WaterSubject

For many fishery management purposes, it is often necessary to estimate the number of fish present in linear water bodies such as rivers, canals and drainage channels. A variety of methods are available to do this, but all require significant physical effort from a team of people and thus impose large costs on the organisation responsible. Cost and practical limitations often mean that variability and error are difficult to estimate for a particular sampling programme, reducing the scientific credibility and the practical utility of the data obtained. This study presents results from an unusually detailed study of the fish populations of a long length of canal, which illuminate the influence of both gear and survey design on the accuracy and precision of the population estimates obtained.

The use of single-pass electric fishing to assess abundance of fish has been reported previously. Such a method is cost effective for fish that occur at a low density but is size selective and cannot reliably estimate fish less than 200 mm. As a recent study found that 54% of fish are less than 60 mm in the canal environment, electric fishing is clearly inadequate to sample all size ranges. An alternative design is to intensively sample a number of sites using a method that can capture smaller fish. A common method is to use nets; these are still size selective but are able to sample much smaller fish than electric fishing. A useful sampling design is to estimate the mean density of fish from a randomly selected number of sites. The true number of fish is estimated from the number found at each site. In this paper, the usefulness of the estimate can be described by knowing its accuracy (amount of systematic overestimate or underestimate of the true value) and precision (how close the estimate is to the true number). In practical terms, a number of sites are selected, and subject to multiple removals. As the catches are kept separate, the total number of fish present can be estimated from the diminishing returns. The accuracy of any survey depends on the fishing gear used. The precision, on the other hand, is dependent on the magnitude of between-site error (spatial variation), within-site error (measurement error) and the sampling effort (number of sites and amount of within-site effort). Consequently, for any given spatial distribution of fish there is an optimum allocation of sampling effort that will achieve a given precision for the least effort. Netting is one sampling technique commonly used. The type of net used may influence the precision of a within-site population estimate for a given number of removals. Consequently, the use of a net that has a higher capture efficiency, per removal, may require less removals to achieve a given precision than one with a lower capture efficiency and a reduced within-site effort. This may reduce effort and hence labour cost. However, few studies have been published where different nets have been compared in similar conditions.

The research presented here aims to assist those designing surveys in linear waters. The objectives of the study were to

Study area

The netting study was undertaken in a system of interconnecting artificial navigation canals located in the Midlands region of the UK. These canals are used for recreational boat traffic, and the physical habitat does not show any marked longitudinal variation. Typically the channels are 8–14 m wide and have a maximum depth of 1.4–1.8 m with shallow margins. Vegetation is sparse, with a few submerged plants, although emergent fringes of macrophytes (e.g. Glyceria maxima (Hartm.) Holmb.) are present in some sections. Water turbidity is variable and much influenced by boat traffic. During this investigation, Secchi disc depth was in the range 0.1–0.4 m in July and 0.3 to >1.2 m between November and March. The channel substrate consisted of soft silt or gravel in the margins and compacted material, usually clay and small stones, in the central channel. There was negligible water flow at all sites on all survey occasions.

Net design and timing of the study

Two net designs were compared: a lightly leaded net (25 m long, 2.5 m wide, mesh diameter 3.0 mm and leadline 120 g/m²) and a heavily leaded net (25 m long, 3.0 m wide, mesh diameter 5.0 mm and leadline 250 g/m²) on 51 occasions. Netting trials took place between September and October in 1995 and also in 1996.

Within-site depletion methodology

At each site a 35 m length of canal was sealed at each end by block nets (25 m long, 2.5 m wide, mesh diameter 3.0 mm and leadline 70 g/m²) drawn across the canal simultaneously. Two or three removals were then undertaken with either the heavy or the light leadlined net. For 11 sites, two or three removals were carried out with the light leadlined net and then a further two or three removals were completed with the heavy leadlined net. For each catch, all fish were identified to species and fork length was measured to the nearest millimetre. If more than 200 fish (<60 mm) of a particular species were caught, a random subsample of not less than 100 fish was measured and the remaining individuals were counted.

Within-site population estimates

The catch data were stratified before estimating population size. From length frequency histograms, size classes of each of the four dominant species were chosen. Roach were divided into three groups (<60, 60–99 and >99 mm), whereas for gudgeon, perch and bream (Abramis brama L.) there were few fish >99 mm and only two size classes (i.e. <60 and ?60 mm) were used. As population estimates based on low numbers of fish tend to be unreliable strata, which consisted of less than 30 fish in total, were excluded from further analyses.

Population size and catchability were estimated using the formulae provided for estimates based on two depletions and those, as given for three depletions. In some cases, the second and/or the third catch was greater than the first and the model failed. When three depletions were undertaken, the assumption of constant depletion was checked by assessing if the third catch was greater or less than expected using ?² analysis.

Size selectivity

The sizes of fish caught from nearby sections of the canal using nets of mesh size 3.0 and 5.0 mm were plotted as a cumulative frequency and compared with the size distribution of fish caught from the same area with a much finer mesh (1.0 mm) net. The sizes of fish caught using a larger mesh net (12.5 mm) from a nearby canal were also plotted as a cumulative frequency.

Accuracy of the removal method using marked fish

On two occasions, the accuracy of the heavy leadlined net in estimating the population size of roach and gudgeon was assessed. Two sites were chosen: one consisting of a gravel substrate (site A, Birmingham and Fazeley Canal, NGR SP 168928) and the other consisting of a silt substrate (site B, Coventry Canal, NGR SP 327953). On each occasion, known numbers of roach and gudgeon >90 mm were marked with a subcutaneous injection of Alcian blue dye, using a panjet inoculator, and introduced into a 35 m section of canal isolated by stop nets. After a wait of 30 min to allow the marked fish to disperse, two removals using the heavy leadlined net were performed.