Introduction

Large-bodied non-native fishes of high trophic position are often introduced into inland waters to enhance recreational freshwater angling (Hickley & Chare, 2004; Ellender & Weyl, 2014). Should these fishes develop an invasive population then in addition to being a major driver of biological and functional homogenisation, they can have substantial negative effects on native fish diversity through both their consumptive effects on prey species and non-consumptive effects on native trophic analogues (Eby et al., 2006; Sih et al., 2010). The release of these fishes into rivers modified by engineered structures (e.g. dams and weirs) often results in their successful establishment and invasion because of the favourable conditions provided by the more homogeneous and stable hydrological environment (Clavero et al., 2004; Johnson et al., 2008). However, weirs and dams can also act to limit the upstream dispersal of invasive species, thus protecting native species in those areas (Burnett et al. 2023).

These engineered riverine environments often also result in the loss of river-floodplain connectivity, despite the importance of this connectivity for enabling fish to access a range of functional habitats in the floodplain (Bolland et al., 2015). Moreover, engineered rivers are often also subjected to channelisation and artificial levee construction that results in a straightened channel (Brookes et al., 1983). These modifications result in episodic high flow and flood events being of higher severity as the water remains in the main channel throughout, with the conditions often preventing fish from accessing off-channel refugia (Bolland et al., 2012). Accordingly, in temperate lowland rivers, summers often provide homogeneous and stable hydrological environments that are favourable for larval and juvenile fish recruitment (Nunn et al., 2007a), increasing the probability of non-native fish establishing. However, during winter periods, these rivers often provide relatively hostile conditions characterised by elevated flows and large in-channel flood peaks (Death et al., 2015), which can be deleterious for juvenile fishes where off-channel refugia is limited and largely inaccessible (Bolland et al., 2015).

The lower River Severn in western England was first modified in the 1850s through construction of multiple of navigation weirs, with the impounded river also subjected to channel straightening to assist flood prevention (Gutmann Roberts et al., 2019). Prior to modification, the river channel included areas of water that was sufficiently shallow for boats to have to pushed upstream manually (IHBC, 2021) and the river is considered likely to have been a heterogenous habitat for fish including multiple natural off-channel refuges. Today, however, the main river channel is characterised by a largely straight river channel with high flood banks, depths that always exceed 2 m and are usually deeper. There are also minimal off-channel habitats for fish, with the most downstream non-tidal impounded section only having one off-channel macro-habitat present in the form of a boat marina of approximately 3 ha in size. The fish assemblage of the river had also been modified by the presence of alien species including the obligate piscivore pikeperch Sander lucioperca (Linnaeus, 1758), which has been present since 1980 (Hickley, 1986). Pikeperch thus coexist with native piscivore Northern pike Esox lucius Linnaeus, 1758 in the river.

Differences in the biology of Northern pike and pikeperch suggest some potential for differences in their habitat use and movement patterns. For example, Northern pike have greater foraging success than pikeperch in the presence of submerged vegetation (Greenberg et al., 1995), with these habitats also being important Northern pike spawning substrate and nursery habitat, especially in shallow water but with spawning also occurring in water up to 6 m deep (Craig, 2008). Conversely, pikeperch spawning involves nest construction and guarding by males, with nests made on sandy, silty or muddy substrates (Lappalainen et al., 2003). Northern pike in temperate waters tend to spawn in early spring (March, April), with pikeperch spawning generally commencing in April (Craig, 2008). The reproductive cycle is at least in part governed by temperature, with active vitellogenesis commencing in autumn and accelerating as water temperatures reduce in winter (Lenhardt & Cakić, 2002).

In anthropogenically modified rivers, off-channel habitats also provide Northern pike with important areas for refugia and foraging (Pauwels et al., 2016), while pikeperch foraging generally involves active searching in open water (Turesson & Brönmark, 2004). Although both species tend to rely on sight for locating prey, juveniles of both species are capable of foraging successfully in turbid water, although energetic costs might be increased (Skov et al., 2002; Zingel & Paaver, 2010). Both species also tend to move more at twilight as this elevates foraging success (e.g. Horký et al., 2008; Baktoft et al., 2012) and, where pikeperch are invasive, they often share prey resources with Northern pike (Jepsen et al., 2000), especially in similar size classes (Nolan et al., 2019a). Consequently, in channelised and impounded river sections of limited off-channel habitat, the two species could potentially express similar movement patterns and habitat use in relation to foraging and spawning. However, pike movements are also scaled by body size, with larger fish using greater spatial areas than smaller pike, thus potentially avoiding interactions with other piscivores (Dhellemmes et al., 2023), and with floods potentially displacing some individuals from areas of river that do not always return (Chevallier et al., 2023).

The aim of this study was to thus simultaneously assess the habitat use and movement patterns of these two piscivorous fishes in the lower River Severn basin, England, using acoustic telemetry. The objectives were to: (i) determine the spatial and temporal movement patterns (including diel movement patterns) for both species; (ii) test the influence of environmental parameters on these movement patterns; and (iii) quantify the importance of the limited off-channel habitat to each species (as an off-channel residency index) versus the main river channel. We predicted that the two species would show similar movement patterns and habitat use, from diel to seasonal patterns that were temperature related, with the off-river refuge providing important habitats for both species throughout the year. The application of acoustic telemetry, where the individual fish were implanted with an internal acoustic transmitter that enabled their detection on an array of acoustic hydrophones (receivers), enabled the fish movements to be recreated through time and space using these detections and tested versus environmental parameters, as has been completed for other fishes in the study river (e.g. Gutmann-Roberts et al., 2019; Davies et al., 2022). Thus, throughout the study, a movement is defined as a tagged fish being initially detected on one receiver and subsequently detected on another receiver located either up- or downstream. The river distance between the receivers allows for determination of the movement distance, and the detection times of the receiver provide information on the direction and duration of the movement.

Materials and methods

Study area

The study area was a section of the lower River Severn, Western England (Fig. 1a) between Diglis Weir (upstream) and Upper Lode Weir (downstream) (Fig. 1b; 52.1819, − 2.2241 to 51.9943, − 2.1735) and encompassing the lower section of the River Teme tributary. The weirs at the up- and downstream limits of the area provided a closed area of 28 km, as neither of the species were assumed to be capable of traversing these weirs because of their height and with no fish passes present at that time; although locks are present next to the weirs to assist navigation, fish have not been detected moving through them in any studies on the river (e.g. Davies et al., 2022, 2024). It was also considered highly improbable that the fish would leave the study area during brief periods of flooding when the weirs were over-topped, and to the best of our knowledge, such events did not occur. The weirs were constructed for navigation and have resulted in the river being highly impounded, with heavy boat traffic in summer (Fig. 1b). The study area is characterised by widths to 40 m, depths to over 4 m (C-MAP, 2019); there is minimal instream vegetation and off-channel areas, with the only exception being a boat marina located at Upton-upon-Severn (Fig. 1).

Fig. 1
figure 1

Map showing a the position of the study area within the UK; b the study area within the River Severn with the receiver and their respective locations shown by circles and the position of the flow gauging station (star) c the off-channel habitat provided by Upton-Upon-Severn marina showing the receiver locations (circles) and the sampling location (triangle). The arrow indicates the direction of water flow, solid lines show the position of weirs, * marks the location of the Saxons Lode gauging station. All receivers were in place for the entire duration of the study with the exception of receiver #2 and #3 which were deployed on the 24 April 2018

Full size image

An acoustic receiver array was established in the study area prior to fish tagging (on the 21 August 2017), except for receivers #2 and #3 which were deployed on 24 April 2018. The receiver array thus comprised a total of 11 acoustic receivers (VR2, Vemco Ltd) in fixed locations (Table 1; Fig. 1b). These receivers remained in place throughout the remainder of the study period, (Fig. 1b). Range testing revealed a maximum detection range of approximately 100 m across the study area (Gutmann Roberts et al., 2019); this exceeded the river width in all locations and thus the receivers functioned as a gated array. Receiver positions were selected to provide equidistant coverage between the upstream and downstream range of the study area, whilst also enabling detection of movements in and out of the boat marina at Upton-Upon-Severn, and the residency of the fish in this marina (Fig. 1c). Receiver batteries were removed and replaced periodically, enabling the stored data to be downloaded for analysis. A temperature logger (Tinytag; Gemini Data Loggers) was deployed at the site of receiver number 6 and recorded temperature (to 0.1 °C) every three hours. Flow data (m3 s−1) were acquired from the flow gauging station operated by the Environment Agency at Saxons Lode (52.0495, − 2.2005, Fig. 1b), with river flow recorded every 15 min.

Table 1 Receiver identification, name and location coordinates (decimal degrees), the total number of detections recorded by each receiver within the study period, and the time and date of first and last detections
Full size table

Fish sampling and tagging

Fish sampling and tagging was completed on 27 September 2017 in the river and marina at Upton-Upon-Severn (Fig. 1c), selected because of its centrality within the study area. Fish were captured using electric fishing from a boat, coupled with rod and line angling. Following their capture, fish were transferred to an aerated tank. Tagging involved general anaesthesia (tricaine methanesulfonate; MS-222) before an acoustic transmitter (69 kHz V9 or V13; Vemco Ltd) was inserted into the peritoneal cavity through a small incision (less than 2 cm wide) which was then closed with a single suture and the application of surgical adhesive. V9 acoustic transmitters were 9 × 21 mm and 1.6 g, whilst V13 transmitters were 16 × 36 mm and 6 g. All transmitters were set to transmit randomly every 60 to 180 s, providing an overall battery life of approximately 22 months (V9) and 36 months (V13). Random repeat pulse rates allowed multiple individuals to be monitored simultaneously within a given area via the fixed receivers with reduced risk of continuous signal overlap and interference. Tag identification numbers were recorded, with the fish measured (fork length, nearest mm) and transferred to an aerated recovery tank where they were held until normal swimming behaviour resumed. The fish were then released close to their location of capture. All surgical procedures were undertaken by a licensed, competent and experienced practitioner following an ethical approval process, with all regulated procedures completed under UK Home Office licence PPL 70/8063. A total of 17 Northern pike and 8 pikeperch were tagged, Northern pike ranged in size from 574 to 958 mm and pikeperch from 356 to 692 mm (Table 2).

Table 2 Summary data for each transmitter, including species, fork length at tagging (mm), the date of last detection, the number of days from first to last detection, the number of days detected, the number of detections, upstream and downstream distance (m), total distance (m), mean daily distance (m) and off-channel residency index during the 12-month study period
Full size table

Data analysis

The fish tracking data were analysed for the period 27 October 2017 to 26 October 2018 that provided data over 365 continual days. Data from the period between tagging and 27 October 2017 were not included to avoid movements that might have been subject to behavioural changes caused by the tagging procedure (Pauwels et al., 2014). Although tagging procedures can reduce fish movements for several days post-tagging (e.g. Sonamzi et al., 2020), this was not considered a major concern. Movements during this period would be indicative of under-estimated distances moved rather than unusually high movement patterns. Northern pike ID 51155 was not detected after 09 November 2017 and so was removed from further analyses.

At the end of the tracking period, the detection data (comprising over 1.6 million individual detections) were initially analysed in the package ‘Vtrack’ in R (Campbell et al., 2012) for residency and non-residency events for each individual. A residency event was defined when a transmitter was detected by a receiver (minimum of 2 detections) and terminated when the transmitter was detected at another receiver, or if the transmitter was not detected by the same receiver within a defined timeout window of 10 min. This time was chosen as a conservative estimate of the time it would take an individual Northern pike or pikeperch to move away from the detection range of a receiver (~ 100 m) based on Northern pike mean swimming speed of 0.23 m s−1 (0.45 body length s−1) (Diana, 1980). No data exists for absolute swimming speed in pikeperch, although swimming speed has been recorded as 1.6 body lengths s−1 (Poulet et al., 2005a). A non-residency event was defined as the movement between the detection fields of two receivers and incorporated measurements of the circuitous distance (river distance) between receivers, with river distance used in all distance calculations. Thus a non-residency event is analogous to a fish movement between two receivers in a specific timeframe). Throughout the study, the receivers remained functional and detection efficiency was > 99% (i.e. minimal instances of where the detection of a moving fish was missed on a receiver between two other receivers that did detect it), and thus no action was needed on this. No false detections were apparent in the dataset. The detection fields of receivers never overlapped, ensuring that individual fish detections occurred on a single receiver and were never detected simultaneously on multiple receivers. Consequently, the detection of a fish on adjacent receivers indicated a movement, with the time between detections on the receiver providing information on the speed of movement.

Individual maximum upstream and downstream distance moved was calculated from the central position of Upton-Upon-Severn marina (‘0’; Figs. 1 and 2) to the most upstream and downstream receivers with detections. Individual total distance moved was calculated for the 12-month study period as the sum of all movements between receivers. Individual mean daily distance was then calculated for the 12-month study period as the total distance travelled by each individual, divided by the length of the study period (i.e. the time between first and last detections for each individual). Although such movement rates are likely to be an underestimate of total movement, they can provide useful insights (Cooke et al., 2001), and are an attempt to reduce the error associated with differences in the number of days individuals were detected. Spearman’s rank correlation coefficient was used to initially test for a correlation between individual total distance moved and mean daily distance moved across the analytical period; as there was significant correlation for both Northern pike (Spearman’s rho (r) = 0.97, P < 0.01) and pikeperch (r = 0.95, P < 0.01), then following testing for normality and homoscedasticity (Shapiro-Wilks and Levene’s tests, respectively), differences between Northern pike and pikeperch in mean daily distances moved during the 12-month study period were tested using a Mann–Whitney U test.

Fig. 2
figure 2

Continuous upstream and downstream movements (km) of Northern pike and pikeperch from 0 (r representing the marina at Upton-Upon-Severn) on the primary axis (solid line), and the total monthly distance moved (km) on the secondary axis (open circle, dashed line); spawning month is represented by the area between the dotted lines and individuals are identified according to transmitter ID (Table 2). Note the difference in scale between the primary and secondary axis

Full size image

To assess seasonal differences in total daily distance moved (cf. Objective i), a mean of the total daily distance moved was calculated for each species for each day in the study period. Differences in this mean total daily distance moved across seasons were tested using a generalised linear model (GLM) with a quasi-Poisson distribution to account for over-dispersion, with the significance of the model tested using a likelihood ratio test.

To assess movements in relation to time of year, water flow and temperature for both species (cf. Objective ii), it was first necessary to ensure that all data were comparable across the study period. For Northern pike this meant removing individuals that were not detected for the entire 12-month study period, resulting in the exclusion of 4 fish (Table 2). For pikeperch, because of a lower number of tagged individuals, the removal of fish without a full 12 months data would have excluded 50% of the sample. Consequently, pikeperch ID 43266 was removed, as this individual was last detected on a receiver on the 25 April 2018 and so would not span the entire spawning period for pikeperch. The analysis of movement for the remaining 7 individuals was completed only for months when all individuals were present (to 04 July 2018; Table 2). Then, a mixed effects logistic regression model tested the binary response of daily movement (as a detected daily movement versus no detected daily movement) against daily mean water temperature and river flow for both species, with season as a fixed effect and individual as a random effect in the model. Individual was included as a random effect to mitigate autocorrelation from repeated measures from the same individual (Harrison et al., 2018). Water temperature and flow were entered as quadratic terms to account for potential non-linear relationships and data were scaled for continuous variables before analyses. As season was defined according to the Northern meteorological season then the pikeperch model did not have a complete dataset for summer and autumn.

As range testing revealed that the receiver located within the boat marina could only detect acoustic transmitters within the marina and not the river, then the proportion of time spent in the marina by each fish was assessed as its ‘off-channel residency’ (cf. Objective iii). This was calculated as the total time of individual residency events within the marina for both the length of the entire study period (i.e. the time between first and last detection for each-individual) and weekly. A GLM (Poisson distribution and negative binomial distribution where over-dispersion was apparent) tested the influence of river temperature and discharge (and their interaction) on the presence of each species in the main river channel (as the number of fish present per day and the total number of detections per species per day).

Daily timings of dawn, day, dusk and night were retrieved for each day in the study period, obtained using the package ‘maptools’ with civil twilight definitions, with the package providing the times of day of these diel periods across the year (and thus the seasonal differences) (Bivand & Lewin-Koh, 2019). Diel periods were defined as dawn, day, dusk and night. A fish movement during each of these diel periods was defined when a fish was initially detected on a receiver and then detected on the next receiver upstream or downstream later in the same diel period. As the receivers did not overlap in their detection fields, fish could not be detected simultaneously on adjacent receivers. Therefore, the detection of a fish on two receivers at different times within the same diel period constituted a movement in that period. Detections of a fish on a single receiver in one diel period and its subsequent detection on an adjacent receiver in the next diel period were not counted as movements here. Diel movements between the marina and the main river—in either direction—were only counted when the fish was detected as moving between both receivers in the same diel period. Once the number of diel movements had been determined for the dawn, day, dusk and night period of each day, they were standardised to the number of movements per hour for both species. These measurements were calculated for the entire 12-month study period, by season and by month for individuals with 12 months of data only. A Kruskal–Wallis rank sum test then tested the overall differences in number of movements over the 12-month study period across diel periods, and a Chi-squared (χ2) contingency table analysis was used to test for an association between dawn, day, dusk and night movements with season (cf. Objective i). Monthly movement within each diel period was used for graphical purposes only. Tests were completed for Northern pike and pikeperch separately.

Analyses and graphical outputs were completed in R (Version 4.2.3; R Core Team, 2023). Logistic regression and generalised linear models were analysed for Northern pike and pikeperch separately and were completed using the package lme4 (Bates et al., 1976; Masters et al., 2005), with optimum spawning temperatures of between 6 and 14 °C (Frost & Kipling, 1967). In mature individuals, movements of Northern pike increase around the spawning period (Baktoft et al., 2012), with migratory spawning movements in spring accounting for differences in seasonal movement patterns (Ovidio & Philippart, 2005; Koed et al., 2006; Pauwels et al., 2014). Indeed, evidence is rare of seasonal differences in movement patterns beyond their migratory spawning behaviour (Kobler et al., 2008). Temperature also usually triggers spawning migration in both sexes (Pauwels et al., 2014), and although Northern pike often show homing to natal spawning sites (Engstedt et al., 2014; Sandlund et al., 2016), this might not be evident across the entire population (Vehanen et al., 2006). It is possible, therefore, that Northern pike making the relatively long-distance movements in spring in this study were moving to specific spawning locations that were suitable for spawning, such as over vegetation in shallow water (Casselman & Lewis, 1996), especially given that these areas were limited given the river engineering and impounded environment. However, a short-coming of acoustic telemetry is that while it generates movement data, it does not currently indicate the activity of the fish in the different habitats. Accordingly, while Northern pike movements in spring have been assumed to relate to their spawning activities, this is an assumption that could not be tested further.

For pikeperch, spawning occurs between 8 and 16 °C and, as it is expected to occur predictably across latitudes, tends to commence in mid-April in central England in most years (Lappalainen et al., 2003). Increased movements of pikeperch in spring is also associated with reproductive behaviour and spawning migrations (Lappalainen et al., 2003), but this can vary between sexes, with male movements being reduced because of their nest guarding behaviours, where males can remain in position for between 2 and 6 weeks (Jepsen et al., 1999; Poulet et al., 2005b). However, it is also common to see seasonal peaks in pikeperch movements that are unrelated to spawning migrations, such as peaks in autumn, summer or winter movements that are linked to feeding migrations, particularly in systems where prey resources are limited (Jepsen et al., 1999; Koed, 2000; Horkỳ et al., 2006). However, the pikeperch movement patterns were relatively consistent across the study, with no apparent peaks of movement. As pikeperch prefer to spawn in still or slow flowing waters (Lappalainen et al., 2003), including in low velocity areas in rivers (Koed et al., 2002), their higher residency index in the marina in spring was likely through them using the area for spawning. Again, as the acoustic telemetry data could not indicate the activity of the fish in the different locations where they were detected then this spawning in the marina can only be assumed. However, exposed pikeperch nests were observed in the marina during a receding flood in April 2018, indicating that some pikeperch did indeed spawn there.

Temperature is an important factor in explaining increased movement for both Northern pike and pikeperch. For example, relative changes in temperature can stimulate movement in pikeperch during the pre-spawning and spawning season, with the extent of temperature changes potentially a more important predictor of movement than actual water temperature (Saulamo & Lappalainen, 2007). However, the relationships of the movement of both species with temperature appear to be strongly nonlinear. In pikeperch, decreased movements with reducing temperature, and elevated movements with increasing temperature, have been detected (Jepsen et al., 1999). In Northern pike, elevated winter temperatures of up to 5 °C can also trigger increased movements (Jepsen et al., 2001; Koed et al., 2006), and with both very high and low temperatures decreasing overall movement rates (Kobler et al., 2008). Northern pike in a Finnish river decreased their movements when water temperatures exceeded 20 °C (Vehanen et al., 2006). The tagged fishes in the River Severn thus had movement patterns that were affected by temperature in a similar non-linear manner to these studies.

On a daily basis, many piscivorous fishes have peaks in their movements at dawn and dusk because of the potential for higher foraging success during these periods (Helfman, 1986). Here, Northern pike had only increased their movements during twilight periods in summer and autumn, with pikeperch having no similar peaks in movements. For Northern pike, these results were similar to those of Kobler et al. (2008), who suggested increased movement in summer twilight periods was a short-term behavioural response to maximise food intake during this period. For pikeperch, a measured increase in their movements in twilight periods has also been found (Horký et al., 2008), but other factors such as water temperature, individual behaviour and sex also influencing their diel movement patterns (Poulet et al., 2005a).

These results suggested that whilst there was some intra-specific variation in the movement patterns and habitat use of both Northern pike and pikeperch in the river, there were also some important inter-specific differences. In particular, the boat marina—the only off-channel macro-habitat available to the fish—was important to Northern pike all year round, where it was assumed it provided an important foraging habitat and, potentially, a key spawning area. In contrast, pikeperch primarily used this off-channel refuge in winter and spring, with it again likely to provide an important spawning area. However, the actual activity of both species could not be measured directly by acoustic telemetry and hence why activities such as spawning have to be assumed. These results suggest that the increased provision of off-channel refugia in this river section would provide both species with increased access to important functional habitats, coupled with their provision of new nursery areas for larval and juvenile fish more generally, given the importance of such refugia for supporting fish recruitment success in temperate lowland rivers (Nunn et al., 2007b; Bolland et al., 2012, 2015). Increasing this habitat provision could thus result in the increased abundance of both piscivorous species, which would be considered as ecologically beneficial for Northern pike. Pikeperch, however, are non-native to the river, albeit they are now established and the feasibility of their eradication is considered as minimal (Britton et al., 2011). Pikeperch increased abundance is thus potentially detrimental to the ecological status of the river. Notwithstanding, the recreational fishery they support is increasing in popularity (Hickley & Chare, 2004) with catch-and-release practices used despite these being contrary to extant legislation (Nolan et al., 2019b). As such, increased pikeperch abundance in this river section would potentially support an increase in the socio-economic value of the fishery.

Conclusion

These results provided important insights into the movements of these co-existing native and non-native piscivorous fishes and can assist the management of this temperate lowland river and the fishery it supports. Both species used the limited off-channel habitat at key times of the year, with Northern pike using it all year round, and thus this outcome can be applied to the consideration of increasing the provision of off-channel habitats, given their highly limited availability in this lowland and highly fragmented river.