John M. Hamley
Ontario Ministry of Natural Resources
P.O. Box 429, Port Dover
Ontario NOA 1NO, Canada
Gillnets are popular for sampling inland waters because of their versatility, low cost and ease of operation. They can be used in lakes of any size, in deep or shallow water, under ice in winter (Figure 4.1), and over bottom too rough for seines and trawls. They can be used on a large or small scale: a Great Lakes fishing boat may haul over 10 km of net a day - at the other extreme, one man can carry a canoe and a few gillnets to sample remote lakes inaccessible by road.
Figure 4.1
A “jigger” for setting gillnets under ice. When rope a
is pulled, metal claw b pushes against ice and moves the
jigger in direction of arrow; when a is relaxed, spring
c takes the claw back, ready for the next pull. Under
smooth ice the jigger moves in a fairly straight line
and can be retrieved by cutting a second hole in ice;
the net is then pulled toward that hole. Imler (1971)
attached a radio transmitter to locate their jigger in
mountain lakes with heavy snow cover. This jigger is
2.44 m long.
(d = iron lever; e = metal runners; f = pin joints)
Gillnets seldom take their fish alive and unharmed, except perhaps when small-meshed nets, set in cold water for a very short time, catch fish “by the nose”. A sampling program can kill many fish, including species not needed for study. This can lead to unpleasant criticism, and needs to be considered when planning to net in waters frequented by the public.
When the purpose of sampling is only to collect qualitative information about a fish community, gillnets do as well as any other gear, as long as a range of mesh sizes is used. (Of course there are species and habitats not easily sampled, but that happens with any gear.) But when the purpose is to estimate fish population parameters quantitatively, the study needs to be planned with careful attention to the sampling characteristics of gillnets. The main problems are:
With some active gear like trawls, one can calculate what proportion of a lake has been swept and, from that, what proportion of its fish has been exposed. (Of course, all fish “exposed” in the path of a trawl are not necessarily caught; they may swim out of the way or escape through the meshes.) But gillnets are passive gear to which fish must swim and become stuck. Some species are active and others sedentary, their behaviour may change with seasons, and fish near nets are not necessarily caught — I once saw an angler take several walleye (Stizostedion vitreum) casting beside my net, yet, next morning when I lifted that net, it was empty. Therefore, one usually cannot tell whether a catch represents residents of a small or large area around the net, or migrants passing by (cf. Treschev, 1973). But this is not serious; estimating absolute abundances of fish is always difficult, and usually one only needs to estimate relative abundances of different species, sexes, ages or sizes of fish.
Selectivity of gillnets, as of other fishing gear, is a well recognized problem (e,g, Pope et al., 1975). Clearly it affects any estimates of size-frequency distributions; less obviously, many other calculations. For example, Kipling (1957) noticed that samples taken with just one mesh size lead to biased estimates of length-weight regression, because they favour the fatter of short fish and the thinner of long fish. And estimates of length -at age, growth, and mortality are affected because larger meshes select larger fish of each age, and are generally more efficient (Section 4.1.3) (Hickling, 1939; Ricker, 1969).
Some of these problems remain even when several mesh sizes are used, because the combination of meshes, too, is selective, although less so than a single one (Figure 4.2). Mark-recapture estimates are affected: if the marking and recapturing gears tend to catch fish of similar sizes, the population is under-estimated; if of different sizes, it is overestimated. The answer to that problem is to estimate the population by size classes (Latta, 1959). The general solution to all these problems is to estimate the appropriate selectivity curves, but that is not easy (the history and techniques of estimation are reviewed in Hamley, 1975).
Figure 4.2 Top: Selectivity curves of 11 mesh sizes to lake whitefish. Bottom: Combined selectivity of those 11 nets when fished together in a gang. (After Hamley, 1972; from data of J.L. Hart)
That is, their selectivity curves are taller. This applies to all species for which I have seen evidence (Figure 4.2; and Hamley, 1975: Figure 4 and 5), and may well be the general rule. Perhaps it is because larger-meshed nets are less visible in water or, perhaps, because larger fish range over larger areas and are more likely to meet nets. In any case, comparing the catches of large fish in large-meshed nets and small fish in small-meshed nets, makes large fish appear too abundant. This is a difficult problem unless the mesh selectivity curves are known (occasionally both can be estimated from the same data, e.g. Hamley, 1972).
The temptation is to use too many mesh sizes, resulting in too little data with each. To avoid this, one should examine the sizes and variability of catches taken in comparable gillnetting studies, and from those, estimate how many sets are needed with each mesh size to estimate the desired averages, ratios or whatever, with acceptable precision (see general theoretical discussions of sample sizes, e.g. in Raj, 1972; or practical calculations with gillnets, e.g. Bagenal, 1972). Then one can estimate how many mesh sizes can be used profitably, by dividing the total number of sets that time or manpower allows, by the number of sets needed per mesh size.
Gillnets vary widely in overall dimensions, colour, mesh size, twine material and thickness, hanging, and rigging of weights and floats. Accordingly, they vary greatly in efficiency and selectivity. Usaully it is best to choose types already found successful by local commercial fishermen or other research groups; this also keeps the results comparable with other data. But when desired, gillnets can be custom-designed to suit a particular kind of fish or experiment. (Several types, including trammel nets, are described by von Brandt, 1972 and 1973, or illustrated in Nédélec, 1975. I treat all enmeshing nets together because of similiarities in their construction and operation; however, von Brandt classifies nets according to their principle of capturing fish, and treats gillnets separately from trammel and other nets specifically designed to tangle fish.)
Commercial nets are often designed for high selectivity, to catch marketable fishes efficiently while minimizing bycatches of others. Research nets are often designed for low selectivity, to obtain samples of all that is there.
Mesh sizes need to be matched to the species and sizes of fish sought. This is not always easy, as growing fish may fit a larger mesh in the fall than in the spring, and girths of same-length fish can differ, especially according to sex and proximity to spawning. Also, the true mesh size is not always easy to determine: it may not come exactly as specified by net manufacturer, and may be stretched 5–10 percent by struggling fish, depending on size of fish and thickness and material of net twine.
The best guide to choosing mesh sizes is previous gillnetting data. If none are available for the species in question, length-girth data can be used, roughly (Hamley, 1975):
Optimum size:
According to many authors, the optimum girth for capture
is about 1.25 times the mesh perimeter (range: 1.08–1.35
times), and the optimum length increases approximately
in proportion to mesh size.
Selection range:
According to Baranov's (1948) rule of thumb, fish more
than 20 percent longer or shorter than the optimum length
are seldom caught.
Alternately, some authors predict selection ranges from head and maximum girths, postulating that to be caught, fish must fit into a mesh past their heads (so that they are “gilled” behind the operculum if they try to back out) but not be able to pass through (so that they are tightly “wedged” if they press forward). This, like Baranov's 20- percent rule, implies that most fish are caught wedged or gilled - when many are “tangled” (without necessarily entering a mesh, caught by their teeth, maxillaries, opercular spines, or other projections), catches include smaller and, especially, larger fish than predicted.
Non-selective gillnets are an attempt to catch all sizes of fish in their natural proportions (e.g. Takagi, 1975). Actually they are “gangs” or “fleets” of nets of such mesh sizes that the sum of their selectivity curves apporximates a horizontal line (Figure 4.3) In that figure all selectivity curves are drawn the same height, but if larger-meshed nets are more efficient, their curves should be drawn taller (Section 4.1.3), and progressively shorter lengths of them included in the gang. Because those curves are species -specific, nets designed to be non- selective to one species may not be so to others.
Instead of trying nets of different mesh sizes end-to-end in one long gang, single nets can be built that contain sections made of different mesh sizes, for example in a Latin square arrangement (Houser and Ghent, 1964). This allows sampling all sizes of fish within a small area, and may be very useful in small lakes.
Figure 4.3 Selectivity curves of 24–79 mm gillnets to pink salmon (Oncorhynchus gorbuscha) and combined selectivity curve of those nets. (After Takagi, 1975)
While fish can use other senses to detect nets, vision seems to be the most important (Parrish, 1969); generally, less visible nets catch more fish. Clear monofilament nets, to humans nearly invisible in water, are usually the most efficient (cf. Jester, 1973). As visibility depends on how net colour and tone contrast with the background, it can be affected by the time of day and seasonal changes in water clarity or colour. As a rule, Andreev (1955) recommended darker nets in good light or clear water, and lighter nets in turbid water.
Fish can distinguish colours, and nets of different colours may show several-fold differences in catches. Several authors, especially in Japan in 1957–58, have conducted aquarium experiments to rank net colours in order of their effectiveness in catching, or repelling, particular species of fish (rainbow trout, Salmo gairdneri; carp, Cyprinus carpio; goldfish, Carassius auratus; etc.) (see reference in Parrish, 1969). The effect of net colour can vary with species, because of differences in behaviour or colour sensitivity, e.g. colour and visibility may not matter to fishes that are active at night. This leads to the idea of choosing net colours that tend to select particular species of fish (Jester, 1973), but much study is still needed before that can become general practice.
Another way to control what is caught is by the net's ability to tangle fish. Nets made of more flexible twine (thin, or soft) or hung more loosely (with more webbing per length of finished net) tangle more fish but catch no more wedged or gilled. This affects catches of species like European perch (Perca fluviatilis) that are easily tangled, but not of species like roach (Rutilus rutilus) that are usually caught only wedged or gilled. Mohr's (1965 and 1965a) studies illustrate what happens:
Hanging:
When he fished 35 mm mesh (bar measure is used throughout
this paper) gillnets of hanging coefficients ⅓, ½ and
⅔, the loosely hung nets(⅓, ½) caught about 30 percent
more perch but no more roach (Figure 4.4). Also, the
perch they caught were of a wider size range, because
tangling depends less on mesh size than the other ways of
capture do.
Flexibility
When he fished nets similarly hung but of differently
treated twine, the number and size rang of perch caught
increased with twine flexibility, while catches of roach
remained about the same.
Figure 4.4 Effect of hanging on the selectivity of 35 mm gillnets to perch and roach (after Mohr, 1965). “Hanging coefficient” is the ratio of the length of completed net to the stretched length of webbing used to make it
Many variations in construction or use of nets have been developed, to increase catches of particular species, or the selection between species. Jester (1977) reported on extensive field studies in a large New Mexico lake, designed to find ways to effectively harvest “commercial” fishes (carp; smallmouth buffalo, Ictiobus bubalus; river carpsucker, Capriodes carpio) while minimizing bycatches of “game” fishes (walleye; channel catfish, Ictalurus punctatus, etc.). He achieved best selection by combining several selective characters of net and fishing method: mesh size, net colour, and fishing on seasonal concentrations of the target species. He also experimented with several substances of “bait” placed in mesh bags or perforated tin cans on the lake bottom near the nets, to selectively attract or repel different species of fish.
Two special types of nets need to be mentioned. Trammel nets are often used to catch species of fish (e.g. flatfish, sturgeon) not easily caught in regular gillnets. They are 3-walled nets; a fish can swim through a large outer mesh without resistance, push part of the loose, small-meshed inner net through the large outer mesh on the other side, and become trapped in the pocket that forms. vertical gillnets are sometimes used to study depth-distributions of fish (e.g. Bartoo et al., 1973). To facilitate handling, they are usually set from floating rollers from which enough net is unwound to reach bottom (Figure 4.5).
 | Figure 4.5 A vertical gillnet system. Above: net and floating roller. Below: boat with rack for holding roller. (After Bartoo et al., 1973) |
 | |
Just as there are many designs of gillnets, there are many ways of fishing them. They may be set perpendicular or parallel to shore; in straight lines, zig-zags, or looped to form traps; anchored in place or drifted with currents; left alone or have fish scared into them by beating water; etc. The choice of method depends on the type of water and species of fish sampled; again, it may be best to follow established local practice.
When only qualitative samples of fish fauna are desired, all that is needed is to set nets of suitable mesh sizes in all likely habitats. (Of course, “likely” habitat may include places where nets cannot be effectively used, e.g. shallow marshy areas.) The most important decision is how to distribute the effort: randomly over the whole lake, or stratified according to types of habitat. The choice depends on what is known about the fish community (see a textbook of sampling, e.g. Raj, 1972).
If the desired fish are known to be absent from a certain part of the lake, e.g. from an anoxic hypolimnion, setting nets in there serves no purpose. This leads to the thought that fish habitats are not just areas on a map, but may also be depth zones. In deep water it may matter greatly whether nets are set on the bottom, floated at the surface, or suspended in midwater.
Most methods of quantitative analysis used in fishery science are founded on the assumption that catch (C) per unit of fishing effort (X) is proportional to fish abundance (N) That is,
| C/X = qN | (1) |
where q = proportionality constant called “catchability”. But under usual fishing conditions this is not likely to hold (see discussion by Ricker, 1975: p. 18–24); therefore one needs to fish deliberately in a way that makes it hold, at least approximately.
Equation (1) implies many assumptions, as “q” sums all processes that affect a fish encountering, getting stuck in, or escaping from, a net (Hamley, 1975: Figure 2). To sort them out, the equation can be factored into simpler constituents by introducing a quantity “l”, the relative number of fish that encounter nets. (This is only for convenience of thought; the value of I would be difficult to determine as it is difficult to tell when a fish becomes aware of a net.) Then equation (1) holds if:
| l α X, | l α N, | and | C α l |
That is, if catch is proportional to the number of fish that encounter nets, and that number is proportional to fishing effort (at particular abundance of fish) and to fish abundance (at specified effort).
l α X: To satisfy this the amount of fishing effort, as seen by fishermen, should be made proportional to the intensity of fishing, as seen by fish. There are three main problems. First, the number of fish that encounter nets may not increase in proportion to net size (length, or especially height) because fish are not distributed uniformly or randomly through the water. Therefore the units of effort, the nets, need to be standardized in size as well as in colour and other details.
Second, the number of fish that encounter nets may not increase in proportion to the number of nets set, if the nets are set close enough to compete for capture of the same fish. A common variation of this problem is that, when different mesh sizes are set in the same gang, the most efficient ones (for available fish) may reduce catches in nearby less efficient ones (Larkins, 1963), or large fish may “lead” along small-meshed nets until they are captured by one with a larger mesh. Little is known about competition between adjacent gillnets, so when efficiencies of different kinds of nets are to be compared, they should be set in separate locations, unless that would increase the sampling variance too much when only a few sets can be made. If the nets must be set in the same gang, at least they should be separated by gaps.
Third, the activity and consequently the vulnerability of fish change through time of day (Figure 4.6) and season, depending on the species. Therefore the times of setting and lifting nets should also be standardized, and if sampling must be prolonged over many weeks, the catch-per-effort data should be examined for trends which would indicate changes in catchability abundance of fish.
l α N: For this to hold, the abundance of fish near the nets should be proportional to that in the lake. This can be achieved by distributing the nets randomly over the lake so that all fish are subject to the same fishing intensity, but that can be wasteful if the fish do not frequent all parts of the lake; then stratified sampling may be better (see Section 4.3.1). (Of course the same spots should not be fished long enough to deplete fish locally.) If during sampling the fish population is known to be changing due to natural mortality, immigration or emigration, one must plan to account for those changes when analysing the data.
C α l: The main problem in satisfying this is that, while nets remain in water, their efficiency can decrease with saturation (accumulation of captured fish) or fouling by a algae or silt. Figure 4.7 shows that, long before saturation, catch per net stops being proportional to fishing time and begins to increase less rapidly than does the number of fish that encounter nets. Then it is no longer a satisfactory index of abundance. To avoid that, the fishing time should be kept short — 24 hours may be a good standard. Shorter times may work well when fitted to the diel cycle of fish activity (Figure 4.6) (e.g. over-night sets for fish active at night); longer times should be used only if catches are extremely low.
Meth (1970) noted that the amount of fish held by a saturated gillnet increases with mesh size (Table 4.1). These are estimates of the saturation asymptote (Figure 4.7), not of the highest catches nets can take. This relation warrants further study. If it is not very sensitive to the species being caught, it may serve as a rough general predictor of saturation catches.
Figure 4.6 Relative numbers of fish caught per experimental gillnet at various hours of the day, Clear Lake, lowa. Each net consisted of five 7.62m long sections, of 19– 51 mm mesh. (Yelloow perch, perca flavescens; northern pike, Esox lucius; bluegill, Lepomis macrochirus; pumpkinseed, L. gibbosus; walleye, Stizostedion vitreum; yellow bass, Morone interrupta; white bass, M. chrysops; black bull-head, Ictalurus melas; crappies, Pomoxis annularis and nigromaculatus; white sucker, Catostomus commersoni; carp, Cyprinus carpio. After Carlander, 1953)
Figure 4.7 Saturation of gillnets when fishing for lake trout and lake whitefish in Great Slave Lake; abundance of fish increased from Area H to Area E (after Kennedy, 1951). Note that if fish abundance is judged by the catch per net, the difference between Areas E to H appears to diminish toward the right
| Mesh size (mm) | Fish per m2 of net (kg) | Main species caught | Data of: |
|---|---|---|---|
| 19 | 0.12 | Alewife,Alosa pseudoharengus Lake chub, Couesius plumbeus | Meth (1970) |
| 32 | 0.29 | Longnose sucker,Catostomus catostomus | Meth (1970) |
| 70 | 0.34 | Lake trout, Salvelinus namaycush Lake whitefish, Coregonus clupeaformis | Kennedy (1951) |
Because our ability to interpret gillnet catch data is limited by our meager knowledge of mesh selectivity, it will be valuable to determine selectivity curves whenever convenient. The best opportunity is when the size-frequency distribution of a fish population is known. For example, Hamley and Regier (1973) tagged and released 650–833 walleye in each of 3 years, then estimated selectivity curves “directly” from the proportions of various size classes of tagged fish caught later. If the population is not known but fishing is very intense, “mortality estimates” (e.g. the de Lury method-Hamley, 1972) can sometimes be used; they rely on comparing the rates of depletion of different size classes of fish.
Figure 4.8 Relative heights of gillnet selectivity curves, as functions of mesh size, for Pacific herring (Clupea pallasi,, data of A.L. Tester), gizzard shad (Dorosoma cepedianum, data of V.W. Lambou), walleye (data of J.M. Hamley and H.A. Regier), lake whitefish (data of J.L. Hart), and striped bass (Morone saxatilis, data of L. Trent and W.W. Hassler). (From curves in Hamley, 1972: Figure 2 and Hamley, 1975: Figures 4 & 5.)
Particularly important is to study the relation of efficiency (height of selectivity curve) to mesh size. It seems linear on a logarithmic plot (Figure 4.8). but why are the observed slopes so different? Are they characteristic to each species, or also affected by net construction, etc.? Finding a general relation would strengthen the “indirect” methods of estimating gillnet selectivity, which rely on comparing size distributions of fish caught by different mesh sizes, and are presently deficient because they do not estimate relative heights of the selectivity curves (Hamley, 1975).
Also worth pursuing are efforts to relate shapes of selectivity curves to shapes of fish body, by measuring girths at places presumed to be critical to capture, e.g. at rear edges of maxillaries and operculum, and at the widest part of body. (Hunter and wheeler, 1972 described a convenient measuring device.) Several authors have worked on it (see references cited by Hamley, 1975: pages 1951–52 and 1955–56), but the results are not yet clear.
Andreev, N.H., 1955 Some problems in the theory of the capture of fish by gillnets. Tr.Vses. Nauchno-losled.Inst.Morsk.Rybn.Khoz.Okeanogr., 30:109-27. Transl.from Russian by Fisheries Laboratory, Lowestoft, England, 30 p.
Bagenal, T.B., 1972 The variability of the catch from gillnets set for pike Esox lucius L. Freshwat.Biol., 2:77–82
Baranov,F.I., 1948 Theory and assessment of fishing gear. Moscow, Pishchepromizdat. (Chapter 7. Theory of fishing with gillnets,translated from Russian by Ontario Department of Lands Forests, Maple, Ontario, 45 p.)
Bartoo, N.W., R.G. 1973 Hansen and R.S. Wydoski, A portable vertical gillnet system. Prog. Fish-Cult., 35:231-3
Brandt, A. von, 1972 Fish catching methods of the world. London, Fishing News (Books) Ltd., 240 p.
Brandt, 1975 Enmeshing nets: gillnets and entangling nets - the theory of their efficiency. EIFAC Tech. Pap., (23) Suppl.1, vol 1:96–116
Carlander, K.D., 1953 Use of gillnets in studying fish populations, Clear Lake, lowa. Proc. lowa Acad.Sci., 60:621-5
Hamley, J.M., 1972 Use of the de Lury method to estimate gillnet selectivity. J.Fish.Res.Board Can., 29(11):1636-8
Hamley, 1975 Review of gillnet selectivity. J.Fish.Res.Board Can., 32(11):1943–69
Hamley, J.M. and H.A. Regier, 1973 Direct estimates of gillnet selectivity to walleye (Stizostedion vitreum vitreum). J.Fish.Res.Board Can., 30(6):817-30
Hickling, C.F., 1939 The selective action of the drift-net on the Cornish pilchard. J.Cons. CIEM, 14:67–80
Houser, A. and A.W. Ghent, 1964 An experimental gillnet based on the principle of the Latin Square. Trans.Am.Fish.Soc., 93:233-42
Hunter, C.J. and C.O. Wheeler, 1972 Device for measuring length and girth of fish. J.Fish. Res.Board Can., 29(12):1784-5
Imler, R., 1971 Radio telemetry as an aid in winter gill-netting. Prog.Fish-Cult., 33:59
Jester, D.B., 1973 Variations in catchability of fishes with color of gillnets. Trans.Am. Fish.Soc., 102:109-15
Jester, 1977 Effects of color, mesh size, fishing in seasonal concentrations, and baiting on catch rates of fishes in gillnets. Trans.Am.Fish.Soc., 106:43–56
Kennedy, W.A., 1951 The relationship of fishing effort by gillnets to the interval between lifts. J.Fish.Res.Board Can., 8(4):264–74
Kipling, C., 1957 The effect of gillnet selection on the estimation of weight-length relationships. J.Cons.CIEM, 23:51–63
Larkins, H.A., 1963 Comparison of salmon catches in monofilament and multifilament gillnets. Commer.Fish.Rev., 25(5):1–11
Latta, W.C., 1959 Significance of trap-net selectivity in estimating fish population statistics. Pap.Mich.Acad.Sci., (44):123-38
Meth,F., 1970 Saturation in gillnets. University of Toronto, M.Sc. Thesis, 39 p.
Mohr, H., 1965 Auswirkung der Einstellung von Kiemennetzen auf die Fängigkeit von Barsch und Plötze. Arch.Fischereiwiss., 16:108-15 (English summary)
Mohr, 1965a Auswirkung der Biegesteifigkeit des Netzmaterials auf die Fängigkeit bei Kiemennetzen für Barsch und Plötze. Arch.Fischereiwiss., 16:215-23 (English summary)
Nédélec, C. (ed.), 1975 FAO catalogue of small-scale fishing gear. West Byfleet, Surrey, Fishing News (Books) Ltd., for FAO, 191 p.
Parrish, B.B., 1969 A review of some experimental studies of fish reactions to stationary and moving objects of relevance to fish capture processes. FAO Fish.Rep., (62) vol.2:233-45
Pope, J.A. et al., 1975 Manual of methods for fish stock assessment, Part 3. Selectivity of fishing gear. FAO Fish.Tech.Pap., (41) Rev.1:65 p.
Raj, D., 1972 The design of sample surveys. New York, McGraw-Hill
Ricker, W.E., 1969 Effects of size-selective mortality and sampling bias on estimates of growth, mortality, production, and yield. J.Fish.Res.Board Can., 26(3):479–541
Ricker, 1975 Computation and interpretation of biological statistics of fish populations. Bull.Fish.Res.Board Can., (191):382 p.
Takagi, K., 1975 A non-selective salmon gillnet for research operations. INPFC Bull., (32):13– 41
Treschev, A.I., 1973 Application of the fished volume method for measuring fishing effort. Coop.Res.Rep.ICES, (79):54 p.
