rx + Rf = 7.9%, tree C, Fig. 1c). In tree A, a large difference in starch content was observed between bark and xylem. The fraction and amount of starch in bark (Sb + Rb) was 3.7 and 2.4 times higher, respectively, than those in xylem (Sx + Rx). However, such a difference was weak or even absent in trees B and C (Fig. 1).

Allocation of July photoassimilate

The fraction and amount of starch in tree D needles were lowest (0.8% and 0.54 × 10 mg) among all needles sampled. This may partly explain the fact that, despite high values of bulk excess 13C in tree D needles, the excess 13C of the needle starch fraction (0.2%, Fig. 2a) was low compared with that of June pulse-labeled tree B (0.8%, Fig. 1b). When tree D was sampled on 5 September, most of the needles had turned yellow and the tree appeared to be on the verge of shedding needles. The low fraction and amount of starch and the low percentage of excess 13C in the starch fraction of tree D reflect resorption of carbohydrates from the needles in autumn (Lal et al., 2001). Of all excess 13C found in the tree D starch pool, only 3% existed in the needle starch fraction. The new needles of trees E and F contained significant amounts of pulse-labeled 13C from the previous year (Fig. 2b,c), verifying the use of carbohydrates stored during the previous summer for the development of new needles. There was less carbon allocated to stem parts, especially to stem xylem, compared with June pulse-labeled trees B and C (Fig. 1b,c).

Figure 2.

Allocation of July pulse-labeled 13C to each tree part of Larix gmelinii saplings, and fraction and amount of starch. Open and closed bars, percentage of bulk and starch excess 13C present in each tree part, respectively; open and closed circles, fraction of starch per tissue dry weight and absolute amount of starch (×10 mg) in each tree part, respectively. Higher allocation of July photoassimilate to below-ground parts was observed.

July pulse-labeled trees (D–F) allocated much higher proportions of photoassimilate to below-ground parts (Rb + Rx + Rf = 32, 44, 33% in Fig. 2a–c, respectively) than June pulse-labeled trees (t-test, P = 0.0047). As already observed in the case of tree A, a higher fraction and amount of starch were observed in bark compared with xylem for all three of these trees.

Allocation of August photoassimilate

As observed in trees E and F, the needles of tree G were found to contain photoassimilate from the previous year, which accounted for 42% of total excess 13C (Fig. 3a). However, after a 2–3-yr chase period, trees H–J showed little excess 13C left in the needles. Despite a high percentage (5.0%, Fig. 3b) and absolute value (25, data not shown) of excess δ13C in the starch fraction of tree H branches, only a small percentage (1.9%) and absolute value of excess δ13C (3.7, data not shown) were observed in tree H needles. This suggests that needles rely mostly on stored material carried over from the previous year, while only a small amount of material is carried over for >2 yr.

Figure 3.

Allocation of August pulse-labeled 13C, and fraction and amount of starch. Open and closed bars, percentage of bulk and starch excess 13C present in each tree part, respectively; open and closed circles, fraction of starch per tissue dry weight and absolute amount of starch (×10 mg) in each tree part, respectively. The highest allocation of August photoassimilate was found in branch parts, except for tree G.

In contrast, significant amounts of excess 13C still remained in non-needle parts after a 2–3-yr chase period (Fig. 3b–d). The highest percentages of excess 13C were detected in the branches of these three trees, suggesting labeled carbohydrate deposition in the branches in autumn and use of the stored material for shoot elongation and needle flushing during the following year. Allocation to stem parts, especially to xylem, remained low (Sx = 2.0–6.4%) except for tree H (Fig. 3a,c,d), again reflecting slow xylem growth in August (Kagawa et al., 2001). The higher allocation to stem xylem in tree H (Fig. 3b) may reflect the active compression wood formation observed in the latewood formed in 2000 (Kagawa et al., 2006). We also observed a lower percentage of excess 13C in the below-ground parts of August pulse-labeled trees (Rb + Rx + Rf = 17, 12, 12 and 24% for trees G–J, respectively, Fig. 3) than July pulse-labeled trees (P = 0.014). All 10 sample trees showed higher fractions and amounts of starch in bark than in xylem parts (Sb + Rb > Sx + Rx), a phenomenon that has been observed in the same species in central Siberia (Milyutina et al., 1998).

Decrease of excess δ13C of bulk and starch samples over chase time

A comparison of excess δ13C values of the trees pulse-labeled with the same amount of 13CO2 (trees A, B, G, I, J, Table 1) reveals that excess δ13C () decreases over time in both bulk and starch samples (Fig. 4). Needles showed the fastest carbon turnover of all tree parts. This is not surprising as needles are replaced every year and, as mentioned earlier, export of starch from needles was observed in 4- to 60-d chase periods (Fig. 1a,b).

Figure 4.

Long-term turnover of pulse-labeled 13C in bulk and starch samples. 60% of starch was replaced with new carbon each year. The needle starch pool showed faster carbon turnover (75% yr−1) than non-needle parts (57% yr−1).


Carbon turnover of spring photoassimilate

Needle and shoot elongation and earlywood formation in L. gmelinii usually occur in early to mid-June in the study region (Kagawa et al., 2001). In spring, when such vigorous needle and shoot growth occurs, trees generally exhibit acropetal phloem translocation (Gordon & Larson, 1968; Kozlowski, 1992; Hansen & Beck, 1994), and temperate Larix is no exception (Schneider & Schmitz, 1989). This explains the high allocation of June photoassimilate to needles and branches, as observed in our study. The quick decrease of spring photoassimilate in needles within the first 2 months after pulse-labeling (Figs 1a,b, 4a) can also be explained by the consumption of labeled photoassimilate caused by higher respiratory demand during times of vigorous growth (Kozlowski, 1992). A large loss of photoassimilate (35–65%) has also been reported within the first month of labeling, and associated with respiration in young Pinus strobus (Ursino et al., 1968). In our study, apart from needles and branches, most of the spring photoassimilate was allocated to stem parts, especially the xylem. June corresponds to the earlywood formation period in the study region (Kagawa et al., 2001), and the high stem proportion reflects active stem growth in spring. It is already known that earlywood formation partly relies on current photoassimilate (Smith & Paul, 1988; Jäggi et al., 2002; Kagawa et al., 2006). Active needle, shoot and stem growth in spring can explain the observed higher allocation of spring photoassimilate to above-ground parts.

Preferential allocation of spring photoassimilate to above-ground parts necessarily causes lower allocation to below-ground parts. In temperate trees there is a seasonal periodicity of root extension, with bimodal maximum extension usually occurring in the spring and autumn (Brown, 1971a; Hansen et al., 1997; Oleksyn et al., 2000). Root cambial activation is closely related to the surrounding soil temperature (Brown, 1971a; Abaimov et al., 1997). The rapid incorporation of labeled carbon caused by active root growth in spring, often observed in temperate trees (Hansen et al., 1997), seems unlikely in the continuous permafrost area of our study. The average soil temperature at the study site in middle–late June is a scant 6.3°C at 0–20 cm depth (Sugimoto et al., 2003), where 80% of L. gmelinii roots are distributed (Kotake & Kubota, 1999; Kuwada et al., 2002). Significant opportunity for root growth did not arise until mid- to late July, when the average soil temperature increased to 13°C at a depth of 0–20 cm. This offers one possible explanation for the particularly low allocation of spring photoassimilate to below-ground parts. Furthermore, boreal trees have little need for root development in spring, as plenty of water is immediately available from snowmelt.

Carbon turnover of summer and autumn photoassimilate

It is known that temperate Larix recovers a higher portion (77%) of nutrients from needles in autumn than other deciduous trees (52%) (Grower & Richards, 1990); like the temperate Larix, L. gmelinii was found to recover starch from needles in autumn efficiently, as reflected in the analysis of tree D needles. Starch stored in the branches was then used for new needle formation the following spring. Tree E needles, which were using 13C-labeled stored carbohydrate for expansion at the time of sampling, had excess δ13C of 89, while that of branch starch was 175 (data not shown). This fact demonstrates that, before becoming autotrophic, the newly expanding spring needles drew carbon from stored carbohydrates in the branches to which they were attached. Assuming the excess δ13C of storage material stayed constant (175) while the tree was developing new needles in June 2001 leads us to conclude that these new needles drew about half (89/175) their carbon from previously incorporated photoassimilate.

The high allocation of labeled photoassimilate to below-ground parts in July can be explained by three factors. First, by the beginning of July radial growth reached 80% of the annual increment, and mid-July is known to correspond to the latewood formation period (Kagawa et al., 2001). Therefore radial growth is slow, there is a lower demand for carbohydrate at the stem in summer and autumn, and part of the photoassimilate is allowed to pass through the stem to the roots (Ursino et al., 1968; Hansen & Beck, 1994; Hansen et al., 1997). Second, in July 2000 the thaw depth of permafrost at the study site reached >80 cm and soil temperature in the active layer was at a maximum (Sugimoto et al., 2003). Higher allocation of 13C to fine roots (Rf, Fig. 2) was observed in July-labeled trees compared with June- and August-labeled trees, and development of white-colored fine roots was observed in other L. gmelinii at the study site in July. It therefore seems likely that carbon was drawn below-ground for the support of active root growth. Third, even more carbon may have been transported below-ground for the deposition and accumulation of storage material, which is also observed in the summer and autumn seasons in temperate areas (Smith & Paul, 1988). It is known that accumulation of storage material begins in roots and then extends distally up through the stem into the branches (Zimmermann, 1971). This may explain the decrease in allocation ratios to below-ground parts from July to August, and the increase in allocation ratios to branches during the same period.

Overall, there was no noticeable redistribution of pulse-labeled 13C among tree parts over chase periods of >1 yr. This observation is in accordance with the fact that once photoassimilate is converted to storage materials such as starch and deposited in parenchyma cells, the transport of remobilized material usually happens via rays towards the cambium, as necessary. Stored material is available more-or-less all along the axis, and the sink organs utilize the nearest carbohydrate sources available (Zimmermann, 1971). Therefore, apart from some exceptions such as mass fruiting, only minor longitudinal transport of stored carbohydrate is necessary. Schier (1970) has also found little redistribution of 14C among non-cell-wall fractions of Pinus resinosa after 2–3-month chase periods: the relative growth rate in each tree part at the time of labeling is the critical factor in determining in which tree parts labeled carbohydrate reserves are eventually stored (Schier, 1970).

Significant allocation of pulse-labeled 13C not only to stem xylem, but also to stem bark, was observed in all pulse-labeled trees. This is reasonable if we consider that L. gmelinii develops very thick stem bark as a defense against frequent forest fires, and shows a higher percentage of bark area in stem cross-section (30%) than, for example, Larix laricina, a boreal larch species in North America (12%) (Sakai, 1973).

Turnover of pulse-labeled 13C as a function of time

As trees A and B were pulse-labeled in June, and G, I and J in August (Table 1), the decrease of excess δ13C between trees A and B and trees G, I and J (Fig. 4) could be attributed partly to the difference in photosynthetic rates between June and August. In August, photosynthetic rate may be limited by water stress in this region (Kuwada et al., 2002); however, such a limitation is absent if the trees are well watered (A.S., unpublished data). Compared with the observed 10–100-fold difference in excess δ13C, we expect the difference in photosynthetic rates to be small in our study, especially as we watered the sample trees before pulse-labeling. We therefore did not correct the δ13C excess data for different photosynthetic rates.

Although there was no noticeable redistribution of percentage excess 13C among tree parts after a 2-month chase period, pulse-labeled carbon was constantly replaced by newly fixed carbon. Part of the decrease in Fig. 4 is associated with respiratory consumption (Lambers, 1985; Kozlowski, 1992), while the rest we attribute to the loss of frequently replaced parts, such as needles and fine roots, and to carbon allocation to mycorrhiza (Qu et al., 2004). Needle and branch starch excess δ13C values at 60-d chase were 14 and 48% of those at 4-d chase (Fig. 4b), suggesting export of labeled photoassimilate from needles to branches within 2 months, on top of respiratory losses. Natural δ13C values of soluble sugars and starch of leaves are known to represent chiefly recently fixed carbon. Sugars reflect an integration timescale of carbon isotope fractionation of approx. 1 d, but starch carbon turns over more slowly with an integration time of up to 10 d in Populus (Brugnoli et al., 1988) and 1 month in Picea abies (Jäggi et al., 2002), similar time scales to our findings.

The rate of decrease did not differ significantly between bulk and starch samples (P = 0.323). This reflects a high percentage of allocation of pulse-labeled 13C to storage material and frequently replaced parts (needles and fine roots), and a low percentage of allocation to annual structural growth. Carbon allometry studies show that more than half (56%) the total net primary production of L. gmelinii goes to needle production and about a quarter (27%) to below-ground production, of which 35% is allocated to fine roots (Kajimoto et al., 1999). High allocation to frequently replaced parts thus leaves little carbon for structural growth, as evidenced by the narrow tree rings of L. gmelinii in the study area.

Larix shows the highest extractive content (16%) compared with five other conifer genera (<7%) studied by Migita et al. (1968), and a higher allocation of pulse-labeled 14C to low molecular-weight substances such as sugars has been observed in L. decidua (Schneider & Schmitz, 1989). In temperate Larix, starch accounts for 35% of nonstructural carbohydrates (Hoch et al., 2003). However, low-temperature environments favor conversion of starch into sugars or fat (Sakai, 1962), as observed in boreal L. gmelinii (Sudachkova et al., 2001). This fact may account for the low excess 13C in starch fractions which we observed in east Siberian L. gmelinii (Figs 1–3).

If a constant carbon-turnover rate in the starch pool is assumed for every growing season (T, yr−1), excess 13C in the starch pool (A, mg 13C) after a given chase period (t, yr) can be expressed as:

A  =  A0(1 − T)t(Eqn 5)

where A0 is the initial excess 13C. The turnover was thus calculated for trees G, I and J, which were pulse-labeled on the same day. This revealed that 60% (= T) of starch was replaced with new carbon each year. The needle starch pool showed faster carbon turnover (T = 75% yr−1) than the pools of non-needle parts (T = 57% yr−1). In other words, 43% of the starch that can be used for xylem formation was composed of the previous year's photoassimilate. Our estimates of starch carbon turnover are summarized in the model tree shown in Fig. 5. The residual (27%) in the figure can be attributed to respiration, loss of fine roots, carbon allocation to mycorrhiza (Qu et al., 2004), etc.

Figure 5.

A model of interannual carryover of starch in Larix gmelinii saplings estimated from excess 13C. After tabulating the percentages for carried-over 13C in the starch pool, needle litter loss, and new needle formation, the loss of 13C caused by respiration, fine root loss, etc. was estimated as the residual. Allocation of 13C to needle formation in the following year was estimated from tree G data.

Stored material is likely to be used for earlywood formation in spring, especially at the beginning of the growing season, while in summer and autumn current photoassimilate is likely to be used for latewood formation. This is because acropetal phloem translocation prevails in spring and the photoassimilate is used preferentially for formation of needles and shoots, while in summer and autumn the phloem translocation shifts to the basipetal direction (Gordon & Larson, 1968; Smith & Paul, 1988; Kagawa et al., 2005; Kagawa et al., 2006). This reasoning is in accordance with the observation of seasonal change of needle starch δ13C and earlywood/latewood δ13C by Jäggi et al. (2002), who found that the δ13C signature of earlywood is closely related to needle starch δ13C, while that of latewood is influenced by environmental conditions.

Growth shows different phenological patterns for needles, shoots, stem and roots (Brown, 1971a, 1971b), and source–sink relationships are constantly changing. This causes changes in the direction of phloem translocation and degree of dependence of cambium on stored material for tree-ring formation (Zimmermann, 1971; Kozlowski, 1992; Hemming et al., 2001). Accordingly, we evaluated the carbon allocation patterns seen in this study in the light of high-resolution δ13C analysis of tree rings from the same sample trees (Kagawa et al., 2006), in order to detail how the phenology in carbon allocation is related to tree-ring isotope ratios. Hemming et al. (2001) have modeled similar responses. Our study revealed that latewood derived its carbon mainly from current photosynthesis, while earlywood relied on both carried-over photoassimilate from the previous year and current photoassimilate (Kagawa et al., 2006). This interannual carry-over of photoassimilate has great significance for isotope dendroclimatology, and is intimately connected to autocorrelation phenomena. However, we stress that carbon allocation patterns of mature trees may be different from saplings such as those used in this study, and there is a need for similar 13CO2-labeling studies of mature trees (such as the currently ongoing study by Helle & Panferov, 2004) – with which (isotope) dendroclimatology is primarily concerned.


The authors thank the staff of the Institute for Biological Problems of the Cryolithozone for helping with the field observations, and appreciate the cooperation with all members of GAME/Siberia. We would also like to thank Andrea Polle and the anonymous reviewers for taking time to give us valuable suggestions for this manuscript. We further extend our thanks to Jennifer Lue for reviewing early drafts of this paper, Maya Jäggi for providing us with detailed starch-extraction methods, and Takayoshi Koike for giving us helpful suggestions. This study was supported by Grant-in-Aid 16403011, 16780119 and 11554017 from the Ministry of Education, Culture, Sports, Science and Technology, Japan.