Buoyancy aptitude of the European wels catfish in relation to vertical migration and hunting behaviour

Abstract

Background

The European wels catfish (Silurus glanis) is a large, physostomous apex predator of freshwater habitats. Its current range expansion both within and beyond its native distribution, is strongly influenced by global warming. This study investigates patterns of diving behaviour of a wels catfish population that has recently increased significantly in a large, deep, cold-water peri-alpine lake.

Methods

Pressure and temperature loggers were implanted into 42 free-ranging wild caught wels catfish in Lake Constance. High-resolution data from 13 recaptured individuals was subsequently analysed.

Results

Two distinct vertical migration modes previously undescribed for this species were identified: fast vertical movements, which lasted for less than 5 minutes, and slower vertical movements in which the vertical movement enacted over a period exceeding 15 min. Swimming depth was shown to be strongly affected by season and water temperature. Fish occupied deeper waters in winter, and shallower depths during the warmer summer months. During summer, vertical movements occurred predominantly around dusk and night with few vertical movements during daytime, indicating nocturnal foraging behaviour. During winter however, vertical movements were more equally divided over the day. Fish making relatively fast vertical movements returned to their original depth, whereas slow vertical movements appeared to be migration between depth layers. The amplitude per period of vertical movement of both fast and slow vertical migrations was mostly restricted to a pressure change of less than 30%. This limitation is likely to be related to the physostomous nature of the species, wherein buoyancy is controlled by a swimbladder with a limited capacity for volume change under pressure.

Conclusions

These results provide new insights into the seasonal timing and diving aptitudes of a highly significant aquatic predator. Vertical movements were mostly made in darkness in summer, and during low light conditions during winter, in agreement with wels catfish having a nocturnal lifestyle in which they forage by detecting and following the wakes of their prey. The narrow amplitude of pressure changes during the vertical movements underscores the importance of tight control over buoyancy during navigation, which may be an important factor allowing it to colonize large deep lakes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40462-026-00634-0.

Background

Temperature and buoyancy control play an important role in fish migration patterns [1–3] and both have been linked to the ability of teleost fish species to adapt to new environments [4]. Teleost fishes with a swimbladder are able to achieve neutral buoyancy by displacing a volume of water with a mass equal to total body mass, which decreases energy expenditure in lift-generating locomotion [1, 5]. This enhances their energetic efficiency and expands their behavioural repertoire [6] over species without such an organ, such as sharks and rays [7, 8] and some teleosts [9]. Maintaining a position in the water column independent from speed [1, 5] and at any depth in the water column [10] is particularly beneficial for species that utilise stalking as part of a hunting strategy since it permits a stealthy approach - a strategy deployed to great effect by species like the European pike (Esox lucius) [11] and the European wels catfish (Silurus glanis) [12].

The volume of gases in the swimbladder, and thus the relative density of the fish which is inversely proportional to hydrostatic pressure, both vary with depth in a linear fashion, according to Boyle-Mariotte’s Law. To control buoyancy, gases are exchanged between the swimbladder and the environment via the bloodstream, but this mechanism is not without drawbacks. The exchange of bladder gasses mostly takes place via capillaries that line the wall of the swimbladder [6, 13–15]. As this is a slow process that may take several hours for an individual to fully adapt, rapid changes in depth or pressure result in compression or expansion of the bladder [2, 16]. In physoclist fishes with a closed swimbladder [6, 15] this may have severe consequences such as embolisms or swimbladder ruptures [17, 18]. Physostome fishes are less affected as they are able to exchange bladder gasses directly via a pneumatic duct connecting the swimbladder to the gut [13, 14].

Maintenance of neutral buoyancy in swimbladdered fish is thus a precarious equilibrium in which any increase or decrease in pressure causes changes in lift that quickly result in accelerating ascent or descent [1]. Such fish are therefore equipped with fast autonomous reflexes [19], to actively correct for loss of buoyancy caused by varying hydrostatic pressure [5, 20], and are obliged to limit rapid vertical movements e.g. while foraging [1, 6]. Many species maintain slight negative buoyancy to assist a vertical migration mode in which gentle descending forward glides alternate with active upward swimming to regain elevation [4] while protecting the fish from accidental upward drift while inactive [6]. The ability of some physostomes, including coregonines and some catfish to release air from their swimbladder permits swift ascent without injury or loss of buoyancy control [21, 22]. This may have significant implications for vertical migration modes and enhances the ability to adapt to new or changing environments [4].

This study measured vertical movement patterns in the European wels catfish Silurus glanis. This species is particularly well-suited for the deployment of loggers for tracking activity [23–25], as it is one of the largest freshwater physostomes globally, at up to 2.92 m long with and over 130 kg in body weight [26]. The species native range is in Middle and Eastern Europe, but following extensive anthropogenic introduction it is now widespread and abundant in Western and Southern Europe [27, 28] and frequently found in freshwater systems outside Eurasia [29–31]. The dramatic impacts of high wels catfish densities outside their native range are well known. For example, invasive wels catfish on the Iberian peninsula have driven small-bodied native species of fish to the brink of extinction [32]. Furthermore, recent scientific evidence shows that wels catfish stocks are strongly increasing within the species native range as well [33, 34], bucking the general trend for apex predators which are globally in decline due to rapidly changing habitats [11, 35]. Further investigation is urgently required to understand the adaptations that underlie the, often problematic, expansion of this apex predator.

Following the major population increase of the European wels catfish in a deep, large lake [34], this study sets out to analyse patterns in vertical movements of free-ranging individuals, using temperature- and pressure-loggers. The daily and yearly activity patterns of this species have been well described in literature [23–25]. This study aimed to investigate the long and short-term changes in depth use and how this might be regulated or limited by buoyancy control at different depths. The study further describes seasonal and daily variation in vertical migrations and how these relate to water temperature. With this data at hand, we describe the diving aptitude of the European wels catfish in the context of its recent expansion success.

Methods

Design: study area and experimental setup

The experiment took place in Upper Lake Constance (ULC, 47 °33’10.1“N, 9 °32’37.5“E) (Fig. 1) a large, deep, oligotrophic lake and the third largest inland body of water in Central Europe. The population density of endemic wels catfish in ULC has increased since the late 1990s, most likely linked to effects of climate warming. Current catches of commercial fishermen exceed 7000 kg a year, an almost 30-fold increase since the 20^th century with no significant change in fishery effort (Baer et al. 2025).

Fig. 1.

Location of lake Constance in Europe and study area; European wels catfish were captured in the Bay of Bregenz. This bay has shallow areas that warm up in summer to ~ 20 °C and areas up to 100 m deep that remain < 10 °C, year-round

A total of 42 wels catfish were caught in September 2022 at two locations in the south-east of ULC (suppl. Table 1): one close to the mouths of the rivers Argen and Schussen (n = 14; 47 °37’43.3“N, 9 °30’20.5“E), and the other in the Bay of Bregenz (n = 28; 47 °30’22.0“N, 9 °37’07.1“E) (Fig. 1). Wels catfish were live-caught by professional fisherman using a fyke net at a depth of approximately 3 m. Fish were pre-selected to a minimum weight of 1 kg to ensure the fitting of data loggers (see below) added less than 2% to total body mass [36], thereby minimising any effect on swimming performance [37].

Data storage tags (G5 Long-Life, Cefas Technology, UK) with dimensions 12 mm × 36.5 mm and a weight (in air) of 6.5 g were implanted intraperitoneal to log pressure and temperature at a frequency of 0.5 min. A moist surgical drape was used as cover to prevent the fish from any drying during the procedure. Surgical instruments were cleaned with a 7.5% povidone-iodine solution (Dynarex, Montvale, USA) and a new scalpel blade was used for each individual fish. The abdominal cavity was opened with a 1–2 cm dorsal incision anterior of the pelvic fins, and the data logger was inserted. The incision was closed with a sterile FS-1 needle and 2–0 Vicryl Plus antibacterial suture (VCP453H, Ethicon, USA). After this procedure, the animal was individually tagged through the pterygiophores of the dorsal fin [38] (Mark II Long Pistol Grip + T-Bar Anchor Tag FD-94, T length = 16 mm, Tag length = 47 mm, Floy Tag, USA), to aid identification of implanted specimens. The entire procedure including surgery and marking took less than five minutes per individual (Fig. 2).

Fig. 2.

Operated European wels catfish with sutured incision at the belly (red arrow)

After surgery and marking, fish were placed in a holding tank with fresh aerated lake water to recover from anaesthesia and monitored until all observable effects on respiratory rate, posture, and swimming behaviour had disappeared. All fish completely recovered from the procedure within 0.5 min, and were released within 1 h post-operation near to the site where they were caught [27, 39, 40].

Recapture of marked fish

Local professional fishermen and angling clubs were alerted to the project at a variety of lecture events, via a website and social media channels and asked to look for tagged fish and report them. Rewards were offered for the return of the loggers. The information on capture location, weight and size (total length) could be submitted online, via a letter or by direct contact with the researchers. Unfortunately, no precise length or weight data of the recaptured fish could be retrieved from anglers and commercial fishermen, meaning no attempt was made to calculate growth data between tagging and recapture. Measures of weight and length are given in suppl. Table 1. Sex of the fish could not be established. Unrecaptured fish will lose their external tag within two years [38], and no effect on performance is anticipated from those continuing to live with an implanted data logger [37].

Identification of descents and ascents

Fish respond to small changes in hydrostatic pressure in the range of 0.5 to 2% and may use such variation to navigate [16, 41]. Therefore, in the current analysis we used a larger change in hydrostatic pressure experienced by the fish to define a diving episode, i.e. ascent or descent. We used a haphazardly selected criterion of more than 3.33% hydrostatic pressure change as a significant change that fish can detect [20] and has physiological consequences in terms of lift or that the fish has to correct for to remain buoyant. This pressure change of 3.33% is achieved with a descent or ascent of 1 m at 20 m depth and of 0.5 m at 5 m depth [16]. To correct for outliers in pressure profiles, such as rapid pressure changes within half-minute intervals, the data were smoothed with a Kalman-filter procedure using the bssm package [42]. For the method of smoothing and selection see the supplement. To provide more detailed information about the frequencies and profiles of ascents and descents, these were classified in four categories of increasing pressure changes that were two, four or eight times higher than the minimum criterion of 3.33% change (i.e. >3.33%, >6.67%, >13.33%, >26.67%). Further they were labelled as descents or ascents based on increasing or decreasing pressure. The frequency of descents and ascents per hour is referred to as vertical movement activity.

To detect variation in vertical movements on a slower time scale than possible with the original dataset, which has a logging frequency of 0.5 min, a second dataset was compiled by averaging 20 logger values thus resulting in a logging frequency of 10 min. Then a similar procedure as above was used to identify slow descents and ascents. In the text these two modes of vertical movements are referred to as slow vertical migration mode (logging frequency = 10 min) vs. fast vertical migration mode (logging frequency = 0.5 min).

Rate of inflation of the swimbladder

The effect of a descent or ascent on the swimbladder was estimated by calculating the theoretical compression or expansion of the bladder gases using the Boyle-Mariotte law [1]:

1

Where (m) was calculated as the mean of depth values 60 to 10 min before the descent or ascent and (m) was the depth arrived at. Hydrostatic pressure on the swimbladder decreases by 1 bar for every 10 meters ascent. Therefore, depth divided by a factor of 10 plus 1 to correct for atmospheric pressure was used to calculate hydrostatic pressure on the volume of the swimbladder. The relative effect on the volume of bladder gases (without ablation of air) (in the formula) was then calculated as a ratio by division of pressure at destination depth and starting pressure at a starting volume of 1 ( in the formula).

Calculations for swimbladder buoyancy were based on general formulas for hydrodynamic lift according to Boyle-Mariotte’s law [5, 14]:

2

3

Where is the force of hydrodynamic lift, is fish volume, is swimbladder volume, is gravitational force, is net density of wels catfish and is the density of lake water.

In this formula, we estimated the net density of wels catfish to be 1.08 kg L⁻¹ based on literature values of the weight to volume ratio of fish carcases [14, 43] and measurements carried out on channel catfish (1.078 ± 0.008 kg L⁻¹ [4]). The density of freshwater in Lake Constance was set at 1.00 kg L⁻¹ (1.000–0.998 kg L⁻¹ between 10 and 20 °C [44]) and acceleration due to gravity was given to be 9.8 m s².

Data management and statistical analysis

Individual and storage tag data were managed in Excel (Microsoft Office Professional Plus 2016, Microsoft Corporation, California, USA) and transferred to R version 4.1.3 for analysis [45]. Pressure (bar) and temperature (°C) data in the raw file were logged at a frequency of 0.5 min. Depth was calculated by subtracting atmospheric pressure (1 bar) from the total pressure and multiplying by 10. Speed at ascent or descent was calculated as the absolute value of the change in depth (in meters) and dividing it by the time interval (in seconds).

Daylight periods at the lake surface were calculated for the study geolocation and for the timeframe of the study with the package suncalc [46] with dusk defined as the period from the start of sunset (bottom part of sun touching the horizon) to the start of night (the sun is at 18 degrees below the horizon), dawn as end of night (sun rises above 18 degrees below the horizon) to the end of sunrise (top edge of the sun appears on the horizon), and night and day as the respective periods between dusk and dawn. Season was defined based on standard astronomical classification.

Prior to statistical analysis, the distribution of each dependent variable was evaluated using the package fitdistrplus [47], and residual plots were deployed post hoc to check for normality, overdispersion and homogeneity of variances using the package performance [48]. Results were considered significant at p < 0.05. For testing linear relationships Pearson’s correlation coefficient was used. Depth and vertical movement activity data were right skewed and were corrected by square root transformation for statistical testing. Linear mixed models (LMM) were used with individual as random effect using the package nlme [49]. The long-term time series of depth and temperature data were expected to show significant autocorrelation. To correct for this, the dataset was reduced to a lower time resolution (per daily period), and a first order ARIMA term was added to the LMM’s:

4

5

in which is a time series of temperature or depth data, represents a fixed factor (one or more), the error term for individual , and residual error term which includes the auto regressive structure with autoregressive parameter and a residual noise term .

Linear models (formula 4 without the term) were used to test relationships between changes in pressure during the vertical movement and changes in pressure from before to after the vertical movement, and for testing speed data. These values were calculated for each vertical movement and aggregated per depth range and amplitude of the movement. In these LM’s, vertical migration mode was added as a factor. A weights factor was added to the LM’s to correct for the frequency of the aggregated data points. p-values of fitted models were evaluated based on likely-ratio tests. The package emmeans [50] was used for post hoc analysis of differences between groups after LMM or LM testing. The performance package was used for calculation of effect sizes for model parameters (Cohen’s f-partial).

Season highly correlated with variation in temperature (p < 0.05). Therefore, season was not added to the LMM. To compare diving patterns between seasons model marginal means were calculated for the mean temperature during summer and during winter.

Results

Recapture of wels catfish

Of the 42 wels catfish that were released, 35.7% (n = 15) were recaptured by professional fishermen (n = 9) and anglers (n = 6). The first wels catfish were released in August 2022 and the last wels catfish was recaptured in August 2024. The time between an individual release and recapture ranged between 6 and 662 days (median = 247 days), amounting to a total of 85,422 h of recorded behavioural data. From two of the data loggers no data could be retrieved. The distance between sites of release and recapture ranged from 0.1 km to 15.4 km (median = 1.6 km), with no significant relationship with time elapsed between release and recapture (n = 15, R = 0.117, p = 0.68). Five wels catfish were recaptured in the pelagic zone of Upper Lake Constance, in areas with a water depth exceeding 200 m (by professional fishermen using drift nets or by anglers trolling for pike Esox lucius). The other 10 individuals were caught near to the shoreline at depths between 0.2 and 40 m, by anglers or by professional fishermen using bottom gill nets and fyke nets.

Individual variation in depth and temperature patterns

The data loggers revealed substantial variation in vertical movement patterns among wels catfish, with six of the recorded wels catfish remaining in the top 10 m of the lake during recording and seven migrating to deeper parts of the lake (suppl. Fig 2). Figure 3 gives an example of these divergent patterns exhibited by two wels catfish released in October 2022: wels catfish #1 showed a shallow diving pattern and was recaptured from a shallow area at the river mouth of the Schussen river while wels catfish #2 showed a large amplitude in depth variation and was captured from the steeper slopes of the Bay of Bregenz. Both fish descended in winter and spring and ascended in summer with increasing body temperature (Fig. 3, higher panels). Summer swimming depth varied slightly (Fig. 3, middle panels), with fish #2 exhibiting a clear diurnal migration pattern with an ascending phase during dusk and a descending phase during dawn, while fish #1 that might have been more restricted in vertical movements on the shallow flat area, showed a less pronounced and inverted diurnal migration pattern (Fig. 3, lower panels). Related to this difference in migration patterns diurnal temperature variation was highest in fish #1. Both fish showed high variation in depth in the dark period.

Fig. 3.

Depth use (black line) and corresponding body temperature (blue line) of two data storage tags retrieved from European wels catfish released in autumn 2022 with different depth profiles. Above: long term profiles with resolution of 1 day; middle: daily diurnal patterns with resolution of 1 h in June; below: daily diurnal patterns with resolution of 10 min during some days in June

During summer, fish data loggers recorded a mean water depth of 4.1 m (n = 9, SD = 2.6 m, maximum depth = 34.8 m), and a mean temperature of 21.0 °C (n = 9, SD = 2.0 °C) (Fig. 3, suppl. Fig 1). In winter, the data loggers recorded a mean depth of 13.1 m (n = 12, SD = 8.8 m, maximum depth = 65.9 m), and a mean temperature of 6.2 °C (n = 12, SD = 0.44 °C).

Analysis of depth profiles

A LMM model was used to analyse the daily variation in depth data. This model showed significant autocorrelation (ϕ = 0.949, X² = 31188, p < 0.001). Controlled for autocorrelation, depth was closely related to logged temperature, recording about 0.65 m descent for each degree of cooling (F_1,14118 = 2360.8, p < 0.001, f = 0.41). Period of day had a significant effect on depth profiles (LMM: period: F_3,14118 = 11.1, p < 0.001, f = 0.049) and this effect showed a significant interaction with temperature (temperature × period: F_9,14118 = 57.8, p < 0.001, f = 0.11). Effect sizes for these results were very small (f < 0.2) which was likely due to widely differing variation in depth patterns between individuals (Fig. 3).

Post-hoc analysis showed that wels catfish habitually dived deeper during the day while ascending towards the surface from dusk to dawn at summer temperatures of 21.0 °C (Tukey-tests: day < dusk <dawn & night: p < 0.01; dawn ~ night: p = 0.21), while at winter temperatures of 6.2 °C no significant variations in daily vertical patterns were found (Tukey-tests, all comparisons p > 0.1).

Analysis of vertical movement activity

The LMM model for vertical movement activity, calculated as the average number of ascents and descents per hour, was corrected for significant autocorrelation (ϕ = 0.539, X² = 4771, p < 0.001). The model showed a significant increase of vertical movement activity with increasing temperature (F_1,14009 = 64,6, p < 0.001, suppl. Fig 2) and varied significantly with period of day (F_1,14009 = 135.2, p < 0.001).

Post-hoc analysis showed that vertical movement activity was lowest during daytime, and highest around dawn and at night at summer temperatures of 21.0 °C (Tukey-tests: day < dusk <dawn & night: p < 0.01; dawn ~ night: p = 0.21), while at winter temperatures of 6.2 °C vertical movement activity was lowest at dawn, intermediate at night and highest at day and dusk (Tukey-tests: dawn < night <day & dusk: p < 0.01; day ~ dusk: p = 0.48).

Depth profiles, speed and vertical migration mode

The data was analysed for vertical movements using a 0.5 min and a 10 min logging frequency. This resulted in two vertical migration modes: one dataset with relative fast vertical movements and one dataset with slower vertical migrations. The deviation in pressure (and thus depth) from the starting point of a vertical movement was significantly positively related to the pressure changes during the vertical movement (suppl. Table 2, Fig. 5, LM: F_1,55 = 233.9, p < 0.001, η² = 0.81). The vertical migration mode (fast vs. slow) had a significant effect on this deviation (Fig. 5, LM: mode: F_1,55 = 0.75, p = 0.39, η² = 0.01; interaction pressure changes and mode: F_1,55 = 143.4, p < 0.001, η² = 0.72). After a fast vertical movement (Fig. 4 A and B) the wels catfish returned to a depth close to the original starting point (Fig. 5: about 1% loss in pressure for every 10% pressure change in depth), regardless of whether the vertical movement began at 2 m or 20 m of depth (Fig. 4 A and B, Fig. 5 red line). In contrast, following a slow vertical movement, wels catfish stayed close to the pressure (and thus depth) they arrived at during the vertical movement (Fig. 4 C and D, Fig. 5 green line). Less than 0.5% of all observed movements exceeded a compression threshold of more than 30% of swimbladder volume (suppl. Table 2).

Fig. 5.

Relative change in pressure during the horizontal movement in relation to resulting long lasting relative pressure change in the period after the horizontal movement, in fast and slow descending and ascending wels catfish. Change was measured between 10 and 20 minutes before and after movement for fast moving wels catfish and 0.5–1.5 hour before and after movement for slow moving examples. Size of the symbols represents n-value (log). For all data see supplemental Table 2

Fig. 4.

Diving profiles for vertical wels catfish movements. A and C are fast descending and ascending movements, B and D are slow descents and ascents. Selected were episodes in which fish did not show an ascent just before descent or vice versa. Solid lines are changes of more than 3.33% − 13.33% in water pressure (equivalent to >1 m at 20 m depth), broken lines are changes of more than 13.33% in water pressure (equivalent to >4 m at 20 m depth)

In accordance to the method of calculation, the speed of vertical movements (suppl. Table 2) was lower in slow migration mode than in fast migration mode (LM: F_1,55 = 144.4, p < 0.001, η² = 0.72). Speed was positively related with depth (LM: F_1,55 = 14.0, p < 0.001, η² = 0.20) and this effect did not differ between modes (interaction effect: LM: F_1,55 = 3.9, p = 0.054, η² = 0.07). Speed varied from 0.0003 m s¹ in slow migration mode to 0.19 m s⁻¹ during fast migration mode (suppl. Table 2). Speed did not differ depending on direction of the vertical movement (addition to model LM: F_4,51 = 1.14, p = 0.35).

Buoyancy

Relative rates of change in the swimbladder were calculated assuming wels catfish were at neutral buoyancy at the start of a vertical movement. This assumption was made as it was found that fish habitually returned close to the starting depth after a fast vertical movement (Fig. 4 A, C, Fig. 5). Based on the formula for hydrodynamic lift (3) and assuming a generalised net fish body density of 1.08 kg L⁻¹ [14, 43], wels catfish would experience a force of gravity in freshwater (density of 1) equivalent to 0.726 * W (weight in kg). To overcome this force, a swimbladder volume of 0.0741 * W is required. When wels catfish descend to a depth resulting in 30% in pressure difference they would experience a force of gravity of 0.207 F * W (N kg⁻¹).

Discussion

This study investigated patterns in vertical movements, and how these might be limited by buoyancy control in free-ranging wels catfish (Silurus glanis) in an environment without relevant physical depth restrictions. Analysis of over 85,000 hours of logged data revealed a strong structuring influence of temperature and period of day on the general vertical distribution of this apex predator. Wels catfish occupied deeper water layers during winter and migrated to warmer shallow water layers in summer. Further additional patterns of diurnal vertical migration are most apparent in the warmer part of the year, with most vertical movement activity taking place around dusk and least during the day. Considerable variations in these patterns were found among the fish (suppl. Figure 1), which may be related to local environmental characteristics such as access to deeper areas. In winter, there is less diurnal variation in vertical movement activity. Two distinct vertical migration modes were found with different consequences for vertical migration: relatively fast vertical movements that rapidly return to the original depth, and slower migratory movements that result in substantial depth alterations. These yearly and daily diving patterns support a nocturnal lifestyle that is responsive to both seasonal temperature changes in the daily movements of prey and point to a well-developed capacity for buoyancy control in this species.

Vertical migration modes in wels catfish

This study analysed two separate vertical migration modes in vertical movements of European wels catfish: fast vertical movements of less than 5 min duration, and slower vertical movements that enacted over a period exceeding 15 min. The speed at which wels catfish moved during these migrations was low with maximum speeds lower than 0.5 m s⁻¹, which is in accordance with these fish being tracing predators that capture their prey by suction [12]. The difference between the vertical migration modes was that after the relative fast vertical movements fish returned close to their starting depth, and after slow vertical movements fish remained at the reached destination depth. For both vertical migration modes, in summer, wels catfish engaged most vertical movements at dusk, and least during the daytime, although this pattern changed in winter when vertical migration during the day relatively increased.

The fast vertical migration mode may serve multiple theoretical functions, including social interaction, predator evasion, and foraging, all of which may occur at all depths in the lake. However most vertical movements were recorded during summer, when wels catfish live in large, solitary home ranges [39, 51, 52], suggesting social interactions including territorial or aggressive behaviours, might play only a marginal role in the observed vertical movements. Furthermore, as apex predators themselves, adult wels catfish are unlikely to engage in extensive predator avoidance [24, 53]. It is therefore likely that fast movements are mostly associated with foraging activity, especially as they took place mainly at dusk and during the night, and wels catfish are strongly nocturnal feeders [54]. The increase in daytime vertical movement activity in winter is interesting, but given the low light levels in the deeper parts of the lake this could also be regarded as pseudo-nocturnal vertical movement activity. Finally, the observation that fish return close to their initial depth after a fast migration mode suggests a hunting strategy in which individual wels catfish scan for prey from a neutrally buoyant starting point, leave neutral buoyancy to attack, and then return to neutral buoyancy again, thereby minimising physiological stress. To our knowledge, these relative fast excursions out and back from a neutrally buoyant origin and end point within their intrinsic boundaries have not been described before.

Slow vertical migration mode was likely used to migrate between layers in the lake. Such movements spanned at least 15 min and wels catfish subsequently remained at the destination depth for an extended period. Moreover, relatively more ascents were observed at dusk and more descents at dawn, a crepuscular pattern that echoes that of nocturnal vertical migration. The fact that the fish remained at the destination depth for at least an hour, strongly suggests its function is to migrate between layers in the lake. A search of the literature suggests that full physiological adjustment to a new depth might be expected to take several hours to days [2, 16], but these studies refer to much higher pressure changes or to adjustments following total ablation of the swimbladder. Thus the 15 minutes period observed here to compensate a relatively minor pressure change (<30%) may well correspond to the time required for buoyancy adjustment via gas exchange in the swimbladder. In contrast to the vertical movements during fast vertical migration mode, these vertical movements during slow migration mode may therefore incur no extra energetic input to counteract positive or negative buoyancy.

Wels catfish are exceptionally well adapted to nocturnal hunting. They have keen senses of smell and electroreception with which to detect and track in darkness the hydrodynamic and chemical traces left by prey [12, 55–58]. The time elapsed between tracking prey and attack can be very short, from just a few seconds to minutes [55]. Most vertical movement activity was recorded in summer at relatively shallow depths in the lake. However, wels catfish were also shown to make fast vertical movements over large depth ranges, i.e. diving from around 22 m to more than 35 m or ascending from the same starting depth to around 11 m. This raises the question whether it is possible that prey wakes, either olfactory or mechanoreceptive [59], in lake with few currents to disrupt them, linger long enough so that such trails may be detected and exploited by foraging wels catfish. It is well known from other large lakes that wels catfish do feed in the pelagic zone [60–62].

Diurnal variation in vertical movement activity

While the fast vertical migration mode identified in this study may be functionally related to tracking prey and foraging behaviour, the slow vertical migration mode may be concerned with transitions between periods of foraging and inactivity. During night, wels catfish [34] hunt in the upper layers of the pelagic zone for prey fish such as bleak Alburnus alburnus [63], roach Rutilus rutilus [64] and others [65], which themselves feed on zooplankton in the upper layers [66, 67]. Thus, it is highly likely that nocturnal feeding habit of wels catfish [54] and their movement into the upper layers of the pelagic zone are driven by prey availability. The diving profiles of single fish suggest that after nocturnal feeding some wels catfish descend to deeper water for the ensuing daylight hours. Such crepuscular migrations likely help conserve energy via lowering metabolism. Moreover, our data shows that wels catfish show significantly more vertical movement activity during dusk and nighttime than during day, except in winter when the daytime vertical movement activity at greater depth increases. Similar diurnal migration patterns have been described before for one wels catfish population in a cold lake in northern Sweden [23] and for wels catfish populations in reservoirs in the Czech Republic [68].

Seasonal variation in vertical movements

During the winter, vertical movement activity was strongly decreased and fish were found in deeper layers in the lake, in comparison to summer. This study chimes with previous reports of lake wels catfish migrating to deeper layers in the colder winter season [24, 60, 68]. A similar seasonal pattern of migration, with greater vertical movement activity from late spring to early autumn and lowest vertical movement activity during winter is seen in studies investigating movement patterns in habitats that are both colder [23] and warmer [24, 25, 68] than Lake Constance. These patterns may be influenced by foraging opportunities but also by behavioral factors like for example spawning which takes place in shallow waters in late spring and summer [69, 70]. The patterns from the loggers did not provide information about the location of the individual in the lake. Some of tagged individuals were recaptured more than 5 km away from the shore and in areas ranging between 150 and 250 m deep. In winter anglers have observed large, motionless wels catfish in deeper water layers during daytime hours (Baer pers. comm.). Similar observations of motionless wels catfish have been reported from reservoirs up to 75 m deep in the Czech Republic (Milan Říha, personnel communication), and it has been found that in winter wels catfish aggregate in large social groups in deeper layers [71, 72]. The function of this behaviour in the pelagic zone of large lakes remains to be studied.

Buoyancy control

The analysis of vertical movements suggests that wels catfish can make rapid vertical movements to significantly different depths, which may play a role in approaching prey swimming several meters below or above their starting position. The resulting positive or negative changes in hydrostatic pressure are of the order of 0–30%, leading to a proportional shrinkage or expansion of the swimbladder. As adjustment of swimbladder gas volume is a slow process potentially requiring hours to correct for pressure changes of one bar [16], vertical movements incurring a change of 0.3 bar but lasting only a few minutes are likely too short to allow swimbladder volume to adjust. Observations have shown that wels catfish swim at 0.17 body lengths per second at night [27], and this may be a suitable cruising speed for sensing and tracking the wakes of prey [55]. However, lift produced by swimming to correct for loss of neutral buoyancy is limited at such slow swimming speed and this might be limiting the amplitude of vertical movements. Therefore, a quick return to the starting depth and reestablishment of homeostatic neutral buoyancy is likely an energy saving measure in wels catfish.

The pressure sensitivity of fish is mediated via the vagus nerve, enabling them to detect and respond behaviourally to changes in the order of 0.5 to 10% [16, 20]. The current study suggests that wels catfish limit their amplitude of vertical movements to a range corresponding to a 1 to 30% change in pressure. Following Boyle-Mariotte’s law this amounts to a larger vertical movement amplitude at deeper layers than in shallow water [1, 5]. Speed during vertical movements was positively related with increasing depth, which further hint to pressure regulation of vertical migrations. Assuming a standard fish body density of 1.08 kg L⁻¹ [14, 43], the species is expected to achieve neutral buoyancy with a swimbladder volume that corresponding to 0.0741 of their weight in kg. This theory accords well with the reported range of swimbladder volumes for the channel catfish, which corresponds to 5.8 - 9.3% of body weight [4]. An average sized wels catfish with a weight of 5 kg may thus be expected to have a swimbladder volume of ±0.370 L at neutral buoyancy. This fish would experience a force of gravity of about 1 N at the limit of a descent (30% decrease in swimbladder volume). The extent to which swimming against this force might impair the species ability to orientate and navigate while tracking prey remains to be investigated. Adjusting buoyancy to compensate for a 30% change in pressure may keep this gravitational force within tolerance such that depth can be maintained by swimming at cruising speed. Overshooting this limit would require the fish to take more energetically costly action to swim against the increasing gravitational force [1].

Relevance for expansion success

The wels catfish has invaded several aquatic ecosystems in southern Europe [24, 61, 73]. Global warming may play a major role in this expansion as the upper layers of large, traditionally cool alpine water bodies, such as Lake Constance, warm up sufficiently in summer to meet the metabolic requirements of the species. The results show that the wels catfish movements effectively utilize this temperature-enabled habitat expansion by adhering closely to seasonal temperature changes, and by exhibiting variation in nocturnal vertical movement activity, likely as adaptation to hunting during different seasonal hydrological and temperature conditions. By limiting the depth of vertical movements, they are able to maintain control of buoyancy and can explore and hunt the pelagic of a deep lake at low metabolic cost. Their nocturnal life style and their apex position in the food web may further facilitate the species expansion [32, 68].

Conclusions

Catfish show clear seasonal shifts in depth (deep in winter, shallow in summer). Vertical movement activity was highest in summer, especially around dawn and dusk. Two modes of vertical movement were identified (fast vs. slow), but their behavioral functions remain uncertain. Wels catfish appeared to limit fast vertical movements within a restricted range of water pressure change and then to return quickly to the original depth – a strategy that has not, to our knowledge, been described before. The slow vertical migration mode was found to result in more prolonged changes in swimming depth suggesting that wels catfish can correct for modest changes in pressure (buoyancy) over a relatively short period of time (<30 minutes). These aptitudes for buoyancy control of wels catfish may optimize their competitiveness and ability to expand, exploit and establish in deep waterbodies as these warm under climate change.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

A.B., J.G. and M.F. made substantial contributions to the conception and design of the work. M.F and J.B to the field work and acquisition of data. A.R, J.B. analysed and wrote the main manuscript text and prepared figures. All authors reviewed the manuscript.

Funding

This project was funded by the Fischereiabgabe Baden-Württemberg.

Data availability

Data are available on Figshare: 10.6084/m9.figshare0.29430329.

Declarations

Ethics approval

All fish were caught by licenced personnel, by permission of the local fisheries administration (Regierungspräsidium Karlsruhe, Freiburg, Stuttgart, Tübingen) and according to the German Animal Protection Law (§ 4) and the experimental procedures were approved by the local authorities (Referat Tierschutz of Regierungspräsidium Tübingen: 1/21; AZ 0350–9185–19/2/7). All recaptured wels catfish sampled by the authors were euthanized with an overdose of clove oil (1 mL L⁻¹) and a gill cut and according to the ordinance on slaughter and killing of animals (Tierschutzschlachtverordnung § 13).

Competing interests

The authors declare no competing interests.