Introduction

‘To try to understand present-day environments without a knowledge of their history since the last ice age would be like trying to reconstruct the plot of a long novel by reading only the last page.’

SA Elias (1995), The Ice-Age History of Alaskan National Parks (p 4)

The many ‘ice ages’ of the last million years have had an enormous effect on the physical landscapes of much of the world, evident in terminal moraines, glacial erratics and deep valleys. Even the sunny climes of New Guinea and Australia have hosted glaciers (Frakes et al., 1987). One of the most influential effects of extreme Quaternary climate change was the many oscillations in sea level, with concomitant alterations in coastlines. This reordering of the landscape had significant flow-on effects for the plants and animals that lived within and shows that neither landscape nor biota is static. As the continental shelf drained, new land and dispersal routes opened up; and as sea levels rose, populations were sundered. This means that the imprint of the ice ages should be evident in genes as well as rocks (Hewitt, 2000).

Continental islands provide an excellent setting in which to study the biological effects of repeated sea-level change because of the intermittent nature of available dispersal routes (Bittkau and Comes, 2005). Southeast Queensland in eastern Australia has four such islands (North Stradbroke, Moreton, Bribie, Fraser) composed almost entirely of sand; the first two of which are in Moreton Bay to the east of Brisbane (Figure 1). The formation of these islands is fascinating in its own right, as their histories, like the rings of a tree, are written in the multiple, sequential sand deposits that formed during various sea-level changes of the Pleistocene (Laycock, 1978; Kelley and Baker, 1984). A combination of prevalent onshore winds from the southeast (Pickett et al., 1985) and changing sea levels (Ward, 1978; Kelley and Baker, 1984) built up a series of sequential and partially overlapped parabolic dunes, each reflecting a specific period of sea-level change (Pickett et al., 1985; Ward, 1985). There are fairly good correlations between dune units on Fraser, Moreton and North (N.) Stradbroke Islands (Ward, 1978; Stock, 1990; Grimes, 1992). The oldest dates obtained for sands at Cooloola (Tejan-Kella et al., 1990) and sediments at Fraser Island (Longmore, 1997) are in the order of 600–730 thousand years ago (‘kya’).

Figure 1
figure 1

Sampling locations for different C. indistincta sp.C lineages (C1, C2, C3) in the sand dune areas of southeast Queensland, eastern Australia (see Table 1 for site details).

Full size image

Moreton Bay has been a flat, broad plain during much of this time, over which many rivers flowed eastward. These rivers would have gone past large sand dune fields and continued a further 25 km to the palaeo-coastline on the continental shelf (Neal and Stock, 1986). Interglacials occurred roughly every 100 000 years (Kershaw et al., 2003). When global climates warmed, the sea came flooding in and Moreton Bay became a ‘drowned landscape’ (Ward, 1978), isolating the composite dune fields as separate islands (as it appears today).

This area hosts the world's largest sand island (Fraser Island, 1653 km2), the highest coastal dune (Mt Tempest, Moreton Island, 280 m) and the largest concentration of perched lakes. Despite the sandy substrate, these islands are well vegetated with a mix of environments from lowland coastal heath to rainforest. Low pH freshwater lakes formed after the dunes stabilised. The water bodies are both ‘water table window’ lakes and swamps, which are an expression of the groundwater at the surface, and ‘perched’ lakes, which are set above the regional water table with their own local aquifers.

The dunes on the islands of Moreton Bay that underlie many of water bodies have been characterised as ‘Awinya’ (Qpa; sensu Ward, 1978) and ‘Yankee Jack’ (Qpy) (Ward, 1978; Stock, 1990). Pickett et al. (1985) used isotope ratios (230Th/234U) to date corals under Awinya sands in northwestern N. Stradbroke Island to 105 kya (revised to 119–132 kya; Tejan-Kella et al., 1990). The ‘Yankee Jack’ dune unit overlies Awinya and so must be younger (105 kya, Kelley and Baker, 1984; ∼90 kya, E Stock, personal communication).

Fully freshwater animals from insular environments are excellent model organisms to study the relationship between biota and landscape because of their dispersal limitations due to the lethal nature of intervening dry land or salt water. The freshwater shrimp Caridina indistincta Calman, 1926 (Decapoda: Atyidae) is widespread and common throughout the area, on all of the islands and the mainland. Chenoweth and Hughes (2003) included a few shrimp from Moreton Island in a phylogenetic study and found they represented a cryptic species (Caridina indistincta sp.C). In another phylogenetic study, Page et al. (2005) identified this species from a small number of sites in other sand areas (Cooloola Sand Mass, Fraser and N. Stradbroke Islands) and found it has large eggs, which is indicative of a fully freshwater lifecycle (Hayashi and Hamano, 1984).

A powerful method for inferring biological history, both evolutionary and biogeographic, is the phylogeographic analysis of DNA sequences, and subsequent dating of genetic divergences. This can then be related to current distributions and known geological and geomorphological history to understand their relationships. This has been done effectively with shrimp to investigate effects of Pleistocene climate change in western Queensland and Malaysia (Carini and Hughes, 2004; de Bruyn et al., 2004).

The climate and sea-level changes of the last 18 000 years (last glacial maximum, ‘LGM’) wrought great changes on the landscapes of the world in general and Moreton Bay in particular, with sea levels 150 m below the present day level, exposing the entire continental shelf and thus providing opportunities for dispersal. This most recent glaciation and the subsequent warming and sea-level rise (17 000–6500 years ago; Neal and Stock, 1986) had significant effects on the intraspecific genetic structuring of many terrestrial and aquatic fauna throughout the world (Hewitt, 2000 and references within), including in the rainforests of Queensland (Schneider et al., 1998).

Barry and Campbell (1977) considered the distribution of various terrestrial vertebrates on N. Stradbroke Island largely determined by the ‘land bridge’ that existed during the last glacial maximum. Given this suggestion, and the demonstrated effect of the last glacial maximum on the population structures of many taxa around the world, we hypothesise that dispersal opportunities during the LGM should have homogenised the genetic structure of freshwater taxa on these islands, especially given the islands’ proximity to the mainland and the fact that they were a part of the mainland until very recently in geological and genetic terms.

A second, alternative, hypothesis is that the LGM, and the Pleistocene itself, did not have the all-encompassing influence that has sometimes been supposed. Interestingly, divergences considerably older than the last glacial maximum and the Pleistocene have also been detected in many populations whose structure had been assumed to have been determined by recent glaciation (Schneider et al., 1998; Hewitt, 2000; Moritz et al., 2000; Bittkau and Comes, 2005).

The two hypotheses of the overriding influence of either the LGM or the Miocene/Pliocene to explain the phylogeographic structure of C. indistincta sp.C on the sand islands of Moreton Bay can be tested by relating current distributions, genetic divergences, climate history and sand dune ages. Should there be little phylogeographic structure, then the relatively recent LGM was likely most influential. Should the structuring be so deep as to predate the dune landscape, then the Miocene/Pliocene divergences seen in many east Australian freshwater taxa (see Page et al., in press) may also exist within these taxa, even at such small geographic scales.

Materials and methods

Sampling

C. indistincta sensu lato were extensively sampled from every basin (135 sites in 36 basins) within an arc of 500 km emanating from Moreton and N. Stradbroke Islands, with an emphasis on these islands and adjacent mainland areas. Twenty-four sites were sampled on N. Stradbroke Island and nine on Moreton Island, and shrimp were found at nine and seven of these sites, respectively. The other sand dune areas of the region were also sampled (Fraser Island: 21 sites, Bribie Island: nine sites, Cooloola Sand Mass: four sites). Shrimp were caught with seine or dip-nets and preserved in liquid nitrogen or 95% ethanol.

DNA extraction, polymerase chain reaction and sequencing

Genomic DNA was extracted and a fragment of the mitochondrial cytochrome oxidase subunit I gene (‘COI’) was amplified and sequenced as per Page et al. (2005). All specimens were sequenced with the forward primer (CDC0.La; Page et al., 2005) and a third of specimens also with the reverse primer COIa.H (Palumbi et al., 1991).

Construction of data sets

A total of 426 C. indistincta specimens were sequenced (combined from this study, Chenoweth and Hughes, 2003; Page et al., 2005 and Page and Hughes, unpublished). The only species found on Moreton Island and N. Stradbroke Island was C. indistincta sp.C (‘Sp.C’) (sensu Chenoweth and Hughes, 2003; Page et al., 2005) and so only this taxon was used for subsequent analyses (150 Sp.C COI sequences of 450 bp). One hundred and eight of these sequences were generated for this study (new Genbank accession numbers DQ656417-DQ656436) to which were added a further eight and 34 sequences, respectively, from Chenoweth and Hughes (2003) and Page et al. (2005) (Table 1). All 33 unique Sp.C COI haplotypes were included in the COI phylogenetic analyses, with 10 Caridina indistincta sp.A (‘Sp.A’) haplotypes as an outgroup (some from Fraser/Bribie Islands) (Table 1). A small data set of the more conserved mitochondrial 16S ribosomal DNA (‘16S’) was also constructed using sequences from Page et al. (2005) and Page et al. (in press) to investigate deeper phylogenetic relationships within Sp.C. The 16S data set of 11 haplotypes of 492 bp was aligned with Sequencher 4.1.2 (Gene Codes Corporation) at default settings (Table 1).

Table 1 Caridina sampling locations and sequences listed by lineage with Genbank accession numbers
Full size table

Deep phylogenetic structure was observed within Sp.C, falling into four distinct lineages (‘C1’, ‘C2’, ‘C3’, ‘C4’); only two of which were found on Moreton and N. Stradbroke Islands (C1, C2). For this reason, data sets for phylogeographic analyses of Moreton and N. Stradbroke Islands were split into separate sections, namely C1 (72 COI sequences) and C2 (45).

Phylogenetic analyses

The model of evolution was selected using the Akaike Information Criterion in Modeltest version 3.06 (Posada and Crandall, 1998) for both COI and 16S data sets. Three methods of phylogenetic analyses (tree building) were carried out in PAUP* version 4.0 b10 (Swofford, 2002) for each data set, namely Parsimony Analysis and Maximum Likelihood (‘PA’/‘ML’; both full heuristic, 100 random repetitions) and Minimum Evolution (‘ME’). Each data set was bootstrapped 500 times (PA: full heuristic, 100 random repetitions; ML: 10 random repetitions). TreeRot v.2 (Sorenson, 1999) was used to calculate Bremer Support Values for both PA data sets.

Phylogeographic analyses

The relationship between local geography on Moreton and N. Stradbroke Islands and small-scale genetic structure within each lineage (C1, C2) was explored using an analysis of molecular variance (AMOVA; Excoffier et al., 1992) and nested clade analysis (NCA; Templeton et al., 1995). We used Arlequin version 2.0 (Schneider et al., 2000) to partition genetic variation and calculate Φ-statistics given various nested geographical hierarchies in an AMOVA (islands/mainland, between islands, between areas of islands) (10 000 permutations for significance testing).

We used NCA to test whether significant associations existed between particular haplotypes and geographic locations, and further to differentiate historical and contemporary processes. A haplotype network was constructed separately for C1 and C2 using TCS version 1.21 (Clement et al., 2000) and the nesting rules of Templeton et al. (1987) and Templeton and Sing (1993) used to assign clades. Geographical locations and haplotype nestings were entered into Geodis version 2.4 (Posada et al., 2000) to test for significant associations, using the November 2005 Inference Key for NCA (available online at: Darwin.uvigo.es/download/geodisKey_11Nov05.pdf) to infer the most likely biological explanation for the significant clades.

Molecular divergence and clock calculations

COI and 16S sequence divergences between monophyletic clades were calculated using a distance matrix constructed in PAUP* using the suggested models of molecular evolution including a correction for within-clade polymorphism (±s.e.) (Avise, 1994). Divergence rates used were 1.4% per million years for COI (Knowlton and Weigt, 1998) and 0.65 and 0.9% for 16S (Sturmbauer et al., 1996; Schubart et al., 1998). A Likelihood Ratio Test was implemented in PAUP* to test for non-clocklike molecular evolution.

The above method is ineffective in dating divergences between areas that may share haplotypes, so we also used the coalescent method of Nielsen and Wakeley (2001) as implemented in MDIV (Nielsen and Wakeley, 2001; online at: cbsuapps.tc.cornell.edu/mdiv.aspx) to calculate the timing of population divergence (t) and expected timing of most recent common ancestor (tMRCA). We used the following MDIV parameters (model=Finite sites; [HKY], cycles=5 000 000, burn in=10%, Mmax and Tmax=various) and converted the highest likelihood value for T (scaled time) and tMRCA into years using the above COI rate as described in Nielsen and Wakeley (2001).

Rhadinocentrus ornatus data set

A data set of a sympatric freshwater fish, Rhadinocentrus ornatus Regan, 1914 (Melanotaeniidae), was constructed to provide a comparison to the shrimp from this study. Mitochondrial ATPase sequences were obtained from a phylogenetic study (Page et al., 2004). Three new fish specimens were collected and sequenced for the same gene (as per Page et al., 2004) (equals accession number AY452216) from Little Canalpin Creek, N. Stradbroke Island. A haplotype network, NCA, AMOVA and MDIV analyses were carried out on the fish ATPase data set as per the shrimp COI data sets above.

Results

Phylogenetic and geographical relationships of lineages

Four distinct lineages were identified within Sp.C, namely C1, C2, C3 and C4. Both data sets and all analyses support C4 as a divergent sister group to C1/C2/C3, which form a strong clade (PA COI phylogram; Figure 2) (other bootstrap values, COI ML: 52%, COI ME: 91%, 16S PA: 90%, 16S ML: 86%, 16S ME: 82%). The relationship between C1, C2 and C3 is less clear. They form a polytomy in the COI data set, but the 16S analyses (not displayed) supports a clade of C2/C3 (bootstrap values: PA: 89%, ML: 68%, ME: 79%).

Figure 2
figure 2

Parsimony phylogram of COI data set, showing specimen geographic location (Bootstrap values above node and Bremer Support values below).

Full size image

None of the lineages were found sympatrically. Lineage C4 is not only very phylogenetically divergent but is also geographically distinct, only having been located well to the south of the islands of interest, in mainland coastal basins of the state of New South Wales. Furthermore, C4 has not been found in similar sandy habitats, but largely in subtropical forest.

The Moreton Bay islands (N. Stradbroke Island, Moreton Island) are the only known home of C1. It is the only lineage thus far found on Moreton Island, and is also on the western side only of N. Stradbroke Island (Figure 1), but not on the mainland. C2 is also found on N. Stradbroke Island, but only on the eastern side. C2 also exists in disparate populations on the mainland in the state of Queensland (Figure 1), the most significant of which is at Lake Freshwater in the Cooloola Sand Mass. C2 is also found in more rainforest-dominated environments to the south of Cooloola (Figure 1), but these appear to be small populations which are always sympatric with more numerous C. indistincta sp.A (‘Sp.A’), which dominate this area. Very few shrimp were found on Bribie Island, and none of these were Sp.C but instead a clade of Sp.A which is common on the nearby mainland.

The final lineage, C3, has only been located on the western side of Fraser Island. There are two geographically distinct clades within C3 (Figure 1). It is unclear whether these represent separate lineages, because the area between them has not been sampled. Fraser is the only island to have both Sp.C and Sp.A (Table 1), including Sp.A ‘TCB’ (sensu Chenoweth and Hughes, 2003) which is well represented on the nearby mainland.

Divergence between lineages

All COI and 16S sequence divergences are in Table 2, with C4 notably divergent from the others. The Likelihood Ratio Test could not reject clock-like evolution in either data set (COI: P=0.0526; 16S: P=0.2527). Molecular clock estimates for lineage divergences (Table 2) vary considerably, but most place divergences in the late Miocene/Pliocene. These estimates should be approached with caution because of many potentially confounding factors. In particular, variation in divergence rates between lineages (Cranston and Rannala, 2005). For example, despite the COI data set passing the Likelihood Ratio Test (barely), C1 COI sequences do seem to have diverged more rapidly than the others (Figure 2). This is likely to have an effect on the deeper interlineage COI comparisons involving C1, but not the more conserved 16S comparisons, nor the shallow intralineage comparisons, where variation between individuals is much less likely. We present the various calculations in Table 2 to place divergences in a broad geological framework, rather than to date them precisely.

Table 2 Genetic divergence estimates between and within lineages
Full size table

Population structure within C1

The network of the 14 COI haplotypes found on N. Stradbroke and Moreton Islands (Figure 3) is centred around a common haplotype, which is found on both islands. It is connected by a single mutational step to half of all the haplotypes, and is thus presumably ancestral (Templeton et al., 1987). According to the NCA inference key, ‘allopatric fragmentation’ explains most of the significant nested clades (Clades 1–3, 1–4, 2–2, Total Cladogram; Figure 3), whereas ‘restricted gene flow with isolation by distance’ explains Clade 2–1, which corresponds to a linear branch of Moreton Island-only haplotypes.

Figure 3
figure 3

Phylogeography of C. indistincta sp.C1 with haplotype network and NCA clades placed in local geographic context. AMOVA statistics and time since divergence (t) are between adjacent areas (mya=million years ago).

Full size image

When Moreton and N. Stradbroke Islands (western side) are defined as geographic groups in an AMOVA, 26.3% of the variation is partitioned between the islands, but with a non-significant ΦCT (0.263, P=0.1011). A re-examination of both the relative geographic locations (Table 1) and the haplotype network (Figure 3) makes it plain that all of the haplotypes from the northern part of Moreton Island and the southern part of N. Stradbroke Island are unique to their respective distinct areas. In contrast, the areas encompassing the northwest of N. Stradbroke Island and the centre of Moreton Island share haplotypes (including the ancestral one) and only have haplotypes one step from the ancestral. When an AMOVA is carried out with these four areas as the groups (north Moreton Island, central Moreton Island, northwestern N. Stradbroke Island, southwestern N. Stradbroke Island), 62.58% of the variation is partitioned among them, with a high ΦCT (0.6258), which is highly significant (P=0.0006). Separate AMOVAs were also carried out between adjacent areas (north and central Moreton Island: 66.23%, ΦCT=0.6623, P=0.0458; central Moreton Island and northwest N. Stradbroke Island: 8.79%, ΦCT=0.0879, P=0.2128); northwest and southwest N. Stradbroke Island: ΦST=0.2292, P=0.0013 (ΦST rather than ΦCT because only one site in southwestern N. Stradbroke Island).

These same four groups were used to calculate time since population divergence (t) and time to most recent common ancestor (tMRCA) in MDIV, and all fall within the Pleistocene (Table 2). The split between north and central Moreton Island is the oldest (t: 0.30 mya, tMRCA: 0.57) and between central Moreton Island and north N. Stradbroke Island the most recent (t: 0.11 mya (millions of years ago), tMRCA: 0.33).

Population structure within C2

The mainland and N. Stradbroke Island (eastern side) share no haplotypes (Figure 4b). Clade 2–1 (mainland) is either ‘allopatric fragmentation’ or ‘inadequate geographical sampling’ depending on whether one assumes there are unsampled shrimp between the observed mainland locations (there probably are, but in low numbers). Therefore, the Total Cladogram (i.e. mainland vs N. Stradbroke Island) suffers the same fate, but ‘allopatric fragmentation’ is a much more likely explanation given the current presence of Moreton Bay between them, blocking the migration of a freshwater shrimp. The Mainland/N. Stradbroke Island groups explain 80.54% of the variation in an AMOVA (ΦCT=0.80537, P=0.0069). MDIV dates the divergence of these groups at 0.23 mya (tMRCA: 0.42) (Table 2).

Figure 4
figure 4

Nested clade analysis (NCA) of (a) R. ornatus and (b) C. indistincta sp.C2 placed in local geographic context.

Full size image

R. ornatus population structure

Moreton Island fish form a distinct branch (Clades 1–1 and 1–2 in Figure 4a), similar to the shrimp C1 (Clade 2–1 in Figure 3). In contrast, N. Stradbroke Island fish have two separate clades; one from eastern N. Stradbroke Island and related to the Moreton Island clades (1–3, and also containing a mainland haplotype), which NCA explains as ‘past fragmentation and/or long distance colonization’. The other is from western N. Stradbroke Island (1–4) and is separately derived from the common mainland haplotype. Clades 1–4, 2–1 and the Total Cladogram are considered ‘allopatric fragmentation’ and Clade 2–2 is ‘contiguous range expansion’.

An AMOVA with two groupings of Mainland and Moreton Bay Islands is not significant (ΦCT=0.1012, P=0.1061), but is just significant with groupings of Mainland, Moreton Island, N. Stradbroke Island (ΦCT=0.2457, P=0.0484), and is highly significant with Mainland, Moreton Island, eastern N. Stradbroke Island, western N. Stradbroke Island (ΦCT=0.4076, P=0.0079). Divergences calculated with MDIV ranged from 0.07 to 0.66 mya (Table 2).

Discussion

The LGM and the Pleistocene

If the LGM was most influential in the small-scale biogeography of these populations, there should be little genetic structure because of the short times involved (∼18 kya). However, the LGM appears to have had little influence on the population structure of these taxa as there is very limited haplotype sharing, and a great deal of genetic structure at many levels, within and between C1 and C2.

The intralineage phylogeographic structure within both C1 (Figure 3) and C2 (Figure 4b) fall within the Pleistocene (Table 2), but well before the LGM. Comparable patterns in which there is little effect of the most recent ice-age cycle, but significant effects of earlier Pleistocene ice ages (∼400 kya), is also visible in some European taxa (Hewitt, 2000).

Theoretically, biological histories of sympatric taxa should correlate with each other as well as with geological history given a large-scale common geological change (‘comparative phylogeography’; Avise, 1994). The sympatric freshwater fish, Rhadinocentrus ornatus, has Pleistocene intraspecific divergences consistent with those within C1/C2 (Table 2), also considerably predating the LGM. Another sympatric fish, Nannoperca oxleyana, shows low levels of divergence among currently unconnected mainland creeks (Hughes et al., 1999) and thus may display the influence of palaeo-connections during the last glacial maximum. However, Moreton and Fraser Island populations of N. oxleyana are distinct and represent isolation from the mainland dating from earlier in the Pleistocene than the LGM.

The Miocene/Pliocene

Although Pleistocene influence, albeit not the LGM, can be observed in intralineage phylogeographic structure, it cannot account for the deeper between-lineage differences. The genetic divergences between all four Caridina lineages date to the Miocene/Pliocene (Table 2), when many other east Australian freshwater taxa appear to have diverged (Chenoweth and Hughes, 2003; Page et al., 2004; Ponniah and Hughes, 2006; discussed in Page et al., in press). This pattern of unexpected Miocene/Pliocene divergences has also been observed in terrestrial fauna of the wet tropics of Queensland (Schneider et al., 1998; Moritz et al., 2000), the Aegean (Bittkau and Comes, 2005) and elsewhere (Hewitt, 2000).

The C1/C2 divergence fits in well with large-scale climatic history, as Australia became colder and drier during this time (Frakes et al., 1987). However, this time-scale does not equate with the current contact zone between C1 and C2, namely the central dunes of N. Stradbroke Island that separate the east and west coasts. The genetic divergence (Table 2) is an order of magnitude older than the ages of the dunes and the water bodies (Table 3), making the shrimp far too old for their current surroundings. A similar situation has been described for stone loach in Lake Constance, as their population subdivision is much older than the lake itself (Barluenga and Meyer, 2005). Plainly these current landscapes cannot explain these levels of differentiation, and we would need to expand our scale in time and space beyond the Pleistocene and Moreton Bay to fully understand this relationship.

Table 3 Sand dune unit and approximate ages of sampling site water bodies
Full size table

Colonisation history of the Moreton Bay Islands

Given that the intralineage divergences date to the pre-LGM Pleistocene, we may be able to use the Pleistocene geological history of the area to understand something of the colonisation history of its freshwater fauna. The timing of the formation of the islands, dunes and water bodies is important in unravelling the changing landscape's effect on the genetic structure of its aquatic fauna. For our purpose, sand has an advantage over rock as its plasticity makes it more amenable to study at these relatively short time-scales. The precise dating of different dune unit sequences has some of the same inherent uncertainties as for molecular clock dating of DNA sequences referred to above. However, relative ages can be assigned by the stratigraphic relationships of the different dunes and the knowledge of how they formed. For example, because of the direction of prevailing winds, newer dunes tend to form on the eastern ocean side, with older dunes to the west in the centre of the islands (Laycock, 1978).

The divergences within each of the Caridina lineages and Rhadinocentrus fall within the pre-LGM Pleistocene (110–430 kya) and so are consistent with sea-level fluctuations. However, the estimates of genetic divergences (Table 2) and water body ages (Table 3) are not precise enough to correlate with specific glacial/interglacials periods. Interestingly, the population divergences within both Caridina and Rhadinocentrus tend to have a periodicity of about 100 thousand years (C1 divergences: 110, 210, 230, 300 kya; Table 2), which is also the period between major sea-level changes associated with the glacial/interglacial cycle (Kershaw et al., 2003). This implies that these divergences may have been structured by these major sea-level changes.

More precisely, how could landscape-change foster vicariance or dispersal of Caridina in Moreton Bay? When the plain of Moreton Bay was above sea level, the various rivers which now flow into Moreton Bay (Figure 5) would have continued past the stranded dune fields and may have provided the necessary freshwater connections between water bodies that are isolated during high sea levels (Jones, 1992; Lang et al., 1998; Lockhart et al., 1998). Ironically, a time of cold, arid conditions may have fostered the dispersal of a tropically derived freshwater-obligate shrimp.

Figure 5
figure 5

Palaeo-channels of southeast Queensland rivers (Jones, 1992; Lang et al., 1998; Lockhart et al., 1998; E Stock, personal communication), palaeo-coastline (Lang et al., 1998), and cross-section of N. Stradbroke Island. Cross-section created from Dunwich Digital Topographic Data (Code: ANZQL0006000609; ID:5255; Department of Natural Resources, Mines and Water) using Profile Extractor v.6.0.

Full size image

This form of past connection may explain why the currently separated central Moreton and northwestern N. Stradbroke Islands are the most similar of the groups within C1, and how C2 can have colonised the eastern side of N. Stradbroke Island from the mainland without an opportunity to also colonise the west. An alternative possibility is that when C2 arrived in eastern N. Stradbroke Island, it outcompeted a previously extant population of C1. Rhadinocentrus also colonised each side of N. Stradbroke Island independently (Figure 4a) during the Pleistocene, emphasizing the independent colonisation history of sympatric taxa, and the independent histories of the two sides of a relatively small island. The large central dunes of N. Stradbroke Island (Figure 5) seem to have kept the different sides of the island isolated for a significant length of time.

If the rivers flowed past these dunes as recently as a few thousand years ago, why should not the Caridina populations be homogenised? Barry and Campbell (1977) point out that a species ‘missing’ from an island may be due to extinction or to a lack of opportunity to colonise. Obviously, the three dimensional structure of a landscape constrains potential paths of dispersal, especially for an obligately freshwater animal. The vagaries of chance of a river path and intervening dune may have determined whether water bodies were connected during specific glacial periods. The palaeo-channels outlined in Figure 5 may only represent the paths during the LGM and not during previous ice ages (Lang et al., 1998). The aridity and low temperatures (Frakes et al., 1987; Kershaw et al., 2003) associated with these times likely also influenced palaeo-distributions significantly (Ponniah and Hughes, 2006).

Biological and earth histories are intimately connected and both can be reconstructed to tease out the paths they may have taken to arrive at their current positions. The ‘legacy’ of ice ages (Hewitt, 2000) before the most recent one is likely visible in the intraspecfic genetic structure of freshwater shrimp of Moreton Bay, whereas deep divergences dating from earlier periods of Miocene/Pliocene climate change are also present. Further interdisciplinary studies utilising both genetic and geomorphological data together should further clarify the complex relationship between landscape and biota in both time and space.