Using a combined method of pulse-labelling trees and analysing detailed distribution of 13C tracer within tree rings, we studied how photo-assimilates incorporated on a given day are then distributed in a tree ring. A branch of a 4-year-old Cryptomeria japonica D.Don tree growing in Tsukuba, Japan was pulse-labelled with non-radioactive 13CO2 on two occasions: 29 May 2001 and 18 September 2001. Two discs were cut from the stem on 4 March 2002, one immediately under and the other 0.5 m below the branch and put through high-resolution δ13C analysis. δ13C peaks were observed in both the earlywood and latewood of the concerned tree ring, corresponding to each pulse-labelling date. The earlywood peaks was broader than the latewood peaks, possibly reflecting seasonal variation of the width of wood developing zone. Half-widths of the peaks were measured and used as indicators for the potential time resolution of tree-ring isotope analysis. The half-widths of the peaks indicated a time resolution no finer than 8.7–28 and 33–42 d in the early and latewood, respectively. Holocellulose extraction yielded only a slight change to the shape of the δ13C peaks. 13C tracer pulse-labelled in May and September reached tangentially different locations in the lower disc, suggesting a seasonal change in the pathway of carbohydrates. Local consumption of spring assimilates and long-distance downward transport of autumn assimilates were also suggested.
Recently, regulatory constraints and concerns associated with use of radioactive isotopes have limited the use of 14C labelling in ecological studies under field conditions. Fractionation of 14C is also much greater than 13C in photosynthetic and metabolic pathways (Van Norman & Brown 1952). This makes the stable isotope 13C an attractive alternative for ecological research. Despite these advantages, the application of stable isotope 13C for the labelling of plants has received relatively little attention. Most preceding 13C research on trees has utilized only natural variation of the 13C/12C ratio. For example, δ13C of C3 plants is known as a good indicator of water-use efficiency (Farquhar & Richards 1984).
The carbon isotope ratio of C3 plants reflects environmental conditions at the time of photosynthesis, and this isotopic information is recorded into tree rings after being modified through the processes in-between. Natural stable isotope ratios of tree rings have been used to reconstruct past climates (Epstein & Krishnamurthy 1990) and to study tree growth response to expected climate change (Barber, Juday & Finney 2000). However, in comparison with carbon isotope fractionation of photosynthetic processes, the method by which photosynthates with specific carbon isotope signatures are transported from leaves to the stem, used for tree ring formation, and eventually reflected in tree-ring isotope profile has not been extensively studied until recently (Damesin & Lelarge 2003; Helle & Schleser 2004). For example, the time lag between the uptake of carbon by photosynthesis at branches, and its incorporation into xylem for tree ring formation is still poorly understood (Schleser 1999). According to carbon allocation studies of trees, photosynthates assimilated for xylem formation at the early part of the growing season are used relatively soon in comparison with those assimilated at the end of the growing season, some of which are stored in the stem or roots before being used for the next year's growth (Hansen & Beck 1994). This should create a seasonal difference in the time lag between the carbon uptake and its use for tree ring formation.
Another crucial issue for tree-ring isotope study is the understanding of carbohydrate transport from leaves to the stem. Although xylem water transport has been studied in great detail, little is known about the axial path of phloem carbon transport from any one point in the crown to the stem. As in the case of xylem, substance flow in phloem is expected to be closely related to cell alignment.
The objective of this study was to clarify how photosynthetic products assimilated on a given day in different seasons flow through phloem and are then distributed in a tree ring, in relation to flow path in the phloem and the time lag between photosynthesis and wood formation. The latter is important for evaluating the time resolution of tree-ring isotope analysis. We also wanted to see the effect of holocellulose extraction conducted in tree-ring isotope research (Leavitt & Danzer 1993), on the 13C tracer distributions. However, we did not extract α-cellulose since the extraction process with NaOH was found to cause disintegration of the thin section samples and lower the sample recovery rate significantly.
For these purposes, we developed a simple combined method of pulse-labelling trees with non-radioactive 13CO2 and later analysing the detailed distribution of 13C tracer in the wood. We applied this method to a Cryptomeria japonica D.Don seedling under field conditions. We also examined the relation between the flow of 13C tracer and the alignment of the vertical cells.
We then measured the half-widths of the δ13C peaks observed in the tree ring to use as indicators of the potential time resolution for intra-annual tree-ring isotope analysis. Each half-width expressed in relative radial distance over the tree ring was converted to a day scale based on growth trend data.
MATERIALS AND METHODS
A 4-year-old Cryptomeria japonica D.Don (height 4.7 m; d.b.h. 10.5 cm) that grew under ambient climatic conditions at a nursery of the Forest and Forest Products Res. Institute in Tsukuba, Japan was selected. The width of the last tree ring, formed in 2001, was 1.3–1.4 cm at the sampled locations. Growth trends of 27 Cryptomeria japonica trees were measured with dendrometer bands in the same year at the Chiyoda experimental station of the Forest and Forest Products Res. Institute, 22 km away from the nursery.
In order to label the tree with 13CO2 tracer via photosynthesis, we assembled the labelling apparatus illustrated in Fig. 1. For application of 13CO2, the whole branch at a height of 2.7 m was enclosed in a 30 cm × 50 cm, 100 µm-thick linear low density polyethylene bag. To prevent water condensation on the inner surface of the plastic bag, air in the bag was constantly circulated with a vertically mounted electric fan. A silica gel cartridge was connected to the fan to continuously remove water vapour generated by transpiration. Pulse-labelling was carried out when there was sufficient sunlight to cause net CO2 uptake, on mostly sunny days, when the photosynthesis photon flux density was assumed to exceed the light saturation point for Cryptomeria japonica (approximately 700 µmol m−2 s−1, Hashimoto & Suzaki 1979).
We conducted two pulse-labelling sessions during the growing season: Once in the earlywood formation period on 29 May 2001 and once in the latewood formation period 18 September 2001. A total amount of 400 mg Ca13CO3 (87 mL 13CO2) was used at each session so that CO2 concentration was well in excess of the CO2 saturation point. We used 98 atom percentage Ca13CO3 (Isotec Inc., Miamisburg, OH, USA) to prepare a 40 mg mL−1 CaCO3 solution. After the branch was sealed in the plastic bag, 13CO2 was released in two equal volumes (200 mg each, once in the morning and once in the afternoon) by injecting H3PO4 solution from a gas-tight syringe into the Ca13CO3 solution. The plastic bag was kept airtight by sealing it to the branch base with fat clay and wrapping the mouth with string.
After each injection, the tree was left for 2 h to allow assimilation of 13CO2 before the bag was removed. During the pulse-labelling session in the fall, cambial cells were injured with a pin so that the location of cambium at the time of pulse-labelling could be later identified (Wolter 1968).
Sample preparation for sectioning
The tree was harvested during the dormant season (4 March 2002). Two discs were taken from the stem, one immediately below the 13CO2-fed branch (Disc 0 m) and the other 0.5 m directly below (Disc −0.5 m). Each disc was divided into six fan-shaped pieces with a base angle of 60° as shown in Fig. 2. After taking cross-sections from the pin-marked block for microscopic observation, the blocks were air-dried. Cross-sections of 100 µm thickness were then prepared from each piece in order to analyse the two dimensional distribution of 13C tracer on the cross surface of each disc. For the two-dimensional analysis of 13C on cross surfaces, the tree-ring area consisting of two 60° fans immediately under the branch base was used for Disc 0 m and the whole tree ring (360°) was used for Disc −0.5 m (Fig. 2). Each tree ring section was separated radially into three parts and again every 12° tangentially before the δ13C analysis. Three blocks (Blocks A, B, and C) of 2 (tangential) × 5 (longitudinal) × 14 (radial) mm size were taken from the areas of highest detected tracer concentrations. This is shown in detail in Fig. 2.
Cross and radial surfaces of the blocks were smoothed with a sliding microtome to ease recognition of grain direction under a stereomicroscope. Serial tangential sections of 60 µm thickness were prepared from the blocks using a rotary microtome (HM 340E; Microm International, Walldorf, Germany) equipped with a stereomicroscope. Both transverse and radial surfaces were observed under the stereomicroscope, and the specimen angle was continually adjusted during sectioning in order to cut the sections parallel to both the grain direction and the growth ring boundary. A universal specimen holder was used in combination with a rotatable knife carrier to set the alignment of the specimen. Special attention was paid when cutting sections close to the ring boundary to prevent mixing of tracheids from the previous year's tree ring.
The 13.6-mm-wide tree ring of Block A was subdivided into 228 pieces of 60 µm thick serial tangential sections. The sections (no.1 – no.228) were separated into two groups, one group of odd-numbered sections (no. 1, 3, 5 . . . 227) and one group of even-numbered sections (no 2, 4, 6 . . . 228). After extracting resin, the first group was analysed as bulk samples for carbon isotopes, and the second group was subjected to holocellulose extraction (Leavitt & Danzer 1993) before carbon isotope analysis. Lignin deposition takes place later and for a longer period than the holocellulose accumulation on cell walls (Takabe et al. 1985). Due to this time lag, lignin removal by the holocellulose extraction process has the potential to cause some changes to δ13C distributions. We compared the holocellulose δ13C values with bulk sample values to see the effects of the extraction process used for tree-ring isotope studies. Two sections were pooled together for each carbon isotope analysis, producing two series (bulk and holocellulose) of 57 (= 228/4) data points from the tree ring of Block A. Blocks B and C produced 53 and 58 data points, respectively, and only bulk samples were analysed. In order to determine the intra-annual background δ13C variation within a tree ring, another block, Block D, was taken from Disc 0 m on the opposite side of the branch – where, from previous research, no tracer was expected to be found – and processed the same way.
Carbon isotope analysis
We used the combined system of an elemental analyser (NC 2500; CE Instruments, Milan, Italy) and an isotope ratio mass spectrometer (MAT 252; Thermo Electron, Bremen, Germany) for the δ13C measurement. The standard deviation for replicate combustions of our internal standards (DL-alanine) was 0.07‰. We determined δ13C of wood tissue and expressed it in the delta notation in per mil units (‰) with respect to the Peedee belemnite (PDB) standard. Average weight for bulk and holocellulose samples was approximately 0.7 and 0.4 mg, respectively. For correction of the raw δ13C data, we first analysed the background intra-ring δ13C fluctuation. Leavitt (1993) reported intra-annual fluctuation of δ13C values of conifer tree rings up to 1–2‰, however, the outermost ring of Block D showed a fluctuation of only 0.4‰. We therefore set the detection limit for 13C tracer at 1‰ over the baseline value and calculated excess δ13C values as deviations from the baseline.
Alignment of vertical phloem cells and tracheids
We followed the method of Kozlowski, Hughes & Leyton (1967) for examination of vertical cell alignment. After removing the outer bark with a knife, we pulled slivers off the inner bark and outermost tracheids originating from the base of the 13C-fed branch. Since slivers were pulled off in parallel to the grain direction, a cell alignment line was continuously drawn following the vertical cell direction from the branch base (Disc 0 m) to 0.5 m directly below (Disc −0.5 m). After checking the cell alignment at phloem and outermost xylem, we used a saw to cut the 0.5 m-long stem part at 7 cm intervals. In order to check the alignment of tracheids in the earlywood that were formed at the time of the first pulse-labelling, the tree ring was dissected with a chisel, and then the surface was planed smooth. The dissection was made at the place of the highest detected tracer concentration in the earlywood, stepwise from 38 to 53% relative distance (Figs 3 & 4), according to the location of the billet from 0 m to −0.5 m.
Thus we investigated the alignment of vertical cells formed around the two pulse-labelling dates and compared the lines with the flow paths of 13C tracer determined in the carbon isotope analysis.
δ13C distributions on cross surfaces of the discs
The highest tracer concentration in the stem was observed in Disc 0 m at the location directly beneath the 13CO2-fed branch up to 43‰ (Fig. 2). However, in Disc −0.5 m, 13C tracer was found at two tangentially separate locations. One of these locations was in the earlywood (maximum 2‰); the other was in the latewood (maximum 6‰) as shown in Fig. 2.
Alignment of the phloem cells
Orientations of tracheids are known to differ according to tree species, due to variance in pseudotransverse divisions of fusiform cambial initials (Heijnowicz 1964). It is well known that Cryptomeria japonica has straight grain, and the cell alignment lines originating from the 13CO2-fed branch run straight down the stem (Fig. 2). Just as spiral carbohydrate flow in the phloem is observed in species with spiral grain (Pinus sylvestris; Hansen & Beck 1990), straight carbohydrate flow was inferred in our Cryptomeria japonica tree. The lines showed seasonally different paths. The early- and latewood cell alignment lines reached near the places of the highest detected tracer concentration in the early- and latewood on Disc −0.5 m, respectively (Fig. 2). Overall, carbohydrate flow followed the orientation of the vertical cells, although there were some discrepancies. The early- and latewood cell alignment lines were found to lead to locations approximately 1 cm away from those of highest tracer concentrations in the early- and latewood at −0.5 m (Fig. 2). However, there were nodes that fell on the cell alignment lines, the presence of which may account for these discrepancies.
Intra-annual distribution of δ13C in the tree ring
A significant amount of 13C tracer was successfully incorporated into the wood by the pulse-labelling method and, in both the bulk and holocellulose measurements, two excess δ13C peaks were found in the tree ring from Block A, one in the earlywood and another in the latewood (Figs 2 & 3). The earlywood peaks were 4.5–5.7 times larger and 2.6–4.1 times broader than the latewood peaks (Table 1), and the location of cambium at the time of the second pulse-labelling (as determined from the pin-marking) coincided exactly with the maximum of the second peak.
Table 1. Data for the excess δ13C peaks
Relative distance (%)
Peak area (‰ · %)
Growth rate (%/day)
0.834 ± 0.047
10.9 ± 0.6
8.7 ± 0.5
0.079 ± 0.035
33 ± 14
42 ± 18
0.672 ± 0.055
28.4 ± 2.6
0.079 ± 0.035
40 ± 17
Of the growth trend curves obtained from 27 trees with dendrometer bands, the six curves that showed growth trends most similar to the pulse-labelled tree (i.e. about 40 and 96% of relative growth by 29 May and 18 September, respectively) were used for estimating xylem formation dates (Figs 3 & 4). The pulse-labelled tree and the six trees whose growth curves were studied were growing under similar conditions in terms of crown exposure to sunlight, temperature and humidity. As confirmed by the 27-tree growth trend data, 29 May and 18 September correspond to early- and latewood formation periods, respectively, for Cryptomeria japonica growing in Tsukuba, Japan. The earlywood and latewood excess δ13C peaks are therefore attributed to the 13CO2 pulse-labelling conducted on 29 May and 18 September, respectively. The coincidence of the pin-marked location and the maximum of the second peak verify this conclusively.
δ13C profile difference between the bulk and the holocellulose samples
Through the holocellulose extraction process, lignin was removed from the samples. In comparison with the holocellulose samples, bulk shows relatively gradual 13C increase at the peak onset. The steepness of the head of the holocellulose peak suggests exceptional time resolution of the δ13C profile. In other words, by cutting serial tangential sections parallel to the cambium plane at the time of the tracheids formation, we have achieved clear separation of tracheids formed before and after the 13C tracer reached the cambium. Earlywood peak area for holocellulose was substantially larger than bulk by 517‰ · %, but there was little difference (55‰ · %) between the bulk and holocellulose latewood peaks.
δ13C peaks observed at Disc −0.5 m
There were also two peaks observed at Disc −0.5 m, again, one in the earlywood and one in the latewood. This time, however, the peaks were found at tangentially different locations on the disc (Figs 2 & 4). The earlywood peak was observed in Block B, and the latewood peak was observed in Block C. This means that the phloem flow followed seasonally different paths in the earlywood and the latewood formation periods. The earlywood peak area (31‰ · %) was much smaller than that of the latewood peak (302‰ · %). In comparison with the earlywood peak at 0 m, the earlywood peak at −0.5 m appeared later in relative distance (38 versus 53%), whereas the latewood peaks at 0 m and −0.5 m appeared at the same relative distance (96 versus 97%). No clear peak was observed either in the Block B latewood or the Block C earlywood.
Excess δ13C peaks observed in the outermost tree ring
Hansen & Beck (1990) also found the highest 14C concentration at the stem area immediately beneath the feeding branch of the tree in their study and Takabe, Fujita & Harada (1981) and Takabe et al. (1984) observed the highest rate of carbon incorporation in the wood developing zone in secondary xylem at secondary wall thickening stage, when active cellulose biosynthesis occurs. This explains the coincidence of the pin-marked location and the δ13C maximum. Short-term experiments with 14C tracer suggest that it takes less than 3 d for the first labelled assimilates in the leaves to reach the stem (Hansen & Beck 1994; Simard et al. 1997a). This quick carbon transport to stem explains the steep increase at the onset of the earlywood peak (left side) in Fig. 3, whereas its gradual tapering off and long tail suggest slower turnover of 13C tracer probably due to the large starch-pool of the branch in spring (Jäggi et al. 2002).
Holocellulose δ13C versus bulk-wood δ13C
Although most cellulose biosynthesis occurs during the secondary wall thickening stage, lignin deposition onto the secondary wall continues over a longer period after the cell wall thickening is completed (Takabe et al. 1981, 1984, 1985). Therefore, lignin is expected to reflect δ13C variation in the phloem carbohydrate flow integrated over a longer period than cellulose. Thus, the difference in peak shape between holocellulose and bulk samples at the onset of the peaks probably reflects the lignin deposition pattern at the wood developing zone. Holocellulose δ13C peak area was larger than the bulk peak area in the earlywood (Fig. 3) but the difference was small for the latewood peaks at Disc 0 m. Longer deposition time for lignin than cellulose may make bulk peaks ‘less focused’, or make the bulk peak area smaller than that of holocellulose peaks. The naturally lower carbon isotope ratio of lignin (Wilson & Grinsted 1977; Leavitt & Long 1991; Loader, Robertson & McCarroll 2003) may also contribute to this phenomenon.
In tree-ring isotope studies, α-cellulose or holocellulose extraction has often been carried out to diminish the effect of the variable ratio of wood chemical components (i.e. cellulose, hemicellulose and lignin) on δ13C values and to remove radially mobile components which may move across growth ring boundaries (Leavitt & Danzer 1993; Wilson & Grinsted 1977). However, with respect to the time resolution of tree-ring isotope analysis, extracting the samples to holocellulose makes practically little difference in relative δ13C variation, as the difference between the bulk and holocellulose δ13C distributions is very little. The enclosed area between Bulk-A and Holocellulose-A curves at the head of the peaks (15–30% relative distance in Fig. 3) was 95 (‰ · %), which accounted for only 3.1% of the bulk-A earlywood peak. This agrees with the observation of Loader et al. (2003) of a very high degree of correlation in the tree-ring δ13C time series among the whole wood, cellulose and lignin.
Different peak widths in the earlywood and the latewood
There are two probable explanations for our observation of broader δ13C peaks in the earlywood than in the latewood. Firstly, this may be caused by slower turnover of the spring starch pool in the branch. During the growing season, a large starch accumulation in needles or sapwood of conifers is observed in spring, followed by a decrease in summer and autumn (Ericsson 1979; Luethy-Krause & Landolt 1990; Höll 2000; Oleksyn et al. 2000; Schaberg et al. 2000). The newly formed starch transported from branches to stem is then used mainly for earlywood formation (Hansen & Beck 1994; Jäggi et al. 2002). This large starch pool in spring may cause slow 13C tracer turnover and a long duration of 13C tracer flow in the phloem in the earlywood as compared to the latewood formation period, potentially making the earlywood peak broader. This is most evident in the long tails of the earlywood peaks in comparison with the latewood peaks in our observations (Fig. 3). Secondly, faster growth and/or a wider wood developing zone in late May than middle September may account for the variation in peak width. The radial growth rate of Cryptomeria japonica is much faster and wood developing zone wider in the earlywood than the latewood formation period. For example, Nagao, Mio & Tsutsumi (1985) observed 60% wider wood developing zone in late May than in mid-September In other words, more cells are incorporating carbon in the earlywood formation period and this can also cause the 13C pulse to be widely diffused over the cells in the wood developing zone.
Insights into the time resolution of tree-ring isotope analysis
These peak widths give us an insight into the time resolution of intra-annual tree-ring isotope analysis. In spectral analysis, resolution of a time series is given by half-width at half height of a peak, as two same-sized peaks separated by an interval of more than the half-width are discernable as separate peaks. By fitting a Gaussian curve, we measured the half-width of each peak (Table 1). The earlywood peaks were narrower than the latewood peaks in relative distance. The half-widths for bulk early- and latewood peaks were 9.1 and 2.2% and those for holocellulose early- and latewood peaks were 7.2 and 2.8%, respectively. We used the growth rate at each peak position of the ring to convert the half-widths to day units, which is shown with standard deviation of estimates from six different growth curves. Half-width in day unit, as we define time resolution, was 8.7–10.9 d in the earlywood at 0 m with smaller relative standard deviation. Due to the coherent growth rate among the six trees in the earlywood formation period, the standard deviation for the time resolution estimates remained small. However, the latewood growth rate showed larger relative standard deviation, leading to the larger standard deviation of latewood time resolution. Although the standard deviations were larger, the time resolution seemed longer in the latewood than in the earlywood. This longer time resolution in latewood probably reflects longer time equivalent of wood developing zone width in September rather than carbon turnover time.
In comparison with 0 m, the time resolution at −0.5 m was 2.6 times longer in the earlywood (10.9 versus 28.4 d) whereas there was little difference in the latewood (33 versus 40 d). The longer time resolution at −0.5 m than 0 m earlywood may be attributed to slower carbon turnover in spring as most clearly seen in the long tail of the 0 m earlywood peak leading up to 70% relative distance (Fig. 3). We should note that these time resolution estimates do not reflect variations of carbon turnover time between branches as the 13C tracer was fed from one branch in this experiment, and the actual time resolution of tree-ring isotope analysis could be lower.
Carbohydrate flow path and phloem cell alignment
Since the orientation of fusiform cambial initials changes seasonally, the water-conducting paths in xylem show annual or seasonal variations (Rudinsky & Vite 1959; Kozlowski et al. 1967). Because of this variation, a given root can be axially connected with various branches of the crown through time. A close agreement between the path of water movement and the vertical run of the grain in conifers was found by comparing the helical path of injected dye with the tracheid alignment (Kozlowski et al. 1967). Hansen & Beck (1990) administered 14CO2 to an upper branch of Pinus sylvestris to study the path of carbohydrates in the phloem and also found the phloem flow to follow the helical grain direction of the conifer in his study. Zimmermann & Brown (1971) speculated that the path of translocation in phloem should be very similar to the path of sap movement in the youngest xylem, except that their respective movement directions are usually opposite. He also speculated that the newly produced sieve cells in the phloem are aligned in the same way as the newly differentiated tracheids in the xylem, since both of them are produced from the same fusiform cambial initials. In other words, just as a given root can be connected with various branches of the crown hydrolically through xylem, a branch might be axially connected through the phloem to different roots over a long period, although the shorter functional life of sieve cells (Esau 1965) may make the seasonality of the conducting path more pronounced than that of the xylem. Our results support this speculation and proved that a given branch is connected with a different part of the stem in different season because of a seasonal change in vertical cell orientation. However, the seasonal change of phloem flow paths suggested shorter functional period of sieve cells (one season) than the one estimated from anatomical observations of callose deposition in conifer phloem (1–2 years; Grillos & Smith 1959). The 13C tracer flow was still confined to limited areas even after flowing 0.5 m down the phloem (Fig. 2) and there was little tangential diffusion, as has been observed also in Hansen & Beck (1990). This may be related to the observed circumferential carbon isotope variation within a tree ring (Leavitt & Long 1986; Kitagawa & Matsumoto 1995), as it may reflect microenvironment variation in different parts of the tree crown. Similar circumferential δ13C variations between leaves in crown (1–2‰) and equivalent tree ring (0.5–1.5‰) were observed (Leavitt & Long 1986).
Carbon turnover difference in the spring and the autumn
Although at −0.5 m, the earlywood peak height (2.5‰) was well above the natural variations, this peak was much smaller than the latewood peak (31 versus 302‰ · %). On the contrary, the earlywood peak was larger than the latewood peak at 0 m (3059 versus 678‰ · %). When earlywood formation is active in spring or summer, considerable amounts of photosynthates are allocated to the radial growth, shoot elongation and maintenance respiration, and thus photosynthates are consumed locally. In autumn, when secondary growth is about to end and the maintenance respiration is lower due to the lower temperature, a fraction of the assimilates can pass through the stem and flow downward to the root system (Ursino, Nelson & Krotkov 1968; Hansen & Beck 1994). Much higher 13C concentrations at 0 m than −0.5 m in the earlywood suggest local consumption of the 13C-labelled photo-assimilates in May near the 13CO2-fed branch, leaving only a small amount of tracer to be transported to the −0.5 m part. Much higher 13C concentrations in latewood than earlywood at −0.5 m suggest smaller local carbon consumption for the radial growth and larger long-distance transport of September carbohydrates downward. The difference in the relative distances of earlywood peaks at 0 m and −0.5 m may reflect either 13C tracer transport time or the different relative growth trend between the two stem heights.
A significant amount of 13C tracer was successfully incorporated into the tree by the pulse-labelling method and two δ13C peaks corresponding to the two pulse-labelling dates (29 May and 18 September 2001) were observed in the outermost tree ring. The steep head of the holocellulose δ13C peaks suggests clear separation of tracheids formed before and after the tracer reached the cambial zone, which was achieved by cutting serial tangential sections parallel to grain direction and ring boundary. The half-widths at half height of the earlywood peaks were wider than the half-widths of the latewood peaks, possibly because of a wider wood-developing zone in spring than autumn and/or faster growth rate. Slower carbon turnover of the starch pool in spring may also explain this, as seen most evidently in long tails of the earlywood peaks in comparison with the latewood ones. The half-widths in day unit, which we used as an indicator for potential time resolution of tree-ring isotope analysis, were 8.7–10.9 and 33–42 d in the early- and latewood at 0 m, respectively. As far as the time resolution is concerned, extracting the samples to holocellulose made practically little difference. Compared to 0 m, the time resolution at −0.5 m became longer in the earlywood whereas there was little difference in the latewood. We should note that these time resolution estimates do not reflect variations of carbon turnover time between branches, and the actual time resolution of tree-ring isotope analysis could be lower. The 13C tracer reached tangentially different places in the earlywood and latewood of Disc −0.5 m. This is explained by a seasonal change in orientation of fusiform cambial initials, which should then create a seasonal change in the orientation of sieve cells in the phloem comparable with that of tracheids in the xylem. In other words, these results suggested that a given side of a stem is connected to different branches in different season. It also suggested shorter functional life of sieve cells of one season than those estimated from anatomical observations of conifer phloem (1–2 years). Much higher 13C concentrations at 0 m than −0.5 m in the earlywood suggested local consumption of spring assimilates and much higher 13C concentrations in latewood than earlywood at −0.5 m suggested long-distance downward transport of autumn assimilates.
In order to better interpret tree-ring isotope data, a clear understanding of the processes between photosynthesis and wood formation is essential, especially the time lag between the two. Using natural carbon isotope ratios, Helle & Schleser (2004) studied how seasonal δ13C variation of starch in leaves – which reflects short-term environmental conditions – is then reflected in intra-annual variation of δ13C in tree rings. Usage of stored starch assimilated in the previous year for the current year's growth was observed, which has also been indicated by a significant correlation of δ13C between the previous year's latewood and current year's earlywood (Lipp & Trimborn 1991; Switsur et al. 1995; Robertson et al. 1997). However, this interannual time lag question may be better answered by the usage of artificially added tracer, which will be the subject of our future studies.
The authors thank Mitsue Fukui for providing a Cryptomeria japonica tree, Shuichi Noshiro and Jennifer Lue for reviewing an early draft of this paper. We also would like to thank D.Yakir and two anonymous reviewers for taking their time to give us valuable suggestions to improve our manuscript. This study was supported by Grant-in-Aid 11554017 and 16780119 from the Ministry of Education, Culture, Sports, Science and Technology, Japan.