Previous Page Table of Contents Next Page



Abundance surveys are frequently used to gain information on stock size and distribution. In marine areas demersal fish species are typically surveyed by bottom trawls whereas pelagic species are predominantly surveyed by acoustic methods calibrated by trawl catches.

In some areas bottom trawling may be impossible due to the sea-bed topography. This is typical of artificial lakes, but large non-trawlable areas also occur in marine areas (e.g. because of reefs or rock outcrops). In West Greenland, for instance, non-trawlable archipelago and fjord areas constituted one-third of the total area of distribution of cod (Hovgård and Riget 1992).

A number of gears can be used to survey such problem areas, including longlines, jigs, gillnets, traps, beach seines, Danish seines and pelagic trawls. Each gear has its own strong points and the choice of research gear should match the scientific objectives (Methven and Schneider 1998). Compared to the towed gears, gillnets and longlines may be used on all bottom types and they may be used from smaller and technologically simpler vessels. Compared to traps they are easier to handle. Gillnets and longline may be set at all depths and may be rigged to either sink or float. They are usually set horizontally but may also be rigged to stand vertically (Rudstam et al. 1987, Matsuoka et al. 1992). The latter design is useful to evaluate the depth distribution of fish.

Gillnet and longlines differ as regards selection properties. They rely on different encounter processes - gillnet fishing depends on a ‘random’ movement of fish whereas longlines attract fish directly by visual and olfactory stimuli from the bait (Engås and Løkkeborg 1994). The gears also differ with respect to ease of handling giving longlines an advantage for work in deeper waters. The implications for survey use are briefly summarised below.

6.1.1 Gillnets

Due to the pronounced size selectivity of gillnets (see Section 5), surveys normally use different mesh-size nets concurrently, typically deployed as ‘gangs’ of short sections tied together. However, special multi-mesh research nets, whereby the nets of different mesh-size are mounted on the same head and foot rope, are also available (e.g. Kurkilahti and Rask 1996). In fresh water research, mesh sizes ranging from 1 cm upward are commonly in use enabling a large range of fish sizes to be caught.

Catch rates in gillnets are usually low (e.g. Methven and Schneider 1998) and therefore many net settings are needed to acquire a sufficiently large sample of fish. In shallow waters, where nets may be hauled by hand, surveys may be carried out using small vessels (e.g. inflatable rubber boats). In such cases, the cost of the effort units may be low.

6.1.2 Longlines

Size selection by hooks is much less pronounced than that of gillnets, and longline surveys have typically been carried out by using only one hook size. In commercial fisheries, longlines are often found to catch more large fish than trawl gear (Sætersdal 1963, Nederaas et al. 1993). Comparison between surveys, where identical distances are fished by different gears, suggests that longlines catch relatively fewer smaller fish, but significantly more larger fish than trawls (Hovgård and Riget 1992, Jørgensen 1995).

Longlines are well suited for surveys covering areas ranging considerably in depth. For instance, Sigler and Fujioka (1988) used a longline survey to cover depth ranges between 100 and 1000 m in the Gulf of Alaska, and Jørgensen (1995) fished as deep as 1500 m in the David Strait.

When different gears are used to cover inshore and offshore areas the difference in selective properties may impede quantitative interpretations (Godø et al. 1989, Methven and Schneider 1998). One approach is to calibrate the different gears against each other. Hovgård and Riget (1992) used between-gear-calibrations to express inshore abundance of cod taken by longline in units used by an offshore trawl survey. Calibrations must be done on a length group basis to account for differences in size selection by gears.


The ultimate aim of an abundance survey should be to estimate the absolute fish abundance in the area surveyed. A number of authors have related catching efficiency to the areas covered by the fishing operations. Two area concepts have been used:

  1. An affected area, which is the area from where catches may be taken and

  2. An effectively fished area, which is a theoretical area where all individuals have 100% probability of capture.

For longlines the affected area may be thought of as the area which is influenced by the odour plume created by leakage of chemical attractants from the bait. For gillnets the affected area is where fish, due to their movements, may be caught by the gear.

Example 6.1 Calculation of the effectively fished area

McQuinn and Gendron (1988) tagged and released the marine snail Buccium undatum at various distances from a baited trap. By recording recapture of animals released at different positions they were able to determine the recapture rate as functions of direction and distance from the trap. The results were presented as isorecapture lines through points with equal recapture rate. An example of such a plot is shown in Fig 6.1.

The affected area (AA) is the sum of the subareas shown, i.e.

where AI is the size of the subareas. The effective area fished (AE) may be calculated as

where RI is the recapture rate per subarea.

The effective area is thus a standardised area from which 100% of the animals are caught.

Figure 6.1

Figure 6.1 Affected and effectively fished areas. The affected area, shown on the left, provides the probability distribution of catches taken at various distances from the gear. The effectively fished area, shown at the right, expresses the equivalent size of an area where 100% of the animals are caught.

If an effectively fished area can be calculated, the density of the surveyed animals (D) can be estimated as:

D = C/AE

where C designates the catch. Subsequently, the total stock abundance may be calculated by raising the densities to the total area.

The concepts of affected and effectively fished areas may be linked to that of the catchability used in the selection equation in Section 5, as:

It is often inferred that larger fish are more mobile than smaller fish (e.g. Eggers et al. 1982, Rudstam et al. 1984, Spangler and Collins 1992, Hansson and Rudstam 1995). In that case the ratio of affected area to effectively fished area depends on fish size:

For highly mobile species like fin-fish, it is difficult to estimate the size of the affected area in practice. The idea of an effectively fished area is, however, valuable from a conceptual viewpoint and has been used in various works for both towed and passive gears (e.g. Dickson 1989, Ramm and Xiao 1995).


In most cases no attempt is made to raise the survey results to absolute abundance, but rather to use the survey results as indices of the stock abundance. The most commonly used index is that of catch per unit effort (CPUE). The CPUE-index is an appropriate measure if it can be related to the stock abundance (N) by proportionality:

CPUE = q N

where q is the survey catchability. One may read this equation as a reduced version of the selection equation where the size dependent selection is ignored. In fish stock assessments, the fish-size effects are often accounted for by expressing CPUE by age:

CPUEage = q ageNage

This formulation is a practical way of viewing selection as a ‘black box’ being absorbed in the qage parameters.

Trawl survey CPUEs are in many cases useful for describing the relative changes in stock abundance. The index may be used in its own right, but may also be used for tuning assessment derived from sequential population analysis (VPAs). Examples of such uses are given in Stefansson et al. (1997).


Size selection of trawl gear is of the sigmoid type and by choosing small mesh-sizes in the codend, it is possible to catch a wide range of fish sizes with an approximate equal efficiency. The selectivity of gillnets, being narrower (Section 5), requires that surveys be conducted using several mesh-sizes simultaneously. Longlines are generally not effective for catching smaller individuals, which are therefore estimated with considerable uncertainty.

Besides the size selection differences, interpretation of gillnet and longline CPUEs is hampered by the fact that both gear types are characterised by a fixed number of positions where the fish may be caught (i.e. the number of meshes or hooks). Over the fishing period an increasing number of these catching sites may be occupied thereby decreasing the fishing power of the gear over time. Existence of such saturation effects implies that catches cannot be expected to increase linearly with fishing time, but that the relationship between CPUE (numbers caught per hour) and fishing time is asymptotic.

6.4.1 Gillnets

For gillnets the importance of gear saturation appears to be relatively weak (see example 6.2). Some studies have indicated a reduction in CPUE for longer setting times. For example, Kennedy (1951) found a clear reduction in the catch per day when nets were allowed to stand for two or three days as compared to their being lifted daily. Hickford and Schiel (1996) observed about the same numbers of fish caught in 6-hour-day settings and 15-hour-night settings for four New Zealand reef species. However, when catch rates were compared no statistical difference was found between the catch per hour in the long and short settings. Engås (1983) compared catches of blue ling (Molva dypterygi) in gillnets lifted at 1, 2 or 3 day intervals and found no differences in catch per day in two experiments whereas a third experiment showed increasing catch rates with increasing set time.

Example 6.2 Saturation effects of cod in a Greenland abundance survey

The cod gillnet selection study by Hovgård (1996a) used data from three inshore areas off West Greenland for an empirical evaluation of the relation between gillnet set duration and CPUE for juvenile cod. Set duration varied between 3 and 20 hours with the majority of sets made (83%) using set times between 5 and 10 hours.

The average catch rates plotted are shown in Fig 6.2. For set durations below 10 hours, the numbers caught appear to be linearly related to set time, i.e. indicating that CPUE is constant, disregarding differences in set duration. However, for longer set time some decline in CPUE is indicated.

Figure 6.2

Figure 6.2 The relation between CPUE and gillnet set times for the Greenland inshore cod survey, 1985–1995. Set durations marked at 14 hours include all sets above 13 hours. The number of settings is indicated by the size of the bubble around the mean. The error bars show twice the standard error of the mean.

6.4.2 Longline

When retrieving longlines it is typically seen that a large proportion of the hooks are either occupied by a fish or cleaned of bait. This suggests that the fishing power of the line has degenerated during the setting as the number of hooks actually fishing is reduced during operation. At the same time, the number of hooks occupied by individuals of other species influence the catches of a species of interest. Empirical studies clearly suggest that catches may not be proportional to set time (e.g. Skud and Hamley 1979, Løkkeborg and Pina 1997).

The catch process for longlines has been discussed by Murphy (1960) and Rothschild (1967) both observing that the decrease of the fishing power of a longline with time can be described by an exponential decay model. This finding was formalised by Somerton and Kikkawa (1995) who note that the instantaneous rate of bait removal, recorded separately by species, may be a more appropriate measure of fish density than CPUE. The rate of bait removal is expressed as

which, after integration, becomes

Bt = B0e-N

The total bait removal rate (A) may be separated into components, i.e.

A = λ12+.......+λno

where λ1, λ2, … λn designate the partial removal rates of n different fish species taken on the line, whereas λo designates the removal by other processes (bait lost due to scavengers or by mechanical processes).

The total instantaneous rate of bait loss is estimated as

where B0 is the number of baited hooks and Bt is the number of hooks still containing bait when the line is hauled at time t.

The partial removal rates caused by species i can subsequently be estimated as

where Ci is the catch in numbers of species i.

Example 6.3 shows the calculations of the instantaneous bait removal rates for a longline fishery and compares this index to that of the traditionally used CPUE measure.

Example 6.3 Using longline CPUE as an abundance index for tusk, off western Norway.

Bjordal (1983a) measured the effect of two bait types on catches in a longline fishery that targeted tusk (Brosme brosme) and ling (Molva molva). When retrieved, mackerel baited hooks had: 17.8% tusk, 3.2% ling, 4.2% other fish, 34.7% bait returned and 40.2% bait lost. Assuming a set time of 6 hours (reported range 3–9 hours) leads to the following total estimates of instantaneous rates of bait removal = 0.1764, which can be separated into:

λtusk = 0.0481; λling = 0.0086; λother fish = 0.0113; λscavengers = 0.1083

where λscavengers designates the empty hooks.

The abundance of the species competing for the bait is expected to differ between years. The effect of such changes is accounted for when using the instantaneous bait removal rates as the measures of stock density.

The CPUE index will be significantly affected by changes in the number of animals competing for the bait. Assume, for instance, that tusk density remains constant (λtusk = 0.0481) but that the numbers of other bait removers (i.e. ling, other fish and scavengers) change between years. This will lead to apparently significant decrease in the tusk CPUE index measured as tusk catch per hour (Fig. 6.3). A halving of the abundance of the other bait predators will result in an apparent increase in tusk CPUE of 57% whereas a doubling of the abundance of other predators results in an apparent reduction in tusk CPUE of 43%.

Figure 6.3

Figure 6.3 The effect of the density of other bait predators on the estimate of tusk density measured as catch per hour. The true tusk density is assumed to be constant.

Somerton and Kikkawa (1994) used ‘intelligent hooks’ that recorded the actual time at which each fish was caught. They were thereby able to document that the captures of pelagic armorhead (Pseudopentaceros wheeleri), of other fish species combined and of the increase in the number of empty hooks, followed the pattern of exponential decay, except for a short initial phase when catches increased. They interpreted the initial phase as a period that the fish used for recognising the longline. A similar exponential decay and also the short initial phase may also be inferred from data on pacific halibut (Skud 1978b, Fig.4).

The use of the instantaneous removal rates provides a simple way to adjust the longline catch information for the effect of gear saturation and competition between species, including that caused by scavengers not brought on deck. The approach may of course also be applied to account for intra-specific competition (i.e. that different age groups of the same species may compete for bait).

The use of instantaneous removal rates does not imply that the duration of setting becomes unimportant. First of all the bait loses its attractive qualities as the leaking of attractants diminishes over time (Løkkeborg 1990). This will cause a decrease of the fishing power of the line with time that is not accounted for by the bait removal. Secondly, the line setting may cover different animal densities of which some may lead to a local saturation of the gear, which may be difficult to discern. For this reason, it is recommended to set the longline for short periods.


Surveys are typically carried out to estimate the relative abundance while usually also providing some measures of the precision of these estimates. The predominantly used approach is stratification where the total survey area is broken down into strata possessing common features regarding the species composition and their densities and age structures. If the strata are well selected, i.e. characterised by a low variability within the strata and clear differences between strata, the stratification is a convenient way of increasing the precision of the abundance estimates (Cochran 1977). Survey effort may be allocated between strata according to their size or may be optimised if a priori knowledge on the variability of interest is known (see a short introduction to the use of stratification schemes in the sampling of fish populations in Sparre and Venema 1992). Within each stratum, the fishing stations are chosen randomly.

For trawl surveys, stations are typically fished by hauls of short duration (0.5–1 hours) allowing the research vessel a high degree of freedom of movement. In contrast, gillnet and longlines require a certain soak time (e.g. the gear is set fishing overnight). The survey vessel has to await commencement of the fishing operations and this restricts the ability to cover a large survey area.

Two major methods of fishing with passive gears are known from the commercial fisheries, which may both be applied in research work.

A) ‘Long-chain settings’ where an extensive length of gillnets or longline are set in a single continuous row and

B) ‘Short-chain settings’ where the gears deployed are separated in relatively short chains (Fig 6.4).

Figure 6.4

Figure 6.4 Long chain setting (left) and short chain setting (right) for gillnets or longline surveys. Possible set and hauling times are indicated before and within the brackets, respectively.

Both approaches have been used in longline surveys. For example, Sigler and Fujioka (1988) used long-chain settings of 15 km (7 200 hooks) whereas Hovgård and Riget (1992) used five short lines of 800 m (400 hooks) set successively as the sketch in Fig. 6.4. In gillnet surveys, short chain settings are usually preferred. Example 6.4 compares the allocation of fishing stations between a trawl survey and two passive gear surveys using short-chain and long-chain settings, respectively.

Example 6.4 Fishing station allocation for hauled and passive gears

Assume that an area with three depth strata is to be covered by a survey confined to six days. Due to dispersal of fish above the bottom at night, the fishing operations are restricted to daylight hours. This allows the research vessel to move freely during the night. A trawl vessel may fish 4 stations of 0.5 hours a day and move between stations while the catch is being processed. The passive gear vessels are restricted to one area each day to allow for the required soak time.

Within these constraints, the trawl survey can cover the area surveyed with 8 stations in each stratum (Fig. 6.5, left). The long-chain scheme (Fig. 6.5, centre) provides 2 stations per stratum. The short-chain approach (Fig. 6.5, right), setting 4 chains per day, provides 8 stations per stratum.

Figure 6.5

Figure 6.5 Selecting the fishing station in a trawl survey and in surveys using fixed gear applying either a long-chain or a short-chain approach.

The trawl station allocation scheme is superior as it allows a better random distribution of a higher number of hauls. The long-chain scheme produces a much lower number of stations fished. The short-chain scheme provides as many stations as the trawl survey but these are not randomly distributed but form clusters of daily sets. Each cluster may be randomly placed. The short-chain approach is better than the long-chain approach as it produces insight as to variability between stations on the spatial scale defined by the within cluster distances.

For covering restricted coastal areas or smaller fresh water habitats the time constraint inherent in the use of passive gears, will usually be of less importance due to the size of these areas. Obviously, the problems encountered for large lakes will mirror those of the marine environment.

Previous Page Top of Page Next Page