Australia, due to its long isolation, lack of major recent geologic activity, and climate, provides an atypical situation in which to investigate freshwater biogeography. It has lacked terrestrial connections with other continents sin ce 95 Ma (Veevers & Eittreim, 1988). Australia also lacks significant relief due to lack of recent major orogenic events. The last one occurred in eastern Australia ca. 90 Ma, and other mountainous areas are substantially older (Veevers, 1984). Pre sent-day drainages were mostly established by Paleocene. Since then, only minor uplift and subsidence has occurred. Pleistocene glaciations were geographically limited, with little probable effect on the aquatic biota.
Having a broad latitudinal spread (ca. 40o), the continent has likely been buffered from complete loss of major climatic types (i.e., temperate, tropical). Furthermore, the direction of continental drift (of ca. 30o) has been lar gely due north during Tertiary, a time of general global cooling, hence on a broad scale maintaining a diversity of climatic types as it moved from the South Pole. The last 15 My has seen increasing aridity, resulting in decreasing surface water. Consid erable attention has focused on Pleistocene events, especially the last glacial maximum. These short-term, regular climatic fluctuations are thought to be controlled by Milankovitch Cycles associated with regular changes in distances of the Earth relativ e to the sun.
How one should approach biogeographic study of such a landmass, and in general can be a contentious issue. The debate between vicariance and dispersalist biogeographers has, however, subsided, with vicariance presently considered more common (Humphrie s & Parenti, 1986). Today's debates mostly center around algorithms, phylogenetic techniques, and artificial three-taxon datasets. A further confounding issue is the lack of independent tests for hypotheses. My philosophy, given the last, is to exa mine data using different techniques (allowing for limitations within each technique) and look for concordance between and among results.
In contrast to vagile terrestrial organisms, freshwater organisms (e.g., fishes) suffer unique biogeographic constraints. A freshwater fish is defined as one which cannot survive more than a short time of any life-cycle stage in seawater. Furthermore , their ability to move in response to climate change or geological event is limited to patterns of connectivity of freshwater bodies. Hence, opportunities for range expansion between isolated drainages is limited to rare events such as drainage re-arran gements, changes in continental shelf width and depth which alters connectivity between regular sea-level changes, and perhaps major pulses of freshwater into oceans. Given the difficulty of dispersal, extinction is far more likely than colonization and can occur due to any number of factors, both deterministic (climatic change) and stochastic (disease, volcanism, interspecific factors, etc).
While the great antiquity of some Australian groups, e.g., lungfishes and bony tongues, has been long recognized, many suggest radiation of other groups was relatively recent (Whitley, 1959; Allen, 1982; Merrick & Schmida, 1984; Williams & Alle n, 1987; Allen, 1989). This is partly due to an unsubstantiated belief that because a majority of families are considered secondarily freshwater (Myers, 1938), they entered freshwater only in the last few million years. Others put forth alternative view s. Based on limited fossil evidence, Hills (1956) suggested "the chief genera of the extant freshwater fishes of Australia were present in the continent during much if not all of Cainozoic time." Crowley (1990) speculated craterocephalids (and likely me lanotaeniids) invaded Australian freshwaters sometime between Mid-Cretaceous and Paleocene. Her belief was based on the conservative nature of modern atherinimorph osteology, and was the first attempt to integrate distributional patterns and phylogenies of Australian fishes specifically to the geological record for a specific family.
This work reports an investigation of biogeographic patterns of the freshwater fish fauna within Australia. Richness and endemism are quantified and patterns identified on a regional scale. The fauna is examined for congruent distributional patterns among species. Shared species between drainages is taken to imply connectivity; hence, biogeographic tracks can be hypothesized and compared with the geologic and climatic record. Furthermore, biogeographic provinces for fishes can be more clearly defin ed. These results (distributional patterns and geological interpretations) should be viewed as working hypotheses to be tested by phylogenetic analysis. While some may argue biogeographic studies lacking phylogenetic data are unfalsifiable, I provide hy potheses that are directly testable and refutable in the hope of stimulating additional research.
GEOLOGIC AND CLIMATIC SETTING
General background A>
The western two thirds of Australia have exposed Precambrian blocks (3.5 - 2.5 Gy) with thin Phanerozoic basins, while exposed Phanerozoic fold belts alternate with younger basins in the eastern third. Except for the Arafura and Tasma n seas, the shoreline has remained in about the same configuration for the last 300 My (Veevers, 1984), with exception of the last continental transgression during Mid-Cretaceous. The most recent major orogeny was 90 Ma along the Eastern Highlands. All major sedimentary basins (hence river basins) of today were established by Paleocene (Veevers, 1991). Since then, Australia has remained relatively quiescent, with only minor, continent-wide uplift and subsidence (Wasson, 1982).
Recent geomorphological work in Australia has seriously challenged long-standing northern-hemisphere-based paradigms of landform evolution (Gale, 1992; Twidale & Campbell, 1995; Nott, 1995; Twidale, 1997). For example, parts of the Kimberleys (WA) have been continuously exposed for the last 700 My, one of the oldest known, continuously exposed landforms in the world (Ollier et al., 1988). Some incredibly low erosion rates have been calculated. Nott (1996) detected only 500 m of escarpment retreat on Groote Eylandt (NT) over the last 100 My; and Bierman & Turner (1995) recorded mean weathering/erosion rates of 0.7 ± 1 mm / My on inselbergs (granitic domes). With few exceptions, erosion rates are typically less than 10 m / My for the entire Mesozoic-Cainozoic (Gale, 1992).
The last major glaciation was during Permian. Since then, glaciation was restricted to Mount Kosciusko (NSW) and parts of the Central Plateau (TAS) during Pleistocene (Colhoun & Fitzsimons, 1990).
One of the greatest basaltic provinces in the world in terms of extent (although not by volume), stretches 4400 km from Torres Strait in northeastern Australia southward along the eastern mountain ranges into TAS and westward into sout heastern SA with some gaps of up to 500 km (Johnson, 1989). Volcanic activity began in eastern Australia in Late Mesozoic and continued through Quaternary, ending as recently as 13 Ka in northern QLD and 4.6 Ka in southeastern SA. Three broad volcano ty pes are recognized: lava fields, leucitites, and central. Most lava fields are on (within 100 km) or east of the Eastern Highlands, leucitite suites are west of the Eastern Highlands, while central volcanoes appear more spatially random. Lava field acti vity began ca. 70 Ma with the major pulse between 55 and 30 Ma. Little activity occurred between 30 and 5 Ma, after which activity again increased. The oldest identified central volcanic activity was 34 Ma in the north, continuing southward as new volca noes were created and northern ones became extinct. By 11 Ma, all volcanic activity had ceased in NSW, with small areas continuing in QLD and SA, as noted above. Causes of volcanic activity are unresolved; however, central volcanoes and leucitites appea r related to a hotspot trail; lava fields are thought related to post-rifting uplift that lead to formation of the Eastern Highlands.
Long-term sea-level change s
Causes of long-term sea-level change include changes in quantity of water, volume of ocean basins, and uplift and subsidence of continental margins (Partridge et al., 1995).
A significant portion of Australia was inundated by shallow seas during Mid-Cretaceous (Aptium-Albian) times (Frakes et al., 1987a). Up to four principal areas remained emergent, southwestern WA, northern WA and parts of adjacent NT, and the ea stern highlands (including TAS), separated in the middle by a narrow seaway in the vicinity of Brisbane (QLD). Withdrawal was likely a combination of broad, gentle, continental uplift and decreasing sea-level. Global mean sea-level (ignoring Milankovitc h-scale (i.e., glacial) fluctuations) fell on average throughout Tertiary and was at its lowest several times over the past 10 My at around -100 m (Haq, Hardenbol & Vail, 1987).
Throughout Tertiary, submergence was largely limited to coastal margins in three major periods (during each of which several transgressions may have occurred). The first was Eocene, when most southern and western coastlines were affected from eastern VIC through the mid-northern coast of WA. The most significant areas inundated include parts of the lower Murray Basin (SA, VIC, and NSW), western Gippsland and Bass Strait (VIC), and the Great Australian Bight (SA and WA) (McGowran, 1989). Oligocene is generally regarded as a time of lower seas; however, a second transgression occurred during Late Oligocene - Early Miocene resulting in flooding of the Murray and Eucla basins (SA and WA). Flooding of the Murray Basin lasted for ca. 20 My (from 32 Ma; B rown & Radke, 1989), then the sea retreated in Mid-Miocene (ca. 10 Ma), to end with a further, final transgression into the Murray Basin in Late Miocene - Early Pliocene (Stephenson & Brown, 1989; Frakes, McGowran & Bowler, 1987b). The Eucla Basin had by then been uplifted and was not affected (Jennings, 1967; Benbow, 1990). Connections of the Australian mainland with New Guinea (Doutch, 1972), and likely also with TAS, occurred throughout much of the Tertiary.
Australia has the lowest relief of any continent. The highest peak is Mt. Kosciusko (NSW) at 2228 m in the southeastern highlands, the lowest is Lake Eyre (SA) at 16 m below sea-level. The continent consists of a plateau in the weste rn portion and a lowland region in the east bordered by the Eastern Highlands, the only major mountain belt in Australia, along the east coast. Relief is shown in Fig. 3.
Long-term global climate c hange
Long-term global climate is influenced by several factors (Partridge et al., 1995), some with slow gradual impacts, others with dramatic, short- and long-term effects. Continental movements may have significant influence on oce an currents, which in turn influence climate. During Tertiary, considerable areas of continents were uplifted, causing both changes in atmospheric circulation and decreased temperature due to higher elevation. Uplift further promotes weathering and eros ion, increasing the exposure of silicate rocks, weathering of which results in removal of CO2 from the atmosphere, decreasing the "greenhouse effect." Throughout Tertiary there was decreased sea-floor spreading, hence less ridge formation, and ocean basins increased in size, leading to sea-level decrease by as much as 200 m and important albedo effects. Shading by dust created by volcanism also causes atmospheric cooling, although its influence is relatively short term.
The overall trend in world climate through Tertiary was of decreasing temperature, with significantly accelerated cooling at 14 and 2.8 Ma. The first is thought to relate to formation of major ice sheets on Antarctica, and the second to establishing i ce sheets in the Northern Hemisphere (Burckle, 1995). Development of polar ice caps had a critical influence on climate, with extremes becoming broader and changing more rapidly than before. Glacial maxima were times of low sea-level, minimal temperatur es, and drier conditions in the tropics; conditions were warmer at high latitudes and wetter at lower ones during interglacials (Williams, 1984). Glacial maxima and minima typically last around 10 Ky, and each recurs around every 100 Ky. This periodicit y appears controlled by Milankovitch Cycles representing variations in revolutions of earth, both on its axis (obliquity and wobble) and around the sun (ellipticity). Three time-scales of Milankovitch Cycles, 23, 41, and 100 Ky are recognized, broadly co rresponding to observed climatic cycles. These cycles have likely been responsible for short-term climatic fluctuations throughout Earth's history (Bennett, 1990).
Geological mechanisms for climate change
Australia had minor areas of uplift over the last 90 My which had little influence on local climates. Northward drift of Australia has, however, had significant long-term influence. While drift alone did not drive climate change, it
brought the continent toward warmer latitudes. At continental breakup (95 Ma), southern-most Australia was around latitude 76oS. The southern coast was near 70oS at the beginning of Tertiary, while northern Australia was near 40
Botanical evidence for cli mate change
Temperate and tropical rainforest dominated Australian landscapes in Early Tertiary. Based on fossil evidence, this flora was never uniform north-south or east-west (Martin, 1994), hence the following summary should be taken only broa dly. With gradual decline in wetter conditions, open-canopy forests dominated by a sclerophylous flora of Myrtaceae, Mimosaceae, and others, gradually became common (Kershaw, Martin & McEwen Mason, 1994). The first grasses are recorded in Pliocene, reflecting development of drier conditions. Vegetative communities were broadly similar to those of today by the end of Pliocene, although their boundaries and composition are now changed due to more extreme climatic fluctuations (Kershaw et al., 1994).
Through modeling plant growth relative to environmental factors, Nix (1982) suggested similar conditions to today have existed for 150 My in Australia, with additional periods of increasing aridity and expansion of arid zones since Mid-Miocene. Althou gh boundaries have repeatedly shifted, similar conditions have persisted in parts of various regions. Hence, with its northward drift into warmer latitudes combined with global climate cooling, Australia has maintained a broad range of climatic types. < /P>
Geologic evidence for deve loping aridity
Examination of paleodrainages in WA led Van de Graaff et al. (1977) to suggest discharge volumes sufficient to modify drainages ceased in Mid-Miocene. This was supported by Clarke (1994), by adding that gypsum precipitation beg inning in earliest Pliocene (5 Ma) denoted an evaporative environment. While palynological data are lacking for the western half of Australia, available evidence suggests aridity developed there first and progressed eastward, e.g., the first evidence of aeolian landforms are sand dunes from Lake Amadeus (NT) dated at 0.91 Ma (Chen & Barton, 1991). No aeolian deposits west of Lake Amadeus have been examined; however, further evidence of aridity progressing eastward was provided by dated deposits from Paleo-Lake Bungunnia (SA, VIC, and NSW) at ca. 0.5 Ma (Zhisheng et al., 1986).
Figures 4 and 5 show today's mean annual rainfall and January maximum temperature. Around 70% of Australia is considered arid. Southern areas, including southwestern WA, VIC, TAS, and southeastern NSW all share similar Mediterranean climates, with cold, wet winters and hot, dry summers. Northern areas have a tropical climate divided into dry winters and wet summers. Cyclones are regular occurrences. Lowest mean annual rainfall is 110 mm in the vicinity of Lake Eyre (SA); highest i s 4252 mm at Tully (QLD).
Past ichthyological work
The Australian freshwater fish fauna has long been recognized as distinctive relative to the remainder of the world. However, few papers have dealt with biogeography of the entire fauna as it relates to other continents. Most contain short references to specific groups which highlight its distinctiveness. Wallace (1876) recognized relationships between Australia and South America based on galaxiids, aplochitonids, and bovichtids (all of which have marine-tolerant species) and also w ith Southeast Asia based on living Scleropages spp. He also noted many of Australia's genera were endemic and the scarcity of most ostariophysan fishes. McCulloch (1925a) added that most Australian groups had marine relatives and several species were shared with New Guinea. Hills (1934) suggested connections between Australia and North America for Eocene fossils of †Phareodus and †Notogoneus found in both continents (although the latter also occurred in Asia and Europe (Grande, 199 6)). Whitley (1943) discussed similarity of species from Australia and New Guinea. Darlington (1957) gave a summary of the distribution of fish families of the World, including Australian groups. Whitley (1959) pointed out the depauperate nature of the Australian fish fauna and its similarity to New Guinea. Darlington (1965) only briefly mentioned the Australian fauna, adding little except emphasizing dissimilarity to South America. McDowall (1981) was first to compare the number of all genera and sp ecies between Australia and other areas of the World. He also categorized Australian families into endemics, pantropicals, southern, Indo-Pacific, and unknown groups, and summarized their occurrences. Little progress has since been made. Allen (1989) u pdated McDowall's (1981) table of the number of genera and species of each family and Banarescu (1990) provided an account of the families discussed by McDowall based on Myers (1938) categories rather than a distributional one.
The first construction of biogeographic regions based on fishes and mollusks divided Australia into nine regions (Iredale & Whitley, 1938). Lake (1971) divided Australia by river basins based upon a hydrologic system (Bauer, 1955). This system, w ith trivial modifications (Australian Water Resources Council, 1976), remains in general use today (Merrick & Schmida, 1984; Allen, 1989).
Fossil fishes< /A>
Few fossil taxa are recorded in the primary literature, and besides Hills' records (1934, 1943, 1946), they are rarely identified beyond family (and then only tentatively) and sometimes not beyond "fish." Fossil fishes from Australia reported in the primary literature are in Appendix I. Neoceratodontid fossils are relatively well known (Kemp, 1982a, 1982b, 1992, 1993, 1997a, 1997b; Kemp & Molnar, 1981). Around thirteen neoceratodontids were present during Tertiary (and earlier) through until Miocene from LEB, MDB, Gulf of Carpentaria (QLD), and coastal drainages between Brisbane and Rockhampton (QLD). The one living species, Neoceratodus forsteri, has existed for at least 100 My (Kemp & Molnar, 1981). Several neocer atodontids had geographic ranges similar to extant fishes, suggesting either present-day fish distributions were attained prior to Miocene or barriers between drainages have since been overcome.
Most knowledge of Australian fossil Tertiary freshwater actinopterygians stems from Hills' work (1934, 1943, 1946) on QLD deposits. Taxonomy of the two species to which Hills compared fossil material has, however, changed. Macquaria (=Perca lates) colonorum was divided into M. colonorum and M. novemaculeata (Williams, 1970) and Maccullochella macquariensis was separated into M. p. peelii, M. p. mariensis, M. ikei, and M. macquariensis I> (Berra & Weatherley, 1972; Rowland, 1993). As a result, his comparisons can be only broadly interpreted.
Minimum dates for a few taxa are provided by fossils. †Macquaria antiquus (Hills), an ancestor to M. colonorum-M. novemaculeata, and Scleropages cf. leichardti, are recorded from Eocene deposits (>45 Ma) (Hills, 1 934; Vickers-Rich & Molnar, 1996). A terapontid is also recorded from Eocene deposits (Turner, 1981, 1982; Henstridge & Missen, 1982). Maccullochella "macquariensis" is recorded from two deposits dating between 13.6-17.1 and 12-21 Ma (Hills, 1946; Johnson, 1989; Tulip, Taylor & Truswell, 1982). Excepting the following examples, all fossil actinopterygians have representatives living today in the same geographic areas. The record of †Macquaria antiquus is about 250 km f urther north than the present distribution of its descendants. Likewise, the record of Scleropages cf. leichardti at Gladstone (QLD) is in the next drainage south of its descendants present range. Ariid catfish were found during Miocene ar ound Lake Eyre (SA), where they no longer occur (Pledge, 1984; Estes, 1984).
Conditions Allowing Fish Movement between drainages< /A>
A frequently cited explanation for fish species occurring on opposite sides of a mountain divide is "river capture." Bishop (1995) provided a geomorphological review of such drainage rearrangements in bedrock systems. Three possible types were identified: beheading, capture, and diversion. Diversion can be further divided into channel migration, divide-topping catastrophic events, and tectonic diversion. He concluded that beheading by headward erosion is unlikely and capture would occur only under rather restrictive conditions. Diversion via channel migration and catastrophic event are unlikely in headwaters, while tectonic diversion would occur only under certain conditions depending upon appropriate trunk and tributary gradients and the dip and strike of the tilting. Hence, he concluded drainage rearrangements are likely not as frequent as some biogeographers suggest.
It was long accepted that drainage rearrangements were common across the Eastern Highlands of Australia (Taylor, 1911; Ollier, 1978, 1995; Ollier & Pain, 1994; and others). However, recent re-evaluation (Young, 1978; Bishop, 1982, 1986, 1988; Nott , 1992; Young & McDougall, 1993; and others) has largely overturned previous views and concluded that drainage divides along the Eastern Highlands have remained essentially unchanged since Mesozoic (Van der Beek, Braun & Lambeck, 1999). This has important consequences for past biogeographic work which has emphasized drainage rearrangements as explanations of observed patterns (Musyl & Keenan, 1992, 1996; Rowland, 1993; Waters, Lintermans & White, 1994; Pusey & Kennard, 1996, Pusey, Ar thington & Read, 1998; Hurwood & Hughes, 1998). Evidence clearly demonstrates some fishes have crossed the Eastern Highlands (Crowley, 1990; Hurwood & Hughes, 1998; this study), but the mechanism by which this was accomplished remains elusive .
Volcanic activity has also been suggested as a means for drainage rearrangement (Hurwood & Hughes, 1998). Volcanic activity on, or very close to a drainage divide might be likely to cause changes via tectonic diversion through associated uplift, c oncurrent to and continuing after extrusion of lava (P. Wellman, pers. comm.). Lava outflows would have little effect as they always move downhill, then the only way such could divert a waterway would be to dam it high enough to overtop its divide, a rar e situation, except where the drainage divide is very low.
Banarescu (1990) identified several ways by which fish move between drainages, including temporary connections between headwaters through low divides, i.e., swampy regions (i.e., Waldai Plateau, Brazil) or connected headwaters which drain both directio ns (i.e., Two-Oceans Pass, Wyoming, USA). "River captures" have been dealt with above. Both causes would primarily affect headwater-dwelling species. Lowered sea-level also may connect the lower reaches of rivers, with degree of connection depending up on local topography of the continental shelf. Also, inland salt lakes may become fresh, allowing connections across them. Both changes would allow primarily lowland fishes to move.
Possibilities for random dispersal of fishes include "rains" and accidental movement of eggs by terrestrial organisms (i.e., birds, mammals, etc.), or whole fish being dropped by birds. "Rains of fish" have been reported in Australia (McCulloch, 1925b ; Whitley, 1972; Glover, 1990) and elsewhere (Gudger, 1929). Some Australian reports are clearly examples of fish moving short distances by overland flow (i.e., Shipway, 1947) rather than falling from the sky, but some clearly indicate fish have fallen. They are presumably picked up and deposited nearby since tornado-strength winds of sufficient velocity are often of short duration, and if transport is only over short distances, moving fish between drainages is unlikely. Clearly, the implications of th is type of dispersal remain poorly understood. Eggs caught on animal feet/feathers/fur are often used to explain fish appearances in formerly uninhabited places. Not all fishes would be susceptible to such transportation; only egg-layers could be transp orted, eggs would need to be adhesive or in adhesive mud and in shallows where birds or mammals are likely to walk, or deposited near the water surface, and would need to survive at least brief periods of aerial exposure. Subsequently, at least one male and one female must reach adulthood and find each other to mate. Movement via this means has never been documented, although clearly possible over short distances.