Linking litter decomposition of above- and below-ground organs to plant–soil feedbacks worldwide


Correspondence author. E-mail:


  1. Conceptual frameworks relating plant traits to ecosystem processes such as organic matter dynamics are progressively moving from a leaf-centred to a whole-plant perspective. Through the use of meta-analysis and global literature data, we quantified the relative roles of litters from above- and below-ground plant organs in ecosystem labile organic matter dynamics.

  2. We found that decomposition rates of leaves, fine roots and fine stems were coordinated across species worldwide although less strongly within ecosystems. We also show that fine roots and stems had lower decomposition rates relative to leaves, with large differences between woody and herbaceous species. Further, we estimated that on average below-ground litter represents approximately 33 and 48% of annual litter inputs in grasslands and forests, respectively.

  3. These results suggest a major role for below-ground litter as a driver of ecosystem organic matter dynamics. We also suggest that, given that fine stem and fine root litters decompose approximately 1.5 and 2.8 times slower, respectively, than leaf litter derived from the same species, cycling of labile organic matter is likely to be much slower than predicted by data from leaf litter decomposition only.

  4. Synthesis. Our results provide evidence that within ecosystems, the relative inputs of above- versus below-ground litter strongly control the overall quality of the litter entering the decomposition system. This in turn determines soil labile organic matter dynamics and associated nutrient release in the ecosystem, which potentially feeds back to the mineral nutrition of plants and therefore plant trait values and plant community composition.


Natural variation in organismal traits is emerging as a powerful tool for understanding both the fitness of organisms in natural ecosystems and their potential effects on them. Consequently, trait-based response-effect frameworks have the potential to predict ecosystem responses to environmental change and the consequences of this for ecosystem processes via plant–soil feedbacks (Chapin, Vitousek & van Cleve 1986; Lavorel & Garnier 2002; Suding et al. 2008). These conceptual frameworks are built on the notion of a single effect of each species via their traits, whereas in reality distinct individuals of the same species and even distinct parts of the same individual can potentially have contrasting impacts on ecosystem processes. This is the case, for instance, of rates of labile organic matter cycling in ecosystems, which largely reflect the relative abundance of plants that have different economic strategies and trait spectra, and that input litter of different qualities to the decomposition subsystem (Garnier et al. 2004; Wardle et al. 2004; Mokany, Ash & Roxburgh 2008).

Plant structural and chemical trait values that influence litter decomposability do not only vary across species but also between plant organs such as leaves, stems and roots (Freschet, Aerts & Cornelissen 2012b). Ecological theories currently used to describe interactions between plant strategies, the decomposability of litters and organic matter cycling in ecosystems nevertheless consider the plant as a homogeneous component and have assumed that plant economic spectra can be described by leaf traits alone (Hobbie 1992; Aerts 1999; Wardle et al. 2004). A large proportion of plant litter inputs are, however, derived from organs other than leaves, notably fine roots (Gill & Jackson 2000; Yuan & Chen 2010), fine stems (Dearden et al. 2006) and large tree stems (Harmon 2009), the importance of which has been frequently overlooked in the development of ecosystem-level organic matter cycling theories. These other tissues do not necessarily impact on ecosystems in a coordinated way with leaves.

Evidence is growing that structural and chemical traits (e.g. concentrations of nutrients and recalcitrant compounds) have largely similar impacts on litter decomposition rates regardless of the plant organ considered (Silver & Miya 2001; Vivanco & Austin 2006; Birouste et al. 2012; Freschet, Aerts & Cornelissen 2012b; but see Hobbie et al. 2010). However, it remains unresolved as to whether plant species show coordinated life strategies (e.g. carbon and nutrient economics) across their distinct organs above- and below-ground (Craine et al. 2005; Tjoelker et al. 2005; Baraloto et al. 2010; Freschet et al. 2010a) and consequently whether decomposability of different organs is correlated across species (Hobbie et al. 2010; Birouste et al. 2012; Freschet, Aerts & Cornelissen 2012b). On one hand, high specialization of one particular organ often appears to be a successful strategy amongst plants (e.g. carnivorous leaves, cluster and mangrove roots, stems of cacti and lianas), suggesting that unbalanced qualitative investments towards distinct organs may be viable strategies – with potential trait ‘afterlife’ consequences for plant organ decomposability. On the other hand, it is probable that most plants would have evolved consistent resource acquisition and defence strategies across their distinct organs to better cope with the dynamic environmental conditions (e.g. resource availability and disturbance) in the course of their evolutionary history (Carroll et al. 2007). It also appears that any potential coordination between traits of different plant organs may be substantially altered by their plastic responses to contrasting influences above- and below-ground (Ryser & Eek 2000). Further, differential senescence processes, including elemental resorption efficiencies, can potentially occur above- and below-ground (Nambiar & Fife 1991; Aerts, Bakker & Caluwe 1992; Freschet et al. 2010b) with likely consequences for coordination among organs of litter decomposability.

Absent or weak coordination between organs with regards to their decomposability would inevitably lead to contrasting impacts of different organs of any one plant on organic matter cycling rates; this would in turn complicate our understanding of feedbacks between plant traits and carbon and nutrient cycling rates (Chapin, Vitousek & van Cleve 1986). Besides, current predictions of ecosystem properties that are usually based on leaf litter characteristics (but see Brovkin et al. 2012) can at best provide approximate measures of ecosystem organic matter cycling because of the potentially substantial differences in decomposability between plant leaves, roots and stems (Taylor et al. 1991; Freschet, Aerts & Cornelissen 2012b). More predictable patterns would emerge from a combination of strong coordination of organ decomposability across species and consistent overall differences in litter decomposition rate between organs, provided that one can also quantify the relative litter mass inputs of each organ into the decomposition subsystem.

Here, we apply meta-analysis techniques to several published and unpublished studies from around the world that have investigated the decomposability of leaves, fine roots and/or fine stems of plants to answer the questions: (i) do plant species show a global coordination – as tested statistically by the strength of correlation – in decomposability across organs? and (ii) are there consistent relative differences in the decomposition rates of these organs? Using global literature data, we also address the question (iii) of how the relative proportion of litter inputs from different organs may influence organic matter cycling in different ecosystem types. Finally, we discuss how integration of the answers to these three questions can advance our understanding of soil-plant feedbacks, at both the plant and the ecosystem scales, in different ecosystems and biomes.

Materials and methods

Data collection

We collected decomposition rate data from 32 published and unpublished studies where several plant tissues of distinct organs (leaves, fine stems and/or fine roots) from the same species had been left to decompose in the same site or type of microcosm. The 13 larger studies, including four or more species, represent all continents except Africa and Antarctica (Fig. 1) and encompass plant species from a wide range of biomes including equatorial, tropical, sub-tropical, semi-arid, Mediterranean, dry and wet temperate, alpine, sub-arctic and arctic regions (Khiewtam & Ramakrishnan 1993; Hobbie 1996; Wardle et al. 1998; Abiven et al. 2005; Jalota et al. 2006; Vivanco & Austin 2006; Hobbie et al. 2010; Wang, Liu & Mo 2010; Birouste et al. 2012; Freschet, Aerts & Cornelissen 2012b; Jackson, Peltzer & Wardle 2013). Coarse woody tissues were not included in this meta-analysis because, to our knowledge, only two previous studies had incubated both coarse wood and other organs of multiple species in the same incubation conditions (Freschet, Aerts & Cornelissen 2012b; Jackson, Peltzer & Wardle 2013).

Figure 1.

Location of the thirteen major litter incubation studies used in this meta-analysis. Pie diagrams indicate the plant parts incubated, with leaves in blue, fine roots in red and fine stems in orange. Nineteen other studies with fewer than four species are shown in smaller grey-scale pie diagrams.

To filter the data available in the literature, we defined ‘fine stems’ as any woody or non-woody elongated structure supporting leaves or reproductive organs of < 5 mm in diameter and ‘fine roots’ as any root of < 2 mm in diameter. In contrast to larger branches and trunks of trees, twigs of woody species were considered closely related to stems of forbs and herbs in terms of function and physiological activity.

Measuring the decomposability of roots is challenging for a number of reasons. Because leaves and stems generally senesce and decompose at the soil surface, while roots senesce and decompose below-ground and surrounded by mycorrhizal fungi and saprophytic organisms, most studies have incubated fine roots under different conditions compared to those for leaves and fine stems. Also, because collecting freshly senesced roots is generally challenging (Hobbie et al. 2010) compared to collecting senesced leaves or stems, most decomposition studies used live fine roots but senesced leaves and/or fine stems. Finally, because fine roots can be of small diameter, some studies have used finer mesh size for roots than for leaves and stems, to prevent the loss of litter fragments. Nevertheless, differences in incubation conditions between tissues were consistent within studies, which allows for comparison of between-species correlation in organ decomposability at the level of individual studies, as performed in meta-analysis. Also, to reduce the degree of uncertainty in our worldwide estimate of fine root decomposability relative to that of other organs, we estimated this relative decomposition rate using only studies where live fine roots were decomposed below-ground.

Because there was a large range in incubation time both between and within studies (from < 1 to 41 months) and the number of separate harvests (from 1 to 12), we collected decomposition records as percentage mass loss (%ML) of the last harvest common to all species and tissues within each study. Nevertheless, when sequential mass loss data were available, we collected %ML before it showed a ‘decomposition limit’ (sensu Berg et al. 1996), that is, before only the most highly recalcitrant fraction remained. These %ML were then transformed into decomposition constants k (y−1) using the negative exponential function %ML = 100−100 e−kt where t is the duration of litter incubation in years (Cornwell et al. 2008). Although k-values do not adequately describe the fate of the particularly recalcitrant fraction at the tail end of the decomposition process (Prescott 2010), they provide good estimates of decomposition rates at initial and middle stages of the decomposition process. This meta-analysis therefore focuses on the labile pool of organic matter in litter inputs that is broken down in the initial to middle stages of decomposition, which is the component generally associated with shorter-term nutrient cycling (Coûteaux, Bottner & Berg 1995).


We used standard meta-analysis techniques (MetaWin v2.1; Rosenberg, Adams & Gurevitch 2000) to test (i) across-species correlations in decomposition rates of leaf, fine root and fine stem tissues and (ii) across-species differences in organ decomposition constant k.

Regarding the first analysis, Pearson's correlation coefficients of (log-transformed) k-values between pairs of plant organs were calculated for each of the 13 individual studies that used four or more species. Effect size estimates were calculated from Fisher's z-transformation of these correlation coefficients. Of these 13 studies, 11 were used for estimating leaf-fine root k relationship (= 104 total data points), eight for leaf-fine stem k relationship (= 128) and five for fine stem-fine root k relationship (= 54). Additionally, the meta-analysis of leaf-fine root k relationship was repeated using only studies that incubated leaves and roots in the same conditions (either ‘above-ground’ or ‘below-ground’; five studies, = 48) and only studies that incubated leaves above-ground versus roots below-ground (six studies, = 56), respectively. We used random-effect models without data structure for all pair-wise plant organ correlations and described uncertainty in effect size estimates with percentile bootstrap confidence intervals (95% CI).

Regarding the second analysis, means and standard deviations of k-values for each plant organ were calculated for each of the 26 individual studies that used two or more species. Effect size estimates of differences between organ k were calculated from Hedge's d standardized mean differences. Of these 26 studies, 15 were used in the meta-analysis of leaf-fine stem (n = 148) comparison. Because we restricted our comparison to live fine roots that were decomposed below-ground, only 14 studies were used for leaf-fine root k comparison (n = 72) and no fine stem-fine root k comparison was performed because of the low number of studies available (four studies, n = 25). Additionally, each test was repeated using only herbaceous species and only woody species, respectively. In each case, we used random-effect models without data structure and described uncertainty in effect size estimates with bootstrap 95% CI.

We estimated the slopes of relationships between (log-transformed) organ k with standardized major axes regressions and used standard slope comparison procedures (SMATR-package; Warton et al. 2006) to test for slope heterogeneity between studies.

Estimates of annual litter inputs

We extracted data from the literature to provide global estimations of average relative annual litter inputs of distinct plant organs for species representative of two contrasting ecosystem types: grassland (species of grass, forb and sedge) and forest (tree species). These estimates were calculated from global data on (i) relative mass fractions of leaf, stem and root parts and (ii) average annual organ turnovers of grassland and forest species (Table S1 in Supporting information). All data represent the mean of numerous species from widely distributed grassland and forest types (including tropical, temperate and boreal/alpine areas). Relative mass fractions of leaf, stem and root parts of typical grassland and boreal, temperate and tropical forests’ species were taken from Poorter et al. (2012). Estimates of average annual below-ground organ turnovers of typical grassland and forest species from boreal, temperate and tropical areas were derived from the global synthesis of Gill & Jackson (2000). Average annual turnover of above-ground organs from grassland species were estimated from Scurlock, Johnson & Olson (2002) as the ratio of annual above-ground net primary productivity to peak biomass averaged over five types of grasslands worldwide including tropical, temperate and boreal areas. Average turnover of leaves from boreal, temperate and tropical tree species was estimated as the inverse of mean leaf life span for a selection of species from these vegetation types using the GLOPNET data set (Wright et al. 2004) where deciduous leaves were attributed an average annual turnover of 1 (i.e., one life cycle per year). Finally, because we were interested in the fine woody fraction of tree stems, relative annual litter inputs of twigs from trees were estimated from leaf to fine stem annual litter input ratios from surveys of boreal (Laiho & Prescott 1999; Dearden et al. 2006), temperate (Sykes & Bunce 1970; Hansen et al. 2009) and tropical (Khiewtam & Ramakrishnan 1993; Sundarapandian & Swamy 1999; Liu, Fox & Xu 2003) forest tree species.


A survey of correlations between plant organ decomposabilities for individual studies (Table 1) shows that virtually all studies including ten or more species generated positive correlation coefficients, but that only half of these deviated significantly from zero. Aggregating and weighting these results through the use of meta-analysis revealed significant correlations between all three pairs of plant tissues (Fig. 2) as shown by the positive mean effect size and 95% CI of the leaf-fine root k relationship (0.46; 95% CI = 27–66), the leaf-fine stem k relationship (0.81; 95% CI = 67–91), and the fine stem-fine root k relationship (0.61; 95% CI = 34–89). Regarding the leaf-fine root k relationship, a meta-analysis of the subset of studies that incubated leaves and roots in the same conditions (either ‘above-ground’ or ‘below-ground’; Fig. S1) also revealed a significant, positive correlation (effect size = 0.44; 95% CI = 29–68). The same was true with the subset of studies that incubated leaves above-ground versus roots below-ground (effect size = 0.49; 95% CI = 19–86).

Table 1. Overview of correlations between organ decomposabilities (k) in studies with four or more species. For each study, k-values are calculated from the last harvest common to all species and tissues and might therefore differ from those shown in the respective publications
 LocationClimateEcosystem typeIncubation typeDuration (months)Leaves – fine rootsLeaves – fine stemsFine stems – fine roots
n r P n r P n r P
  1. Significant (P < 0.05) and marginally significant (< 0.10) Pearson's correlations are indicated in bold, with n the number of species available and r the correlation coefficient.

Abiven et al. 2005S. BrazilTropicalCroplandField3.550.430.4715−0.040.9445−0.430.476
Birouste et al. 2012FranceMediterraneanGrasslandLab 2 17 0.59 0.013       
Freschet et al. 2012ab; N. SwedenSub-arcticForestField24 11 0.67 0.025 38 0.75 <0.001 11 0.77 0.006
Hobbie 1996AlaskaSub-arcticTundraLab4.550.110.85550.750.14550.690.193
Hobbie et al. 2010PolandTemperateForestField1711−0.020.953      
Jackson et al. 2013S. New ZealandTemperateForestLab4    27 0.66 <0.001    
Jalota et al. 2006E. AustraliaSemi-aridCroplandField2440.490.508      
Khiewtam & Ramakrishnan 1993N.E. IndiaTropicalForestField12   40.180.821   
Vivanco & Austin 2006ArgentinaTemperateGrasslandField12100.530.114      
Wang et al. 2010S. ChinaSubtropicalForestField12 4 0.99 0.014       
Wardle et al. 1998N. New ZealandTemperateGrasslandLab1170.330.193 17 0.59 0.012 20 0.39 0.091
Unp. data (China)ChinaSubtropicalForestField12100.230.519 14 0.58 0.031 13 0.67 0.013
Unp. data (Russia)CaucasusAlpine temperateGrasslandField24100.310.384 18 0.66 0.003    
Figure 2.

Relationships between decomposition constant k of leaves, fine roots and fine stems across species and studies. Each point represents one species. Large points and standardized major axis regression lines characterize studies with four or more species and therefore included in the meta-analysis. Small (pale) points represent species from studies with fewer than four species. Mean effect sizes and 95% confidence intervals (CI) of organ k correlations are displayed graphically. The common slope β (±95% CI) of standardized major axis regressions across all studies is given for the leaf-fine root k relationship and the fine root-fine stem k relationship, but not for the leaf-fine stem k relationship because the probabilities (P) to wrongly reject the slope homogeneity hypothesis were 0.40, 0.15 and 0.04, respectively.

Across all plant functional types, the average ratio of leaf to fine stem k was 2.8 (n = 148; effect size = 1.27, 95% CI = 0.66–1.86; Fig. 3). However, this concealed large differences in this ratio between woody species (3.4; = 83) and herbaceous species (1.9; n = 65). Across all plant species data for below-ground incubation of live fine roots versus above-ground incubation of senesced leaves, the average ratio of leaf to fine root k was 1.5 (n = 72; effect size = 0.47, 95% CI = −0.11–1.17; Fig. 3). This ratio increased to 1.8 (n = 38) for herbaceous species and fell to 1.2 (n = 34) for woody species.

Figure 3.

Ratios of leaf to fine root and leaf to fine stem decomposition rates (k), representing how many times slower decomposition is for fine roots or fine stems than for leaves of the same species. Box plots represent median, first and third quartiles and 95% intervals, and mean ratios are displayed as numbers. n is number of species. † leaf: fine root k ratios are based on species for which live fine roots were incubated below-ground. * and ns indicate effect sizes of organ k comparison significantly and non-significantly different from zero at = 0.05, respectively.

By bringing together global literature data of relative mass fractions and annual turnovers of leaf, fine stems and fine roots of plants, we estimated that a grassland species would on average shed only 26% of its total annual litter production as leaves, against 41% as fine stems and 33% as fine roots (Fig. 4). In contrast, an average forest species would shed 41% of its total annual litter production as leaves, as compared to 11% as fine stems and 48% as fine roots.

Figure 4.

Litter decomposability and relative litter input from different plant organs control the quality of organic matter entering the decomposition subsystem, as shown for two representative species of contrasting growth-form. Relative litter inputs of each organ (iorgan) are expressed in percentage of the total plant annual litter input. Mean residence time of litter of an organ (torgan) is expressed relative to that of leaf litter. Average forest and grassland species inputs are estimated from data from contrasting climatic areas (see main text). Forest and grassland root annual litter inputs are displayed in relation to the mean residence time of their respective fine root fractions only, which assumes that fine roots represent the bulk of the annual turnover of root systems.


These analyses revealed global within-site coordination among leaf, fine root and fine stem decomposability. However, they also provided evidence for a large scatter around the regression lines that describe these relationships, and for faster decomposition rates of leaves than of fine roots and fine stems. Because the same litter traits to a large extent determine decomposability across different plant organs, and because most of these traits are conserved through organ senescence, our results suggest that the traits responsible for decomposability of plant litters are most likely to be correlated across leaves, stems and roots. These results support the view that natural selection favours plants with coordinated nutrient and carbon investment strategies above- and below-ground. In other words, balanced structural and chemical investments towards distinct organs appear as a relatively successful strategy for plants in order to cope with the dynamic environmental conditions (e.g. resource availability and disturbance) in the course of their evolutionary history (Carroll et al. 2007). From a functional perspective, our results provide partial support of the idea that plants can be coarsely characterized by a whole-plant life strategy (e.g. Grime 2001; Díaz et al. 2004; Freschet et al. 2010a). The coordinated ‘afterlife’ effects of nutrient concentrations and carbon investments on decomposability have tremendous consequences for how plants interact with their environment and more specifically, how plants control biogeochemical cycling in ecosystems, with potential feedbacks to plant community traits and composition. Nevertheless, the extent to which current plant–soil feedback frameworks need to be further developed depends on the strength and scale at which plant organ coordination occurs.

At the ecosystem scale, these results suggest that differences in the dynamics of labile organic matter observed between any two coexisting plant species above-ground are probably paralleled below-ground. In addition, we provide some preliminary global estimates highlighting that the contribution of fine stem and fine root litters represent a large part of the overall annual plant litter production. Finally, we show that these litters display lower decomposabilities than those of leaf litter. Together, these results show that allocation, especially with regard to the relative proportion of leaf versus non-leaf organs, is a central aspect of whole-plant effects on labile soil organic matter dynamics. For instance, an average grassland species (e.g. grass, forb, sedge) should shed only 26% of its annual litter as leaf litter, as compared to 41% as fine stems (e.g. culm, flowering stem) and 33% as fine roots (Fig. 4). Considering that fine stem and fine root litters of grassland species decompose roughly twice as slowly than their leaf litter, organic matter cycling in grasslands is likely to be much slower than predicted by estimates of leaf litter only. In contrast, with respect to an average forest (tree) species, the percentage annual litter input represented by leaves increases to 41%, against 11 and 48% for fine stem and fine root litters respectively, and the decomposition rate of roots is only slightly slower than that of leaves. As a consequence, labile organic matter dynamics may in this case be relatively well inferred from estimates of leaf litter decomposition only. However, we emphasize that these numbers represent only approximate global averages for typical ecosystem types and that considerable variation is likely to exists across species and ecosystems (Gill & Jackson 2000; Poorter et al. 2012).

Our estimates of litter inputs in forested systems assume that fine roots and fine stems represent the bulk of the annual turnover of root systems and stem components. Whereas the impacts of coarse woody debris on organic matter cycling may be negligible in young forests that are aggrading biomass, this assumption is inappropriate for forests that have high inputs of wood such as old-growth forests or forests that are periodically disturbed (Harmon 2009). In contrast to our results on finer parts of plants, recent studies have suggested that coarse stem components of trees might not show coordinated functional trait syndromes with those of other organs (Baraloto et al. 2010) with consequently low coordination in their afterlife effects on decomposability (Jackson, Peltzer & Wardle 2013). This suggests that, in forests with substantial amounts of coarse woody debris (including branches, stumps and coarse roots), organic matter dynamics cannot be adequately predicted from leaf decomposition rates only. We recognize that coarse woody debris also play an important role in the carbon cycling and soil organic matter build-up of forested ecosystems (Harmon 2009; Cornelissen et al. 2012). Nevertheless, because of its low nutrient content, slow turnover rate and low decomposition rate, its influence is less important in the short term for the cycling of nutrients, particularly N and P (Laiho & Prescott 2004).

The range of observed plant economics and consequently afterlife effects on organ decomposability is likely to be greater at the among-ecosystem than at the within-ecosystem level. Further, among ecosystems, plant above-ground and below-ground litter decomposition rates generally decrease as temperature and humidity conditions decline as a consequence of both direct effects (microbial enzyme kinetics; Fierer et al. 2005) and indirect effects (litter quality, decomposer community composition and activity; Cornwell et al. 2008; Wall et al. 2008; Freschet, Aerts & Cornelissen 2012a). Consequently, because our meta-analysis is based on the observations at the within-ecosystem scale, this suggests a high likelihood of observing consistent rankings of organ decomposability across ecosystems between pools of species. Further, ecosystem-level feedbacks between soil fertility and plant economic strategies may be amplified when considering the effects of shifts in allocation on the proportion of above- and below-ground litters. For example, the root mass fraction, which generally decomposes more slowly than the leaf mass fraction, tends to increase with increasing climatic stress due to lower temperature and water availability (Poorter et al. 2012). Nevertheless, as discussed above, such global patterns of ecosystem-level plant–soil feedbacks should be highly variable across ecosystem types, and particularly among those dominated by contrasting growth forms.

Large variation exists among plant species in above- versus below-ground biomass allocation (and organ mass fractions), both within and across ecosystems, as well as within each of tropical, temperate and boreal areas (Cairns et al. 1997; Poorter et al. 2012). This high variability is likely to drive highly contrasting patterns of litter production above- versus below-ground. Also, we show that fine stem and fine root litters decompose approximately 1.5 and 2.8 times slower, respectively, than leaf litter derived from the same species and that these differences in decomposability across organs are highly variable across species. Consequently, large differences in above- versus below-ground litter production across species will strongly control the quality of litters entering the decomposition system (Fig. 5). For instance, two species that do not differ from each other in the qualities of each of the litter fractions that they produce, but which show contrasting above- versus below-ground litter production (e.g. Fig. 5a vs. 5b) will produce overall litter pools of contrasting quality. Consequently, litter quality (i.e. afterlife effects from plant functional traits; Cornelissen 1996) and litter production patterns co-determine labile organic matter dynamics and influence nutrient release for both microbial and plant uptake (Kaye & Hart 1997; Schimel & Bennett 2004). As described in previous theories of plant–soil feedback, this partial control by plants on nutrient availability via litter quality feeds back to the mineral nutrition of plants, and therefore to plant functional trait values and, at larger scale, plant community composition (Hobbie 1992; Aerts 1999; Wardle et al. 2004). Our work develops on previous frameworks on plant–soil interactions by recognizing that differences among plant species in their (fine) litter production across their different organs may be as important as among-species differences in functional traits for driving nutrient cycling in ecosystems.

Figure 5.

Plant economics and relative above- and below-ground litter inputs co-determine the overall quality of the litter entering the decomposition system, and therefore the dynamics of the labile organic matter pool, and associated nutrient release. This in turn affects the supply rate of mineral nutrients for the nutrition of plants, which in turn feeds back to plant trait values and plant community composition. While this framework is illustrated here at the patch scale with one species only, it is also able to represent the ‘average litter quality and quantity’ of inputs from whole plant communities containing multiple coexisting species in line with the ‘mass ratio hypothesis’ (Grime 1998). A resource acquisitive plant community (a, b) produces litter of consistently higher decomposability than a resource conservative plant community (c, d). In parallel, the relative inputs of litter from different organs, for instance, leaf and root components, which contrast in their decomposability, co-determine overall litter decomposability. These two factors together determine the sign and strength of the feedback.

Within this framework, it is important to recognize that the relationships between litter decomposition, nutrient release and plant nutrient uptake can also vary between sites as the environmental conditions that drive competition for nutrients between microbes and plants change (Kaye & Hart 1997; Schimel & Bennett 2004). In nutrient-rich soils, there are few nutrient-poor soil microsites where microbes are in competition with plants for nutrient uptake (Schimel & Bennett 2004) and most of the nutrients released through decomposition of litters should be readily available for plants (Schimel & Hättenschwiler 2007). In contrast, in nutrient-poor soils, nutrients released from litters should be subject to strong plant-microbe competition (Schimel & Bennett 2004). At the scale of plant communities, because nutrient-rich and nutrient-poor sites are usually dominated by acquisitive (nutrient-rich) and conservative (nutrient-poor) plants respectively, site differences in plant-microbe competition for nutrients should reinforce ecosystem-scale positive and negative feedbacks in nutrient-rich and nutrient-poor sites, respectively. Nevertheless, at the scale of individual plants, because there is large variation in plant economics and litter qualities between species and organs within most ecosystems worldwide (Cornwell et al. 2008; Hättenschwiler et al. 2008; Freschet, Aerts & Cornelissen 2012b), negative feedbacks potentially initiated by nutrient conservative species in nutrient-rich soils, and positive feedbacks by acquisitive species in nutrient-poor soils, should be generally diminished.

Several aspects of this framework should be further developed. First, it is unclear how the respective influences of above- and below-ground litters interact to determine soil organic matter dynamics. Large differences in organic matter quality between the litter layer (typically comprised of leaf and fine stem litter), the topsoil layer (dominated by fine roots) and deeper soil layers should lead to very different parallel food webs, with consequences for soil organic matter stabilization and nutrient bioavailability (Schmidt et al. 2011). Second, the role that phenotypic variation plays in these feedbacks through influencing the functional trait values, the total biomass, and the relative mass fraction of the distinct plant functional parts is likely to be substantial and merits further attention. As an example, an intense water stress can drive a 50% within-species decrease in total plant biomass relative to non-stressed conditions and lead to a readjustment of more than 10% of its total biomass from above- to below-ground organs (Poorter et al. 2012). Finally, although the short-term influence of coarse woody debris on nutrient cycling appears minor relative to that of fine plant tissues which are of higher quality and therefore decompose much faster (Laiho & Prescott 2004), its indirect effects on environmental conditions for decomposition of fine litters are potentially large (Harmon et al. 1986; Nordén et al. 2004). As such, a better understanding of its role within the framework that we propose would require further studies that include spatially explicit, long-term approaches.

This quantitative illustration of differential plant–soil feedback strengths in carbon and nutrient cycling, as determined by (i) the integration of relative mass inputs of above-ground versus below-ground litters derived from different organs, (ii) their overall differences in decomposability and (iii) the coordination of organ decomposabilities within several key ecosystem types, will help advance our understanding and predictive power of global biogeochemical cycling rates in a world with rapidly shifting vegetation composition.


We are grateful to Richard van Logtestijn for his help in setting up the Caucasus litter decomposition study. GTF and DAW were supported by a Wallenberg Scholars Award to DAW. The litter study in the Russian Caucasus was supported by NWO grant 047.018.003 to JHCC and other NWO participants GTF, TGE, NAS and VGO. TGE and VGO acknowledge support from the Russian Foundation for Basic Research (project 11-04-01215). The litter study in sub-tropical China, as well as WL and JT, was supported by the Natural Science Foundation Project CQ CSTC (2010BB1011).