Relative abundance and habitat association of three crayfish (Orconectes virilis, O. rusticus, and O. immunis) near an invasion front of O. rusticus, and long-term changes in their distribution in Lake of the Woods, Canada

We documented the abundance and habitat association of the native virile crayfish [Orconectes virilis (Hagen, 1870)] and two invasive species, the rusty crayfish [O. rusticus (Girard, 1852)] and the papershell crayfish [O. immunis (Hagen, 1870)] in a 38 km area in the Lake of the Woods (LOW; Canada). From 20 June to 1 September in 2006, traplines were set at 376 sites for approximately 24 hours. A total of 9833 crayfish were captured, of which 68% were O. virilis, 23% were O. rusticus, and 9% were O. immunis. The mean catch per unit effort (CPUE) was 4.4 crayfish per trap with a maximum of 29. At least one invasion front of O. rusticus was detected, where its CPUE decreased sharply from a mean of 1.3 to 0.03. The mean CPUE of both O. virilis and O. immunis was significantly higher outside than within the front. For the entire catch, CPUE values of all three species were negatively correlated. Water depth (0.5-12 m) was significantly related to crayfish abundance, positively for O. virilis and negatively for O. rusticus and O. immunis. The distribution and abundance of the two invasive species on islands and adjacent mainland sites indicated that water depths near and deeper than the thermocline limit the colonisation of islands by O. rusticus and, particularly, O. immunis, but that islands surrounded by shallower water may serve as shortcuts in crayfish expansion circumventing convoluted shorelines. Mean CPUE of all three crayfish species differed between bottom substrates, being highest on mainly rocky substrates for O. virilis and O. rusticus, and on organic and inorganic fines for O. immunis. Mean CPUE of O. virilis near macrophytes was similar to that in areas free of aquatic plants; CPUE near macrophytes was significantly lower for O. rusticus and significantly higher for O. immunis. In conjunction with crayfish captures during lake-wide fish surveys between 1973 and 2006, we show that by 2006 O. rusticus has spread into most parts of LOW since its first record from Long Bay on the east side of the lake in 1963, expanding its distribution by an average of 2.1 km per year. Our results further indicate that O. immunis is more widespread and more abundant in LOW than previously assumed.


Introduction
Recent interest in the ecological impacts and associated economical costs of nonindigenous organisms, including aquatic invasive species (AIS) (Pimentel et al. 2000;Colautti et al. 2006), and recognition of the impacts of invasive crayfish (Berrill 1978;Lodge and Lorman 1987;Gherardi 2007) have lead to a growing concern for the potential effects of introduced crayfish in North America and Europe (Momot 1996;Harlioglu and Harlioglu 2006;Rosenberg et al., in press). For example, the rusty crayfish [Orconectes rusticus (Girard, 1852)] has largely expanded its North American range since the 1960s and has the potential to become an AIS in Europe, following the example of its North American congeners, the papershell crayfish [O. immunis (Hagen, 1870)] and the spiny cheek crayfish [O. limosus (Rafinesque, 1817)] (Holdich 2002).
The rusty crayfish originates from the Ohio River basin and has invaded, largely aided by bait bucket transfer, most states in the north central US (Capelli and Magnuson 1983;Hobbs III et al. 1989) and the Canadian Province of Ontario (Momot et al. 1988), where it was first recorded in Lake of the Woods (LOW) in 1963 (Crocker and Barr 1968). Since then O. rusticus has also been found in other Ontario lakes and rivers (Berrill 1978;Momot et al. 1988;Momot 1996), including Lake Superior (Momot 1996). The native range of O. immunis includes southern Ontario (Crocker and Barr 1968), but not other areas of the province (Winterton 2005). O. immunis was first reported in LOW from Snake Bay sometime after 1975 (Schueler, pers. comm. 1993 in Dubé and Renaud 1994). The papershell crayfish appears to be native to southern Manitoba, where it was first recorded in 1969 (Popham and Hancox 1970). This species seems to be well established in the Red River drainage, but does not occur in the far eastern part of the province (Hamr 1998;Bill Watkins, Manitoba Water Stewardship, pers. comm., April 2007) except for the Winnipeg River (Watkinson and Batty, in press). O. immunis is an AIS in Europe where it was first reported as a likely aquarium release in south-west Germany (Geiter 1998;Dehus et al. 1999) and where it has since spread north and southwards along the Rhine River corridor (Gelmar et al. 2006).
Whereas no or little information exists about the dispersal ability or the interspecific competitive interactions of O. immunis (Chucholl et al. 2008), O. rusticus has the potential of rapid dispersal within lakes and rivers once a new water body has been reached (Byron and Wilson 2001;Kuhlmann and Hazelton 2007), is known to be aggressive towards and extirpate other crayfish species, and to restructure aquatic ecosystems (Lodge and Lorman 1987;Lodge et al. 1994;Taylor and Redmer 1996;Hill and Lodge 1999;Wilson et al. 2004). These traits make the presence of O. rusticus in LOW particularly relevant. The lake is directly connected to the Nelson River drainage and, thus, provides a potential gateway for a substantial range expansion of any AIS into northern waters. Such an invasion has already been documented for rainbow smelt [Osmerus mordax (Mitchill, 1814)]; Remnant et al. 1997). However, despite the long (30-40 years) documented presence of O. rusticus and O. immunis in Lake of the Woods, relatively little is known about their current distribution, particularly their northern limits and the location of invasion fronts.
This study examines the distribution and the abundance of O. rusticus and O. immunis and that of their native congener the virile crayfish [O. virilis (Hagen, 1870)] in an approximately 38 km 2 area in the northwest of LOW (Figure 1). This location was chosen because O. rusticus had not been detected there by 2004, but was known to occur in nearby areas of LOW. The study further documents the association of the three crayfish species with important habitat variables, such as substrate type, macrophyte presence, and water depth. Finally, literature data and catches of crayfish from fish surveys over the past 34 years were used to trace the historical distribution and rate of expansion of O. rusticus and O. immunis in LOW.

Study area
LOW is located between 93°54' and 95°21' west and between 48°42' and 49°47' north ( Figure 1), covering an area of over 3,840 km 2 . Two-thirds of LOW are located within the Province of Ontario with smaller parts falling within the jurisdiction of the Province of Manitoba and the US state of Minnesota (Figure 1). For a detailed description of the geology, morphometry, and water chemistry of LOW see Mosindy (in press). The lake forms part of the Winnipeg-Nelson River drainage system which eventually empties into Hudson Bay ( Figure 1). Some headwaters within the LOW watershed are less than 100 km from Lake Superior or from the Mississippi drainage. The location of LOW and its popularity as a tourist destination make the lake vulnerable to introductions of aquatic organisms. The 2006 crayfish survey covered approximately 62 km of mainland and island shoreline within Clearwater Bay, Corkscrew Channel and Ptarmigan Bay in the northwest of LOW ( Figures  1, 2-4).

Crayfish sampling
Sampling took place from 20 June to 1 September, 2006, when LOW surface water temperatures can be expected to be at 20C or higher, and crayfish activity is not strongly affected by water temperature, reproduction, or moulting (Capelli 1975;Somers and Stechey 1986). Sampling was mainly based on a protocol Science Center (DESC) which has been shown to produces CPUE estimates with good precision and repeatability (Somers and Reid, in press). Minnow traps made with vinyl coated metal mesh (6 mm high, 12.5 mm wide diamond shaped openings) and a trap opening of approxi-mately 3.5 cm were baited with fish flavoured cat food inside a perforated plastic film canister. Traps were arranged along 15-m long traplines each containing a set of six traps. Each trapline was treated as an individual site and its starting location was recorded using GPS. The shoreline was divided into approximate 500-m long sections, in which three sites were randomly chosen to represent the existing diversity in substrate types. A minimum distance of 50 m between traplines was maintained except for some of the small islands. The first of the six traps of each trapline was anchored near shore at a minimum depth of 0.5 meters, and the remaining traps were deployed from a boat perpendicular to shore with a distance of three meters between traps. Marker buoys were attached to both terminal ends. Traplines were left overnight and retrieved 19-25 hours later. Traplines were set on 33 days at 376 sites (Figures 2-4) for a total of 2256 traps. Thirteen traps that had opened upon retrieval were not considered in the analysis. Several factors affect catch per unit effort (CPUE) of crayfish using stationary, baited traps (Somers and Stechey 1986;Rach and Bills 1987;Somers and Green 1993;Ogle and Kret 2008) and CPUE cannot always be directly related to crayfish density, particularly at high densities and in densely vegetated, shallow water habitats (Capelli 1975;Dorn et al. 2005). However, we feel that for our study area trap catches provided at least an adequate estimate of relative crayfish abundance, and CPUE and abundance are used synonymously in the text.
Catches from each trap were identified to species, counted, and expressed as a CPUE per trap or trapline per night. Crayfish were then released at the site of capture. Water depth was recorded for the first and last traps at all sites; and for the middle depth of the trapline at most sites. The depth of traps 2-5 was estimated using linear interpolation between first trap, mid (if available), and last trap depths. For each trapline, the set time and lift time was recorded and the surface (0.5 m depth) water temperatures (C) was measured with a digital thermometer. For the section of the trapline where the bottom substrate was visible (usually to water depth of less than 2 m), its composition was assessed according to the presence or absence of the following types: bedrock, boulders (>25 cm), rubble and cobble (8-25 cm), gravel (0.3-<8 cm), sand (<0.3 cm), silt (inorganic material finer than sand with noticeable structure), clay (inorganic material without noticeable structure), detritus (decayed organic material), muck (soft organic material), and man made (mainly concrete pieces). The presence or absence of submergent, emergent, and floating macrophytes was also noted.
Historical crayfish distribution records were obtained from the literature and, for the Ontario waters of LOW, from catches of crayfish in gill nets and seines used by the LOW Fisheries Assessment Unit (LWFAU) for the years 1973, 1985, 1987, 1997, 1998, and 2001-2006. The LWFAU covers all areas of the Ontario portion of the lake during a six-year rotation.

Data analysis
Differences in mean CPUE among species and between sites east and west of the southern invasion front of O. rusticus (see results) were ascertained by one way analysis of variance (ANOVA). Because data sets had unequal variances and/or their distribution could not be normalized by transformation, Kruskal-Wallis ANOVA on ranks was used, applying Dunn's method for multiple comparisons.
To determine the effect of substrate type and macrophyte presence/absence on crayfish CPUE only traps from depths ≤2 m were used, because the above habitat descriptors could only be meaningfully assessed to that depth. Furthermore, for O. rusticus all sites to the west of a southern invasion front of this species were excluded from all habitat related analyses because abundances decreased rapidly to almost zero past the front. For the calculation of mean crayfish CPUE by substrate, the type classifications (e.g., silt, clay) recorded in the field were used to represent the 13 most common substrate groups (e.g., "inorganic fines") that were represented by at least five sites and 12 traps. To evaluate how trapline locations and crayfish species separated on gradients of key environmental variables in a multivariate context, substrate, macrophyte, water depth, and species distribution data were subjected to a redundancy analysis (RDA), using SYNTAX 2000 software (Podani 2001). For this, substrates types were reduced to six categories: bedrock, large rocks (>8 cm diameter), small rocks (gravel and sand), inorganic fines, organic fines including detritus, and manmade (concrete pieces). Macrophytes were scored as either present or absent.
To determine differences in CPUEs for each crayfish species among substrate types, macrophyte presence-absence, and water depth, a repeated measures negative binomial regression model (in the following referred to as "the regression model") fit via generalized estimating equations (Molenberghs and Verbeke 2005) was used. CPUEs were modeled as a function of substrate group (n=13) or macrophyte presenceabsence as categorical predictors, trap depth as a continuous predictor, and their interaction. If the interaction was not statistically significant (P>0.05), the model was simplified to a maineffects only model and the simplified model rerun. Sites acted as the experimental unit for the substrate effect, while individual traps acted as the experimental unit for the depth effect and interaction between trap depth and substrate code. An autoregressive order-1 covariance structure was assumed for the dependence in the observations obtained from the six traps within a site, with data from traps closer to each other assumed to be more correlated than traps at more disparate depths. For O. rusticus and O. immunis model convergence problems arose from substrate categories and trap depth combinations with no individuals observed. Therefore sites with inorganic fines only or with bedrock and small rocks were removed from the analyses. Bonferroni error-controlled pairwise comparisons of least square means were used to establish significant differences between CPUEs for particular substrate groups. All means presented in the text and figures are given with their standard errors (SE).

Results
Water temperature ranged between 18.2 and 26.6C over the study period with a low daily mean temperature of 18.8C on 22 June and a maximum mean daily temperature of 25.2C on 11 July. Thereafter, water temperatures decreased gradually to a daily mean of 19.6C on 24 August, and increased slightly to approximately 20.5C on the last two days of the study.

Crayfish distribution, abundance, and relative species composition
A total of 9833 crayfish were captured, for a mean CPUE of 4.38 per trap ( Table 1). The maximum number of crayfish captured in one trap was 29, and 303 of the 2243 working traps did not catch any crayfish. No crayfish were caught from two of the 376 sites and the maximum CPUE per trapline was 85. Three crayfish species were identified from the study area: the native virile crayfish, captured with 6718 individuals, and the two AIS species, the rusty crayfish (n=2253) and the papershell crayfish (n=861). The mean CPUE for all crayfish from sites west of the southern invasion front of O. rusticus (see below) was significantly higher than the corresponding value for sites east of the front (Table 1). Orconectes virilis was the most widespread and abundant crayfish in the study area. This species was caught at all but 14 of the 376 sites with a CPUE of up to 29 per trap (Table 1) and 72 per trapline ( Figure 2). The average CPUE of virile crayfish per trap was 3.00, significantly (P<0.001) higher than the CPUE of either of the other two species (  Overall, O. virilis contributed more than 50% to the total crayfish CPUE per trapline at 260 sites. Notable exceptions to this pattern were the aforementioned island areas, including Island 23, and both island and mainland sites along the southernmost 2-3 km of Corkscrew Channel ( Figure 2).
In contrast to O. virilis, the distribution and abundance of O. rusticus was more patchy and generally centered in the northern half of the study area ( Figure 3). The mean CPUE per trap was 1.00, and the maximum number of rusty crayfish caught was 16 for a single trap and 51 for a trapline. O. rusticus had the highest relative CPUE at 78 sites and at 47 of these sites this species contributed more than 80% to the crayfish catch. Seventy-three of the 78 sites were located in the northern half of the study area, mainly near the northeast tip of Corkscrew Island, including Islands 22 and 23, and the southern end of Corkscrew Channel, including Island 6. The rusty crayfish was not found at 135 sites. These sites fell into two major groups: several clusters of adjacent sites on the island and mainland side of Corkscrew Channel, including five consecutive sites at the northwestern edge of the study area on Corkscrew Island (Figure 3). The second group of sites were all located in the southwest part of Ptarmigan Bay and adjacent sites in Ash Bay. In fact, no individuals of O. rusticus were captured in Ash Bay and Ptarmigan Bay southwest of Copper Island, including Islands 7 and 13 which are located south and east of the eastern tip of Copper Island ( Figure 3). Based on the distribution of O. rusticus catches of less than three individuals per trapline in this area, a line extending from Island 13 to the mainland between sites 189 and 190 ( Figure 3) is referred to as the "southern invasion front". Sites located east and north of this invasion front had a mean CPUE of 1.28 per trap, whereas only 0.03 O. rusticus were captured on average in traps west of the front (Table 1).
Orconectes immunis was found with relatively low abundance at 168 sites throughout the study area ( Figure 4). The mean CPUE per trap was 0.38 and the maximum number of papershell crayfish caught in a single trap or by an entire trapline was 10 and 32 individuals, respectively. The distribution of O. immunis was patchy and its abundance was generally low in the northern part of the study area where only two out of 207 sites yielded more than 10 individuals ( Figure 4). Despite its overall low abundance, the papershell crayfish had the highest relative CPUE at five northern sites and at 12 of the 169 sites in the southern part of the study area. The mean CPUE of O. immunis for sites west of the invasion front of O. rusticus was significantly higher than the CPUE for sites east of the front (Table 1). In contrast to the other two species, O. immunis was found on only a few of the small islands in the study area. This species was not captured from any of the seven islands sampled within Corkscrew Channel and from six of the 13 small islands within Ptarmigan Bay southeast of Copper Island (in the following referred to as 13-Island Bay; Figure 4). Furthermore, the mean CPUE for all traps set on any of the seven islands within 13-Island Bay where O. immunis was captured, was, with one exception significantly lower than the mean CPUE from adjacent mainland sites. Within 13-Island Bay, the maximum depth of the surrounding water of islands without papershell crayfish was at least 9.1 m and mostly 12.2 m. In contrast, six of the seven islands with O. immunis were separated from the mainland or an adjacent island by water less than 3.1 m deep; only one island was surrounded by water of up to 9.1 m deep. The three small and the one large islands at the western edge of the study area where O. immunis was captured at similar CPUEs as at the adjacent mainland sites on Sherlock Point were surrounded by water with maximum depths of between 3.0 and 9.1 m The maximum depth of water which separated the small islands without O. immunis captures within Corkscrew Channel from the mainland or Corkscrew Island ranged between 3.0 and 6.1 m.
Overall, the mean trap CPUE of O. immunis and O. virilis from island sites (regardless whether Corkscrew Island was included or excluded) was significantly lower than the mean CPUE from mainland sites ( Figure 5). In contrast, the mean trap CPUE of O. rusticus for island sites was almost identical to that for A significant negative correlation existed between the CPUE values per trap among the three crayfish species and best followed an exponential function for all three species pairs. However, whereas the regression equations explained 38% and 46% of the variation in CPUE between O. rusticus and each of its two congeners, respectively, only 15% of this variability could be explained for the species pair of O. virilis and O. immunis.

Crayfish habitat associations
RDA produced an ordination that was significantly different from random ( Figure 6). Axis 1, representing the gradient between large rock/bedrock and macrophyte presence explained 73.5% of the species-environment relationship, and axis 2, representing the gradient between manmade (concrete) substrates and small rocks and water depth explained 21.6% of the relationship. However, the correlations between  crayfish species and environmental parameters were generally weak. The rusty crayfish was more closely associated with large rock and bedrock substrates, with a negative correlation with inorganic and organic fines (including detritus), and the presence of macrophytes. The associations of O. immunis with environmental parameters were largely opposite to those of O. rusticus. O. virilis showed an association with small rocks (including sand) and deeper water, and its CPUE was negatively correlated with manmade substrates.
The associations between the distribution of crayfish species and environmental parameters were further quantified by examining mean species-specific CPUEs for each of the three main parameters measured: water depth, substrate composition, and macrophyte presence/ absence. The repeated measures regression models showed that the interaction between water depth and substrate group or macrophyte presence/absence were not significant for all three species, and the main-effects only models were used in the following analyses

Water depth
Water depth significantly affected the abundance of all three crayfish species. Mean trap CPUE of O. virilis increased to 6.0 m then fluctuated and subsequently declined sharply for depths 6.0 m to 10.0 m. No virile crayfish were captured in the two traps set at 10.2 and 11.0 m (Figure 7). The regression model indicated the effect of water depth on O. virilis CPUE was significant (P<0.0001) and there was a 16% increase in CPUE for each 1-m increase in water depth.
The mean CPUE of O. rusticus fluctuated for depths up to 8.0 m and no individuals were captured beyond 8.5 m depth (Figure 7). The regression model indicated the effect of water depth on O. rusticus was significant (P=0.008) and there was a 9% decrease in CPUE for each 1-m increase in water depth.
The mean CPUE of O. immunis decreased rapidly from depths of less than 1.5 m to 3.5 m and then fluctuated to depths of up to 6.0 m. Only a single individual was captured at 8.8 m (Figure 7). The regression model indicated the effect of water depth on O. immunis was significant (P=0.0002) and there was a 27% decrease in CPUE for each 1-m increase in water depth.

Substrate composition and macrophyte presence
Mean CPUE of all three crayfish species differed among substrate groups (Figure 8), but the overall effect of substrate composition was only significant for O. virilis (P=0.04) and O. rusticus (P=0.01), but not for O. immunis (P=0.08). Highest CPUEs of O. virilis were found on rocky (gravel to bedrock) substrates that also contained organic or inorganic fines, on gravel and sand, and on a mix of large and small rocks. Lowest CPUEs were associated with concrete pieces, a mix of bedrock and small rocks, or substrates that consisted entirely of organic and inorganic fines ( Figure 8). Except for bedrock and small rocks, the difference in CPUE for each substrate between the above "low" and "high" groups was significant.
Orconectes rusticus was most abundant on substrates that were made up of large rocks (also in mixture with other rocks) and that could also contain inorganic fines (Figure 8). This species was not found on a combination of bedrock and small rocks. The rusty crayfish occurred in very low abundance on substrates that entirely consisted of organic fines or a mix of organic and inorganic fines (Figure 8). The CPUE of the latter group was significantly lower than those from the five substrates with the highest CPUE.
In partial contrast to the substrate association of O. rusticus, mean CPUE of O. immunis were highest on substrates containing organic and inorganic fines, followed by mixed rocks and organic fines (Figure 8) Figure  9). The type of macrophyte (floating, submergent, emergent) did not significantly affect crayfish CPUE.

Historic and current distribution of crayfish in LOW
Since its first record from Long Bay on the east side of LOW in 1963 (Crocker and Barr 1968), O. rusticus has spread into most parts of the lake by 2006 with the possible exception of the southwest (Figure 10). The earliest records (1973) of the LWFAU show a concentration of captures within Long, Lobstick, and Regina bays. They also indicate that rusty crayfish had moved at least 40 km into Whitefish Bay to the southwest and 16 km into Yellow Girl Bay to the northwest within 10 years after its first capture. The next records from 1985 show a further northwest expansion of O. rusticus to two island sites approximately 8 km (Shore Island) and 12 km (Ferrier Island) north and west of Yellow Girl Bay (Figure 10, Annex 1). Another 12 years later, based on the extensive records for 1997/1998, rusty crayfish were found at several locations throughout the large body of water north of the Aulneau Peninsula, except the northernmost bay near Kenora and the area west of the eastern entrance into Ptarmigan Bay that included the 2006 study area. The distance between the most western and northern of the sites where O. rusticus was found in 1997 and the north-westernmost location in 1985 is 27.7 and 16.4 linear km, respectively. In most cases the LWFAU records only indicate species presence. However at one of the 1997 sites presently has no open-water connection to Whitefish Bay, Sabaskong Bay was directly connected to Whitefish Bay by a man-made navigation channel at Turtle Portage from the 1950s until 1992 when it was filled in. Thus, the recent records of rusty crayfish from Sabaskong Bay should not be interpreted as a secondary introduction by humans, but likely were the result of a natural range expansion from locations just north in Whitefish Bay prior to the closure of the navigation channel. Until the intensive 2006 study, captures of O. immunis in LOW have been very rare and were restricted to an area in the eastern part of the lake. The first published record (Dubé and Renaud 1994) does not give an exact capture date (sometime "since 1975") and location ("Snake Bay" without reference to LOW), and refers to a personal communication (F. W. Schueler) as the source of this information. Others (e.g., Hamr 1998) have subsequently referred to Snake Bay as Snake Bay in LOW. The LWFAU has captured O. immunis on only four separate occasions, at three sites in 2006 and at one of these sites also in 2005 ( Figure 10). All three sites are located within Whitefish Bay, within approximately 5-20 km distance of the first published record for the papershell crayfish from LOW. The linear distance between the northernmost site in Whitefish Bay and the eastern limit of the 2006 study area is approximately 50 km.

Discussion
The abundance of crayfish in the Ptarmigan Bay area of LOW differed substantially among the three species and showed several distinct spatial patterns. With a mean CPUE of 3.0 individuals per trap, the native O. virilis was the most abundant species overall but throughout the study area clusters of sites existed where mainly O. rusticus, and also O. immunis occurred at the highest relative abundance. The CPUE of O. virilis in LOW is higher than reported from other studies. Somers and Stechey (1986) caught 0.1-0.4 O. virilis per trap, however, water temperatures in their study ranged from only 6-11C. These cold temperatures likely contributed to the low capture rates, as Somers and Green (1993) found that highest catches of O. virilis coincide with maximum temperatures of 18-23C in July and August. These temperatures are similar to the daily means of 19-25C recorded in the Ptarmigan Bay area in 2006 and fall within the range of 17-23C for which trap catches of orconectid crayfish seem to be independent of water temperature (Capelli 1975). Despite adequate temperatures, O. virilis CPUE was only 0.4-1.1 for four southern Ontario lakes (Somers and Green 1993), and 0.7-2.2 for an 18-year record from Blue Chalk Lake, Ontario (Somers and Reid, in press), values considerably lower than for the Ptarmigan Bay area. Several factors other than temperature may be responsible for differences in crayfish CPUE between lakes, such as predation pressure (Stein and Magnuson 1976;Somers and Green 1993), or perhaps LOW provides a relatively productive habitat for O. virilis and crayfish in general.
Some caution must be applied to the above and all following interpretations of trap catches with regard to species abundance. Several factors affect CPUE of crayfish with stationary, baited traps, including the size and species identity of crayfish already present in a trap, water temperature, lunar phase, bait type, and the presence of other crayfish species and of fish predators near the traps (Somers and Stechey 1986;Rach and Bills 1987;Somers and Green 1993;Ogle and Kret 2008). Furthermore, traps primarily catch adult male crayfish (Somers and Stechey 1986;Rach and Bills 1989;Momot 1996) and may particularly select for males of O. rusticus while underestimating the abundances of O. virilis and O. propinquus (Olsen et al. 1991). However, this may not always be the case, as Capelli and Magnuson (1983) have shown that trap catches of crayfish that included all species found in LOW can provide a reliable estimate of relative species abundance. If there was a positive bias for O. rusticus in the 2006 catches, then the abundance of rusty crayfish within the study area was overestimated. However, it can be assumed that if such a bias existed, it acted similarly at all sites within the study area. Thus, the observed spatial pattern in crayfish CPUE should accurately reflect associated differences in relative abundance and species composition.

Crayfish habitat associations
Although crayfish were found on all of the 12 substrate groups considered in this study, their abundance differed significantly between some substrates and these differences were not uniform among species. O. virilis and O. rusticus were mainly found on rocky substrates, with the rusty crayfish being relatively more abundant on the larger size fractions, including pieces of concrete. An affinity of the rusty crayfish for broken concrete has previously been documented by Taylor and Redmer (1996), suggesting that shorelines affected by human construction can provide suitable O. rusticus habitat. The results from LOW are similar to several other field studies, indicating that O. rusticus tolerates a wide range of bottom sediments, but mainly occurs on firm substrates that provide some kind of shelter (Lorman 1980;Hill and Lodge 1994;Kershner and Lodge 1995;Garvey et al. 2003). However, there are exceptions to this general pattern. For example, Smiley and Dibbel (2000) reported highest abundance of all size-classes and both sexes of O. rusticus on fine grained soft bottom substrate (i.e., detritus or sand), and Wilson et al. (2004) found that total crayfish abundance after the invasion of the rusty crayfish increases mostly on sandy sediments.
In LOW, O. immunis preferred sediments consisting partially or entirely of organic or inorganic fines and occurred mainly within or near macrophyte beds. In contrast, the abundance of O. virilis was similar at sites with or without macrophytes, and O. rusticus abundance was lowest at locations where macrophytes were present. The habitat associations of the papershell crayfish in LOW fit well with the account by Crocker and Barr (1968) Capelli and Magnuson (1983) documented substantial overlap in orconectid species distribution among different types of waterbodies Similar to the results for LOW, Magnuson et al. (1975) found that the abundances of submerged macrophytes and O. rusticus were inversely related. However, Garvey et al. (2003) reported that O. rusticus commonly occurred in macrophytes. The reduction of macrophyte species richness and abundance by grazing and non-consumptive destruction is one of the primary impacts of O. rusticus invasions (Lodge and Lorman 1987;Olsen et al. 1991). Thus, the low abundance of the rusty crayfish near macrophytes found in the present study may reflect ongoing reductions of macrophytes by O. rusticus consumption, and not habitat avoidance. Our findings for O. immunis are consistent with other studies showing that the papershell crayfish may feed on submersed macrophytes (Seroll and Coler 1975;Letson and Makarewittz 1994), but does little nonconsumptive destruction and only reduces macrophyte standing crop at high crayfish biomass of >140 g/m 2 (Letson and Makarewittz 1994).
Some of the inconsistent results regarding the substrate preference of orconectid crayfish may be due to factors other than preference. Both O. virilis and O. rusticus have been observed to modify their habitat use in the presence of congeners (Hill and Lodge 1994;Garvey et al. 2003;Roth and Kitchell 2005), indicating a plastic behavioural response to reduce the frequency of aggressive encounters and to avoid competition. For example, in the absence of O. rusticus, O. virilis has been shown to occupy both rocky and soft substrates (Capelli 1975;Hobbs andJass 1988, cited in Roth andKitchell 2005), and Garvey et al. (2003) suggested that the virile crayfish was restricted to macrophyte habitat due to replacement by O. rusticus on cobble. Furthermore, O. rusticus may increase the use of cobble substrate with increasing predator density (Kershner and Lodge 1995), and substrate specific abundance can differ between crayfish size classes (Lorman 1980;Smily and Dibble 2000;Garvey et al. 2003) and sexes (Smily and Dibble 2000). Parameters such as these were not quantified in the present study, but may have affected crayfish substrate use. However, there was no clear evidence for habitat segregation or habitat specific species displacement by crayfish in LOW, as two or all three species were frequently captured by the same trapline or even in the same trap.

Invasion fronts and expansion rates of alien species
There was at least one spatial gradient in the abundance of the AIS O. rusticus in the Ptarmigan Bay area that is highly indicative of an invasion front. Rusty crayfish CPUE dropped markedly to values of considerably less than 1 individual per trap in southern Ptarmigan Bay, when moving towards the eastern shoreline of the peninsula that separates Ash Bay, and no rusty crayfish were caught anywhere on or past the northern and western shore of the peninsula. Wilson et al. (2004) defined an invasion front as <1 individual per trap day in a study of dispersal patterns of O. rusticus in Trout Lake, Wisconsin. A potentially second, northern invasion front of the rusty crayfish existed on the north shore of Corkscrew Island. There, O. rusticus CPUE declined continuously from 6.2 per trap at a site on the eastern tip of the island to 0 over the next 10 sites in the westerly direction, and no rusty crayfish were caught at five further sites which Orconectes rusticus was found with relatively high (up to 4.7) CPUE values per trap on the only island (number 22) sampled in Clearwater Bay in 2006, just across a <7-m deep channel from the northern invasion front. For most sites on the north shore of this island, rusty crayfish contributed more than 75% to the total crayfish catch. Furthermore, in 2007, rusty crayfish were captured with mean CPUE values of 1.3 and 0.5 on the very small and slightly larger island, respectively, in Clearwater Bay west of Island 22 and just north of the northern invasion front on Corkscrew Island (Tom Mosindy, unpublished data). Unless there was a secondary introduction by humans, the islands in Clearwater Bay must have been colonized very recently and presumably later than the north shore of Corkscrew Island. The substantially higher CPUE values compared to the adjacent invasion front on Corkscrew Island and the rapid establishment of numerical dominance by O. rusticus over the other two crayfish species on Island 22 suggests that the factors that modulate the rate of population expansion and speed of colonization of O. rusticus may differ between the islands in Clearwater Bay and the northern invasion front, or that the island's habitat is generally more favourable for rusty crayfish. However, none of the habitat features documented in this study differed substantially between Island 22 and the north shore of Corkscrew Island, and other, undocumented habitat features may favour O. rusticus over O. virilis and O. immunis on Island 22. Also possible, biotic factors such as different predation pressures between sites at the Clearwater Bay islands and the northern invasion front may account for the above differences in relative abundance of the three crayfish species. Fish predation can substantially reduce crayfish abundance (Taub 1972;Mather and Stein 1993 Garvey and Stein 1993;Roth and Kitchell 2005).
Based on the entire 2006 study area, the overall abundance of O. rusticus was relatively low, indicating that this population has arrived relatively recently in the Ptarmigan Bay area and is not yet well established. The mean CPUE per trap for the rusty crayfish was 1.3 when all sites west of the southern invasion front were excluded. This value is well below the 4.2-18.0 individuals per trap and the >15 individuals per trap for established populations of O. rusticus in small lakes near Thunder Bay, Ontario (Momot 1996) and in Trout Lake, Wisconsin (Wilson et al. 2004), respectively. Conversely, the CPUE of O. rusticus in LOW in 2006 is considerably higher than the 0.05 individuals for a newly developing population of the rusty crayfish in Pigeon Bay, Lake Superior (Momot 1996). However, because crayfish CPUE is affected by many biotic and abiotic variables (Somers and Stechey 1986;Somers and Green 1993), it is difficult to directly compare O. rusticus abundance between lakes without some knowledge of the habitat types sampled. Furthermore, the CPUE of O. rusticus at newly invaded locations has been shown to increase slowly at first, but often rapidly at sites behind the invasion front (Wilson et al. 2004), making the timing of between lake or between area comparisons in CPUE critical.
Several locations existed within the southern end of Corkscrew Channel where one or more traps per site caught nine or more rusty crayfish. A CPUE of >9 may represent a threshold abundance above which O. rusticus reduces numbers of competing fish species, the abundance of Odonata, Trichoptera, Amphipoda, and Gastropoda, and the species richness and biomass of submerged macrophytes (Wilson et al. 2004). These authors further concluded that high abundance is the main reason for the large impacts caused by O. rusticus. The maximum trap CPUE for rusty crayfish in LOW was 16, well below the more than 60 reported for Trout Lake after O. rusticus had achieved total dominance within the crayfish community, and its catches were 2-18 times greater than historical catches of native crayfish (Wilson et al. 2004). No published quantitative data on crayfish abundance exist for LOW and the impacts of AIS crayfish species have not been assessed. However, if O. rusticus did not arrive in Ptarmigan Bay prior to 2003, then its abundance in at least some parts of this bay in 2006 reached levels that potentially result in ecological impacts only three years after its first detection. In the relatively small (16 km 2 ) Trout Lake it took O. rusticus approximately 20 years from first detection to totally dominate O. virilis and O. propinquus (Wilson et al. 2004). Perhaps the rate of O. rusticus population growth and the speed of its expansion are relatively high in LOW. With 2.1 km/year the average rate of net movement of O. rusticus in LOW is three times the rate of 0.68 km/year calculated for Trout Lake (Wilson et al. 2004) and more than five times as high as the rate of 0.40 km/year for a small lake in northwest Ontario (Momot 1996). Only the 1.5 km/year reported by Momot (1996) for the first year of expansion of O. rusticus into Lake Superior comes close to the rate observed in LOW. Perhaps the very complex, generally rocky shoreline and the many islands that are separated from the mainland by only shallow channels favour a rapid expansion of O. rusticus (see below). This argument is indirectly supported by Momot (1996) who found that the access to sheltered bays was critical for the speed of expansion of O. rusticus in Lake Superior.
Daily (24 hour) travel distances of O. rusticus have been recorded as 28 m on average and as high as 110 m (Byron and Wilson 2001). A similar maximum distance of daily movement (124 m) has been observed for O. virilis (Momot and Gowing 1972). Based on these data, and conservatively assuming that O. rusticus is only active between June and August (Lorman 1980), Wilson et al. (2004) have calculated the maximum potential invasion rate of rusty crayfish at 2.7 km/year. This rate of expansion is between 29% (LOW) and almost 400% (Trout Lake) higher than actual field estimates. Several factors exist that might explain the discrepancy between observed and potential invasion rates of O. rusticus. These have been discussed in detail in Byron and Wilson (2001) and Wilson et al. (2004) and include crayfish population density, predation pressure by fishes, habitat availability and connectivity, and interactions with congeners. Very little information about these processes is available for the LOW and their relative importance in determining invasion rate and success of AIS crayfish species in this lake is largely unknown.
Considering the entire 2006 study area, O. virilis was the dominant crayfish species based on trap CPUE. However, the strong negative association between the CPUE of O. rusticus and that of O. virilis or O. immunis indicates that the rusty crayfish has started to replace his two congeners in the Ptarmigan Bay area east of the southern invasion front. A similar negative relationship in the trap catches of these three orconectid species has been documented by Wilson et al. (2004). Furthermore, species replacements with the primary loss of O. virilis have occurred in many other lakes in Illinois (Taylor and Redmer 1996), Wisconsin (Capelli 1982;Lodge et al. 1986;Olsen et al. 1991;Wilson et al. 2004), and northern (Momot 1996) and southern (Berill 1978) Ontario, as O. rusticus has expanded its distribution northwards and eastwards. These replacements by O. rusticus of congeners, including O. virilis and O. immunis, are thought to be mediated by inappropriate mate selection and hybridization Butler IV and Stein 1985; but see Perry et al. 2001 for lack of evidence for O. rusticus-O. virilis hybridization), aggressive dominance (Capelli and Munjal 1982;Butler IV and Stein 1985), higher memory capabilities (Hazlett et al. 2002), competitive exclusion (Hill and Lodge 1994;Garvey et al. 2003), and lower rates of fish predation (Butler IV and Stein 1985;DiDonato and Lodge 1993;Mather and Stein 1993;Roth and Kitchell 2005).
Assuming that O. rusticus arrived in Ptarmigan Bay in 2003 (see above), the CPUE data for 2006 suggest that the displacement of O. virilis progressed quite rapidly in this area of LOW. The rate of displacement by the rusty crayfish varies among lakes (Olsen et al. 1991;Capelli 1982), but can be rapid. For example, Capelli (1982) found that O. rusticus completely replaced O. virilis and O. propinquus just 5 years after accounting for only 8% of total crayfish abundance in a relatively small (<3.4 km 2 ) Wisconsin lake. A complete replacement of O. virilis by O. rusticus appears not to occur in LOW or progresses much more slowly than in other lakes. Although no quantitative data exist on long-term crayfish species abundance in LOW, the catches by the LWFAU indicate that O. virilis is still present in sometimes large numbers at many locations more than 30 years after these sites were first invaded by O. rusticus.
The widespread presence of O. immunis that was documented in Ptarmigan Bay area in 2006 is remarkable. Until that year, this species was only known from the east side of LOW where it had been captured by the LWFAU in 2005 and 2006 within 10 km of the location of its first record from sometime after 1975 (Schueler, pers. comm. in Dubé and Renaud 1994). The most western of these locations in Whitefish Bay is at least 46 km distant from the eastern edge of Ptarmigan Bay. Either O. immunis occurred at very low abundance in other areas of LOW during its expansion from Snake Bay to the Ptarmigan Bay area and was not detected during more than 20 years of sampling by the LWFAU, or the papershell crayfish expanded along a path not covered by the sampling locations of the LWFAU. Both these scenarios are unlikely considering that this species is susceptible to gillnet and seine catches, the many catches of O. rusticus by the LWFAU in the areas of the likely expansion path of O. immunis, and the extensive distribution and often relatively high abundance of the species within the 2006 study area. The papershell crayfish was present at most sites very close or at the limits of the 2006 study area and there were no obvious strong spatial gradients in the CPUE of O. immunis. This pattern was confirmed by regular but lowabundance catches of the papershell crayfish on the northwestern tip of Corkscrew Island, the two small islands north of this area in Clearwater Bay, and the southwestern tip of Corkscrew Island in 2007 (Tom Mosindy, unpubl. data). Compared to the northern part of the 2006 study area, the distribution of O. immunis among mainland sites in the southern part was almost continuous and CPUE values were generally higher. These geographic differences in abundance of the papershell crayfish seemed to be less indicative of a temporal pattern in dispersal or the availability of specific habitats (see below), but were likely related to the distribution of its congeners, particularly O. rusticus. Many of the sites with relatively high abundance of O. immunis were located west of the southern invasion front of O. rusticus. Overall, these results do not indicate the existence of an invasion front for O. immunis. Furthermore, unless this species disperses very rapidly, the distribution data suggest a longer history of residence in the study area for O. immunis than for O. rusticus. One alternative explanation that is consistent with both the general distribution of O. immunis within LOW and its distribution and abundance within the 2006 study area is that the papershell crayfish has been recently (i.e., 1998-2002) introduced into the Ptarmigan Bay area and currently has a disjunct distribution within LOW.

Islands and the distribution of invasive crayfish
The papershell crayfish was absent from most small islands in the 2006 study area or occurred at very low CPUE. This was particularly obvious in 13-Island Bay where this species was not found on islands surrounded by water deeper than 9.1 m. Catches throughout the entire study area show that O. immunis generally does not enter traps set at >6 m depth, and support the hypothesis that the absence of this species from many island is at least partially due to an avoidance of deeper water. Although CPUE also declined with increasing water depth for O. rusticus, this decline was much more gradual and rusty crayfish were caught at depths of up to 8.4 m in the Ptarmigan Bay area. Furthermore, O. rusticus occurred on all islands but two, and was the only species with a higher overall mean CPUE on islands than on the mainland. The two islands where rusty crayfish were not captured lay within the southern invasion front of this species and were surrounded by water of at least 9.1 and 12.1 m depth, respectively. Thus, water depth and invasion dynamics likely combined to delay the expansion of O. rusticus onto these two islands. Wilson et al. (2004) found that rusty crayfish invaded islands two years later than corresponding sections of mainland. These authors also found that O. rusticus will travel to depths of >12m, and Taylor and Redmer (1996) reported SCUBA observations of rusty crayfish at 14.6 m depth in Lake Michigan, indicating that this species may be able to reach islands separated by relatively deep channels on its own. Because crayfish locomotory activity is strongly dependent on temperature (Van den Brink et al. 1988), it is negatively affected at temperatures below 17C (Capelli 1975) and substantially reduced at 10C (Momot 1996), the lower depths frequented by O. rusticus likely coincide with the depth of the metalimnion in larger, dimictic lakes such as LOW. This hypothesis is supported by data collected from 31 lakes in Ontario, Québec, and New York State showing that the maximum depth where crayfish were collected depended on the slope of the bottom, with steeper slopes having crayfish (no species identity provided) as far as the thermocline (Lamontagne and Rasmussen 1993). Three temperature profiles taken within 4-6 km of the study area on 21 August, 2006 indicate metalimnion depths of 7-9 m, 8-10 m, and 10-12 m, respectively (Tom Mosindy, unpubl. data). Thus, it is unlikely that rusty crayfish will move into water of >12 m depth in most areas of LOW, and the littoral zones of isolated islands that are separated by deep (>12 m) channels may not become O. rusticus habitat for some time after the species has expanded its distribution into the general area of such islands. Conversely, islands not separated by deep channels from the mainland may serve as stepping stones for reaching mainland areas, for example, across bays faster than via the potentially much longer route along the mainland shoreline, and thus will promote the expansion of the rusty crayfish. Hill and Lodge (1999) suggested that O. rusticus will replace O. virilis where they come into contact, that no lake or fisheries management strategy will be able to reverse a replacement of the virile crayfish by the rusty crayfish, and that the only way to protect the native species is to prevent the spread of the invasive species. Given the current distribution and abundance of O. rusticus in LOW, a lake that is ideally located to act as a potential epicentre for human mediated transfers of AIS into Lake Winnipeg and the Nelson River watersheds, it seems only a matter of time until the rusty crayfish will expand further north and west. The documented presence of O. rusticus in the Winnipeg River in 2006 where it was found as far as 45 km north of the outflow from LOW and well past the first hydroelectric generation station (Watkinson and Batty, in press), indicates that such an expansion is already in progress.

Future expansion of invasive crayfish
It will be insightful to see whether a large river environment with multiple dams will slow the downstream progression of the rusty crayfish. This could be expected based on observations by Momot (1996) who found that, at least in smaller streams, beaver dams and weirs greatly impeded the movement of O. rusticus. Conversely, some sections of the Winnipeg River are heavily used by humans and several fishing camps represent potential hotspots for introductions of invasive crayfish. Capelli and Magnuson (1983) have shown that the presence of O. rusticus in lakes is, among other factors, positively correlated with human activities. More recently, Puth and Allen (2004) demonstrated that humans are vectors enabling the rusty crayfish to move along discontinuous routes. Considering that a single egg-carrying female may be sufficient to establish a new population, the rusty crayfish will likely move past dams on the Winnipeg River with the help of humans. Lotic habitats may also further increase the competitive advantage of O. rusticus over O. virilis and O. immunis. Maude and Williams (1983) collected rusty crayfish immediately downstream of large dams where current speeds often exceeded 1 m s -1 and showed in laboratory experiments that O. rusticus has superior station-holding abilities in currents compared to its two congeners. It is also possible that O. rusticus will circumvent some of the dams on the main stem of the Winnipeg River by reaching the section downstream of the confluence with the Whiteshell River via a different route. In July 2007 it was discovered that the rusty crayfish is well established in Falcon Lake (Doug Leroux and Martin Erikson, Manitoba Water Stewardship, Fisheries Branch, pers. comm., 28 August, 2007), which is located in Manitoba approximately 20 km west of Shoal Lake adjacent to LOW (Figure 10), to which it is connected via the Falcon Creek. The apparent replacement of the native species O. virilis in the western half of the lake, and a population structure that includes a large range of size classes (Doug Leroux, Manitoba Water Stewardship, Fisheries Branch, pers. comm., 28 August, 2007), suggests that O. rusticus has likely been a resident of Falcon Lake for several years. The lake is heavily used by cottagers and campers, and thus provides a hotspot for further expansion of the rusty crayfish into water bodies of the Winnipeg River drainage, including West Hawk Lake and the Whiteshell River. Irrespective of the exact route, it is conceivable that the rusty crayfish will relatively soon reach Lake Winnipeg, where this species could present an additional stressor to an ecosystem that is currently undergoing eutrophication (Lake Winnipeg Stewardship Board 2006) and that has recently been invaded by other AIS (see Franzin et al. 2003 for fish species).
The speed of expansion of O. immunis is less well known. However, its unexpectedly widespread distribution in the 2006 study area in LOW and its occurrence in the Winnipeg River in 2006 (Watkinson and Batty, in press) and 2007 (Doug Watkinson, DFO, Freshwater Institute, Winnipeg, unpubl. data) at several locations within Eaglenest Lake more than 100 km downstream of LOW suggests that the papershell crayfish is equally capable of rapid dispersal or is transported by humans at similar frequency and over comparable distances as the rusty crayfish. If this hypothesis is correct, O. immunis will also invade Lake Winnipeg in the near future, if it has not already reached the lake via the Red River system. This latter