Tree diversity effects on litter decomposition are mediated by litterfall and microbial processes
Data files
Abstract
Forest ecosystems are critical for their carbon sequestration potential. Increasing tree diversity was shown to enhance both forest productivity and litter decomposition. Litter diversity increases litter decomposability by increasing the diversity of substrates offered to decomposers. However, the relative importance of litter decomposability and decomposer community in mediating tree diversity effects on decomposition remains unknown. Moreover, tree diversity modulation of litterfall spatial distribution, consequently, litter decomposition has rarely been tested. We studied tree diversity effects on leaf litter decomposition and its mediation by the amount of litterfall, litter species richness and decomposability, and soil microorganisms in a large-scale tree diversity experiment in subtropical China. Furthermore, we examined how litter functional identity and diversity affect leaf litter decomposability. Finally, we tested how leaf functional traits, tree biomass, and forest spatial structure drive the litterfall spatial distribution. We found evidence that tree species richness increased litter decomposition by increasing litter species richness and the amount of litterfall. We showed that soil microorganisms in this subtropical forest perform 84–87% of litter decomposition. Moreover, changes in the amount of litterfall and microbial decomposition explained 19–37% of the decomposition variance. Additionally, up to 20% of the microbial decomposition variance was explained by litter decomposability, while litter decomposability itself was determined by litter functional identity, diversity, and species richness. Tree species richness increased litter species richness and the amount of litterfall (+200% from monoculture to 8-species neighborhood). We further demonstrated that the amount of species-specific litterfall increased with increasing tree proximity and biomass and was modulated by leaf functional traits. These litterfall drivers increased the spatial heterogeneity of litter distribution, thus, litter decomposition. We highlighted multiple biomass- and diversity-mediated effects of tree diversity on ecosystem properties driving forest nutrient cycling. We conclude that considering spatial variability in biotic properties will improve our mechanistic understanding of ecosystem functioning.
README: Tree diversity effects on litter decomposition are mediated by litterfall and microbial processes - DATA
Three main dataset used for the analyses in the article "Tree diversity effects on litter decomposition are mediated by litterfall and microbial processes" published in Oikos
Related article DOI: 10.1111/oik.09751
Analyses scripts release: 10.5281/zenodo.8039140
Description of the data and file structure
Three datasets are present in this release related to the three main analyses of the manuscript. The columns description can be found in the analyses scripts. In short:
df.csv: experimental data measured to test the effect of tree species richness on litter production and decomposition
$ TSP : (chr) Sampling point ID
$ C.loss_Mi1 : (num) Carbon total decomposition (% of Closs) measured at the sampling point
$ C.loss_Ma1 : (num) Nitrogen total decomposition (% of Nloss) measured at the sampling point
$ N.loss_Mi1 : (num) Carbon microbial decomposition (% of Closs) measured at the sampling point
$ N.loss_Ma1 : (num) Nitrogen microbial decomposition (% of Nloss) measured at the sampling point
$ C.loss_CG : (num) Carbon decomposability (% of Closs) measured in the common garden experiment
$ N.loss_CG : (num) Nitrogen decomposability (% of Nloss) measured in the common garden experiment
$ lit.rich : (num) number of species in the littertrap
$ neigh.sp.rich: (num) number of species in the 12 trees surrounding the sampling point
$ fall : (num) total litterfall (g)
$ plot : (chr) plot ID
$ comp : (chr) plot species compositiondf-2.csv: experimental data measured to identify species-specific litterfall drivers
$ TSP : (chr) Sampling point ID
$ plot : (chr) Plot ID
$ spe.rich : (num) Plot species richness
$ species : (chr) Species name
$ litter.biomass : (num) Species specific litter biomass (g)
$ biomass : (num) Species specific tree biomass (sum of neighboring trees biomass)
$ dist : (num) Sum of inverse distances from the trees
$ n.rich : (num) Species richness of the neighboring trees
$ area : (num) Litter trap area (m2)
$ litter.biomass.area : (num) Species specific litter biomass corrected for area (g/m2)
$ SLA : (num) Species SLA
$ LDMC : (num) Species leaf dry matter content
$ C : (num) Species leaf carbon content
$ N : (num) Species leaf nitrogen content
$ Mg : (num) Species leaf magnesium content
$ Ca : (num) Species leaf calcium content
$ K : (num) Species leaf potassium content
$ P : (num) Species leaf phosphorus content
$ pca.1 : (num) Species leaf trait PCA axis 1
$ pca.2 : (num) Species leaf trait PCA axis 2
$ log.litter.biomass.area: (num) Log of litter per area biomass
$ log.biomass : (num) Log of tree biomassdf-3.csv: experimental data measured to identify decomposability drivers
$ ID : (num) Sample ID
$ CG_ID : (chr) Common garden sample ID
$ TSP : (chr) Sampling point ID
$ weight.bf : (num) Initial leaf mass content
$ weight_af : (num) Final mass content
$ ini.SLA : (num) CWM of SLA from initial leaf
$ ini.LDMC : (num) CWM of SLA from initial leaf
$ ini.C : (num) Initial carbon content (mg)
$ ini.N : (num) Initial nitrogen content (mg)
$ ini.Mg : (num) Initial magnesium content (mg)
$ ini.Ca : (num) Initial calcium content (mg)
$ ini.K : (num) Initial potassium content (mg)
$ ini.P : (num) Initial phosphorus content (mg)
$ var.SLA : (num) Variance between leaves SLA
$ var.LDMC : (num) Variance between leaves LDMC
$ var.C : (num) Variance between leaves carbon content
$ var.N : (num) Variance between leaves nitrogen content
$ var.Mg : (num) Variance between leaves magnesium content
$ var.Ca : (num) Variance between leaves calcium content
$ var.K : (num) Variance between leaves potassium content
$ var.P : (num) Variance between leaves phosphorus content
$ CN : (num) Initial CN ratio
$ NP : (num) Initial NP ratio
$ compo.pca.1 : (num) first pca axis of initial nutrient contents
$ compo.pca.2 : (num) second pca axis of initial nutrient contents
$ div.pca.1 : (num) first pca axis of initial nutrient contents variance
$ div.pca.2 : (num) first pca axis of initial nutrient contents variance
$ var.x : (num) variance between species all chemical properties
$ var.CNP : (num) variance between species CNP
$ var.t : (num) variance between species all traits
$ var.tough : (num) variance between species toughness traits (SLA LDMC)
$ ash : (num) decomposed sample ash content
$ dry.content : (num) sample dry content
$ sample.C.mg.g : (num) decomposed sample carbon content (mg/g)
$ sample.N.mg.g : (num) decomposed sample nitrogen content (mg/g)
$ lit.rich : (num) litter species richness
$ sample.C.mg : (num) decomposed sample nitrogen content (mg)
$ sample.N.mg : (num) decomposed sample nitrogen content (mg)
$ sample.soil : (num) calculated soil content (g)
$ sample.soil.C.mg: (num) calculated soil C content (mg)
$ sample.soil.N.mg: (num) calculated soil C content (mg)
$ sample.Cc.mg : (num) corrected litter C content (mg)
$ sample.Nc.mg : (num) corrected litter N content (mg)
$ sample.weigh : (num) corrected litter final mass (mg)
$ C.loss : (num) carbon loss during decomposition (%)
$ N.loss : (num) Nitrogen loss during decomposition (%)
$ plot : (chr) Plot ID
Sharing/Access information
NA
Code/Software
R software, see Zenodo release for the related scripts.
Methods
Cf. Joined paper
Study site
The study site is located in south-east China near the town of Xingangshan (Jiangxi province, 29.08-29.11° N, 117.90-117.93° E). Our experimental site is part of the BEF-China experiment (Site A, Bruelheide et al., 2014), which was planted in 2009 after a clear-cut of the previous commercial plantations. The region is characterized by a subtropical climate with warm, rainy summers and cool, dry winters with a mean annual temperature of 16.7 °C and a mean annual rainfall of 1.8 mm (Yang et al., 2013). Soils in the region are Cambisols and Cambisol derivatives, with Regosol on ridges and crests (Geißler et al., 2012; Scholten et al., 2017). The natural vegetation consists of species-rich broad-leaved forests dominated by Cyclobalanopsis glauca, Castanopsis eyrei, Daphniphyllum oldhamii, and Lithocarpus glaber (Bruelheide et al., 2011, 2014).
Study design
To identify the effect of tree spatial organization on litterfall distribution and decomposition, we measured litterfall and decomposition between tree species pairs (hereafter, TSP) across various neighborhoods. Each TSP consisted of two trees next to each other (~1.28 m), and we defined its neighborhood as the ten trees directly adjacent in the planting grid. Each TSP was replicated three times in five tree species richness levels (1, 2, 4, 8, and ≥ 16 species), when available according to the experimental design (see “broken stick design”, Bruelheide et al., 2014). In total, we surveyed 24 combinations of tree species resulting in a total of 180 TSPs in 52 plots (Suppl. S1).
Litterfall sampling
In September 2018, a litter trap of 1 m2 was set up at a height of 1 m above the soil surface between each TSP (Suppl. S1). Litter was collected in December 2018 to cover the main litterfall season in the region (Huang et al., 2017). To measure litterfall composition (i.e., species-specific litter biomass), each leaf of the litter trap was sorted and identified to species level. Each species' litter was dried at 40⁰C for two days and weighed (± 0.1 g). Litter species richness was assessed as the number of species identified in the trap, and the total amount of litterfall was calculated as the sum of the dried biomass of all species in 1 m2.
Litter decomposition experiments
We performed two complementary decomposition experiments: one in the TSPs to measure microbial and total decomposition (H1-2), and one in a Common Garden experimental field site to assess decomposability (H3-4; Suppl. S1).
For both experiments, litterbags (10 cm x 10 cm), with different mesh sizes (see details below) were filled with 2 g (± 0.01 g) of dried litter according to litter trap species composition (i.e., amount of species-specific litter) of the different TSPs. Therefore, the litter composition of the litterbags matched the litterfall composition collected in the corresponding TSP. The litterbags for both experiments were installed in December 2018 and collected in September 2019 before litterfall, i.e., after nine months of decomposition when about 30 to 50% of mass loss is expected in this area (Lin et al. 2021) while avoiding interaction with freshly fallen litter. The litterbags were water-cleaned by dissolution and gentle removal of soil particles as well as dried at 40⁰C for two days. The residual litter was weighed (± 0.01 g) and milled for further chemical content analyses. The water-cleaning treatment of the bag was considered neglectable in comparison to heavy rain and sediment runoff in the region (e.g., up to 250 mm of precipitation on average in May between 2009 and 2012, with up to 130 mm in 24 h in 2010 and significant runoff, see Seitz et al., 2015).
Decomposition experiment in between the TSPs
To assess total litter decomposition (total C and N loss, including fauna-mediated decomposition) and microbial decomposition (microbial C and N loss, excluding fauna-mediated decomposition), two large-mesh (5 mm mesh, total litter decomposition) and two small-mesh (0.054-mm-mesh, microbial decomposition) litterbags were set up between the TSPs, respectively, with plot-specific litter. Small-mesh litter bags excluded meso- and macro-detritivores by using a fine mesh size (0.054-mm-mesh) to assess microbial decomposition, while large-mesh litter bags were built using a 5-mm-mesh in the upper half of the bag to provide access to macro-decomposers, and a 0.054-mm-mesh only at the bottom to prevent loss of fine leaf litter particles to assess total litter decomposition (see Bradford et al., 2002). All litterbags were covered by a 50 cm x 50 cm grid to prevent heavy rainfalls from dislocating the litterbags (1 cm mesh size, see Suppl. S1).
Decomposition experiment in the Common Garden
The Common Garden decomposition experiment was performed in a monoculture from the BEF China site to ensure decomposition experiments were performed under comparable environmental conditions (e.g. seasonal macroclimatic fluctuations). The setting consisted of a monoculture stand of Schima superba, a species that was not included in the TSP experiment; thereby, we were able to exclude any home-field advantages (Fanin et al., 2021). Schima superba was chosen to maximize the phylogenetic distance with our target species and minimize environmental heterogeneity within the plot (i.e., productive species with closed canopy). Schima superba's litter was removed from the ground before deploying the litterbags at a distance of 10 cm from each other in two blocks (one TSP replicate per block, Suppl. S1). To measure litter decomposability, two small-mesh litterbags (0.054-mm-mesh) representing the litter composition of each TSP were incubated in the Common Garden experiment. Only small-mesh litterbags were considered to avoid interaction between litterbags from different TSP (e.g., composition over attracting the macro-decomposers).
Leaf and litter trait measurements
Leaf functional traits were assessed at the species- and plot-level in September 2018, following Davrinche and Haider (2021). For each TSP species, several fresh leaves were collected, and the reflectance spectra were measured using ASD FieldSpec® 4 Wide-Resolution Spectroradiometer (Malvern Panalytical Ltd., Malvern, United Kingdom). Leaf functional traits were predicted from the reflectance spectra of a calibration dataset of the same species, where both reflectance spectra and leaf functional traits were measured. For leaf morphological traits – specific leaf area (SLA, leaf area divided by dry weight) and leaf dry matter content (LDMC, ratio of leaf dry weight to fresh weight) were measured before and after drying for 72 h at 80°C. Leaf areas were measured from scans with a resolution of 300 dpi of the fresh leaves using the WinFOLIA software (Regent Instruments, Quebec, Canada). Leaf chemical contents, i.e., carbon (C), nitrogen (N), phosphorus (P), magnesium (Mg), calcium (Ca), and potassium (K) contents, were measured from dried leaves ground into a fine powder (Mixer Mill 400, Retsch, Haan, Germany). About 5 mg of leaf powder was used to determine C and N contents with an elemental analyzer (Vario EL Cube, Elementar, Langenselbold, Germany); a 200-mg subsample was used to measure P content via nitric acid digestion and spectrophotometry using the acid molybdate technique. The filtrate resulting from nitric acid digestion was analyzed with atomic absorption spectrometry (ContrAA 300 AAS, Analytik Jena, Jena, Germany) for Mg, Ca, and K contents. Of these calibration samples, the relation between the leaf spectra and the measured leaf traits was analyzed with the software Unscrambler X (version 10.1, CAMO Analytics, Oslo, Norway) to predict leaf functional traits of each leaf, then averaged at species- and plot-specific. For each litterbag, we calculated the total amount of nutrients (i.e., C, N, P, Mg, Ca, K) as the sum of all species’ contributions, and leaf morphological traits community weighted mean (i.e., CWM SLA and LDMC) using species-specific litter dry weight and species- and plot-specific leaf functional traits. In addition, we calculated the variance of each litter functional trait (i.e., C, N, P, Mg, Ca, K, SLA, LDMC) within the litterbags.
Litter C and N content after decomposition were measured from the residual litter with an elemental analyzer (Vario EL Cube, Elementar, Langenselbold, Germany). To estimate soil contamination, the ash content of the sample was measured as it represents the amount of soil using the loss on ignition method. Refer to methods in the paper for formulae.
Decomposition measures
C and N loss (%) in the litterbags between December 2018 and September 2019 were used as a measure of the total decomposition (i.e., measured via the large mesh-size in the TSP experiment), microbial decomposition (i.e., using small mesh size in the TSP experiment), and litter decomposability (i.e., using small mesh size in the Common Garden experiment).
Statistical methods
A description of all the variables used in this study can be found in Suppl S1. All data handling and statistical calculations were performed using the R statistical software version 4.1.0 (R Core Team 2021). All R scripts and statistical models used for this project can be found in a GitHub repository. All following linear multiple-predictors models were tested in R using the 'lm' function (R Core Team, 2021), and statistical hypotheses (i.e., residuals normality, homoscedasticity, homogeneity of variance) of the following linear models were tested using the 'model_check' function from the 'performance' package (Lüdecke et al., 2020, Suppl. S3).
Tree diversity effect on carbon and nitrogen loss (H1)
We used linear models and normal distribution assumptions to test the effects of neighborhood tree species richness on total decomposition ("C loss" and "N loss" measured between the TSPs) and microbial decomposition ("C loss" and "N loss" measured between the TSPs, when soil meso- and macro-fauna were excluded). In addition, we used linear models and normal distribution assumptions to test the effects of litter species richness on litter decomposability ("C loss" and "N loss" measured in the Common Garden experiment).
Tree diversity effect on the amount of litterfall and litter species richness
We used linear models and normal distribution assumptions to test the effect of neighborhood tree species richness on the amount of litterfall, and litter species richness.
Mediation of tree species richness effects on litter decomposition
To test the effects of litter species richness on litter decomposability ("C loss" and "N loss" in the Common Garden experiment), we used linear models and normal distribution assumptions. To test the effects of litter species richness, amount of litterfall, and decomposability ("C loss" and "N loss" in the Common Garden experiment) on litter microbial decomposition ("C loss" and "N loss" between the TSPs, when soil meso- and macro-fauna were excluded), we used linear multiple predictor models and normal distribution assumptions, where all predictor values were rescaled using the R function 'scale' (R Core Team 2021). To test the effects of litter species richness, amount of litterfall, and litter microbial decomposition ("C loss" and "N loss" between the TSP, when soil meso- and macro-fauna were excluded) on litter decomposition ("C loss" and "N loss" between the TSP, when soil meso- and macro-fauna were included), we used linear multiple predictor models and normal distribution assumptions, where all predictor values were rescales using the R function 'scale' (R Core Team, 2021 - H2). All previously cited model outputs can be found in Suppl. S3.
To test the mediation of tree species richness effects on litter decomposition by the amount of litterfall and litter species richness effects on decomposability, we implemented the previous relationships in a Structural Equation Model (SEM) framework (see Suppl. S3 for model structure); this comparison being possible as (1) each TSP litter composition was replicated in all experiments, (2) all the three on-site experiments were temporally synchronous, and (3) all variables were centered and reduced to compare effect sizes. Our SEM was fitted using the 'sem' function from the 'lavaan' package (Rosseel, 2012). The quality of our model fit on the data was estimated using three complementary indices: (i) the root-mean-squared error of approximation (RMSEA), (ii) the comparative fit index (CFI), and (iii) the standardized root mean squared residuals (SRMR); a model fit was considered acceptable when RMSEA < 0.10, CFI > 0.9 and SRMR < 0.08.
Litterfall composition effect on litter decomposability (H4)
To test the effects on litter functional identity and diversity on litter decomposability, we first summarized changes in litter functional identity (i.e., total amount of C, N, P, Mg, Na, K, and the CWM of the litter SLA and LDMC in the litterbag) using a principal component analysis (PCA). Second, we summarized changes in litter functional diversity (i.e., variance of C, N, P, Mg, Na, K, SLA, and LDMC in the litterbag) using a PCA (R function 'prcomp'), and third, we tested the effects of litter species richness and litter functional identity and diversity on litter decomposability.
The first two axes of the litter functional identity PCA covered 76% of the litter functional identity variance between the litterbags (Suppl. S3). The first axis (i.e., "Litter nutrient content" axis) was correlated with the chemical content (total amount of C, N, P, Mg, Na, K) of the material in the litterbag, while the second axis (i.e., "Litter morphology" axis) was correlated with the litter morphological traits (i.e., CWM of SLA and LDMC within the litterbag). We extracted the first two axes of the PCA ("Litter nutrient content" and "Litter morphology") for the following analyses. The first two axes of the litter functional diversity PCA explained 91% of the variance in litter functional diversity between the litterbags (Suppl. S3). We extracted the first two axes of the PCA ("Litter fun. diversity 1" and "Litter fun. diversity 2") for the following analysis. To test the effects of litter species richness, litter nutrient content, morphology, and functional diversity on litter decomposability (i.e., "C loss" and "N loss" in the Common Garden experiment), we used linear multiple predictor models and normal distribution assumptions, where all explanatory variables were rescaled using the R function 'scale'. Explanatory variables were selected using forward and backward step selection based on AIC, R 'step' function from 'stats' package (R Core Team, 2021).
Tree biomass, functional traits and planting pattern effects on litterfall composition (H5)
To test the effects of tree biomass, the tree proximity to the traps ("1/dist") and tree ecological strategies (Pierce et al., 2017) on the amount of species-specific litterfall in our traps, we first summarized changes in leaf functional traits (i.e., C, N, P, Mg, Na, K, SLA and LDMC) using a principal component analysis (PCA, Pierce et al. 2017). The first two axes of the PCA covered 77% of the leaf functional identity variance. Second, we extracted the first two axes of the PCA (i.e., Leaf Economics Spectrum axes: "LES 1" and "LES 2") for the following analysis. Third, we fitted linear effect multiple predictor models with normal distribution assumptions using the R 'lm' function, where the explanatory variables were rescaled using the R function 'scale' (R Core Team 2021) and selected using forward and backward step selection based on AIC (R 'step' function).