Professor Ian Hodkinson, School of Biological and Earth Sciences, Liverpool John Moores University, Liverpool L3 3AF, UK. Tel: 0151 2312030; E-mail: email@example.com
1Invertebrate community assembly is described for two contrasting proglacial chronosequences (over 1900 years) at Kongsfjord, W. Spitsbergen, Svalbard.
2Three hypotheses were tested: (1) community assembly is deterministic, directional and predictable; (2) succession is inextricably linked to plant community and soil development; and (3) dispersal is a rate-limiting factor for community assembly.
3Communities, dominated by omnivores and detritivores, are more complex than supposed previously. Herbivore species were few but predators, parasitoids and hyperparasitoids were abundant.
4Species fell within one of eight defined groups with respect to colonization. This relates to ecophysiological tolerances, need for facilitation or dependence on other species. Spiders, surface-active Collembola and drought-resistant cryptostigmatic mites arrived before vascular plants established and soils developed. Later colonizers required site facilitation through soil development or host availability.
5Comparisons between the proglacial area of Midtre Lovénbre, a land-terminating glacier, and three Lovén Islands, released from beneath a tidewater glacier, showed similarities in community composition and species abundance with respect to successional stage, suggesting determinism and direction in community development.
6Most common species on Midtre Lovénbre colonized the Lovén Islands rapidly, suggesting that dispersal does not seriously restrict community development on the time scales involved. Some minor species associated with older soils appear not yet to have reached the islands.
Our primary objectives were to describe, measure and model patterns and rates of invertebrate community assembly following glacial retreat, using selected proglacial areas of known chronology. In particular, we tested three linked hypotheses, namely that (1) invertebrate community assembly is deterministic and directional and therefore predictable; (2) invertebrate succession is linked inextricably to plant community and soil development and (3) dispersal, linked to ecophysiological adaptability, is a rate-limiting factor for invertebrate community assembly.
Community assembly for a 1900-year chronosequence along the proglacial area of a land-terminating glacier, the Midtre Lovénbre, was compared with that (over c. 1000 years) on the nearby Lovén Islands (see Hodkinson et al. 2003). These small islands in Kongsfjord were released successively from beneath the ice during the regression of the main valley glacier, the Kongsbre. They lie 3 km from the mainland shore and are separated by c. 1 km of open water. Midtre Lovénbre communities developed on bare nutrient-poor moraines in proximity to the cooling effect of a large glacier but with unrestricted access to colonizing organisms from adjacent areas. The ‘isolated’ Lovén Island communities, by contrast, developed on glacial diamicton, receiving significant nutrient enrichment from surrounding marine ecosystems (Alsos, Elvebakk & Gabrielsen 1998; Hodkinson et al. 2003).
Sites (Table 1) were adjacent to Ny-Ålesund, Kongsfjord, West Spitsbergen, Svalbard. On Midtre Lovénbre foreland, seven sampling plots (approx. 20 × 30 m), were established at roughly equal distances along a transect of 1·7 km across the proglacial area. Starting at the glacier snout, sites 1–5 lay on the moraine, site 6 on the sandur below the terminal moraine and site 7 a little to one side on a raised rock ridge beyond the sandur. Plot 7 was relatively undisturbed by recent glaciation (Hodkinson et al. 2003). The Lovén islands chronosequence comprised sites on three islands, Storholmen, Midtholmen and Leirholmen. Each island plot (approximately 30 m diameter) was situated on the shallow down slope below the crest of the island. For both chronosequences plant cover, plant species richness, depth of soil organic horizon, percentage soil organic matter, soil nitrogen and soil moisture increased over time, whereas soil pH and average soil particle size decreased (Hodkinson et al. 2003). By contrast with the oldest Midtre Lovénbre site, the oldest Lovén Islands sites had a slightly lower vascular plant species richness and higher Bryophyte cover: the dominant species, nevertheless, tended to remain the same. Organic soil depth, however, was greater on the Lovén Islands, resulting presumably from nutrient enrichment and impeded drainage. Table 1 gives locations, altitudes and ages of sites and plots. Plot ages were estimated from land-based and aerial photography, coupled with radiocarbon dating of soil organic matter (Hodkinson et al. 2003).
Table 1. Study sites on Kongsfjord, Spitsbergen, Svalbard. Sites were aged in 2000 by either aerial/ground photography (P) or radiocarbon dating (RC); see Hodkinson et al. (2003) for full details
Midtre Lovénbre 7
Midtre Lovénbre 6
Midtre Lovénbre 5
Midtre Lovénbre 4
Midtre Lovénbre 3
Midtre Lovénbre 2
Midtre Lovénbre 1
Several techniques were used to sample invertebrates. Four-centimetre-diameter cores were taken randomly in late July 2000 from within each plot (n = 40 or 20 per plot for Midtre Lovénbre and Lovén Islands, respectively) and extracted for soil microarthropods using modified Macfadyen high gradient heat extractors (Leinaas 1978). Parallel sets of identical soil cores (n = 20 per plot) were extracted for Enchytraeidae and larval Chironomidae using wet funnel extraction (O’Connor 1962). On the shallow soils (< 4 cm depth) of Midtre Lovénbre and Leirholmen each core, comprising the top 4 cm of substrate containing a complete soil profile, was inverted before extraction. On the deeper organic soils of Midtholmen and Storholmen (> 4 cm) cores were divided horizontally before extraction. Midtre Lovénbre and Lovén Island sites were sampled in mid July to early August 2000 and 2001, respectively.
Pitfall and yellow sticky traps provided additional data on species present. Water-filled pitfall traps (5 cm diameter, 3 × 5 per site, 1 m spacing) were operated continuously from 16 July to 3 August 2000 on Midtre Lovénbre plots and from 17 July to 6 August 2001 on the Lovén Islands plots. Yellow sticky traps (double-sided, each area 400 cm2, 4 per plot) were operated continuously from 17 July to 6 August on 2001 on Midtre Lovénbre (Coulson et al. 2003a).
Species richness of soil testate amoebae was obtained from six soil cores per plot on Midtre Lovénbre. Soils were homogenized and subsamples examined on glass slides under a compound microscope. Numbers of species found during a 2-h search of six slides per plot were recorded. Lower than expected species richness at sites 6 and 7 was confirmed by examining four extra slides for 1 h.
Canonical correspondence analysis (CCA) was used to examine relationships between invertebrate community composition and site variables. Models based on linear regression were not significantly different from those using polynomial relationships (Makerenkov & Legendre 2002) and results presented use the linear model. CCA graphs were plotted using the Multivariate Statistical Package (mvsp) 3·1. (Kovach Computing Services). Species present at one site only were excluded from the analysis. CCA models, along orthogonal axes, the distribution of animal species (response variables) across sites with respect to environmental data for each site (explanatory variables). Results are displayed as diagrams plotting the correspondence (proximity) of species and sites to vector lines representing the explanatory variables. The closer the proximity, the greater the effect of that vector in determining the species distribution. Analyses are based on six explanatory variables, chronological age of site, soil surface pH, percentage vegetation cover, percentage cover by blue-green bacteria, plant species richness and percentage organic matter (OM) in the top 4 cm of soil. Variables are described fully in Hodkinson et al. (2003). The list was optimized from a larger set in which several variables, such as soil depth, soil moisture, soil clast size and soil nitrogen were serially correlated with those used. Adding these extra variables had minimal effect on the variation explained.
species richness during community assembly
The trophic composition of communities across the Midtre Lovénbre and Lovén Island chronosequences, derived from core, pitfall and yellow trap samples, is summarized in Fig. 1, with species listed by guild in Table 2. Taxonomic nomenclature follows Coulson & Refseth (2004) and Coulson et al. (2003a). All identified taxa with established populations are included. In addition, many Diptera (Nematocera and Muscomorpha) were caught by pitfalls and sticky traps (Coulson et al. 2003a) but taxonomic uncertainties prevented these groups being identified precisely and they are excluded from detailed analysis. They comprise predominantly non-resident adult Chironomidae (c. 18 species) blown in from surrounding aquatic breeding areas (Hodkinson et al. 1996b). However, one or two Svalbard chironomid species, particularly Smittia spp., are terrestrial, breeding in the transect soil (see Sendstad, Solem & Aagard 1976).
Table 2. Presence or absence of all higher invertebrate and testate amoeba (Rhizopoda) species at sites along Midtre Lovénbre and Lovén Island chronosequences. Note that the higher invertebrates are divided into trophic groupings. For testate amoebae* indicates the most common species
Some species caught by sticky and water traps parasitize or predate both resident and allochthonous prey species. These include several species of hymenopterous parasitoids, such as Aclastus borealis and Gelis glacialis (parasitizing spider cocoons), Plectiscidea hyperboreus and Stenomacrus groenlanidicus (parasitizing Diptera: Nematocera) and Atractodes pusillus (parasitizing Diptera: Muscomorpha) (Coulson et al. 2003c), as well as species attacking sawfly and aphid hosts.
The striking feature about community composition (Fig. 1) was the low number of invertebrate herbivore species. Initial colonization was almost exclusively by detritivores/omnivores. Obligate herbivores, feeding on vascular plants, did not establish for 60 years, and then only a few species. Some Collembola, however, include cyanobacteria (blue-green bacteria) in a mixed diet (Hodkinson et al. 1994; Birkemoe & Liengen 2000). By contrast, predators and parasitoids feeding on the limited prey species were surprisingly species-rich. Communities on the oldest Lovén Islands contained a total species number intermediate between sites 6 and 7 on Midtre Lovénbre. For the most mature communities on Midtre Lovénbre and Lovén Island chronosequences, species in each guild represented, respectively: herbivores 5% and 8%, detritivores/omnivores 47% and 42%, predators 21% and 29% and parasitoids 26% and 21% of total species.
Initial colonization was by detritivores/omnivores and their predators and parasitoids (Table 2). The first incomers at site 1 (2 years) were the surface-active Collembola Isotoma anglicana, Hypogastrura tullbergi and Sminthurides inaequalisarmatus,together with the cryptostigmatic mite Camisia anomia and, surprisingly, the occasional terrestrial chironomid larva. Common predators included one species of gamasid mite and two species of spider, Meioneta nigripes and Erigone psychrophila. The single resident parasitoid was S. groenlandicus. Herbivores were absent.
Community complexity increased slowly (Table 2) and after 60 years a further soil-dwelling collembolan, Folsomia quadrioculata and two additional cryptostigmatic mites, Diapterobates notatus and Tectocephus velatus, were added to the deritivore/omnivores. Enchytraeidae first appeared after 37 years. It was, however, 60 years (site 4) before the first herbivore of higher plants, the sawfly Prontoprista amentorum, established together with its parasitoid Ichneutes hyperboreus. Over the same time interval, a single predatory mite species was added, one spider species disappeared and a further two, Erigone arctica and Halorates spetsbergensis, entered the community.
Community development continued equally slowly between sites 5 and 6 (100–150 years) with two detritivore/omnivores, the collembolans Folsomia bisetosa and Sminthurinus concolor added, together with a further predatory gamasid mite and the spider Hilaira glacialis. A major transition, however, occurred at site 7 (1900 years) with the inclusion, of the relatively sedentary soil dwelling collembolans Onychiurus groenlandicus and Ceratophysella longispina and several additional mites including Ceratoppia hoeli, Oppiella neerlandica and Brachythonius sp. Establishment of the aphid Acyrthosiphon svalbardicum, feeding on Dryas octopetala, was accompanied by the addition of several aphid parasitoids and hyperparasitoids, namely Diaeretiella rapae(M’Intosh) (Braconidae) (=Trioxys sp. of Coulson et al. 2003a), Dendrocerus bifoveatus (Kieffer) (Ceraphronoidea, Megaspilidae) and Alloxysta sp. (Cynipidae), as well as larvae of the aphidophagous Parasyrphus tarsatus (Syrphidae).
Testate amoebae (Rhizopoda) (Table 2) were present throughout the Midtre Lovénbre transect. A single species, Difflugia sp., was found at site 1 and the species that was to become dominant at remaining sites, Centropyxis cassis, appeared at site 2. This was followed by increasing species richness up to site 5 (100 years), but richness then declined. This decline appears real, as confirmed by extra samples from the older sites revealing no additional species.
Comparisons between the Lovén Islands and Midtre Lovénnre reveal many similarities but some differences. Leirholmen is chronologically of similar age to site 5: Midtholmen and Storholmen are intermediate between sites 6 and 7.
The same four common Collembola species that first colonized Midtre Lovénbre were present on all islands but some of the later soil-dwelling species, such as O. groenlandicus, were not found. An additional species, Hypogastrura viatica, absent from Midtre Lovénbre, was abundant on Leirholmen but not found on the older islands.
Mites showed similar species consistency, with the early colonizers of Midtre Lovénbre present on all islands. Similarly, T. velatus was present at the younger but missing from the oldest sites on both chronosequences. Some Midtre Lovénbre species, notably Mycobates sarakensis and C. hoeli, were absent from all islands. Terrestrial chironomid larvae and Enchytraeidae were present throughout the island sites. Predatory mite species were well represented on all islands and E. psychrophila and H. spetsbergensis were again the first spider colonists, with the former preceding the latter. The herbivores P. amentorum and A. svalbardicum were not found on Leirholmen, but were present on Midtholmen and Storholmen.
Hymenoptera appeared less diverse than on Midtre Lovénbre. S. groenlandicus was again ubiquitous on the islands but the remaining ichneumonids were sporadic (P. hyperborea on Midtholmen, A. pusillus on Leirholmen) or absent (A. borealis, G. glacialis). Of the remaining species I. borealis, D. bifoveatus and Alloxysta were again associated with herbivores only on Midtholmen and Storholmen, but D. rapae was not found.
Table 3 summarizes Kruskal–Wallace tests for significant differences in invertebrate population densities among sites on Midtre Lovénbre and Lovén Island chronosequences. Abundances of all taxa differed significantly among sites across Midtre Lovénbre. Similar significant differences were found among Collembola, Enchytraeidae and predatory mites on the Lovén Islands but cryptostigmatic mite and chironomid populations were similar.
Table 3. Results of Kruskal–Wallace tests for significant differences in the abundance of soil invertebrates among sites on Midtre Lovénbre and Lovén Island chronosequences. Species present at one site are excluded. For mites, Collembola and Chironomidae n = 40 or 20 on Midtre Lovénbre and Lovén Island sites, respectively. For Enchytraeidae n = 20 throughout
Total Collembola on Midtre Lovénbre (Fig. 2a) (maximum = 12 000 m−2) were consistently lower than on the Lovén Islands (Fig. 2b) (maximum = 80 000 m−2) for sites of equivalent age. In general, Collembola densities increased with site age, becoming maximal at or beyond 100 years. On Midtre Lovénbre H. tullbergi tended to be numerically dominant, with F. quadrioculata second and I. anglicana and S. inaequalis almost invariably present but at lower densities (< 1000 m−2). Remaining species formed minor components of the community. By contrast, numerical dominance was reversed on the Lovén Islands, with F. quadrioculata dominant, I. anglicana and S. inaequalis generally present at low populations densities, and other species occurring sporadically. H. tullbergi remained the second dominant species on Midtholmen and Storholmen but was largely replaced by H. viatica on Leirholmen.
Total cryptostigmatic mites on Midtre Lovénbre (Fig. 3a) increased gradually from sites 1–4 (maximum = 15 000 m−2) and then declined steadily to site 7 (6000 m−2). C. anomia was dominant, particularly at the youngest sites (1–4), with D. notatus or T. velatus second or third in rank abundance (< 4000 m−2). Among remaining species, only Hermannia reticulata reached significant densities at site 7 (2000 m−2). Species abundances were more equitable on the Lovén Islands with C. anomia, and D. notatus displaying similar densities (2–3000 m−2) and O. neerlandica becoming relatively abundant on Midtholmen and Storholmen (3–400 m−2). Populations of H. reticulata, where present, were less than at site 7 (< 1000 m−2).
Densites of larval Chironomidae (Fig. 4a) were surprisingly higher in heat-extracted soil samples than in O’Connor wet extractions and data for the former are presented. Densities on Midtre Lovénbre varied among sites with a maximum (600 m−2) at site 2. Site 1 supported a pioneer population of 25 m−2 but other sites were variable within the range 200–400 m−2. Lovén Island populations, however, were similar (900–1100 m−2) and about 2–3 times those on Midtre Lovénbre.
Enchytraeidae populations (Fig. 4b) were consistently higher in wet extractions. Densities differed significantly among sites on Midtre Lovénbre, resulting primarily from a higher population at site 7 (800 m−2). Sites 3–6 supported broadly similar densities (< 200 m−2). Populations on the Lovén Islands were consistently higher (500–2100 m−2) than at comparable sites on Midtre Lovénbre, with Midtholmen significantly higher than the remaining islands.
Populations of predatory mites (Gamasida + Prostigmata) were present throughout both chronosequences. Densities on Midtre Lovénbre generally increased from a small number of scattered individuals m−2 at site 1, through measurable populations (4–111 m−2) at sites 2, 3 and 5 to significant populations at sites 6 and 7 (350–550 m−2). Site 4 was unusual in that only a sparse population was found. Population density on Leirholmen (500 m−2) was comparable with site 6 but those on Midtholmen and Storholmen (1200–1500 m−2) were twice the Midtre Lovénbre maximum and probably reflected the higher numbers of Collembola present.
Many parasitoid Hymenoptera were taken in water traps at all sites. The Midtre Lovénbre fauna (Fig. 5a) was dominated by four Ichneumonidae, S. groenlandicus, P. hyperborea, A. borealis and A. pusillus. Brachypterous forms of S. groenlandicus indicate its resident status. The abundance trend for these species was a progressive increase, from a low at site 1 to a peak at the midpoint of the transect (sites 3–5) and then a gradual decline at sites 6–7. G. glacialis and S. lineiger, were taken at low densities on sticky traps but were absent from water traps. Remaining Hymenoptera were less abundant and their distribution reflected that of potential herbivore hosts. Thus, I. hyperboreus was consistently present at low densities at sites 4–7, but S. lineiger was infrequent and only at site 7. Similarly, the aphid parasitoids/hyperparistoids all occurred only at site 7. By contrast, S. groenlandicus was again the dominant species on all Lovén Islands (Fig. 5b), the herbivore parasitoids/hyperparasitoids occurred at similar densities to Midtre Lovénbre but catches of the remaining species were lower and more sporadic.
Total spider catches in water traps on Midtre Lovénbre tended to increase gradually from sites 1–7 (Fig. 6a). This reflected catches of indeterminate juveniles and masked individual species trends related to distribution. Numerically dominant species were H. spetsbergensis and two Erigone species. M. nigripes and H. glacialis became increasingly important in the early and latter stages of succession, respectively. The Lovén Islands (Fig. 6b) followed similar trends, with highest spider numbers trapped on the oldest islands and the same rankings of species abundance. One species associated with the early succession on Midtre Lovénbre, M. nigripes, was not found.
variations in species demography
Some abundant collembolans, such as H. tullbergi, differed in population body size distribution across the Midtre Lovenbre (Fig. 7). Body length correlates with Collembola age (Petersen 1971), and on this criterion the population age structure of H. tullbergi varied among sites. At younger sites (1–3) the population comprised mainly of older adult individuals (length > 0·85 mm; Birkemoe & Sømme 1998) compared with a greater proportion of juveniles at sites 4–7.
canonical correspondence analysis
Figure 8 compares CCA ordinations for Midtre Lovénbre sites alone and with the addition of the Lovén Island data. Analyses demonstrated strong relationships (> 85% of variation explained by the first three axes), confirming directional trends already apparent in the data. Thus, most species and sites ordinated along the first and second axes adjacent to vector lines representing site age, percentage cover and percentage OM. This suggests that site age and the associated factors, percentage cover and percentage OM were the best predictor variables for many, but not all invertebrate taxa. Thus O. neerlandica (8), Enchytraeidae (3), F. quadrioculata (12), predatory mites (10) and H. reticulata (7) all tend to associate strongly with older sites. D. notatus (5), and Chironomidae (1) lie along the time axis but closer to the centroid, suggesting less dependence on time but no strong association with the other factors. By contrast, some earlier colonizers, such as I. anglicana (10), S. inaequalis (11), T. velatus (6) and C. anomia (4), lie closer to other vectors such as vascular plant species richness, percentage blue-green and pH.
One problem in chronosequence studies that substitute space for time (Pickett 1989) is that prevailing conditions at any location have probably varied over the long time scales involved. For example, Kaufmann (2002) demonstrated faster community development on Rotmoos Glacier foreland over the last 30 y compared with the previous 30–50 years, linked to marginally warmer temperatures. Another confounding factor is rapid colonization of glacier forelands from lateral vegetated areas, as occurs on Rotmoos. Lack of such areas around Midtre Lovénbre, however, reduces this possibility significantly for soil animals, at least at sites 1–5, where the direction of colonization most probably corresponded with the long time axis of the chronosequence. Furthermore, because of low community diversity, there was little evidence of strong topographical variation in community composition at sites of equivalent age, as evident on Rotmoos (Kaufmann & Raffl 2002). Despite such potential limitations chronosequences provide unique insights into community assembly that cannot be replicated by short-term experimentation.
Time scales for community assembly were slow, reflecting harsh environmental conditions extending over hundreds to thousands of years, longer by an order of magnitude than most studies elsewhere (Matthews 1992). Nevertheless, the composition of animal communities developing on Midtre Lovénbre and Lovén Island chronosequences were broadly similar, despite their contrasting histories and situations. Thus the fauna of Leirholmen after 100 years corresponded closely with that of Midtre Lovénbre site 5, paralleling the established similarities in the soil and vegetation (Hodkinson et al. 2003). Thus I. anglicana, F. quadrioculata, Hypogastrura spp., C. anomia and T. velatus were the dominant soil microarthropods, H. spetsbergensis and E. psychrophila the dominant spiders, S. groenlandicus and A. pusillus the dominant parasitoids and Enchytraeidae and terrestrial chironomids were present. Differences included the substitution of H. tullbergi by H. viatica on Leirholmen and the absence of some of the common Midtre Lovénbre species, such as S. inaequalis and E. arctica. These latter species, nevertheless, were present on the older islands. Similarly, the faunas of Midtholmen and Storholmen largely mirrored that of site 7, but lacked some minor species. Common additional elements included the herbivores P. amentorum and A. svalbardicum, together with the parasitoids I. hyperboreus, D. bifoveatus, D. rapae and Alloxysta sp., the spider H. glacialis, the syrphid P. tarsatus and the microarthropods H. reticulata and O. neerlandica. Lovén Island communities contained many flightless species that have dispersed effectively across the marine water barrier to and between islands over the last 100–1000 years (Coulson et al. 2002). Some apparently more sedentary soil-dwelling microarthropod species, however, such as F. bisetosa, O. groenlandicus, M. sarekensis, C. hoeli and a Brachythonius sp., together with parasitoids such as A. borealis, G. glacialis and S. lineiger, were absent from island samples. Enchytraeidae and F. quadrioculata, were significantly more abundant on the older islands than on Midtre Lovénbre, probably reflecting their better exploitation of deeper, moister organic soils.
Colonization rates by testate amoebae on Midtre Lovénbre contrasted strongly with those for volcanic tephra on Deception Island, South Shetland (Smith 1985). Difflugia sp. occurred within 1 year and three species were present after 60 years compared with the elapse of 150 years before first colonization of unfavourable sites on Deception. The apparent decline in species richness between sites 5–7 is, however, difficult to explain.
In addition to variations between chronosequences, more subtle differences may occur within species populations across the chronosequence. Varying population age structure of H. tullbergi across Midtre Lovénbre may thus result from poor recruitment related to low food availability and high desiccation stress, to which hatching Collembola are particularly susceptible (Birkemoe & Leinaas 1999). Alternatively, it may reflect differential dispersal of mature animals from favourable adjoining sites (Johnson & Wellington 1983; Hertzberg 1997; Hertzberg & Leinaas 1998). Mature Collembola often have a greater tendency to disperse than juveniles.
The chronosequences chosen showed strong directional trends for several factors and it is unsurprising that these often emerged as good predictors of species distributions. Factors linked positively to site age, or distance from the glacier, included soil OM, moisture, depth, clast size and total nitrogen as well as vegetation cover and vascular plant species richness. Other factors, such as cyanobacteria cover, were less dependent on site age. Together these factors provide the environmental template for invertebrate colonization. Community assembly is, however, complex depending on the interplay between the ecology, ecophysiology and dispersal characteristics of individual species. Eight categories of colonizers, each with their characteristic demands can be recognized, namely:
(a) Species independent of vascular plant establishment. Includes surface-active species subsisting on allochthonous inputs of prey items or organic detritus. These are the first colonizers, including spiders such as H. spetsbergensis, M. nigripes and E. psycrophila and darker pigmented Collembola with protective pigmentation such as I. anglicana, S. inaequalis and H. tullbergi. Species are often highly dispersive and show some drought resistance (Hodkinson et al. 1996a; Coulson et al. 2002). The mite C. anomia is an exception that is slow-moving but it is, nevertheless, highly robust and drought-resistant and may be dispersed passively by other agencies (Hodkinson et al. 1996a; Coulson et al. 2002). These Collembola and mite species may feed on cyanobacteria or lichens as they become available.
(c) Species dependent on cyanobacteria or lichen cover. Embraces mites such as T. velatus and D. notatus whose population peaks corresponded with the maximum cover by blue-green bacteria or lichen. These are again desiccation=resistant surface-dwelling species (Hodkinson et al. 1996a).
(d) Species dependent on colonization by the host plant. Obligate herbivores whose appearance corresponded with the first colonization by their plant host. Included P. amentorum on S. polaris and A. svalbardicum on D. octopetala (Strathdee & Bale 1995).
(e) Specialized predators/parasitoids/hyperparasitoids dependent on resident host species. Includes species such as I. hyperboreus, S. lineiger, D. bifoveatus, D. rapae, Alloxysta sp.and P. tarsatus associated with obligatory prey. Other species such as A. borealis and G. glacialis are associated more closely with spider species and were distributed more broadly (Coulson et al. 2003a).
(f) Less specific general predators dependent on general availability of prey. Includes predatory mites that reflect the abundance of potential prey species such as F. quadrioculata. This may give the misleading impression in CCA that prey and predator are responding similarly to the same environmental variables.
(g) Poor dispersers. Species occurring only in the oldest soils and having poor dispersal powers, being apparently absent from sites, where on the basis of similar soil characteristics they might be expected to occur. Includes the mites H. reticulata, C. hoeli and O. neerlandica.
(h) Vagrants. Species of unknown biology that are not permanently resident, e.g. allochthonous Chironomidae that may be present at any site at any time.
Thus, the concept of facilitation, used widely in plant succession (McCook 1994; Callaway & Walker 1997), is equally applicable to animal community assembly. Some species are facilitated by the presence of specific plants, others by the presence of suitable animal hosts. Some require no facilitation, yet others require long-term general facilitation through soil development associated with plant growth.
High Arctic terrestrial ecosystems have low net primary productivity (< 140 g C m−2 year−1) and low vascular plant species richness, with an essentially two-dimensional vegetation structure (Bliss & Matveyeva 1992). Ecological theory predicts low diversity communities with relatively simple food chains and low connectivity (Bazely & Jefferies 1997; Morin 1999). Is this realized? The food web is dominated by detritivore and omnivores throughout community assembly, with herbivores of vascular plants relatively late entrants, such that they are unlikely to impact significantly on the early direction of plant succession as observed elsewhere (Brown & Gange 1992). However, the food web, particularly at the older sites, is surprisingly complex (cf. Summerhayes & Elton 1923). It comprises relatively high proportions of hymenoperous parasitoids, some hyperparasitoids and a complex of predators, including several spider species, some gamasid and prostigmatic mites and an aphidophagous syrphid. Several larger Hymenoptera were captured at densities that appear to exceed those of their potential prey (Coulson et al. 2003a). This raises the question of how the system sustains such extended food chains with an apparently high parasite/predator loading. The answer appears to lie in the spatial structure of the community and the vagility of parasitoids. The low diversity of plant species produces an extensive vegetation mosaic of relatively uniform species composition. Some prey species exist at low densities across this mosaic and their location requires widespread searching, particularly by the larger parasitoids. Trap catches reflect aggregate numbers of vagile parasitoids searching through the area rather than the absolute numbers present at one time. Furthermore, allochthonous inputs of both insects and detritus provide significant further continuing subsidies to the terrestrial food web in areas of low productivity (Polis et al. 1997; Hodkinson et al. 2002).
Returning to our initial hypotheses, the evidence suggests that community assembly was both deterministic and directional (Ward & Thornton 2000), with similar common predictable patterns, involving common suites of species, across both chronosequences. This relates to species requirements as listed previously, which determine their point of entry into the community. However, as for the plant community (Hodkinson et al. 2003) it is difficult to distinguish between chance and determinism by default, when the community progressively recruits from a limited pool of effectively dispersed species. Kaufmann (2002) reached similar conclusions for his more complex community on Rotmoos Glacier foreland − that community assembly was an ordered and predictable process linked to the varying biologies and tolerances of the colonizing species. Nevertheless, several earlier colonizers exist independently of vascular plants and soil development. Later colonizers depend more directly on characteristics of the developing soil or the presence of particular plant types or species, which facilitate their establishment.
Another conclusion common with the Rotmoos study is that for the more abundant invertebrates rates of dispersal do not appear seriously to limit colonization (Kaufmann 2001; Kaufmann et al. 2002). This is evidenced by the fact that the colonization is almost instantaneous. For many other species requiring facilitation, colonization of three separate islands 3 km from the mainland did not present an insurmountable barrier given the time scales involved, and there was high commonality of species at comparable successional stages. Several species survive extended passage on or in seawater (Coulson et al. 2002) and only in the case of the less common soil-dwelling microarthropods restricted to Midtre Lovénbre site 7 does weak dispersal ability appear to limit their realized distribution.
Invertebrate community assembly on proglacial chronoseqences in the High Arctic thus shares many common features with successions elsewhere, but with a greatly extended time scale. This raises concern over the long-term vulnerability of these isolated island communities when subjected to rapidly accelerating climate change. Future change will almost certainly occur more rapidly than in the past (IPCC 2001).
We thank Nick Cox and Maggie Annat for their boat handling skills in landing us safely on the Lovén Islands, often in difficult conditions. The work would have been impossible without the invaluable taxonomic contributions of Rowley Snazell, Arne Fjellberg, Josef Stary, Matt Colloff, Matti Viitasaari, Endre Wuillassen, Nigel Fergusson, Peter Neerup Buhl, Kees van Achterberg, Mark Shaw and Reijo Jussila. In particular, David Wilkinson kindly processed the Rhizopoda samples. This research is based upon work supported by the National Science Foundation under agreement no. OPP-0002239 and the National Oceanic and Atmospheric Administration under agreement no. NA67RJ0147.