Living in the edge: the fate of individually marked pike (Esox lucius) stocked in a hyper-eutrophic lake with frequent winter hypoxia

ABSTRACT

Eutrophication of lakes has increased the risk of fish kills during winter under ice cover. If such a lake is isolated, the fish assemblage consists of only oxygen-tolerant species without predatory species. Here, our main aim was to analyze the fate of 99 marked pike (Esox lucius) stocked to a pikeless lake that is known to have low oxygen concentrations under ice cover. These pike (total lengths 33.5–76.9 cm; weights 260–3050 g) were stocked in May 2008 and were followed 7 years. The first marked pike was caught in August 2008 and the last one in September 2012. Between these dates, 21 pike were caught at least once, five pike were caught twice, and two pike were caught three times. Relative condition was based on the length–weight relationship. Both length and weight affected significantly the recapture probability, whereas relative condition had no effect. The growth of marked pike was estimated based on length or weight increments in two separate models with either starting length or weight together with the number of days between observations. In these models, the number of days had positive, and size, either length or weight, had a negative relationship with growth. According to marginal r², length and the number of days explained 65.5% of variation in length increments, whereas the marginal r² for weight and the number of days was 49.3% for weight increments. The length increment model was used to compare pike growth in Lake Savijärvi with other Finnish lakes. The comparison showed that during the first four growing seasons after stocking, the pike growth in Lake Savijärvi was faster than in the other lakes studied.

KEYWORDS:

• Recapture probability
• growth rate
• growth model
• Esox lucius
• winter fish kill

Introduction

Eutrophication is one of the most critical problems in temperate freshwaters. Along with increasing nutrient concentrations, the biomass of cyprinid fish species increases (Olin et al. 2002), which is usually an unwanted change in fish assemblage (Tammi et al. 1999). In temperate eutrophic lakes, wintertime oxygen depletion under ice cover can induce dramatic changes in fish assemblage: species demanding high oxygen concentrations disappear, while hypoxia-tolerant species remain and may fill all available niches of the lake (Tonn et al. 1990; Holopainen et al. 1997). In connected lakes, the recovery of fish assemblage starts immediately after hypoxia due to immigration and the lake returns gradually to the earlier species structure and level of eutrophication (Ruuhijärvi et al. 2010). In isolated lakes, however, the fish assemblage may consist of small-bodied species such as crucian carp (Carassius carassius) and lacking typical predators like perch (Perca fluviatilis) and pike (Esox lucius; Öhman et al. 2006). Thus, no top-down control of these cyprinid species exists, which can intensify zooplanktivory and bioturbation of sediment causing cyanobacterial blooms (Horppila & Kairesalo 1990; Scheffer 1991; Zambrano et al. 2005).

Pike is a common circumpolar species in the Northern hemisphere occurring in small ponds to large lakes, in slowly running freshwaters, and in brackish waters (Crossman 1996). Pike is tolerant of low oxygen concentrations down to 0.3 mg/L (Raat 1988). Hence, it is one of the most suitable predatory species to be stocked in eutrophic lakes that suffer from wintertime hypoxia (Petrosky & Magnuson 1973). Pike stocking is also implemented in biomanipulation projects which target either to reduce unwanted fish species or to improve water quality (Berg et al. 1997; Skov & Nilsson 2007). Besides, due to its importance both in commercial and recreational fisheries, it is regularly stocked both to increase and to stabilize catches (Raat 1988). In general, pike stocking is done with larvae and juveniles; older individuals are not used probably because handling and transporting are more difficult, necessitating a lower number of individuals than when stocking juveniles. In studies where larger pike have been stocked, the focus has been in other, more specialized topics, such as movement (Vehanen et al. 2006), diets in previously fishless lakes (Venturelli & Tonn 2006), or possible avoidance of duck species (mallard Anas platyrhynchos, teal Anas crecca, and goldeneye Bucephala clangula) due to increased predation risk by introduced mature pike (Dessborn et al. 2011). Several studies with different species including 0+ pike have shown that the size at stocking is positively correlated with the survival, especially if stocked juveniles are larger than naturally occurring juveniles (Grønkjær et al. 2004; Skov & Nilsson 2007). In older pike (age >1), the size may be of less importance but this has not been studied.

The main aim of the present study was to follow the fate of the adult, marked pike stocked in Lake Savijärvi. At the time of stocking, there were no pike in Lake Savijärvi based on a 5-year removal-seining targeting crucian carp. Other observed fish species in the lake were roach (Rutilus rutilus) and tench (Tinca tinca). The dominance of crucian carp and tench reflects the fact that regular anoxia occurs during long winters (Tonn et al. 1990; Lappalainen et al. 2013). According to an older report, however, there have been pike in the lake (Helsingin vesipiirin vesitoimisto 1986). The main aim included two parts: first, the recapture probability was analyzed in relation to the size and condition of pike at stocking and, second, the growth of marked pike was evaluated and then compared with growth of pike in other Finnish lakes.

Methods

Lake Savijärvi (60° 21′ N, 25° 20′ E) is a small (40 ha) and shallow (mean depth of 1.6 m and maximum depth of 2.6 m) lake in southern Finland. The lake is turbid (from clay) and highly eutrophic. The Secchi disk depth varied between 20 and 60 cm during June–August (n = 28) in 2008–2014, while at the same time the mean total phosphorus (µg/L) and chlorophyll-a concentrations (µg/L) were 169 (± 96 SD, n = 28) and 125 (± 100 SD, n = 28), respectively.

As a habitat for pike, Lake Savijärvi is challenging because under the ice cover, water can become anoxic, and during summer, the whole water column can warm up to 25 °C. During the period when marked pike were caught (2008–2012), oxygen concentrations were low in late winter, but not low enough to cause substantial mortality. However, severe hypoxia during the winter 2012–2013 (Table 1) caused substantial pike mortality. The reason for the high mortality was that the ice cover lasted for almost a month longer than, for example, in the previous winter 2011–2012.

Table 1. Mean oxygen concentrations (mg/L; n = 1–4) in three depth layers with ice cover thickness (m) at each sampling date in Lake Savijärvi in southern Finland.

	Year
	2009	2010	2011	2012	2013	2014
Sampling date	26 March	27 March	17 March	12 March	6 March	14 March
Ice thickness (m)	0.5	0.45	0.5	0.4	0.5	0.13
Oxygen (mg/L)
Depth ≤ 1 m	4.7	5.7	3.6	2.0	2.5	9.1
Depth 1–1.9 m	1.6	0.8	1.6	0.0	0.0	0.5
Depth ≥ 2 m	0.7	0.6	0.0	0.0	0.0	nd

nd = no data.

The adult and spawned pike were stocked in May 2008 and were marked with T-bar anchor tags (n = 88) or Carlin tags (n = 11). Most of the T-bar tagged pike came from two lakes (n = 78), whereas those marked with Carlin tags (n = 11) and the other T-bar anchor tags (n = 10) came from four different lakes and were tagged earlier (Carlin tagging in 2006 and T-bar anchor tagging in 2007; Kuparinen et al. 2012). The possible difference between tagging methods in recaptures was tested with Fisher's exact test. Besides marked adult pike, newly hatched pike larvae (11,000) and juveniles (230) were stocked but these were not tagged. Pike sampling was done by removal seining (3–6 hauls per year during August and September). The seine was 7 m high, 295 m long and had 6 mm mesh size (from knot to knot) in the cod end. The area covered by each seine haul ranged between 3 and 18 hectares. About 25 hectares (62.5%) of the surface area of Lake Savijärvi is suitable for seining. When a marked pike was caught, it was weighed, the total length was measured, and scales for age determination were taken before returning it to the lake.

Logistic regression models were used to estimate whether the size at stocking, as length or weight, combined with relative condition was connected with the recapture probability. Because lengths and weights were strongly correlated together (r_s = 0.984, p < 0.0001, n = 98) but relative condition was not correlated with either length or weight at stocking (length: r_s = 0.038, p = 0.713, n = 98; weight: r_s = 0.162, p = 0.111, n = 98), the size effect was estimated using two separate models that employed either length or weight with relative condition. Relative condition was estimated based on residuals from the length–weight equation and was calculated as real weight minus the estimate based on the length–weight-relationship (e.g. Fechhelm et al. 1996; Lappalainen et al. 2005b; Gascho Landis et al. 2011; Milardi et al. 2014).

A stepwise backward elimination was used in the two logistic regressions and selection was based on the AIC value. In the elimination, a variable is dropped if the χ² falls below twice its degrees of freedom (Harrell 2014). The performance of two logistic regression models was evaluated first by the validate option with 1000 bootstrapping with the rms-package (Harrell 2014) in R (version 3.1.2; R Development Core Team 2014). Two different fit indexes were used. First, Somers’ D_xy rank correlation between predicted probabilities and observed responses was estimated. D_xy is the difference between concordance and discordance probabilities. When D_xy = 0, the model is making random predictions and when D_xy = 1, the predictions are perfectly discriminating. Second, the generalized r²_N-index of Nagelkerke was estimated. This has the same general interpretation as the usual r² but it is more appropriate for a binary logistic regression model (Harrell 2014). In addition, the le Cessie–van Houwelingen–Copas–Hosmer unweighted sum of squares test for global goodness of fit was evaluated (Hosmer et al. 1997) using the rms-package.

Scale-based ageing of marked pike was challenging because of several false annuli (Neumann et al. 1994). Therefore, the growth of marked pike was based on size increments. The increments were estimated based on the starting size of pike, both in length and weight, and the time passed (in days) between two successive observations. Some marked pike were caught only once, but because there were also pike that were caught up to three times (Supplemental Table 1), the marking code was used as a subject in modelling to take into account possible autocorrelation (Dorn 1992; Lappalainen et al. 2005a). The data were fitted by linear mixed-effects model using length (model 1) or weight (model 2) as a dependent variable and the marking code as a subject (lme4-package in R; Bates et al. 2014). The R equation was as follows:where size was either weight (g) or length (cm) and Day_no was the number of days between observations.

The method suggested by Nakagawa and Schielzeth (2013) was used to compare the model fits between length and weight increments. They suggested the use of marginal r², which is associated with fixed effects, here the size and number of days, and the conditional r², which is associated with both the fixed effects and the random effects, the latter being the marking code in our case. Thus, the marginal r² gives the estimate of how much variation the fixed variables describe of the size variations. These analyses were done with MuMIn-package in R (Barton 2014).

The growth of marked pike in Lake Savijärvi was compared with that in other Finnish lakes. The obtained model (1) was used to calculate three annual length increments (365 days each). The starting length chosen was 43 cm and then one annual length increment, based on the growth model (1), was added to the starting length, and so on for the two subsequent years. This procedure enabled the comparison of three years with other lakes without ageing the pike in Lake Savijärvi. The starting length of 43 cm was selected because (1) the starting lengths of marked pike (33–77 cm) at stocking and those of fish recaptured two or three times (53–77.5 cm) were close to the length range (43–73.4 cm) estimated based on model 1 and (2) pike are generally mature at that length (e.g. Kotakorpi et al. 2013). The other Finnish pike lakes were Alinen Mustajärvi (Rask & Arvola 1985), Pohjois-Päijänne (Alaja 2008), Ontojärvi and Lentua (Heikinheimo & Korhonen 1996), and Kemijärvi (Heikinheimo-Schmid & Huusko 1987). In lakes Ontojärvi, Lentua, and Alinen Mustajärvi, length-at-age shown were originally presented separately for both sexes, and because pike were not sexed in Lake Savijärvi, an average length-at-age was calculated for these lakes taking into account the number of specimens of each gender.

Results

Recapture probability of marked pike

As shown in Figure 1, the length–weight relationship of pike at stocking was

Figure 1. Length–weight-relationship of marked adult pike stocked to Lake Savijärvi in May 2008.

Of the 99 stocked and marked pike, no recaptures were made from 78 pike. Their average length and weight were 46.4 cm (±8.75 SD, range 36–76 cm) and 0.688 kg (±0.485 SD, range 0.26–3.05 kg), respectively. The first marked pike was caught in August 2008 and the last one in September 2012. Between these dates, 21 pike were caught at least once, five were caught twice, and two were caught three times. According to a Fisher's exact test, recapture rates did not differ between tagging methods (Carlin: 3 recaptures from 11 tagged; T-bar: 18 recaptures from 88 tagged, p was not significant). Therefore, all the data were pooled.

In logistic regression with backward selection of significant variables with pike caught after stocking (0/1) as dependent and condition, either with length or weight as independent and continuous variables, condition was removed from both models (Table 2). Size of pike was significant in both models (length: χ² = 4.90, p = 0.0269; weight: χ² = 4.15, p = 0.0416, Figure 2) but the model with length (D_xy = 0.323, r² = 0.0618) fit better than that with weight (D_xy = 0.314, r² = 0.0512) based on 1000 bootstrapping. According to global goodness of fit, both final models showed good fit (length: Z = 0.487, p = 0.626; weight: Z = 0.877, p = 0.381).

Table 2. Recapture probability of marked adult pike according to logistic regression analyses showing coefficients (Coef) for length (cm TL) model and weight (g) model with standard errors (SE), Wald test statistic, and the significance level (p).

	Coef	SE	Wald Z	p
Model: caught (0/1) = length
Intercept	−3.942	1.243	−3.170	0.0015
Length	0.056	0.025	2.221	0.0272
Model: caught (0/1) = weight
Intercept	−1.930	0.413	−4.680	<0.0001
Weight	0.827	0.406	2.040	0.0418

Figure 2. Probability of recapture in relation to pike stocking length and weight estimated with logistic regression (markers are jittered around 0 and 1).

Growth

In the growth models based on length or weight increments, the starting size and the number of days were both significant (Table 3). In models, increments increased as a function of number of days and decreased with size (length and weight, Figure 3). According to marginal r², starting length and the number of days explained 65.5% of the variation in the length increments, whereas the same marginal r² for starting weight and the number of days was 49.3% for the weight increments.

Table 3. Results for linear mixed-effects model for pike growth using length (model 1) or weight (model 2) as dependent variable and day number and starting length or weight as independent variables with standard errors (SE).

	Estimate	SE	df	t-Value	p
Model 1: length increment
(Intercept)	33.25	3.792	20	8.766	<0.0001
Day number	0.0182	0.002	7	8.304	0.0004
Starting length	−0.5159	0.048	7	−10.607	<0.0001
Model 2: weight increment
(Intercept)	788.4	324.1	20	2.432	0.0245
Day number	1.861	0.331	7	5.625	0.0008
Starting weight	−0.3407	0.101	7	−3.361	0.0121

Marking code was used as a subject. Day number: number of days between stocking and first recapture or between subsequent recaptures. Starting length: length (cm) at stocking or at previous recapture. Starting weight: weight (g) at stocking or at previous recapture.

Figure 3. Model surfaces for pike weight (a) and length (d) increments with corresponding residuals for independent variables weight (b and c) or length (e and f). Day is the number of days between stocking and each recapture.

The length increment estimated with the growth model (Table 3 and Figure 3) was 17.7 cm for the starting length of 43.0 cm (during the first 365 days), 8.6 cm for 60.7-cm pike, and finally 4.1 cm for 69.3-cm pike. The growth of pike in Lake Savijärvi was substantially faster than those found in other Finnish lakes (Table 4).

Table 4. Pike growth in Lake Savijärvi and in lakes Alinen Mustajärvi (Rask & Arvola 1985), Pohjois-Päijänne (Alaja 2008), Ontojärvi and Lentua (Heikinheimo & Korhonen 1996), and Kemijärvi (Heikinheimo-Schmid & Huusko 1987).

	Lake Savijärvi	Lake Alinen Mustajärvi	Lake Pohjois-Päijänne	Lake Ontojärvi	Lake Lentua	Lake Kemijärvi
Start	43.0	44.0 (age 5)	42.7 (age 4)	44.0 (age 4)	43.5 (age 5)	43.0 (age 4)
Year 1	60.7	48.2	50.2	50.4	49.5	49.0
Year 2	69.3	nd	60.5	57.9	nd	55.0
Year 3	73.4	nd	67.6	68.2	nd	59.0

In Lake Savijärvi, length increments (in cm TL) were estimated for three 365-day periods and each increment was added to previous starting length. In parenthesis is pike age at the starting length in other lakes. nd = no data.

Discussion

The results showed that the size of marked pike at stocking was positively related to subsequent recapture probability. Of the size variables, length showed better fit than weight, whereas condition was not significant. The difference between length and weight is probably due to larger variability in weights due to seasonal differences along with possible stomach contents. Size has been found to be a very important factor for survival when stocking juvenile pike and, typically, mortality has been much lower within larger juveniles (Grønkjær et al. 2004; Cucherousset et al. 2007). Two main differences between our study and those studies where marked juveniles were stocked are: all of our marked pike were over 35 cm and their number was low – only 99 individuals. However, as the logistic regression showed, size was a significant predictor but not as important as in studies when marked juveniles are stocked. Lorenzen (1996) showed that, in general, the increase in weight at juvenile stage (weight 1–100 g) increased annual survival rapidly at first, but later at adult stage (weight > 500 g), the increase decelerates. In Grønkjær et al. (2004) study, the growth in length from 2.2 to 3.1 cm increased the survival of juvenile pike by a factor of 3.3, whereas in our study, the increase in length at stocking from 33 to 77 cm increased recapture probability by a factor of 2.4.

The number of recaptured individual pike during the whole study period (21) from the 99 adult pike stocked yielded a recapture rate of 21%. In other studies, the same percentage has varied between 5.6% and 76.2% (Kipling & Le Cren 1984; Clark 1990; Kekäläinen et al. 2008; Koch & Steffensen 2013). In Lake Erie, 2659 adult northern pike (36.4–89 cm) were tagged over an eight-year period and, of these, 698 were recaptured one to seven years later (Clark 1990). In another lake in Kansas, a total of 110 northern pike were tagged (56–94 cm) between February and March 2011 and 16 were recaptured during that period (Koch & Steffensen 2013). In a northern Baltic river, of 142 marked pike (40.5–105 cm), eight individuals were recaptured within one month (Kekäläinen et al. 2008). The highest percentage was observed in Lake Windermere (Kipling & Le Cren 1984), where 1431 pike were marked during the years 1954–1973 and 1091 pike were recaptured, one individual 12 years after marking. More precise comparison among these studies is difficult, because both the time frame and the number of pike marked are quite different. In most cases, where adult pike have been marked, the focus has been on movements or on behavioural aspects (Jepsen et al. 2001; Vehanen et al. 2006; Kekäläinen et al. 2008).

Factors affecting the number of recaptured pike could be connected with the lake, the number of stocked pike, marking methods, and recapturing effort and methods (Kipling & Le Cren 1984; Kuparinen et al. 2012). Because no pike were caught during five years of seining before the stocking described here, the lake can be judged to be isolated. There is one ditch outflow from which marked pike could have left the lake but immigration seems less probable. Tag loss may have reduced recaptures (Gurtin et al. 1999) but that was not evaluated directly. During the first two years after stocking, all large pike (>1 kg) caught (n = 10) with seine were found with tags, suggesting low tag loss (Pierce & Tomcko 1993). The sampling was done with seine, which operates in the pelagic areas of the lake and, therefore, pike staying in the littoral zone will not be caught. Originally negligible recreational pike fishing increased towards the end of the study period and hence some of the marked pike might have been caught by recreational fishermen. However, because recreational fishing licences in Finland are sold for large areas that include hundreds of lakes, it is not possible to evaluate the increase in fishing effort or catches in certain lakes based on fishing licences sold.

The last observation of marked pike was made in 2012, when a 5.7-kg pike was caught for the first time. In the following two years, no marked pikes were caught. The winter of 2012–2013 was long and the whole water column was nearly anoxic in March. In the spring of 2013, local environmental officers collected 450 kg of decomposing pike from the lake shores, which is more than 10 kg per hectare. Most of these dead pike were large (1.5–3 kg), confirming earlier findings that large pike are more vulnerable to low oxygen levels than smaller ones (Casselman & Harvey 1975). Unfortunately, no special attention was paid to check if marked pike were among the dead pike. Also, some smaller pike died (<1.5 kg), because these were seen on the lake shore, but their number is not known. In the following autumn, both juveniles and older pike were caught with seine (length range 10–60 cm, five individuals over 1 kg with a maximum of 1.6 kg) confirming that some pike survived over the previous winter.

The growth of marked pike was fast, especially during the first years in Lake Savijärvi, and it surpassed growth in other lakes in Finland (Rask & Arvola 1985; Heikinheimo-Schmid & Huusko 1987; Heikinheimo & Korhonen 1996; Alaja 2008). At the beginning of the study period in 2008, crucian carp catches per seine hectare was 243 kg. Thus, food resources for the stocked pike were extremely abundant. Pike were the only piscivorous species in the lake and during the first years after stocking, the intraspecific competition was low. However, the number of pike started to increase, because 0+ juveniles were found in the autumn of 2009 revealing that the marked pike started to reproduce during the next spring after stocking. These juveniles grew rapidly (Lappalainen et al. 2013). Since then, 0+ and older pike were always found in autumn seining. With the increasing pike population in Lake Savijärvi, the exceptionally high growth rate started to stabilize towards the average found in other lakes (Lappalainen et al. 2013).

Pike marking combined with the growth model enabled the estimation of growth without actually ageing the pike. This was very useful, because ageing from scales was hampered by several false annuli found in the scales (Neumann et al. 1994). In general, pike ageing is done from cleithrum bones (Neumann et al. 1994; Faust et al. 2013), but these could not be used because all pike were returned to the lake after measurement. Formation of false annuli in pike scales has been connected with the challenging environment (Gerdeaux & Dufour 2012). In Lake Savijärvi, the mass occurrence of blue-green algae is typical during summer months (Vinni 2005) and could negatively affect the visual predation of pike (Casselman 1996), while water temperature can reach 25 °C and more in the whole water column (Lappalainen et al. 2013). Though the lethal temperature for pike is as high as 28 °C–29 °C, the optimum temperature in adult pike is only 19 °C–21 °C (Casselman 1978). Bevelhimer et al. (1985) gives higher values for lethal (34 °C) and optimum (24 °C) temperatures. They suggested that this was due to adaptation to warm climate in Ohio (∼40° N). For the same reason, suggesting adaptation to colder climate, Heikinheimo and Korhonen (1996) used lower lethal (24 °C) and optimum (17 °C) temperatures in a bioenergetics model applied to pike in two Finnish lakes (∼64° N).

To conclude, the stocking of adult pike was successful, because a self-reproducing population was established. The size was a positive predictor of recapture probability for stocked mature pike, but was not as important as for stocked juveniles. Although the lake was a very challenging habitat for pike, the growth was rapid during the first years after stocking and surpassed those found in other lakes in Finland.

Acknowledgments

We thank Petri Savola and Miska Etholén and their teams for letting us analyze the pike caught during the removal seining.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Maa-ja Vesitekniikan tuki ry [grant number 15312].

Notes on contributors

Jyrki Lappalainen

Jyrki Lappalainen gained his Ph. D. degree at the University of Helsinki in 2001. He is a researcher and University Lecturer at the Department of Environmental Sciences. His research interest focuses on environmental changes and its effects on fish, fisheries and aquatic ecosystems.

Mika Vinni

Mika Vinni is a consultant in aquatic and environmental topics. His research interest is on lake fish, their growth and diet, and on lake restoration.

Tommi Malinen

Tommi Malinen gained his M. Sc. degree at the University of Helsinki in 1995. He is a researcher at the Department of Environmental Sciences and his research focuses on fish stock assessment and food-web interactions