Is the home range concept compatible with the movements of two species of lowland river fish?
David A. Crook
The Johnstone Centre, School of Science and Technology, Charles Sturt University, Wagga Wagga, New South Wales, Australia, and the Cooperative Research Centre for Freshwater Ecology, Murray–Darling Freshwater Research Centre, Albury, New South Wales, Australia
*Present address: Arthur Rylah Institute for Environmental Research, Department of Sustainability and Environment, 123 Brown Street, Heidelberg, Victoria, Australia. E-mail: David.Crook@dse.vic.gov.au
*Present address: Arthur Rylah Institute for Environmental Research, Department of Sustainability and Environment, 123 Brown Street, Heidelberg, Victoria, Australia. E-mail: David.Crook@dse.vic.gov.au
1Many studies of the movements of riverine fish have found that most individuals are sedentary and occupy very restricted home ranges. Recently, this ‘Restricted Movement Paradigm’ has been challenged and there is currently a need for tests of the home range concept as a theoretical basis for describing the movements of riverine fish. In this paper, I describe a radio-tracking study of golden perch (Macquaria ambigua) and carp (Cyprinus carpio) in the Broken River, Australia that aims to assess the home range concept as a means of describing the movements of these species.
2A random movement analysis and a translocation experiment were conducted to test for site fidelity and home range occupation. Both golden perch and carp exhibited strong site fidelity and occupied restricted home ranges. Carp had larger total home ranges than golden perch, and both species had areas of concentrated use (core areas) within the home range.
3Several golden perch and carp exhibited shifts in the locations of their home ranges during the study. To incorporate such shifts into a theoretical framework, a ‘home range shift’ conceptual model is proposed and the need to consider the temporal stability of site fidelity when describing home range movements is discussed.
The home range concept was applied first to terrestrial mammals (Burt 1943), and has since been influential in the development of theoretical models to explain the movements and distributions of a wide variety of animals. In the 1940s and 1950s, Gerking (1950, 1953, 1959) conducted a series of mark–recapture studies that led to the application of the home range concept in riverine fish ecology. In these studies, Gerking found that the majority of recaptured fish were collected from the original capture site and that experimentally translocated fish exhibited a strong tendency to return home. A series of similar mark–recapture studies followed Gerking's work and, although certain types of specialized movement were acknowledged (e.g. spawning migrations), most studies concluded that adult stream fish are normally sedentary (see reviews by Northcote 1978; Gowan et al. 1994; Rodríguez 2002). Many of these studies also found that a small proportion of fish were ‘strays’ that moved away from the site of capture or did not return home after translocation. This finding led some workers to propose that riverine fish populations consist of a behavioural dichotomy, comprising a dominant sedentary component and a smaller mobile component (e.g. Stott 1967; Northcote 1992).
The body of research relating to the restricted movements of adult stream fish forms the central tenets of what has become known as the ‘Restricted Movement Paradigm’ (RMP) (sensuGowan et al. 1994). However, aspects of the evidence underlying the RMP have recently been called into question. Gowan et al. (1994), for example, reviewed literature relating to the movements of riverine salmonids. They argued that mark–recapture studies used to support the RMP may be biased against detecting movement, because conclusions are based mainly on recaptures from the original capture site, and the fate of non-recaptured fish is largely unknown. Several recent studies have suggested that mobile behaviour by some species of stream fish is more common and biologically important than recognized previously (e.g. Gowan & Fausch 1996; Young 1996), although Rodriguez (2002) suggested that there remains strong evidence for the main tenets of the RMP.
It is evident that the movements of fish vary considerably between taxa and life-history stages (see Lucas & Baras 2001). However, environmental variables, including temperature, flow conditions, habitat heterogeneity, channel size and location within a watershed, have also been shown to strongly influence movements (Linfield 1985; Bruylants, Vandelannoote & Verheyen 1986; Gorman 1986). The RMP, with its emphasis on the occupation of restricted home ranges, has been based largely upon studies of fish in relatively small streams (Gowan et al. 1994). The tenets of the RMP therefore do not necessarily apply to fish in larger lowland rivers, where physical restrictions to movement (e.g. shallow riffles, see Schaefer 2001) are not as common. In a study of cyprinids in several large, lowland rivers in England, Linfield (1985) proposed an alternative model to the home range concept in which patterns of movement are viewed as ‘an infinitely variable combination of active movements’ that are influenced by environmental factors such as season, water temperature and flow. More recent radiotelemetry studies of fish in larger rivers have both provided support for the home range concept (Allouche, Thévenet & Gaudin 1999; Snedden, Kelso & Rutherford 1999) and expressed doubt as to its applicability (Baade & Fredrich 1998; Clough & Beaumont 1998).
A major impediment in assessing the applicability of the home range concept for describing riverine fish movements lies in the use of the term ‘home range’. Commonly cited definitions of home range (e.g. the area over which an animal travels in its day-to-day activities; Hayne 1949) are problematic because they fail to distinguish between free-ranging animals and animals that maintain fidelity to particular areas (Munger 1984). This important distinction has rarely been addressed in studies of riverine fish movements. For an animal to possess a measurable home range, it must exhibit ‘site fidelity’ − i.e. the extent of its displacement must be smaller than if its movements were random (Spencer, Cameron & Swihart 1990). In some instances, the existence of site fidelity is evident based on qualitative criteria − a fish recaptured after a year from the same pool where it was originally collected is likely to have exhibited fidelity to a home range that included the pool. In other instances, however, the existence of site fidelity is less clear − for example, would a fish be considered to exhibit fidelity to a home range if it was recaptured 500 m from the original capture site after a year?
In view of these scaling and definitional issues, as well as recent studies demonstrating high levels of mobility by some riverine fish, there is a requirement for tests of the home range concept as a theoretical basis for describing the movements of riverine fish (Matthews 1998). In this paper, I describe a radiotelemetry study that examines the movements of two species of fish [golden perch (Macquaria ambigua Richardson) and common carp (Cyprinus carpio L.)] in a lowland river in Australia. The study aims to determine whether the movements of golden perch and carp are compatible with the home range concept, or if alternative models of movement might be more appropriate.
The movements of golden perch and common carp have been the subjects of previous studies, although there has been little detailed examination of home range occupation or homing in rivers. Studies of golden perch in the Murray and Ovens rivers in south-eastern Australia revealed high levels of mobility at times (up to 10 km per day; Reynolds 1983), and provided little direct evidence of home range occupation or site fidelity (Reynolds 1983; Koehn & Nicol 1998). In a study of carp in the Murray River, Reynolds (1983) found that most fish were relatively sedentary and possessed a homing ability. Schwartz (1987) also reported homing by carp in a lake in Pennsylvania/Ohio, USA. In contrast, Koehn & Nicol (1998) found that carp made frequent upstream and downstream movements of > 100 km per month and they concluded that carp are a highly mobile species. The specific objectives of the current study are to compare the movements of these species and to determine whether individuals exhibit site fidelity and occupy measurable home ranges. In discussing the results of the study, I propose a conceptual model to explain the observed movement patterns, and discuss the potential implications of this model in defining the movement patterns of riverine fish.
Materials and methods
Golden perch are found throughout inland south-eastern Australia and are one of the most common large native fish within the Murray–Darling Basin (MDB). The maximum size recorded for golden perch is 23 kg and 760 mm TL, although most weigh less than 5 kg (Cadwallader & Backhouse 1983). Golden perch are believed to spawn usually in spring and summer and it has been suggested that they undertake extensive upstream migrations prior to spawning (Reynolds 1983). In the Broken River (site of the current study), golden perch exhibit a strong preference for deep, slow-flowing pool habitats and are often associated with woody debris and other types of cover (Crook et al. 2001).
Common carp are native to Asia and were introduced into Australia as early as the 1860s (Shearer & Mulley 1978). Carp became widespread and abundant throughout the MDB after extensive floods during the 1960s and 1970s, and are currently considered a major pest species in Australia (Roberts & Tilzey 1997). Carp usually spawn during spring in the MDB, with both sexes reaching maturity at ∼125–150 mm TL (Brumley 1996). In the Broken River, carp do not exhibit strong preferences for either pool or run habitats, although they are often associated with relatively deep, slow-flowing habitats at smaller scales (Crook et al. 2001).
The study was conducted in a low gradient region of the Broken River in the south-eastern region of the MDB (Fig. 1). The Broken River has a mean annual discharge of 2 × 108 m3 and the channel in the study region ranges from ∼10–50 m in width and up to ∼4 m in depth. The dominant habitat types within the study region are relatively shallow, fast-flowing runs interspersed by deep, slow-flowing pools. The study was conducted between late summer (February) and midwinter (July), with water temperatures ranging from ∼8 to 25 °C (Fig. 2). Although there were several freshes, no major flood events occurred during the study period (Fig. 2). The upstream limit of the study site was a river crossing consisting of a concrete roadway that prevented fish movement during the study period. A large weir ∼35 km downstream served as the downstream limit of the study area.
fish collection and radio-tagging
Twenty golden perch [329 ± 32 mm FL, 511 ± 171 g (mean ± SD)] and 15 carp (368 ± 93 mm FL, 1136 ± 818 g) were collected from the study reach by angling and backpack electrofishing. Fifteen golden perch and all the carp were collected between 29 February and 3 March 2000, and a further five golden perch were collected on 6–7 April 2000. The collected fish were held for up to 12 h in cages placed along the river margins prior to tagging and release. Single-stage 151 MHz radiotransmitters with estimated battery lives of 180–210 days were used (Titley Electronics, Australia). These tags weighed 4·3 g in air and were attached externally to the left dorsal surface of the fish using a method similar to that described by Beaumont et al. (1996).
release and translocation of fish
To determine if fish exhibited fidelity to home ranges within the vicinity of the original capture location, 10 of the golden perch and five of the carp were released at the point of capture after tagging. In addition, a translocation experiment was conducted to determine if fish displayed homing abilities. For the translocation experiment, five individuals of each species were transported by car in opaque plastic bins and released 2250–2710 m upstream of the point of capture. Five individuals of each species were similarly translocated 1870–2720 m downstream of the point of capture. Translocated fish were deemed to have homed successfully if they returned to within 100 m of the original capture site.
A single operator tracked the fish between 4 March and 11 July 2000 by walking along the riverbank with a scanning radio receiver (Titley Electronics). Tracking was conducted daily for the first 10 days after the initial release of fish (4–13 March) and was then conducted twice per week until 31 May and at 7–19-day intervals thereafter. Normal tracking surveys were conducted during the day by scanning from the upstream barrier to a point ∼5 km downstream and then returning upstream. In addition to the normal surveys, extended surveys covering the area 10 km downstream were conducted on 20 and 25 April. To determine whether any fish had moved even further downstream, two radio receiver operators conducted a 35-km survey between the upstream and downstream barriers on 9–11 May 2000.
Quantitative tests for site fidelity were conducted using a method similar to the random walk analyses described by Spencer et al. (1990). This type of analysis models the amount of movement that would be expected if movement was random with regards to its direction. The observed amount of movement is then compared with the modelled movement patterns to determine whether the animal moved less than would be expected at random (i.e. exhibited site fidelity). Estimates of the hourly movements of eight golden perch and six carp measured during a separate radiotelemetry study (Crook 2002) were used as representative hourly movements to generate random movement paths in the current study. The use of representative movements, rather than observed movements (see Spencer et al. 1990), was required because the temporal resolution (time between observations) and the spatial resolution (measurement error) of the current study were coarse relative to the generally small distances moved by the fish. Data collected at short time intervals correspond more closely with actual movements than data collected at larger time intervals and therefore provide a more realistic measure of the potential for movement (see Kindvall 1999). The representative distances were calculated for randomly selected individuals of each species during the day (11.00–14.00 h) and at night (21.00–24.00 h). Linear distances moved in upstream or downstream directions were measured at 30–60-min intervals and then adjusted to hourly rates of movement. As there were no significant differences between mean day and night rates of movement by individuals of either species (Crook 2002), day and night distances were pooled for the analysis (golden perch: n = 36, mean 8·4 m, range 0·1–49·6 m; carp: n= 36, mean 29·9 m, range 0·4–224·5 m).
Several fish were relatively mobile in the first few days of tracking, after which they exhibited restricted movement patterns (see below). To determine if these fish exhibited significant site fidelity after the mobile behaviours had ceased, frequency plots of the estimated locations were examined for each fish and the site fidelity analysis was repeated for spatially and temporally clustered locations. Estimates of home range size were calculated using locations for periods during which significant (P < 0·05) site fidelity was demonstrated. Frequency plots of the estimated locations for each fish were also examined to determine whether fish shifted the location of home ranges during the study. If the frequency plot for an individual was multimodal, the temporal sequence of locations was examined to determine whether the modes represented core areas within the same home range, or a shift in home range location. In the case of a home range shift, the site fidelity analysis and home range size calculations were conducted for each home range separately.
Total linear home ranges were estimated by determining the distance along the river channel between the outermost location coordinates for each individual fish (Young 1999). In addition, 90% and 50% home ranges were estimated for each fish by calculating the minimum distance containing at least 90% and 50% of the locations, respectively. As mentioned, a previous radiotelemetry study found no significant differences in movement rates between day and night for individuals of either species, suggesting that use of only daytime observations did not cause appreciable underestimation of home range size. In addition to the arbitrary home range measures, percentage plots of home range length vs. radiolocation inclusion were used to determine whether there were core areas of disproportionate use within home ranges (Wray et al. 1992).
Data were collected for 33 of the 35 fish tagged with radio-transmitters. However, transmitter failures occurred frequently and resulted in incomplete data sets for many of the tagged fish. Two transmitters failed before any data could be collected and 21 other transmitters failed before the end of the study period (Tables 1 and 2). Slowing of the transmitter pulse rate usually preceded the failure of transmitters. Transmitter failure, as opposed to movement out of the survey area, was further indicated by the failure to locate missing fish during surveys between the upstream and downstream barriers and through the recovery of non-functioning transmitters from three recaptured golden perch.
Table 1. Summary of site fidelity analysis and home range estimation for golden perch. Distances refer to the distance from the original capture location (positive values = upstream direction, negative values = downstream direction). ‘Site fidelity day’ is the day that significant (P < 0·05) site fidelity was exhibited and ‘site fidelity location’ is the location where significant site fidelity was exhibited. aHome range estimate omitted from mean (< 10 locations contributing to estimate). bData between day 8 and day 50 missing. cNo data collected due to transmitter malfunction
Table 2. Summary of site fidelity analysis and home range estimation for carp. Distances refer to the distance from the original capture location (positive values = upstream direction, negative values = downstream direction). ‘Site fidelity day’ is the day that significant (P < 0·05) site fidelity was exhibited and ‘site fidelity location’ is the location where significant site fidelity was exhibited. aHome range estimate omitted from mean (< 10 locations contributing to estimate). bShift in home range after day 56: first estimate is for days 4–56, second estimate is for days 59–72. cNo data collected due to transmitter malfunction
All 10 of the non-translocated golden perch exhibited significant site fidelity during the tracking period (Table 1, Fig. 4a). Six of the 10 golden perch (G1–G6) occupied home ranges near the original point of capture after release. One of the non-translocated golden perch (G7) moved ∼150 m from the pool in which it was collected to the next pool upstream, and then occupied a home range in the upstream pool for over a month before it was captured by an angler. The other three non-translocated fish (G8–G10) made large-scale movements (> 1500 m) soon after release in either upstream or downstream directions before exhibiting fidelity to restricted home ranges. One of these fish (G10) had moved out of the normal survey area by day 8 and was next located during an extended survey on day 50, 7500 m downstream from the point of release.
Four of the five non-translocated carp (C1–C4) exhibited fidelity to home ranges near the point of capture after release (Table 2, Fig. 4b). One of these fish (C3) was not located until day 16 because of a slight change in the frequency of its transmitter. When first located, this fish was ∼750 m upstream of the point of capture. However, when next located on day 20, the fish was 5 m downstream of the point of capture and remained within this area until its transmitter failed after day 51. One of the non-translocated carp (C5) made a rapid upstream movement from the point of capture before establishing a home range ∼2800 m from the original capture site.
Three of the four golden perch with functional transmitters that were translocated upstream (G11–G13) returned to within 100 m of the original capture location within a few days of release (Table 1, Fig. 4c). The single fish that failed to return to the site of capture (G14) remained relatively mobile for the first 10 days after release, commonly moving more than 500 m between surveys. Unfortunately, the transmitter on this fish failed after day 10 and it is impossible to determine whether the fish remained mobile or eventually established a more restricted home range. Of the five golden perch translocated downstream, two (G15, G16) returned to the point of capture. One of these fish (G15) was tracked for 45 days before the transmitter failed. The other fish (G16) was located on only one occasion before its transmitter failed, but was recaptured by angling on day 13 from the original capture location. Two of the golden perch translocated downstream (G17, G18) returned ∼1300 m in the direction of the original site of capture before exhibiting fidelity to restricted home ranges. Both of these fish established home ranges within the same pool, although they arrived independently on days 3 and 6. The other golden perch translocated downstream (G19), moved ∼1000 m upstream towards the original point of capture by day 5 and had begun to move slowly back downstream when its transmitter failed after day 9.
Two of the four carp with functional transmitters translocated upstream (C6, C7) returned to the original point of capture within a few days of release and occupied home ranges in that region for the rest of the study (Table 2, Fig. 4d). Of the two carp that did not return to the point of capture, one (C9) established a home range near the release point and the other (C8) established a home range ∼800 m upstream of the release point. Three of the five carp translocated downstream (C10–C12) returned to the original capture site within a few days of release and occupied home ranges in that area. Of the two that did not return, one (C13) established a home range ∼1000 m downstream of the release point. The other (C14) established a home range near the release point and exhibited fidelity to this area until day 56, before shifting the location of the home range ∼240 m upstream.
home range estimation
Estimates of longitudinal home ranges for fish that exhibited significant (P < 0·05) site fidelity are presented in Tables 1 and 2. Most estimates of home range length with ≥ 10 radiolocations appeared to approach an asymptote (Fig. 5). Home range estimates with < 10 radiolocations often did not appear to approach an asymptote and were therefore excluded from the analysis of home range length. The 100% home ranges of golden perch were significantly smaller than those of carp (Mann–Whitney U-test; P < 0·01), as were the 90% home ranges (0·01 < P < 0·05). However, the 50% home ranges of the two species were not significantly different (P > 0·05). More than 50% of locations for both species were concentrated within 25% of the mean total home range, suggesting that home ranges were not used uniformly and that there were core areas of intensive use within the home ranges of both species (Fig. 6). These results suggest that although carp had larger total home ranges than golden perch, both species exhibited strong fidelity to relatively small core areas within their home ranges.
home range occupation
The random movement analyses showed that golden perch and carp exhibited site fidelity and occupied measurable home ranges at the spatial and temporal scales analysed. The occupation of restricted home ranges by golden perch contrasts with the findings of previous studies that have emphasized high levels of mobility by this species (Reynolds 1983; Koehn & Nicol 1998), although it is in agreement with many other studies of riverine fish movement (see Gowan et al. 1994; Lucas & Baras 2001). A possible reason for the contrast with previous studies of golden perch is that the current study was conducted over a period that was characterized by generally low flow conditions and that did not include the spawning season. Reynolds (1983) suggested that mature golden perch undertake large-scale spawning migrations in response to major flood events, and Mallen-Cooper et al. (1997) found that immature golden perch undertake upstream migrations in response to rises in river level.
The restricted movement patterns found in the current study might also be explained by the fact that other studies of golden perch movement were conducted in larger systems than the Broken River, where the geomorphic characteristics may be less confining to movement (Lonzarich et al. 2000). The results for carp are in general agreement with the findings of Reynolds (1983), who described relatively restricted movements and homing abilities. The current study did not reveal evidence of high levels of mobility by carp as described by Koehn & Nicol (1998). Again, however, this may be due to the fact that the current study was conducted over a shorter period and in a smaller river system.
More than half the translocated golden perch and carp in the current study exhibited strong site fidelity, as demonstrated by homing to the original point of capture. The fact that these fish were capable of negotiating relatively large distances back to the point of capture shows that the generally restricted movements of both species were not simply the result of confinement due to physical barriers, such as shallow areas or wood debris jams. Both the random movement analyses and the translocation experiment results suggest strongly that both golden perch and carp exhibited fidelity to restricted home ranges. Although homing abilities were demonstrated clearly by individuals of both species, almost half the translocated fish failed to return to the original point of capture. This variation in behaviour suggests that there are individual differences in the abilities and/or motivations of fish to return home after translocation. Similar variations have been noted in other fish translocation experiments (e.g. Kennedy 1981; Armstrong & Herbert 1997). It is possible that factors within the original home range such as food availability, predation risk or behavioural interactions with conspecifics affect the motivation of individual fish to attempt to return home after translocation. The availability of suitable habitats near the release point, or between the release point and the original home range, may also influence homing behaviours. For example, two of the golden perch translocated downstream established new home ranges in a large pool approximately halfway between the release point and the original capture site. Large pools of this type are the preferred habitat of golden perch in the Broken River (Crook et al. 2001) and it is possible that these fish would have continued on towards the original home range if this habitat had not been encountered.
home range shift
Although the vast majority of radio-tracked fish exhibited fidelity to restricted home ranges, fidelity was not necessarily exhibited to the original capture site. Several of the non-translocated fish moved away from the site of capture and exhibited fidelity to other regions of the river. Additionally, several of the translocated fish did not return to the original capture site and established new home ranges at or near the release location. One of the carp shifted the location of its home range after 56 days at liberty, showing that changes in home range location were not limited to the immediate post-release period. Several previous telemetry studies have also reported patterns of initial post-release mobility followed by more restricted movements (Baras 1997; Chilton & Poarch 1997; Hilderbrand & Kershner 2000). In a study of cutthroat trout (Oncorhynchus clarki Richardson) in a small stream in Idaho and Utah, USA, Hilderbrand & Kershner (2000) removed the first 10 days of data from their home range estimates because of high rates of post-release mobility that they attributed to a capture effect. However, they also observed less frequent shifts in the location of home ranges after the post-release period, and suggested that the ‘degree of mobility in individuals appears flexible within stream salmonid populations’.
Observations of home range shift have potentially important implications for the interpretation of data relating to riverine fish movements. Many mark–recapture studies, for example, have concluded that fish collected outside the original site of capture are ‘mobile’ (e.g. Stott 1967; Jackson 1980; Hesthagen 1988; Bridcut & Giller 1993; but see Harcup et al. 1984; Smithson & Johnston 1999). Shifts in the location of home ranges, however, raise the possibility that fish recaptured outside the original capture site might be normally sedentary fish that had relocated their home range in the time between marking and recapture. Although home ranges are often quite spatially consistent throughout an animal's life span (Wauters & Dhondt 1992; White, Saunders & Harris 1996), a number of studies of terrestrial animals have also demonstrated shifts in the locations of home ranges. Doncaster & McDonald (1991), for example, found that the locations of red fox (Vulpes vulpes L.) territories drifted continually and that the rate of drift varied considerably between seasons. Red squirrels (Sciurus vulgaris L.) in poor quality habitats have also been shown to shift the location of their home ranges relatively frequently, although squirrels in high quality habitats generally remain within the same home range throughout their lives (Wauters & Dhondt 1992; Lurz, Garson & Wauters 1997).
a‘home range shift’model of fish movement in rivers
Based on the results of the current study and on an evaluation of previous studies (including those cited above), I propose a simple conceptual model to describe movements associated with home range occupation and shift by riverine fish (Fig. 7). It should be noted that, apart from the current study (which is based on data for a relatively small number of individuals over a 3-month period), this model has not been evaluated empirically and its utility and generality require further investigation. The model is not intended as a replacement for the RMP or the behavioural dichotomy model, but rather as a theoretical framework within which to further examine and define the movements of riverine fish.
Under the tenets of the proposed model, riverine fish exhibit fidelity to restricted home ranges for extended periods. These home ranges may contain core areas where the majority of normal activities occur (e.g. foraging, resting). For individuals that defend territories actively, core areas would contain those defended territories. Home ranges may also contain intermediate areas where normal activities are undertaken less frequently than in core areas, and outer areas where infrequent exploratory behaviour occurs. Extended periods of home range occupation would be interspersed by relatively short periods of mobility associated with home range shift. Home range shift might occur in response to events including disturbances (e.g. floods, drought, capture and release by humans), behavioural interactions (e.g. arrival of new predators or competitors), decreases in habitat profitability or the location of a new core area during exploration. Home range shift might also occur following specialized movements associated with the species’ life history, such as spawning or feeding migrations or occupation of overwintering habitats. The process of home range shift would include an emigration phase, a mobile phase and a home range establishment phase (Fig. 7).
Previous studies of the movements of riverine fish have generally measured home ranges at arbitrary temporal scales, including diel (e.g. Young 1999), seasonal (e.g. Snedden et al. 1999) and annual scales (e.g. Hesthagen 1990). Under the tenets of the proposed model, the temporal scale of measurement is vitally important to the measurement of home range size. Home range measurements conducted over a 1-year period, for instance, might include several shift events and would suggest a very large home range. Measurements taken over a shorter time span, on the other hand, might not include any shift events and would indicate a small home range. Rather than basing home range estimates on measurements taken over arbitrary time scales, an alternative approach is to determine the appropriate temporal scale for measuring the home range according to the temporal stability of site fidelity. Measurement of an individual's dispersive potential therefore would need to include information regarding the frequency and magnitude of movements associated with home range shift, as well as the size of the home range itself. The analytical approach used in the current study provides a potential means of measuring these parameters, although repeated observations of individuals using telemetry were required to detect shifts in home range location. Such repeated observations are usually not possible using mark–recapture techniques. It should be noted, however, that telemetry studies (including the current study) usually have lower sample sizes than mark–recapture studies due to the logistical difficulties in collecting repeated observations of individuals.
Switzer (1993) provides a detailed discussion of factors that may influence the stability of site fidelity by terrestrial animals. These factors include the temporal stability of habitat quality within patches, the predation risks and energetic costs associated with changing locations and the probability of mortality within high quality habitats. Studies of riverine fish have similarly emphasized the importance of habitat quality, mortality risk and the stability of habitat quality in determining rates of movement (Bruylants et al. 1986; Gilliam & Fraser 1987; Railsback et al. 1999; Lucas 2000; Bélanger & Rodríguez 2002; Gowan & Fausch 2002). Consideration of the existence of site fidelity has implications for the application of optimization models (e.g. the ideal free distribution; Fretwell & Lucas 1970) because it will result potentially in the failure of predictions that individuals will utilize high-quality habitat patches that are located outside the area to which fidelity is exhibited (Railsback et al. 1999). While the current study has demonstrated the existence of site fidelity for two lowland river fishes, further work is required to elucidate the limitations this behaviour may place on the responses of fish to changes in habitat quality, and to incorporate site fidelity behaviour into models of space utilization by riverine fish.
Alistar Robertson, Alison King, Paul Humphries and two anonymous referees provided helpful comments on earlier versions of this manuscript. Chris Medlin, Simon McDonald and Zygmunt Lorenz assisted with the analyses and David Titley designed the radio-transmitters. Alison King, Paul Humphries, Helen Gigney and Nigel Anthony assisted with fieldwork. Paul Humphries provided the water temperature data and Goulburn-Murray Water provided the flow data. I thank Geoff and Lois Heaney, David and Julie Allnut and Doug Farley for allowing access to the study site and for their generous hospitality. This study was supported by a Charles Sturt University Postgraduate Scholarship and a CRC for Freshwater Ecology Postgraduate Top-up Scholarship.