Variation in Gill-Net and Angling Catchability with Changing Density of Northern Pike in a Small Minnesota Lake

Introduction

Knowledge of catchability and how it varies is critical to interpreting fish catches. Yet catchability coefficients (q) have seldom been determined because of the difficulty of obtaining valid density estimates for fish. The simplest theoretical form of the relationship between catch per unit effort (CPUE) and fish abundance (N) is

where q is the proportion of the stock removed by one unit of effort (Richards and Schnute 1986). If catchability is constant, then CPUE is a reliable index of relative abundance. Such a proportional relationship was found for quillback rockfish Sebastes maliger in the Strait of Georgia, British Columbia (Richards and Schnute 1986).

Other studies, however, have reported that catchability changes with changes in fish abundance, thus requiring a nonlinear model of catchability such as the power curve

where β is a shape parameter. When β < 1, catch rates can remain high, or hyperstable (Hilborn and Walters 1992; Harley et al. 2001), even as fish abundance drops. In a commercial gill-net fishery in the Connecticut River, Crecco and Savoy (1985) found increases in catchability with decreased abundance of American shad Alosa sapidissima. Hyperstability is often attributed to operational characteristics of the fishery that increase its efficiency in the face of lower fish densities. In contrast, catch rates can also drop faster than abundance (β > 1). Such hyperdepletion has been attributed to animal behavior or nonuniform distributions of animals (Hilborn and Walters 1992). In a study of Pacific salmon Oncorhynchus spp., Shardlow (1993) reported increases in catchability with an increasing density of salmon caught by trolling lures and attributed the changing catchability to fish behavioral interactions.

Gill-net and angler creel surveys have historically been used to assess fisheries in Minnesota. Since 1941, one of the primary techniques for sampling fish during lake surveys has been with standardized multimesh experimental gill-net sets (Moyle 1950; Pierce et al. 1994). Gill-net locations in most Minnesota lakes have been fixed over time. In general, nets are set on the bottom, often above the thermocline, in water depths with sufficient dissolved oxygen concentrations to support fish. The nets are usually raised in the morning after they have been fished about 24 h and are only fished once in each location. Angler creel surveys have also been used during the same period to sample the recreational fishery. In both types of survey, mean CPUE has been used as a measure of the relative abundance of fish, an approach that is based on the intuitive notion that higher catch rates indicate higher fish densities. One of the most important uses has been to track population trends over time within individual lakes. In spite of the importance attached to gill-net and angling CPUE, the relationship between CPUE and the actual population density of the northern pike Esox lucius, an important sport fish in Minnesota, has never been explored. The purpose of our study was to determine how closely gill-net and angling catch rates tracked the changing density of northern pike within a single lake and to examine how catchability varied with the density of the population.

Methods

The relationship between CPUE and northern pike population density was examined in a small north-central Minnesota lake. We used population estimation methods that were validated by Pierce (1997) to obtain fish density estimates. Variations in catchability coefficients for both gill netting and angling were then monitored while manipulating the density of the stock through a series of nettings that removed nearly one-half of the northern pike population.

Northern pike were trap-netted and tagged during ice-out and spawning in Camerton Lake, Minnesota. Camerton Lake is a 28.3-ha lake located on private property in Itasca County (47°29′N, 93°25′W). With a maximum depth of 3.0 m, the lake is isothermic, with wind mixing the entire water column. The lake is surrounded by deciduous forest and bog and has low total alkalinity (3 mg CaCO₃/L). Trap-netting occurred during 6–15 April 1998 and entailed a total of 87 net-days of effort (up to 12 nets/d at 14 sites). The trap nets were 19-mm-bar mesh and had two throats, a 0.9-m × 1.8-m rectangular frame opening into the trap, and 12-m leads that extended to shore. Trapped northern pike were measured for total length (mm) and tagged with a numbered Floy tag and an unnumbered Dennison tag. Tags were inserted on opposite sides of the anterior base of the dorsal fin (Pierce and Tomcko 1993). Scales for estimating fish age were sampled from five individuals of each sex per 25-mm length-class.

A series of removals of northern pike was used to artificially manipulate the population size. Removals were by gill netting and angling during 15 June–30 July 1998 when water temperatures ranged from 21.0°C to 25.7°C. The experimental gill nets were 76 m long and consisted of five 15.2-m × 1.8-m panels with graded-bar-mesh sizes of 19, 25, 32, 38, and 51 mm. Daily gill-net efforts were 2–6 overnight net sets. Nets were set during the morning and retrieved the following morning. Total gill-net effort during the removals was 75 overnight sets. Gill-net locations were randomized each day using a random number table and a numbered grid system superimposed on a lake map. A restriction on the randomization was that net sets needed to be at least two grid locations from each other to minimize gear interference. Each grid was slightly larger than 1 ha in surface area. Angling catch rate (fish/h) was monitored during the removals, and 3 d before and 3 d after the removals. Angling was by a single angler using one line and casting artificial lures throughout the lake, and effort consisted of 1–3 h/d during a total of 19 d. Total length, weight (g), and scale samples were obtained from nearly all fish that were removed, and cleithra were obtained from large (>530 mm) fish as an additional structure for estimating the age of each fish.

The scale samples and cleithra from each fish were examined independently by two people to estimate age; a third person resolved discrepancies between the first two. The age structure of the tagged population was estimated with an age–length key constructed from scale information obtained from tagged fish during both the trapping and removal periods.

Single mark–recapture population estimates were calculated for each age-class by considering the trapping in April as a single marking event and the gill netting in June–July as a single recapture event. A Chapman modification of the Petersen estimate was used to compute the early spring population sizes, and confidence limits were approximated from the Poisson frequency distribution (Ricker 1975). The numbers of marked fish in each age-group were corrected by an estimated 2.4% handling mortality rate (Pierce and Tomcko 1993). The total mortality of the northern pike in Camerton Lake was estimated from the slope of a regression of age on the log_e transformed Petersen population estimates for each age-group. This method relied on some of the same assumptions as catch curve analysis (Ricker 1975). Some other recreational angling by private individuals also occurred at Camerton Lake. To estimate the recreational exploitation of northern pike that occurred outside of our experiment, voluntary tag returns were obtained from a locked tag return box located at the only private access to the lake.

Daily estimates of catchability (q) were obtained by tracking population size and gill-net CPUE by age-class. Because the total mortality rates of northern pike are typically high (Pierce et al. 1995), we used the following recursive formula to account for mortality and fish removals during the sampling periods:

where N_s = the estimated population size at the end of sampling period s, t = the date of sampling period s, r_s = the number of fish removed during sampling period s, and Z = the instantaneous total mortality rate. We assumed that a given day's CPUE data were collected before any fish were removed for the day and that fish mortality occurred after netting and removal. Daily catchability was computed as the mean CPUE for each sampling day divided by the estimated population number at the beginning of the day.

Relationships between population density, mean daily gill-net and angling catch rates, and daily q were explored with linear regression. Regressions using log_e transformations of daily catch rate and population size were used to estimate β. Regressions of the mean lengths of fish in each age-class on the week of sampling (time) were used to detect fish growth during the removal period. Kruskal–Wallis one-way nonparametric analysis of variance (ANOVA) was used to test for differences in daily gill-net q among the age-classes of northern pike. Changes in age and size structure during the removal period were detected by comparing the mean age and length from the first 3 d and the last 3 d of gill netting with a two-sample t-test.

Results

In 1998, a total of 683 northern pike were caught during spring trap-netting, 758 were removed by gill netting, and 66 were removed by angling. An additional 33 fish were caught and released alive during angling before and after the removal period. The total lengths of northern pike caught in Camerton Lake ranged from 244 to 721 mm. No fish smaller than 288 mm were tagged, and we were unable to measure two dead fish that floated away from the gill nets as the nets were retrieved. The length distributions of the fish caught by the three different gears were similar in modal length-groups and range of lengths (Table 1), except that the largest length-groups were not represented in angling catches. Regressions of the mean length of fish in each age-class on the week of sampling (time) showed no apparent changes in fish size during the removal period (r²= 0.05–0.49; F = 0.15–2.91; df = 1, 3 for each age-class; P = 0.19–0.72). The characteristics of the northern pike population in Camerton Lake are more fully described in Pierce et al. (2003).

Table 1. Length-frequency distributions for northern pike caught by trap-netting (N = 683), gill netting (N = 756), and angling (N = 99) in Camerton Lake, Minnesota, in 1998

The estimates of spring population size showed that the bulk of the northern pike in Camerton Lake were age 3 or younger (Table 2) and that less than 3% of the population was older than age 5. The sum of the population estimates for northern pike age 2 and older was 1,673 fish, for an estimated density of 59 fish/ha. During gill netting and angling in 1998, we removed 43% of the estimated spring population of northern pike age 2 and older.

Table 2. Petersen population estimates, numbers of fish tagged during trap-netting, and recapture-to-capture ratios (R/C) for northern pike caught by gill netting in Camerton Lake in 1998

The total mortality of northern pike in Camerton Lake was high, especially compared with the recreational fishing mortality. The instantaneous total mortality from the slope of the regression of log_e transformed population estimates by age was 0.987 (ages 2–7; r² = 0.92), yielding a total annual mortality rate of 0.63. In contrast, the instantaneous fishing mortality rate calculated from tag returns by recreational anglers was 0.053. Total annual tag returns were estimated to be only 3.4%, but recreational fishing effort undoubtedly declined after fish removals from mid-June to July due to the large numbers of fish removed from the lake.

Gill-net catch rates declined during the removals for the whole population of fish age 2 and older (Figure 1) and also for individual age-classes (Figure 2). Assuming that gill-net catch rates should approach zero as the population size approaches zero, gill-net catch rates for fish age 2 and older appeared to decline in a curvilinear manner (Figure 1). The slope of the relationship between population size and catch rate represents q. Therefore, q also appeared to decline during the removal period because the slope changed with decreasing fish density. To fit a power function to these data, a regression was done using log_e transformations of catch rate and population size (r² = 0.78; F = 44.86; df = 1, 13; P < 0.0001). The shape parameter of the power function, β, was 1.70 (SE = 0.25), indicating hyperdepletion of gill-net catch rates.

The daily estimates of q were variable, even within age-classes, but generally declined during the removals. The mean daily q for fish age 2 and older was 0.009665/net (SE = 0.000770; range = 0.005518–0.0154). Gill-net q increased significantly with increasing population size of northern pike age 2 and older pooled (Figure 1; r² = 0.52; F = 13.80; df = 1, 13; P = 0.003). Within individual age-classes, there was also some evidence of increases in gill-net q. Although the slopes of all regressions of q and population size appeared to be positive, age-4 and age-5 fish showed significant (P < 0.05) increases in q with increasing population size (Table 3).

Table 3. Results of regressions of daily gill-net catchability coefficients (q) on numbers of northern pike age 2–6 (df = 1, 13)

Gill-net q differed significantly among age-classes of northern pike, with differences found in mean daily q among ages 2–6 (Figure 3; P = 0.021 for Kruskal–Wallis one-way nonparametric ANOVA of daily catchability between ages). Catchability was greatest for age-4 fish, which had a mean length at capture of 519 mm (SE = 4 mm) during the gill-net removals. The largest difference in q among ages was between age 4 and age 2 (Figure 3). As a result, changes in age and size structure of the northern pike population occurred during the removal period. Significant decreases were found in mean age (t = 3.96; df = 305; P = 0.0001) and mean length (t = 4.67; df = 305; P < 0.0001) between the first 3 d and the last 3 d of gill netting. Mean age in the gill-net catches decreased from 3.0 (SE = 0.1) to 2.5 (SE = 0.1), and mean length decreased from 465 mm (SE = 5) to 428 mm (SE = 6).

The experimental angling catch rates for northern pike were variable but showed a trend of decreasing catch rate with declining population size (Figure 4). Angling catch rates ranged from 0 to 7 fish age 2 and older per angler-hour, with a mean of 3.60 fish/angler-hour (SD = 0.46; coefficient of variation [CV, defined as 100·SE/mean] = 56.27). Log_e catch rate was positively related to log_e population size (F = 4.14; df = 1, 16; P = 0.059), although the relationship was not as strong (r² = 0.21) as that between log_e population size and log_e gill-net catch rate (r² = 0.78).

The daily angling q was also highly variable (Figure 4) and showed no apparent trend in relation to population size during the removal period. The mean of the daily angling catchabilities was 0.00375/angler-hour (SE = 0.00055; range = 0.0–0.0106). The coefficient of variation for angling q (63.87) was more than twice as great as that for gill-net q of fish age 2 and older (30.85). Regression of daily angling q on population size showed no significant relationship (r² = 0.01; F = 0.17; df = 1, 17; P = 0.69).

Discussion

The differences in catchability among ages of northern pike were important, as they illustrate the size-selective nature of Minnesota's multimesh experimental gill nets. Age-4 fish were of sizes that were most vulnerable to the gill nets. The mean length of age-4 fish (519 mm) corresponded to the peak of an indirectly estimated selectivity curve (Pierce et al. 1994) as well as to the statewide mean length of 507 mm for northern pike sampled in experimental gill nets during lake surveys (N = 137,838 fish from 4,196 lake surveys). Most northern pike harvested by the recreational fishery in Minnesota are age 2–5 (Cook and Younk 1998). With the possible exception of age-2 northern pike, the mesh sizes used in experimental gill nets in Minnesota target the ages of fish that are most commonly harvested in the recreational fishery.

Although the variation in gill-net catchability was evident from one day of sampling to the next, gill-net catch rates tracked the declining population size of northern pike in Camerton Lake relatively closely. Further, catchability declined over the removal period. We attributed some of the hyperdepletion in catch rates to the size selectivity of the gill nets. Size selectivity would tend to remove the most catchable individuals in the population at a faster rate, leaving a larger proportion of fish with lower catchabilities. Changes in age and size structure of the northern pike population from our intensive netting helped cause changes in catchability. Size selection should not be a factor in routine gill-net surveys in Minnesota because of the small amount of effort and the low numbers of fish sampled. Size selection could, however, have an important influence on long-term catchability in commercial fisheries subject to large amounts of effort and harvest. The hyperdepletion in our study was not, however, solely attributed to size selectivity because there was also some evidence of changing catchability within age-classes.

Hyperdepletion in catch rates in Camerton Lake could also have been influenced by behavior patterns in the fish. Examples include (1) learning to avoid the net, (2) a reduction in the population of fish exhibiting behavior patterns (such as a greater propensity to move) that would make them more vulnerable to netting, and (3) a reduction in prey species such as yellow perch that would bait northern pike into the gill nets. Though there has been no previous evidence of northern pike avoiding nets, it has been shown that northern pike have learned to avoid other fishing gear. Beukema (1970) showed that the susceptibility to angling with spinners decreased to very low levels after one-half of the population had been caught and released. For movement behavior, telemetry studies by Diana (1980) illustrated that northern pike can be relatively inactive, with displacements of 5 m or more occurring during less than 20% of all 5-min observation intervals. We observed a decline in yellow perch catch rates during our fish removals that may have also influenced the catchability of northern pike. Yellow perch were the predominant prey item in northern pike stomachs in Camerton Lake and seemed to be sustaining much of the growth and secondary production of the latter species (Pierce et al. 2003). Our own visual observations suggested that northern pike were often tangled in nets near yellow perch and that struggling yellow perch lure northern pike into the two smallest mesh sizes.

Gear saturation was not a factor in our study because such saturation would lead to lower catchability at high fish densities. Even with the unusually high density of northern pike found in Camerton Lake (Pierce et al. 2003), catchability was greatest at the highest densities. Changes in size due to the growth of fish during the removal period also did not contribute to changes in q. Because removals occurred during a relatively short period, the mean length in an age-class did not change significantly over the netting period.

Our catchability data showed a trend opposite to the hyperstability in catch rates found by others. Bannerot and Austin (1983) suggested that hyperstable catchability could be influenced by the spatial dispersion of a fish population or by a gradient of skills among fishers, and Crecco and Overholtz (1990) concluded that hyperstability is a general phenomenon among commercial fisheries. Borgstrom (1992) found that the catchability of brown trout Salmo trutta in gill nets depended on fish behavior and declined with increasing density. Crecco and Savoy (1985) found an inverse relation between catchability and the density of American shad in the Connecticut River. Density-dependent catchability in the shad fishery was promoted by the patchy distribution of shad runs, which were differentially exploited by commercial fishing. Henderson et al. (1983) modeled an inverse relationship between catchability and population size with historical catch data for lake whitefish Coregonus clupeaformis in Lake Huron. Using historical angling catch rates and run size information, Peterman and Steer (1981) found an inverse relation between catchability and abundance for chinook salmon O. tshawytscha in two sport fisheries in British Columbia and one in Oregon. However, the use of longer-term catch records, as in the latter three studies mentioned above, masks the shorter-term interactions between fish density, behavior, and fishing effort.

We found only one other study that observed a positive relationship between catchability and fish abundance. In a shorter-term study, Shardlow (1993) was able to directly observe interactions between the fish and the fishing gear. In that study, catchability increased with increasing density in a Pacific salmon hook-and-line fishery in the Strait of Georgia. The positive relation between catchability and density was attributed to aggressive interactions between feeding fish. Both our study and Shardlow (1993) show that the hyperstability reported in several previous studies (listed in Bannerot and Austin 1983) is not a universal phenomenon.

Our direct manipulation of population density had advantages over many of the other studies that used historical time series or commercial fisheries to examine catchability. We were able to use randomized, controlled, and nonoverlapping fishing efforts. Furthermore, our study did not have to account for technology changes or longer-term climatic or anthropogenic effects on the fishery. Because we had known numbers of removals, much of the error in our estimates of northern pike abundance was reduced. However, this study may have suffered from the fish's learning to avoid nets, a reduction in the abundance of prey fish in the lake, and changes in the size or age structure of the population due to the size selectivity of the nets. Nonetheless, this study was unique in that we were able to manipulate the northern pike population in a private lake; most agencies would not have natural lakes available for such experimental purposes. We encountered no other work that used direct manipulation of population density in a natural environment to observe variation in catchability.

Angling catch rates were less useful for tracking changes in northern pike population density, especially compared with gill-net catch rates. Angling catchability, even for a single angler, was much more variable than gill netting and was probably influenced more by changing weather patterns and the fish's learning to avoid angling gear. Our angling data showed no relationship between density and catchability. Similarly, the catchability of walleyes Stizostedion vitreum in angling was found not to be density dependent in two studies from northern Wisconsin. Newby et al. (2000) concluded that angling catchability was not significantly related to walleye population density in Escanaba Lake during 1980–1995. Nor was a relationship apparent to Hansen et al. (2000) across 111 northern Wisconsin lakes surveyed during 1990–1997. Providing some contrast was an analysis of the catchability of lake trout Salvelinus namaycush across 54 Ontario populations. Shuter et al. (1998) concluded that the angling catchability of lake trout increased significantly at low population densities.

Our study monitored the variation in catchability within a single population of northern pike over a relatively short time period. We found some influences of size selectivity and possibly fish behavior on gill-net catch rates. In spite of the variability in gill-net catchability that we observed, gill-net catch rates were useful for tracking the declining population size of northern pike in Camerton Lake. Our results confirm that gill nets can be used as a tool for monitoring changes in abundance, and knowledge of this type of variation in catchability contributes greatly to our ability to interpret trends observed in gill-net catch rates within a single lake. However, changes in catchability within a single lake are only one facet of variation in catchability. Variation in catchability among different lakes or populations of northern pike was not addressed in this study and remains a subject for future research.

Acknowledgments

We thank D. Arola, M. Hagen, J. Hagen, J. Tillma, D. Pereira, S. Shroyer, C. Stertz, P. Wingate, and D. Yule for help with the manuscript, the math, or field work. Thanks to D. Sherman for additional support during this study. We would particularly like to thank D. Schupp, who has been both a mentor and an inspiration for this work. This study was funded in part by Federal Aid in Sport Fish Restoration Project F-26-R in Minnesota.