Deriving a consistent δ13C signature from tree canopy leaf material for palaeoclimatic reconstruction


Author for correspondence:
Chris Turney
Tel: +44 (0)28 90273980
Fax: +44 (0)28 90335354


  • • Contemporary studies of stable carbon isotope values (δ 13 C) from leaves in the canopies of mature forest indicate that reduced irradiance and respired CO 2 might bias palaeoenvironmental reconstructions developed using the 13 C content of unsorted plant material formed within comparable ancient environments.
  • • Here, we investigated whether any simple morphological and/or chemical characteristics could identify the position of the leaf within a canopy of the evergreen mountain beech ( Nothofagus solandri var. cliffortiodes ).
  • • Leaf mass per unit area, carbon content and δ 13 C values of both bulk and lignin components of the leaves increased exponentially from the ground up through the canopy. Nitrogen, remobilized before death, was unsuitable as an indicator of canopy position. Leaf mass per unit area on the forest floor indicated that leaves from the sunlit part comprised approximately 30% of the fallen leaves; leaves originating from the upper canopy made up the remaining litter.
  • • Application of leaf mass per unit area to a 22 000-yr-old sequence dominated by mountain beech leaves from Mount George (New Zealand), demonstrated that this signal was preserved in the fossil record. Sunlit leaf material can be successfully differentiated from that originating deeper within the palaeo-canopy, thereby allowing robust δ 13 C analyses for palaeoenvironmental reconstruction.


The measurement of stable carbon isotopic values (δ13C) within organic matter shows promise as a tool for reconstructing palaeoenvironmental change (Stuiver, 1975; Håkansson, 1986; Hammarlund, 1993; Hatte et al., 1998; Turney, 1999), with significant shifts being recorded in fossils from horizons distributed on either side of known climatic boundaries. However, many studies have focused on the measurement of δ13C within sedimentary material that is inherently heterogeneous in makeup (e.g. lake sediments) and therefore difficult to interpret (Deines, 1980; Hammarlund, 1994). By limiting measurements to species- and organ-specific terrestrial plant macrofossils, uncertainties in stable carbon isotopic sources can be restricted and more robust δ13C data sets can be obtained (Beerling, 1996; Turney et al., 1997). Investigations in this subject area have largely focused on relatively low-lying plants which are known to have little storage of carbon over the winter period (i.e. a negligible ‘memory’ effect), are generally well ventilated, and therefore assumed to be representative of ambient, environmental conditions. However, relatively low-growing plants may not reflect the conditions found above the boundary layer they inhabit. Therefore, well-exposed tree foliage, held in a turbulent wind field may furnish a better connection between leaf-derived parameters and the general climate.

The discrimination of 13C relative to 12C in C3 terrestrial plants is related to the isotopic composition of the air and the ci : ca ratio of a leaf, where ci and ca are the substomatal internal and external CO2 concentrations (Farquhar et al., 1989; Read & Farquhar, 1991). Although there are fractionation factors associated with diffusion through air and carboxylation, these can be considered essentially constant. As a result, any environmental stress that influences the leaf stomatal conductance and/or net assimilation will affect the stable carbon isotope ratio of the plant material. Recently, a calibration of δ13C variations in contemporary cladodes (photosynthetic leaf-like organs) of Phyllocladus alpinus (a shrub/tree) in New Zealand against modern climate data-sets indicated a strong correlation with the leaf-to-air vapour pressure deficit (VPD) (Turney et al., 1999), suggesting this approach can be used to quantify changes in palaeoatmospheric moisture content.

Within a mature, closed canopy forest, terrestrial plant carbon isotopic fractionation is complicated by additional factors, the most important probably being changes in irradiance through the canopy (Ehleringer et al., 1986; Schleser, 1990). Leaves exposed at the top of the canopy, under high irradiance will be more closely coupled to the imposed environmental conditions than leaves deeper in the canopy. They will, therefore, have reduced boundary layers and have higher imposed VPDs (i.e. more representative of the climate outside the forest canopy). In addition, 13C-depleted respired CO2 from decomposing humus on the forest floor may also alter the plant δ13C, if recycled within the tree canopy (Medina & Minchin, 1980; Schleser & Jayasekera, 1985; Brooks et al., 1997).

Taphonomic studies indicate that transport and diagenesis of plant tissues can be selective (Spicer, 1991) with a predominance of sunlit morphotypes in the fossil record (Kürschner et al., 1996). However, in fossil-rich deposits where the material appears to directly underlie the palaeovegetation stand, it is problematic to assume the transportation of leaves by wind and water has biased the samples to such an extent that shade material can be considered negligible. If fossil leaves from mature, closed-canopy forests are to be used for robust palaeoclimatic reconstruction, it is essential to identify whether the above mechanisms play a significant role in determining and biasing the δ13C values and, if so, whether the outer envelope of leaves can be identified in fossil deposits (Lockheart et al., 1998). Some workers have attempted to use stomatal counts to distinguish sunlit leaves in the fossil record but this methodology is relatively labour intensive and requires experienced analysis (Kürschner, 1997).

The aims of this study were therefore twofold: (1) to investigate contemporary processes operating through a canopy in a mature stand of mountain beech in New Zealand; (2) to apply the results and principles of δ13C measurement to a fossil-rich site at Mount George, South Island, New Zealand, laid down during the Late Otiran Glaciation (c. 22 000 yr ago).

Materials and Methods

Contemporary samples

Contemporary samples were collected within a mature, closed canopy (maximum canopy height of 18 m) of mountain beech (Nothofagus solandri var. cliffortiodes (Hook. f)) in the Craigieburn Range, South Island, New Zealand (43°08′ S, 171°41′ E; Fig. 1). Since mountain beech holds its leaves for up to 14 months (i.e. there are only 2 months where there are different-aged leaves on the tree) and we can distinguish the age of litter fall on the basis of colour and degree of preservation, we were able to collect the following leaf material: (1) samples of leaves from the tree that were 4 months old (Year 0T); (2) newly senesced material aged approximately 13 months on the tree (Year 1T); (3) newly senesced material within the leaf litter aged between 13 months and 2 yr (Year 1L); and (4) approximately 2-yr-old dead material in the leaf litter (Year 2L). Samples of both live and newly senesced leaves were hand picked (covering different leaf angles and aspects) at the five levels through the canopy (1.5, 5.0, 12.0, 16.5 and 18.0 m above ground level). Sampling was completed during the late summer of 1998 (March) following the 1997–1998 period of growth. Leaf litter samples were collected immediately below the canopy. All samples were cold-stored at 4°C until processed.

Figure 1.

Location of mountain beech forest (Craigieburn) and Late Otiran Glaciation (Mt George) sampling sites, New Zealand. The principal lithostratigraphic units of the Mt George site are also shown. All fossil samples were collected in the C and D/E horizons. Radiocarbon sample is shown as solid square.

For each height, 20 live and senesced leaves were randomly sampled and individually stored. Leaf area was determined using a LI-3100 leaf area meter (Li-Cor, Lincoln, NE, USA). Samples were then dried overnight at < 70°C. The mass was recorded after drying on a Precisa 40SM-200 A meter (PAG Oerlikon AG, Zurich, Switzerland) and the leaf mass per unit area was then calculated. The error values obtained indicate that 20 samples per horizon are a sufficiently robust data set for the purposes of this exercise. An identical protocol was followed for the litter, for which 20 leaves were sampled as representative of the amount of fossil material available from the Mt George site (see below). Based on colour and their state of preservation, leaves were randomly sampled for Year 1L and Year 2L, representing two different years’ senescence (the growing seasons of 1995–1996 and 1996–1997). Mass and leaf area were determined as above and because nondestructive, allow for preselection of exposed leaves before time consuming and expensive analysis.

The light environment through the mountain beech canopy was modelled using an ellipsoidal leaf distribution, leaf optical transmission properties (adaxial transmission and reflectance) and leaf area distribution from data previously collected at the site (obtained from Hollinger, 1989). The extinction coefficient at the bottom of each 1-m layer (K) was calculated following Campbell (1986) and combined with the transmission and reflectance data of the layer above to produce a light transmission equation for any tier

Ii   =  Ii +1 e math image  +  Ti   −   Ri(Eqn 1)

where, for each layer, I is the irradiance at the base of the layer i, L is the leaf area index, T is the proportion of light transmitted through the leaves and R is the proportion of light reflectance from the surface of the leaves.

Fossil samples

Mount George (42°20′ S, 171°15′ E) is a 12-m exposure of seven weakly developed soil horizons (Units A–G; Fig. 1) adjacent to the coast road between Westport and Greymouth. Between these soil horizons are inorganic units, consisting of greywacke clasts and in places, large blocks of conglomerate (C. Burrows et al. unpublished). Lenses of gravel and fine gravel layers are recorded between organic units B–C, C–D, D–E and E–F, suggesting water-laid sediment, although it is possible these units may represent a complex colluvial/alluvial sedimentation. The organic units C, D and E were relatively rich in plant macrofossils, of which N. solandri var. cliffortiodes dominated and exhibited a high degree of preservation. Samples were soaked in distilled water prior to sieving through mesh size 250 µm. Twenty leaves were extracted for analysis from each horizon. A radiocarbon age of 22 280 ± 150 14C BP (wk-3797) using an acid-wash pretreatment was obtained on terrestrial wood taken from unit E (University of Waikato).

Nitrogen, carbon and δ13C measurements

Contemporary leaves were selected for both lignin and bulk analyses (1) to identify the degree to which different plant tissue carbohydrate constituents have distinct stable isotopic signatures (cellulose being typically more enriched in 13C values relative to lignin; Deines, 1980), and (2) determine the diagenetic alteration of the leaf material following death (Benner et al., 1987). Resins were removed by sequential treatment in benzene–ethanol (2 : 1 by volume), ethanol and distilled water solutions within a Soxhlet apparatus (TAPPI, 1987). The lignin component is defined here as the acid-insoluble component when digested in 72% H2SO4 (the ‘Klason’ lignin) (TAPPI, 1988). All lignin samples were dried overnight at < 70°C.

Samples for stable isotope analysis were ground to a fine powder. Between 1 mg and 5 mg of sample was weighed into tin capsules for automated combustion in a ANCA–SL elemental analyser interfaced to a continuous flow Europa Geo 20/20 isotope ratio mass spectrometer at the Institute for Geological and Nuclear Sciences (IGNS), Lower Hutt, New Zealand. The carbon dioxide and nitrogen gases were resolved using chromatographic separation on a gas chromatography column at 85°C, and analysed for isotopic abundance as well as total organic carbon and nitrogen. Standards and blanks were included during the run for internal calibration. The δ13C values are reported as per mille (‰) relative to the Cretaceous belemnite sample recovered from the Peedee formation in South Carolina (PDB). Analytical precision for δ13C at the 1σ level is reported as 0.15‰.

Results and Discussion

Contemporary study

Canopy Leaf mass per unit area (g m−2) was found to increase exponentially through the canopy for both live (Year 0T) and senesced (Year 1T) leaves, from 125 to 220 g m−2 (Fig. 2) with a significant a significant drop in the leaf mass per unit area in the uppermost 1.5 m, indicating a strong change in imposed environmental conditions. There does appear to be a partial offset between the mean values obtained from the two leaf types (approximately 12% at 18 m height), with slightly lower leaf masses per unit area recorded in the senesced leaves (Year 1T), suggesting some remobilization of organic compounds. However, the overall trend indicates that fully sunlit leaves are generally thicker and smaller than those in the shaded, lower parts of the canopy, confirming the findings of Hollinger (1989). The relatively large error bars (shown at 1 SE) at the top of the canopy are probably a reflection of some partly shaded leaves being included in the analysis. Nevertheless, it can be clearly seen that there is a significant marked change in leaf mass per unit area through the canopy at 15 m, independent of whether the leaves are live or in a state of senescence (i.e. the relationship appears to be preserved following death).

Figure 2.

Variation of leaf mass per unit area in live (open circles; samples of leaves on the tree that were 4 months old) and senesced (closed circles; newly senesced material aged approximately 13 months on the tree) mountain beech leaves through a closed canopy, Craigieburn, New Zealand. Error bars are defined as 1 SE.

Over 80% of light incident on the top of the canopy is absorbed or reflected in the top 3 m of the canopy (i.e. above 15 m; Fig. 3). This is an extremely strong environmental gradient over a relatively short distance and probably also reflects equally large gradients in temperature and VPD over a similar distance. Therefore, only the upper few metres of the canopy are exposed to the wider regional environmental conditions and leaves deeper in the canopy become progressively decoupled from the above canopy climate. A marked change in leaf mass per unit area, along with C and N per unit mass of leaf material occurs at about 15 m above ground. This is associated with cumulative leaf area index exceeding 2 and light transmission falling below 12% of the above-canopy value (Fig. 3), making the former an unsuitable surrogate for contemporary and palaeo-leaf position.

Figure 3.

Leaf area index by layer (a, data from Hollinger, 1989 ), cumulative leaf area index calculated from the top of the canopy (b) and modelled light transmission from the top of the canopy (c), for a mountain beech stand at Craigieburn, New Zealand. LAI, leaf area index.

The nitrogen content per unit area of live leaves also shows an exponential relationship within the canopy (Fig. 4), with a generally exaggerated profile for the lignin samples. A difference of 2.5 g m−2 is recorded between the lower and upper parts of the canopy for lignin derived from live leaf material (Year 0T; Fig. 4). Lignin is enriched in nitrogen compared with bulk material by up to approximately 1.8 g m−2. Nitrogen is largely associated with photosynthetic proteins, such as ribulose 1,5-biphosphate carboxylase, and is commonly taken to be an indicator of photosynthetic capacity (Evans, 1989), hence the highest values in the upper part of the canopy. However, this signal is not preserved in the senesced leaves (Year 1T; Fig. 4), with both the bulk and lignin nitrogen contents having low values through the canopy, averaging 0.5 g m−2 and 0.6 g m−2, respectively, indicating that the remobilization of nitrogen from lignin and bulk samples occurs prior to death. Thus, nitrogen does not appear suitable for identifying the original position of fossil leaves within a canopy.

Figure 4.

Nitrogen (N) content per unit area in lignin and bulk live (circles; samples of leaves that were 4 months old on the tree) and senesced (squares; newly senesced material aged approximately 13 months on the tree) leaves through a closed canopy of mountain beech, Craigieburn, New Zealand: open symbols, bulk; closed symbols, lignin.

An exponential increase with height is recorded in the carbon content (per unit area) of the live and senescing leaves through the canopy (Fig. 5), with a generally more exaggerated profile recorded in the lignin samples. A higher carbon content is found in the live material compared with senescing leaves at all points through the canopy with values ranging from between 80 g m−2 and 155 g m−2 from the lower to the upper parts of the canopy for live lignin material. Lignin is enriched in carbon compared with bulk material by up to 50 g m−2 (Fig. 5). By contrast to nitrogen, the carbon content profile of the leaves appears to be maintained during senescence.

Figure 5.

Carbon (C) content per unit area in lignin and bulk live (circles; samples of leaves that were 4 months old on the tree) and senesced (squares; newly senesced material aged approximately 13 months on the tree) leaves through a closed canopy of mountain beech, Craigieburn, New Zealand: open symbols, bulk; closed symbols, lignin.

An enrichment in 13C up through the canopy leaves occurs in both the bulk and lignin of live and senesced samples (Fig. 6), though whether this is driven by irradiance, atmospheric moisture content or respired CO2 is uncertain. For both live and senesced leaves, the lignin component was always more depleted in 13C (by 3‰) as expected (cf. Benner et al., 1987). In addition, the δ13C values of sunlit material can be seen to be significantly heavier (3‰ for both bulk and lignin) for the most recent material (Year 0T) compared with the previous years growth (Year 1T). Such a difference in δ13C is not unexpected and is almost certainly related to changes in environmental conditions between the two growing seasons. The season 1997–98 was a particularly intense El Niño period (Takayabu et al., 1999) which, at the Craigieburn site, was manifested as increased aridity, causing a high VPD and therefore resulting in reduced stomatal conductance and enriched 13C content of plant tissues (cf. Turney et al., 1999).

Figure 6.

δ 13 C in lignin and bulk live (circles; samples of leaves that were 4 months old on the tree) and senesced (squares; newly senesced material aged approximately 13 months on the tree) leaves through a closed canopy of mountain beech, Craigieburn, New Zealand: open symbols, bulk; closed symbols, lignin.

Leaf litter

To test whether the trends observed through the canopy are maintained following leaf fall, analysis was made on samples identified in the litter produced in the previous two years’ growth. A frequency histogram of leaf litter for this period is given in Fig. 7. In a sample of 20 leaves for the most recent leaf fall (Year 1L), the leaf mass per unit area values ranged between 160 g m−2 and 270 g m−2. These values are represented in the senesced leaves within the uppermost 3 m of the canopy (Fig. 2), and suggest that leaf litter fall is dominated by material originating from this part of the canopy. Values > 220 g m−2 are considered here to be representative of full sunlit conditions (Fig. 2) and therefore comprise 30% of the total litter fall for that year.

Figure 7.

Frequency histogram of leaf mass per unit area of leaf litter from 1996–1997 (black, Year 1L: newly senesced material within the leaf litter aged between 13 months and 2 yrs) and 1995–1996 (grey, Year 2L: approximately 2-yr-old dead material in the leaf litter) for mountain beech, Craigieburn, New Zealand.

A similar-shaped histogram to the above can be seen in leaf material identified from the previous growing season, 1995–96 (Year 2L) (Fig. 7), although the range in values was between 60 g m−2 and 200 g m−2. Assuming that the difference in leaf mass per unit area is due solely to diagenetic processes and that all leaves decompose at the same rate, a value of 30% was used to identify a value of > 130 g m−2 as representative of sunlit leaf fall material for the Year 2L litter. The mean value of Year 2L sunlit samples (158 g m−2) experienced a loss of 94 g m−2, or 37% relative to live material at a height of 18 m (Year 0T), suggesting that leaf material will experience rapid loss of carbon content if not buried relatively quickly.

The nitrogen content per unit area of these leaves, as discussed above, cannot be used to identify whether they originated from the upper part of the canopy and this can be seen in the negligible differences in values between samples (Fig. 8). In contrast, the carbon content of the heaviest leaf mass per unit area litter fraction (> 220 g m−2; Fig. 9) was always the largest, and this is consistent with the view that these leaves originated from the uppermost, sunlit part of the canopy.

Figure 8.

Nitrogen (N) content per unit area in lignin (closed circles) and bulk (open circles) leaf litter material from Year 1L (newly senesced material within the leaf litter aged between 13 months and 2 yrs) and Year 2L (approximately 2-yr-old dead material in the leaf litter). Each year is divided into sunlit and shaded leaves by leaf mass per unit area. Mean values of canopy sunlit material are also shown. Year 1T, newly senesced material aged approximately 13 months on the tree.

Figure 9.

Carbon (C) content per unit leaf area in lignin (closed circles) and bulk (open circles) leaf litter material from Year 1L (newly senesced material within the leaf litter aged between 13 months and 2 yrs) and Year 2L (approximately 2-yr-old dead material in the leaf litter). Each year is divided into sunlit and shaded leaves by leaf mass per unit area. Mean values of canopy sunlit material are also shown. Year 1T, newly senesced material aged approximately 13 months on the tree.

The δ13C values (Fig. 10) for the > 220 g m−2 fraction were consistently more 13C-enriched than those of the lighter leaf material, confirming a full sunlit origin, representative of the ambient, atmospheric conditions during carbon fixation. These values support the proposal that sunlit leaf material from above 15 m can be separated purely by their leaf mass per unit area. Intriguingly, however, the absolute values obtained from litter leaves, when compared with the senescent leaves remaining on the tree, are more enriched in 13C (by up to 1.7‰) (Figs 6 and 10), despite supposedly being formed during the same growing season (Year 1). Such a shift in isotopic values between both bulk and lignin material obtained from Years 1T and 1L indicates a fractionation operating immediately prior to leaf fall. The mechanism for this is uncertain, but it is possibly related to carbon remobilization operating within the tree during the final stages of senescence (Proe et al., 2000; Yang et al., 2001), with the utilization of lighter carbon to other essential organs on or in the tree, resulting in an enrichment in 13C within the leaf material. Such an isotopic shift is unlikely to have been increased as a result of decreased carbon assimilation during an El Niño (Yang et al., 2001). The difference between the leaf components (lignin vs bulk, sunlit vs shaded) during Years 1 and 2 are statistically indistinguishable (Fig. 10), despite the latter not senescing during an El Niño year.

Figure 10.

δ 13 C in lignin (closed circles) and bulk (open circles) leaf litter material from Year 1L (newly senesced material within the leaf litter aged between 13 months and 2 yrs) and Year 2L (approximately 2-yr-old dead material in the leaf litter). Each year is divided into sunlit and shaded leaves by leaf mass per unit area. Mean values of canopy sunlit material and isotopic differences (triangles) between the components are also shown. Year 1T, newly senesced material aged approximately 13 months on the tree.

Fossil study

Although the leaf mass per unit area decreased owing to burial of the fossil material, it is still valid to use leaves with the highest specific leaf weights and treat them as sunlit leaves. Nitrogen plays an important role in decomposition, but since both sunlit and shaded leaves have the same nitrogen content when they fall, then we made the assumption that they decompose at the same rate.

The leaf mass per unit area from Mt George range between 15 g m−2 (Unit D/E; 6.24–6.26 m) and 115 g m−2 (unit D/E; 6.04–6.06 m) (Figs 1 and 11). These values, which are approximately half those recorded for present-day leaf litter mountain beech noted earlier, are surprisingly high given their age. Field observations indicate that even under dry conditions, beech lose 80% of their leaf mass in 2 yr, and quickly lose structural integrity. The excellent state of preservation and the structural integrity of the fossil leaves we examined indicates that the leaves must have been buried within a year of falling from the trees. There is a remarkable similarity between the frequency distributions of leaves in each sampling unit, as defined by their mass per unit area, with almost all showing the same distribution. Using the protocol outlined above, the heaviest 30% of leaves from each horizon were interpreted as having originated from a sunlit position within the canopy.

Figure 11.

Frequency distribution of fossil leaf mass per unit area for 11 specific levels collected from Mt George, West Coast, New Zealand.

Observation and modelling studies (Hedges et al., 1985; Benner et al., 1987; Spiker & Hatcher, 1987) indicate that upon burial, the stable carbon isotopic composition of bulk plant tissues is altered following the preferential removal of 13C-enriched polysaccharides (e.g. cellulose) by biogeochemical processes. The remaining bulk tissue is therefore believed to become lignin-enriched, which is reflected in lighter δ13C values (typical of lignin). To test the above, carbon isotopic values have been reported for lignin and bulk terrestrial plant macrofossils preserved in sediments approximately the same age as the deposits investigated at Mt George. δ13C in fossil cladodes of P. alpinus (a small coniferous tree) preserved under the Kawakawa Tephra (22 600 14C BP) on North Island, New Zealand, were statistically indistinguishable from that of the bulk whole tissue (Turney et al., 1999). Thus, despite the significantly different isotopic signatures of organic compounds within the tissue during growth, the fossil bulk material is dominated by the lignin δ13C signature and can be considered a relatively robust measure of the latter. Fractionation of lignin δ13C following burial plays a minor role (Marino & DeNiro, 1987; Spiker & Hatcher, 1987) and is not considered to have played a significant role here.

At Mt George, in almost all instances, the δ13C values from leaves interpreted to be of sunlit origin were invariably the most enriched in 13C (Fig. 12). This is further supported by the higher carbon contents in the ‘sunlit’ leaves. Differences in δ13C between sunlit and shaded leaves were at most 1.5‰ between extreme values. Overall, the δ13C relatively complacent, though with a partial enrichment in sunlit 13C of 1‰ between 6.16 m and 6.19 m (soil horizon unit D/E) which is not so clearly recorded in the shaded leaves. ‘Sunlit’ leaves do, therefore, appear more responsive to climatic variability than the ‘shaded’ leaves, and possibly indicate a shift to a greater VPD (Turney et al., 1999) at the time of deposition (Fig. 12). Although the rate of sedimentary deposition (and therefore sample duration) is uncertain owing to the paucity of ages, these results are suggestive of relatively stable environmental conditions (Turney et al., 1999) during the period of leaf formation.

Figure 12.

Variations in leaf δ 13 C and carbon content on sunlit (open circles) and shaded (closed circles) leaves extracted from sequence at Mt George, West Coast, New Zealand. Best-fit (mean) line for each data set is shown. Radiocarbon sample shown as a solid square.

Unfortunately, the identification of shifts in δ13C values of the mountain beech leaves immediately prior to leaf fall during senescence precludes a quantified reconstruction of environmental change during the Late Otiran Glaciation in this area, although relative changes may be possible. It is uncertain as to what leaf component is most suitable for palaeoclimate reconstruction. Previous work has focused on stable isotopic values obtained from plant material remaining on the plant prior to fall (Turney et al., 1999). However, the results from Craigieburn demonstrate that absolute values differ considerably between the canopy and the fall material, indicating a change in isotopic composition before fall, and calling into question calibrations of δ13C vs climatic parameters that are based on ‘live’ material. It is possible that we can assume that changes occurring during senescence are relatively constant both temporally and spatially and therefore calibrations are robust enough for determining relative changes in climate. Conversely, such a process may be only significant in long-lived mature, plant stands and relatively insignificant for shorter-lived vegetation with little or no carbon storage over the winter period. It must be assumed that the fossil material is dominated by leaves that have senesced naturally. It therefore seems likely that that any contemporary calibration data set must therefore be based on a senesced data set rather than material collected from the canopy.


The results presented here indicate that several parameters can be used to identify the original position of leaf material within a closed evergreen canopy. These parameters include the leaf mass per unit area, carbon mass per unit area and the δ13C value, and are relatively less labour intensive to obtain than alternative methods (e.g. stomatal counts). For carbon mass per unit area and the δ13C, this relationship holds for both bulk and lignin material, suggesting that during decomposition, fossil material can still potentially record its former position within the canopy, despite complex changes in both carbon content and δ13C between living and senescent leaves. In a practical sense, the leaf mass per unit area provides the most appropriate, nondestructive means to identify the former position within the canopy (and therefore have a closer representation of past climate, which is less confused by leaves that occupy the lower portions of the canopy), prior to stable isotope measurement with only a few assumptions about the original canopy, as shown in the results from the Mt George site. Although the δ13C values obtained from Mt George indicate relatively constant climatic conditions during part of the Late Otiran Glaciation, the results from the contemporary study suggest that carbon is remobilized during senescence, altering the δ13C signature of the leaves and therefore potentially biasing any quantitative environmental reconstruction.


Thanks to David Whitehead and Mike Ryan who collected the contemporary samples for us and to Karyne Rogers (IGNS, Lower Hutt, New Zealand) who ran the samples for C, N and δ13C. In addition, we would like to thank Dan Hammarlund, Matt McGlone, Neville Moar, Janet Wilmshurst and two anonymous referees for reading earlier drafts of the manuscript and providing informative discussions. This research was funded by the New Zealand Foundation for Research, Science and Technology (Grant No. C09525).