Fish Eating Birds and Salmonids in Scotland

5. Potential responses by juvenile salmonid populations to predation
J D Armstrong, R Gardiner & R Laughton*
*Spey District Fishery Board
SUMMARY
The dynamics of salmon populations are briefly reviewed with reference to general population theory and potential impacts of predators. It is concluded that in some areas of streams, losses after a critical period during the first summer and autumn of growth may not be compensated. Even so, it is difficult to estimate the overall impacts of predation on parr because of uncertainties about the relationships between carrying capacities and sizes of fish.
Predation on smolts may cause a near proportional reduction in numbers of returning adult salmon. However, the impact on a fishery is likely to depend on characteristics of the smolts such as size.
Movements within populations of salmon parr are important for two reasons. First, they may, under some circumstances, increase the scope for populations to compensate for losses. Second, if levels of movement are sufficiently low and the home range of salmon parr is sufficiently small then it may be possible to study the effects of predators on populations by replicated field trials.
Experiments to determine movements of salmon parr are described. Movements were measured in terms of their potential to recolonise areas of stream in which populations had been artificially and severely depleted. Communities of fish were depleted at eight sites during summer months in Scotland. Salmon parr densities in the depleted sites were sampled 28-43 days after manipulation and compared with sites that had not been depleted. The mean maximum recolonisation was estimated as 13.8% (95% limits: 6.6-20.9). Because there are currently no robust published data on the home ranges of salmon parr it is not possible to calculate more realistic levels of recolonisation. An index of recolonisation that accounted for changes in salmon parr densities in undepleted sites (termed "adjusted recolonisation") was 21.1% (2.0-39.1).
The levels of immigration are probably sufficiently low that large-scale exclusion experiments may be considered as a practical means of directly estimating the impact of sawbill ducks on populations of salmon parr in sections of some rivers during summer.
INTRODUCTION
In the previous Chapter simple models were used to predict the numbers of salmon and trout consumed by goosanders and red-breasted mergansers on the Rivers Dee and North Esk. In order to put these numerical estimates into perspective, they were compared with estimates of fish abundance from the same rivers. The calculated proportions of salmon standing stock consumed by sawbills on these two rivers were usually considerably lower than 1% per day but occasionally higher (Chapter 4). Such comparisons allow us to estimate the numbers of fish eaten by birds in relation to the standing stock of prey. However, they do not on their own allow us to assess the impact of predation on populations of fish because predator-prey interactions can be complex (Russell et al. 1996). To move from estimates of consumption to estimates of impacts of predators it is necessary to consider the population dynamics both of predators and prey. We begin with a brief review of some of the key ideas and concepts associated with predator-prey relationships drawn from general texts (e.g. Cherrett 1989, Crawley 1992, Begon et al. 1996) as well as specific works on fishes (e.g. Pitcher 1986, Wootton 1990). This is followed by a short review of the potential for juvenile salmonids to compensate at the population level for losses caused by predation. Finally, we report on a series of field experiments undertaken to investigate movements within populations of wild salmon parr in Scotland. Understanding movements of fish is a key precursor to developing a sound understanding of the likely responses of their populations to predation.
SOME BASIC CONCEPTS
The size of any population of any species is determined by the balance between gains brought about through birth and immigration and losses through death and emigration. Four specific terms are associated with the relationship, or lack of it, between population density (i.e. the number of individuals per unit area) and the rates of change of these demographic processes - birth, death, immigration and emigration.
If the rates of either gains or losses expressed as a proportion of the population are unrelated to population density, processes are said to be density independent. If, as population density increases, the proportional rate of gain decreases or the rate of loss increases, these processes are said to be density dependent. In addition, if the reverse is true and rates of gain increase and rates of loss decrease as density increases, the processes are said to be inversely density dependent.
Ultimately, populations cannot increase unchecked. In general, populations are thought to become constrained by their environment at some upper limit, the so-called carrying capacity. This involves both density-dependent and density independent factors. For example, territory size has been identified as a factor that may limit the populations of several species of salmonid fishes (Grant & Kramer 1990, Keeley & Grant 1995). A population is said to be at equilibrium if gains are equal to losses. If for any reason a population is shifted from its equilibrium, only density dependent processes can act to return it to equilibrium through a process termed regulation.
In practice, a number of processes (both density-dependent and density-independent) can act on a population at any one time. The relative importance of each of these processes is likely to vary in both space and time, thus altering the position of the equilibrium density both within and between populations. Furthermore, these processes are likely to affect specific populations differently depending on when they occur in the animal's life-cycle.
NUMERICAL AND FUNCTIONAL RESPONSES
In order to determine the effects of predation on prey populations, it is important to understand the way in which predators respond to changes in prey density. As prey density varies (at a specific site or generally over some defined geographical area), the numbers of predators are likely to vary too because higher numbers of prey can be expected to support higher numbers of predators. Changes in the number of predators in response to the number of prey is termed a numerical response. There may also be changes in the number of prey eaten per predator as a result of variation in prey density; this is termed the functional response. There are at least three theoretical functional responses of predator feeding to changes in prey density, resulting in a variety of possible trends in predation levels ('percentage kill' - the proportion of the population eaten). At it's simplest (Type I), a straight-line relationship between prey density and the number of prey taken by a predator will result in a constant percentage kill. An asymptotic relationship between prey density and the number encountered, indicating a limit to the number of prey eaten per predator (Type II), results in a declining (i.e. inversely density-dependent) percentage kill. A sigmoid (S-shaped) relationship where foraging is inefficient at low prey densities but increases to a limit thereafter (Type III) produces direct density-dependence in percentage kill at low prey densities - as prey density increases so too does percentage kill - but inverse density-dependence at high prey densities because percentage kill declines.
Although such functional responses are hypothetical there is good evidence for their existence in nature, albeit from studies largely focusing on invertebrate predators. These relationships provide us with a useful conceptual framework for considering predator-prey interactions and suggest that determining the functional response for any particular interaction is likely to be important as each will have different effects on the dynamics of the populations of predators and prey concerned.
OTHER FACTORS THAT MAY AFFECT PREDATION
Considering predator-prey interactions within the functional response framework leads on to predictions that several factors besides densities of predators and prey are important. Predators may be specialists, consuming a single, or small number of prey types, or generalists feeding on a wide variety of prey. Generalist predators may show a weak numerical response to changes in density of a particular prey compared with a specialist, preying almost exclusively on that prey. In theory, generalist predators could have a considerable impact on their preferred prey species because their numbers are buffered to some extent against declines in these prey by their ability to switch to other types. It is clear that studies of predation should therefore not merely consider the preferred prey in isolation but should also consider other acceptable prey.
A further important consideration is that several species of predator may feed on a given species of prey. Removing the predation pressure exerted by one predator (e.g. by reducing its numbers) may lead to compensated mortality, an increase in predation from other predators.
Simple hypothetical functional responses assume that individual prey are equally vulnerable to predation. This may not be the case in natural situations where predators may select old, young or sick prey which are easier to catch because of their size, lack of stamina or reduced vigilance. Such individuals are unlikely to make a significant reproductive contribution, so the effect of predation on them is less likely to affect the overall population dynamics of prey. Conversely, predators may select individual prey that are particularly conspicuous when they are feeding actively or breeding. Had they not been eaten, such individuals may have made a disproportionately large reproductive contribution. In many cases, individuals may have access to refuges from predator attack. However, it is unlikely that all prey will have equal access to refugia and there is evidence to suggest that in some animal populations under conditions of intra-specific competition, it might be the 'fittest' (e.g. territory holders) which do. Under these circumstances, those individuals making the biggest reproductive contribution to the prey population may be the least vulnerable to predation. Conversely, in species that rely heavily on crypsis to avoid predation, prominent territory holders may be most vulnerable.
THE SCOPE FOR COMPENSATION IN JUVENILE SALMONIDS
It might be expected that in the presence of predators annual mortality of prey would increase and lead to a lower equilibrium density. However this might not always be the case, particularly wherever prey density is high enough for intra-specific competition to occur. In populations at carrying capacity, growth of individual fish appears to be associated with a general increase in their requirement for space and with a decrease in local density (Grant & Kramer 1990) in a process termed self-thinning. In such populations, 'out-competed' fish must either move (and so potentially may colonise depleted areas) or die in situ. If populations of salmon and trout do self-thin, then losses to predators in any given area may, to some extent, be compensated either by immigration and/or increased growth and/or survival of those fish that remain (see Grant & Kramer 1990) through the resulting reduction in intra-specific competition.
Relationships between mean fish weight and population density consistent with self-thinning have been reported for populations of salmonids in their first year of life stocked in flume tanks (Grant 1993). Inverse relationships between weight and density reported for many wild populations of salmon and trout (Bohlin et al. 1994) lend support for the occurrence of self-thinning but do not indicate at what stage of the life-cycle such thinning may take place (Armstrong 1997).
Although Elliott (1993) concluded that long-term data for brown trout in an English Lakeland stream were consistent with the idea of the population self-thinning throughout the freshwater stage of the life-cycle, a re-analysis (Armstrong 1997) has suggested that, for most cohorts, density-dependent mortality was likely only during a 'critical period' early in the first summer of life. Armstrong (1997) argued that a change from density-dependent to density-independent mortality might be due to the carrying capacity effectively increasing as trout grow during their first summer of life. This could be because larger fish are able to use a greater proportion of the stream area. An analysis of data from long-term studies of the Shelligan Burn (Gardiner & Shackley 1991) shows that the critical period for Atlantic salmon parr may extend into the autumn of the first year of life. There is little growth but continuous mortality of both salmon and trout during winter (Elliott 1993, Egglishaw & Shackley 1977) so fish numbers in spring may be low in relation to the carrying capacity, which appears to be determined mainly by size of fish and physical structure of the stream bed (Grant et al. in press, Gardiner, unpublished). Such a situation would explain an absence of density-dependence of growth and mortality of parr in their second year of life and older (Armstrong 1997) and implies that losses of these older fish to predators may not be compensated.
The concept of self-thinning is important in assessing impacts of predators on salmon and trout parr because it emphasises that a stream that may produce a given number of smolts may hold much larger numbers of smaller parr. The impact of a predator should be assessed as the difference between the actual number of smolts produced and the potential number produced (in the absence of the predator) rather than the number of parr consumed.
Given the crucial nature of density-dependent factors in regulating populations and compensating for such things as predation, it is important to determine at what stage in the fishes' lives these factors operate. Although there is at present no strong evidence for density-dependent mortality of salmonid parr of one year and older in wild populations of these fishes in the UK, the possibility that it occurs in some years in sections of some rivers and streams should not be ruled out (Armstrong 1997). Hence, compensation through reduction in intra-specific competition in sections of some rivers and streams similarly cannot be ruled out.
There is no evidence for strong density dependence in the survival of salmon at sea (e.g. Hansen et al. 1996, Crozier & Kennedy 1993, Gibson 1993). Therefore any predation on salmon from the time they leave fresh water as smolts might be expected to result in a proportional reduction in the number of adults returning to the fishery. Considering that birds take mainly smaller than average-sized smolts, the relationship (and any evidence for density-dependence) between smolt output and adult return needs to be examined for smolts of various sizes to develop accurate models of the impacts of such predation.
EXPERIMENTAL INVESTIGATIONS INTO RECOLONISATION BY SALMON PARR
Because most of the salmon taken by birds are parr of one year and older (see Chapters 3 & 4), and because the dynamics of salmon populations may be complex at this stage of their lives, it is difficult to derive estimates of the impacts of predators. Such impacts depend on whether populations can compensate for losses and on how carrying capacities, in terms of densities of parr, change as they grow. Compensation may occur locally if there is density-dependence of mortality and/or growth and/or immigration. Estimates of movements of fish within populations are needed to establish what potential there may be for local compensation by immigration into regions of river that become depleted. If there is relatively little movement of salmon parr within populations, then it may be possible to devise replicated experiments to study compensatory growth and mortality and to estimate directly impacts of predators in natural river systems. It is important to realise that the occurrence of movement is not on its own evidence of compensation for predation because propensity to move may not be density-dependent.
The degree of mobility of salmonid fishes is likely to vary with time of year. Movements of salmonids might occur in association with (1) fry dispersal (e.g. Gustafson-Greenwood & Moring 1990); (2) increased intraspecific competition (Jenkins 1969, Hesthagen 1988, though see also Heggenes 1988); (3) change in habitat preferences as fish grow (Symons & Heland 1978, Egglishaw & Shackley 1982, Kennedy & Strange 1982, Elliott 1986); (4) on some tributaries, with search for over-wintering habitat (Bjorrn 1971, Rimmer et al. 1983, 1984), and (5) spawning by mature male salmon parr (Buck & Youngson 1982, Garcia de Leaniz 1989) and trout (Stuart 1957).
Several studies suggest that the potential for juvenile salmonid fish to move and recolonise depleted areas of streams may be high. In Alaska, areas of streams that were artificially depleted of salmonids were rapidly recolonised (Bjornn & Kirking 1991). In a Norwegian stream, brown trout showed a tendency to redistribute away from areas of relatively high population density (Hesthagen 1988). In stream tanks, several species of salmonids made occasional exploratory excursions before returning to their focal points (Fausch 1984), and juvenile Atlantic salmon redistributed in response to changing food supply (Symons 1971) albeit at very high population densities. Gowan & Fausch (1996) recorded brook trout moving extensively through streams and questioned whether it was valid to assume that salmonids generally exhibit restricted movements (Gowan et al. 1994). The potential for recolonisation of depleted areas by salmonids has also been suggested from movements detected in mark-and-recapture experiments (e.g. Harcup et al. 1984, Heggenes et al. 1991). However, the process of removing fish in mark-and-recapture studies is likely to disrupt the population structure (Huntingford et al. 1998) and therefore results of such studies are difficult to interpret.
Other studies however suggest a lower recolonisation potential. Brown trout become widely distributed throughout some streams (Elliott 1986, Mortensen 1977, Solomon 1982), but in others, distribution from natal areas is insufficient to use much of the available habitat (Beard & Carline 1991). Hesthagen (1988) found no evidence of movements by an introduced cohort of one-year and older Atlantic salmon from areas of high to low population densities (varying over a three-fold range). Even where movements have been detected, it is not clear that roaming fish would choose to settle in depleted areas. On the other hand, the absence of movements by fish does not imply that individuals would not move if they were adjacent to, and could identify, a severely depleted area.
Because of the between-study variation in estimates of mobility within populations of salmonid fishes, and hence the poor ability to predict likely levels of recolonisation in Scottish streams, we examined recolonisation by salmon parr (here defined as fish of one year and older) directly. We depleted areas in several Scottish streams and rivers during early- and mid-summer, a time when fish grow quickly and when the carrying capacity of their habitat in terms of numbers of fish is likely to decrease. We chose to attempt to remove as many salmonid fishes as possible from areas of stream. Such removals could be expected to give a maximum estimate of immigration because there would be few resident fish to repel potential colonisers through aggressive defence of the manipulated areas. We chose to focus on colonisation by salmon parr because they usually predominate in the diets of birds (Chapters 3 & 4).
The proportion of a population removed during depletion of a site depends on the efficiency of the fishing method. Recent work (Armstrong & West, unpublished) suggests that some individual salmon parr evade capture during depletion but can be caught on some subsequent sampling occasions. Such fish would appear to have colonised and would lead to an overestimate of immigration. During depletion of a site, only a fraction of those individual fish with home ranges overlapping the edges of the site would be caught. Some fish in the fraction that was not captured may be caught during resampling. Such fish would not have colonised in the sense of having established a new home range within the site and will lead to an overestimate of immigration. In the absence of sufficient data regarding evasion of capture and home range of salmon parr we present here maximum estimates of colonisation, which may later be modified to more realistic estimates when further data on fish behaviour are available.
METHODS
Trials were conducted in 1991, 1992 and 1993. In the first of these years, mark-and- recapture experiments showed that few movements between undepleted "comparison" sites in which fish were sampled but returned and depleted sites could be attributed to handling parr (Armstrong et al. 1994). Therefore, comparison sites could be monitored without causing a large increase in immigration into depleted sites. In 1992 and 1993 comparison sites were monitored in tandem with each depletion experiment. We use the term "comparison" rather than control because we had insufficient information on the population dynamics of the sites to infer that they genuinely reflected what would have occurred in the treatment sites had they not been depleted.
Details of the starting dates and estimated fish densities, areas of depleted sites, durations of experiments and numbers of comparison sites are summarised for each site in Table 5.1.
In 1991 study sites were selected on the Clunie Water, a tributary of the Aberdeenshire River Dee; the Banvie Burn, a tributary of the River Tay and Raitts Burn, a mid-river tributary of the River Spey (Table 5.1). Each study site comprised a length of fairly regular habitat dominated by riffle/glide and without large pools. On each site, 60m lengths of stream were isolated by stop nets and fished six times by electrofishing. All captured salmon and trout were displaced to remote locations. Each site was resampled and the parr density was estimated by the Zippin (1958) method. Detailed monitoring of comparison sites was conducted on the Banvie Burn (see also Armstrong et al. 1994) throughout the experiment. Comparison sites were also sampled on the other burns at the end of the experiment.
In 1992 four sites (20m length) were depleted during early June: two on the River Tervie and two on the Banvie Burn. Fish were removed to a remote location. The physical characters of the sites were similar to those studied in 1991. The sites included areas near (<100m, Banvie 3 and Tervie 1, Table 5.1) and remote (>350m, Banvie 2 and Tervie 2) from the confluence of the burn with its mainstem. Each site was depleted by six fishings on each of two successive days. Control sites were sampled at the beginning of trials and whenever the depleted sections were resampled.
In 1993 a square section (10m x 10m) of the main stem of the River Dee was isolated using pegged stop nets and was electrofished six times. All the captured fish were transferred to a remote location. The depleted area was resampled after 28 days. Two control sites (10m x 10m areas, 4.5m from the cleared section) were sampled at the beginning and end of the trials.
Changes in population density were summarised by two indices. The recolonisation index expresses the number of parr during resampling of the depleted area (D2) as a percentage of the number of parr that were removed (D1):
recolonisation index = (D2/D1).100	(1)
An adjusted recolonisation index was also calculated to estimate apparent immigration in relation to "expected" densities at the end of each experiment. Expected densities were estimated from the comparison sites. Where there were data from comparison sites over the duration of experiments:
adjusted recolonisation index = (D2/[{C2/C1}.D1]).100	(2)
where C1 is the mean starting density and C2 is the mean final density in comparison sites. For those sites where there was no estimate for C1 (Table 5.1) C1 was assumed
= D1 and the index was calculated as:
adjusted recolonisation index = (D2/C2).100	(3)