Invasive legumes fix N₂ at high rates in riparian areas of an N-saturated, agricultural catchment

Summary

1. This study examined the seemingly paradoxical proliferation of invasive, N₂-fixing broom (Cytisus scoparius) and gorse (Ulex europaeus) in N-saturated riparian areas of intensive agricultural land in Canterbury, New Zealand.

2. A field study of natural abundance δ¹⁵N suggested that broom and gorse along the Selwyn River fix approximately three times more N than they take up from soils, and are thus a potentially large source of N in the landscape. Broom N fixation rates based on mass balance calculations from a glasshouse study were similar.

3. In the controlled glasshouse study, broom grown at both c. 1× and 6.5× field NO₃⁻ supply fixed N at the same rate per unit biomass (0.061 mg N day⁻¹ g⁻¹ dry wt) over a 9-month period. Broom plants grown under the high-N supply, however, grew c. 1.6 times larger, and thus fixed more N per plant. Above-to-below-ground biomass ratios and %N in above- and below-ground pools were the same under the two levels of N supply.

4. Each broom plant in the greenhouse study contributed at least 0.02 g N year⁻¹ to soils, but leaching from the soils was surprisingly low (<2% of total plant and soil stocks) suggesting that plants less than 1 year old are not contributing substantially to high NO₃⁻ concentrations in Selwyn ground and surface water.

5. Synthesis. This study shows that both broom and gorse growing in the Selwyn riparian area are an additional source of bioactive N in this N-saturated ecosystem. Additionally, broom grows more quickly as N availability increases and therefore fixes more N per plant. This suggests a positive feedback whereby agricultural nutrient pollution leads to increased per-plant N₂ fixation in broom, and probably in gorse, given the taxonomic and physiological similarity of the species. The Selwyn is representative of a large number of New Zealand rivers with riparian zones that are dominated by invasive N₂ fixers. The likelihood that these invasive plants increase the amount of bioactive N in rivers and downstream ecosystems presents new considerations and challenges for management.

Introduction

In the last two decades New Zealand has experienced dramatic increases in land-use intensity, the area of farmland under irrigation and a c. 10-fold increase in the import and use of synthetic fertilizers (Parfitt et al. 2006). These changes contribute to growing concerns about nutrient pollution in freshwaters and nearshore areas. Exotic, fast-growing N₂-fixing species, particularly broom (Cytisus scoparius) and gorse (Ulex europaeus), have proliferated in riparian areas and have largely replaced native plants, most of which were not N₂ fixers. These invasive plants are thus potentially a major new source of bioactive N in rivers and downstream ecosystems.

This study examines the role of broom and gorse in the N dynamics of a degraded, nutrient-enriched river. The Selwyn (Canterbury Region, Fig. 1) is an alluvial river that flows through an intensively agricultural landscape before entering Te Waihora/Lake Ellesmere, a hypertrophic coastal lagoon. On average, c. 100 kg fertilizer N ha⁻¹ year⁻¹ and c. 150 kg fixed N ha⁻¹ year⁻¹ from cultivated legumes are added to soils in Canterbury (2001 values, Parfitt et al. 2006). Atmospheric deposition of reactive N is low (1–3 kg N ha year⁻¹) in New Zealand’s South Island (Galloway et al. 2004) and has not increased substantially since the late 1800s. The Selwyn catchment is, ostensibly, N-saturated. For example, intensive water quality monitoring in 2004 showed that NO₃-N concentrations increased by 10–100 fold in the c. 50 km between the foothills and the mouth of the Selwyn throughout the year. Such elevated concentrations of NO₃⁻ in ground and surface waters are a primary symptom of N excess at the catchment scale (Aber et al. 1998; Fenn et al. 1998). Increasing groundwater withdrawal for irrigation has led to reductions in the Selwyn summer discharge (McKerchar & Schmidt 2007) and may also contribute to increasing concentrations of NO₃⁻ in surface and ground waters. Although degraded, rivers such as a Selwyn still provide ecological services such as transport, transformation and sequestration of nutrients. For example, when the dissolved inorganic N load was experimentally increased by a factor of >30 over natural concentrations in a New England tidal river, the ecosystem continued to process c. 54% of the inorganic N load through biological uptake and denitrification (Drake et al. 2009).

Figure 1.  Map of the Selwyn River. Three major reaches, the upstream perennial (losing) reach, the intermittent reach, and the downstream perennial (gaining) reach are demarcated by dashed lines. Plant or water samples were collected at the numbered Rkm locations (distance in km from the river source).

The riparian vegetation of many degraded rivers in New Zealand is dominated by broom and gorse – closely related legumes of the subfamily Papilionoideae. These species were introduced to New Zealand for use as hedgerows in the early 1800s, but quickly escaped cultivation and became nuisance species. The ability of broom and gorse to fix N₂ may contribute to their success in recently disturbed alluvial riparian areas which, under natural conditions, are relatively low-N environments (e.g. Rhoades et al. 2008). N₂-fixing plants, however, are rare in the native riparian flora of New Zealand (Williams & Wiser 2004), suggesting that other life-history traits contribute to the proliferation of broom and gorse. For example, both are aggressive pioneer species that set large numbers of seeds, are adapted to frequent disturbance, drought, and inorganic soils (Pojar & MacKinnon 1994), and both have photosynthetic stems that improve water-use efficiency (Nilsen & Sharif 1983).

N₂ fixers should have an advantage in low-N environments, but the proliferation of N₂-fixing vegetation in the N-saturated landscape of Canterbury seems paradoxical. If NO₃⁻ is abundant in soils, uptake should be more efficient than N₂ fixation which requires expenditure of c. 35% more C than root uptake (Pate & Layzell 1990). Given the physiological expense of N₂ fixation and the high availability of NO₃⁻ in the Selwyn riparian area, one would expect plants in this habitat to derive N from soils. If the success of broom and gorse is based on their ability to fix N₂, these plants should lose their competitive advantage and be replaced by non-N₂ fixers as the availability of NO₃⁻ increases.

N₂ fixers are the world’s largest natural source of bioactive N and they play a critical role by regulating N availability in early successional stages of ecosystem development (Chapin, Matson & Mooney 2002). A review of regulation of N₂ fixation cited evidence that it is influenced by N availability (Hartwig 1998), while other studies have shown no effect of N supply on N₂ fixation rate (e.g. Binkley, Senock & Cromack 2003). Few clear patterns are evident, but, in general, NO₃⁻ tends to inhibit fixation more than NH₄⁺, and NO₃⁻ is very abundant in the Selwyn system.

The continued competitiveness of broom and gorse under high-N conditions may be at least partially explained if they were facultative N₂ fixers, fixing less N as soil N availability increases. It is unknown, however, whether broom and gorse are obligate or facultative N₂ fixers. Many N₂ fixers appear to use an obligate strategy at natural N concentrations, especially in temperate zones (e.g. Menge & Hedin 2009), but frequently disturbed, high-light conditions such as those in the Selwyn should favour a facultative strategy according to modelled cost–benefit analysis (Menge, Levin & Hedin 2009).

Ultimately, the contribution of N₂-fixing riparian species to N dynamics in the nutrient-polluted streams of Canterbury is unknown. This study was designed to describe and quantify N uptake and contribution by broom and gorse. I used two approaches – a field study examining N isotope composition of broom, gorse, surface water and ground water, and a controlled glasshouse mass balance study in which broom seedlings were grown under high and low-N supply – to answer three questions:

• 1
   Do broom and gorse function as a source or a sink of N in the Selwyn River system?
• 2
   If broom and gorse are fixing N₂, do fixation rates decrease as NO₃⁻ concentrations in ground and surface water increase downstream, and do growth rates, above-ground : below-ground biomass ratios, or %N in plants change with NO₃⁻ availability?
• 3
   If broom and gorse are fixing N₂, what is their annual contribution of N to riparian soils and leaching?

Materials and methods

Study Site

The Selwyn River originates in the foothills of New Zealand’s Southern Alps, and flows east for 58 km through the alluvial Central Plains of Canterbury (Fig. 1). It drains a catchment of 97 354 ha and is the largest tributary of Lake Ellesmere/Te Waihora, a hypertrophic coastal lagoon. In recent years the middle reaches of the Selwyn have dried completely during summer resulting in three distinct reaches: a 10-km perennial-flow reach from the headwater to where the river flows onto the plains, a 43-km intermittent reach which is frequently dry, and a 15-km perennial-flow groundwater-fed gaining reach near the river mouth (Larned et al. 2008; Fig. 1). Monthly monitoring during 2004 showed a consistent >10-fold increase in stream water NO₃⁻ concentrations between the headwaters (always <0.3 mg L⁻¹) and the mouth (always >5.0 mg L⁻¹).

The Selwyn’s riparian flora is dominated by exotic plants: broom, gorse, willow (Salix spp.), lupin (Lupinus spp.) and non-native grasses comprise >95% of vegetation cover (A. Hendry & D. C. Drake, unpubl. data). Only a few scattered native species such as willow weeds (Epilobium sp.) and mat-forming cushion plants (Raoulia sp.) persist. The average density of broom and gorse in the riparian area was 5.5 plants m⁻², and the mean age was 4.5 years with no plants older than 7 years (A. Hendry and D.C. Drake, unpubl. data). For the purposes of this study, the riparian area is defined as all of the area surrounding the channel that is not actively farmed – this generally extends 50–500 m from the active channel, and is maintained in a disturbed state by channel migration.

Field Methods

Broom and gorse leaves were collected from seven sites [river kilometre (Rkm) 3, 7.4, 13.3, 38, 49, 55.4 and 59.5, Fig. 1] in September 2007 (spring). It is in spring when the highest concentration of N tends to be found in the leaves of most deciduous trees (e.g. Chapin & Kedrowski 1983) and therefore collection at this time should minimize the effect of seasonal N isotope fractionation resulting from N resorption. New growth foliage was collected from at least five different plants at each of two locations [near (<5 m) and far (>40 m) from the active channel] at each of the seven sites. Leaves were dried for at least 48 h at c. 100 °C and then ground to a fine powder. Equal weights of ground leaf material from the 5+ plants at each location were composited for δ¹⁵N analysis.

Surface water was collected for δ¹⁵NO₃⁻ and NO₃-N determination in the late winter (September) 2007 at Rkm 3, 13.3 and 59.5, and mid-summer (December) 2008 at Rkm 3, 13.3, 49 and 59.5 (Fig. 1). Groundwater from a depth of c. 6 m was collected for δ¹⁵NO₃⁻ and NO₃-N determination from one well at Rkm 12 in November 2007 and three wells at Rkm 12, 40 and 58 in December 2008.

Water was filtered through combusted GF/F filters and frozen in acid-washed bottles until analysis. Surface and ground water δ¹⁵NO₃⁻ was determined using protocols developed for the Lotic Inter-site Nitrogen eXperiment (LINX; http://www.biol.vt.edu/faculty/webster/linx/). Briefly, NO₃⁻ was isolated from water samples by driving off NH₄⁺ by boiling water samples in an open container with MgO (conversion to gaseous NH₃). Boiling reduced the sample volume to c. 100 mL. Devardas alloy, salt, additional MgO and acidified GF/F filter material sealed in waterproof but gas-permeable Teflon tape were added to the samples, and bottles were quickly sealed. The sealed samples were incubated for c. 48 h at 70 °C, and then for 14 days on a shaker table for c. 30 °C. Devardas alloy reduces NO₃⁻ to NH₄⁺, which then is converted to gaseous NH₃ and is adsorbed to the acidified filter material in the head space of sample containers. After incubation, the filter material was removed from Teflon pouches and stored in a desiccator for at least 5 days prior to shipping for analyses. All isotope samples were shipped to the University of California, Davis, Stable Isotope Facility for determination of δ¹⁵N and δ¹³C.

Glasshouse Methods

Study design

Only broom was used in the glasshouse study. All N inputs, standing stocks and outputs except gaseous losses were quantified for 40 different plants and their respective pots of soil between 2 June 2008 and 17 March 2009. This included: above- and below-ground broom biomass and N pools, extractable soil N, N fertilizer added to each pot, water volume and the cumulative mass of glasshouse tap water N (determined from concentrations of NH₄⁺ and NO₃⁻) added to each pot, and NH₄⁺, NO₃⁻ and total N (TN) in leachate from each pot. Because soils contained almost no organic matter (only a trace was detected), total soil N was not measured.

On 1 June 2008 eighty-five 10-L plastic pots, 25 cm in diameter, were filled with newly deposited river gravel and sand (henceforth ‘soils’) collected from Rkm 38–39 of the Selwyn. These were transported to a glasshouse at the University of Canterbury, Christchurch. The following day, c. 90 broom seedlings (above-ground height <30 cm, less than 2 months old) with their entire root systems were carefully excavated from within c. 500 m of the soil collection area. Sixty seedlings were planted, one per pot, and a random selection of 20 seedlings were kept for initial biomass and δ¹⁵N determination (described below). Fifteen pots were not planted and were used as controls for determination of soil N accumulation and leaching. Seedlings were allowed to acclimatize for 1 week before any fertilizer was added, and several seedlings that appeared to be unhealthy were replaced. An automatic drip irrigation system was used to water plants and the control pots individually. The volume of water added over each c. 2-week period was estimated from two hoses that drained into plastic jugs.

The low-N (tap water sourced from local groundwater) and high-N (tap water + fertilizer) treatments were loosely based on N concentrations observed in the Selwyn River, although it was not possible to measure the bulk flow of groundwater to quantify root exposure to NO₃⁻ in the field. Each week 5.0 mg N as KNO₃ and 1.0 mg P as K₂HPO₄ was added to ‘high-N’ plants, and 1.0 mg P only was added to the ‘low-N’ plants. Fertilizer δ¹⁵N was +2.98‰. Low-N plants received only N dissolved in tap water, and tap water δ¹⁵N was not determined.

Leachate from all plants was collected in DI-rinsed plastic trays under each pot. Throughout most of the 9-month experiment the leachate trays contained only a small amount of water or were dry, but on two dates early in the experiment several trays filled with water, which was collected and frozen for later analysis. Glasshouse tap water was collected approximately once per month for N concentration analyses.

Broom plants were destructively harvested 2, 4, 7 and 9 months after planting. On each date five high-N plants, five low-N plants and three unplanted controls were harvested. Soils were sieved to retrieve all root material. Leachates containing N that seeped out of pots were collected by scrubbing trays out with DI water. The wash water volume was measured and a subsample was used for total N, NO₃⁻ and NH₄⁺ determination. Soils in each pot were homogenized and subsamples of 100 g (particles <10 mm diameter) were collected and refrigerated until analysis.

Laboratory Methods

Above- and below-ground portions of each plant were washed thoroughly with tap water, dried at 100 °C and weighed. From each pot, 20-g soil subsamples were extracted with 100 mL 1M KCl. Samples in KCl were shaken vigorously and allowed to stand overnight. The following day the samples were re-shaken and allowed to settle, then filtered through pre-leached, 10-cm diameter GF/F Whatman filters and frozen for nutrient analyses. Soil subsamples were weighed for loss-on-ignition and bulk density determination.

For the first three harvests, entire above- and below-ground fractions of each plant were ground for δ¹⁵N determination so that results would be representative of the entire above- or below-ground pool. At the fourth harvest above-ground pools had become very large, so each plant was divided roughly into quarters radially, and one quarter was ground for isotope analysis.

Quantitative analyses

N isotope model

The two major N sources available to riparian broom and gorse in the Selwyn system are (i) fixed N₂ and (ii) dissolved N in soil water. The δ¹⁵N values of the two sources were used as endmembers to estimate source contribution to plant N using a double endmember mixing model (eqn 1). NO₃⁻ comprised 97.0–99.8% of dissolved inorganic nitrogen in the Selwyn system during monthly sampling in 2004 (S. T. Larned, unpubl. data). The high permeability of Selwyn gravels suggests that shallow ground water (within 6 m of the surface) and surface water are intimately connected – i.e. dissolved NO₃⁻ in surface water is representative of the N that riparian plants access through uptake. The isotopic endmember value for N₂ fixation was taken from a controlled glasshouse study which determined that broom grown in N-free soils (forcing 100% N fixation) had leaf δ¹⁵N of −0.64‰ (Watt et al. 2003). Equation 1 was used to calculate %N fixed from leaf δ¹⁵N (δ¹⁵N_leaf), mean δ¹⁵NO₃⁻ in ground or surface water collected from the site nearest to plant collection, and the 100% broom fixation value (δ¹⁵N_fix).

• (eqn 1)

The mass of N fixed by each plant upon destructive harvest in the glasshouse experiment was calculated using mass balance (eqn 2). Initial N pools and NO₃-N in water and fertilizer (N_fert) were subtracted from the sum of all N pools at the end of the experiment:

• (eqn 2)

Paired t-tests were used to determine whether field δ¹⁵N values differed between species or with distance to the river (<5 m vs. >40 m from the active channel) and to compare plant size and N stocks over the course of the glasshouse experiment. A single factor anova and Tukey’s multiple comparison of δ¹⁵N values pooled by site (each site n = 4) was used to determine whether plant δ¹⁵N differed along the length of the river and plant characteristics varied over time in the glasshouse experiment.

Results

Field Study

N₂ fixation estimates based on leaf δ¹⁵N of plants growing in the field suggest that broom and gorse in the Selwyn riparian area fix 66–88% of their foliar N (Table 1). All broom and gorse δ¹⁵N values from the field study were between the dissolved NO₃⁻ and 100% N fixation endmembers (eqn 1), lending further support to the assumption that these were the main sources of N in the Selwyn riparian area. The δ¹⁵N of both species was significantly lower (among sites anovaP < 0.01) in the middle reaches (Rkm 13.3, 37.9 and 48.9, Fig. 1, Table 1) relative to upper and lower reaches (Rkm 3, 7.4, 55.4 and 59.5, Tukey’s multiple comparisons), and both broom and gorse δ¹⁵N values were statistically equal within species between the upper and lower reaches.

Table 1.   δ¹⁵N and resulting estimates of %N (±SE) fixed in broom and gorse foliage along the length of the Selwyn River. Values for broom and gorse are averages of two samples (near and far from the channel), each composited from five plants – i.e. averages are representative of 10 plants. % N fixed was calculated based on N fixation and dissolved NO₃⁻ endmembers (eqn 1). Groups a and b were determined using an anova and Tukey’s multiple comparison of broom and gorse δ¹⁵N values pooled by site (each site n = 4)

River km from headwaters	Average δ^15N (‰)		Foliar N fixed (%)
	Broom	Gorse	Broom	Gorse
3^a	1.0 ± 0.09	1.3 ± 0.17	74.9 ± 6.7	69.5 ± 11.8
7.4^a	1.3 ± 0.04	0.8 ± 0.04	68.2 ± 2.7	76.5 ± 3.05
13.3^b	0.8 ± 0.37	0.6 ± 0.09	76.8 ± 28.4	80.5 ± 7.2
37.9^b	0.3 ± 0.05	0.1 ± 0.03	86.3 ± 4.3	88.4 ± 2.7
48.9^b	1.0 ± 0.27	0.4 ± 0.02	74.7 ± 20.1	84.5 ± 1.7
55^a	1.8 ± 0.51	1.3 ± 0.06	66.7 ± 34.0	73.4 ± 4.4
59.5^a	1.7 ± 0.04	1.4 ± 0.25	67.7 ± 2.7	72.4 ± 18.1

Average gorse δ¹⁵N was 0.83‰– slightly but significantly lower than the average broom δ¹⁵N of 1.12‰ (two-tailed, paired t-test P = 0.04). There was no statistical difference in broom or gorse δ¹⁵N with distance from the river (within 5 m and >40 m from the active channel), and there was no pattern in broom or gorse δ¹³C measured in a subset of 15 samples – all values were within 0.5‰ of −25.5 ‰. Average foliar %N was 3.69 ± 0.52 for broom and 2.46 ± 0.48 for gorse.

In September (late winter) 2007, Selwyn River water δ¹⁵NO₃⁻ values ranged from 5.53‰ to 6.63‰, and groundwater δ¹⁵NO₃⁻ was within this range at 6.03‰. The standard deviation of five dissolved NO₃⁻ standards was 0.77‰. Surface water δ¹⁵NO₃⁻ in December 2008 (mid-summer) was also similar along most of the length of the river (3.05–3.30‰), and groundwater δ¹⁵NO₃⁻ at Rkm 40 and 58 was similar to surface water (2.13‰ and 3.09‰, respectively), although at Rkm 13.3 groundwater δ¹⁵NO₃⁻ was 8.25‰. On both dates surface water NO₃-N concentrations increased from 0.3 to 0.5 mg L⁻¹ in the headwaters to 6.2–7.1 mg L⁻¹ in the lowest reaches. Groundwater NO₃-N concentration also increased downstream from 0.9 mg L⁻¹ at Rkm 12–6.3 mg L⁻¹ at Rkm 58.

Glasshouse Study

N pools and conditions at the onset of the experiment

Soils collected from the Selwyn River riparian area contained <1% organic matter by weight. Nutrient concentrations were also very low; extractable (plant available) NH₄⁺ was <0.001 mg N g⁻¹ dry soil, extractable NO₃⁻ was generally 0.001–0.003 mg N g⁻¹ dry soil, and extractable PO₄³⁺was usually below the detection limit of c. 0.0001 mg P g⁻¹ dry soil. The mean, initial soil N standing stock in each pot was 6.3 ± 2.4 mg (Table 2).

Table 2.   Nitrogen pools, additions and a mass balance calculation of N fixation under high and low-N supply after 7 and 9 months of growth in the glasshouse. Plant and soil values are mean values of five replicates ± standard errors. Water N concentrations were measured in single samples over time and have no error estimates. Leachate samples were pooled for analysis and have no error estimates

		7 months (January 2009)		9 months (March 2009)
		High N	Low N	High N	Low N
	Treatment	N (mg)	N (mg)	N (mg)	N (mg)
N supplied to each pot + plant	Seedlings	10.8 ± 1.1	10.8 ± 1.1	10.8 ± 1.1	10.8 ± 1.1
	Soil	6.3 ± 2.4	6.3 ± 2.4	6.3 ± 2.4	6.3 ± 2.4
	Water	28.1	28.1	32.7	32.7
	Fertilizer	145	0	190	0
N pools at harvest	Plants	717.4 ± 75.6	287.4 ± 46.5	863.9 ± 59.9	438.8 ± 12.5
	Soil	30.2 ± 4.5	22.7 ± 6.3	68.5 ± 9.1	53.0 ± 12.3
	Leached N	5.4	5.5	13.9	5.7
N fixed per plant (pools at harvest – supplied; mg)		562.8 ± 12.3	270.3 ± 69.2	706.6 ± 95.5	447.8 ± 50.5

The mean dry weight of whole broom seedlings collected in July 2008 was 0.52 ± 0.08 g, and the average N content was 0.0108 ± 0.0012 g (c. 2% of dry wt). Above-ground : below-ground seedling biomass was c. 1:1.

N added over the experiment

Glasshouse tap water contained almost no NH₄-N throughout the experiment (<0.001 mg L⁻¹) while NO₃-N increased over the summer and autumn from 0.6 to 3.2 mg L⁻¹. Monthly average NO₃-N concentrations multiplied by the volume of water added to each pot comprised the low-N treatment: 4.0 mg NO₃-N had been added from tap water by the first destructive harvest, 11.9 mg by the second, 28.1 mg by the third, and 32.7 mg by the fourth (Table 2). High-N treatments received tap water N and were also fertilized with 5 mg NO₃-N per week.

Broom growth and N mass balances

By the end of the 9-month glasshouse study, broom under high-N supply had grown 1.6 times larger than broom under low-N supply (Table 2, Fig. 2a). Above-ground : below-ground biomass increased from an initial value of c.1 to an average of c. 1.8 by the first destructive sampling (3 months), and remained at that level in both treatments until the end of the experiment (Fig. 2b). No consistent pattern in root or shoot δ¹⁵N was attributable to N supply (Fig. 3). %N was similar between treatments with the exception of the January (mid-summer) 2009 harvest in which the average N content of below-ground tissues in low-N plants was significantly lower than in high-N plants (2.17 ± 0.24% vs. 1.66 ± 0.24% respectively). Over the experiment, the average N content of stems was 2.37 (± 0.26) % and below-ground tissues was 1.92 (± 0.23) %.

Figure 2.  Average (±SE) biomass (a) and above-ground:below-ground biomass ratios (b) of broom grown under high () and low () N supply. *Significant differences (t-test P < 0.05) between N treatments and differences over time (anovaP < 0.05) in biomass ratios are denoted by a,b.

Figure 3.  δ¹⁵N of above-and below-ground pools (±SE) in low-N () and high-N treatments ().*Significant differences (t-test P < 0.05) between treatments.

N fixation per gram of plant dry weight (gdw) derived from mass balances was remarkably similar in the high- and low-N treatments at 0.092 (±0.017) and 0.091 (±0.021) mg gdw⁻¹ day⁻¹ after 7 months and 0.061 (±0.007) and 0.062 (±0.006) mg gdw⁻¹ day⁻¹ at 9 months (Table 3). Broom δ¹⁵N in the high- and low-N treatments (Fig. 3) were also similar throughout the experiment providing additional evidence that N₂ fixation rates per unit biomass were similar irrespective of N supply. I note that gaseous losses were not measured, and would result in higher estimated rates of N fixation. Average broom δ¹⁵N in the glasshouse experiment (excluding initial values) was −1.0 (±0.7) ‰, notably lower than the 100% fixation value from Watt et al. (2003) of −0.64‰. Fertilizer δ¹⁵N was +2.98‰ and cannot be responsible for the low broom δ¹⁵N, suggesting that the N fixation endmember here may be lower than the value determined by Watt et al. (2003).

Table 3.   Mean dry weights, N fixation rates and % of N fixed in broom (±SE) after 7 and 9 months of growth under controlled glasshouse conditions

	7 months (January 2009)		9 months (March 2009)
	High N	Low N	High N	Low N
Mean dry weight whole plant (g)	28.19 ± 1.91	13.67 ± 4.60	41.4 ± 3.69	25.8 ± 1.35
N fixed (mg day^−1)	2.58 ± 0.39	1.24 ± 0.24	2.52 ± 0.26	1.60 ± 0.14
N fixed per g plant dry weight (mg day^−1)	0.092 ± 0.017	0.091 ± 0.021	0.061 ± 0.007	0.062 ± 0.006
% N fixed estimated using mass balance	74.7%	85.7%	74.7%	90.0%

Soils subjected to all treatments (including unplanted controls) accumulated N, but soils in the high-N treatment accumulated significantly more N than those in the low-N treatment (Table 2, t-test P < 0.05). At the end of the experiment soils in high-N treatment contained, on average, c. 4.7 ± 0.62 g N m⁻³ (c. 10.9×) more extractable N than they did initially. Soils in the low-N treatment contained 3.5 ± 0.80 g N m⁻³ (c. 8.4×) more, and unplanted control soils contained 2.5 ± 0.33 g N m⁻³ (c. 6.1×) more than they did initially. Each young plant, therefore, contributed c.1.3–2.9 g N m⁻³ year⁻¹ to riparian soils, and plants contributed more N to soils as N load increased.

Soil leachates from pots with plants contained <2% of plant + soil N standing stocks. Thus plants and soils were retaining most of the N added to the pots and produced by N fixation. Leaching rates in the glasshouse experiment were c.1.45 g N m⁻² year⁻¹ in the control soils (which received only N in tap water), 0.76 g N m⁻² year⁻¹ in the low-N treatment soils, and 1.39 g N m⁻² year⁻¹ in the high-N treatment soils, suggesting that the presence of plants decreases N leaching rates. A statistical comparison is not possible, however, because samples were composited for analyses.

Discussion

Only broom was used in the glasshouse study. It is closely related to gorse, it exhibited similar field δ¹⁵N values to gorse, and the two species grow interspersed and are equally abundant in the Selwyn system. I therefore apply findings of the glasshouse study to gorse, and in this discussion assume that both species are acting similarly in the Selwyn environment.

A positive feedback between N supply and per-plant and per-unit area N₂ fixation

Introduction of new N sources (such as agricultural pollution) to an ecosystem frequently reduces the competitive ability of legumes and inhibits symbiotic N₂ fixation (Hartwig 1998; Menge, Levin & Hedin 2009). Under assumptions of self-regulation or facultative N₂ fixation I expected to observe reduced rates of N fixation in plants grown under higher N supply, but this was contradicted by both field and glasshouse experiments. Instead, under the increased N supply, N fixation rates per unit biomass remained the same while the plants grew faster, resulting in increased per-plant N₂ fixation. This, coupled with high rates of N₂ fixation measured in the field, suggests that broom and gorse in the Selwyn riparian are N-limited even under the current conditions of high NO₃-N availability. Under the range examined here, a 10% increase in NO₃⁻ supply resulted in an additional 34.5 mg of N₂ fixed by each plant per year. Agricultural nutrient pollution is therefore almost certainly stimulating growth of broom and gorse and resulting in an increase in the amount of N₂ fixed in the ecosystem per unit time.

Other factors that may limit the growth of broom and gorse in the Selwyn riparian include herbivory, limitation by nutrients other than N, and light and water availability. Herbivory on broom and gorse is very light as these species are unpalatable to livestock. There is also little visible evidence of insect damage (D.C. Drake, pers. obs.). Phosphorus is the most likely alternative limiting nutrient, but P fertilizers are used heavily in New Zealand pasture agriculture (e.g. they comprise c. 91% of total P concentrations in fresh water; Gillingham & Thorrold 2000), and P was added to broom in the glasshouse to simulate agricultural pollution. This and the fact that broom and gorse fix such a large fraction of the N in their tissues strongly suggest N-limitation in the Selwyn riparian. Light may limit growth in the field where plant densities are high and plants are older and larger. The density of broom in the glasshouse, however, was c. 15 plants m⁻², or about three times the average density in the Selwyn riparian (A. Hendry & D. C. Drake unpubl. data) and plants showed no sign of light limitation. I conclude that light limitation is rare or unlikely during at least the first year of growth in the field. Broom and gorse are phreatophytes and are probably not severely limited by water availability. This is supported by broom and gorse δ¹³C – severe water stress would probably be indicated by higher δ¹³C in the drier, intermittent reaches, but this was not observed.

The N₂ fixation strategy in the Selwyn River environment

High channel migration rates in many of New Zealand’s alluvial rivers maintain riparian areas in an early successional state, populated mainly by young, fast-growing, invasive legumes. Frequently disturbed environments should favour facultative over obligate N₂ fixers as long as the costs of being facultative (regulation of physiology and mechanisms required to sense and respond to N availability) are relatively low (Menge, Levin & Hedin 2009). Broom, however, used an obligate strategy under a wide range of NO₃-N supplied in the glasshouse study, and field results suggest obligate N fixation in both species. It is unknown whether a facultative N-fixing strategy might be triggered as N availability continues to increase or as the relative availability of P, light, CO₂ or other potentially limiting resources changes.

This study raises a new concern about the interplay of invasive N-fixing species and nutrient pollution, and is of particular relevance given the recent, anthropogenic doubling (or more) of bioactive N availability at a global scale (Vitousek et al. 1997). The sequestration of nutrients by vegetation is used to manage nutrient pollution and is an important part of the ability of natural river ecosystems to maintain water quality. This study, however, shows that invasive broom and gorse in early successional riparian areas are likely to be a major, additional N source under nutrient-polluted conditions. These obligate N₂ fixers can interact with N pollution in a positive feedback, with increased N availability from agricultural pollution or other sources resulting in higher plant growth rates, and a net increase in the amount of N₂ fixed at the ecosystem scale.

Acknowledgements

I sincerely thank Ann Hendry, Jo Bind, Clive Howard-Williams, Trevor Partridge, Dave Arscott, David Conder, John Carter, Greg Kelly, Rose Sephton-Poultney, the Editors at Journal of Ecology and several anonymous reviewers for their help and input through the process of designing, conducting, and writing up this study. This research was supported by NIWA Capability Funds CPAB075 and CPAB085.