Introduction

Global climate warming is a significant ecological and environmental issue (Frölicher et al 2014). Enhancing the carbon (C) sink of terrestrial ecosystems, especially forests, is an effective way to mitigate global warming (Wang et al 2021). Trees in forest ecosystems are the largest and most dynamic C reservoirs and have the potential to sequester large amounts of C (Erb et al 2018; Girardin et al 2021). Forest carbon storage is estimated by multiplying tree biomass by the corresponding C concentration (i.e., the mass of C per unit dry biomass, Thomas and Martin 2012; Doraisami et al 2022). In addition, C levels are also indispensable to accurately assess ecological stoichiometry and nutrient use efficiency (Elser et al 2010; Zhang et al 2018). Therefore, plant C is a fundamental trait for estimating carbon storage and nutrient utilization.

Although C is one of the most abundant and conservative elements in living plants compared with other life elements (Zhao et al 2016; Tang et al 2018), its concentration varies significantly among plant tissues (organs), tree species, plant functional groups, and climatic zones (Ma et al 2018; Martin et al 2018; Paroshy et al 2021). Carbon concentrations of living tree stems ranged from 305 to 607 mg g–1 (Doraisami et al 2022), i.e., the highest value was almost double the lowest one. Neglecting the C concentration differences between tissues and species by using the canonical value (e.g., 500 mg g–1 (Lamlom and Savidge 2003; Doraisami et al 2022) to estimate forest C stocks can cause significant errors, and may exceed 10% in some cases (Elias and Potvin 2003; Bert and Danjon 2006; Zhang et al 2009; Thomas and Martin 2012; Martin et al 2018). Therefore, accurately determining the C concentration of various tree species can provide essential data for improving carbon stocks and carbon cycling studies at global, regional, and local scales (Doraisami et al 2022).

The patterns and mechanisms of intra- and inter-specific variations in C concentration of trees remain unclear (Paroshy et al 2021). First, there is controversy regarding C concentration variations among different tissues. Foliage is often considered to have the highest C levels (Zhang et al 2009; Widagdo et al 2021), but this is not supported by some studies (Pompa-García et al 2017). The stem is a highly heterogeneous organ, with bark generally having higher carbon levels than wood (Gao et al 2016; Doraisami et al 2022); and heartwood generally slightly higher than sapwood, although there are exceptions (Widagdo et al 2021). Unfortunately, most studies have not differentiated between stem tissues; global dataset on wood tissue C concentration includes only 33 sapwood and 28 heartwood samples (Doraisami et al 2022). Therefore, for accurate C accounting, there is a need for more detailed and systematic data to further understand the differences in carbon levels among tree tissues.

Second, the mechanisms underlying inter-specific variation in tree C, even at a local scale, are not well understood. Wood of coniferous species generally has higher carbon levels compared with broadleaved species (Lamlom and Savidge 2003; Martin et al 2018). However, there is limited research on the comparison of other tissues between coniferous and broadleaved species. Tree size and age have been found to have a positive effect on C concentration (Ma et al 2020; Bārdule et al 2021), but these relations may also be influenced by shade tolerance of the species (Gao et al 2016). In addition, wood density (or specific gravity) may be positively (Elias and Potvin 2003) or negatively (Martin et al 2018) related to C concentration in wood, but reasons for these discrepancies are unclear. Growth rate has a negative (Zhang et al 2009) or insignificant (Martin and Thomas 2011) effect on carbon levels, while diameter at breast height (DBH) or height (H) has significantly positive (Ma et al 2020) or is insignificant (Bert and Danjon 2006). For these controversies, one key issue is that few studies have considered the effect of wood porosity (non-porous, diffuse-porous, ring-porous), a well-known property that indicates functional differentiation in transport, support and storage tissues (McCulloh et al 2010; Tyree and Zimmermann 2002; Zhang et al 2014). Non-porous wood has only tracheids for both transport and support, and has low storage space (parenchyma cells) of nonstructural carbohydrates and nutrients. Diffuse-porous wood has vessels for more efficient transport and fibers for support, and large storage space. The most advanced wood type, ring-porous wood, has the largest vessels in the earlywood and thus the largest storage space for collaboration of transport (Spicer 2016). However, how wood porosity relates to C concentrations remains unclear.

Temperate forests in northeast China are an important part of East Asian temperate forests (Wang 2006). Nine tissues were sampled, foliage, branches, bark, sapwood, heartwood, stump, coarse roots, medium roots, and fine roots, from 216 individuals of 32 tree species. The objectives of this study were to: (1) explore the inter-tissue and inter-specific variations in C concentration; (2) investigate how it is affected by wood density, tree size and growth rate; and (3) determine how wood porosity mediates C concentration variations and correlation.

Materials and methods

Site description

The sampling site is in the Muling Taxus cuspidata National Nature Reserve in Heilongjiang Province, China (130°00′ − 130°28′ E, 43°49′ − 44°06′ N). The Reserve is in the Laoyeling Mountains of the Changbai Mountain Range, with average elevations of 500–700 m a. s. l. The region has a temperate continental monsoon climate of the mid-latitude zone, characterized by long, cold, dry winters, humid and rainy summers, and rapid temperature changes during the transitional seasons of spring and autumn. In 2020, mean air temperature was 3.8 °C, and the total annual precipitation 683 mm (Dong et al 2022). The typical vegetation is a temperate mixed forest of conifers and broadleaved species, with Alfisols being the main soil type based on the United States Soil Taxonomy.

Species and sample collection

The richness in species made it possible to explore the effect of wood porosity on tissue C concentration of a wide variety of species. A total of 32 tree species were selected, with a minimum of six healthy and mature individuals for each species (Table 1). Samples of nine tissues (foliage, branch, stem bark, sapwood, heartwood, stump, coarse roots (diameter > 5 mm), medium roots (diameter 2–5 mm), and fine roots (diameter < 2 mm)) were collected from late July to mid-August 2020. For each tree, the diameter at breast height (DBH) and height (H) were measured, and nine tissues sampled. To collect foliage and branch samples representing the entire canopy, a hired tree climber worked at the lowest to the mid canopy, and a standard first-order branch from the mid to upper part of the canopy was sawn off. Different diameter branch segments were mixed (not only the small twigs), and sunny and semi-sunny foliage randomly collected and mixed as foliage samples. A bark sample was taken at breast height using a chisel. Three to five increment cores were extracted at breast height with an increment borer (with an inner diameter of 5.15 mm), one core was used to determine annual ring widths and the relative growth rate (RGR); the remaining cores were used for sapwood and heartwood C measurements. Stem tissues sampled at breast height ignored any vertical C variation (Bert and Danjon 2006; Herrero de Aza et al 2011). Stump samples were also collected with an increment borer and root systems excavated using a pickaxe and shovel in the top 20 cm. All samples were numbered and placed in sealed bags for transport to the laboratory for further treatment.

Table 1 Characteristics of sample trees
Full size table

The sapwood and heartwood of the wood cores were distinguished based on transparency and color (Dong et al 2022), with B. platyphylla and T. amurensis considered as full sapwood species (Zhang et al 2013). The root samples were washed to remove soil and separated into coarse (diameter > 5 mm), medium (diameter 2–5 mm), fine (diameter < 2 mm). Fresh samples were microwaved at 600 W for 90 s to eliminate enzymatic activity, then dried to a constant weight at 70 °C, ground to pass through a 0.2-mm sieve, and stored in sealed containers. The 32 species were classified as non-porous, diffuse-porous, and ring-porous based on wood properties (Dong et al 2022), reflecting the water conducting tissue evolution (Tyree and Zimmermann 2002).

Tissue density of foliage, sapwood and heartwood were also measured. Foliage volume was measured using the drainage method (Chen et al 1997), after rehydration by soaking three leaves in distilled water and avoiding light for at least 6 h (Garnier et al 2001). Sapwood and heartwood volumes were calculated by the diameter and length of the wood core (considered as a cylinder). The samples were then oven-dried at 75 °C for 48 h. Dry mass was weighted to the nearest 0.001 g. Tissue density (in g cm–3) was calculated as the ratio of dry mass to the fresh volume (Pérez-Harguindeguy et al 2013).

Carbon measurement

C concentration was determined using the common dry combustion method (Zhang et al 2009). Approximately 40 mg of powdered sample was dried at 60 °C for a minimum of 2 h to remove water and weighed using an analytical balance with an accuracy of 0.1 mg. The carbon was measured using a total organic carbon/total nitrogen analyzer (Multi N/C 2100 S, Analytik Jena AG, Germany) at 1200 °C for 90 s for each sample. For every 20 samples, a measurement of calcium carbonate was used to check the accuracy of the analyzer. C concentration was expressed as mg C mass per 1000 mg dry mass (mg g–1). This method did not include volatile C components (Jones and O'Hara 2016).

Measurement of mean annual ring width and biomass

The width of each annual ring in the last 10 years was measured for each tree. After scanning the polished core samples with a scanner, the images were measured for ring width using Image J 1.52i, with a measurement accuracy of 0.001 mm. Because biomass allometric equations were not available for each species, the absolute rate of growth was expressed as the mean ring width, which had the advantage of being direct but was influenced by the size of the individual tree. The relative growth rate (RGR) was calculated as the ratio of the annual increment in basal area to the basal area of the previous year (Bevilacqua 2002), which partially eliminated the effect of tree size. To minimize the influence of interannual fluctuations, the mean width of the tree rings in the past 10 years was used to calculate both absolute and relative growth rates.

Statistical analysis

Inter-tissue and inter-specific variations were expressed by the coefficient of variation. For inter-specific variation, all individual tissues of each species were averaged to reduce the influence of different sample numbers among species. Pearson correlation was used to test the inter-correlations of carbon concentrations between tissues. Differences in carbon levels between wood porosity types were tested with ANOVA and the Duncan Post-Hoc multiple comparison. Relationships of tissue C with foliage and wood density, tree size (DBH and H) and growth rate (annual ring width, and relative basal area growth rate, RGR) were tested with the Pearson correlation for each porosity type. Phylogenetic effects were analyzed using the “Picante” R package (Kembel et al 2010). The results were insignificant (P > 0.07) and generally weak (Blomberg’s K values < 0.15) for all tissue types, thus the phylogenetic effect was not considered. For the final step, the random forest model (Breiman 2001) was used to quantify the relative importance of ten exploratory variables on carbon variation. The ten variables were: tissue type, species, DBH, H, foliage density, sapwood density, heartwood density, annual ring width, and RGR. The importance of each variable was calculated based on the increase in node purity of 1000 regression trees and was standardized to sum to 100%.

Results

Inter-tissue variation

When the 32 species were pooled together, there were significant differences in C levels among different tissues (P < 0.001). For all species (Fig. 1), it was clear that carbon was higher in metabolically active terminal tissues (foliage and fine roots) compared with less metabolically active tissues (branches, stem, coarse roots). For each porosity group, the trend above was evident in diffuse-porous species, while ring-porous species showed a gradual increase in C from foliage to fine roots (Fig. S1). Tissues with the highest average C in non-porous, diffuse-porous, and ring-porous species were foliage, medium roots, and fine roots, respectively, while the corresponding lowest carbon concentrations were in sapwood, heartwood, and bark.

Fig. 1
figure 1

Carbon concentration in different tissues of the 32 temperate forest species. [CRoot: coarse roots, MRoot: medium roots, FRoot: fine roots.] Error bars are standard errors; different letters indicate significant differences between tissues

Full size image

The coefficient of variation among tissues in C (CVT) for the 32 species was within 4% for most species (Fig. 2). Among the non-porous wood species, CVT was higher in P. koraiensis and lowest in A. holophylla. Among the diffuse-porous wood species, CVT was highest in T. amurensis and lowest in A. tegmentosum. And for the ring-porous wood species, the CVT was highest in A. elata and lowest in M. amurensis.

Fig. 2
figure 2

Comparison of the coefficients of inter-tissue variations of carbon concentration by species

Full size image

Inter-specific variation

In contrast to our expectation, there was no consistent pattern in the magnitude of C levels among the nine tissues of the 32 species (Figs. S2 and S3), indicating a generally poor coordination among tissues (Fig. 3). The correlation coefficient was highest between foliage and coarse roots (R = 0.62), but that of sapwood or heartwood with other tissues was poor.

Fig. 3
figure 3

Inter-tissue correlation matrix of carbon concentration across 32 tree species

Full size image

Among the 216 sample trees of the 32 species, the fluctuation of C in aboveground tissues was between 367 and 536 mg g–1, while the variation in belowground tissues was 403 and 538 mg g–1. The mean coefficient of inter-specific variation (CVS) was within 3% for most tissues (Fig. 4), the CVS for sapwood and fine roots was relatively high compared with other tissues.

Fig. 4
figure 4

Comparisons of coefficient of inter-specific variation of carbon concentration in different tissues

Full size image

Determinants of carbon variations

The 32 species were grouped into three porosity types. C levels followed the pattern of non-porous > diffuse-porous > ring-porous (Fig. 5). Specifically, the mean C concentration of non-porous species was significantly higher than that of ring-porous species, excluding sapwood and heartwood. There were significant differences in carbon content in branch, heartwood, and stump between non-porous and diffuse-porous species. In addition, carbon in foliage, bark, coarse roots, and medium roots was higher in diffuse-porous species than in ring-porous species.

Fig. 5
figure 5

Comparisons of carbon levels between wood porosity types by tissue. Different lower case letters indicate significant differences between porosity types within tissues

Full size image

Correlations between C concentration and tissue density diverged among the three porosity types (Fig. 6). For ring-porous species, foliage and sapwood C concentration was positively related to corresponding tissue densities (Figs. 6a, b). However, C levels were positively related to heartwood density for non-porous and diffuse-porous species (Fig. 6c).

Fig. 6
figure 6

Divergent correlations of carbon levels with tissue density among porosity types. Density was measured only for foliage, sapwood, and heartwood, thus only these tissues were analyzed

Full size image

The carbon levels of most tissues, except for the heartwood, was positively correlated with tree size and negatively with growth rate (Table 2). DBH was positively correlated with carbon levels in foliage, bark, sapwood, tree stump, coarse roots, medium roots, and fine roots, but negatively correlated with heartwood. Tree height was significantly positively correlated with carbon concentrations in bark and medium roots but negatively with heartwood. Mean annual ring width was negatively correlated with carbon levels in sapwood, heartwood, coarse roots, and medium roots, although it was positively correlated with bark carbon. Relative growth rate was negatively correlated with carbon levels in sapwood, heartwood, coarse roots, and medium roots, while it was positively correlated with branch carbon values.

Table 2 Correlation coefficient (R) of carbon concentration in different tissues with tree size and growth rate for 32 species
Full size table

The relationship between carbon levels and tree size differed among the three porosity types. For non-porous species, foliage carbon, was positively correlated with DBH, while carbon in the sapwood and stump was negatively correlated. For diffuse-porous species, bark and sapwood carbon was positively related to DBH (Fig. S4). For ring-porous wood species, carbon levels in branches, stumps, and roots were positively correlated with DBH, while heartwood carbon was negatively correlated with DBH. Most tissues showed a positive correlation between C concentration and height but again, there were distinct differences among the three porosity types (Fig. S5). The relationship between carbon in different tissues and annual ring width varied considerably among different wood types (Fig. S6). The number of tissues with a negative correlation between C concentration and RGR increased from non-porous to ring-porous wood (Fig. S7).

The results of the random forest model indicated that the most important single variable to tissue carbon variations was tissue type (contributed 33% to the total explanations), and the second important determinant was species (contributed 12%). Tissue density (together 21%), tree size (DBH and H together contributed 14%), growth rate (together 13%) and wood type (7%) had relative low importance (Fig. 7).

Fig. 7
figure 7

Relative importance of each factor in the random forest model to the variability in tissue carbon of 28 coexisting species. The coefficient of determination (R2) is 31.32%. T. amurensis and B. platyphylla were excluded due to an absence of heartwood, and A. arguta and A. elata were excluded because of a lack of growth rate or foliage density data

Full size image

Discussion

Our dataset is the current richest in species and woody tissue on carbon concentration in northeast China. The previous most species- and tissue-rich dataset were 14 (Widagdo et al 2021). Others are for tissues of 10 species in Maoershan region (Zhang et al 2009), and wood C values of 14 species in Changbaishan Nature Reserve (Thomas and Malczewski 2007). Our dataset will help to improve regional estimates and comparison of C stock among species.

Using the C concentrations of nine tissues from 216 individuals of 32 species and the auxiliary data, we found that metabolically active tissues had higher carbon levels except for the foliage of ring-porous species. Unexpectedly, the inter-correlations among tissue C concentrations were generally poor. As xylem evolved from non-porous wood to ring-porous wood, carbon decreased in most tissues. Inter-specific variations were positively correlated with tissue density, tree size and negatively with growth rate. These findings provide insights on the variability of tissue C concentrations of temperate forest species, and have important implications for species in other biomes.

Inter-tissue carbon concentration variation

Metabolically active tissues (foliage and fine roots) had higher carbon levels compared with less metabolically active ones (branches, stem, coarse roots) for non-porous and diffuse-porous wood species. This is consistent with our results for ten broadleaved species in a natural secondary forest (Zhao et al 2021). The highest average C value was consistently in the foliage from a global synthesis (Ma et al 2020). In the largest database of woody tissue C concentration (Doraisami et al 2022), the highest carbon value (751 mg g–1) was in fine roots. The lowest foliage C of ring-porous species might be partly due to the three U. davidiana varieties (Fig. S2). The low foliage C of P. amurense and U. laciniata were consistent with those of Widagdo et al (2021). Overall, the C variation among the nine tissues in this study was more important than inter-specific variation based on the random forest model (Fig. 6), which is consistent with that for woody tissues, including bark, based on the nested ANOVA (Martin et al 2015). This finding indicates that, using the more commonly measured stem C value (Doraisami et al 2022) to represent the whole-tree carbon content might be worse than using a biomass-weighted mean C concentration of dominant species (Zhang et al 2009; Widagdo et al 2021) when estimating stand carbon density.

For non-porous species, the bark had higher C levels compared to wood, and heartwood generally had higher C concentration than sapwood (Figs. 1 and S2). There are few studies comparing stem components, such as bark versus wood, sapwood versus heartwood (Doraisami et al 2022). Studies in Canada (Martin et al 2015) and northeast China (Widagdo et al 2021) support these findings. However, for ring-porous species, bark C was relatively lower compared with wood, which is also supported by Widagdo et al (2021). The biomass proportion of bark, sapwood and heartwood varied among species and changed as trees grew (Wang et al 2010; Zhang et al 2014), and carbon levels changed across the stem cross section and diverged among wood porosity types. These findings indicate that, using a single increment core to represent the whole stem would introduce additional uncertainties in carbon accounting, similar to nutrient concentration estimation (Augusto and Bert 2005).

We also found that C concentration had weak linear relationships among the nine tissues examined (Fig. 3). Moderate correlations between tissues were found in a recent synthesis (Ma et al 2020), and high correlations have also been reported (Thomas and Martin 2012; Martin et al 2015). The weak correlations between tissues in this study are consistent with those between bark and sapwood across 17 temperate tree species in Canada (Martin et al 2015), between bark and wood (Ma et al 2020), and between foliage and wood in a previous global dataset (Thomas and Martin 2012). Martin et al (2015) found that sapwood C levels were good predictors of concentrations in heartwood, small branches, and coarse roots across 17 temperate tree species. We suggest that the differences in the range of C values and sample size may have an apparent role on the divergences between studies. However, the mechanisms are still poorly understood further studies are required at various spatial scales. Nevertheless, the weak coordination in C concentration among tissues requires the measurement of multiple organs or tissues, the same as for nutrients such as nitrogen and phosphorous (Martin et al 2015; Zhao et al 2021).

Effect of wood porosity type

It is well known that conifers have higher carbon levels than broadleaved species (Ma et al 2018), at least for wood (Thomas and Martin 2012; Pompa-García et al 2017; Ma et al 2018). We further found that diffuse-porous wood species had higher C concentrations versus ring-porous species for foliage, bark and roots, a pattern in angiosperms not previously addressed. The nearly systematic differences among the three wood porosity types (Fig. 1) may be related to xylem vascular tissue and supporting tissues. As wood evolves from non-porous to ring-porous, the xylem changes from tracheid (vascular and supporting function) to small size vessel and then large size vessel (Tyree and Zimmermann 2002; Spicer 2016), and lignin decreases while cellulose increases. Because lignin has higher C content than cellulose (Lamlom and Savidge 2003), C generally declines from non-porous to ring-porous species.

Because ring-porous species are mainly distributed in northern temperate zones (Wheeler et al 2007), we suggest that considering wood porosity will improve biomass C estimation in temperate forests of the Northern Hemisphere. In addition, exploring potential differences in other aspects among wood porosity types, and not only restricted to water transport and storage of water and nutrients (Spicer 2016), will enhance our understand of xylem evolution and functional adaptation.

Relations to wood density, tree size, growth rate

We found carbon levels often related positively to tissue density across the 32 sympatric temperate species (Fig. 5). This is consistent with that of nine species in central Panama (Elias and Potvin 2003) but contrary to global negative relationships (Martin et al 2018). Other studies have also reported no significant correlations between C concentration and wood density (Thomas and Malczewski 2007; Zhang et al 2009; Martin and Thomas 2011). Our results indicate that the correlations of C values with wood density was the result of wood porosity type, which partly addresses the contradiction among studies with various proportions of wood porosity types. In fact, the global negative relationships between C concentrations and wood density exist among angiosperms but not among gymnosperms (Martin et al 2018). However, the underlying mechanism linking wood density with carbon content is still unclear. A better understanding of the nature of wood density as an integrative wood or plant trait (Chave et al 2009; Larjavaara and Muller-Landau 2010) will help to interpret the relationships of wood density with C concentration in the context of C, water and nutrient economics.

Tree size often had positive effects on carbon levels (Table 2). Positive size effects on stem C were reported for a boreal forest (Gao et al 2016), a tropical rain forest (Thripob et al 2022) and at a global scale (Ma et al 2020). However, there were no significant correlations of carbon content with tree size for Pinus pinaster (Bert and Danjon 2006). Our negative correlation of heartwood C with DBH was mainly found in ring-porous species (Fig. S4e), and with height, it was mainly with non-porous species (Fig. S5e). Again, except for sample size, our results stress that wood porosity plays a role on the tree size effect on C concentrations.

The negative correlation of carbon with absolute growth rate is consistent with our previous study (Zhang et al 2009). In addition to the absolute growth rate, there were comparable correlations of C concentration with the relative growth rate. This was overall consistent with the positive correlations between carbon values and wood density, because trees with high growth rates have low wood density (Chave et al 2009; Larjavaara and Muller-Landau 2010). This negative correlation also helps to explain the negative relationship between C concentrations and temperature globally (Paroshy et al 2021), because trees in warmer climates and without water limitations have potentially faster growth rates.

Conclusion

We explored the variations and determinants of tree-tissue carbon concentrations using a comprehensive dataset for a temperate forest. Metabolically active tissues (i.e., foliage and fine roots) had higher carbon levels, but wood porosity type affected these inter-tissue patterns. For almost all tissues of non-porous species, carbon was generally higher than for ring-porous species. Tissue carbon was often positively correlated with tissue density and tree size (DBH, height) while it was negatively correlated with growth rate. However, it was also divergent among the three wood porosity types. The effect of wood porosity on the variations in temperate tree tissue C levels should be considered in carbon accounting research.