Australian mulga ecosystems –13C and 15N natural abundances of biota components and their ecophysiological significance

Authors


J. S. Pate Fax: + 61 08 93801001; e-mail: johnpate@cyllene.uwa.edu.au

Abstract

Samples of recently produced shoot material collected in winter/spring from common plant species of mulga vegetation in eastern and Western Australia were assayed for 13C and 15N natural abundance. 13C analyses showed only three of the 88 test species to exhibit C4 metabolism and only one of seven succulent species to be in CAM mode. Non-succulent winter ephemeral C3 species showed significantly lower mean δ13C values (– 28·0‰) than corresponding C3-type herbaceous perennials, woody shrubs or trees (– 26·9, – 25·7 and – 26·2‰, respectively), suggesting lower water stress and poorer water use efficiency in carbon acquisition by the former than latter groups of taxa. Corresponding values for δ15N of the above growth and life forms lay within the range 7·5–15·5‰. δ15N of soil NH4+ (mean 19·6‰) at a soft mulga site in Western Australia was considerably higher than that of NO3 (4·3‰). Shoot dry matter of Acacia spp. exhibited mean δ15N values (9·10 ± 0·6‰) identical to those of 37 companion non-N2-fixing woody shrubs and trees (9·06 ± 0·5‰). These data, with no evidence of nodulation, suggested little or no input of fixed N2 by the legumes in question. However, two acacias and two papilionoid legumes from a dune of wind-blown, heavily leached sand bordering a lake in mulga in Western Australia recorded δ15N values in the range 2·0–3·0‰ versus 6·4–10·7‰ for associated non-N2-fixing taxa. These differences in δ15N, and prolific nodulation of the legumes, indicated symbiotic inputs of fixed N in this unusual situation. δ15N signals of lichens, termites, ants and grasshoppers from mulga of Western Australia provided evidence of N2 fixation in certain termite colonies and by a cyanobacteria-containing species of lichen. Data are discussed in relation to earlier evidence of nitrophily and water availability constraints on nitrate utilization by mulga vegetation.

INTRODUCTION

The term mulga is used to describe a biome or biogeographic region of arid and semiarid (200–500 mm annual rainfall) parts of Australia in which mulga (Acacia aneura F. Muell. ex Benth.) and other Acacia spp. of similar physiognomy typically comprise the dominant arboreal elements (Johnson & Burrows 1981; Nelder 1986). Mulga ecosystems normally experience brief periods of moderate to high rainfall in winter through to early spring when ephemeral herbaceous species flourish transiently. Summer conditions encompass long periods of relatively severe drought and high temperatures, when biomass normally becomes restricted to woody shrubs, trees and a few perennial succulent species. The biome includes what is generally known as ‘hard mulga’ which occupies rocky dissected tablelands, and ‘soft mulga’, typical of red earths or deep sands (see biogeographic and floristic descriptions of Nelder 1986 for Eastern mulga and Beard (1990) for Western mulga). Soils are typically depauperate in organic matter and key nutrients such as phosphorus, although nitrate concentrations in soil (Barnes et al. 1992) and groundwaters (Murray & Siebert 1962) can exceed 100 μg NO3 -N g–1.

A recent paper (Erskine et al. 1996) describing the nitrogen relationships of mulga woodlands of Queensland and Western Australia examined nitrate in soil and xylem sap and concentrations of osmoprotective compounds and activity of nitrate reductase in foliage of common plant species. Both ephemeral and perennial components were shown to respond to hydration of soils of their habitat by markedly increasing their ability to utilize nitrate, while reducing foliar pools of nitrate and of compatible osmotica. In this paper we further explore ecophysiological relationships within the same soft and hard mulga associations, using δ13C analyses to prescribe photosynthetic pathways and apparent water-use efficiencies of the taxa and growth forms represented, and δ15N values for soil mineral N, plants and certain faunal components to assess possible inputs of fixed N in the ecosystems.

MATERIALS AND METHODS

Study sites

All plant material used came either from mulga vegetation 1000 km west of Brisbane, Queensland, Australia (Currawinya National Park, 28°51’ S, 144°30’ E) or from physiognomically similar vegetation on Thundelarra Station (28°53’ 36’ S, 117°07’ 45’ E) ≈ 500 km north of Perth and in the Murchison Region of Western Australia. The Queensland site comprised mulga vegetation typical of the eastern biogeographic region, and the Western Australian site equivalently representative mulga flora of the western counterpart (Beard 1990). Samples of plant biomass, soil and certain faunal components were collected at each site from soil:community typings deemed typical of either ‘soft’ or ‘hard’ mulga.

A special study was also undertaken on the vegetation on a 500-m-long, 5-m-high dune of wind-blown sand on a southern shore of Lake Mongers, Western Australia, 15 km south of the main sampling site at Thundelarra. While containing mostly the same species as in surrounding soft mulga, the open-textured, heavily leached sands of the dune were clearly highly atypical of mulga generally within the region.

Monthly mean climatic data for the sites are collated in Fig. 1, and annual rainfall averages (for a 50 year period) 294 mm for Currawinya and 262 mm for Thundelarra.

Figure 1.

. Climatic data for mean monthly rainfall and monthly mean maximum and minimum temperatures (50 year averages) for two sites in Western Australia (a) and Queensland (b) at which mulga vegatation was examined in this study.

Sampling of soils, plants and other material

Differences in the availability and the δ15N of soil NH4+ and NO3 were assessed at the ‘soft’ and ‘hard’ mulga sites at Thundelarra in an effort to ascertain which of these two forms of N the principal plant components were utilizing. The first series of soil samples (May 1995) were taken from three replicate pits in the depth zones 0–5, 30–35 and 70–75 cm at a ‘soft’ mulga site, and at 0–5 and 5–10 cm depth from a ‘hard’ mulga site. In August 1996 the ‘soft’ site was sampled again, using five replicate pits and soil recoveries at 0–5 and 40–50 cm. Samples were refrigerated immediately and their NH4+ and NO3 extracted by shaking for 1 h in 4 volumes (w/v) of 2 kmol m–3 KCl. Extracts were filtered and stored frozen until required for analysis of NH4+ and NO3 and determinations of their respective δ15N signatures.

Plant material from the eastern and western sites was sampled in July to September of 1994 and 1995, when ephemeral ground cover was at peak biomass and most perennial species were resuming active growth following wetting of soil by winter rains. Each site, including the dune at Lake Mongers, Western Australia, was first surveyed floristically to establish which common species collectively comprised 80% or more of the standing plant biomass, represented, respectively, by life and growth types comprising herbaceous winter (wet season) ephemerals, herbaceous perennials, woody shrubs and trees. Pooled samples of shoot biomass were then harvested from 30 to 40 different individual plants of the common higher plant taxa identified for each of the above categories. Whole shoots were collected from winter active ephemerals, whereas harvests were restricted to recently extended leafy shoot material of herbaceous perennial, woody shrub and tree species. All material was oven-dried at 60–80 °C.

Thalli of two unidentified species of crustose lichens were sampled at the hard mulga site at Thundelarra in 1995, and the same site was also utilized for collection of termites, ants and grasshoppers. Sampling of lichens involved sets of 50 randomly selected thalli at each site, while insect species were bulk-harvested from at least 10 sampling stations throughout the habitat. δ15N values for these fauna were then assessed to test for possible N2 fixation by termites and lichen symbionts.

Stable isotope analyses of plant material

Oven-dried samples of plant and animal material were ground to a fine powder in a vibratory ball mill (Retsch MM-2, Haan, Germany). δ15N analyses were conducted on the ammonia recovered by steam distillation of Kjeldahl digests. The distillate (50 cm3) of each digest was reduced to ≈ 1 cm3 volume using a microwave oven and the (NH4)2S04 converted to N2 by on-line oxidation with sodium hypobromite (Unkovich et al. 1993). The isotope analyses were performed on a dual inlet, triple collector mass spectrometer (VG Isogas, SIRA 9, Cheshire, UK). Analytical procedures and the potential errors of measurement involved in each were as detailed by Unkovich et al. (1993). The compounded error of measurement of δ15N between assays on replicate samples of plant material was ± 0·2‰ (SD).

δ13C analyses of matching replicates of all plant material were conducted in Perth (Western Australian material) using the VG Isogas SIRA9 following combustion of 1–1·5 mg samples of dry matter in an Isoprep 13 apparatus (VG, Isogas). 13C/12C ratios of the resulting CO2 were then compared with that obtained by combustion of a hydrocarbon δ13C standard, NBS-22. Precision of measurement based on triplicate assays of a single dry matter sample was ± 0·11‰ (SD). Comparable δ13C analyses of Queensland material were carried out using an automated 15N/13C analyser:mass spectrometer (ANCA-MS, Europa Scientific, Crewe, UK). The precision of measurement based on multiple analyses of a standard plant dry matter sample (leaves of Eucalyptus crebra) was 0·21‰ (SD).

Analyses of soil extracts for ammonium and nitrate

Soil extracts were assayed for nitrate using the cadmium method of Sloan & Sublett (1966) and for ammonium using the method described by McCullough (1967). δ15N natural abundances of the ammonium and NO3 (as NH4+) of the soil extracts were conducted following the steam distillation procedure described by (Keeney & Nelson 1982), using 1 g MgO and 2·5 g Devardas Alloy and a collection of 50 cm3 of distillate from each replicate extract. Analysis of extracts in this way is routinely performed with a precision of ± 0·12‰ (SE for n = 5). This analytical procedure has been shown to be reliable where plants have been grown solely on a defined N source, and, analysed in this way, the δ15N of shoots has been shown to be not significantly different from the measured δ15N of the N supplied (see Pate et al. 1994).

RESULTS

δ13C and δ15N signals of study species

Table 1 lists all plant species sampled from the hard (H) and soft (S) mulga sites in Western Australia and Queensland, segregating the flora of each region into (a) non-succulent winter ephemerals, (b) leaf or stem succulent ephemerals or herbaceous perennials, (c) putative C4 species, (d) non-succulent herbaceous perennials, (e) woody shrubs and (f) trees. For each taxon, mean δ15N and δ13C values for shoot dry matter are given, along with mean values (± SE) for sites and growth forms below each entry in the table. Note that data for the dune at Lake Mongers, Western Australia are listed separately, since this habitat was deemed non-representative of general mulga of the region.

Table 1.  . δ13C and δ15N signals of dry matter of whole shoots (ephemerals and herbaceous perennials) or recently produced shoot growth (woody shrubs and trees) collected from major species components of hard (H) and soft (S) mulga of Eastern Australia and Western Australia and from a ridge of wind-blown sand at Lake Mongers, Western Australia. All material sampled in July–September 1993–1995 Thumbnail image of

The δ13C values (Table 1) show that both eastern and western ecosystems carried floras comprised predominantly of C3 species. Thus, all woody perennial species, which collectively formed upwards of 90% of community biomass, were designated C3, as also were the vast majority of herbaceous perennials and ephemerals. Amongst the small groupings of leaf or stem succulent ephemerals and perennials, members of a few genera (Asclepias, Calandrinia, Dysphania and Zygophyllum) showed leaf succulence characteristics indicating C3/CAM, and one species (Salsola kali) exhibited typical C4 features with agranal chloroplasts and starch in its bundle sheath. The only other C4 species encountered were two non-succulent ephemerals, Eragrostis dielsii and Euphorbia drummondii. Both showed classic Kranz anatomy in their leaves and starch confined to bundle sheath parenchyma in the case of E. drummondii.

δ13C values reflected those photosynthetic pathways presumed to be operating, namely largely within the range – 25 to – 30‰ for non-succulent C3 species, – 13 to – 16‰ for C4 species and – 22 to – 27‰ for presumed facultative C3/CAM succulents currently functioning predominantly in C3 mode. Amongst the potential CAM species, Asclepias sp. was exceptional in exhibiting a strongly C4-type discrimination value of – 11·9‰.

Analysis of C3 species by site and growth form (Fig. 2a) showed significantly lower mean δ13C values for winter ephemerals (– 28·7‰ for Western Australia and – 27·8‰ for Queensland) than for woody perennials (mean for shrubs and trees at sites lying in the narrow range – 25·3 to –26·2‰). Differences between eastern and western sites proved to be non-significant for each life form.

Figure 2.

. Mean values for δ13C and δ15N of dry matter of shoots of species of different life and growth forms harvested during the spring flush of growth following winter rains at sites comprising typical mulga vegetation in Western Australia (Thundelarra) and Queensland (Currawinya). Standard errors are given for mean values of each species grouping. Note the separation within the analysis of succulent and C4 ephemeral species from other ephemerals and herbaceous perennials.

Corresponding mean values for δ15N of shoot dry matter of the same growth form categories (Fig. 2b) spanned the range from 7·5 to 15·5‰, with consistent evidence of higher δ15N values for shallow-rooted ephemerals (succulent, non-succulent and C4) than for deep-rooted perennial woody species at Thundelarra. This difference was not evident for the taxa sampled at Currawinya.

N concentrations and δ15N of KCl-extracted nitrate and ammonium of soil

Average concentrations of extractable NH4+ (8·2 mg kg–1) in the soils sampled at Thundelarra were 4 times those of NO3 (1·8 mg kg–1) (Table 2). Corresponding δ15N data (also in Table 2) showed substantial differences between the two forms of nitrogen, namely δ15N values in the range of 2·5–7·2‰ for NO3 versus 9–22·5‰ for NH4+. Sampling down to 1 m at the soft mulga site in 1995 showed substantial variation with depth in the δ15N of NO3, but no corresponding downward variation in respect of the δ15N of NH4+.

Table 2.  . Soil mineral N and δ15N of soil NH4+ and NO3 in western mulga at Thundelarra sites in 1995 and 1996 Thumbnail image of

Evidence for nitrogen fixation at study sites

Judging from the literature on the distribution of N2 fixing capacities amongst taxa of higher plants, one would have expected a potential for symbiotic N2 fixation in any of the eight study species of Acacia (Mimosaceae) and the genera Swainsona and Jacksonia (Fabaceae) (see Tables 1 & 3). By contrast, all other angiosperm species studied, including species of Senna (Cassia) (Caesalpiniceae), would be classed tentatively as non-fixing, in view of no known previous records for nodulation and N2 fixation in the genera in question.

Table 3.  . δ15N values for putative N2-fixing flora and fauna components and matching mean δ15N values for non-fixing reference species from an atypical sand dune system (Lake Mongers, Western Australia) and from general soft and hard mulga of Western and Eastern Australia Thumbnail image of

Excavations surrounding putative N2-fixing legumes were made in the soft mulga at Thundelarra and in the sand dune ridge at Lake Mongers. No cases of nodulation were recorded for the soft mulga site, whereas nodulation proved to be highly prolific on all specimens of Swainsona sp., Jacksonia sp., Acacia acuminata and A. aneura excavated at the dune site at Lake Mongers.

Of the two unidentified lichens sampled from rock pavements or trunks of tree species in hard mulga at Thundelarra, one association had a noncyanophyte algal component, and the other a heterocyst-bearing cyanophyceous symbiont in its thallus. The latter association was accordingly classified as putatively N2-fixing.

Table 3 assembles data for 15N natural abundances of all of the above-mentioned legumes (Queensland and Western Australian sites) and compares the values with means for comparable growth form groupings of non-fixing taxa from general mulga and from the atypical dune site at Lake Mongers. Based on the 15N natural abundance technique developed for measuring N2 fixation in natural ecosystems by Shearer & Kohl (1986) and Virginia et al. (1988), our data would indicate appreciable N2-fixing activity at Lake Mongers, where the δ15N values of Swainsona (2·2‰) Jacksonia (2·0‰) and two of the three acacias (3·0‰ and 2·3‰) were several units lower than the mean (9·1 ± 1‰) of the assigned group of four woody reference species. We accordingly concluded that these legumes were likely to be meeting a high proportion of their requirements for N from symbiotic N2 fixation. By contrast, all nine Acacias from regular hard or soft mulga at Thundelarra and Queensland showed an overall mean δ15N (9·10 ± 0·6) indistinguishable from that of partner or shrub reference species (9·10 ± 0·5, see Table 3) and were accordingly considered essentially non-N2-fixing.

A similar analysis of the data for lichens (Table 3) suggested appreciable dependence on atmospheric N2 by the cyanophyceous lichen association (mean δ15N = 2‰) but not the other lichen (15·5‰). Whole termite δ15N values ranged from 1·3 to 9·1‰ (mean 6·2‰) and for other phytophagous or carnivorous insect fauna from 11 to 12·1‰, i.e. a range of values close to that of the major woody taxa (9·0‰) from which they were likely to be directly or indirectly deriving their sustenance.

DISCUSSION

Our present study of two mulga communities of closely matching physiognomies and life and growth form characteristics shows closely similar respective sets of values for natural abundance signals of 13C and of 15N between western and eastern assemblages of major species. Despite geographical separation, the two sites experience similar mediterranean-type patterns of rainfall and temperature, and, when shoot materials were sampled in July to September, ephemeral and herbaceous perennials were completing a short, wet season life cycle, while woody shrubs and trees were mostly engaging in a spring flush of growth and flowering. By sampling whole shoots of the former herbaceous species and identifiable new shoot extension growth of the latter large woody members, it was assumed that resulting δ13C and δ15N signals of dry matter of the species would provide definitive information on patterns of recent assimilation of both carbon and nitrogen. However, while the above supposition was obviously true of the ephemeral species whose growth had been confined to the 4–5 month period between germination on onset of autumn rain and harvest the following spring, it might not have been fully valid for the dry matter sampled from perennial species. For example, discrepancies for perennials would result were the δ13C and δ15N signals of previously stored C and N used to initiate the season's new growth to differ appreciably from that subsequently synthesized during the season of sampling.

Assuming that the above complications did not apply or had only a minor effect on the natural abundances recorded for perennial species, the study suggested that carbon of dry matter accumulated by C3 ephemerals was consistently lower (more negative) by some 2‰ than that fixed at supposedly the same time by woody species. This applied to both eastern and western mulga and indicated, at face value, that subtle differences in carbon acquisition strategies might exist between the life forms concerned. Essentially similar differences in δ13C ratings between different growth and life forms are indicated for other ecosystems in the studies of Ehleringer & Cooper (1988), Ehleringer (1988) and Brooks et al. (1997).

Following the classic interpretations of the significance of δ13C discrimination values developed by Farquhar (1980) and Farquhar & Richards (1984), fast-growing winter ephemerals, drawing upon well-wetted upper soil, would be expected to exhibit only slight water stress and correspondingly poor instantaneous water-use efficiencies. Indeed, this was reflected in generally highly negative δ13C values across C3 members of this life form. By contrast, one would have anticipated slow-growing woody perennial forms to be geared to more conservative patterns of stomatal behaviour and water use, presumably as adaptations to generally poor water availability at depth following shallowly penetrating episodes of rainfall. Following this line of argument, these perennials would be expected to show generally less negative δ13C values than counterpart ephemerals.

A widely held presumption is that differences in carbon and water acquisition strategies between life and growth form groupings represent genetically fixed characteristics subject to little modulation from vacillations in availability of water between sites, seasons and years. Running counter to this were results from a preliminary study comparing δ13C signals of dry matter of field-sampled newly formed shoot material of mulga (Acacia aneura), harvested on three occasions during a year, with corresponding dry matter of well-watered glasshouse-cultured seedling plants. This study indicated marked adaptive responses to water status in this dominant tree species. Thus, the comparisons showed not only substantial (– 24 to – 28‰) differences in δ13C of new shoot material harvested in wet cool versus dry hot seasons for the field populations, but also an unusually low, uniform value of –31‰ for the continuously unstressed glasshouse plants. Arguing again from the suppositions of Farquhar and colleagues (see references above), it would appear that concentration gradients of CO2 between the internal leaf gas space and the atmosphere, and hence corresponding δ13C signals, can indeed be varied across wide limits within A. aneura, presumably reflecting targeted stomatal responses of the species in terms of stomatal response to daily and seasonal changes in water potentials of soil and xylem.

Finally for the 13C data, one finds evidence of a surprisingly high incidence of C3 flora in mulga, to the extent that potential CAM-type functioning associated with leaf or stem succulence is displayed by only a handful of relatively non-prominent species. Taken together with similar deficiencies in respect of C4 taxa, one concludes that the severe mediterranean-type climate experienced by mulga fails to provide a forum conducive to metabolic characteristics other than regular C3 behaviour [see discussion on adaptive qualities of C4 and other species in arid ecosystems by Henderson et al. (1995) and Ehleringer et al. (1997)]. Unfortunately, the number of species surveyed in this project was deemed not sufficient to conduct a phylogenetic analysis of the flora surveyed.

Turning to the δ15N data (see summary in Fig. 3), the strong indication is that symbiotic N2 fixation by Acacia spp. is virtually non-existent in regular mulga at both the eastern and western sites. This conclusion is based on high (≈ 9·0‰) δ15N values for both putative fixers and associated non-leguminous species at each site, together with no evidence of nodulation in any of the root excavations performed on the leguminous tree species. Interestingly, similar evidence of poor nodulation or apparent incapacity to nodulate by A. aneura was recorded over three decades ago by Beadle (1964) in his extensive field and glasshouse studies of mulga communities of Eastern Australia. This being generally so, we would view the unrivalled dominance of Acacia spp. in mulga as stemming principally from competitive advantages other than N2 fixation. A phyllodinous habit, with attendant advantages over normally leafed species in terms of drought tolerance and perhaps exceptional abilities to collect rainwater and tap underground reserves of water, come to mind as beneficial attributes in this connection.

Figure 3.

. Summary scheme depicting the ranking of various components of the biota of western mulga at Thundelarra in terms of δ 15N signatures of dry matter. Note the distribution between data for an atypical lakeside dune ridge at Mongers lake, where only N2-fixing woody shrubs and trees and reference non-N2-fixing shrubs were collected, and values for a range of biota obtained in general mulga of the surrounding region.

The finding of good nodulation of papilionoid (Swainsona, Jacksonia) and mimosoid (Acacia aneura and A. acuminata) legumes at the sand dune site at Lake Mongers, and evidence of a mean δ15N value of 2·4 ± 0·2‰ for these legumes versus 9·1 ± 1·0‰ for companion woody non-fixing species (see Fig. 3), argue strongly for relatively high proportional dependences of the above legumes on N2 fixation in this unusual situation. It seems reasonable to conclude that low soil N availability may have been responsible for determining the prevalence of nodulation and associated symbiotic N2 fixation at the lake site. The corollary to this for adjacent general mulga would then be that relatively high levels of NO3, and especially NH4+ in upper soil layers (Table 2), would select strongly against symbiosis by the same putatively fixing legumes, in view of the extra energetic costs (see Pate & Layzell 1990) likely to be incurred were such species to perform symbiotically when competing with non-fixing counterparts enjoying easily accessible N sources from soil and possibly ground water.

It is interesting to note in the above connection that all ephemeral floras surveyed so far in eastern and western mulga are lacking or extremely deficient in Fabaceae. This may possibly reflect the inherently poor capacity of these potential N2 fixers to utilize inorganic N in comparison with ephemeral taxa of other families such as Asteraceae, Poaceae, Brassicaceae, Amaranthaceae, Portulacaceae, Solanaceae and Goodeniaceae. Indeed, non-legumes of the above families have been shown in our recent companion study (Erskine et al. 1996) to exhibit strongly nitrophilous capacities in respect of utilization, transport and storage of nitrate.

Our limited observations on the soils of soft mulga at Thundelarra have indicated higher levels of NH4+ than NO3 and much higher δ15N values for ammonium (mean value of 22‰) than for nitrate (mean of 5·2‰). The calculated total mineral δ15N of 18·1‰ takes into account the predominance of NH4+-N. It is highly possible that such large differences in the δ15N of NO3- and NH4+-N may be generated during nitrification, a process which discriminates greatly in favour of the lighter isotope (Hogberg 1997), and accordingly leaves a relatively 15N-enriched NH4+ substrate in the soil solution or adsorbed onto soil surfaces.

Interpretation of soil extract δ15N data is problematic since it is our experience, and that of others, that the choice of extractant and the ratio of extractant to soil (e.g. Lindau & Spalding 1984) may sometimes alter the quantity of NO3- and NH4+-N extracted (Wheatley, MacDonald & Smith 1989) and the individual δ15N signals (Herbel & Spalding 1993). To date we have used KCl as extractant since it is generally considered the most reliable indicator of exchangeable N (Bremner & Hauck 1982) and equates to seasonally available N. However, KCl is likely to extract more NH4+ from clay soils than, say, a water extraction (Stanford 1982), and, furthermore, NH4+ bound to clay particles would be expected to be higher in δ15N than free NH4+ in the soil solution (Yoneyama 1996). Therefore, differences between the δ15N of NO3 and NH4+ fractions may be greater with a KCl than a water extraction procedure. At Thundelarra the clay content of surface soils is 21%. In the soils that we have examined to date the choice of extractant has made little difference to the 15N signature of NO3. Clearly, further work is required to identify which extraction conditions will give the most reliable indication of the δ15N of plant available N.

Despite these concerns, and those raised by Handley & Scrimgeour (1997) relating to difficulties in the interpretation of matching δ15N data for soils and plants, the differences in NH4+ and NO3δ15N signals for Thundelarra may prove a model system for following the N dynamics of an ecosystem. This opportunity is provided by the soils in question being uniformly low in organic matter (0·3‰ organic C and 0·02% total N for 0–5 cm at Thundelarra; see also Graetz & Tongway 1986) compared to more heterogeneous soils with larger pools of intractable organic N (e.g. Handley & Scrimgeour 1997), and having surface layers dry for much of the year, two factors which would greatly reduce those biological and chemical processes which incur substantial isotope fractionation.

Matching the divergent NH4+ and NO3δ15N signals for soil with the mean δ15N value of 13·3‰ for shoot N of ephemerals growing at the very site at which soil was sampled, the simplistic conclusion is reached that more NO3-N than NH4+-N is currently being utilized by the plants in question. This supposition fits well with (a) the much greater mobility and hence accessibility of NO3 than NH4+, particularly in clay soils, (b) the observation of Barnes et al. (1992) that NO3 concentrations in mulga are especially low under mulga trees and therefore possibly indicative of high rates of plant uptake of this form of N, and (c) the data of our companion paper (Erskine et al. 1996) demonstrating uniformly high in vivo nitrate reductase activities in leaves and high proportional levels of nitrate in xylem sap N of a number of key mulga species.

Extending our data set for soil δ15N even further, the observed decline in δ15N of soil nitrate with depth at the Thundelarra sites (Table 2) can be related to the general observation of appreciably lower δ15N values for newly formed shoot dry matter of deep-rooted perennial woody species than for shallow-rooted ephemerals at the same sites (Table 1). The obvious interpretation is that access to nitrate by ephemerals would be restricted to upper soil layers where NO3δ15N values are higher, whereas perennials would also have access to NO3 of less positive δ15N signal from lower down the profile. Herman & Rundel (1989) made similar observations in chaparral soils of California. The above conclusions are somewhat speculative in assuming that perennials can indeed access NO3 from all soil layers and that the proportional amounts of NH4+ and NO3 contributing to shoot δ15N signals are similar for each life and growth form grouping. The conclusions become even more fallacious were significant isotope fractionations to occur during uptake of different forms of soil N or were further discriminations to occur, for example, in root:shoot interchanges and cycling of assimilated N. The recent reviews of Handley & Scrimgeour (1997) and Hogberg (1997) highlight these and the many other possible complications which can beset interpretations of δ15N data, and thus our present conclusions need to be tested in greater detail than possible for the present study.

Our project poses interesting questions concerning the origin of the high amounts of soil mineral N found in mulga soils and the equally high levels of nitrate leaching from such soils into groundwater (Murray & Siebert 1962; Barnes et al. 1992). One viewpoint is that much of the nitrate-N currently present in mulga ecosystems might have formed in the distant past when N2-fixing cyanobacterial and lichen crusts predominated, prior to introduction of grazing sheep and cattle (Rogers et al. 1966; Graetz & Tongway 1986; Hodgins & Rogers 1997). While present-day land use tends to destroy this cryptobiotic crust and thereby possibly reduce N2 fixation inputs from such sources, the counter influence of grazing in reducing plant biomass and returns of nitrogenous nutrients in urine and dung (Graetz & Tongway 1986) might well have the opposite effect of sponsoring retention of large pools of soil mineral N.

Alternative candidates for N2 fixation inputs would be free-living cyanobacteria (Tchan & Beadle 1955; Smith et al. 1990) and termites (Collins 1981 and see δ15N values in Fig. 3). The latter are considered the more likely major contemporary source, since they still clearly comprise major consumers of dead woody residues within mulga and other arid zone vegetation (Watson et al. 1973). It is well authenticated that the capacity of termites to subsist on intractable plant residues of exceptionally high C:N ratio revolves around their possession of gut floras of synergistically operating cellulose-decomposing and N2-fixing bacteria (Collins 1981; Prestwich & Bentley 1981; Chappell & Slaytor 1986). A highly plausible suggestion is, then, that primary inputs of fixed N from termites, coupled to release of inorganic N from turnover of faecal and other residues of termite colonies, might comprise the principal current source of replenishment of nitrogen within the ecosystem (Barnes et al. 1992). In any event, the highly nitrophilous characteristics of mulga vegetation of Western Australia (Erskine et al. 1996), and the high nitrate levels in associated groundwater (Murray & Siebert 1962), stand in stark contrast to the situation in surrounding sandplain (kwongan) vegetation where proteaceous and myrtaceous rather than Acacia species dominate and appreciable stands of legumes flourish transiently only after fire. Unlike the mulga systems described here, soil and ground water levels of nitrate in the sandplain system are generally at barely detectable levels (M. J. Unkovich, J. S. Pate & G. R. Stewart, unpublished results; Stewart et al. 1993). Detailed comparative studies of the very different N dynamics of these ecosystems are clearly called for.

Acknowledgements

We are grateful to a programme grant to two of us (J.S.P. and G.R.S.) from the Australian Research Council for funding of this project. We thank David Arthur and Julie Crosbie for assistance with the Western Australian segment of the project and Tom and Roxanne Morrisey for allowing us to sample at Thundelarra Station.

Ancillary