SEARCH

SEARCH BY CITATION

Keywords:

  • chronosequence;
  • decomposition;
  • ecosystem;
  • litter mixing;
  • nutrient release;
  • retrogression;
  • species effects;
  • succession

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • 1
    Following major disturbances ecosystem development occurs but in the prolonged absence of disturbance a decline (retrogressive) phase follows in which productivity and nutrient availability diminishes. Although it is recognized that litter quality and decomposition rates decrease as retrogression proceeds, little is known about the extent to which this is driven among- vs. within-species variation across these sequences.
  • 2
    We selected six long-term chronosequences that each included retrogressive stages, in New Zealand, Hawaii, Sweden, Alaska and Australia. Two involve significant species turnover across the sequence so that different species dominate at different stages, two involve low species turnover so that the same dominant species occur at all stages, and two involve some turnover of species but with certain species persisting throughout most of the sequence.
  • 3
    For each chronosequence, we collected litter from each dominant plant species at each stage of that sequence. For each litter collection we measured concentrations of N and P, and performed laboratory decomposition bioassays to measure mass loss, N and P loss, and the response of mass loss to mixture with litters of coexisting species.
  • 4
    We found that litter N and P concentrations often declined with increasing ecosystem age, both among- and within-species. However, the relative importance of among- and within-species effects varied across the six chronosequences. Rates of litter mass, N, and P loss during decomposition sometimes decreased with increasing ecosystem age, but most often at the among-species rather than the within-species level.
  • 5
    Litter mixing effects often varied across chronosequence stages, but the magnitude and direction of these effects was inconsistent among sequences. Variation in litter mixing effects across chronosequence stages was driven mainly by among- rather than within-species variation.
  • 6
    Although several recent studies have emphasized the role of within-species variation on ecosystem properties, our results point to among-species variation as a consistently important ecological driver, with within-species variation being important only for some variables and in some instances. As such they highlight that decomposition processes are most likely to be highly responsive to gradients of soil fertility (such as across chronosequences) when significant species turnover occurs across the gradient.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Following major ecological disturbances, primary or secondary succession takes place, and this involves an initial period of ecosystem development to a maximal biomass phase. Most research on succession has focused on this developmental period, which usually involves distinct shifts in primary productivity, soil processes, and accumulation of biomass and soil organic matter (Odum 1969; Walker & del Moral 2003). However, in the prolonged absence of catastrophic disturbance and as soils age, a decline or retrogressive phase often follows during which there is a decrease in vegetation biomass and ecosystem productivity (Vitousek & Farrington 1997; Walker et al. 2001; Wardle et al. 2003; Vitousek 2004). This decline in vegetation biomass is often associated with reduced availability of major soil nutrients, particularly phosphorus (P) (Walker & Syers 1976; Vitousek 2004), and an increase in the soil nitrogen (N) to P ratio (Wardle et al. 2004). This pattern is relatively consistent across several long-term chronosequences (Wardle et al. 2004) and only a disturbance of sufficient magnitude to rejuvenate the soil will reverse these effects (Walker et al. 2001).

The decline in soil fertility during ecosystem retrogression causes a reduced supply of plant-available nutrients from the soil (Vitousek & Farrington 1997; Wardle et al. 2008). As a consequence, plants produce litter of poorer quality with lower concentrations of nutrients (Richardson et al. 2004; Coomes et al. 2005) that show reduced rates of mass loss and nutrient release during decomposition (Crews et al. 1995; Wardle et al. 2003; Vitousek 2004). This reduction in litter quality during retrogression could potentially occur either among- or within-species. Among-species effects involve different plant species with contrasting physiological attributes occupying different chronosequence stages (Cortez et al. 2007; Quested et al. 2007). Thus, as retrogression proceeds, species that are less adapted for nutrient poor conditions should be replaced by species that are better adapted but that produce inherently poorer quality litter (Grime 2001; Kazakou et al. 2006). Within-species effects involve species that can potentially occupy a broad range of conditions and that show a high degree of phenotypic plasticity in their leaf and litter quality (see Vallandarez et al. 2007). For example, Metrosideros polymorpha, which dominates across long-term chronosequences in Hawaii, shows marked declines in litter quality and decomposability as retrogression proceeds (Crews et al. 1995). However, little is known about the relative contribution of among- and within-species variation in determining the overall decline of litter quality and decomposability during ecosystem retrogression.

In the vast majority of situations, litters of different species occur together in the litter layer, and therefore do not decompose in isolation from one another. A growing number of studies have investigated interactive effects among different litter types (Blair et al. 1990; Wardle et al. 1997a, Schädler & Brandl 2005), and a range of effects from strongly positive to strongly negative have been reported (Gartner & Cardon 2004; Hättenschwiler et al. 2005). However, whether or not litter quality is an important determinant of litter mixing effects remains poorly understood, and those studies that have explicitly addressed this issue have yielded contrasting results (e.g. Wardle et al. 1997a; Hoorens et al. 2003; Quested et al. 2005; Liu et al. 2007; Schimel & Hättenschwiler 2007). Further, the issue of how litter mixing effects may vary across environmental gradients or contrasting habitats has been investigated in only a handful of studies (Dearden et al. 2006; Gartner & Cardon 2006; Madritch & Cardinale 2007; Jonsson & Wardle 2008). It is well known that retrogressive chronosequences represent strong gradients of resource availability (Vitousek 2004), and if litter quality and environmental context are important drivers of litter mixing effects, then we would expect these effects to change predictably during retrogression.

In an earlier study (Wardle et al. 2004), we measured changes in above-ground and below-ground properties across each of six long-term chronosequences from around the world. We showed that plant biomass consistently declined during ecosystem retrogression, and that this was generally associated with reductions in soil fertility, substrate and litter N : P ratios, decomposer activity, and mass and nutrient loss rates from decomposing litter. In the present study, we further explored the litter decomposition data obtained from each of these chronosequences to answer the following two questions: (i) To what extent are overall changes in litter properties (i.e. nutrient concentrations, and loss of mass and nutrients during decomposition) across these chronosequences explained by changes among species vs. within species for these properties?; (ii) Does the magnitude of litter mixing effects among coexisting species within stands change in a predictable manner across these chronosequences, and is this explained by across-chronosequence changes at the among-species level or at the within-species level? Our ultimate goal is to better understand the roles of among-species and within-species variability in determining how ecosystem processes change during retrogression, and whether broadly similar trends occur during retrogression for vastly contrasting chronosequences.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

study sites and sampling

For this investigation we focused on the same six long-term chronosequences used by Wardle et al. (2004, 2008); all sequences are at least several thousand years (and up to 4·1 million years) in duration (Supporting Information, Appendix S1). Two of these sequences are in the Boreal zone, that is, the Arjeplog sequence in northern Sweden (Wardle et al. 1997b, 2003; Wardle & Zackrisson 2005) and the Glacier Bay sequence of southeastern Alaska (Noble et al. 1984; Chapin et al. 1994). Two are in the temperate zone, that is, the Franz Josef sequence of Westland New Zealand (Walker & Syers 1976; Richardson et al. 2004) and the Waitutu sequence of southern New Zealand (Ward 1988; Coomes et al. 2005). The remaining two are in the sub-tropical zone, that is, the Hawaiian island sequence (Crews et al. 1995; Vitousek & Farrington 1997; Vitousek 2004) and the Cooloola sequence of Queensland, Australia (Thompson 1981; Walker et al. 2001). These sequences are formed on vastly different substrates and have been created by different agents of disturbance (Supporting Information Appendix S1). In all six cases, long-term ecosystem development involving soil weathering and aging has occurred after a catastrophic disturbance event or an event that has substantially reset the successional clock. Five of these chronosequences are primary successional; the sixth sequence, Arjeplog, is a secondary succession, but we maintain that ecological processes in the retrogressive phase of primary and secondary succession are comparable (Walker & del Moral 2003; Wardle et al. 2004). Previous work on these chronosequences has shown that for each sequence there is a substantial decline in tree basal area and forest stature following the maximal biomass phase, and that this is associated with increasing limitation by nutrients, notably P, as soils age (Crews et al. 1995; Wardle et al. 2003, 2004; Richardson et al. 2004; Coomes et al. 2005).

Each chronosequence was sampled once, between February 1998 and April 2001, during late summer. For each chronosequence, we identified several stages of ecosystem development or decline in that sequence (Supporting Information Appendix S1), and then established replicate plots (circular, and typically 10 m radius) within each stage (Wardle et al. 2004). The number of chronosequence stages and total numbers of plots measured are presented in Supporting Information Appendix S1. Numbers of plots per stage were not equal for the Arjeplog and Waitutu sequences because of insufficient space within some stages to locate several replicate plots. For each plot, fresh litter (i.e. the most recently produced litter that we could find on the forest floor) from the most dominant (typically 4–6) vascular plant species were collected and kept in an air-dry condition before analysis.

litter analyses

For each plot, concentrations of N and P in litter from each collected species were determined using colorimetric methods (Technicon Instruments 1977). The decomposability of each litter sample was assessed using a standard laboratory bioassay as described by Wardle et al. (1998, 2002). Here, litters are decomposed in Petri dishes containing a standardized soil or humus substrate containing a resident community of decomposer microbes. For the present study, for each litter sample, three 9 cm diameter Petri dishes were each two-thirds filled with a standardized humus substrate (with 1·8% N and a pH of 4·5%) collected from a mixed podocarp-hardwood forest in the southern South Island of New Zealand and amended to 250% moisture (dry wt basis); a disc of nylon mesh with 1 mm holes was placed on the humus surface. Litter (1 g, oven-dried) was cut into 5 mm fragments and placed on the surface of the mesh of each Petri dish; the dish was then sealed with tape to minimize water loss and incubated at 22 °C. To assess litter mixing effects among species from each plot, Petri dishes were set up in triplicate for every possible two-way mixture of litters of all the species sampled from that plot, as described by Wardle et al. (2002). These were set up as for the monospecific Petri dishes, but with each dish containing 0·5 g of each of the two litter types.

All incubations were performed for 90 days [found by Wardle et al. (2002) to be of appropriate duration for measuring decomposition of litter from forest species], except for the litter collected from the Cooloola sequence which was incubated for 180 days because of its poorer quality (and therefore slower decomposition). At the end of the incubation, all remaining (undecomposed) litter was removed from each Petri dish and rinsed. For the Petri dishes containing two species mixtures of litter, the remaining litter was visually sorted into the component species. All litter was then oven-dried (80 °C, 24 h), and its dry mass determined. For all litter samples that were decomposed in monoculture, the concentrations of N and P in the remaining litter were determined as described above.

The rate of litter decomposition was determined as the percentage mass lost during incubation. Loss of N and P from the litter was calculated as the total mass × nutrient concentration before incubation minus that after incubation (Wardle et al. 2002). The percentage mass loss of each litter type in each two-way mixture was compared with its percentage mass loss in monoculture. These values were used to determine two data values for each litter type as described by Wardle et al. (2002). First, the net effect of each litter type on the rate of decomposition of all the other litter types from that location (hereafter called the ‘litter mixing effect’) was calculated as the mean enhancement of all other litter types by that species in mixture relative to their decomposition in monoculture. Second, the net response of decomposition rate of each litter type to all the other litter types from the same location (hereafter called the ‘litter mixing response’) was calculated as the mean enhancement of the rate of decomposition of that species in all mixtures in which it occurred relative to its decomposition rate in monoculture. Within any given plot, the mean ‘litter mixing effect’ and ‘litter mixing response’ should yield a similar value, but within any given species the ‘litter mixing effect’ and ‘litter mixing response’ are able to differ within plots.

data analyses

For each litter variable, we averaged data for all the litters collected from each plot to provide a single data value for that plot (Wardle et al. 2004). The response variable was then analysed using one-way anova to test for the effect of chronosequence stage, with individual plots serving as the units of replication, as described by Wardle et al. (2008). This analysis tested for the overall response of each litter variable to chronosequence stage, which could be due to either among-species or within-species variation across the chronosequence. All data were tested to confirm that they satisfied the assumptions of parametric analysis.

For each litter variable we used regression analyses to assess among-species and within-species variation across each chronosequence. For these analyses, chronosequence stage was a discrete variable and was therefore treated as a rank variable with a rank of 1 representing the youngest stage (Wardle et al. 2004). The use of rank transformed values to represent chronosequence stage is justified because of the high level of uncertainty in assigning precise ages for many of the chronosequence stages (Wardle et al. 2004). To assess among-species variation, we focused only on those chronosequences for which litter was collected for at least 10 species, that is, Cooloola, Glacier Bay, Franz Josef and Waitutu. Within each chronosequence, for each species we calculated the mean chronosequence stage that the species occurred in. Within that chronosequence, regression analyses were then used to test for the relationship (linear or quadratic) between each response variable and mean chronosequence stage, with each species serving as an independent data point. To assess within-species variation, we considered only those species within each chronosequence for which litter was collected from at least 10 separate plots and was present on at least half the stages of that chronosequence. This enabled us to focus only on those species that maintain a presence in a large proportion of the total range of conditions represented by the chronosequence, which allowed us to perform within-species analyses for 13 separate plant species from the Cooloola, Arjeplog, Glacier Bay and Hawaii chronosequences. For each species within each chronosequence, regression analyses were then used to test for the relationship (linear or quadratic) between each response variable and chronosequence stage, with each plot serving as an independent data point (Wardle et al. 2004).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

litter n and p

The litter N concentration when data from all species in each plot were averaged varied significantly across chronosequence stages for four of the six sequences (i.e. Cooloola, Glacier Bay, Hawaii, Franz Josef) (Fig. 1). For the Glacier Bay and Hawaii sequences, N concentration was maximal at intermediate stages and declined at the later stages; there were no clear trends over time for the Franz Josef and Cooloola sequences. However, at the among-species level, N concentration was significantly related to mean chronosequence stage for only one sequence, that is, Franz Josef (Table 1), and even then only weakly. Meanwhile at the within-species level, litter N concentration was significantly related to chronosequence stage for 7 of the 13 individual species (Table 2). For four of these species (from the Cooloola, Glacier Bay and Hawaii sequences), the relationship was negative linear or quadratic, pointing to declines in N concentration at least during late retrogressive stages. The other three (from the Arjeplog and Hawaii sequences) showed a positive linear or quadratic relationship (Table 2).

image

Figure 1. Changes in litter N, P and N : P ratio across each of six chronosequences, with data from all species in each plot being averaged before data analysis. Details on each successional stage are given in Supporting Information Appendix S1. F and P values are derived from anova with individual plots serving as the units of replication; within each panel bars that are topped by the same letter do not significantly differ at P = 0·05 (least significant difference test); NS = no significant effects of chronosequence stage. Data for the N : P ratio for Hawaii has been log transformed before analysis. Degrees of freedom for each chronosequence are 8,26 for Cooloola, 5,29 for Arjelog, 7,24 for Glacier Bay, 5,23 for Hawaii, 8,17 for Franz Josef, and 6,16 for Waitutu.

Download figure to PowerPoint

Table 1.  Relationships among species between the mean value of a given nutrient property (decomposition rate or nutrient release) for each species, and the mean chronosequence stage that the species occurs in, with each species serving as an independent data point. Data are R2 values (P values in brackets) between mean within-species chronosequence stage and mean within-species value for the response variable. All R2 values are for linear regressions, except when stated as quadratic. ‘n’ = the number of independent data points (i.e. number of different species) for each chronosequence
ChronosequenceNitrogen (N) (%)Phosphorus (P) (%)N : P ratio
  1. Positive linear and negative linear indicate trends of increasing or decreasing values respectively across the chronosequence; positive quadratic and negative quadratic indicate greater or lesser values respectively at intermediate chronosequence stages relative to early or late stages. Values in bold are statistically significantly different to 0 at P = 0.05.

Cooloola (n = 27)0·001 (0·870)0·071 (0·180)0·047 (0·276)
Glacier Bay (n = 10)0·229 (0·161)0·597 (0·041)0·579 (0·049)
negative quadraticpositive quadratic
Franz Josef (n = 24)0·165 (0·048)0·523 (< 0·001)0·286 (0·029)
negative linearnegative linearpositive quadratic
Waitutu (n = 14)0·004 (0·841)0·001 (0·917)0·398 (0·050)
positive quadratic
Table 2.  Response of litter nutrient status to chronosequence stage for individual plant species. Data are R2 values (P values in brackets) between chronosequence stage and response variable. All R2 values are for linear regressions, except when stated as quadratic. ‘n’ = the number of data points (i.e. plots sampled) for each species
SequenceSpeciesNitrogen (N) (%)Phosphorus (P) (%)N : P ratio
  • Data for Stage 3 excluded because of anomalous values resulting from domination of plots by a N-fixing tree species that was absent from all other plots.

  • Positive linear and negative linear indicate trends of increasing or decreasing values respectively across the chronosequence; positive quadratic and negative quadratic indicate greater or lesser values respectively at intermediate chronosequence stages relative to early or late stages. Values in bold are statistically significantly different to 0 at P = 0.05.

CooloolaBanksia integrifolia (n = 11)0·408 (0·035)0·461 (0·021)0·802 (< 0·001)
negative linearnegative linearpositive linear
Banksia serrata (n = 10)0·008 (0·802)0·082 (0·420)0·197 (0·199)
Eucalyptus intermedia (n = 12)0·254 (0·094)0·284 (0·074)0·183 (0·164)
Monotoca sp. (n = 12)0·633 (0·011)0·854 (< 0·001)0·744 (0·001)
negative quadraticnegative quadraticpositive quadratic
ArjeplogBetula pubescens (n = 30)0·001 (0·967)0·002 (0·837)0·008 (0·873)
Empetrum hermaphroditum (n = 30)0·010 (0·598)0·004 (0·749)0·132 (0·048)
negative linear
Picea abies (n = 25)0·000 (0·986)0·002 (0·840)0·001 (0·907)
Pinus sylvestris (n = 25)0·272 (0·006)0·305 (0·003)0·175 (0·033)
positive linearpositive linearnegative linear
Vaccinium myrtillus (n = 30)0·056 (0·205)0·112 (0·070)0·332 (< 0·001)
positive linear
Vaccinium vitis-idaea (n = 30)0·191 (0·016)0·169 (0·024)0·031 (0·349)
positive linearpositive linear
Glacier BayPicea sitchensis (n = 19)0·405 (0·005)0·040 (0·411)0·143 (0·122)
negative linear
Tsuga heterophylla (n = 16)0·079 (0·310)0·151 (0·154)0·122 (0·200)
HawaiiMetrosideros polymorpha (n = 20)0·688 (< 0·001)0·678 (< 0·001)0·505 (0·003)
negative quadraticnegative quadraticpositive linear
Cibotium glaucum (n = 16)0·945 (< 0·001) 0·672 (< 0·001)0·437 (0·023)
positive quadraticpositive quadraticpositive quadratic

When data for all species were averaged for each plot, litter P concentration was significantly related to chronosequence stage for the Glacier Bay and Hawaiian sequences where it was greater during some intermediate stages than for the initial and final stages, and for the Franz Josef sequence where it declined sharply at the later stages (Fig. 1). These trends were supported by analyses at the among-species level for the Glacier Bay and Franz Josef sequences, for which P concentration showed negative linear or quadratic relationships with mean chronosequence stage (Table 1). At the within species level, litter P concentrations were significantly related to chronosequence stage for 6 of the 13 species; three of these relationships (in the Cooloola and Hawaii chronosequences) were negative linear or quadratic, while the other three (in the Hawaii and Arjeplog chronosequences) were positive linear or quadratic (Table 2).

When data for all species in each plot were averaged, the litter N : P ratios varied according to chronosequence stage for all sequences except Arjeplog and Waitutu (Fig. 1). For the Cooloola and Franz Josef sequences, N : P ratios increased sharply at the retrogressive phases, and for the Glacier Bay sequence it decreased and then increased across the sequence. For the Hawaiian sequence, there was a general pattern of increasing N : P ratio, but this was only evident when the very high values for the stage three site (dominated by N fixing trees) were not considered. Domination by N-fixing species did not occur elsewhere along this sequence, or at any stage on any of the other sequences. At the among-species level, three of the four chronosequences tested showed a positive quadratic relationship between N : P ratio and mean chronosequence stage (Table 1). At the within-species level, two species in each of the Cooloola and Hawaii sequences showed positive linear or quadratic relationships between litter N : P ratio and chronosequence stage, while for the Arjeplog sequence one species showed a positive linear relationship and two showed a negative linear relationship (Table 2).

rates of loss of litter mass, n and p

When litter decomposition rates were averaged among all species in each plot, a range of responses to chronosequence stage was detected (Fig. 2). For Hawaii, litter decomposition rates increased with chronosequence stage, while for Franz Josef there was a general pattern of decline. The rate of decomposition was maximized at intermediate stages for the Waitutu sequence and at the early and late stages for the Glacier Bay sequence. There were no clear trends over time for the Cooloola sequence. At the among-species level, two chronosequences, that is, Franz Josef and Cooloola, showed a negative relationship between decomposition rate and mean chronosequence stage (Table 3). At the within-species level, two of the 13 species (from the Cooloola and Arjeplog sequences) showed significant negative relationships between decomposition rate and chronosequence stage, while three species (from the Glacier Bay and Hawaii sequences) showed positive linear or quadratic relationships (Table 4).

image

Figure 2. Changes in litter mass loss and N and P release across each of six chronosequences, with data from all species in each plot being averaged before data analysis. Details on each successional stage are given in Supporting Information Appendix S1. F and P values are derived from anova with individual plots serving as the units of replication; within each panel bars that are topped by the same letter do not significantly differ at P = 0·05 (least significant difference test); NS = no significant effects of chronosequence stage. Degrees of freedom for each chronosequence are 8,26 for Cooloola, 5,29 for Arjelog, 7,24 for Glacier Bay, 5,23 for Hawaii, 8,17 for Franz Josef, and 6,16 for Waitutu.

Download figure to PowerPoint

Table 3.  Relationships among species between the mean value of a given decomposition variable (decomposition rate, or N or P release) for each species, and the mean chronosequence stage that the species occurs in, with each species serving as an independent data point. Data are R2 values (P values in brackets) between mean within-species chronosequence stage and mean within-species value for the response variable. All R2 values are for linear regressions, except when stated as quadratic. ‘n’ = the number of independent data points (i.e. number of different species) for each chronosequence
SequenceDecomposition rate (%)Release of N (mg/g litter)Release of P (mg/g litter)
  1. All data based on 90 days incubation, except for Cooloola which is based on 180 days incubation. Positive linear and negative linear indicate trends of increasing or decreasing values respectively across the chronosequence; positive quadratic and negative quadratic indicate greater or lesser values respectively at intermediate chronosequence stages relative to early or late stages. Values in bold are statistically significant different to 0 at P = 0.05.

Cooloola (n = 27)0·167 (0·034)0·006 (0·705)0·146 (0·049)
negative linearnegative linear
Glacier Bay (n = 10)0·070 (0·461)0·402 (0·048)0·587 (0·046)
negative linearnegative quadratic
Franz Josef (n = 24)0·427 (< 0·001)0·074 (0·198)0·330 (0·003)
negative linearnegative linear
Waitutu (n = 14)0·037 (0·350)0·047 (0·458)0·012 (0·761)
Table 4.  Response of litter decomposition rate and nutrient release to chronosequence stage for individual plant species. Data are R2 values (P values in brackets) between chronosequence stage and response variable. All R2 values are for linear regressions, except when stated as quadratic. ‘n’ = the number of data points (i.e. plots sampled) for each species
SequenceSpeciesDecomposition rate (%)Release of N (mg/g litter)Release of P (mg/g litter)
  1. All data based on 90 days incubation, except for Cooloola which is based on 180 days incubation. Positive linear and negative linear indicate trends of increasing or decreasing values respectively across the chronosequence; positive quadratic and negative quadratic indicate greater or lesser values respectively at intermediate chronosequence stages relative to early or late stages. Values in bold are statistically significantly different to 0 at P = 0.05.

CooloolaBanksia integrifolia (n = 11)0·009 (0·782)0·206 (0·161)0·133 (0·133)
Banksia serrata (n = 10)0·005 (0·854)0·497 (0·022) positive linear0·263 (0·130)
Eucalyptus intermedia (n = 12)0·690 (0·005) negative linear0·103 (0·308)0·003 (0·868)
Monotoca sp. (n = 12)0·028 (0·617)0·365 (0·049) negative linear0·244 (0·121)
ArjeplogBetula pubescens (n = 30)0·010 (0·593)0·166 (0·423)0·001 (0·951)
Empetrum hermaphroditum (n = 30)0·004 (0·727)0·001 (0·929)0·004 (0·734)
Picea abies (n = 25)0·005 (0·745)0·007 (0·707)0·047 (0·319)
Pinus sylvestris (n = 25)0·008 (0·673)0·003 (0·807)0·042 (0·332)
Vaccinium myrtillus (n = 30)0·168 (0·024) negative linear0·007 (0·656)0·133 (0·027) negative linear
Vaccinium vitis-idaea (n = 30)0·007 (0·659)0·137 (0·044) positive linear0·160 (0·028) positive linear
Glacier BayPicea sitchensis (n = 19)0·310 (0·050) positive quadratic0·453 (0·002) positive linear0·259 (0·026) positive linear
Tsuga heterophylla (n = 16)0·166 (0·020) positive linear0·000 (0·954)0·261 (0·050) positive linear
HawaiiMetrosideros polymorpha (n = 24)0·055 (0·277)0·042 (0·331)0·020 (0·508)
Cibotium glaucum (n = 16)0·439 (0·023) positive quadratic0·440 (0·023) positive quadratic0·019 (0·610)

Nitrogen release from litter averaged among all species in each plot showed a generally negative trend for Cooloola, and tended to be optimized at intermediate stages for Glacier Bay, Hawaii and Franz Josef; no response occurred for Arjeplog or Waitutu (Fig. 2). At the among species level, the Glacier Bay sequence showed a significant negative relationship between N release and mean chronosequence stage, while the other chronosequences showed no relationship (Table 3). At the within-species level, four of the 13 species showed a positive linear or quadratic relationship, that is, one each from the Cooloola, Arjeplog, Franz Josef and Hawaii chronosequences (Table 4). Only one species (from the Cooloola sequence) showed a negative relationship.

When data were averaged among all species for each plot, the release of P from litter declined during some of the later phases for Franz Josef and Hawaii, although there were inconsistencies at earlier stages of the sequences (Fig. 2). Meanwhile, P immobilization by litter was minimized at intermediate stages for the Cooloola sequence. Phosphorus release for the other three sequences was not related to chronosequence age. At the among-species level, three chronosequences (Cooloola, Glacier Bay and Franz Josef) showed significant negative linear or quadratic relationships between P release and mean chronosequence stage (Table 3). However, at the within species level, three species (from the Glacier Bay and Arjeplog sequences) showed significant positive relationships, while only one (from the Arjeplog sequence) showed a significant negative relationship (Table 4).

litter mixing effects and responses

When the litter mixing effect was averaged among all species in each plot, four of the six chronosequences showed significant responses to chronosequence stage (Fig. 3). For the Cooloola and Glacier Bay sequences mixing effects were strongly negative at some intermediate stages, while for the Franz Josef and Waitutu sequences mixing effects were strongly positive at intermediate stages. At the among-species level, relationships between litter mixing effect and mean chronosequence stage were negative linear or quadratic for the Franz Josef and Waitutu sequences, and positive quadratic for the Cooloola sequence (Table 5). At the within-species level, none of the 14 species showed a significant relationship between litter mixing effect and chronosequence stage (data not presented).

image

Figure 3. Changes in litter mixing effect (i.e. mean effect of litter of a given species on decomposition rates of all the other species collected from the same plot), with data from all species in each plot being averaged before data analysis. The litter mixing response (i.e. mean response of decomposition rate of a given species to all the other species from the same plot) yields approximately the same response to chronosequence stage at the within-plot scale (not presented). Details on each successional stage are given in Supporting Information Appendix S1. F and P values are derived from anova with individual plots serving as the units of replication; within each panel bars that are topped by the same letter do not significantly differ at P = 0·05 (least significant difference test); NS = no significant effects of chronosequence stage. Degrees of freedom for each chronosequence are 8,26 for Cooloola, 5,29 for Arjelog, 7,24 for Glacier Bay, 5,23 for Hawaii, 8,17 for Franz Josef, and 6,16 for Waitutu.

Download figure to PowerPoint

Table 5.  Relationships among species between the mean value of a given litter mixing variable (effect or response) for each species, and the mean chronosequence stage that the species occurs in, with each species serving as an independent data point. Data are R2 values (P values in brackets) between mean within-species chronosequence stage and mean within-species value for the litter mixing variable. All R2 values are for linear regressions, except when stated as quadratic. ‘n’ = the number of independent data points (i.e. number of different species) for each chronosequence
SequenceLitter mixing effectLitter mixing response
  • Mean effect of litter of given species on decomposition rates of all other species collected from the same plot, and mean response of decomposition rate of given species to all other species from the same plot, in litter-mixing bioassays.

  • All data based on 90 days incubation, except for Cooloola which is based on 180 days incubation. Positive linear and negative linear indicate trends of increasing or decreasing values respectively across the chronosequence; positive quadratic and negative quadratic indicate greater or lesser values respectively at intermediate chronosequence stages relative to early or late stages. Values in bold are statistically significantly different to 0 at P = 0.05.

Cooloola (n = 27)0·315 (0·010)0·001 (0·929)
positive quadratic
Glacier Bay (n = 10)0·217 (0·118)0·400 (0·049)
negative linear
Franz Josef (n = 24)0·311 (0·018)0·325 (0·016)
negative quadraticnegative quadratic
Waitutu (n = 14)0·269 (0·050)0·075 (0·343)
negative linear

When data were averaged within each plot, the litter mixing responses showed the same results as for the litter mixing effects (Fig. 3). At the within species level, the Glacier Bay and Franz Josef sequences showed significant negative linear or quadratic relationships between mixing response and chronosequence stage (Table 5). At the within-species level, only three of the 13 species showed significant relationships between litter mixing response and chronosequence stage. Two of these were from the Cooloola sequence and both showed negative linear relationships, i.e. Banksia integrifolia (R2 = 0·350, P = 0·050, n = 11) and Banksia serrata (R2 = 0·626, P = 0·006, n = 10). The third was Tsuga hererophylla from the Glacier Bay sequence which showed a positive quadratic relationship (R2 = 0·451, P = 0·028, n = 16).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

nutrient properties

Our results from certain chronosequences support previous work (e.g. Vitousek 2004; Wardle et al. 2004) in showing that during retrogression, litter N and P concentrations decline and litter N : P ratios increase. This is consistent with soil nutrient availability declining during retrogression, and with P becoming increasingly unavailable relative to N (Walker & Syers 1976). It is recognized that reductions in the availability of P relative to N occur over time because a substantial proportion of initial P is converted to unavailable forms during pedogenesis (Walker & Syers 1976; Vitousek 2004), and because N can be replenished biologically during retrogression (by biological fixation of atmospheric N) while P cannot (Lagerström et al. 2007).

Four of the six chronosequences each involved significant turnover of species across the sequence (with different species dominating at different stages), and four of the six contained species that occurred across most or all stages of the sequence. Thus, for our data set we were able to explore the role of both among- and within-species variation in influencing the overall shifts in litter nutrients. There were several instances in which among-species variation was important in determining overall changes in litter nutrient properties during retrogression. For example, the decline in litter N and P concentrations during some stages of the Franz Josef sequence (Fig. 1) was due in large part to species with low litter nutrient concentrations dominating at latter stages of the chronosequence (Table 1). Similarly, the overall maximal N : P ratio at the initial and late stages for the Glacier Bay sequence was caused by species with high N : P ratios dominating at these stages. Such examples point to different plant species being adapted for occupying different portions within a given chronosequence (Grime 2001; Cortez et al. 2007), and to species that dominate in different stages of the sequence producing litters of contrasting quality.

There were also several instances for which within-species variation was important. For example, the maximal values for overall N and P concentrations at intermediate chronosequence stages in the Hawaiian sequence (Fig. 1) follow the same pattern as that found for a single species that dominates the entire sequence, i.e. M. polymorpha (Table 2). Similarly, increases in overall N : P ratios for the Hawaii and Cooloola sequences during late successional stages match the increases observed for some of the individual species as succession proceeds. These results show that shifts in overall litter nutrient properties across entire chronosequences can sometimes be driven by phenotypic plasticity of dominant species that occur across large portions of the sequence. This is consistent with the findings of other studies that point to within-species variability in litter quality being an important ecological driver (Madritch & Hunter 2002; Schweitzer et al. 2004; Classen et al. 2007). These within-species differences across the gradient can be due to either plasticity at the within-individual level (Vallanderez et al. 2007) or differences in genotypes across the chronosequence (Treseder & Vitousek 2001).

It is apparent that the role of within-species and among-species variability in determining how retrogression affects litter nutrient properties may vary among chronosequences, and may not always be important. For the two sequences in which among-species and within-species effects could be directly compared, contrasting patterns were observed: for the Cooloola sequence, within-species, but not among-species effects led to the overall increase in the litter N : P ratio, whereas the reverse was true for the Glacier Bay sequence. Further, there was considerable variation among species in their responses to chronosequence stage, even within sequences. Although soil fertility and vegetation biomass and productivity is known to decline during retrogression for each of these sequences (Vitousek & Farrington 1997; Wardle et al. 2004; Coomes et al. 2005), the concentration of litter N and P was unresponsive to retrogression for some species, and even positive responses to retrogression sometimes occurred. Our results are indicative of high variability among species in the degree of phenotypic plasticity that they show (Valladarez et al. 2007). In total, our results reveal that depending on the gradient, overall changes in litter quality across chronosequences can be driven both by those species that have the greatest phenotypic plasticity in litter nutrient concentrations, and by variability among species that occupy different chronosequence stages.

decomposition, nutrient release and litter mixing

We found overall rates of litter decomposition declined during retrogression for three chronosequences and increased for two others (Wardle et al. 2004). However, the overall rate of N and P release from litter either declined during retrogression or was unresponsive. This decline in nutrient release points to greater retention of nutrients by litters collected from sites with greater nutrient limitation, and is consistent with the lower fluxes of nutrients observed in late successional ecosystems (Vitousek 2004; Parfitt et al. 2005). The overall reduction in the rate of decomposition and release of N and P during retrogression was more consistent with the patterns observed at the among-species level than at the within-species level. The effects of among-species variation on these variables was always negative or neutral (Table 3), meaning that those species that dominated in late successional stages often produced litter that broke down more slowly and released less nutrients than those that dominated earlier in succession. In contrast, at the within-species level, a range of responses to retrogression was found (Table 4), and there were several instances in which the rate of mass loss and N and P release actually increased during retrogression. The reason for greater mass loss from litter collected from infertile retrogressed systems is unclear. However, given that synthesis of recalcitrant structural carbohydrates such as lignin requires N (Osunkoya et al. 2007) and is sometimes produced in lower amounts when N is more limiting (Aerts, De Caluwe & Beltman 2003), it may be that litter from retrogressive stages sometimes contained lower concentrations of recalcitrant compounds. In any case, our data indicate that the overall declines in litter decomposition and nutrient release that occur during retrogression (Fig. 2) emerge largely (though not entirely) from variation among species rather than phenotypic variation within species.

Litter mixing frequently had important effects on litter decomposition rates. The overall litter mixing effect and litter mixing response was highly variable and varied from strongly negative to strongly positive depending on the chronosequence and successional stage that the litter was collected from. This is consistent with other studies showing the magnitude and direction of litter mixing effects to be highly context-dependent (Wardle et al. 1997a; Garter & Cardon 2004). Although both overall litter mixing effect and response varied significantly within at least half of the chronosequences, very different patterns were found for different sequences (Fig. 3). These differences both among chronosequences and among stages within sequences must be due to variations in litter quality, although there is no simple explanation as to how litter quality may have determined these differences. This is consistent with several previous studies that have found variable relationships between litter quality and litter mixing effects (Wardle et al. 1997a; Hoorens et al. 2003; Quested et al. 2005; Liu et al. 2007). Despite this apparent unpredictability, the magnitude of variation in litter mixing effects and responses across stages within some sequences should be sufficient to cause important among-stage differences in overall litter decomposition rates. However, one limitation of this (and indeed most) litter mixing studies is that they involve addition of equal amounts of all litter types, and when litters are mixed in the proportions that they occur in nature then mixing effects may be considerably weaker (Dearden et al. 2006).

The effects of chronosequence stage on mixing effects and mixing responses appear to be largely due to among-species rather than within-species variability. For five of the seven instances in which overall litter mixing effects or responses were significantly related to chronosequence stage, we found similar relationships between mixing interactions and chronosequence stage at the among-species level. In contrast, at the within-species level, there were few instances in which litter mixing interactions varied according to chronosequence stage. It is therefore apparent that the large overall variations we observed for litter mixing interactions between stages within chronosequences are due largely to different species with contrasting litter qualities occurring at different stages, rather than because of within-species variability in litter quality. This means that different species that are adapted for occupying different portions of a given chronosequence will produce litters that differ not only in litter quality and decomposition rates (Grime 2001; Wardle et al. 2004; Cortez et al. 2007), but also in their interactions with litters of coexisting species.

conclusions

We were able to show that several measures of litter quality and decomposability varied across chronosequence stages, and develop insights as to whether this variation could be explained best by variation among- or within-species across the sequences. Litter N and P concentrations often declined with increasing chronosequence stage both among- and within-species. Meanwhile, loss rates of litter mass, N, and P during decomposition frequently declined with increasing chronosequence stage among species, but less often within species. Litter mixing effects varied across chronosequence stages but in a less predictable manner; this variation was mainly apparent at the among- rather than within-species level. As such, among-species variation is the main driver of the overall response of the litter properties that we measured to chronosequence stage (and therefore soil fertility), except for litter nutrient concentrations which can be driven by both among- and within-species variation. This points to among-species variation as a consistently important driver of ecosystem properties across ecological gradients (Binkley & Giardana 1998), and to within-species variation (Madritch & Hunter 2002; Classen et al. 2007) only being important for some variables and only in some instances.

By considering not only trends within chronosequences but also different sequences, we were able to determine whether consistent responses to long-term chronosequence development emerged for vastly contrasting locations (Wardle et al. 2004, 2008). Although there were some similarities in results across the chronosequences there were also important differences. Some sequences involve significant replacement of species across chronosequence stages (notably Franz Josef and Waitutu) while others involve the same dominant species occupying all stages (notably Hawaii and Arjeplog). It is noteworthy that the overall responses sevenof decomposition-related variables to chronosequence stage (Figs 2, 3) was often weaker or less consistent in direction for the sequences for which there was little species turnover across stages (Arjeplog and Hawaii) than for the others. Our results provide evidence that decomposition, litter nutrient release, and the strength of litter mixing effects, are all most likely to be responsive to gradients of soil fertility (such as occurs across long-term chronosequences) when there is significant turnover of species across the gradient. They also highlight that understanding the role of species in driving ecosystem properties and processes is essential for enhancing our knowledge of how ecosystems change during succession and retrogression.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

For assistance in accessing field sites and sampling authors thank: C. H. Thompson and J. Walker (Cooloola sequence), L. Sharman, M. Kravolec, G.P. Streveler, and P. Haygarth (Alaska sequence), H. Farrington and P. M. Vitousek (Hawaii sequence), and D. A. Coomes (Waitutu sequence). The authors also thank G. Rattray for technical assistance, and two anonymous referees and the handling editor for helpful comments.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Appendix S1. Ages and dominant plant species for each stage of each of the six chronosequences. For further information see Wardle et al. (2004, 2008).

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

FilenameFormatSizeDescription
FEC_1513_sm_AppendixS1.pdf38KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.