Differential facilitation by a nitrogen-fixing shrub during primary succession influences relative performance of canopy tree species


  • P. J. Bellingham,

    Corresponding author
    1. Landcare Research, PO Box 69, Lincoln 8152, New Zealand,
      Peter Bellingham, Landcare Research, PO Box 69, Lincoln 8152, New Zealand (tel. + 64 3325 6700, fax + 64 3325 2418, e-mail bellinghamp@landcare.cri.nz).
    Search for more papers by this author
  • L. R. Walker,

    1. Landcare Research, PO Box 69, Lincoln 8152, New Zealand,
    2. Department of Biological Sciences, Box 454004, University of Nevada, Las Vegas, 4505 Maryland Parkway, Las Vegas, Nevada 89154–4004, USA
    Search for more papers by this author
  • D. A. Wardle

    1. Landcare Research, PO Box 69, Lincoln 8152, New Zealand,
    Search for more papers by this author
    • Present address: Department of Plant and Animal Sciences, University of Sheffield, Sheffield S10 2TN, UK.

Peter Bellingham, Landcare Research, PO Box 69, Lincoln 8152, New Zealand (tel. + 64 3325 6700, fax + 64 3325 2418, e-mail bellinghamp@landcare.cri.nz).


  • 1 The facilitative and inhibitory effects of a nitrogen-fixing shrub, Carmichaelia odorata, during primary succession were studied using both field measurements in a New Zealand temperate montane valley, and manipulative glasshouse experiments on seedlings of the three dominant tree species, Griselinia littoralis, Metrosideros umbellata and Weinmannia racemosa.
  • 2During a stand development chronosequence of < 100 years in which Carmichaelia colonized, dominated and senesced, there was significant development of soil organic horizons and a large build-up of soil nitrogen, especially in the organic horizon. Soil organic matter and nitrogen levels across the sequence were strongly correlated with the main DCA axis of vascular plant species composition, along which there was change in dominance from herbaceous to woody species. Vegetation increased in height and light levels declined with stand development.
  • 3Similar responses to shade that mimicked that in mature Carmichaelia stands suggested that inhibitory effects were likely to be uniform across the three tree species.
  • 4Nitrogen, either added via Carmichaelia litter or in solution, enhanced shoot biomass and foliar nitrogen concentrations of all three tree species. Growth in soils of increasing development increased foliar nitrogen concentrations for Griselinia and Weinmannia, but not Metrosideros.
  • 5Overall, Metrosideros was the least responsive to potential facilitative effects of Carmichaelia, and Griselinia exhibited the highest degree of plasticity of response. Future forest composition and spatial patterning of species in mixed stands here, as elsewhere, is likely to result from differential facilitative responses during early primary succession.


Much remains unknown about how initial interactions during succession drive the subsequent development of vegetation and composition of plant communities. The biotic interactions that influence successional development are facilitation, tolerance and inhibition (Connell & Slatyer 1977). Facilitation can occur by several mechanisms, including resource and substrate modification and protection from herbivores (Callaway 1995). In primary successions, plants that are early colonists may facilitate growth of other plants that establish concurrently or subsequently, for instance by improving soil nutrient status through colonization by nitrogen-fixing plants (e.g. Vitousek & Walker 1989; Fastie 1995), or by improving fertility and organic matter (Adachi et al. 1996). Inhibition may be caused by competition for nutrients and light by the resident vegetation (Tilman 1988). Facilitation and inhibition often act simultaneously during succession (Chapin et al. 1994; Callaway & Walker 1997) and the balance of these processes is thought to be a major determinant of subsequent vegetation composition (Huston & Smith 1987; Pickett et al. 1987; Walker & Chapin 1987).

Many investigations of how interactions between plant species influence succession have been conducted in relatively simple systems, i.e. between an early colonist species and a single tree species (e.g. Chapin et al. 1994). However, many forests that develop during primary succession are multispecies systems (e.g. Tagawa 1964; Whittaker et al. 1989). If processes of facilitation and inhibition do not act uniformly on different colonizing tree species (Callaway 1998), this is likely to have long-term consequences on future vegetation development and forest composition (e.g. Vitousek & Walker 1989; Chapin et al. 1997).

We investigated interactions between plant species during primary succession (i.e. on substrates that entirely lack prior soil development, Walker 1999) and their consequences for development of species-rich forests. Our study sites on river terraces and landslides in a montane valley in New Zealand are colonized by a nodulated, nitrogen-fixing, thicket-forming leguminous shrub, Carmichaelia odorata (sensuHeenan 1996; hereafter Carmichaelia), at the same time as three long-lived tree species that dominate mature forests in the study area. We conducted field measurements to determine how resources that are likely to be limiting during succession, i.e. light and nutrients (principally nitrogen), vary across a chronosequence of stand development. We manipulated light and nutrients to simulate field measurements, and measured their effects on growth and allocation of seedlings of the three tree species in controlled glasshouse experiments. We tested two hypotheses:

The ultimate goal of this study was to understand better how the relative importance of inhibition and facilitation during early succession may contribute to determining forest composition in the longer term.


Site description

The field work was conducted during February 1999 in the Kokatahi Valley in the Hokitika River catchment of the South Island of New Zealand (42°57′ S, 171°14′ E, 410–500 m a.s.l.). The main geological substrate of the area is foliated schist. The study area is in an area of rapid geological uplift (> 10 mm year−1) along the boundary between the Indian and Pacific Plates (Whitehouse 1988), with high rates of erosion and periodic large earthquakes (Bull & Cooper 1986; Wells et al. 1999). Rainfall is estimated at c. 7000 mm (Griffiths & McSaveney 1983), with much occurring as high-intensity events (Whitehouse 1985), and the average annual temperature is c. 10 °C.

Early successional vegetation in these valleys following landslides, flooding and earthquakes typically proceeds from mosses, lichens, small herbs (e.g. Raoulia spp., Epilobium spp.) and grasses to colonization by shrubs including Carmichaelia odorata, Hebe salicifolia and Olearia spp. (Wardle 1980; Reif & Allen 1988; nomenclature for all species except Carmichaelia follows Allan 1961). Mature forests in the valley are cool-temperate montane rain forests dominated by evergreen angiosperms (principally Weinmannia racemosa and Metrosideros umbellata) with some conifers (principally Podocarpus hallii).

We examined the effects of Carmichaelia on soil characteristics and documented associated vegetation in four successional stages. The stages represented sites where Carmichaelia was just beginning to establish (‘open’: at least one Carmichaelia individual per 50 m2), where it was coalescing (‘young’: 20–40 individuals per 50 m2), where it formed a dense stand (‘vigorous’: > 50 individuals per 50 m2), and where it was present but senescent (‘mature’: at least one senescent individual per 50 m2, < 50 total individuals) (see Table 1 for details). The maximum age of this successional sequence is c. 100 years (Wardle 1980; Basher et al. 1985). We established six 50-m2 circular plots (4-m radius) in the open stage, and five plots in each of the young, vigorous and mature stages; plots were located sufficiently far apart (> 100 m) to be considered independent of one another. The plots were all on rock slides (two open, two vigorous plots) or floodplains (all other plots), where the major habitat difference appeared to be slope. Slope ranged from 1° to 38° and was independent of stages.

Table 1.  Mean vegetation characteristics and soil chemical properties (± SE) in four developmental stages of Carmichaelia odorata stands from the Kokatahi Valley, New Zealand. n = 5 plots per stage (except the open stage: n = 6, and the organic soils of the vigorous stage: n = 3). Values within the same row were compared by anova; F statistic, degrees of freedom and P-values are listed. Those numbers in the same row followed by the same letter are not significantly different (P > 0.05, SchefféF-tests). No organic soils were found in the open and young stages
 Developmental stage of Carmichaelia Fd.f.P
 OpenYoungVigorous Mature 
Canopy height (cm)
 Mean     32 ± 5.7a     33 ± 5.2a    180 ± 8.49b     396 ± 39.2c 61.983, 17< 0.001
 Maximum     82 ± 9.6a    117 ± 2.97a    272 ± 19.2b     524 ± 31.3c 96.303, 17< 0.001
Species richness (no. plot−1)
 Total 28.7 ± 3.8 25.4 ± 2.2 24.0 ± 3.3  24.0 ± 1.1  0.493, 17      0.692
 Non-woody 21.3 ± 3.0a 18.0 ± 1.7ab 13.6 ± 1.5ab   9.8 ± 0.9b  4.833, 17      0.013
  Ferns  2.0 ± 1.0  2.6 ± 1.2  4.4 ± 1.2   6.6 ± 0.5  3.153, 17      0.052
  Monocots  8.8 ± 1.6a  7.6 ± 1.3ab  5.6 ± 0.6ab   1.8 ± 0.3b  5.583, 17      0.007
  Dicots 10.5 ± 1.4a  7.8 ± 0.8ab  3.6 ± 0.6bc   1.4 ± 0.7c 14.193, 17< 0.001
 Woody  7.3 ± 1.2a  7.4 ± 0.7a 10.4 ± 1.9ab  14.2 ± 0.7b  5.383, 17      0.009
Woody stem (no. plot−1)
 Total 18.0 ± 3.60a 73.6 ± 13.9ac169.0 ± 21.42b 127.6 ± 13.28bc 18.843, 17< 0.001
 Carmichaelia  5.0 ± 1.3a 38.8 ± 3.55a 82.2 ± 15.8b  15.2 ± 3.78a 15.133, 17< 0.001
 Griselinia      0 ± 0 0.20 ± 0.18  8.4 ± 3.9   6.4 ± 1.5  3.783, 17      0.030
 Metrosideros 0.16 ± 0.15  6.2 ± 3.4  2.0 ± 1.4   2.8 ± 1.0  1.583, 17      0.231
 Weinmannia 0.67 ± 0.30a  1.2 ± 0.52a  9.8 ± 5.3a      54 ± 12b 12.633, 17< 0.001
Light (% transmission of total) 80.3 ± 5.20a 89.6 ± 2.44a 20.9 ± 3.26b  6.13 ± 0.859b 82.333, 17< 0.001
Fine particle (% of total mass) 14.0 ± 2.66 12.2 ± 2.60 13.4 ± 4.99  16.1 ± 2.97  0.253, 17      0.858
pH 5.36 ± 0.06a 5.38 ± 0.07a 5.12 ± 0.17ab  4.90 ± 0.35ab 4.54 ± 0.14b  4.38 ± 0.14b 8.145, 22< 0.001
Organic C (%) 0.02 ± 0.0032a 0.16 ± 0.024a 2.19 ± 0.785ab 24.31 ± 4.003b 6.29 ± 1.35ab 30.81 ± 3.022b50.685, 22< 0.001
 g m−2 17.1 ± 3.53a 24.3 ± 6.12a247.0 ± 120.7a2265.1 ± 481.4b570.2 ± 177.8a2282.8 ± 223.9b26.505, 23< 0.001
Total N
 % 0.27 ± 0.06a 0.22 ± 0.03a 0.34 ± 0.04ab  1.12 ± 0.15ab 0.59 ± 0.14ab  1.44 ± 0.13b26.405, 22< 0.001
 g m−2 22.5 ± 4.49a 32.9 ± 11.6a 29.9 ± 8.41a 104.3 ± 11.21b 56.5 ± 16.6b 106.5 ± 8.937b 9.445, 23< 0.001
C : N ratio0.661 ± 0.155a0.847 ± 0.176ab 5.96 ± 1.88ab  21.1 ± 3.44bc 11.9 ± 2.36b  21.6 ± 1.68c23.485, 23< 0.001
Available P
 µg g−1 21.4 ± 4.6a 18.2 ± 1.2a 29.2 ± 2.4ab  36.3 ± 2.3b 24.6 ± 2.5ab  29.8 ± 2.5ab 4.495, 22      0.006
 g m−2 0.18 ± 0.038 0.26 ± 0.066 0.22 ± 0.035  0.34 ± 0.018 0.22 ± 0.042  0.22 ± 0.016 1.025, 23      0.430
Cation exchange capacity (me/100 g)  2.1 ± 0.16a  2.1 ± 0.26a  5.9 ± 0.85a  13.2 ± 0.93ab 14.6 ± 3.61ab  24.5 ± 4.14b13.315, 22< 0.001
Minerals (µg g−1)
 Ca169.4 ± 21.4a175.2 ± 31.8a436.2 ± 26.7ab 890.7 ± 305.7ab782.4 ± 213.9ab1034.8 ± 154.1b 7.275, 22< 0.001
 Fe 86.8 ± 12.3a 91.6 ± 3.4a167.2 ± 22.8ab 133.3 ± 46.3ab236.8 ± 28.7b 136.4 ± 19.1ab 6.685, 22      0.001
 K 13.2 ± 1.8a 14.6 ± 0.7a 46.4 ± 7.1ab 217.3 ± 19.5b 55.8 ± 6.9ab 166.8 ± 224.8b38.445, 22< 0.001
 Mg 36.6 ± 4.4a 40.0 ± 5.3a 70.4 ± 12.4a 145.0 ± 12.8b139.8 ± 19.3b 261.4 ± 21.9c35.965, 22< 0.001
 S  7.0 ± 0.4a  6.6 ± 0.5a 10.2 ± 1.0a  16.3 ± 0.7ab 13.6 ± 0.8ab  21.0 ± 4.1b 8.495, 22< 0.001
 Zn  2.0 ± 0.2  2.5 ± 0.2  3.1 ± 0.6   6.4 ± 2.5  5.3 ± 1.5   9.6 ± 4.1 2.085, 22      0.107

Soils and light

At each plot we analysed physical and chemical characteristics of the soils to evaluate the possible effects of Carmichaelia. Volume and mass of three size fractions of inorganic particles (gravel, > 16 mm diameter; cinder, 2–16 mm; and fines, < 2 mm) were determined from one pit 20 × 20 cm wide × 10 cm deep in each plot. After discarding any surface litter, we removed the entire organic layer (A0 horizon) and the top 10 cm of the mineral soil (A1 horizon) to sample depth. The fractions were sieved and the volumes (by flotation) and mass of gravel and cinder were determined. The volume of fines (including air pockets in the soil) was determined by subtraction. The mass of the organic horizon and the fines were determined from subsamples brought back to the lab and dried at 30 °C for 4 days. We determined mineral bulk density (A1 horizon; g cm−3) in two ways: the dry mass of fines per volume of fines, and the dry mass of gravel, cinder and fines combined per pit volume (total bulk density). Fractions of sand, silt and clay were determined by the Malvern Laser Sizer (Singer et al. 1988).

Five subsamples (0.5 L, 10 cm deep) of mineral soil were taken from just outside each plot and depths of leaf litter and A0 horizon (when present) recorded. Soils were sieved (2-mm mesh), weighed and oven-dried at 30 °C for 4 days before re-weighing to determine gravimetric moisture content. All subsequent analyses were performed on sieved, dry soils, including pH (1 : 1 ratio of soil and distilled water), organic C (organic matter from mass loss following combustion multiplied by 1.7), total N (CHN analyser), available P and mineral elements (Mehlich 3 extractant, Mehlich 1984; Brookside Laboratory Association Inc., Knoxville, Ohio, USA). Areal estimates of soil N, C and P were made by multiplying their concentration by the density of the fines times soil volume (1 m2 by 10 cm depth).

Availability of light in each plot was determined in April 2000 and February 2001 using hemispherical photographs taken 1 m above the ground at the centre of the plot. Images were digitized and analysed using Gap Light Analyser 2.0 (Frazer & Canham 1999) to determine percentage transmission of total light.


In each plot we counted individuals of each woody species (open and young: > 10 cm tall; vigorous and mature: > 50 cm tall) and measured the height of the five tallest individuals of each species. We also recorded all vascular species present, measured total canopy height (average per plot), and estimated the cover of individual vascular species and of mosses and lichens in six cover classes (1 = < 1%, 2 = 1–5%, 3 = 6–25%, 4 = 26–50%, 5 = 51–75%, 6 = > 75%), with that for vascular plants recorded in tiers (< 0.3 m, 0.3–2 m, 2–5 m, > 5 m; methods of Allen 1992). Five Carmichaelia plants were selected randomly in each plot, and samples of foliage, photosynthetic stems and roots (stems and roots both < 1 cm diameter, roots to 10 cm depth) were collected for analysis of tissue concentrations of N and P using automated colourimetric methods (Technicon Instruments 1977).

Relationships between vascular plant species cover and environment were examined using detrended correspondence analysis (DCA) and canonical correspondence analysis (CCA) and partial CCA in Canoco for Windows 4.0 (ter Braak & Šmilauer 1998) using default settings. Correlations between environmental variables and ordination axes were evaluated using Pearson’s correlation coefficients, with probabilities assessed using Bonferroni corrections (Kleinbaum et al. 1988). Following examination of initial correlations, we conducted a CCA of the data using stepwise forward selection of site variables to determine which of the 24 variables best predicted observed vegetation patterns. Monte Carlo permutations were applied to each site predictor to assess whether its contribution to species sorting was greater than would be expected by chance. All variables that were significant predictors (P < 0.05) were retained in a CCA.

To determine the role of N as an environmental gradient related to species composition, the correlations between soil N levels and the other 23 environmental variables were assessed using Pearson’s correlation coefficients, with probabilities assessed using Bonferroni corrections. To partition out the effect of N as a predictor of observed vegetation patterns, we first conducted a CCA using only N as the sole constraining variable. We then conducted a partial CCA using N as a constraining variable, and five variables determined to be significant as constraining variables from forward selection in CCA as covariables. In this way, the variance explained by N alone that was not correlated with the other five variables could be calculated. Finally, we carried out a partial CCA using the five variables determined to be significant as constraining variables from forward selection in CCA as constraining variables, and N alone as a covariable. In this way, the variance explained by those five variables that was not correlated with N could be evaluated (Økland 1999).

Glasshouse experiments

Two glasshouse pot experiments were set up at Lincoln, New Zealand (43°38′ S, 172°29′ E, 15 m a.s.l.) in early March 1999 to examine the potential impact of Carmichaelia and of nitrogen on three forest tree species commonly found in the four stages at our study site. In experiment 1, we examined the effects of developmental stage (as represented by soil properties), shade and Carmichaelia litter. The full factorial design was: four soil types × three species × two light levels × two litter levels with six replicates, yielding 288 experimental pots. These were arranged into six replicate shaded blocks and six replicate unshaded blocks (main effect = shade). We collected approximately 125 kg of soil from representative sites of each of the four stages and hand-removed roots and rocks > 16 mm in diameter, and used a 50 : 50 mixture with 10-mm diameter river gravel to fill 1-litre pots that were planted with a single seedling of either Griselinia littoralis, Metrosideros umbellata or Weinmannia racemosa (hereafter referred to by genus). Seedlings were raised from seed except for Metrosideros (from cuttings) and were 1 year old when planted. Before planting, roots were washed clean and the roots and shoots of Griselinia and Weinmannia seedlings trimmed to a uniform size (initial transplant dry mass in grams; mean ± SE, n = 8: Griselinia: roots, 0.56 ± 0.05; shoots, 1.03 ± 0.08; Metrosideros: roots, 0.30 ± 0.03; shoots, 0.67 ± 0.05; Weinmannia: roots, 0.58 ± 0.10; shoots, 0.51 ± 0.07). Seedlings were either placed under optically neutral shade cloth that reduced ambient light to 5% (to imitate light conditions under Carmichaelia thickets in the vigorous and mature stages, Table 1) or left uncovered (i.e. normal glasshouse lighting conditions; 30% of ambient). Carmichaelia leaf/photosynthetic stem litter was collected from vigorous stages in the Kokatahi Valley and was air-dried, chopped and sieved (3-mm mesh) before addition to litter-treatment pots. Photosynthetic stems < 3 mm diameter were used with leaf litter, because they form a substantial fraction of the litter under Carmichaelia plants, and because we surmized that a mix of leaves (fast decomposition) and stems (slower decomposition) would provide both fast- and slow-release fertilization effects (both are rich in N, Table 2). A mean 27.2 g dry mass (mean = 0.80 g N pot−1) was applied in equal parts over three occasions (March, May and November 1999) and lightly mixed into the surface of the soil in each pot to speed decomposition and to avoid surface mulch effects.

Table 2.  N and P concentrations (mean ± SE) in tissues of the shrub Carmichaelia odorata sampled from four stages of stand development on primary seres in the Kokatahi Valley, New Zealand. Values within the same row were compared by anova. Those followed by the same letter are not significantly different (P > 0.05, SchefféF-tests)
N (mg g−1)
 Leaf25.64 ± 1.075a25.20 ± 1.177a31.30 ± 2.163ab37.49 ± 3.255b5.9733, 160.006
 Stem17.42 ± 0.7969a22.06 ± 0.6775ab21.44 ± 1.321ab24.90 ± 1.834b5.5133, 170.008
 Root22.78 ± 2.05726.30 ± 1.66921.49 ± 2.16721.42 ± 3.3370.7213, 170.553
P (mg g−1)
 Leaf 1.49 ± 0.0851ab 1.37 ± 0.0396a 1.75 ± 0.105ab 1.87 ± 0.136b4.3353, 160.020
 Stem 1.26 ± 0.0833 1.32 ± 0.0691 1.52 ± 0.146 1.52 ± 0.2100.8233, 170.499
 Root 1.30 ± 0.0990 1.28 ± 0.0578 1.22 ± 0.03000.983 ± 0.05633.5193, 170.038

Plants were watered as necessary to maintain soil moisture (approximately once daily) and the air temperature was kept at approximately 5 °C of ambient (range: −4 °C to +33 °C). All pots and all blocks were rotated in the glasshouse once monthly to offset position effects on seedling performance. The experiment was harvested between 13 and 21 March 2000, approximately 1 year after set-up. Leaves, stems and roots of each seedling were separated and their dry mass determined. We calculated total leaf area on fresh leaves from each seedling (leaves > 2 mm length only) using a LI-3100 area meter (Li-Cor Inc., Lincoln, Nebraska, USA), and calculated specific leaf area (total leaf area/total leaf mass; SLA) and root : shoot biomass ratios. Nitrogen and P were measured on 0.2 g of leaf material (which included petioles) from each seedling using automated colourimetric methods (Technicon Instruments 1977).

Experiment 2 was set up to investigate the response of the same three forest species to four levels of N addition to assess whether any positive effect of Carmichaelia litter additions were likely to be due to addition of N in the litter. The full design was three species × two light levels × four N levels with six replicate shaded and unshaded blocks (as experiment 1) yielding a total of 144 pots. In early March 1999 the seedlings (source and size as experiment 1) were transplanted into 1-L pots in a 50 : 50 mix of fine river sand and 10-mm-diameter river gravel, and watered to field capacity (50 mL per pot) three times a week over the duration of the experiment with a nutrient solution (Millard & Proe 1991) that contained either 0.0, 0.25, 1.0 or 4.0 mMol N but was complete for other nutrients. The total N added over 52 weeks was 0, 0.023, 0.091 and 0.36 g pot−1, respectively, for each of these treatments. Harvest procedures and measurements were identical to those described for experiment 1.

Statistical analyses

Field data on soils, vegetation parameters, vegetation and microhabitat cover values were compared among stages with one-way analyses of variance (anovas – using rank variables where necessary) followed by multiple comparison tests. Percentage data were arc-sine square-root transformed before analysis. Significance of effects for field data was determined at the P < 0.05 level. Data from glasshouse experiments on biomass and allocation parameters for seedlings of three tree species were compared initially by multivariate anova (manova). Where these were significant we analysed these parameters further using three-way (experiment 1) and two-way (experiment 2) anovas. Significance was determined at the P ≤ 0.005 level for all experiments because we measured multiple response variables.


Field studies


Organic horizons, absent in the open and young stages of Carmichaelia development, were less deep in the vigorous stage than in the mature stage (0.48 ± 0.18 [SE] cm vs. 1.3 ± 0.23 cm, t10 = 2.42, P = 0.042). Bulk densities did not differ across stages; over all stages total bulk density was 2.06 ± 0.06 g cm−3, and fine particulate density was 0.88 ± 0.11 g cm−3. Gravel constituted the largest percentage (47.5 ± 2.53%) of total mass, with cinder next (38.5 ± 2.15%) and fines the least (13.9 ± 1.03%), of which > 80% was sand and < 2% was clay (pooled values across all stages). Soils were significantly drier (F3,17 = 18.78; P < 0.001) in the open or young stages (0.8 ± 0.3% and 0.5 ± 0.04% gravimetric moisture, respectively) than in the mature stage (28 ± 7.9%).

Soil pH declined through the chronosequence while organic C matter percentages and pools increased over 100-fold (Table 1). Mineral soils in the open and young stages had significantly less C and higher pH than organic soils in the vigorous and mature stages (Table 1). Nitrogen pools increased twofold in the mineral soils during the chronosequence and were two to four times higher in organic than mineral soils in the same stage (Table 1). Available P in open or young stages was less than in organic soils of the vigorous stage (but not of the mature stage, Table 1). Cations (Ca, K, Mg) also increased from open and young soils to mineral and organic soils of later stages (Table 1).


Total stem density and Carmichaelia stem density increased until the vigorous stage and then declined (Table 1). Mean canopy heights increased from open or young stages via vigorous to mature stages (Table 1). Lichens had greatest cover in the open stage (9 ± 6%) but differences among stages were not significant (F3,17 = 2.40; P = 0.104). Moss cover reached > 25% in several of the early successional plots but was generally much lower and did not differ among stages. Although the number of vascular plant species per plot did not vary, the number of herbaceous species per plot decreased through the stages, while the number of woody plant species increased (Table 1). The proportion of exotic vascular plant species declined across the chronosequence (open: 13.4 ± 1.8%; young: 10.0 ± 2.8%; vigorous: 4.9 ± 1.6%; mature: 0%; F3,17 = 11.36; P < 0.001).

The successional gradient was reflected in Axis 1 scores in DCA: open and young plots had low scores, while mature plots had high scores (Fig. 1a). Species with low Axis 1 scores were typically herbs (especially grasses), while woody plant species had higher Axis 1 scores (Fig. 1b). Woody plant and fern species-richness scores per plot were strongly positively correlated with plot Axis 1 scores, while those for monocotyledonous and dicotyledonous herbs were strongly negatively correlated (Pearson’s correlation coefficients (Bonferroni corrected), P ≤ 0.002). There were strong positive correlations between Axis 1 plot scores and cation exchange capacity, S, Mg, K, Fe and A0 depth, and a strong negative correlation between Axis 1 scores and soil pH (Pearson’s correlation coefficients (Bonferroni corrected), P < 0.001). There were no clear patterns of species distribution along DCA Axis 2, and Axis 2 plot scores were not correlated with any of the environmental variables for which we collected data.

Figure 1.

DCA ordinations of vegetation in 50-m2 plots (n = 21) along a stand development sequence of Carmichaelia odorata in the Kokatahi Valley, New Zealand. (a) Ordination of plots. Filled circles = open plots, open circles = young plots, filled triangles = vigorous plots, and open triangles = mature plots of Carmichaelia. Bi-plot scores are given for six environmental variables, all of which except N were determined as significant in constraining ordinations in CCA. (b) Ordination scores of 42 species that occur in at least 60% of plots within at least one developmental stage. Filled circles = ferns, open circles = monocotyledonous herbs, filled triangles = dicotyledonous herbs, and open triangles = woody dicotyledons. Eigenvalues are the same for both graphs: axis 1 eigenvalue = 0.5311, axis 2 eigenvalue = 0.2323.

Five of 24 site variables were significant (P < 0.05, Monte Carlo permutations) in constraining CCA, i.e. slope, soil cation exchange capacity, and soil S, P and Zn. The sum of all canonical eigenvalues in CCA ordination constrained by these five variables was 1.248. Bi-plot DCA scores of these variables are shown in Fig. 1(a), along with that of soil N. Soil N used alone was significant in constraining CCA (P = 0.005, Monte Carlo permutations). However, soil N was strongly positively correlated with two of the variables determined through forward selection procedures, i.e. slope and soil S, as well as with soil C, Ca, Mg, Mn, Al and water content, and strongly negatively correlated with soil pH (Pearson’s r-values all > 0.7, Bonferroni-adjusted P < 0.003). The sum of all canonical eigenvalues in a CCA ordination constrained by soil N alone was 0.348. From partial CCAs we determined that 9% of the variation explained by soil N was not correlated with slope, soil cation exchange capacity, and soil S, P and Zn, and that 75% of the variation explained by those five variables was not correlated with soil N.

Foliar and stem N concentrations of Carmichaelia increased along the successional stages (Table 2), so that leaf N was greater in mature stages than in open and young stages, and stem N was greater in mature stages than in open stages. Similarly, foliar P of Carmichaelia was greater in mature stages than in young stages. Foliar N concentrations exceeded stem N concentrations in the open, vigorous and mature stages (paired t-tests, P < 0.01, Table 2), and foliar P concentrations exceeded stem P concentrations in the mature stage (paired t = 3.33, P = 0.029). There were no differences in Carmichaelia root N or P concentrations among the stages (Table 2).

Of the three tree species examined in glasshouse experiments, Weinmannia stem density was less in all stages other than mature, where it was also greater than densities of Griselinia and Metrosideros (Table 1). However, densities of the three species were not different in the earlier three stages.

Glasshouse experiments

Experiment 1

Shade, added Carmichaelia litter and successional stage of soils all had highly significant effects on the biomass and allocation parameters of all three tree species. Overall manovas for each parameter measured for each of the three species were highly significant, as were differences among the three species (P < 0.001). There were no significant block effects on any of the biomass or allocation parameters measured. Only one three–way interaction was significant (shade × litter × stage for foliar P in Weinmannia;F3,79 = 3.90, P = 0.012), and we therefore confine results to main treatment effects and two-way interactions. Survivorship was high for Griselinia and Metrosideros (only one seedling of each died during the experiment) and for Weinmannia, except for shaded seedlings on vigorous-stage soils amended with Carmichaelia litter, in which all but one seedling died.

Of the three tree species, Griselinia was by far the most responsive to the treatments, with about twice as many significant (P < 0.005) responses as either Metrosideros or Weinmannia (Table 3). Seedlings of Griselinia and Weinmannia grew more relative to initial total seedling mass (mean 345% and 315% increases, respectively, across all treatments) than did Metrosideros(mean 200% increase).

Table 3. anova results (F-values, with P-values in parentheses) for responses of seedlings of three species to shade, addition of Carmichaelia litter, and soil from under four stages of Carmichaelia stands. Degrees of freedom for terms in anova model are given in brackets except for shade × litter × stage (3, not reported in this table) and error (79)
 Treatment effect
 Shade (1)Litter (1)Stage (3)Shade × litter (1)Shade × stage (3)Litter × stage (3)
Shoot mass
 Griselinia79.6 (< 0.001)141 (< 0.001)7.35 (< 0.001)48.2 (< 0.001)2.97 (0.037)1.25 (0.297)
 Metrosideros89.9 (< 0.001)10.6 (0.002)1.84 (0.147)6.04 (0.016)2.52 (0.064)0.487 (0.692)
 Weinmannia78.1 (< 0.001)103 (< 0.001)9.37 (< 0.001)52.3 (< 0.001)0.280 (0.839)0.875 (0.458)
Root : shoot ratio
 Griselinia421 (< 0.001)67.7 (< 0.001)2.92 (0.039)2.71 (0.104)0.679 (0.568)0.257 (0.856)
 Metrosideros115 (< 0.001)1.33 (0.253)2.77 (0.047)1.80 (0.184)5.55 (0.002)0.862 (0.465)
 Weinmannia70.8 (< 0.001)44.9 (< 0.001)11.9 (< 0.001)7.73 (0.007)0.861 (0.465)4.00 (0.011)
Specific leaf area
 Griselinia954 (< 0.001)38.9 (< 0.001)6.97 (< 0.001)9.25 (0.003)4.19 (0.008)0.611 (0.610)
 Metrosideros8.29 (0.005)2.01 (0.161)0.483 (0.695)0.158 (0.692)0.492 (0.689)0.299 (0.826)
 Weinmannia104 (< 0.001)0.651 (0.423)1.29 (0.284)0.462 (0.499)1.41 (0.248)0.124 (0.945)
Foliar N
 Griselinia558 (< 0.001)129 (< 0.001)6.00 (0.001)39.3 (< 0.001)1.41 (0.247)1.29 (0.284)
 Metrosideros166 (< 0.001)86.1 (< 0.001)1.64 (0.188)35.4 (< 0.001)1.65 (0.185)0.846 (0.473)
 Weinmannia261 (< 0.001)104 (< 0.001)6.91 (< 0.001)8.10 (0.006)2.03 (0.117)2.87 (0.042)
Foliar P
 Griselinia243 (< 0.001)48.7 (< 0.001)10.1 (< 0.001)25.3 (< 0.001)3.16 (0.029)1.98 (0.124)
 Metrosideros21.6 (< 0.001)0.011 (0.918)2.12 (0.105)14.6 (< 0.001)0.078 (0.972)0.957 (0.418)
 Weinmannia0.480 (0.491)58.6 (< 0.001)3.75 (0.015)2.99 (0.088)1.33 (0.270)4.65 (0.005)

For all three species, shade significantly depressed shoot mass and root : shoot ratio (Table 3, Figs 2 and 3), but increased specific leaf area (SLA) and foliar N (Figs 4 and 5). Shade also increased foliar P, but only for Griselinia and Metrosideros (Fig. 6). Carmichaelia litter addition significantly increased shoot mass and foliar N in all three species (Table 3, Figs 2 and 5), reduced root : shoot ratio and foliar P of Weinmannia (Figs 3 and 6), and increased SLA and foliar P of Griselinia (Figs 4 and 6). Shade × litter interactions were significant for shoot mass of Griselinia and Weinmannia, SLA of Griselinia, and for foliar N and P of Griselinia and Metrosideros (Table 3).

Figure 2.

Shoot mass (g) of three tree species (Griselinia, Metrosideros, Weinmannia) grown for 52 weeks in two glasshouse experiments with and without shade (both experiments) and with and without additions of Carmichaelia odorata litter (experiment 1) or nitrogen additions (experiment 2). In experiment 1, tree seedlings were grown in soils from under four stages of Carmichaelia stands in the Kokatahi Valley, New Zealand, with or without Carmichaelia litter. In experiment 2, complete nutrient solutions (varying only in the amount of nitrogen present) were added three times each week to the sand/gravel mixture in each pot. Vertical bars represent one standard error. Note different y axis scaling on different panels.

Figure 3.

Root : shoot biomass ratio of the same three species in two experiments as described for Fig. 2.

Figure 4.

Specific leaf area (cm2 g−1) of the same three species in two experiments as described for Fig. 2. Note different y axis scaling on different panels.

Figure 5.

Foliar N concentrations (mg g−1) of the same three species in two experiments as described for Fig. 2.

Figure 6.

Foliar P concentrations (mg g−1) of the same three species in two experiments as described for Fig. 2. Note different y axis scaling on different panels.

The successional stage of soils also influenced biomass and allocation parameters in Griselinia and Weinmannia but had no significant effects on Metrosideros (Table 3). With increasing soil development, shoot mass and foliar N increased in Griselinia and Weinmannia (Figs 2 and 5), SLA and foliar P increased in Griselinia (Fig. 4), and root : shoot ratio decreased in Weinmannia (Fig. 3).

Experiment 2

As for experiment 1, overall manovas and differences among species for all measured parameters were highly significant (all P < 0.001), and there were no significant block effects on any of the parameters. Shade and added N had a similar number of significant effects (Table 4). Again, Griselinia was by far the most responsive species, with nearly twice as many significant effects of treatments as either Metrosideros and Weinmannia. Survivorship of all three species was high.

Table 4.  Analysis of variance results (F-values, with P-values in parentheses) for responses of seedlings of three species to shade and four levels of nitrogen addition (0, 0.25, 1.0 and 4.0 mMol N) treatments. Terms for anova model (degrees of freedom in brackets) are shade (1), N (3), shade × N (3), error (38)
 Treatment effect
 ShadeNShade × N
Shoot mass
 Griselinia213 (< 0.001)105 (< 0.001)59.0 (< 0.001)
 Metrosideros84.4 (< 0.001)9.20 (< 0.001)5.95 (0.002)
 Weinmannia10.2 (0.003)5.67 (0.003)2.74 (0.059)
Root : shoot ratio
 Griselinia102 (< 0.001)26.2 (< 0.001)2.12 (0.114)
 Metrosideros18.6 (< 0.001)3.75 (0.019)1.69 (0.185)
 Weinmannia9.25 (0.005)12.1 (< 0.001)0.525 (0.668)
Specific leaf area
 Griselinia288 (< 0.001)12.5 (< 0.001)2.28 (0.096)
 Metrosideros7.85 (0.008)1.21 (0.318)0.256 (0.857)
 Weinmannia28.6 (< 0.001)2.18 (0.109)0.048 (0.986)
Foliar nitrogen
 Griselinia137 (< 0.001)122 (< 0.001)7.36 (0.001)
 Metrosideros30.7 (< 0.001)25.2 (< 0.001)4.92 (0.006)
 Weinmannia3.21 (0.082)54.0 (< 0.001)1.42 (0.255)
Foliar phosphorus
 Griselinia129 (< 0.001)9.85 (< 0.001)8.87 (< 0.001)
 Metrosideros0.111 (0.741)5.54 (0.003)4.67 (0.007)
 Weinmannia0.257 (0.615)7.29 (0.001)1.98 (0.137)

Shade depressed shoot mass and root : shoot ratio in all three species (Table 4, Figs 2 and 3), and increased SLA in Griselinia and Weinmannia (Fig. 4), increased foliar N in Griselinia and Metrosideros (Fig. 5), and increased foliar P in Griselinia (Fig. 6). Added N increased shoot mass and foliar N in all three species, often substantially so (Figs 2 and 5). Added N substantially reduced root : shoot ratios for Griselinia and Weinmannia (Fig. 3) and increased SLA in Griselinia (Fig. 4). While N levels were varied in experiment 2, P levels were added in constant amounts and were not limiting. In this experiment, foliar P concentrations declined in response to increased added N in the case of Metrosideros and especially Weinmannia (Fig. 6), but increased for Griselinia (Fig. 6). There were significant interactions between shaded and added N treatments in the cases of shoot mass for Griselinia and Metrosideros (Table 4, Fig. 2), and foliar N and P for Griselinia (Table 4, Figs 5 and 6).


Our study provided evidence in support of both of our hypotheses. First, both facilitation and inhibition appeared to be important in influencing tree seedling biomass and nutrient acquisition during early successional stages. Secondly, seedlings of different tree species that establish in the same habitat differed markedly in their relative response to manipulation of nutrients and light.

Biotic mechanisms affecting seedling performance

During the successional sequence that we studied, Carmichaelia increased in abundance, then senesced, while total soil organic matter and N pool size showed a large increase and then plateaued (Table 1). This is consistent with studies in other early successional systems that show nitrogen-fixing shrubs to be responsible for rapid accretion of soil N (Van Cleve et al. 1971; Luken & Fonda 1983; Vitousek & Walker 1989). Many of the soil changes were correlated with changes in vegetation. Increasing soil N, S and P corresponded closely with the principal DCA axis, which in turn reflected a change in vegetation structure from herbs and grasses in stands with little or no Carmichaelia present to shrubs and trees in Carmichaelia-dominated thickets (Fig. 1).

Increasing soil nitrogen along this successional gradient (Fig. 1) corresponded with an increase in foliar and stem N concentrations in Carmichaelia (Table 3). However, there are also sources of nitrogen input other than Carmichaelia during early succession. In particular, the lichen Stereocaulon colensoi Church Bab. that was common in open and young stages probably fixes N given that its congener S. vulcani serves as a source of N input in primary seres in Hawai’i (Kurina & Vitousek 1999). While our study cannot show conclusively whether changes in soil N are caused entirely by Carmichaelia stand development, it appears likely that Carmichaelia plays an important role in ecosystem N aggradation given our experimental evidence that N-rich Carmichaelia litter differentially enhances biomass and foliar N concentrations of seedlings of dominant tree species (Figs 2 and 5). The absolute amount of N added through Carmichaelia litter (0.80 g N pot−1) in experiment 1 exceeded the highest levels of N applied as solution over the course of experiment 2 (0.36 g N pot−1 in the case of 4.0 mMol N solutions). However, microbial use of N in the litter due to associated labile sources of C present in the litter would have reduced the total amount of nitrogen available to seedlings (Wardle 1992). Beneficial effects of litter on tree seedling growth and allocation were shown in both unshaded and shaded treatments, particularly for Griselinia (Figs 2–6). These effects are similar to those shown for addition of N-rich litter on biomass growth of tree seedlings in other primary seres (Walker & Vitousek 1991; Chapin et al. 1994).

Mineral soils from the chronosequence under Carmichaelia differed in nutrient concentrations (Table 1) and had significant effects on biomass and allocation in Griselinia and Weinmannia seedlings, with greater shoot mass and foliar N concentrations of seedlings grown on soils from under mature stands than under open stands (Figs 2 and 5). These results may reflect facilitative effects of Carmichaelia through its influence on soil development, although we did not measure this directly. Our results are similar to changes in biomass and allocation of Picea sitchensis seedlings that became more pronounced with increasing soil development under Dryas drummondii and Alnus sinuata during primary successions following deglaciation in Alaska (Chapin et al. 1994).

Shade, at a level comparable to that found under vigorous and mature stands of Carmichaelia, reduced allocation to shoots and decreased root : shoot ratios and increased SLA in both glasshouse experiments (Figs 2–4). Shading has been shown to produce the same changes in allocation patterns in other studies of temperate tree seedlings (Loach 1970; Latham 1992; Grubb et al. 1996). The number of significant effects of shade across all species in our experiments was comparable with the number of significant effects of added N, either via litter (Table 3) or added N solutions (Table 4). This points to the dual effects of inhibition through shading and stimulation through N build-up during succession, and the success of tree seedlings in early successional sites will be determined by the relative balance of these two opposing mechanisms.

Differential responses of tree species

All three species used to assess effects of Carmichaelia on future forest development colonize sites simultaneously (Table 1, see also Wardle 1980). However, our study shows that biomass and allocation patterns are substantially different among the three species (e.g. shoot mass after 1 year, Fig. 2). Griselinia and Weinmannia were much more responsive in absolute terms to any of the treatments we imposed than was Metrosideros (Figs 2–6). Wardle (1971) noted that Metrosideros does not compete with faster-growing species such as Griselinia and Weinmannia in more fertile conditions (Fig. 2), and as a consequence tends to dominate only in infertile sites.

Inhibitory effects of shade (at levels equivalent to that measured under developed Carmichaelia stands) were uniform across species (Tables 3 and 4), although interactions between shade and added N (via litter or solution) were variable between species, with the strongest interactions occurring for Griselinia and weakest for Weinmannia (Tables 3 and 4). In contrast, facilitative effects of Carmichaelia (either through added litter or expressed through soil development) resulted in different responses among species.

Given that shading responses were consistent across the three tree species, the effects of N relative to light in influencing seedling growth are most variable for Griselinia, intermediate for Weinmannia and least for Metrosideros (Fig. 2). We could expect this differential response to contribute to variation in relative performance when all three species co-occur, with the relative success of species depending on the amount of N made available following Carmichaelia colonization. That all three species are common in mature forest in the study area suggests there is variation across early successional sites with respect to the relative balance of light and nutrient availability, leading to coexistence of all three species being possible at the stand and landscape scale (Stewart & Veblen 1982; Reif & Allen 1988).

Nitrogen-rich Carmichaelia litter increased foliar N concentrations of all three species, but increased foliar P concentrations only of Griselinia across all stages. In a comparable system, addition of nitrogen-rich Myrica faya litter increased shoot N concentrations of Metrosideros polymorpha in early successional systems in Hawai’i, but did not increase shoot P concentrations (Walker & Vitousek 1991). We observed similar effects of Carmichaelia litter on its congener Metrosideros umbellata (Fig. 6). Foliar nutrient concentrations increased across the chronosequence for Griselinia (N and P) and Weinmannia (N only). In contrast foliar nutrient concentrations of Metrosideros umbellata showed no response (Figs 5 and 6), which is the same result as was shown for its congener M. polymorpha during the earliest stages of succession in Hawai’i (Crews et al. 1995).

Specific leaf area of Griselinia increased across the soil chronosequence, and with added Carmichaelia litter or added N (Fig. 4), which is in accordance with trends found by Poorter & de Jong (1999) where SLA across species increased with site productivity. In contrast there were no relationships between SLA and nitrogen or litter addition for Metrosideros or Weinmannia (Fig. 4). SLA of Metrosideros had a substantially lower range than that for Griselinia, which was in turn lower than that for Weinmannia (Fig. 4). Meanwhile foliar N concentrations were similar across the three species (Fig. 5). Turner (1994) found a positive relationship between foliar nitrogen concentrations and SLA across species, and that increasing nitrogen availability during succession could increase SLA; this in turn could increase productivity per unit plant mass (Poorter & Remkes 1990). There was a strong positive relationship in experiment 1 between foliar N and SLA for Griselinia (R2 = 0.65, P < 0.001), a weaker positive relationship for Weinmannia (R2 = 0.42, P < 0.001) and no relationship for Metrosideros (R2 = 0.04, P = 0.052). It appears therefore that SLA is a suitable assay for leaf construction costs with respect to resource availability for some species (e.g. Griselinia), as suggested by Poorter & Remkes (1990) but not others.

Overall, Metrosideros was the least responsive of the three tree species to facilitative effects of Carmichaelia. Compared with Griselinia and Weinmannia, its unresponsiveness to soil development and its much lesser response of stem biomass growth to added Carmichaelia litter suggest that it lacks the phenotypic plasticity to take advantage of nutrient build-up in the way the other two species are able to. The comparatively low plasticity of Metrosideros demonstrated in this study is consistent with its tendency to establish in nutrient-poor sites compared with the other two species (Wardle 1971). Metrosideros can dominate some primary successions in these valleys (e.g. on landslides, Stewart & Veblen 1982) and this may occur most often on sites not colonized by Carmichaelia. Reif & Allen (1988) considered that Weinmannia tends to dominate as a canopy species on more fertile sites than Metrosideros and our experimental evidence provides further support for this.

Fine-scale habitat heterogeneity may be a consequence of primary successions in these valleys such that proximity to neighbouring Carmichaelia shrubs and their litter zone and rhizosphere may favour growth and development of Weinmannia and Griselinia, while Metrosideros may be favoured away from Carmichaelia shrubs, so that mixed stands involving all three tree species can develop (Wardle 1980; Stewart & Veblen 1982; Reif & Allen 1988). Our field studies show that Weinmannia is much more abundant in mature stands of Carmichaelia than the other two species (Table 1) despite no differences in abundance at earlier stages of succession. Our results suggest that forest composition in the long term may be determined by initial colonization conditions, and that the relative success of the three tree species will depend on the degree of N accumulation early in succession (e.g. through facilitative effects of Carmichaelia) and on the level of shading. All three species should be able to coexist as dominants in the same stand where there is initial spatial variability in these facilitative and inhibitory effects, and clumped distributions of stems of Metrosideros and Weinmannia found in old-growth forests (Stewart & Veblen 1982) may be a legacy of spatial heterogeneity of facilitative effects during early successions. Differential effects of facilitation and inhibition on future forest trees may be widespread in successions and may be an additional mechanism for coexistence in species-rich forests.


We thank Richard Bardgett, Kate Orwin, Phil Suisted and Deborah Zanders for field help; Karen Bonner, Melissa Brignall-Theyer, Gaye Rattray, Stuart Oliver, David Purcell and Nadejda Zviaguina for greenhouse assistance; and Karen Bonner and David Whitehead for help with the nutrient solutions. Kate Dorling (University of Western Australia) conducted foliar nutrient analyses and Blair Lynch-Blosse (University of Waikato conducted panticle size analyses). David Glenny determined dicot and cryptogam specimens and Kerry Ford determined monocot specimens. Susan Wiser assisted with ordination analysis and David Coomes and Laura Fagan with analysis of the hemispherical photographs. Christine Bezar, Craig Palmer, Duane Peltzer and Peter Wardle provided helpful comments on the paper. LRW panticipated in this research while funded by a sabbatical leave from the University of Nerada, Las Yegas. This research was funded by the New Zealand Foundation for Research, Science and Technology (C09×0006 and NSOF).

Received 22 August 2000 revision received 22 March 2001