Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
1The uptake of free amino acids by the grass Puccinellia phryganodes was investigated in soils of an Arctic coastal salt marsh, where low temperatures and high salinity limit inorganic nitrogen (N) availability, and the availability of soluble organic N relative to inorganic N is often high.
2Following the injection of 13C15N-amino acid, 15N-ammonium and 15N-nitrate tracers into soils, rates of soluble nitrogen turnover and the incorporation of 13C and 15N into plant roots and shoots were assessed. Chloroform fumigation-extraction was used to estimate the partitioning of labelled substrates into microbial biomass.
3Free amino acids turned over rapidly in the soil, with half-lives ranging from 8.2 to 22.8 h for glycine and 8.9 to 25.2 h for leucine, compared with 5.6 to 14.7 h and 5.6 to 15.6 h for ammonium and nitrate, respectively. 15N from both organic and inorganic substrates was incorporated rapidly into plant tissue and the ratio of 13C/15N incorporation into plant tissue indicated that at least 5–11% of 13C15N-glycine was absorbed intact.
4Microbial C and N per unit soil volume were 1.7 and 5.4 times higher, respectively, than corresponding values for plant C and N. Plant incorporation of 15N tracer was 56%, 83% and 68% of the comparable incorporation by soil microorganisms of glycine, ammonium and nitrate ions, respectively.
5These results indicate that P. phryganodes can absorb amino acids intact from the soil despite competition from soil microorganisms, and that free amino acids may contribute substantially to N uptake in this important forage grass utilized by lesser snow geese in the coastal marsh.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Plant amino acid uptake may play an important role in the N dynamics of goose-grazed Arctic coastal marshes, where high salinity, in addition to low temperatures, limits N mineralization and inorganic N uptake (Wilson & Jefferies 1996). However, amino acid turnover in saline terrestrial systems and their uptake by salt-marsh plants in situ have not been examined. In goose-grazed coastal marshes, the regrowth of grass following defoliation by geese is dependent on droppings, yet estimated amounts of soluble inorganic nitrogen in faeces are inadequate in some cases to account for N sequestered in new plant growth (Hik et al. 1991). Based on bulk soil solution concentrations and the results of a continuous flow nutrient addition experiment, amino acid uptake by the salt-marsh grass, Puccinellia phryganodes, was estimated to be as high as 57% of the uptake of ammonium ions (Henry & Jefferies 2002). Thus free amino acids may provide an important source of nitrogen for the regrowth of salt-marsh plants grazed by geese, provided that the plants can compete effectively with soil microorganisms for amino acids.
In the present study, heavy-isotope tracers were used to characterize the partitioning of soluble organic and inorganic N between plants and microorganisms in soils of an Arctic salt marsh situated on the south-west Hudson Bay coast. Dual-labelled 13C15N-amino acids, 15N-ammonium and 15N-nitrate were injected into soil to estimate rates of N turnover and to compare rates of intact amino acid and inorganic N uptake by P. phryganodes, the dominant goose forage grass. Chloroform fumigation-extraction was used to estimate microbial immobilization of N substrates.
Materials and methods
In situ experiments were conducted in a grazed Arctic coastal salt marsh at La Pérouse Bay, Wapusk National Park, located approximately 30 km east of Churchill, Manitoba, Canada (58°43′ N, 94°26′ W), on the south-west coast of Hudson Bay. The marsh can be subdivided into an intertidal zone, that is tidal in late summer and autumn, but not from snow melt (mid-June) until late July in most years, and a supratidal zone that is rarely flooded with sea water. The vegetation of the intertidal marsh is dominated by the grass, Puccinellia phryganodes (Trin.) Scribn. and Merr., and the sedge, Carex subspathacea Wormskj which form discontinuous swards because of early spring grubbing by lesser snow geese that leads to sward destruction (Jefferies 1988a,b). Net above-ground production of intact swards has been estimated to be between 100 and 150 g m−2 year−1 (Cargill & Jefferies 1984b; Hik & Jefferies 1990). Further inland, the supratidal zone also is dominated in low-lying areas by swards of P. phryganodes and C. subspathacea, but low shrubs, predominately Salix brachycarpa Nutt., and grasses, especially Festuca rubra L. and Calamagrostis deschampsioides Trin., colonize frost-heave sites. Nomenclature follows Porsild & Cody (1980).
Soils were classified as Regosolic Static Cryosols (Agriculture Canada Expert Committee on Soil Survey 1987; Wilson & Jefferies 1996). Soils in the intertidal zone have a greyish mineral horizon of marine sediment above which lies the top 3–4 cm that is rich in organic material. Soils are often highly reduced (Eh−50 mV) in spring, especially where they are close to drainage channels (Wilson & Jefferies 1996). In the supratidal marsh, a mineral base is covered by 5–7 cm of dark brown-black highly humified organic material. Over the growing season from late May to late July, mean amino acid concentrations in soil solutions of the intertidal zone ranged from 32 to 45 µm, while ammonium and nitrate concentrations ranged from 55 to 160 and 10 to 31 µm, respectively (Henry & Jefferies 2002). Amino acids present at high concentrations are alanine, proline, glutamic acid, leucine, tyrosine, gamma amino-butyric acid and glycine. Soil pH averages 7.12 ± 0.04 (n = 24) in both the intertidal and supratidal marshes. The salinity of extracted soil solutions can rise to as high as 40 g Na+ L−1 in late summer in supra- and intertidal sites devoid or nearly devoid of vegetation (Iacobelli & Jefferies 1991). However, beneath intact swards the salinity of the bulk soil solution in summer is at, or less than that, of sea water (c. 12 g Na+ L−1) (Srivastava & Jefferies 1995; Wilson & Jefferies 1996). Turnover and uptake experiments were conducted in two contrasting years 2000 and 2001, which were character-ized by late and early springs, respectively. Temperature, soil moisture and plant biomass data that correspond with the sampling dates are provided in Table 1.
Table 1. Summary of temperature, soil moisture and plant biomass data (n = 6) for dates on which N turnover and uptake experiments were conducted in the intertidal marshes at La Pérouse Bay, Manitoba
Degree days calculated by taking the sum of daily temperature maxima above 0 °C since 1 June
Soil moisture (% by dry mass)
Root biomass (g m−2 of soil core)
Shoot biomass (g m−2 of soil core)
One pair of soil cores per treatment was collected from each of three intertidal and three supratidal sites on 25 June and 12 July 2000, and 17 June and 9 July 2001. The distance between sites was a minimum of 100 m. Cores were collected using 7.6 cm diameter × 6 cm deep galvanized steel rings, then wrapped in plastic film and transported back to the field laboratory at La Pérouse Bay. All cores were injected in a 7-point hexagonal pattern with a total of 7 mL of 1 mm solutions of 15NH4Cl, K15NO3, 15N-glutamic acid, 13C15N-glycine or 15N-leucine, using a 15-cm double-sideport spinal syringe (Popper and Sons, Hyde Park, New York). The amount of substrate added to each core was approximately equal to the background concentrations of free amino acids, ammonium or nitrate in the soil solution and soil moisture was increased by < 5%. Syringes were raised while injecting to evenly distribute the solution at all depths of the core. Immediately, soil from one core from each pair was mixed and a 15-g subsample was extracted in 75 mL of chilled 2 m KCl on ice for 1 h, then filtered through Whatman GF/A filter paper and frozen. A second subsample was collected to determine the fw/dw ratio. The second core from each pair was incubated at soil temperature for 24 h prior to mixing and extraction, as outlined above.
Ammonium (15) was recovered from KCl extracts using a modified diffusion technique (Stark & Hart 1996). Ammonium traps were constructed by pipetting 5 µm of 2.5 m KHSO4 onto 7 mm disks of KCl- and deionized water-leached Whatman no. 1 filter paper and by encasing the disks in folded strips of Teflon thread sealing tape (Flou-Kem, Scienceware, Pequannock, New Jersey). For each sample, one trap and 0.2 g MgO were added to 8 mL of KCl extract and 20 mL of a 0.14-m NH4Cl/2 m KCl carrier solution placed in a sealed 75 mL plastic specimen container. Solutions were mixed daily and diffused for 6 days, after which the traps were dried in a dessicator over H2SO4 and analysed for their 15N/14N ratios and N contents using an Isochrom continuous flow stable isotope mass spectrometer (Micromass) coupled to a Carla Erba elemental analyser (CHNS-O EA1108). Isotopic analyses were performed by the University of Waterloo Environmental Isotope Laboratory, Waterloo, Ontario.
Nitrate (15) was recovered from KCl extracts using an extension of the diffusion method (Hart et al. 1994). First, 0.2 g MgO was added to open containers of extract to volatilize NH3. After 6 days, containers were sealed and provided with ammonium traps and 0.4 g Devarda's alloy to reduceto , which was diffused as described above. 15N-amino acids were recovered by adding 0.4 g of Devarda's alloy to alkaline persulphate oxidations of KCl extracts (see below) and by reducingto to determine total 15N (NaOH was substituted for MgO). 15and 15were subtracted from this total to give an estimate of 15N amino acids. Blanks, N standards and diffused and non-diffused 15N standards (the latter applied directly to filter paper disks) were included for all diffusion analyses (the coefficient of variation of standards was < 3%).
Roots and shoots of P. phryganodes were collected from 15, 15and13C15N-glycine-injected cores used in the N turnover experiments on 12 July 2000 and 9 July 2001. Plant materials were washed three times in 0.5 mm CaCl2 to remove isotope in the Donnan free space, dried, ground, then analysed for 13C and 15N content by the use of mass spectrometry. Separate subsamples were collected to determine the dry weight of roots and shoots per unit area of soil core (Table 1).
microbial biomass and15n incorporation
Microbial C and N were estimated on 6 June, 4 July and 26 July, 2001 using a modified chloroform-fumigation extraction method developed for wetland soils (Witt et al. 2000). Two soil cores were collected from each of the three intertidal and three supratidal sites used in the N-turnover experiment. Two 70 g subsamples were collected from each core and cleared of live roots to minimize contamination. The first subsample was immediately extracted in 140 mL of 0.5 m K2SO4 for 1 h, filtered through Whatman GF/A filter paper, then frozen. The second subsample was mixed in a sealed 250 mL glass Schott bottle with 2 mL of ethanol-free chloroform. Following incubation in the dark at room temperature for 24 h, the bottles were opened and allowed to evaporate for 30 min, then extracted as described above.
Total extractable N was measured following alkaline persulphate oxidation (Cabrera & Beare 1993). Briefly, 1 : 1 mixtures of samples and oxidizing reagent (25 g K2S2O8 and 15 g H3BO4 dissolved in 50 mL of 3.75 m NaOH and made up to 500 mL with deionized water) were pipetted into glass tubes sealed with a Teflon-lined screw cap and autoclaved for 30 min at 120 °C. Nitrate concentrations were then determined colormetrically using an auto analyser following reaction with Marshall's reagent after reduction to nitrite (Pulse Instrumentation Ltd. Saskatoon, Saskatchewan). Cadmium was used to reduce nitrate to nitrite (Keeney & Nelson 1982).
Extractable C was determined using a dichromate method (Nelson & Sommers 1996). Soil extracts (1 mL) were pipetted into boiling tubes and treated with 1 mL of 0.07 m K2Cr2O7, 2 mL of 98% H2SO4, and 1 mL of 88% H3PO4. Samples were mixed and digested at 150 °C for 30 min. After cooling, samples were titrated with 0.01 m Fe(NH4)2(SO4)2·6H2O in 0.4 m H2SO4 using 120 µL of 4.7 mm N-phenylanthranilic acid and 0.01 m of Na2CO3 as an indicator. A glucose standard was used for calibration, and values were adjusted for Fe2+ content (which interferes with the dichromate method), which was measured using the phenanthroline method (Loeppert & Inskeep 1996). Microbial C and N were also determined for N turnover cores injected with 15NH4Cl, K15NO3 and 13C15N-glycine on 9 July, 2001. Excess 15N content in fumigated samples was determined following alkaline persulphate oxidation.
For diffused samples, a minimum of 97% and 70% were recovered. Therefore, 15N enrichment was blank corrected using the following isotope dilution equation that does not assume complete recovery (Stark & Hart 1996):
Es = Em + Mb(Em − Eb)/Ms(1)
where Es is the corrected 15N enrichment of the sample expressed as percent abundance, Em is the enrichment measured in diffused standards expressed as percent abundance, Mb is the mass of N in the blank in grams, Eb is the enrichment of the blank expressed as percent abundance and Ms is the mass of N in the sample in grams. Half-lives of 15N loss were estimated based on concentrations of 15N in the extractable fraction of the soil at 0 and 24 h according to the equation:
t1/2 = ln 2/ln (N0/Nt)(2)
where Nt is the amount of 15N enrichment at time t in micrograms, No is the amount of 15N enrichment at time 0 in micrograms and t1/2 is the half-life in hours. Equation 2 is a rearrangement of the exponential decay function:
15N enrichment data conformed well to this decay function in pilot trials where soils were extracted at 0, 6 and 24 h. Tracer incorporation into roots and shoots was estimated by subtracting background enrichment in control samples from the total enrichment present in test samples. Extractable microbial C and N were determined by subtracting non-fumigated soil concentrations from those of fumigated soil. Correction factors of 2.64 and 2.22, as estimated by Vance et al. (1987) for a wide range of temperate soils, were used to estimate total microbial C and N from extractable C and N, respectively (cf. Witt et al. 2000). For dependent variables, site means were compared using anova followed by Tukey's multiple comparison tests to resolve pair-wise group differences. Half-life data were log10-transformed prior to analysis to improve normality. 15N and 13C uptake data did not differ significantly between sites (P > 0.7). Therefore, these data were pooled between sites to increase the statistical power of comparisons between substrate treatments.
Overall, 51–87% of 15N glycine, 48–85% of 15N leucine, 68–95% of 15N-ammonium and 62–95% of 15N-nitrate were removed from the soil solution during the 24 h incubation. In general, rates of ammonium turnover were most rapid, followed by rates of nitrate and amino acid turnover (Fig. 1). Turnover of 15N substrates was slowest on 25 June, 2000, the coldest day tested (1.9–7.4 °C) (Table 1). 15N from ammonium, nitrate and glycine was incorporated equally rapidly into plant roots and less rapidly into plant shoots, where incorporation of 15N from ammonium was highest (Fig. 2). Significant 13C enrichment was present in roots exposed to 13C15N-glycine but no 13C enrichment was detected in shoots. Overall 13C enrichment as a percentage of 15N enrichment was 5% in 2000 and 11% in 2001.
Extractable microbial C and N and the C/N ratio were generally higher in intertidal sites than in supratidal sites (Fig. 3). Extractable microbial C and N increased during the growing season, although this increase was delayed for intertidal sites. On 9 July, 2001, extractable microbial C and N per unit soil volume were 1.7 and 5.4 times higher than root C and N, respectively. Microbial 15N enrichment did not differ significantly among treatments (Fig. 2). On average, from 17 to 29% of 15N injected into soil cores was recovered in plant tissue, from 28 to 34% was recovered in microbial N and from 11 to 29% remained in the soil solution (Fig. 4). Approximately 25% of added label was unaccounted for (not shown).
Free amino acids turn over rapidly in soil primarily as a result of uptake by microorganisms and plant roots (Lipson et al. 2001; Vinolas et al. 2001). In these Arctic salt marsh soils, free amino acids turned over rapidly in summer, with half-lives ranging from 8.2 to 22.8 h for glycine and 8.9 to 25.2 h for leucine, compared with 5.6 to 14.7 h for 15N-ammonium and 5.6 to 15.6 h for 15N-nitrate. These amino acid turnover rates are within the range of 1.7–28.7 h reported for the turnover of amino acids in other Arctic, alpine and temperate soils (Hadas et al. 1992; Martens & Frankenberger 1993; Kielland 1995; Lipson et al. 2001; Jones & Kielland 2002). However, unlike in temperate sites, where microbial mineralization is largely responsible for the turnover of amino acids, it appears, based on the substantial incorporation of stable isotope tracers by plant roots, that amino acid turnover is also partly driven by plant uptake in Arctic salt-marsh soils in summer.
In addition to microbial and root uptake, soluble nitrogen is removed from the soil solution by sorption to the soil solid phase (Jones & Hodge 1999). High rates of adsorption are typically observed for cations, such as ammonium and acidic amino acids (Hodge et al. 1999; Jaeger et al. 1999). Because of their adsorption to the solid phase, these amino acids and ammonium are typically protected from loss in the soil (Lipson & Monson 1998; Vinolas et al. 2001). In contrast, neutral and basic amino acids and nitrate are susceptible to losses through runoff and drainage of the soil solution. In saline soils, however, high concentrations of sodium ions and divalent cations such as calcium and magnesium may exclude other monovalent cations from exchange sites (McBride 1989). Therefore, ammonium and acidic amino acids may behave more like anions in salt-marsh soils, resulting in steep depletion zones of ions close to the root surface in the absence of buffering from exchangeable ions attached to the solid phase.
Although 15N from 13C15N-glycine was incorporated as rapidly as 15N from ammonium and nitrate into roots of P. phryganodes in situ (Fig. 2), mean 13C incorporation was only 5% of 15N incorporation in 2000 and 11% in 2001. This low ratio of 13C incorporation suggests that a large portion of glycine may have been deaminated by soil microorganisms prior to uptake of the 15N fraction. However, the ratio of 13C/15N incorporation is expected to provide an underestimate of intact amino acid uptake as a result of plant respiratory losses of 13CO2 following the decarboxylation of amino acids or their breakdown products (Schimel & Chapin 1996). Respiratory 13C losses were minimized in the present study by employing a relatively short incubation time of 24 h and by utilizing amino acids labelled at the 2-C position which is decarboxylated less rapidly that the 1-C position (Fokin et al. 1993). Nevertheless, the carbon in the 2-C position can be respired following deamination and breakdown of the C-skeleton in the Krebs cycle (Näsholm & Persson 2001). Therefore, some 13C may have been lost to respiration within 24 h. Even when substantial respiratory losses of 13C do not occur, isotopic enrichment may be more difficult to detect for 13C than for 15N because of the high background concentration of 13C in plant tissue relative to 15N (Näsholm & Persson 2001).
The incorporation of 13C relative to 15N may also have been low as a result of the deamination of glycine by plant extracellular deaminases at the root surface. Although extracellular deaminases have not been localized on plant roots, they are present on other photosynthetic organisms such as algae (Paul & Cooksey 1979, 1981; DeBusk et al. 1981). Hypothetically, their presence on the surface of roots could explain why 13C incorporation was 50% lower than 15N incorporation for excised roots of P. phryganodes incubated in sterile hydroponic media in short-term (20 min) 13C15N-glycine uptake experiments (Henry & Jefferies 2003).
The simultaneous recovery of 13C and 15N in roots provides only indirect evidence for intact amino acid uptake. For example, if the amino acid glycine is deaminated by extracellular enzymes, both breakdown products (13C-glyoxylate and 15N-ammonium) could be absorbed independently by plant roots (Näsholm & Persson 2001). Nevertheless, results from the use of gas chromatography-mass spectrometry have verified the presence of intact 13C15N in roots following uptake by wheat (Näsholm et al. 2001). In the present study, excess 13C was recovered in roots of P. phryganodes but not in shoots. However, when glycine is used as a tracer, no relation between 13C and 15N label is expected in the shoot because transport of N from roots to shoots occurs in the form of specific amino acids, such as the amides, asparagine and glutamine.
Rates of amino acid turnover are typically low at low soil temperatures (Vinolas et al. 2001), which are associated with low plant root and microbial biomass (Fig. 3) and low rates of N transport per unit biomass (Jones 1999; Henry & Jefferies 2003). This is consistent with the recorded rates of N turnover in salt-marsh soils, which were slowest on 25 June, 2000, the coldest of the four sampling dates. However, in late fall and early spring, high soluble N concentrations are observed in salt-marsh soils (Henry & Jefferies 2002). These peaks in soluble nitrogen are probably the result of the release of soluble N from lysed roots and microbial cells during freeze-thaw cycles in fall and spring (Skogland et al. 1988; Hobbie & Chapin 1996; Brooks et al. 1998; Lipson & Monson 1998; Lipson et al. 1999; Jonasson et al. 1999; Grogan & Jonasson 2003). Collectively, the results of these studies strongly suggest that in alpine and Arctic regions there is a seasonal separation in the relative demand for N by plants and microbes. The latter immobilize N particularly during the winter months and plants mostly absorb N during the snow-free season. Plant uptake of nitrogen, although substantial in early fall, declines rapidly in winter only to be resumed at spring thaw when soluble N is released from lysed microbial cells. In contrast, microbial biomass builds up in winter in soils provided a well-developed snow base is in place that insulates the soil. It declines, however, in late winter as cells lyse, resulting in a short-lived pulse of soluble N in spring. With the onset of warmer temperatures, microbial growth resumes in summer. Although the results presented in this study do not include the partitioning of soil N in both winter and summer, they are consistent with the above interpretation for the summer months.
The C/N ratio of soil microbial biomass was higher in intertidal sites than in supratidal sites, which may indicate a high proportion of fungal biomass relative to bacterial biomass in intertidal soils (Paul & Clark 1989). Estimates of microbial C and N per unit soil volume were, respectively, 1.7 and 5.4 times greater than corresponding values for plant root C and N. However, the reliability of these estimates is questionable, given that correction factors to convert extractable microbial C and N to total microbial C and N have not been developed specifically for wetland soils (Witt et al. 2000).
Although soil microorganisms typically capture a relatively large proportion of the injected substrate in short-term in situ tracer experiments (Jackson et al. 1989; Schimel et al. 1989; Zak et al. 1990; Schimel & Chapin 1996), rates of substrate capture by plants comparable to those of soil microorganisms are observed in some systems (Norton & Firestone 1996; Lipson & Monson 1998). The latter was true of salt-marsh soils, where 17, 29 and 19% of the total 15N injected as glycine, ammonium and nitrate (respectively) were recovered in plant material and 30, 34 and 29% were recovered as microbial N. This relatively high capture of 15N by plants is likely to have resulted from the high density of roots in the upper 5 cm of soil. Approximately a quarter of 15N remained unaccounted for in soil cores. This incomplete recovery of 15N may reflect error associated with the estimation of total microbial N, as discussed above. Given that the extraction efficiency of in these soils is high (unpublished data), it is unlikely that large quantities of 15N substrates remained sorbed to the surface of soil particles following extraction in 2 m KCl. Furthermore, the recovery of equalled that of , which is highly mobile in the soil.
In addition to potential problems associated with incomplete 15N recovery, chloroform fumigation-extraction does not distinguish between symbiotic microorganisms and those in competition with plants roots for nitrogen, which may cause an overestimation of microbial uptake and an underestimation of root uptake, particularly in short-term experiments (Lipson & Näsholm 2001). Residual fine root material in fumigated soil also could provide an overestimate of microbial 15N uptake. However, the effects of root contamination in fumigation-extraction experiments are generally minimal (Witt et al. 2000), and in the present study every attempt was made to remove roots prior to fumigation, although some rootlets may have remained in the soil cores. Microbial biomass estimates are also relatively insensitive to the unintentional removal of root associated microorganisms during root removal (Witt et al. 2000). Likewise, although 15N uptake by non-symbiotic microorganisms present at the root surface may have resulted in an overestimation of root 15N uptake, the biomass of root-associated microorganisms would have been minimal relative to the total root biomass, and root contamination by microorganisms was minimized by washing roots in CaCl2.
Overall, the results of the in situ labelling experiments indicate that P. phryganodes competes effectively with soil microorganims for both organic and inorganic N and that amino acids are probably an important source of N for P. phryganodes in coastal salt marshes, where N limits plant growth. These results are consistent with the observation that the quantity of inorganic N derived from mineralization and goose faeces is not adequate to explain the observed regrowth of plants following defoliation by geese (Hik et al. 1991). The relative contribution of amino acids to plant nitrogen acquisition may be particularly high from mid- to late summer, when ammonium availability is low and high soil salinity interferes with the uptake of inorganic N (Henry & Jefferies 2002, 2003). Therefore, plant amino acid uptake appears to compensate for low ammonium availability over the period in the season when plant growth is most rapid and the demand for forage by geese is greatest.
This work was supported by the Natural Sciences and Engineering Research Council of Canada through a research grant to RLJ and a postgraduate scholarship to HALH. Additional funding was provided by the Association of Canadian Universities for Northern Studies and the Northern Scientific Training Program. We thank Rachel Sturge and researchers at the La Pérouse Bay Field Station and staff at Wapusk National Park for assistance and logistical support. Helpful comments on an earlier version of the manuscript were provided by Lindsay Haddon, Knut Kielland and an anonymous reviewer.