Factors controlling bark decomposition and its role in wood decomposition in five tropical tree species

Organic matter decomposition represents a vital ecosystem process by which nutrients are made available for plant uptake and is a major flux in the global carbon cycle. Previous studies have investigated decomposition of different plant parts, but few considered bark decomposition or its role in decomposition of wood. However, bark can comprise a large fraction of tree biomass. We used a common litter-bed approach to investigate factors affecting bark decomposition and its role in wood decomposition for five tree species in a secondary seasonal tropical rain forest in SW China. For bark, we implemented a litter bag experiment over 12 mo, using different mesh sizes to investigate effects of litter meso- and macro-fauna. For wood, we compared the decomposition of branches with and without bark over 24 mo. Bark in coarse mesh bags decomposed 1.11–1.76 times faster than bark in fine mesh bags. For wood decomposition, responses to bark removal were species dependent. Three species with slow wood decomposition rates showed significant negative effects of bark-removal, but there was no significant effect in the other two species. Future research should also separately examine bark and wood decomposition, and consider bark-removal experiments to better understand roles of bark in wood decomposition.

authors have suggested that the presence or absence of bark may inhibit or enhance the decomposition of dead wood, but few studies have tested these ideas. For example, the bark of some species contains compounds that are known to inhibit microorganisms (e.g., Tectona grandis bark 18 ). Physiological and mechanical functions of bark have been investigated in living trees and several distinct functions, including defense against pests and herbivores, reduction of desiccation, protection against fire, control in trunk girth seasonal changes, mechanical support, storage, and photosynthesis, have been attributed to bark 6,[19][20][21][22][23] . However, to what extent these functions, especially the defensive ones, are still important after tree death remains poorly understood 2 . For example, a recent article highlights the potential role of bark traits in influencing faunal assemblages during the early stages of wood decomposition 3 . Arthropods or invertebrate fauna in general are recognized to influence organic matter decomposition not only directly through the amount of matter consumed (mainly by termites in tropical and subtropical environments) but also through indirect effects such as grazing on fungi sporocarps 5 .
Bark is the outer layer beyond the living cambium, which often has corky structure and is composed of dead cells 24 . It can represent up to 20% of the dry weight of the tree stem 13 . Bark and wood differ markedly in chemical composition and structure. Bark is mostly suberin. Among the functions attributed to suberin are desiccation reduction and inhibition of attack from microorganisms 25 . Bark also contains higher nutrient concentrations and C than wood 19,26,27 . Since chemical composition of litter is the most important factor after climate in determining decomposition rates 28 , one may anticipate that bark and wood will differ substantially in their decomposition rates. Indeed, some studies have already shown that nutrient dynamics during decomposition differ between bark and wood 11 . A recent study also showed that lignin and polyesters such as cutin and suberin responded differently to temperature in a warming experiment 29 . In addition, for logs there is a substrate size -decomposition rate relationship 30 , whereby larger logs decompose more slowly. This relationship may in part be linked to bark effects on decomposition rates, because smaller logs have a larger proportional surface area and hence a higher proportion of bark. Thus, coupling bark decomposition studies with additional experiments investigating the decomposition of logs with and without bark is likely to enhance our understanding of the mechanisms affecting wood decomposition.
We therefore investigated factors affecting bark decomposition and the effects of bark on wood decomposition for five important tree species in a secondary seasonal tropical rain forest in Xishuangbanna, SW China. We employed a common litter-bed approach to minimize environmental variation and thus better understand the effects of species traits on bark decomposition. We examined the following hypotheses: (i) Bark contains higher C and N than wood among the study species, (ii) bark decomposition rates decline across species as the proportions of lignin, cellulose and tannins in bark tissue increase; (iii) meso-and macro-fauna promote bark decomposition, both through enhancing microbial decomposition and fragmentation of bark; and (iv) logs retaining bark decompose faster than logs without bark, and this effect decreases with bark decomposability.

Results
Initial characteristics of bark litter and branches. When bark and wood chemistries are compared, in four of the five species (the exception being Kleinhovia hospita) wood tissue contained more carbon and cellulose but less N and P than bark tissue (Table 1). Kleinhovia hospita had substantially the highest N, P and cellulose contents in both bark and wood, and the wood of Kleinhovia hospita contained less cellulose and water-soluble sugar than bark. The patterns for lignin, hemicellulose and K were mixed ( Table 1). The tannin content of bark tissue increased in the following order: Kleinhovia hospita << Tectona grandis < Cunninghamia lanceolata < Dipterocarpus turbinatus < Toona ciliata. Meanwhile, wood specific gravity increased in the following order: Kleinhovia hospita < Tectona grandis < Toona ciliata < Cunninghamia lanceolata < Dipterocarpus turbinatus (Table 1).

Nutrient dynamics in bark litter during decomposition.
Regardless of the species and litter-bag treatment, sugar, cellulose and carbon concentrations in bark litter decreased between the initial (0 month) and final  samples (after 12 months of incubation), whereas nitrogen concentrations increased (Fig. 1, Fig. S1). For other nutrients, no general trends were found. The first two axes of a principal component analysis (PCA) explained 96% of the variation in bark chemistry among species and samples taken at both the start and end of the experiment (PCA1, 82.01% and PCA2, 14.33%). C, N, lignin and tannin were important in explaining the variation bark litter chemistry (Fig. 1). The PCA reveals a strong species effect in the change in bark chemistry through decomposition, but the mesh size effect was species dependent. There were large differences in nutrient concentrations between bark litter remaining at the end of the experiment in coarse mesh and fine mesh bags for three species, but not for Tectona grandis and Cunninghamia lanceolata.
Bark litter decomposition. The optimal model for bark litter decomposition retained number of days (polynomial 2 nd degree), species, litter-bag type and all two-way interactions as independent variables. To correct for heteroscedasticity, the exponential function (varExp (form = ~fitted(.)) was used. After 1 yr incubation, bark litter mass loss ranged from 96.45 ± 0.54% (mean ± standard error, N = 5) for Dipterocarpus turbinatus to 58.49 ± 4.15% for Toona ciliata in coarse mesh bags and from 81.96 ± 4.39% for Kleinovia hospita to 43.78 ± 2.63 for Cunninghamia lanceolata in fine mesh bags (Fig. 2). Across all species, bark litter in coarse-mesh bags had higher mass loss than in fine-mesh litter-bags ( Table 2). The variance in decomposition rates was much higher for coarse mesh bags than for fine mesh bags. There was a strongly significant species effect ( Table 2 and Table S1), a significant mesh size:time interaction (Table 2), and a significant species:time interaction ( Table 2). All species had rapid decomposition over the first six months, which was during the wet season. However, Kleinhovia hospita decomposed more rapidly over the first three months, compared to other species. Post-hoc pairwise tests for bark decomposition among species revealed three significantly distinct groupings irrespective of litter-bag treatments; decomposition rates increased in the following order: (i) Cunninghamia lanceoleta and Toona ciliata (ii) Tectona grandis, and (iii) Kleinhovia hospita. Dipterocarpus turbinatus could not be distinguished from groups (i) and (ii) ( Table S2).
Log decomposition. The optimal model of branch mass loss over two years retained number of days (polynomial 2 nd degree), species, bark-removal treatment and all two-way interactions as independent variables. To correct for heteroscedasticity, we employed the variance identity (varIdent) function to permit variances among both species and bark treatment to vary. We found that main effect of bark removal was not significant ( Table 3, Fig. 3, Table S3). However, there was a significant bark-treatment:species interaction. Logs with bark decomposed faster than those without for Toona ciliata (1.33 fold increase), Cunninghamia lanceolata (1.46 fold increase) and Tectona grandis (1.92 fold increase), but not for Kleinhovia hospita and Dipterocarpus turbinatus (Fig. 3). As expected, there was also a strong species effect on the rate of branch decomposition and also a significant species:time interaction ( Table 3). The species with the fastest decomposition rate was Kleinhovia hospita. Regardless of the treatment, Kleinhovia hospita completely decomposed within 12 mo. The slowest to decompose was Tectona grandis (53.1 ± 5.5% for branches with bark and 27.6 ± 25.70% for branches without bark over 24 mo). Post-hoc pairwise tests for decomposition of branches with bark among species revealed three distinct groupings. Decomposition rates increased in the following order: (i) Tectona grandis and Cunninghamia lanceolata, (ii) Toona ciliata and Dipterocarpus turbinatus, and (iii) Kleinhovia hospita. For branches without bark four distinct groups were found. Decomposition rates increased in the following order (i) Tectona grandis, and Cunninghamia lanceolata, (ii) Toona ciliata, (iii) Dipterocarpus turbinatus, and (iv) Kleinhovia hospita (Table S4).

Discussion
This experiment examined the decomposition of both bark litter and branches of five important tropical tree species, using a common litter-bed approach. Previous studies have revealed that climate 28,31 , litter substrate quality 28 and decomposer community composition 32,33 are principal drivers of variance in decomposition rates. Through using a common litter bed approach we controlled for environmental effects, including climate and decomposer community composition. For bark litter, we found that litter in mesh bags that permitted meso-and macro-faunal access (i.e. coarse mesh) decomposed faster, regardless of wood species. There was also substantial variation in bark decomposition rates among species. Kleinhovia hospita with substantially lower bark tannin content decomposed fastest, especially in fine-mesh bags. However, contrary to our hypothesis differences in decomposition rates among the other four species did not follow variation in tannin content. For bark litter in coarse-mesh bags, there was very high within-species variance in decomposition rates particularly after 6 mo, which appeared to reflect termite activity. For wood decomposition, as predicted branches with bark decomposed faster than without bark in three species and in the remaining two species there were no significant effects of bark removal. The species with significant bark removal effects had the slowest decomposing wood.  Our study is the first to have examined differences in chemistry between bark and wood in tropical tree species. Among temperate species, Harmon et al. 34 and Martin et al. 27 found C content to be higher in bark than wood. In our study, we found the opposite. Likewise, for N content, Martin et al. 27 reported higher concentrations in bark than in wood, which was opposite to the pattern in our study. These differences between our results and those reported previously from temperate species may simply reflect the small number of species examined. However, if the results do reflect differences in tissue-construction strategy between tropical and temperate species, it would suggest a wider sampling of tree species would be rewarding.
The relative increase in the concentration of N through decomposition is perhaps surprising, but can possibly be explained by the high N content of tannins and other secondary metabolites that are resistant to decomposition. The two species for which mesh size did not have any effect on the changes in relative nutrient concentrations were the two species with the most recalcitrant bark litter.
Bark litter in coarse-mesh decomposed faster (1.11-1.76 fold increase) than in fine-mesh bags across all species, indicating an important effect of litter meso and macro invertebrates on decomposition of bark in this environment. In a recent review, Ulyshen 35 found that arthropods are capable of consuming up to 20% of decomposing wood litter. Colonisation by invertebrates is also likely to facilitate microbial decomposition through the increase in substrate surface area resulting from fragmentation and microbial inoculation. High within-species variation in decomposition rates of bark in coarse-mesh bags could be explained by differences in arthropod colonization patterns, most likely termites. Other bark factors, such as water-holding capacity 5 Table 3. Model results for decomposition (mass loss) of branches. Percentage mass loss (logit transformed) was modelled over 24 months as function of number of days (2 nd order polynomial), bark treatment, species and all two-way interactions (for the full model and summary see Table S3 in supplementary material). There were five species: Kleinhovia hospita, Tectona grandis, Cunninghamia lanceolata, Dipterocarpus turbinatus, Toona ciliata, and bark treatment had two levels: branches with bark and branches without bark. DF denotes degree of freedom. Denominator DF = 98.

Figure 3. Percent mass loss of branches after 6 mo, 12 mo, 18 mo and 24 mo incubation in a common litter bed in a secondary rain forest at Xishuangbanna Tropical Botanical Garden for five tree species (Kleinhovia hospita, Tectona grandis, Cunninghamia lanceolata, Dipterocarpus turbinatus, and Toona ciliata) with respect to bark removal treatment (bark intact vs bark removed).
Three branches were collected every six months. Data represent means ± SE. and surface structure of bark tissue, may together affect the attractiveness of bark litter to litter fauna and the rate of microbial decomposition. However, in our experiment there was no significant interactive effect between tree species and mesh size, indicating that specific characteristics of the litter did not alter effects of meso-and macro-faunal access. Nevertheless, there was a strong direct effect of species on bark litter decomposition rates. Recent findings from Zuo et al. 3 highlight that for the early stages of decomposition, bark traits potentially determine the faunal assemblage on logs, which perhaps could explain part of the variation in wood decomposition earlier explained by species identity. Litter quality has long been identified as key controlling factor for the decomposition of organic matter 37 . Decomposition by fungi is generally limited by N availability 37 , which may explain why the bark of Kleinhovia hospita, with the highest initial N content, decomposed most rapidly in fine-mesh bags. Kleinhovia hospita bark also had the highest cellulose to lignin ratio and microorganisms tend to perform well in presence of high cellulose to lignin ratios 30 . Conversely, the low cellulose to lignin ratio in the bark of Cunninghamia lanceolata may explain why it had the slowest decomposition rate in fine-mesh bags. Moreover, C. lanceolata is a gymnosperm and as Weedon et al. 37 and Cornwell et al. 28 point out, gymnosperms usually contain lignin made from guaiacyl, which is a highly resistant compound. As expected, in three species branches with bark decomposed faster than branches without bark. In the other two species bark removal did not have any significant effect. Recently Johnson et al. 11 suggested that variation among species in rates of wood decomposition may be a consequence of differences in bark mineralization. Bark may have a moisture-retention function in woody debris due to the presence of suberin in bark, thereby enhancing microbial decomposition through providing a favorable micro-environment. Ulyshen et al. 5 , who studied bark removal in a single species (Liquidambar styraciflua L.) also found the presence of bark enhanced wood decomposition rates. Bark may also alter overall substrate quality. It is noteworthy that in our experiment the effect of bark removal was significant in the three species with the slowest decomposing branches: Tectona grandis, Cunninghamia lanceolata and Toona ciliata. Perhaps, for these species, chemical constituents released from the decomposition of bark tissue facilitate the decomposition of wood, which is otherwise highly recalcitrant.
As expected, we found a significant effect of species on the rate of decomposition of branches. Kleinhovia hospita, whose branches were the fastest to decompose regardless of bark treatment, had the lowest wood specific gravity, highest concentration of N, P, and K, and lowest carbon to nitrogen ratio, while its bark had the lowest tannin content and highest cellulose to lignin ratio. Cunninghamia lanceolata branches with and without bark were the second slowest to decompose, which may be explained by the slow colonization and establishment of fungi on gymnosperms as compared to angiosperms 28,38 . We found a significant species:time interaction in the decomposition of branches, with the slower decomposing species evidencing slower initial decomposition (Fig. 3). These results corroborate those recently reported by Schilling et al. 39 who tracked the decomposition of eight woody species in a tropical dry forest. Moreover, previous studies recognized Cunninghamia lanceolata as a highly decay-resistant species and especially resistant to fungal attack 40 , perhaps because of the high lignin content of its bark and wood and also because its heartwood contains antifungal properties 41 . Branches of Tectona grandis were the slowest to decompose likely because as reported by several studies this species both wood and bark contain some antifungal properties [42][43][44] . Our Tectona grandis branches with bark decomposed faster than those without bark. Previous research also found that the decay resistance of Tectona grandis wood increased with the depth (i.e. sapwood decompose faster than heartwood which was correlated with the concentration of antifungal extracts 42 ). Moreover, Tectona grandis wood itself is also well recognised for its resistance to termite attack 45 .
In this study we did not investigate the effect of bark removal with respect to log diameter, which is likely to be important 21 . In addition to surface area:volume effects, the thickness, physical structure and chemical composition of bark may vary among woody tissues of different size, although Rosell et al. 24 recently showed that bark from branches and twigs was representative of bark on the trunk among 85 species of angiosperms. Our chemical analyses also did not include important (but difficult to measure) bark components, such as lignin monomers or suberin. Improved understanding of the distribution and role of these substances in decomposition may be important to understanding bark decomposition and its role in the decomposition of wood. In addition to repeating our experiments across an expanded sample of species, a comprehensive survey of functional relationships among bark and wood characteristics, and their evolutionary and ecological patterns would be very informative.
In summary, organic matter decomposition, carbon and nutrient cycling are important topics for ecological study. However, recently, the need to further parameterize and provide better mechanistic understanding of decomposition processes, has resulted in calls for more tissue specific approaches. Until recently, researchers have usually regarded wood and bark as a single substrate. Indeed, assignment of wood decay classes has relied mostly on the status of bark (presence/absence, state of decomposition). Here, we investigated bark and wood decomposition in five tropical tree species. We found that species traits and litter fauna access had important effects on bark litter decomposition. Through a bark removal experiment, we also found that bark enhanced the rate of wood decomposition in some species, but had no effect in others. Species with a significant bark removal effect were those with the slowest rates of wood decomposition.

Materials and Methods
This study was conducted in a secondary rainforest (UTM 47Q, 0734280 E, 2425546 N, 589 m als.) at Xishuangbanna Tropical Botanical Garden (XTBG), a seasonal moist tropical forest. Total annual rainfall is 1463 mm but rainfall is highly seasonal with 87% of total precipitation falling between May and Oct 46 . However, fog during the cool dry season (Nov-Mar) reduces plant water deficits, permitting tropical rain forest to persist 46 . Monthly mean temperature varies from 15.9 °C in Jan to 25.7 °C in June 47 .
We conducted an experiment using litter bags to assess bark decomposition over 1 yr from May 9 th 2014 to May 9 th 2015. In addition, we monitored the decomposition of 5-cm diameter branches with and without bark over 2 yrs from Nov. 23 rd 2013 to Nov. 4 th 2015.
Species selection. We selected tree species which met at least two of the following criteria: (i) important timber producing tree species for the Asian timber-trade market, (ii) species that are ecologically important in Asian tropical forests and (iii) tree species used extensively for reforestation or in silvicultural plantations in Asia. Species were also selected to represent a range of wood densities and supposed decay resistance. Finally, the species also all had to be present at XTBG, so that their wood was available. In total, five species; Kleinhovia hospita L. Bark collection and preparation. Bark was removed from ~5 cm diameter branches of these tree species. Branches were from one tree for Cunninghamia lanceolata, Toona ciliata, and Tectona grandis, two trees for Kleinhovia hospita and from three trees for Dipterocarpus turbinatus. The number of trees used was determined by the amount of material produced during the regular pruning operations of the horticultural department at XTBG. The removed bark material was cut into pieces of ca. 1 cm × 5 cm. This was to ensure that bark from different species was presented in more or less same physical shape and also so that it could be fitted into the litter-bags. The collected bark litter was then dried at 80 °C for 24 hrs, frozen at − 80 °C for 24 hrs and dried again at 80 °C for 24 hrs to defaunate it 48,49 . Litter-bags. To study effects of fauna on the decomposition of bark, we used two types of litter-bags. Coarse mesh (mesh size 4.75 mm for upper side, and mesh size 0.5 mm lower side to reduce loss of small particles from fragmentation) bags allow litter fauna access to the bark litter. Our fine mesh (mesh size 0.068 mm) bags excluded all arthropods and mesofauna 32 . Litter-bags were incubated within a 3.6 m × 2.4 m common litter-bed. At 3-mo intervals we harvested litter-bags. Five replicates were harvested at 3, 6 and 12 mo after incubation, and three replicates were harvested at 9 mo.
Each litter-bag was initially filled with 20 g of dry bark litter. Upon collection, each litter-bag was checked visually for termites and then the remaining material in each bag was carefully and gently cleaned with tap water before oven drying at 105 °C to constant mass. Any fine mesh bags with evidence of termite holes were rejected, as in addition to termites access by other meso-and macrofauna may have occurred. Coarse-mesh bags attacked by termites were retained for analysis. The remaining bark litter dry mass was measured to 0.01 g using an electronic balance.
Branches collection, preparation, incubation and harvest. Fresh branches were cut from the target species while the garden's horticultural staff were conducting routine pruning operations. Branches were transported to the laboratory, where an electric saw was used to cut branches into pieces of ~0.5 m length. The saw was sterilized with ethanol and flamed between cuts to minimise transfer of microbes, in particular fungi, between samples. Branches with irregular shapes were not used. Each branch was tagged with a uniquely numbered metal tag and both ends were coated with black enamel paint to reduce arthropod colonization into cuts 50 . We recorded the (i) diameter (at both ends and in the middle) as measured using a diameter tape, (ii) bark thickness measured at both ends with a digital caliper, (iii) fresh weight of the entire log, and (iv) the total length. For branches with bark removed, these metrics were recorded before and after bark removal. Disks taken at the base of each branch prior to preparation were stored at − 20 °C until chemical analyses.
Three replicates of each species per treatment (with/without bark) were collected at 6, 12, 18 and 24 mo after incubation. If branches became fragmented, we carefully collected each fragment that was visible to naked eye using forceps. After harvesting, we recorded the presence/absence of termites and then mud, sand, insect frass and mosses were carefully brushed off the branch. The fresh weight of each branch was measured to 0.01 g using an electronic balance. Next, three disks of 2 cm thickness were cut from the middle of the branch and approximately 5 cm from each end (to avoid end effects) (Fig. S2). For branches that were fragmented, we used several pieces instead of three disks. Each disk was weighed before and after drying at 105 °C to constant mass. The dry volume of each disk was measured using the water displacement method 51 . To calculate wood specific gravity (WSG) of a sample we used the mean derived from the three disks (or pieces). Using the final fresh mass and water content we calculated the final dry mass, which enabled us to derive the mass loss from the initial dimensional measurements and initial WSG.
Chemical analysis. We assessed initial chemical properties of the wood and bark. Disks samples stored at − 20 °C were dried and ground to a powder for chemical analysis. Bark and wood were analysed separately. In addition, final chemical properties of remaining bark were also determined. The total carbon (C), nitrogen (N), phosphorus (P), potassium (K), water soluble sugar content, neutral detergent fiber (NDF), acid detergent fiber (ADF), and acid detergent lignin (ADL) were measured for both bark litter and wood. In addition, tannin content was measured for bark and fiber content was measured for wood. C and N were determined by Vario Max CN element analyzer (Elementa Analsensysteme, Germany). For P, and K measurements, samples were diluted with HCl and digested with HNO 3 -HClO 4 . These were measured with an inductively coupled plasma atomic emission spectrometer (iCAP 6300, IRIS Advantage, E R, Thermo Fisher Scientific, USA). Water-soluble sugars were analysed with Analyzer 3 (SEAL Analytical GmbH, Germany). Tannin content was measured with UV-visible spectrometer (UV 2450, SHIMADZU, Japan). NDF, ADF, ADL, fiber contents were determined using Fibertec ™ 2010 (FOSS Analytical AB, Sweden). ADL, NDF and ADF were later converted to lignin, hemicellulose and cellulose following equations in Chen et al. 52 .
Scientific RepoRts | 6:34153 | DOI: 10.1038/srep34153 Data analysis. We assessed the decomposition of both bark litter and branches through the loss of dry mass over time. We modeled the percentage of mass loss of bark litter (logit transformed following 53 over 12 months as function of number of days incubation, species, treatment (litter-bag type) and all two-way interactions. We modeled the percentage mass loss from wood branches over 24 months as function of number of days incubation, species and bark treatment (with/without bark) and all two-way interactions.
We used Akaike information criteria (AIC) to select optimal models, the function ANOVA to examine effect of each independent variable. All the analyses were performed in R version 3.3.1 (R Foundation for Statistical Computing 54 ) and the nlme package 55 , using the modeling function gls (generalized least squared). Where differences among species were significant, we conducted post-hoc multiple comparison pairwise tests using the functions cld and lsmeans from the lsmeans package 56 . Model fit was investigated by plotting residuals against standardised effect sizes.