SEARCH

SEARCH BY CITATION

Keywords:

  • abiotic stress;
  • environmental adversity;
  • Eucalyptus;
  • meta-analysis;
  • Nothofagus;
  • physical stress;
  • resource availability;
  • the habitat templet

Summary

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

1. Our aim was to develop a quantitative proxy for environmental adversity (abiotic stress) in temperate Eucalyptus and Nothofagus forest and woodland ecosystems.

2. Samples and measurements were collected at 42 sites across a rainfall gradient in southern Australia, an elevation gradient in south-eastern Australia, and a longitudinal transect (temperature gradient) in Patagonia, Argentina.

3. We compared the ability of (a) abiotic variables (14 soil and 21 climatic variables) and (b) the stable carbon isotope (δ13C) values of soil organic matter (SOM), to predict variation in leaf area index (LAI; a forest productivity variable).

4. The δ13C of SOM (soil aggregates) explained more variation (57%) in LAI than multivariate statistical models that integrated information on many abiotic variables. W* (a climatic water balance model) was also a powerful predictor variable, explaining 37% of the variability in LAI.

5Synthesis. The stable carbon isotopic signature of soil aggregates is a powerful explanatory variable that may help us to quantify environmental adversity (abiotic stress) in temperate forest and woodland ecosystems.


Introduction

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

Spatial and temporal heterogeneity in the abundance of resources and in the physical conditions for life can have a profound effect on plant and ecosystem function and is an important part of the explanation for the diversity of life and species coexistence (Tilman 1982; Silvertown 2004; Facelli et al. 2005). A large number of abiotic factors (and stresses) can directly impact plant growth and ecosystem production through effects on resource availability and/or uptake. Examples include, but are not limited to: the concentration of nitrogen, phosphorus, potassium and micro-nutrients in the soil; the pH, salinity, texture and depth of the soil; attributes of the landscape (i.e. topography), rainfall, temperature and the quantity of photosynthetically active radiation.

Such abiotic complexity has been a major barrier to synthesis and progress in ecology. Our observation is certainly no revelation. In the 1960s, Charles Elton suggested that a lack of definitions of habitat were a ‘chief blind spot’ in ecology (Elton 1966 cited in Southwood 1977). The fact that habitat characteristics are difficult to quantify and the subsequent problems this has created are exemplified by four leading conceptual models in ecology: the Habitat Templet (Southwood 1977) (∼950 citations; Fig 1a), the C-S-R model (Grime 1977) (∼1400 citations; Fig 1b), Bertness & Callaway’s (1994) Stress Gradient Hypothesis (∼550 citations; Fig 1c) and the Growth Differentiation Balance Hypothesis (Herms & Mattson 1992) (∼1000 citations; Fig 1d). Central to each of these models is a prediction about how biotic interactions may change as the physical environment becomes more adverse. In this study, we follow Southwood (1988) and define environmental adversity in its most general sense as the external constraints on dry matter production. For higher plants, environmental adversity may be due to a low availability of resources (e.g. water, nitrogen), and/or due to inclement physical conditions (e.g. extremes of temperature), that is conditions which ‘impose a particular cost for the maintenance of normal protoplasmic homeostasis’ (Southwood 1988).

image

Figure 1.  Conceptual models relating ecological and environmental factors. Redrawn from: (a) Southwood (1988) (b) Grime (1977) (c) Bertness & Callaway (1994) (d) Herms & Mattson (1992). Environmental adversity is a central component in each of these models and is referred to as: (a) environmental adversity, (b) abiotic stress, (c) physical stress and (d) resource availability. Note that this habitat attribute is never quantified, with the exception of Grime (1977), who uses an arbitrary scale (from zero to one).

Download figure to PowerPoint

A second feature the four models in Fig. 1 have in common is that environmental adversity is always defined qualitatively. The fact that we are, at present, limited to qualitative definitions of adversity has precluded the possibility of synthesizing the wealth of empirical data relevant to each of these models. The heated debate on competition intensity, for example, and how this should be affected by environmental adversity could have, almost certainly, been resolved if we possessed a definition of environmental adversity that allowed comparisons across gradients and studies (Keddy 1994). More recently, researchers have been measuring facilitation (positive species interactions) across adversity gradients (see Lortie & Callaway 2006). However, as was the case in studies of competition, it has been impossible to compare the level of adversity across these facilitation studies. Meta-analyses of studies on facilitation have therefore been restricted to the use of ‘two undefined and arbitrary classifications of stress’ (Lortie & Callaway 2006). Being able to define environmental adversity in a way that is meaningful across studies should enable robust comparisons across experiments, and therefore the synthesis (meta-analysis) of ecological data.

In this study, we measure a large number of abiotic variables in an attempt to quantify environmental adversity for temperate forest ecosystems. We also hypothesize that the isotopic composition (δ13C) of soil organic matter (SOM) may prove useful as a continuous and convenient variable for describing environmental adversity in temperate forest ecosystems because it integrates adversity effects over time and reflects long-term site conditions. The δ13C composition of plant tissues is affected by a wide range of environmental variables (i.e. rainfall, temperature, soil water content, irradiance and soil nitrogen availability) (Song et al. 2008) and by the photosynthetic pathway of the relevant vegetation (Ehleringer et al. 2000). C4 plants (e.g. grasses from semi-arid and tropical regions) possess a δ13C isotope signature (−15‰ to −9‰) distinct from that of C3 plants which form the great majority of the planet’s vegetation (approximately −21‰ to −40‰) (Bender 1968; Smith & Epstein 1971; Staddon 2004). For sites strongly dominated by C3 plants, as those in this study, compositional effects (the relative abundance of C3 vs. C4 species) are negligible compared with photosynthetic discrimination.

Photosynthetic discrimination against 13CO2 in C3 plants is related to the pi/pa ratio, the internal partial pressure of CO2 (pi) relative to that in the atmosphere (pa) (Farquhar et al. 1989). This ratio is negatively related to intrinsic water use efficiency A/g, the rate of CO2 assimilation (A) relative to stomatal conductance (g) (Korner et al. 1991). Factors affecting either A or g disproportionately to the other in C3 plants will influence the pi/pa ratio and thus photosynthetic discrimination against 13CO2. For instance, stomatal constraints associated with soil moisture deficits are likely to be lower in wetter habitats (Farquhar et al. 1989). In wetter habitats, g is therefore more likely to operate nearer to its maximum more often than in drier habitats, reducing A/g and thereby promoting more negative values of δ13C in plant matter. In habitats where water is in adequate supply, the physical environment will also tend to support a higher leaf area index (LAI, m2 leaf area m−2 ground cover) and forest productivity and therefore we expect δ13C of SOM (δ13CSOM) and forest productivity to be correlated.

The density of the forest canopy per se, for which LAI is a reliable proxy (Macfarlane et al. 2007a), may also affect the δ13C composition of vegetation biomass in two key ways. Relatively high values of LAI will result in increased humidity (a lower vapour pressure deficit) for a large number of the individual leaves within the forest canopy (McNaughton & Jarvis 1983). This will again favour g operating nearer to its maximum more of the time, and thus decrease A/g. Greater light extinction through a dense (high LAI) canopy will also occur, and will result in shading for a greater proportion of the canopy, i.e. those individual leaves not positioned in the outer canopy. Thus, although more photosynthesis may be occurring at the canopy scale, due to a greater LAI, a large portion of the canopy will tend to operate at lower levels of photosynthesis, and thus A, than is occurring in the sun-exposed leaves positioned in the outer canopy. Such ‘canopy effects’ on discrimination (Ometto et al. 2006; Drucker et al. 2008) may be small for low-productivity woodland ecosystems (low LAI) compared with more productive forest ecosystems. Thus photosynthetic discrimination against the heavier stable carbon isotope (13C) may reflect physiological functioning (A/g) as it is affected by a diverse range of abiotic factors (i.e. any factor that affects the productivity and density of the forest canopy). These effects will be recorded in the isotopic composition of the source vegetation and, because of plant litter deposition over time, in the SOM (Nadelhoffer & Fry 1988; Ehleringer et al. 2000).

Based on the above we predict that the δ13C values of SOM will be a useful variable for quantifying environmental adversity in C3-dominated forest and woodland ecosystems. We specifically test the following hypotheses: (i) δ13C values of SOM will correlate strongly with LAI, a key ecosystem productivity variable, in temperate Eucalyptus and Nothofagus forest and woodland ecosystems; and (ii) δ13C of the SOM, as an integrator of environmental information, will outperform the abiotic variables (climatic and soil fertility parameters) used in this study, singularly or in combination, as a predictor of variation in forest productivity (i.e. LAI).

Materials and methods

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

Study sites

We measured 35 different abiotic variables at 42 locations (22 sites in Australia and 20 in Southern Argentina) (see Table S1 in Supporting information). At each sampling location, plant, soil and ecosystem properties were measured in a 20 × 40 m quadrat. Data collection in Australian Eucalyptus forests took place in March 2007 and was conducted across a rainfall gradient in southern Australia and over an elevation gradient in south-eastern Australia. Nothofagus forests in Patagonia were sampled in February 2007 (at the peak of the austral summer), and sampling was conducted across a latitudinal (temperature gradient) and longitudinal transects (rainfall gradients) (see Table S1 and Fig. S1). We sampled forest and woodland communities that occupied a diverse range of topographic positions within the landscape. In this way we maximized the variability in the physiognomic composition of the plant communities sampled. The average height of mature trees, for example, ranged from 2 to 40 m across the three transects (see Table S1). All sampling sites were either forest or woodland communities dominated by C3 species and all sites were undisturbed for at least 30 years. The degree of C3 domination was almost complete; there are no C4 grasses in the south of Argentina and at our semi-arid woodland sites in southern Australia C4 grasses (e.g. Themeda spp.) were present but were a very minor component of the understorey (<1%) at the time of data collection. As a result, the δ13C values measured in SOM in this study reflect the δ13C signal of C3 plants (Staddon 2004). Records covering more than 30 years, established and kept by the National Parks Service in Australia and Argentina, were used to identify sites that had been undisturbed for at least 30 years. In Argentina, sites with a long absence of disturbance were also identified with tree ring data collected during a previous study (Peri et al. 2006).

Plant community characterization

Leaf area index was measured using a digital photography method. For each sampling quadrat, we took digital photographs of the forest canopy at 20 points randomly located throughout that quadrat. LAI values were calculated using a modified version of the Beer-Lambert light extinction coefficient (equation 3 of Macfarlane et al. 2007a). A detailed description of the method can be found in Macfarlane et al. (2007a). The digital photography method yields estimates of LAI similar to measurements made with a Licor LAI-2000 canopy analyzer (Licor, Nebraska, USA) (Macfarlane et al. 2007b). LAI, measured with the digital photography method, also scaled closely to key functional traits, such as the height of mature trees (Table S2). A digital clinometre (Haglöf, Långsele, Sweden) was used to measure the height of the 10 largest trees within each LAI quadrat.

Climatic variables

Climatic parameters for each site were derived from the WorldClim data set (Hijmans et al. 2005; http://www.worldclim.com). WorldClim contains geographic surfaces for 19 different climatic parameters that describe rainfall, temperature and their variability. Incoming solar radiation (Wh m−2) was calculated using the Solar Radiation tool in ArcGIS version 9.2 (ESRI, California, USA), with topography data from the NASA Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM) of the globe (http://www2.jpl.nasa.gov/srtm). We also calculated two composite climatic variables (W* and E-T) for use in statistical analyses because they integrate climatic conditions highly relevant to plant growth. W* represents mean annual water availability (Wynn et al. 2006) (eqn 1), while E-T is the climatic index derived by Liski et al. (2003) (eqn 2).

  • image( eqn 1)

where MAP is mean annual precipitation (mm year−1), Q is mean annual global solar radiation (J m−2 year−1), r is the density of liquid water at 25 °C (1000 kg m−3), and L is the latent heat of evaporation of water at 25 °C (2.5 × 106 J kg−1 H2O).

  • image( eqn 2)

where MAT is mean annual temperature (ºC) and D is mean annual precipitation minus mean annual potential evapotranspiration (mm).

Soil sampling and analysis

Replicate soil samples were collected from 0 to10 cm depth, using a hand auger, at nine random points within each 20 × 40 m quadrat used for LAI measurement. To reduce the number of geochemical analyses we pooled individual soil samples into bulk samples (after Bird et al. 2004). From the nine samples collected within each quadrat we created three composite samples so that each composite sample contained an equal proportion of soil from three auger holes (= 3 for each site). Soil samples were stored in airtight bags, and posted back to the laboratories periodically during the field campaigns. During field work, we collected soil samples up to a maximum sampling depth of 90 cm, the limit of our hand auger. The samples below 10 cm have been archived at the University of New South Wales, but have not been chemically analysed. Shallow soil can be an important contributing factor to site water balance, with subsequent effects on vegetation distribution (Stephenson 1998). Thus we include soil depth in our statistical analyses, using a default value of 90 cm when soil depth exceeded the maximum sampling depth.

Upon return to the laboratory, soil samples were oven-dried at 40 °C to a constant weight and then stored in the laboratory. The processing of samples for analyses began in May 2007, and the analyses were completed by the end of October 2007. Soil aggregates were extracted from the 10-cm depth samples and finely ground to below 2 μm using a tungsten-carbide mill (= 3 for each site). The ground aggregates were used to measure the percentage carbon and nitrogen (elemental C & N) using a LECO auto-analyzer (St. Joseph, Michigan, USA) at the Australian Nuclear Science and Technology Organisation (ANSTO). After determination of elemental C and N in the 10-cm aggregates, a portion of homogenized sample was used to measure the isotopic composition (δ13C) of the sample. The ratio of heavy to light isotopes in the sample material (Rsample) was measured using mass spectrometry as the deviation from the isotopic ratio of a standard (Rstd); where R denotes the ratio of stable carbon (13C/12C) isotopes. These measurements are defined as: δ = 103(Rsample - Rstd) /Rstd; where Rstd is the value of the Pee Dee Belemnite (PDB) standard. δ13C was measured at the Stable Isotope Facility of the Research School of Biological Sciences at the Australian National University.

Major cations (Na, Al, P, K, Ca) and effective cation exchange capacity (eCEC) were measured for the 10-cm soil samples, after the samples had been passed through a 2-mm sieve. Standard analytical techniques were used, and the analyses were performed by a commercial laboratory (CSBP Ltd., Western Australia). Available phosphorus was measured using the Colwell method (Rayment & Higginson 1992), in which soils are tumbled with a 0.5-M sodium bicarbonate solution (adjusted to pH 8.5) for 16 h at 25 °C employing a soil:solution ratio of 1:100. The acidified extract was then treated with an ammonium molybdate–antimony trichloride reagent and the concentration of phosphorus measured colorimetrically at 880 nm. Exchangeable cations (Ca, Mg, Na, K) were measured using the Gilman and Sumpter method (Rayment & Higginson 1992) which employs a 0.1 M BaCl2 and 0.1 M NH4Cl extraction of exchangeable bases. Extractable aluminium was measured by extraction with a 0.01-M calcium chloride solution and subsequent analysis of the extract by a colorimetric method using a catechol violet reagent. The effective cation exchange capacity is the sum of the measured cations. The pH of soil samples was determined with an electronic meter immersed in a 1:5 mixture of soil and deionized water. The percentages of clay, silt and sand in each sample were determined using a Malvern Mastersizer 2600 laser particle size analyzer (Malvern Instruments Ltd., Worcestershire, UK).

Statistical analysis

The normality of independent variables was assessed using JMP version 5 (Cary, North Carolina, USA). Potential nonlinear univariate relationships between LAI and the independent variables were also assessed using JMP. Nine soil variables were subsequently log-transformed (after Molofsky et al. 2000): percentage nitrogen, the C:N ratio, phosphorus, exchangeable calcium, magnesium, sodium, potassium and aluminium, and the eCEC. Nine climatic variables were also log-transformed: mean temperature in the driest quarter, mean temperature in the warmest quarter, mean annual precipitation, precipitation in the wettest month, precipitation in the driest month, precipitation seasonality, precipitation in the driest quarter, precipitation in the warmest quarter and precipitation in the coldest quarter (see Table S2 for further description of the climate variables). The interactive effects of the 35 independent variables on LAI were analysed by principal components analysis (PCA) on the complete data set (Patagonia and Australia combined) (see Appendix S1). Following PCA, we used S-Plus (Insightful Software, Seattle, Washington, USA) to perform multiple linear regression to assess the interactive effects of the principal components on LAI. Following Quinn & Keough (2002), only those principal components with eigenvalues >1 were included in the multiple regression model. The percentage of variation accounted for by each principal component within the regression model was determined by expressing the sum of squares for each component as a percentage of the total sum of squares, a statistic described as the η2 value (Plaistow et al. 2006). Finally, we used a negative exponential regression model to assess the univariate relationship between LAI and δ13C.

Results

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

Seven principal components had eigenvalues >1 (see Appendix S1). The multiple linear regression in which these seven principal components were regressed against LAI show that the seven principal components explain 53% of the variance in LAI across continents (Table 1, Fig. 2). The η2 statistic indicated that principal components 1, 2 and 5 were the most important determinates of variation in LAI (Table 1). For Principal Component 1 (PC1), a range of soil fertility (5 variables) and climate (7 variables) variables had factor loading scores >0.7 (Appendix S1), indicating that they were strongly correlated with this principal component (Quinn & Keough 2002). PC2 was mainly associated with climatic variables, and the most important of these was W*, the climatic water balance model (Appendix S1). The percentage sand, the percentage silt and Temperature Seasonality (Bio 4) were the independent variables most closely associated with PC5 (Appendix S1). The δ13C of the soil aggregates explained (in the statistical sense) 57% of the variance in measured LAI; y = 0.0002e0.3358x, r2 = 0. 57; where x is the δ13C of soil aggregates (Fig. 3). The univariate relationship between W* and LAI is also noteworthy. There were two outliers for W* vs. LAI (Fig. 4), both from alpine Australia. When the alpine data were excluded, W* performed nearly as well as δ13C as an independent variable using a linear (y = 0.0021x - 4.3117, r2 = 0.51) or a nonlinear (y = 0.0171e0.0015x, r2 = 0.48) model.

Table 1.   Summary statistics from the multiple-regression on principal components (PC) that had eigenvalues >1
Source of variationd.f.SSF-valuePr  > Fη2
  1. Patagonian and Australian data combined, Model, LAI (Leaf Area Index) = 1.63 + 7.14PC1 + 0.14PC2 - 5.88PC3 - 1.63 PC4 + 0.17PC5 + 9.09PC6 - 9.68PC7. The multiple r2 is 0.53, the F-statistic is 5.488 on 7 and 34 degrees of freedom and the P-value is 0.00027. SS = sum of Squares, Pr = probability, and η2 values indicate the relative importance (in percentage terms) of each independent variable in explaining variation in the response variable (LAI) (see Plaistow et al. 2006).

PC112.316.160.0188.51
PC217.13219.050.000126.3
PC310.531.4130.2431.95
PC410.000.0010.9760.001
PC513.318.850.00512.2
PC610.681.810.1872.5
PC710.411.100.3011.5
Error3412.73   
image

Figure 2.  Modelled versus observed leaf area index (LAI) (mean values ±SE) for Australian and Patagonian data combined. The regression model combined the first seven principal components. See Table 1 for further details on the statistical model.

Download figure to PowerPoint

image

Figure 3.  δ13C vs. LAI (leaf area index); increasing (less negative) δ13C values indicate that the relevant sample [in this case soil organic matter (SOM)] contains more 13C molecules and more closely approximates the carbon isotope composition of the Pee Dee Belemnite (PDB) standard.

Download figure to PowerPoint

image

Figure 4.  Leaf Area Index (LAI) vs. W*, a climatic index that incorporates information on rainfall and radiation thus reflecting water balance (Wynn et al. 2006).

Download figure to PowerPoint

Discussion

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

Measurements of natural isotope abundance have been suggested as a convenient, continuous and integrated metric for quantifying the breadth of the trophic niche in animals (Bearhop et al. 2004). In a similar vein, the results reported here suggest that the isotopic signature of SOM (δ13CSOM) could be an important quantitative variable for describing environmental adversity in plant communities. Throughout this study we have assumed that LAI (m2 of forest canopy per m2 of ground) would be optimized in undisturbed forest and woodland ecosystems to maximize productivity according to the availability of resources. The LAI determines how much photosynthetically active radiation can be absorbed (Williams & Rastetter 1999). Further, the uptake of nutrients in higher plants is closely associated with transpiration (McDonald et al. 2002), and transpiration, in turn, relates to the evaporative surface area (LAI) of the canopy (Eamus et al. 2006). Therefore, a high LAI is expected to result in high productivity, and potentially high carbon sequestration relative to a low LAI (Street et al. 2007). Furthermore, LAI in this study was strongly correlated to key functional traits which define plant productivity, such as tree height (Table S2), which has long been used by foresters as a site quality and/or productivity index (Koch et al. 2004; Peri et al. 2006).

What then do we gain from measuring the isotopic composition (δ13C) of SOM that could not be ascertained from direct measurements of LAI? The LAI returns limited information about ecosystem parameters such as environmental adversity since any disturbance will act to lower the LAI, but not the potential for that habitat to produce plant biomass. In contrast, the δ13CSOM signal in soil aggregates can provide information about habitat potential in spite of short term variations in disturbance history and/or in environmental conditions (e.g. drought). Soil aggregates are a passive, relatively inert form of soil carbon (Brodowski et al. 2006; Schwendenmann & Pendall 2006; Skjemstad et al. 2008). δ13CSOM reflects the (average) contribution from the source vegetation over many years, it is static from year to year, and it can provide a temporally integrated measure of ecosystem properties and/or functioning (Ometto et al. 2006).

This study was restricted to temperate regions. However, research in tropical rainforest (Williams et al. 2002) suggests that the relationship between LAI and δ13CSOM might be robust for a wider range of habitats. Williams et al. (2002) report that tropical rainforest LAI values ranged between 5 and 7, and δ13C values in leaf litter between −29.28 and −31.14, which suggests that the capacity for δ13CSOM to describe environmental adversity extends beyond temperate forests and woodlands. It is important to note that the comparison between leaf litter and SOM is reasonable because leaf litter is the dominant input into SOM and has a δ13C signal representative of other plant tissues (wood and fine roots). We expect also that additional data from environmentally adverse tropical ecosystems will further strengthen the relationship between LAI and δ13C. Recent work in Brazilian savanna (Silva et al. 2008) supports this prediction. Environmental adversity in the tropics is generally attributable to seasonal drought, intense radiation, frequent fires or, commonly, a combination of these factors. A common ecosystem type where these conditions prevail is the savanna. In tropical savanna, grasses with C4 photosynthesis co-dominate with woody (C3) species. Savanna plant communities with a significant C4 grass component are likely to have LAI values at the lower end of the range reported in Fig. 4 (Asner et al. 2003), and C4 plants have less negative δ13C values, which results in carbon isotope values in SOM that are closer to zero (less negative) (Wynn & Bird 2007). However, in the tropics primary production (Williams et al. 2002) and SOM (Ometto et al. 2006) are both highly variable. Therefore, more tropical ecosystems must be studied to see whether the relationship between LAI and δ13C in soil aggregates can be extended to tropical ecosystems more generally.

We expect also that additional sampling in tundra ecosystems will further strengthen the predictive relationship between LAI and δ13CSOM. LAI values in alpine, tundra and/or steppe ecosystems are at the lower end of the range presented in Fig. 3 (van Wijk & Williams 2005), and reported values of δ13CSOM in such communities types [∼−24‰] (Huber et al. 2007) tend to fall at the less negative end of the δ13C range reported here (Fig. 3). Finally, recent work by Neill et al. (2008) on the physiology of stress tolerance provides additional and independent evidence that δ13CSOM may be useful for describing environmental adversity in a wider range of habitats. Many different abiotic stresses (e.g. low temperature, salt stress, anaerobic conditions [flooding], mineral nutrient imbalance and drought) cause stomata to close (Neill et al. 2008). This in turn affects the isotopic composition of plant tissues (Farquhar et al. 1989), and, ultimately, the δ13C of SOM (Ehleringer et al. 2000).

In our study, δ13C, as an integrator of environmental information, was a better predictor of LAI than multivariate statistical models that included many abiotic variables. There are many possible reasons why the abiotic variables measured in this study were not powerful predictors of LAI, either individually, or collectively in multivariate analyses (see also Leuschner et al. 2006). Soils, for instance, are inherently complex and form in response to climatic, geologic, hydrologic and biotic influences. Soil fertility, in particular, is a notoriously difficult variable to quantify (Neatrour et al. 2008). It is entirely possible that soil fertility variables that are more sophisticated than total static nutrient content (e.g. Adams & Attiwill 1986; Turner 2008) could, in combination with δ13CSOM, allow for more precise estimates of environmental adversity across ecosystems. This may be of particular relevance for Australia, where soils are old and infertile by global standards (Turner 2008) and where the correlation between δ13C and LAI was weakest (Table S2).

Climatic variables were generally better predictors of LAI than soil variables (Table S2), though still inferior to δ13C. The one possible exception to this was W*, a climatic water balance index (Stephenson 1998; Wynn et al. 2006). W* combines information on rainfall and radiation (Wynn et al. 2006), and compared with other abiotic variables it was a relatively powerful predictive variable, except with the data from alpine Australia. This exception may be because snow was not differentiated from rainfall in the records from which the precipitation dataset was developed (Stephenson 1998). Stand age could also be a contributing factor; forest biomass and LAI change during succession, increasing linearly in the early stages of succession, and then beginning an asymptotic decline as forest stands age (Peri et al. 2006). W* is a potentially convenient variable that can be calculated from existing data sources (http://www.worldclim.com & http://www2.jpl.nasa.gov/srtm) and should be considered in any future analyses.

Conclusions

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

The results from this study support the hypothesis that δ13C in soil aggregates correlate strongly with the key ecosystem productivity variable, LAI, in temperate Eucalyptus and Nothofagus forest and woodland ecosystems. δ13C in soil aggregates could therefore be an integral component of a quantifiable definition of environmental adversity in C3-dominated forest and woodland ecosystems. Support for the second hypothesis was less conclusive; although δ13C was a better predictor of LAI than multivariate models that included several abiotic variables, climatic water balance (W*) warrants further investigation, and to increase precision it will also be important to incorporate an adequate approach to quantifying soil fertility (Neatrour et al. 2008; Turner 2008). A quantitative proxy for environmental adversity should greatly facilitate the quantitative meta-analysis (synthesis) of ecological data (Lortie & Callaway 2006). Further research is now required to represent soil fertility more precisely and determine whether the correlation between LAI and δ13CSOM and/or W* is robust in a wider diversity of habitats.

Acknowledgements

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

We are grateful to the City of Onkaparinga who helped fund the analytical work. AINSEE (grant # 4013) provided 24 analyses of δ13C which were used in a pilot for this study, P.L.P. was supported by INTA Conicet, S.P.B. was supported by a UNSW vice-chancellor’s research and teaching fellowship and a Faculty Research Grant, J.R.L. was supported by a UOW postgraduate award and an ANSTO scholarship, D.A.P. acknowledges grant support from the Australian Research Council and the Australian Department of Climate Change. Thanks to National Parks in South Australia and New South Wales, and in particular to D Woods and M Buekers for freely sharing their impressive knowledge of the natural history of SE-NSW. We thank also Richard Field, Wulf Amelung, the anonymous reviewers and the Editorial Team at the Journal of Ecology for critiquing early versions of the MS.

References

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

Supporting Information

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

Table S1. Basic site information.

Table S2. Description of the independent variables and Pearson correlation coefficients.

Appendix S1. Statistical summary of the principal components analysis.

Figure S1. Location of sites in this study.

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
JEC_1530_sm_Appendix S1.xls1157KSupporting info item
JEC_1530_sm_Fig. S1.doc209KSupporting info item
JEC_1530_sm_Table S1.xls19KSupporting info item
JEC_1530_sm_Table S2.xls29KSupporting 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.