Fish Eating Birds and Salmonids in Scotland

6. Methods of control of sawbill duck predation

M Marquiss

SUMMARY

A review of methods of controlling predation by fish-eating birds suggests that population turnover, influenced by the attractiveness of the site and the mobility of birds, reduces the effectiveness of control measures.

In 3 of 4 well-documented sawbill reduction case studies, intensive shooting reduced duck abundance at a site. In all cases the reduction in duck abundance was much less than the numbers shot and sawbills were not eliminated, even from small river sections.

The implications for sawbill 'control' procedures are that (i) shooting has to be intensive to reduce local sawbill abundance, (ii) sawbill reduction programmes need to be monitored to quantify effectiveness, and (iii) it cannot be assumed that shooting has only local effects.

INTRODUCTION

This Chapter reviews methods of reducing sawbill duck predation. Using existing knowledge and some guesswork, models of duck predation on juvenile salmon could become valuable fisheries management tools, but only if they can be shown to resemble reality (Chapter 7). The predictions of such models can only be tested by experiment so for this reason at least, we need to investigate efficient methods of controlling sawbill predation. Here we discuss potential methods of control and review case studies of control measures undertaken in Canada and in Scotland.

POTENTIAL METHODS OF SAWBILL CONTROL

Sawbill ducks are wide-ranging birds, well dispersed at some times of year and over some habitats, but aggregating to exploit abundant food whenever and wherever it arises (Wood 1985a). They are attracted to places where foraging is easiest, that is, where foraging is easier than at nearby alternative sites. There are two potential ways to control sawbill predation - the whole population of ducks might be reduced (e.g. White 1939) or the local foraging circumstances changed so that ducks find it easier to fish elsewhere (discussed in Wood 1985a, 1985b, Wood & Hand 1985).

Population reduction

The simplest way to reduce the population of birds is by shooting but this is not easy because birds become vigilant and escape even co-ordinated efforts. Persistent shooting throughout the daylight hours might reduce the number of birds using a small site such as a stocked pond but on larger sites, such as lochs or whole catchments, shooting merely moves survivors from place to place. This makes the process progressively more difficult as the remaining birds become increasingly wary.

Aside from these technical difficulties, the main factor influencing the efficiency of shooting is the turnover of birds at a site. A site might support ten birds at any one time but they could be drawn from a population of 100 birds that come and go. In this situation, shooting five birds might be expected to reduce the fishing flock by half, but in fact would only reduce it by five percent (i.e. 5/100). This is one explanation proffered for the ineffectiveness of killing cormorants at Loch Leven, Kinross-shire. There, at least 1200 cormorants were killed between 1983 and 1993 with no obvious effect on the numbers of birds using this heavily stocked loch (Carss et al. 1997a). Cormorants forage over very wide areas and the shooting at Loch Leven could have been drawing on a population numbering thousands of birds from adjacent waters, estuaries and coasts.

The turnover of birds at a site is largely determined by its carrying capacity (reflected in the numbers of birds present at any one time) together with its attractiveness. The site may be more attractive than all the alternative foraging sites, in which case shooting there produces a 'sink' effect so that, in the above example, 95 birds out of the overall population of 100, must be shot before the numbers there can be reduced to half the former level. This situation is thought to occur at unprotected cage-farms where serious efforts to reduce predation by herons and cormorants have failed (Carss 1994).

The turnover of birds is also influenced by season because populations are less settled at some times of year than at others. For example, in spring and early autumn goosander counts on the Dee were more variable than at other times of year associated with the influx of breeding birds into the watershed in April and the September dispersal of juveniles (Marquiss & Duncan 1994a). Goosander populations are least mobile in midsummer because many full-grown birds are moulting and ducklings are unable to fly. Similarly, red-breasted merganser populations in fresh waters are probably most volatile in April prior to their settling to breed in May (Marquiss & Duncan 1993).

Shooting sawbills can only be justified if it reduces bird use of a specific site in a cost-effective way. Cost effectiveness is influenced by the number of birds to be killed and so must be considered within the context of three main factors - the carrying capacity of a site, its attractiveness and the potential population from which shot birds might be drawn. These factors are dependent on the characteristics of a particular site relative to adjacent ones, and on the overall size of the bird population.

An additional consideration must be the risk that the process of removing birds encourages turnover so that many birds that would not otherwise use a site will be drawn in and killed. Fish-eating birds have value as part of the nation's natural heritage and its biodiversity, so they cannot be assumed to have zero value when evaluating the cost-effectiveness of shooting activities. Theoretically, this would be of less concern if birds could be caught and released some distance away instead of being killed. However, this is almost certainly impractical in the present circumstances because sawbill ducks are difficult to catch even under optimal circumstances and would have to be moved substantial distances to prevent their return. Sawbill ducks commute over 20 km on a daily basis between foraging sites and overnight roosts (Marquiss & Duncan 1994b) and move much greater distances (100s of kilometres) during migration (Little & Furness 1985) and cold weather movements (Chandler 1981).

Manipulating foraging conditions

For each duck, a foraging site represents a trade-off between the rate at which it can get food and the risks to both life and future breeding success associated with its presence at the site. Sawbill foraging is probably most influenced by the abundance of their potential prey, and whether these fish are easy to see and catch (Chapter 2). In turn, this is related to fish habitat - fish foraging conditions and the amount of cover or refuge offered by the substrate (reviews in Milinski 1986, Mittelbach 1986, Dill 1987).

Where fish are vulnerable to sawbill predation, the attractiveness of a site might be changed by manipulating fish abundance or habitat, to decrease fish availability to ducks at that site, or conversely to increase fish availability at a nearby site. This may be practical to protect small waterbodies where fish density and habitat might be easily manipulated but it is difficult to imagine ways in which natural fish populations could be protected by manipulating foraging conditions on a river catchment basis. The amount of cover for fish could be 'improved' by increasing riparian vegetation, preventing bank erosion and providing undercut banks, or boulder substrates. Wood & Hand (1985) found that goosander predation on juvenile Pacific salmon was less in experiments providing such cover. In Scotland, such 'improvements' may be more applicable to trout habitat, as juvenile Atlantic salmon tend to inhabit riffles.

The provision of alternate foraging sites where less valuable fish are easier to catch might lure sawbills from sensitive sites in the short term, but there is a longer term risk of increasing the number of birds in the area to the potential detriment of wild fish stocks. Another possibility is to provide 'buffer' foods by introducing stocks of less valuable and more easily caught prey (Barlow & Bock 1984, Mott & Boyd 1995). However, the introduction of alternative 'exotic' forage fish is not to be recommended as many fish introductions have had dramatic, often deleterious, effects on natural fish communities (Maitland & Campbell 1992).

The concept of reducing site attractiveness by manipulating duck foraging efficiency seems to have little to offer in practical terms but it nevertheless suggests artificial situations that should be avoided. Herons, cormorants and sawbills are attracted to the very high densities of small fish in rearing facilities and cage-fish farms so these sites need to be protected by antipredator nets (Carss 1989, 1993b, 1993c). Releasing large numbers of naive hatchery fish can also attract cormorants (Kennedy & Greer 1988) and goosanders (Wood 1986). Also, marking fish with Carlin tags can make them vulnerable to predation, probably because they are easier to see and to catch (Feltham & MacLean 1996). The loss of hatchery fish to sawbill predation could defeat attempts to artificially enhance wild fish populations through stocking and might also lead to artificially high predation of wild fish in the vicinity (Wood 1986).

Factors that influence sawbill foraging efficiency include the need for birds to be vigilant of their predators. Disturbances can be distracting and much effort has gone into devising mechanisms to scare birds from small sites (reviewed in Mott & Boyd 1995). Some of these do work, but rarely for long because birds habituate to regular disturbances unless they are endorsed by chasing and shooting to kill persistent individuals. The presence of people in itself is not necessarily a deterrent to sawbills because in towns and villages, unmolested sawbills continue to forage apparently oblivious to walkers and their dogs.

The principles discussed for shooting also apply to scaring. The amount of effort needed to reduce the use of an area by sawbills is not necessarily related to the number of birds in the area, but will be proportional to the attractiveness of the area, and the size of the population of birds that will be attracted to it. There is also the additional consideration that scaring birds increases their time spent flying and consequently their daily food intake (Gremillet et al. 1995). We found no published studies of experiments to scare sawbill ducks from rivers.

Case Studies of Sawbill Shooting

CANADIAN RIVERS

In Canada, the numbers of goosanders (common mergansers) have been reduced on rivers on several occasions in attempts to assess their impact on stocks of juvenile Atlantic salmon. None of these field trials showed that killing sawbills increased salmon harvest, mainly because of poor experimental design or implementation (reviewed in Marquiss & Carss 1994, Russell et al. 1996). Nevertheless, they are examples of how the reduction in sawbill numbers has been conducted.

The Margaree River, 1937-8 (White 1939)

From the spring of 1937 to July 1938, common mergansers were shot on the Forest Glen Brook (about 6km), and on the adjacent tributaries and mainstem of the Northeast Margaree River in Nova Scotia (White 1939). Wardens patrolled the brook and all fish-eating birds, including 6 common mergansers, were shot. An additional 50 mergansers were killed on adjacent waters and 120 on the main river downstream. Many of the birds were shot before they could breed, though females with broods were shot and their ducklings netted and drowned.

The effect of killing was said to be a "fairly complete prevention of their feeding on Forest Glen Brook, and incidently a great reduction in their feeding on a large part of the Northeast Margaree". No details were given of the effort required to kill the birds or the cost of the operation. There were no systematic counts before killing started, but it was stated that in the previous year there had been a single female with about a dozen young, and that a warden had reported flocks several miles upstream. In the previous August, a fish survey had not located any birds, or any sign of their droppings. After the duck kill a survey was conducted to assess fish but there was no mention of birds. Flocks of common mergansers had been seen on the adjacent mainstem two years previously, but only one brood was on the same stretch after bird control. In the absence of systematic count data, it was difficult to know whether bird control was effective.

The Petitcodiac, 1947-53 (Elson 1962)

From 1947 to 1953, a field trial was conducted on the Pollett River, a 58 km tributary of the Petitcodiac in New Brunswick (Elson 1962) to eliminate predation on juvenile salmon by shooting and trapping common mergansers (and kingfishers) on a 16km stretch where salmon stock enhancement was taking place. Prior to the bird kill there was no systematic recording of birds, but casual records listed 2-6 common merganser broods and up to 6 females each year, 1942-46. Flocks of mergansers were also seen on open water sections in the winter of 1944 and in September and October 1946.

Bird control was started in late June 1947 along the length of the Pollett River, involving shooting by two pairs of hunters, one pair driving ducks towards the other. In subsequent shoots to 1953, the areas where shooting occurred included the 16km study stretch and neighbouring upstream and downstream sections. In the first year, all ducklings were shot or netted and drowned, and in subsequent years persistent shooting around the study area prevented broods settling there, so predation by unfledged broods was eliminated. Outside the summer, shooting involved two hunters patrolling the riverbanks on foot, at least twice a week and more often when occasions demanded.

Each year, between 77 and 428 ducks were seen and between 20 and 99 killed. Altogether 364 ducks were killed; this was about 25% of the sightings and, according to the author, probably about half of the ducks visiting the area. Elson believed the shooting had an additional, though unquantified, effect in scaring birds so that the reduction in fish predation was possibly greater than what would be expected from a decline in ducks to half of what it would otherwise have been. As the fish stock enhancement programme progressed, the numbers of juvenile salmon increased and there was a parallel increase in duck sightings, but both salmon numbers and duck sightings decreased in 1952/3 as the enhancement project ended. The shooting efforts were apparently able to cope with this fluctuation in duck numbers because more birds were shot as more were seen.

The study site was obviously attractive to ducks and, taken at face value, the level of shooting applied seemed to be able to reduce duck abundance to about a half of what it otherwise might have been. However, there were no measurements from unshot areas so there is no means of quantifying this properly. Erskine (1972) reported that shooting on the Pollett river continued from 1947 until 1965, and that P. F. Elson had told him that the area had regained its former population of common mergansers within 4 years.

The Margaree River 1957-68 (Erskine 1972)

An even longer experiment in sawbill reduction took place on the Margaree River system during the 1960s. The effects of shooting on duck populations were documented (Erskine 1972) but no other results published. The breeding population was estimated at 15 pairs of common mergansers and 2 pairs of red-breasted mergansers from 1960 to 1962, and between 15 and 17 common merganser broods were produced annually from 1957 to 1962.

Shooting took place from August 1962 to 1968, and was conducted year-round by a Fisheries Research Board shooting crew of six staff who were active 20 days a month, visiting much of the river at least once a week. Intensive shooting had to be maintained throughout the study period. In the first year 223 birds were shot and in the second year 128. Thereafter there was no evidence of a decline in the numbers shot annually (92 to 168 per annum). Each year the numbers of birds shot represented about 18% of bird sightings (range 16% to 24%) and there was no significant trend in this figure, suggesting a relatively even shooting success rate. About 75% of the ducklings seen were killed (range 40% to 91%).

Such systematic intensive shooting apparently reduced the breeding population of common mergansers by about 90%, to 1 or 2 pairs and only a single brood each year from 1965 to 1968. Numbers of transient migrants and wintering birds were also shot and their "numbers reduced probably by 75% or more". The effect was to reduce the merganser use of the area after shooting (1963-68), by about 90% compared with before shooting (1957-61). Ducks on adjacent rivers were counted and their numbers seemed unaffected, suggesting that the decline in numbers on the Margaree could be directly attributable to shooting. Observations were not continued after 1968 but subsequent casual records from fisheries wardens suggested the population was still below its former level three years later.

The Miramichi River 1950-61 & the St Mary's River 1953-67 (Erskine 1972)

Two other merganser elimination experiments took place in Canada at the same time, on the Miramichi in New Brunswick and on the St Mary's River in Nova Scotia. These experiments were planned to investigate whether the shooting of mergansers increased the subsequent numbers of returning grilse and salmon. The results were not published but they were described briefly by Erskine (1972).

The St Mary's River experiment apparently obtained little usable data on either mergansers or salmon. The Mirimachi experiment failed because another agency sprayed the watershed with DDT to combat spruce budworm. Mergansers had been shot, but the spraying eliminated invertebrates and young salmon so the experiment was aborted. Merganser shooting had stopped by 1961 but few mergansers were seen afterwards presumably due to the lack of fish; only two broods were recorded on the whole system in 1968, and none in 1969.

The Restigouche, 1982-5 (Anderson 1986)

From 1982 to 1984, common mergansers were shot on the Restigouche River system (New Brunswick and Quebec), in an experiment to quantify the impact of predation by ducks on Atlantic salmon catches. The river catchment comprised 840 km of stream, but ducks were shot on only 173 km of the wider sections; ducks were deliberately not shot on 71 km of similar habitat. There was no detectable effect of duck predation on salmon parr populations. Furthermore, any effect on subsequent salmon harvest became untestable as the methods of harvest changed in the course of the experiment. Nevertheless the results of shooting on duck abundance were informative.

Mergansers were present on the river each year from late April to late November, and were shot on experimental sections from June to October 1982-4, with increased effort each year in the hatching period, from mid-June to mid-July. In 1982, 465 ducks were killed on the mainstem, 105 in 1983 and 162 in 1984. On one tributary, 129 were killed in 1983 and 64 in 1984, and on another tributary 29 were killed in 1984. Birds were killed by some members of the study team and by wardens of the Restigouche Riparian Association, but methods were not detailed in the report so effort was not quantified.

Altogether, 954 common mergansers were killed over the three years and the spring population on the whole river system fell from about 205 in spring 1982, to 132 in 1983 and about 98 in 1984. Including censuses from other times of year, annual average duck counts fell from 255 in 1982, to 222 in 1983, and to 128 in 1984. Most of this decline occurred on the shot areas (from 140 to 54 birds), but the bird count also declined on the unshot areas (from 65 to 43). This was possibly a result of natural variations in counts but alternatively could have been due to shooting in one area affecting duck numbers in adjacent areas. The same argument was used to explain a similar inconsistency in the decline in brood numbers. In shot areas, brood numbers fell from 35 in 1982 to 14 in 1983, then to 10 in 1984,whereas on unshot areas numbers fell from 23 in 1982, to 22 in 1983, and to 18 in 1984.

No statistical tests were applied to the data, but taken at face value, it could be claimed that shooting over the three years reduced the merganser spring population on the shot areas by at least 27%, and at most 61%. Similarly, broods might have been reduced by between 49% and 71%. Thus in a labour intensive effort to eliminate ducks from more than half of the river, about 950 ducks were killed over three years to reduce spring bird numbers by less than 100, and to reduce duckling numbers by less than 125.