Wood plays an important role in the performance and life-history strategies of trees, conferring mechanical stability, defence against pathogen and herbivore attacks, tree architecture and hydraulics (Falster Reference FALSTER2006, Hoeber et al. Reference HOEBER, LEUSCHNER, KÖHLER, ARIAS-AGUILAR and SCHULDT2014, Iida et al. Reference IIDA, POORTER, STERCK, KASSIM, KUBO, POTTS and KOHYAMA2012, Poorter et al. Reference POORTER, MCDONALD, ALARCÓN, FICHTLER, LICONA, PEÑA-CLAROS, STERCK, VILLEGAS and SASS-KLAASSEN2010). Recent studies have also found that wood density and wood chemical traits such as lignin and nitrogen (N) concentrations and carbon-to-nitrogen ratio (C/N) are good predictors of above-ground biomass and wood decomposition rates (Chave et al. Reference CHAVE, RÉJOU-MÉCHAIN, BÚRQUEZ, CHIDUMAYO, COLGAN, DELITTI, DUQUE, EID, FEARNSIDE, GOODMAN, HENRY, MARTÍNEZ-YRÍZAR, MUGASHA, MULLER-LANDAU, MENCUCCINI, NELSON, NGOMANDA, NOGUEIRA, ORTIZ-MALAVASSI, PÉLISSIER, PLOTON, RYAN, SALDARRIAGA and VIEILLEDENT2015, Freschet et al. Reference FRESCHET, WEEDON, AERTS, VAN HAL and CORNELISSEN2012, van Geffen et al. Reference VAN GEFFEN, POORTER, SASS-KLAASSEN, VAN LOGTESTIJN and CORNELISSEN2010, Weedon et al. Reference WEEDON, CORNWELL, CORNELISSEN, ZANNE, WIRTH and COOMES2009). Given the fact that a large amount of biomass is stocked as wood in forests, quantifying wood-trait variation among species will contribute to a better understanding of C stock and nutrient cycling in forest ecosystems.

Wood density and wood chemical traits have often been measured in stems and branches in the tropics (Poorter et al. Reference POORTER, MCDONALD, ALARCÓN, FICHTLER, LICONA, PEÑA-CLAROS, STERCK, VILLEGAS and SASS-KLAASSEN2010, van Geffen et al. Reference VAN GEFFEN, POORTER, SASS-KLAASSEN, VAN LOGTESTIJN and CORNELISSEN2010). Because roots comprise up to c. 20% of the total biomass in tropical forests (Niiyama et al. Reference NIIYAMA, KAJIMOTO, MATSUURA, YAMASHITA, MATSUO, YASHIRO, RIPIN, KASSIM and NOOR2010, Poorter et al. Reference POORTER, NIKLAS, REICH, OLEKSYN, POOT and MOMMER2012), differences in wood density and wood chemical traits between above- and below-ground parts could affect C pool and wood decomposition rate, and consequently, C fluxes and nutrient cycling. However, few studies have focused on the relationship between stem/branch and root wood (but see Fortunel et al. Reference FORTUNEL, FINE and BARALOTO2012, Reference FORTUNEL, RUELL, BEAUCHȆNE, FINE and BARALOTO2014; Freschet et al. Reference FRESCHET, CORNELISSEN, VAN LOGTESTIJN and AERTS2010, Schuldt et al. Reference SCHULDT, LEUSCHNER, BROCK and HORNA2013). Wood density also depends on adult stature (Poorter et al. Reference POORTER, MCDONALD, ALARCÓN, FICHTLER, LICONA, PEÑA-CLAROS, STERCK, VILLEGAS and SASS-KLAASSEN2010), suggesting that the relationships of wood density and wood chemical traits between stems and roots might vary with adult stature. However, sampling coarse roots is challenging. If researchers could estimate root wood traits by studying stems, sampling effort and damage to trees could be reduced, which would be a major technical advance in plant ecology. Here, we examined the relationships of wood density and wood chemical traits between the stems and roots of 53 coexisting Bornean rainforest tree species. We focused on stems and coarse woody roots because of their large biomass (Niiyama et al. Reference NIIYAMA, KAJIMOTO, MATSUURA, YAMASHITA, MATSUO, YASHIRO, RIPIN, KASSIM and NOOR2010). Specifically, we investigated the following hypotheses: (1) Root wood traits can be estimated from stem wood traits, (2) The stem-root relationships differ with adult stature.

This study was carried out in a mixed dipterocarp lowland forest at Lambir Hills National Park, Sarawak, Malaysia (4°12′N, 114°02′E). We selected 90 individuals from 53 species representing 12 families and 21 genera, including abundant and common taxa at the study site. Tree species were classified into three life forms according to the adult stature: tall trees, > 30 m in height; medium–tall trees, 15–30 m; and small trees, 5–15 m (Jawa & Chai Reference JAWA and CHAI2007; Table 1). In September 2013, we took 5–15 cm wood cores from the stems and roots of each individual using an increment borer (5.15 mm radius). These were sampled from trunks at heights of 0.5–1.0 m, and from coarse roots within a radius of 1.5 m from trunks after excavation of the base.

Table 1. Wood density and wood chemical traits of stems and coarse roots for each life form and all species measured on 53 Bornean tropical tree species (mean ± SE). Different letters indicate significant differences between stems and coarse roots for all species (P < 0.05).

We measured four wood traits of stems and coarse roots: wood density, N concentration, lignin concentration and C/N. Wood density (g cm⁻³) was calculated as dry mass to fresh volume. Carbon and N concentrations (g kg⁻¹) were measured using a CN analyser (Macro corder, JM1000CN, Yanaco, Japan). Lignin concentration (%) was determined as acid-insoluble lignin by the common Klason lignin procedure. We measured all of the wood traits in each tree, and averaged them for each species.

To examine relationships of wood density and wood chemical traits between stems and coarse roots considering taxonomic lineage, we performed correlations including phylogenetically independent contrasts (PICs). We constructed a phylogenetic tree for the 53 species using PhyloMatic v.3 utility (Webb & Donoghue Reference WEBB and DONOGHUE2005) based on the phylogenetic hypothesis of Davies et al. (Reference DAVIES, BARRACLOUGH, CHASE, SOLTIS, SOLTIS and SAVOLAINEN2004), with polytomies applied within most families and genera. For this analysis, branch lengths were scaled to one. Each wood trait was also compared between stems and coarse roots using Welchʼs t-tests. We developed regression equations for each wood trait by the standardized major axis (SMA) method. To explore differences in the stem-root relationships among life forms, we used tests for common slope and elevation (intercept) among SMAs. When no significant difference was detected in the SMA, a common equation was regressed using pooled data for all species irrespective of life forms. All of the analyses were conducted using R 3.0.2 (packages ape and smatr).

Correlations with PICs showed that stems and coarse roots were strongly correlated for each wood trait (Figure 1). Most regression equations fitted well, and on average, explained 56% of the variation in wood traits (range = 0.18–0.81). The lignin concentration in coarse roots was significantly greater than that in stems (P < 0.05; Table 1). SMA regression analyses detected significantly different slopes in the regression between N concentration of stems and roots among life forms; small trees had a lower slope than tall and medium–tall trees (Figure 1).

Figure 1. Relationships between stems and coarse roots for 53 Bornean tropical tree species, with Pearson correlation coefficients for PICs (R; P < 0.05 using Bonferroni correction) and equations in SMA regression analyses. Wood density (a); lignin concentration (b); N concentration (c); and C/N (d). Circles, triangles, and crosses indicate species of tall, medium-tall, and small trees, respectively. Thin dashed lines are 1:1 lines. Regression lines by SMA are also shown in thick lines for all species data (a, b, d). For (c), solid, dashed, and thick lines are SMA regression lines for tall, medium-tall, and small trees, respectively.

We found that wood density and wood chemical traits of roots mirrored those of stems; thus they can be estimated from stem wood traits in Bornean tropical trees. Similar relationships between stem and root wood traits have been reported in Neotropical trees (Fortunel et al. Reference FORTUNEL, FINE and BARALOTO2012) and subarctic plants (Freschet et al. Reference FRESCHET, CORNELISSEN, VAN LOGTESTIJN and AERTS2010). Our results will aid a greater understanding of underground C pool, C fluxes and nutrient cycling. The difference between stems and coarse roots was found only in lignin concentration. Because lignin is known to inhibit microbe and insect attacks (Bhuiyan et al. Reference BHUIYAN, SELVARAJ, WEI and KING2009, Wainhouse et al. Reference WAINHOUSE, CROSS and HOWELL1990) and roots can be infested by various herbivores (Blossey & Hunt-Joshi Reference BLOSSEY and HUNT-JOSHI2003), a higher lignin concentration in roots might be effective for defence against root feeders. Higher concentration of lignin in coarse roots than in stems also indicates that wood decomposition rates of coarse roots are slower than those of stems in Bornean tropical rain forests. This hypothesis has been rarely examined, but is supported by a study of two species in a boreal forest (Freschet et al. Reference FRESCHET, WEEDON, AERTS, VAN HAL and CORNELISSEN2012). Since substantial stocks of woody debris are provided underground after trees die, the differences in wood chemical traits and wood decomposability between stems and coarse roots might affect C and nutrient cycling in a forest. The different stem–root relationship in N concentration among life forms might reflect interspecific variation in N allocation between above- and belowground parts (Millard & Grelet Reference MILLARD and GRELET2010). To ensure generality and elucidate mechanisms, N allocation between stems and coarse roots in adult trees in situ should be examined. There is an increasing need for accurately estimating C stocks and fluxes in a forest ecosystem due to global climate change. We thus must quantify wood density and wood chemical traits of both stems and coarse roots in various biomes.