Large-scale terrestrial gastropod community composition patterns in the Great Lakes region of North America

Study region

Land snail faunas were sampled across a 1400 × 800 km area centred on the western portion of the Great Lakes basin in eastern and central North America (Fig. 1). This area covers a wide range of bedrock, climate and vegetation types. Both Palaeozoic sedimentary and Precambrian igneous bedrock outcrops in the region. One of the more prominent sedimentary exposures is the Niagaran Escarpment, a band of Silurian limestones and dolomites extending from western New York state to north-eastern Iowa. Outcrops along the western Lake Superior shore typically represent late-Precambrian mafic igneous rocks associated with the Keewenawan mid-continental rift system (Anderson, 1983). Average minimum winter temperatures range from −25 °C in northern Minnesota to −10 °C in western New York State. Average maximum summer temperatures range from 25 °C along the western Lake Superior shore to 30 °C in south-eastern Iowa. The average length of the 0 °C growing season varies from 100 to 110 days in northern Minnesota and Michigan to 180–190 days in southern Iowa, southern Ontario, and western New York state. In areas adjacent to or downwind from (east of) the Great Lakes (especially the Lower Peninsula of Michigan, southern Ontario, and western New York State), the climate tends to be buffered over that normally experienced in the continental interior, being warmer in the winter, cooler in the summer and having a longer growing season with more constant precipitation (Eichenlaub, 1979). Matrix vegetation varies from tallgrass prairie in the west to deciduous forest in the east to mixed boreal–hardwood forest in the north (Barbour & Billings, 1988).

Study sites

Four hundred and forty-three sites (Fig. 1) were surveyed across the range of habitats known to support diverse land snail assemblages (Nekola, 1999). The 26 habitats surveyed were broadly grouped into five major categories: rock outcrops, upland forests, lowland forests, upland grasslands and lowland grasslands (Table 1). While sites generally represent undisturbed examples of their respective habitats, an effort was also made to sample some (25 rock outcrop, 12 upland forest, nine lowland forest, four upland grassland and eight lowland grassland) that had been disturbed anthropogenically by grazing, logging, recreational/urban development or bedrock/soil removal. Examples of such sites include field-edge stone piles, abandoned agricultural fields, abandoned building foundations, old quarries, pastures, road verges and exploited forests.

Field methods

Documentation of terrestrial gastropods from each site was accomplished by hand collection of larger shells and litter sampling for smaller taxa from representative 100–1000 m² areas. Soil litter sampling was primarily used as it provides the most complete assessment of site faunas (Oggier et al., 1998). As suggested by Emberton et al. (1996), collections were made at places of high micromollusc density, with a constant volume of soil litter (approximately 4 L) being gathered from each site. For woodland sites, sampling was concentrated: (1) along the base of rocks or trees; (2) on soil covered bedrock ledges; and/or (3) at other places found to have an abundance of shells. For grassland sites, samples consisted of: (1) small blocks (c. 125 cm³) of turf; (2) loose soil and leaf litter accumulations under or adjacent to shrubs, cobbles, boulders and/or hummocks; and (3) other locations observed to have an abundance of shells.

The latitude–longitude location of each sample was determined using either USGS 7.5 minute topographic maps or a hand-held GPS. To minimize bias from use of polar-coordinates, these locations were converted subsequently to Cartesian UTM Zone 16 coordinates using arcinfo.

The presence or absence of anthropogenic disturbance and soil surface architecture (duff vs. turf) was also recorded from each site. For purposes of this study ‘duff’ soils represent sites where the organic horizon was deep (> 4 cm) and subtended by a friable upper A horizon consisting largely of humus and mineral soil. ‘Turf’ soils represent sites where the organic horizon is thin (< 4 cm) and immediately subtended by an upper A horizon firmly bound together by living plant roots. While many habitats only supported a single soil architecture type (e.g. all carbonate cliffs were duff, and all bedrock glades were turf), some (such as white cedar swamps) could possess either turf or duff surface layers, depending upon individual site conditions. Thus, habitat type could not be used as a surrogate for soil surface architecture.

Laboratory procedures

Samples were dried slowly and completely in either a low-temperature soil oven (c. 80–95 °C) or in a greenhouse. Dried samples were then soaked in water for 3–24 h, and subjected to careful but vigorous water disaggregation through a standard sieve series (ASTME 3/8′ (9.5 mm), #10 (2.0 mm), #20 (0.85), and #40 (0.425 mm) mesh screens). These fractions were then dried and passed again through the same sieve series, and hand-picked against a neutral-brown background. All shells and shell fragments were removed.

All identifiable shells from each site were assigned to species (or subspecies) using the author's reference collection and the Hubricht Collection at the Field Museum of Natural History (FMNH). Identification of some additional specimens representing Holarctic taxa more common in western Europe were verified by Robert Cameron of the University of Sheffield, UK. All specimens have been catalogued and are housed in the author's reference collection at the University of Wisconsin — Green Bay. Nomenclature generally follows that of Hubricht (1985), with updates and corrections by Frest (1990, 1991) and Nekola (2002).

Statistical procedures

Site ordination

One hundred and eight terrestrial gastropod taxa were identified from the 443 inventoried sites (Appendix I). Twenty-two sites were species poor (four or fewer taxa) and removed from further analysis. These included eight igneous cliffs, three lakeshore forests, three tamarack wetlands and single shale cliff, oak–hickory forest, maple–basswood forest, floodplain forest, sand dune, old field, sedge meadow and cobble beach sites.

NMDS of the remaining 421 sites demonstrated that the only stable solution occurred along two axes of variation, where one was achieved in 50% of starts. The minimum stress configuration of this solution was 0.197412. In other dimensions, the most stable solution(s) were achieved in five (one dimension), three (three dimensions) and one (four dimensions) runs (Table 2).

Identification and description of compositional clusters

Visual observation of the chosen NMDS ordination solution demonstrated apparent natural clustering in faunal composition, with at least one well-defined group existing in the lower-centre of the diagram (Fig. 2). AWE_k analysis demonstrated that the maximum score (2131.5) was achieved at the 53rd cluster. As this result provides too many groups to be useful for generalization of faunistic trends, AWE_k scores from k= 1–10 were calculated, along with the percentage increase in AWE_k from k to k + 1 clusters (Table 3). These data demonstrate that over 50% of maximum AWE_k was achieved by the 6th cluster. The percentage increase in AWE_k fell by almost 50% for cluster 7 (7.3%), and decreased steadily to the 3.4% range by cluster 10. Based on this, the optimal number of clusters was set at six (Fig. 3), even though it does not represent maximum AWE_k.

Contingency table analysis (Table 4) demonstrates that habitat representation significantly (P < 0.00005) varies between the six nonoverlapping kmeans clusters. Clusters A–C were equally (P = 0.3778) represented by rock outcrop and upland forest sites, while cluster D was dominated by lowland forests, cluster E by lowland grasslands and cluster F by upland grasslands. The 10 most frequent taxa also varied greatly, with approximately 50% turnover occurring between even the most similar groups (Table 5). Spearman's rank correlations of species occurrence frequencies indicated that clusters D and E were most similar (0.831), while clusters A and F were the most different (0.316). Six species possessed modal occurrence frequencies in cluster A, 25 in cluster B, 32 in cluster C, five in cluster D, 20 in cluster E and 20 in cluster F (Table 6). Forty-one total taxa were encountered in cluster A, 78 in cluster B, 87 in cluster C, 58 in cluster D, 60 in cluster E and 60 in cluster F (Appendix I).

Analysis of environmental co-variables

Monte Carlo testing of the maximum correlation vectors for the four recorded environmental variables (Fig. 4) demonstrated that all were highly significant (P < 0.0005), having maximum r-values ranging from 0.2111 to 0.7909 (Table 7). In duff sites, only UTM N coordinate and anthropogenic disturbance were found to correlate significantly (P < 0.0005) with the ordination diagram (Table 7). Northern sites tended to occur in the lower left of this group (maximum r= 0.8409), while disturbed sites tended to occur further to the right (maximum r = 0.4812; Fig. 5). In turf sites, only UTM N and E coordinates were found to correlate significantly (P < 0.0005) with the ordination diagram (Table 7). In this group, more northern (maximum r = 0.6690) and eastern (maximum r= 0.5067) sites tended to occur to the lower left (Fig. 5).

Discriminant analysis demonstrated that the location of duff and turf sites in ordination space differs significantly (P < 0.0005) (Table 8), with duff sites being essentially limited to the upper left half of the diagram, and turf sites being found largely in the lower right (Fig. 6). The classification summary for this analysis indicates that only 33 of the 421 sites (7.8%) were classified improperly when soil surface type was used as the sole predictor variable.

Discriminant analyses conducted separately on duff and turf sites demonstrated that disturbed duff sites were significantly (P < 0.0005) shifted to the right of undisturbed ones (Fig. 7). The classification summary for this analysis indicates that only 50 of the 295 duff sites (16.9%) were improperly classified when anthropogenic disturbance was used as the sole predictor variable. However, no significant differences (P = 0.758) were noted in the location of disturbed vs. undisturbed turf sites (Table 8).

Faunistic turnover between duff and turf soils

As differences in occurrence frequency between duff and turf sites were analysed for all 108 species, the significance threshold was lowered using a Bonferroni correction to P = 0.000463. Species with P-values ranging from 0.05 to 0.000463 were considered to possess statistically nonsignificant trends in their response to soil surface architecture. Species with P-values exceeding 0.05 were considered generalists.

Thirty-six species demonstrated no significant differences in their occurrence frequencies between duff and turf soils (Appendix I). Sixteen of these (Cochlicopa lubricella, Deroceras spp., Discus cronkhitei, Gastrocopta armifera, Gastrocopta contracta, Haplotrema concavum, Hawaiia minuscula, Helicodiscus parallelus, Helicodiscus singleyanus, Planogyra asteriscus, Striatura ferrea, Striatura milium, Triodopsis multilineata, Vallonia costata, Vertigo pygmaea, Vitrina limpida) were found in 10 or more sites, and clearly represent generalists. However, the remaining 20 (Carychium nannodes, Cepaea nemoralis, Discus patulus, Glyphyalinia wheatleyi, Helicodiscus inermis, Mesodon pennsylvanicus, Mesomphix cupreus, Mesomphix inornatus, Oxychylus cellarius, Oxychylus draparnaudi, Oxyloma peoriensis, Pomatiopsis lapidaria, Pupilla muscorum, Succinea indiana, Trichia striolata, Triodopsis denotata, Triodopsis fosteri, Vallonia excentrica, Vertigo modesta parietalis, Zonitoides limatulus) are known from fewer locations, making Type 2 errors a significant concern. Additional data will be needed to adequately assess the response of these species to duff vs. turf soils.

Eighteen species (Cochlicopa lubrica, Cochlicopa morseana, Discus macclintockii, Mesodon clausus clausus, Mesodon thyroidus, Punctum minutissimum, Punctum vitreum, Stenotrema barbatum, Strobilops aenea, Strobilops labyrinthica, Succinea ovalis, Triodopsis albolabris, Triodopsis alleni, Triodopsis tridentata, Vallonia perspectiva, Vertigo meramecensis, Vertigo modesta modesta, Vertigo tridentata) nonsignificantly favoured duff sites. Another eight (Gastrocopta procera, Gastrocopta rogersensis, Gastrocopta similis, Hawaiia n.sp., Helicodiscus n.sp., Striatura exigua, Vallonia pulchella, Zonitoides nitidus) nonsignificantly favoured turf sites. While a number of these (e.g. Discus macclintockii, Gastrocopta procera, Gastrocopta rogersensis, Vertigo meramecensis) demonstrated very strong absolute preferences, their few total occurrences in combination with the conservative Bonferroni correction prevented them from exhibiting significant responses.

The remaining 46 species demonstrated clear and significant soil surface preferences. Twenty-eight species (Allogona profunda, Anguispira alternata, Carychium exile, Catinella ‘gelida’, Columella simplex, Discus catskillensis, Euconulus fulvus, Euconulus polygyratus, Gastrocopta corticaria, Gastrocopta holzingeri, Gastrocopta pentodon, Glyphyalinia indentata, Glyphyalinia rhoadsi, Guppya sterkii, Helicodiscus shimeki, Hendersonia occulta, Nesovitrea binneyana, Paravitrea multidentata, Stenotrema fraternum, Vallonia gracilicosta, Vertigo bollesiana, Vertigo cristata, Vertigo gouldi, Vertigo hubrichti, Vertigo n.sp., Vertigo paradoxa, Zonitoides arboreus, Zoogenetes harpa) favoured duff soils, while another 18 (Carychium exiguum, Catinella avara, Catinella exile, Catinella ‘vermeta’, Euconulus alderi, Gastrocopta tappaniana, Nesovitrea electrina, Oxyloma retusa, Punctum n.sp., Pupoides albilabris, Stenotrema leai leai, Strobilops affinis, Vallonia parvula, Vertigo elatior, Vertigo milium, Vertigo morsei, Vertigo nylanderi, Vertigo ovata) favoured turf soils.

Habitat type

The six compositional clusters significantly differ in their habitat representations. Clusters A–D generally consist of forested sites while clusters E–F generally consist of grasslands. These results are in agreement with previous studies from other regions, including north-western Spain (Ondina & Mato, 2001), southern France (Magnin et al., 1995), western Switzerland (Baur et al., 1996), Croatia (Stamol, 1991, 1993), Hungary (Bába, 1989) and north-eastern Nevada (Ports, 1996). The contrast between open-ground and forest faunas is not limited to terrestrial gastropods. Other soil invertebrate groups that demonstrate this pattern include fungus-eating microarthropods (Branquart et al., 1995), carabid beetles (McCracken, 1994), terrestrial amphipods (Taylor et al., 1995) and collembola (Greenslade, 1997). In an ordination of global earthworm communities, Lavelle et al. (1995) demonstrated that open-ground and forest assemblages were distinct from the warm-tropics to the arctic. The distinction between forest and grassland faunas thus appears to be a general driving factor in soil biota community composition.

Soil surface architecture

The greater similarity of most lowland forest faunas to lowland grasslands, as opposed to upland forests and rock outcrops (Fig. 2), suggests additional factors underlie observed land snail composition patterns. The potential importance of soil surface architecture is implied as many lowland forest sites (e.g. tamarack, white cedar and most black ash swamp forests), and all lowland grasslands, possess turf soils. Only 8% of sites were improperly classified when soil surface type was used as the sole predictor variable (Table 7). Even this rate may be exaggerated, as most misclassifications were limited to two specific instances. First, even though having turf soils, igneous shoreline habitats had faunas almost identical to surrounding rock outcrop sites. Snails in this habitat, however, were largely restricted to friable accumulations of organic matter under stunted white cedar trees. Secondly, almost all duff sites with faunas similar to upland grasslands had experienced severe levels of anthropogenic disturbance.

Striking differences exist between the species of duff and turf sites: 43% of taxa significantly favoured one soil surface type over the other (even with use of a conservative Bonferroni-corrected significance threshold), while only 15% of frequent taxa (10 + occurrences) showed no preference. Inspection of these data indicate that for eight groups of closely related taxa within the same genus (24 total), one or more significantly favour duff soils, while the other(s) significantly favour turf (Table 9). In another four groups (10 additional taxa), one or more taxa significantly favour one of these soil types, while the other(s) exhibit a nonsignificant preference (Table 9). These 12 groups represent a wide variety of phylogenetic stocks (representing nine families: Carychiidae, Helicarionidae, Polygyridae, Punctidae, Pupillidae, Strobilopsidae, Succineidae, Valloniidae, Zonitidae), shell shapes (five wider than tall, five taller than wide and two equally tall as wide), and maximum shell dimensions (0.8 mm– 12 mm). The presence of so many pairs of closely related duff- and turf-specialist taxa across such a wide range of phylogenies, shell shapes and dimensions suggests that very strong selective pressures between these soil surface types have extended over evolutionary time scales.

It is not possible via the current analyses to definitively identify what such factors might be. They must be limited to the detritusphere (Coleman & Crossley, 1996), as almost 90% of snails live within 5 cm of the soil surface (Hawkins et al., 1998). One possible mechanism is increased competition with living plant roots in turf soils for inorganic nutrients (Lavelle et al., 1995). Another may be the greater organic litter thickness in duff soils, as the abundance (Berry, 1973), diversity (Cain, 1983; Locasciulli & Boag, 1987) and composition (Cameron & Morgan-Huws, 1975; Baur et al., 1996; Barker & Mayhill, 1999) of land snail communities often correlates positively with litter depth. The architecture of organic litter (Burch, 1956; Cameron, 1986; Young & Evans, 1991; Alvarez & Willig, 1993), and the underlying soil (Cameron, 1982; Hermida et al., 2000) may also have strong impacts on land snail community structure.

Similarly, the composition and abundance of other soil taxa communities can be influenced by organic litter depth and architecture, including amphipods (Taylor et al., 1995), microarthropods (Borcard & Matthey, 1995; Branquart et al., 1995; Kay et al., 1999; Whitford & Sobhy, 1999), collembola (Kovac & Miklisova, 1997) and ground beetles (McCracken, 1994). Thus, like habitat type, upper soil layer architecture appears to be another vital factor driving soil biota composition.

Geography

The geographical location of sites also influences community composition, particularly in duff soils (Fig. 5). Each of the three duff clusters have an unique geographical affiliation, with cluster A being largely restricted to the most northern sites, cluster B to sites in the northern half of the Lake Michigan–Huron basin, and cluster C to sites in Iowa, Illinois, and southern Wisconsin.

While a significant correlation with both latitude and longitude was also observed in turf sites, this result is almost certainly an artefact of the limitation of upland grassland sites to the south-west of the study region. Fens and lowland forests, found throughout, exhibited little geographical trends inside of the ordination diagram. Geographical location was presumably less important for these sites due to the overriding importance of habitat type and soil moisture.

Soil moisture and temperature

Soil moisture and sunlight levels also appear to influence land snail community composition in turf sites, with the driest and sunniest habitats (upland grasslands) being most different in composition from wet, shaded lowland forests. However, temperature and relative humidity, not sunlight, are probably the important driving factors (Suominen, 1999), as in both duff and turf sites the coolest and wettest habitats (northern cliff, upland forest and lowland forest) were most different in composition from the hottest and driest sites (southern cliff, upland forest and upland grassland).

Disturbance and conservation

Anthropogenic disturbance influences snail composition differentially between duff and turf sites. While turf faunas were not impacted, the most disturbed duff sites had faunas more characteristic of upland grasslands. Typical species found on such disturbed sites include Cochlicopa lubrica, Pupilla muscorum, Vallonia costata, Vallonia excentrica, Vallonia pulchella and Vertigo pygmaea. These faunistic differences may be related to differential changes in soil surface architecture with disturbance. Because undisturbed turf soils usually have thinner and less structurally complex organic litter layers, they may be less susceptible to soil compaction (and changes in composition) as compared to duff sites.

As anthropogenic soil compaction negatively impacts soil invertebrates more severely than plants in the same communities (Duffey, 1975), conservation of duff-specialist land snails will likely require protection of the soil litter layer architecture, perhaps by limiting forestry and recreation activities in duff soil sites of conservation importance. While turf sites appear to be more tolerant of human disturbance, this should not indicate that their land snail communities are immune to human activity. For instance, heavy grazing can negatively impact grassland snails (Cameron & Morgan-Huws, 1975), while the use of fire-management can lead to significant reductions in both species richness and abundance (Nekola, 2002b).