Post-photosynthetic fractionation processes during translocation, storage and remobilization of photoassimilate are closely related to intra-annual δ13C of tree rings, and understanding how these processes affect tree-ring δ13C is therefore indispensable for improving the quality of climate reconstruction. Our first objective was to study the relationship between translocation path and phloem grain. We pulse-labelled a branch of Larix gmelinii (Rupr.) Rupr. and later analysed the δ13C distribution in the stem. A 13C spiral translocation path closely related to the spiral grain was observed. Our second objective was to study the use of remobilized storage material for earlywood formation in spring, which is a suspected cause of the autocorrelation (correlation of ring parameters to the climate in the previous year) observed in (isotope) dendroclimatology. We pulse-labelled whole trees to study how spring, summer and autumn photoassimilate is later used for both earlywood and latewood formation. Analysis of intra-annual δ13C of the tree rings formed after the labelling revealed that earlywood contained photoassimilate from the previous summer and autumn as well as from the current spring. Latewood was mainly composed of photoassimilate from the current year’s summer/autumn, although it also relied on stored material in some cases. These results emphasize the need for separating earlywood and latewood for climate reconstruction work with narrow boreal tree rings.
Tree-ring parameters such as ring widths, density and stable isotope ratios have long been used as environmental proxy record by many investigators (Fritts 1976; Schweingruber 1987; Epstein & Krishnamurthy 1990; Leavitt 1993a; McCarroll & Loader 2004). However, our physiological understanding about tree-ring formation is still insufficient to adequately explain the observed climate–ring parameter correlation. Improved understanding of the limitations and potential problems associated with the use of tree rings as a climate proxy is needed to make climate reconstruction based on tree rings more reliable. For example, Kagawa et al. (2005) have shown a limit on the time resolution of tree-ring isotope proxy in a study that revealed that tree-ring δ13C levels do not offer climatic information for time periods of less than 1 week.
Another limitation of tree rings as a climate proxy is autocorrelation, or correlation of ring parameters to the climate in the previous year(s) (Fritts 1976), which can interfere in climate reconstruction attempts. Among other proposed causes for autocorrelation, such as effects of climate on formation of buds, leaves, roots, fruits, hormones (Fritts 1976) or availability of nutrients (Jarvis & Linder 2000; Vaganov & Hughes 2000), the use of stored carbohydrates has been suggested as the most likely cause (Kozlowski 1992). In fact, first- and second-order autocorrelations are frequently also observed in isotope dendroclimatological studies (Monserud & Marshall 2001). One example is the high correlation reported between the δ13C of earlywood and previous year’s latewood in Quercus robur (Hill et al. 1995; Robertson et al. 1997). Quantitative information on how much photoassimilate is stored in a previous year, carried over, and then remobilized in the spring for tree-ring formation is necessary for understanding autocorrelation and improving the interpretation of (isotope) dendroclimatological data.
Two structural phenomena are of particular interest. Firstly, translocation path is closely related to the grain (the arrangement of phloem sieve cells with respect to the tree axis) of a tree species. For example, in straight-grained Cryptomeria japonica, pulse-labelled 13C fumigated to a branch is translocated straight down the stem (Kagawa et al. 2005), while in spiral-grained Pinus sylvestris, a spiral translocation path of pulse-labelled 14C has been observed (Hansen & Beck 1990). Such differences in translocation paths may cause discrepancies among δ13C chronologies obtained from different stem directions or tree species with different grains. Sieve cells in phloem function as a conducting system for photoassimilate; they are produced from the same cambium initials that produce xylem cells on the other side. Recently formed phloem and xylem cells close to the cambium should therefore be arranged parallel to each other (Zimmermann & Brown 1971) and the grain of wood should be reflected by phloem cells (Heijnowicz 1964).
Secondly, there is a time lag between the photosynthetic uptake and the use of the photosynthate for tree-ring formation in the stem (Schleser et al. 1999), and this time lag may differ among photoassimilates incorporated in different seasons and trees living in different climatic zones. In fact, a fast-growing plantation species (C. japonica) in a temperate region consumes spring pulse-labelled 13CO2 in 1 month or less (Kagawa et al. 2005). In autumn, however, a portion of the photoassimilate can be stored and then remobilized for the following year’s earlywood formation, thereby extending the time lag by several months. The high correlation between δ13C of earlywood and the previous year’s latewood observed in Q. robur (Hill et al. 1995; Robertson et al. 1997) is explained by the fact that deciduous ring-porous species form large vessels before bud break, and therefore the earlywood must rely at least partly on stored carbohydrates (Brown 1971). We will henceforth refer to tree rings formed in the year of, prior to and following pulse-labelling as the current, previous and following rings, respectively.
Compared with deciduous tree species, evergreens are less likely to use stored carbohydrates for early season xylem formation (Kozlowski 1992). Because acropetal phloem translocation prevails in spring and spring photoassimilates are preferentially used to produce needles and shoots, stored material is likely to be used for earlywood formation in the stem, while in summer and autumn, phloem translocation shifts to the basipetal direction (Gordon & Larson 1968; Smith & Paul 1988) and current photoassimilate is likely to be used for latewood formation. This argument is in accordance with the observation 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 largely influenced by climatic conditions.
Branch-level 14C labelling and autoradiography of Pinus sylvestris has proved the use of photoassimilate from the previous year for earlywood formation (Hansen & Beck 1990). After bud break, 14C-labelled photoassimilate from October is used for the formation of the first layers of earlywood, while late earlywood is produced from freshly incorporated photoassimilate. However, exactly how much 14C is actually used for the following earlywood formation remains unclear, as 14C allocated to radially mobile substances such as resin and storage carbohydrates in ray parenchyma would also have contributed to the autoradiograph. Unfortunately, other preceding carbon allocation studies conducted with 14C labelling have focused on short-term carbon turnover with chase periods of less than 1 year (Hansen et al. 1997) and interannual carbon carry-over, which has great significance for dendrochronology, remained unexplored. In order to improve our physiological knowledge on (isotope) dendroclimatological issues such as autocorrelation, a long-term carbon allocation study is necessary, with chase periods of up to a few years.
The magnitude of the dependence on previously incorporated storage material for tree-ring formation may also vary according to climate zone (Kozlowski 1992). We accordingly chose the study site in boreal zone. Larix gmelinii (Rupr.) Rupr. is distributed in eastern Siberia [some systems classify this species as Larix cajanderi Mayr. (Abaimov et al. 1998)] and plays an important role in the global carbon cycle (Tseplyaev 1965; Schulze et al. 1999; Benkova & Schweingruber 2004). The growing season of L. gmelinii in continuous permafrost regions of eastern Siberia lasts only about 3 months (Kagawa, Naito & Sugimoto 2001), and the spring, summer and autumn periods correspond roughly to June, July and August, along with the short growing season, natural disturbances such as frequent forest fires (Osawa, Maximov & Ivanov 1994), moth attack of needles (Galkin 1992; Gninenko & Sidelnik 2003) and cold waves, make photosynthetic production of this species very unreliable. In order to survive the extreme environment, the survival strategy of boreal trees such as L. gmelinii is expected to include a particularly slow rate of carbon turnover. Boreal coniferous forests have a very low net primary production compared to temperate and tropical forests (Gower et al. 1997; Jarvis et al. 1997; Malhi, Baldocchi & Jarvis 1999). Indeed, L. gmelinii keeps stored carbohydrates for longer periods and consumes less carbohydrates for annual structural growth of the above-ground part compared with temperate Larix (Kajimoto et al. 1999; Kagawa, Sugimoto & Maximov 2006), because sunlight is not a strong limiting factor for growth in boreal regions and there is less need to reach the crown for receiving sunlight. Earlywood formation of L. gmelinii is consequently very likely to involve the usage of stored material, and therefore L. gmelinii is an ideal candidate for the study of interannual carbon carry-over.
As a preliminary study, we measured the 13C content in the needles, branches, stem and roots of pulse-labelled L. gmelinii trees to study the long-term carbon allocation to different body parts and also the long-term carbon turnover in the starch pool (Kagawa et al. 2006). We found that 43% of carbon in the starch pool of L. gmelinii is carried over to the following year, while autocorrelation analysis of δ13C time series of temperate conifers revealed that approximately 30% of the current-year value depends on the previous year (Monserud & Marshall 2001). The spring (June) photoassimilates of L. gmelinii were mainly allocated to above-ground parts, especially to the stem xylem, while summer (July) and autumn (August) photoassimilates went to both above- and below-ground parts, with a lesser amount allocated to the stem xylem.
Our current study was conducted with two chief objectives. The first objective was to examine whether the translocation path of photoassimilate is related to the spiral grain of L. gmelinii phloem cells. This was carried out by pulse-labelling one branch of L. gmelinii and analysing the circumferential distribution of the pulse-labelled 13C at different stem heights. The second objective was to examine how the seasonal carbon turnover difference observed in our preliminary study is related to intra-annual variation of tree-ring δ13C. This was carried out by conducting a high-resolution δ13C analysis of the same 13CO2 pulse-labelled saplings used in our preliminary study (Kagawa et al. 2006). Especially, we focused on clarifying to what extent the trees rely on stored carbohydrates from previous year(s) for current ring formation, which has been suggested to be a major physiological mechanism behind the autocorrelation observed in (isotope) dendroclimatogical studies (Fritts 1976; Lipp & Trimborn 1991; Lipp et al. 1991; Hill et al. 1995; Switsur et al. 1995; Robertson et al. 1997; Monserud & Marshall 2001). In order to observe interannual usage of stored carbohydrates for tree-ring formation, we set chase periods of up to 3 years for the trees pulse-labelled at the end of the growing season (August).
MATERIALS AND METHODS
We studied 10 saplings of L. gmelinii (Rupr.) Rupr. (Table 1), naturally growing in gaps created by fire disturbances. The study site was located about 20 km north of Yakutsk city (62°15′N, 129°37′E), in the Spasskaya Pad experimental forest of the Institute for Biological Problems of Cryolithozone in a continuous permafrost zone. The trees and tree-identification letters correspond to those in the paper of our preliminary study (Kagawa et al. 2006).
Table 1. Sample tree data
5.4 × 5.8
6.1 × 5.7
8.4 × 8.3
6.2 × 5.5
13 & 31/7/2000
7.1 × 6.6
13 & 31/7/2000
9.0 × 9.0
13 & 31/7/2000
6.8 × 6.9
9.6 × 9.6
10.5 × 9.2
9.0 × 8.9
Total tracer used (mL 13CO2)
Baseline value (‰)
Starch excess δ13C (‰, stem)
Starch excess δ13C (‰, stem + roots)
Tree C was found to be dead at the time of sampling. Judging from microscopic observation and 13C analysis of the tree rings formed, the time of the tree death was estimated to be around August of 2001. Chase periods for July pulse-labelled trees D, E & F were calculated as the time elapsed from the middle of the two labelling dates.
The trees were watered the evening before labelling. On the days of labelling, each whole tree was enclosed in a cylindrical chamber into which CO2 gas was injected. The chambers were 100 µm-thick polyethylene bags, 42 cm in diameter and 60 cm in height (volume 83 L). Smaller chambers (volume approximately 1 L) were used for pulse-labelling trees D and H. The interior air was circulated while water vapour was removed by silica gel cartridges throughout pulse-labelling. 13CO2 was produced by mixing 98 atom per cent Ca13CO3 (Isotec Inc., Miamisburg, OH, USA) with phosphoric acid and was injected in three equal volumes on each day of pulse-labelling. For example, 240 mL (Table 1) was separated into 40 mL × 3 (injections d−1) × 2 (d), yielding an expected initial concentration of 482 ml L−1 of 13CO2. Trees A, B and C were pulse-labelled in June; D, E and F in July; and G, H and I in August. They were sampled after chase periods ranging from 4 d to 3 years and immediately oven-dried at 70 °C for 24 h. Initially, we injected total amounts of 3.8–32 mL 13CO2 for the two-day pulse-labelling (trees D, E, F, H and S). However, the amount was later increased to 240 mL (trees A, B, C, G and I) to ensure sufficient incorporation of 13CO2 for detection in tree rings (Table 1). Further details of the pulse-labelling apparatus are described in Kagawa et al. (2005).
The tenth tree, tree S, was reserved for examination of phloem translocation. A branch of tree S directly attached to the stem at 68 cm above ground was enclosed in a plastic bag and pulse-labelled in the same way as the other trees.
Spiral grain measurement and analysis of circumferential 13C distribution in stem and branches
After a 45 day chase period, each branch of tree S was numbered and removed from the stem, oven-dried, and ground to 40 mesh (350 µm particle size) in a steel ball mill (Wig-L-Bug Model 30; International Crystal Laboratories, Garfield, NJ, USA). For examination of the spiral grain of phloem cells, we followed the method of Kozlowski, Hughes & Leyton (1967). We first removed the bark to expose cells near the cambium and then pulled the slivers off the surface originating from the base of the 13CO2 fumigated branch. The slivers were pulled off parallel to the spiral grain and a continuous phloem cell alignment line was drawn from the base of the pulse-labelled branch (at 68 cm above ground) down to the base of the stem. Then 13 stem disks of 1 mm thickness were taken at the heights of 10, 5 and 1 cm above, and 1, 5, 10, 15, 20, 25, 30, 35, 40 and 45 cm below the pulse-labelled branch. The stem xylem diameter ranged from 4.2 to 6.1 mm at distances of 1–45 cm, respectively, below the pulse-labelled branch. After the removal of the bark, each disc was divided into eight fan-shaped pieces with a base angle of 45° as shown in Fig. 1. In order to determine the phloem translocation path, 13C in each piece was analysed as described later.
After sampling, trees A–I were partitioned according to body part: needles, branches, stem bark, stem xylem, root bark, root xylem and fine roots (diameter < 0.5 mm). These parts were ground to 40 mesh in the same manner as tree S and then homogenized (Vibrax mixer; IKA, Staufen, Germany). Starch was extracted from the tissues according to Jäggi et al. (2002) as follows: about 250 mg of each body tissue was mixed with 5 mL of 30% ethanol and centrifuged to remove polar substances. This process was repeated twice. To remove non-polar substances, each pellet was then mixed with 5 mL of methanol/chloroform/water (12 : 5 : 3) mixture and centrifuged. This process was repeated more than three times until the supernatant was clear. After washing each pellet with 30% ethanol, the starch was solubilized by mixing the pellet three times with 2 mL, 2 mL and 1 mL of 20% HCl. Then the hydrolysed starch was precipitated by adding 20 mL of 100% ethanol (the final concentration of the final solution exceeded 80%), followed by overnight refrigeration. After decanting, this gelatinized starch was immediately freeze-dried.
Serial tangential sectioning
For microscopic observation of the tree rings formed after pulse-labelling, a sliding microtome was used for the preparation of 15-µm-thick cross sections from the stem bases at 5 cm above ground from trees A–I. The cross sections were stained with safranin and photographed by a microscope (Eclipse E600; Nikon, Tokyo, Japan) equipped with a digital camera (DXM1200F; Nikon). Then a stem disk of 1 mm thickness with the surface smoothed during sectioning was separated from each stem base. From each disc, one or two rectangular blocks containing the outermost tree rings [ca. 0.4–0.6 mm (tangential) × 0.8–1.2 mm (longitudinal) × 2.0–3.0 mm (radial)] were prepared by a razor knife under a stereo microscope (SMZ1500; Nikon). With the exception of tree I, when formation of compression wood was observed, one block was taken from the compressed side and the other was taken from the normal side. The rings of tree I were too narrow for sectioning on the normal side. Each block was glued to a wooden stage with a cyanoacrylate adhesive soluble to acetone (Aron alpha Gel 10; Toa Gosei, Tokyo, Japan). From each block, serial tangential sections of 15–45 µm thickness were prepared by a rotary microtome (HM 340E; Microm International, Walldorf, Germany) equipped with a stereomicroscope. Both upper and lower cross surfaces of the block were observed under the stereomicroscope in order to align the sectioning plane parallel to both the grain direction and the growth ring boundary. The location of each tangential section on the block was successively marked, accurate to the cellular level, on the printed image of the cross surface. Special attention was paid to prevent mixing of tracheids from the previous ring. Six sections at a time were sealed in separate polyester filter bags with 50 µm pores (ANKOM Technology Corp., Fairport, NY) for the removal of the adhesive and extractives of the wood. The sections were placed in a Soxhlet extractor (4312-01, Sibata, Tokyo, Japan), extracted with acetone, toluene/ethanol, and ethanol each for 48 h, and then boiled in distilled water for 6 h. The final mass of each section was between 4 and 12 µg. We did not extract α-cellulose because the extraction process with NaOH was found to cause disintegration of the thin section samples.
Carbon isotope analysis
Carbon isotope analysis was carried out with the combined system of an elemental analyser (NC 2500; CE Instruments, Milan, Italy) and an isotope ratio mass spectrometer (MAT252; Thermo Electron, Bremen, Germany). One or two sections at a time were weighed into a small tin capsule (ϕ 3.2 × 4 mm; Lüdi Co., Regensdorf, Switzerland) which had been rinsed beforehand in a methanol/dichloromethane 1:1 solution to lower the amount of background carbon. For the starch analysis, 1 mg of the samples was measured out. Five standards were used for the calibration of the data [C-13 labelled UL-glucose IAEA-309 (A) + 535.3‰ & (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; alanine −23.5‰, Center for Ecological Research, Otsu, Japan].
We determined δ13C of the sections and starch samples (δ13Csample) in delta notation in permil units (‰) with respect to the Pee Dee Belemnite (PDB) standard. A natural baseline δ13C (‰) value for each tree was calculated by averaging the intra-annual δ13C values of both the tree-ring part formed before the pulse-labelling in the current year, if present, and the previous year’s tree ring (Table 1). Excess δ13C (‰) was calculated as deviations from these baseline values (δ13Cbaseline) as
Excess δ13C = δ13Csample − δ13Cbaseline(1)
Excess δ13C values of stem starch were calculated as weighted averages of the excess δ13C of the xylem and bark starch according to the relative amount of starch present in the respective parts. The excess δ13C value of (stem + roots) was calculated the same way with the data of stem xylem, stem bark, root xylem, root bark and fine roots as listed in Table 1.
The SD for replicate combustions of our internal standards (alanine) at the 14–78 µg range was 0.084‰ (n = 8). Natural δ13C variation of L. gmelinii tree rings at this site was 1.1‰ (± 1σ) during 1996–2000 (Kagawa et al. 2003) and we therefore set the detection limit for excess δ13C (later described as E-δ13C) at 2.0‰ over the natural δ13C value.
Circumferential distribution of excess δ13C and spiral grain
As expected, the observed spiral translocation path of pulse-labelled 13C in tree S (Fig. 1) roughly matched the spiral grain. However, the abnormal grain around the nodes that fell upon the cell alignment line made the exact evaluation of the spiral grain difficult, which may explain the discrepancies between the spiral cell alignment line and the locations of the E-δ13C maximum. Pulse-labelled 13C was confined within a single 45° fan even after translocating 45 cm down the stem, indicating minimal tangential diffusion of 13C-labelled sucrose across the sieve cells. The grain showed a steep S helix with a pitch (the vertical distance for the spiral to go around a stem) of ca. 15–30 cm (Fig. 1), equivalent to 5.7–9.3° grain angle characteristic of juvenile L. gmelinii (Takizawa et al. 1990). The autumn photoassimilate pulse-labelled in late August in a branch of tree S (Table 1) was not detected in any other branches, nor was it detected in the stem part at distances of 5 cm or more above the pulse-labelled branch (Fig. 1), indicating prevalence of basipetal translocation in autumn. The pulse-labelled branch showed E-δ13C values of 45 and 69‰ for needles and branch, respectively (data not shown).
Radial distributions of June photoassimilate in tree rings (trees A, B and C)
Tree ring formation may partly rely on stored carbohydrates translocated from the roots depending on the season (Kozlowski 1992). However, starch excess δ13C value in the stem and stem + roots showed similar values in reference to respective tree-ring excess δ13C maximum (Table 1, Figs 2–4). We therefore decided to use excess δ13C values of the stem starch instead of stem + roots for comparison with the excess δ13C of tree rings, because we do not know the relative contribution of storage carbon from the roots for the tree-ring formation at the stem.
June photoassimilate reached the stem base within 4 days of pulse-labelling and was found in the four outermost tracheids of the normal wood and the 16 outermost tracheids of the compression wood in tree A (Fig. 2a). The E-δ13C maximum was 7.3 times higher in compression wood than normal wood.
Two months after pulse-labelling (tree B, Fig. 2b), however, E-δ13C values in recently formed tracheids had decreased, suggesting a decrease of E-δ13C in sucrose and other substrates of the cambium. No significant E-δ13C was detected in the tracheids formed in 2000 (year prior to labelling), except in the last two sections of the rings, thus indicating a successful removal of radially mobile substances such as resin or starch by the organic solvents and hot-water extraction.
Tree C died of unknown reasons. A comparison of the ring formation (Fig. 2c) to the growth trend curve of L. gmelinii in this area placed the date of death around August 2001 (Kagawa et al. 2001). Like tree B, tree C showed a decrease of E-δ13C in recently formed tracheids corresponding to the earlywood/latewood boundary, and no significant E-δ13C was detected in the tracheids formed in 2000, except in the last two sections of the rings.
In short, spring photoassimilate pulse-labelled in June was found mainly in the earlywood. Maximum E-δ13C values of xylem were higher than that of starch in the stem in all cases except for tree A normal wood.
Radial distributions of July photoassimilate in tree rings (trees D, E and F)
July photoassimilate was found in both early and latewood in trees D and E (Fig. 3a & b). Normal and compression wood of tree D showed different E-δ13C distributions, indicating different growth trends between them. As in the case of June pulse-labelled trees, current xylem of trees D and E showed higher maximum E-δ13C values than those of their respective stem starch pools. A small but significant amount of E-δ13C was detected even at the early midpart of the earlywood of trees D and E, where lignin deposition is assumed to have occurred (Takabe et al. 1985). Interestingly, the carry-over of July photoassimilate pulse-labelled in 2000 into the 2001 ring was observed in tree F (Fig. 3c).
Assuming E-δ13C is constant within each tangential section(s), we integrated E-δ13C (‰) over distance (µm) to find E-δ13C (‰µm) peak area in each tree ring (Table 2). Of all the E-δ13C detected in the tree rings, only 9–14% was allocated to the current ring (formed in 2000), while an overwhelming 86–91% was allocated to the following ring (2001) and especially the following earlywood. In one of the two radial directions studied, a decrease of E-δ13C was observed in the following latewood (Fig. 3c). The outermost section in the second direction was lost during the extraction process and it is unknown if a similar decrease existed in the other direction. The following ring, which we expected to have been formed by the mixing of the carry-over of stored carbon assimilated in 2000 with spring photoassimilate of 2001, showed a lower E-δ13C value than that of the starch.
Table 2. Amount and percentage of pulse-labelled 13C allocated to each ring
Area (percentage) of excess δ13C in each ring (‰µm or percentage)
Excess δ13C in each section(s) measured as a single data point was assumed to be constant along the radial direction. The areas calculated are not weighted with density of wood.
ND, not detected, or excess δ13C did not significantly exceed natural δ13C variations of tree rings (excess δ13C < 2.0‰).
F (direction 1)
F (direction 2)
G (direction 1)
G (direction 2)
Radial distributions of August photoassimilate in tree rings (trees G, H and I)
Like tree F, the carry-over of August photoassimilate to the following ring was more or less observed in all August pulse-labelled trees. In tree G, only 9–18% of the E-δ13C was allocated in the current ring (2001; Table 2), while the remaining 82–91% was allocated in the following ring (2002). E-δ13C of the following ring was again lower than that of starch, indicating the use of stored carbohydrates for 2002 ring formation (Fig. 4a). In contrast to tree G, August photoassimilate in tree H was mainly allocated to the current ring (67–73%) and only a small portion (24–33%) was allocated to the following ring (Fig. 4b, Table 2). Like tree A, a higher E-δ13C maximum was observed in compression than normal wood. The E-δ13C maximum of the current ring (2000) exceeded the E-δ13C of stem starch, while the E-δ13C of the following ring (2001) fell below that of stem starch, except for the early part of the following ring on the compressed side. Tree I formed an unusually wide area of compression wood on one side in the following ring as compared with the rings formed in the other years (Fig. 4c). E-δ13C was found mainly in the following ring, but a significant amount of E-δ13C was also detected in the ring formed 2 years after pulse-labelling (Fig. 4c, Table 2). Tree I was also unusual in that the maximum E-δ13C of the following ring exceeded the starch E-δ13C.
Phloem translocation path in relation to spiral grain
The observed match of the spiral grain of tree S with the phloem translocation path of pulse-labelled 13C (Fig. 1) supports the hypothesis that phloem translocation parallels the grain of sieve cells. Furthermore, stem translocation of autumn (August) photoassimilate in tree S was in the basipetal direction. This is in accordance with the general statement that the translocation of shoots shifts from acropetal to basipetal direction in summer or autumn (Hansen et al. 1997), when shoot elongation slows down and xylem formation shifts from the early to latewood formation stage (Gordon & Larson 1968).
There was almost no circumferential diffusion of pulse-labelled 13C (less than 45°), even after translocating 45 cm down the stem. A similar phenomenon was also observed in our previous study with a straight-grained evergreen conifer, C. japonica (Kagawa et al. 2005). The sectorial phloem translocation observed in the two species supports the notion of integrated physiological unit (Watson & Casper 1984). In gymnosperms, sieve cells are interconnected by sieve areas on the radial walls (Esau 1965), and tangential diffusion of 14C-labelled photoassimilate to neighbouring sieve cells of Larix has been documented by autoradiography (Schmitz & Schneider 1989). However, because tree S had branches on the stem equally in all circumferential directions, we conclude that the sieve cells adjacent to the 13C labelled strip must be supplied with non-labelled assimilate, creating little diffusional potential across the stem circumference. Furthermore, a large conductivity difference may exist between transverse and longitudinal directions of sieve cells, analogous to the hydraulic conductivity difference in xylem tracheids (Cutter & Guyette 1993).
The maxima of 14C peaks in wood-developing zones are known to match the locations of secondary wall development zones at the time of labelling (Takabe et al. 1985). Judging from the locations of 13C peaks, we therefore conclude that the wood development zone was located in the earlywood at the time of June pulse-labelling, and spring (June) photoassimilates were used for current earlywood formation in trees A, B and C (Fig. 2a–c), confirming earlier findings (Smith & Paul 1988, Warren, McGrath & Adams 2001). Apart from tree A normal wood, earlywood E-δ13C maxima were higher than respective E-δ13C of the stem starch, suggesting a direct use of translocated June photoassimilate for earlywood formation before the photoassimilate enters the starch pool and the E-δ13C signal becomes diluted by 12C. Because tree A was sampled and oven-dried only 4 days after pulse-labelling, we can assume that the radial growth during the chase period was negligible. Therefore, the widths of E-δ13C peaks in Fig. 2a should reflect widths of wood-developing zones. The larger number of labelled tracheids (16 versus 4) and higher E-δ13C in the compressed versus the normal side (Fig. 2a) reflect a wider wood-developing zone and higher sink strength on the compressed side. The stronger sink strength would explain lower E-δ13C of normal wood than that of the starch pool in tree A, as labelled photoassimilate would have been preferentially allocated to the compression side, causing higher dependence of the normal wood on non-labelled stored reserves.
The observed decrease of earlywood E-δ13C (Fig. 2b & c) can be explained by a gradual decrease of 13C concentration in sucrose exported from branches, as a result of the export of pulse-labelled June photoassimilate from branches and replacement of the labelled carbon with newly incorporated July photoassimilate in the branch carbohydrate pool. The E-δ13C maxima, that is, the locations of wood-developing zones, in tree A (Fig. 2a) and trees B and C (Fig. 2b & c) were different at the time of pulse-labelling, reflecting the growth trend variation among individuals. Incorporation of the July photoassimilate into the late earlywood and latewood, and the higher E-δ13C of the current ring compared to starch in trees D and E (Fig. 3a & b) suggest a direct use of translocated July photoassimilate for late earlywood and latewood formation before it enters starch pool.
Judging from the date of pulse-labelling and the growth trend curve (Kagawa et al. 2001), we conclude that the wood-developing zone of tree F was located in the 2000 latewood at the time of labelling. The observation that current latewood E-δ13C did not exceed starch E-δ13C is explainable if the tree had almost finished latewood formation at the time of labelling and the sink strength of the wood-developing zone was weak. A similar phenomenon was also observed in August – labelled tree G. The lower E-δ13C of the following earlywood than the starch in trees F (Fig. 3c) and G (Fig. 4a) suggests a mixing of spring photoassimilate with stored material for earlywood formation.
Overall, earlywood was made of both current spring and previous summer/autumn photoassimilate. This agrees well with Jäggi et al. (2002)’s observation that earlywood δ13C of Picea abies is well correlated to starch and bulk needle δ13C sampled in May and not to climatic factors in spring. The dampening of the climate signal in earlywood δ13C (Lipp et al. 1991) could be explained by the mixing of current spring photoassimilates with storage starch assimilated in the previous year.
Unfortunately, because tree C died during the experiment, we could not clarify how much June photoassimilate is used for the current latewood or the following earlywood formation. However, judging from the decreasing E-δ13C trend in the earlywood of trees B and C (Fig. 2b & c), we believe that the contribution of June photoassimilate to the current latewood formation is relatively small. July and August photoassimilate was found to be used more or less for the current latewood formation in all six trees (Figs 3 & 4), and direct use of translocated summer/autumn photoassimilate for the current latewood formation was clearly suggested in trees D, E and H, where latewood E-δ13C was higher than the starch E-δ13C. In tree F, E-δ13C decreased near the following latewood at least in one direction (Fig. 3c), suggesting a diminishing dependence on storage material carried over for the wood formation in the course of the following growing season. The following latewood of tree H was also affected diminishingly less by carried-over storage material (Fig. 4b). However, such decreasing trend was absent in August – labelled tree G (Fig. 4a) and even the following latewood showed strong evidence of the influence of storage material carried over from 2001.
Our observation of latewood composed of current summer/autumn photoassimilate again agrees well with Jäggi et al. (2002)’s observation that latewood δ13C is correlated to climatic factors such as global radiation and has no relation to starch and bulk needle δ13C. Apart from tree G, our observation is also in line with Gordon & Larson’s (1968) reports that the cell wall synthesis of latewood in Pinus resinosa mostly relies on current summer/autumn photoassimilate because the onset of latewood formation is phenologically related to the onset of 14C-labelled photoassimilate export from needles in summer. Anatomical observation has also proved that secondary wall development of latewood takes place when temperature increases during summer and depends on the supply of current photosynthates (Antonova & Stasova 1993).
Carry-over of assimilated carbon into the tree ring(s) of the following year(s)
There was no clear relation between the percentage of pulse-labelled 13C carried over into the following ring (Table 2) and the size of starch pool. However, the carry-over percentage was related to starch E-δ13C value: the higher the carry-over percentage, the larger the E-δ13C ratio of the following ring/stem starch, ignoring E-δ13C area in early earlywood observed in tree H compression wood. Among the trees with chase periods greater than one year, both carry-over percentage of pulse-labelled 13C into the following ring and following ring/starch E-δ13C ratios increased in the order of trees H, F and G, I.
Tree rings are made by mixing carbon from two different sources, that is, current photoassimilate exported directly from needles and storage carbohydrate remobilized from parenchyma cells (Kozlowski 1992). We found higher contribution of carried-over storage from the previous year in earlywood than in latewood. The direct use of spring photoassimilate was evident in the extremely high absolute tree-ring E-δ13C maxima of June pulse-labelled trees, exceeding 800‰ (Fig. 2). However, once the labelled photoassimilate enters the starch pool, the label gets ‘diluted’ by mixing with the starch pool’s unlabelled carbon. The formation of the following earlywood would then draw on a mixture of the already diluted carbon from the starch pool and the current unlabelled spring photoassimilate, causing further dilution. This was reflected in the low absolute E-δ13C values in the following rings of trees G and I, and also as high E-δ13C of the current latewood and low E-δ13C of the following ring of tree H. If tree rings depend more on (labelled) storage material and less on (unlabelled) current photosynthate, then the following ring E-δ13C would be expected to be closer to starch E-δ13C. This hypothesis, however, cannot explain why the following ring E-δ13C exceeded starch E-δ13C in tree I and also why higher E-δ13C was observed in the following early earlywood than in the starch E-δ13C of tree H compression wood (Fig. 4b). In fact, we expected the starch E-δ13C of tree I to have been higher when the 2002 ring was formed, because starch E-δ13C is constantly decreasing as a result of carbon turnover in the starch pool (Kagawa et al. 2006).
There are two possible routes for the interannual carry-over of carbon assimilated near the end of the growing season into the following ring. The first route – which might explain the E-δ13C pattern observed in tree H compression wood – is most apparent in branch-level labelling experiments, where labelled carbon is detected only in the first layer of the following earlywood (Hansen & Beck 1990). In this process, preformed earlywood cells (cells that are destined to become the first layer of the following earlywood) already exist in the cambium at the end of the growing season (Larson 1994). In branch-level labelling (Fig. 1), labelled carbon becomes diluted with carbon from non-labelled branches once it enters storage pool and the contribution of labelled carbon from the storage pool to the following ring is minimal. In this case, labelled carbon should appear only within the preformed cells. Only four to seven cell layers are present in the overwintering cambium of Larix (Meier 1973; Imagawa & Ishida 1981). This is consistent with the observation of high E-δ13C observed in the first two cell layers at the beginning of the following earlywood in the compressed side of tree H. Again, contamination from current latewood tracheids is unlikely as both the upper and lower cross surfaces of the sectioning block were inspected during sectioning near the 2000–2001 ring boundary.
The second route is storage of carbohydrate and its subsequent use for formation of the following ring, which can be seen with whole-tree-level labelling. We found evidence of this process in trees F–I. The E-δ13C patterns observed in these trees were consistent with the argument that E-δ13C of a tree ring formed by this second route should be lower than E-δ13C of storage carbohydrate at the time of the following ring formation, while the earlywood cells formed with the first route could either show higher or lower E-δ13C than starch E-δ13C, depending on sink strength of the cambium in autumn. The use of stored carbohydrate for the following earlywood formation observed in this study can also partially explain the results of our previous study at the same site, in which a significant positive correlation was observed between earlywood width and natural δ13C of L. gmelinii in eastern Siberia (Kagawa et al. 2003).
However, Leavitt (1993b) found that no indication of the carry-over of storage carbohydrate for the following ring formation is observed in intra-annual tree-ring δ13C from Pinus strobes and Acer saccharinum with wider tree rings (average ring width = 4.6 and 5.8 mm, respectively) from south-eastern Wisconsin. A quick turnover of pulse-labelled 13CO2 in autumn to a branch of fast-growing plantation species (C. japonica) with a wide ring (13–14 mm) is observed, where excess δ13C quickly decreased to the baseline value at the outermost tracheids of the current latewood (Kagawa et al. 2005). However, our sample trees had narrow rings (0.03–0.66 mm), and, compared with these temperate trees with wider rings, our trees showed a slower rate of carbon turnover (Kagawa et al. 2006). The difference in carbon turnover rates may explain the discrepancy with Leavitt’s studies.
Care must be taken when applying the above explanations of interannual carbon carry-over to compression wood. Because of the stronger sink strength in the compression wood-developing zone, compression wood is likely to use carried-over storage material as opposed to normal wood. Such phenomenon was seen in the 2002 ring of tree I, which was made of unusually wide compression wood (Fig. 4c). Furthermore, the growth pattern of compression wood was sometimes different from normal wood (tree D, Fig. 3a), and therefore we caution against the use of isotope ratios of compression wood for climate reconstruction. In fact, unusual δ18O values have also been reported in compression wood of Picea engelmannii (Luckman & Gray 1990).
In conclusion, we offer the following recommendations for isotope dendroclimatological works with boreal trees of narrow rings where slow carbon turnover is expected, such as Larix in Siberia.
1Early and latewood should be separated in climate reconstruction work because the latewood relies mainly on current photoassimilate and influence of carried-over carbohydrate from the previous year is minimal, while the earlywood relies on a mixture of current and carried-over photoassimilate.
2Compression wood should be avoided.
As further motivation for our first recommendation, latewood δ13C of L. gmelinii is reportedly better correlated to growing season precipitation and soil water conditions than earlywood δ13C (Kagawa et al. 2003). A significant correlation between δ13C values of earlywood and the previous latewood (r = 0.42; P < 0.01) was also found in a 100-year δ13C chronology of P. sylvestris from the same site (Kagawa unpublished). However, carbon allocation patterns of mature trees may be different from saplings such as the ones used in this study, and similar 13CO2-labelling studies of mature trees, such as the currently ongoing study by Helle & Panferov (2004), with which (isotope) dendroclimatology is primarily concerned, are necessary.
The authors thank the staff of the Institute for Biological Problems of Cryolithozone for helping with the pulse-labelling in the field, Maya Jäggi for providing us with detailed starch extraction methods and all the members of GAME-Siberia for their cooperation. We would also like to thank Graham D. Farquhar and two anonymous reviewers for taking time to give us valuable suggestions for this manuscript. The authors also thank Jennifer Lue for reviewing early drafts of this paper and Takayoshi Koike for the helpful suggestions. This study was supported by Grant-in-Aid nos. 16403011, 16780119 and 11554017 from the Ministry of Education, Culture, Sports, Science and Technology, Japan.