Fisheries Management and Ecology Fisheries Management and Ecology, 2008, 15, 189–198
Reducing fish losses to cormorants using artificial fish refuges: an experimental study I. RUSSELL Cefas, Lowestoft Laboratory, Pakefield Road, Lowestoft, Suffolk, UK
D. PARROTT Central Science Laboratory, Sand Hutton, York, UK
M. IVES & D. GOLDSMITH Cefas, Lowestoft Laboratory, Pakefield Road, Lowestoft, Suffolk, UK
S. FOX Central Science Laboratory, Sand Hutton, York, UK
D. CLIFTON-DEY Environment Agency, Kings Meadow House, Kings Meadow Road, Reading, UK
A. PRICKETT & T. DREW Central Science Laboratory (CSL), Sand Hutton, York, UK
Abstract The hypothesis that the introduction of artificial refuges might provide protection for fish and reduce the level of cormorant predation was tested in two, paired-pond, cross-over trials during the winters of 2003 and 2004, using a ÔrefugeÕ pond and an adjacent equivalently stocked ÔcontrolÕ pond. There were 77% fewer cormorant visits to the refuge pond than the control pond, on average. There was also a 67% fall in the mean mass of fish consumed per cormorant visit and 79% less fish mass lost in the refuge pond. The results are discussed in the context of interactions between cormorants and fish and the potential use of the tool in fisheries management. KEYWORDS:
cormorant, freshwater fish, habitat, management, predation, predator–prey interactions.
Introduction Due to recent large increases in cormorant populations, interactions between these fish-eating birds and fisheries have a high profile in many parts of the world, including Europe (Carss 2003), Japan (Kameda, Ishida & Narusue 2003) and North America (US Fish &
Wildlife Service 2003). This has highlighted the need for effective management tools to address the resulting conflicts (e.g. Mott & Boyd 1995; McKay, Furness, Russell, Parrott, Rehfisch, Watola, Packer, Armitage, Gill & Robertson 1999). In the UK, conflicts between cormorants and fisheries mainly occur on inland, freshwater sites as a
Correspondence: Ian Russell, Centre for Environment, Fisheries and Aquaculture Science (Cefas), Lowestoft Laboratory, Pakefield Road, Lowestoft, Suffolk NR33 0HT, UK (e-mail:
[email protected]) Cefas and CSL are Executive Agencies of the Department for Environment, Food and Rural Affairs (Defra)
2008 The Authors. Journal compilation 2008 Blackwell Publishing Ltd.
doi: 10.1111/j.1365-2400.2008.00600.x
190
I. RUSSELL ET AL.
consequence of both a rise in cormorant numbers (Bayliss, Austin, Musgrove & Rehfisch 2005), changes in the seasonal distribution and range extension of the birds (Kirby, Gilburn & Sellers 1995) and the development of inland breeding colonies (Sellers, Ekins, Hughes & Kirby 1997). The inland breeding population, currently about 2100 pairs in England (Newson, Marchant, Ekins & Sellers 2007), is increasing at about 19% per year (i.e. roughly doubling every 5 years) (Newson, Ekins, Hughes, Russell & Sellers 2005). However, conflicts mainly occur during the winter months (November to March) when cormorant numbers at inland sites are the highest (Hughes, Bevan, Bowler, Still, Carss, Marquiss, Hearn & Bruce 1999). The population of cormorants wintering in Britain is currently estimated at about 30 700 birds (Jackson, Austin & Armitage 2006). The susceptibility of fish to predators is affected by factors ranging from their size and shape (de Nie 1995), swimming speed and shoaling behaviour (Pitcher 1986; Magurran 1990a; Videler 1993) to their use of aquatic habitat features (Savino & Stein 1989a,b; Lovvorn, Yule & Derby 1999). Indeed, the use of aquatic vegetation and other submerged structures by many freshwater fish species is widely regarded as an adaptation to reduce the risk of predation by piscivorous fish (Persson & Eklov 1995; Jacobsen & Perrow 1998). Prey accessibility and prey density have been shown to influence the foraging success of piscivorous birds, and consequently the selection or abandonment of individual feeding sites (Gawlik 2002). Thus, underwater habitat complexity is also likely to influence the predator–prey dynamics between fish and piscivorous birds. However, few studies have considered this in relation to the foraging success of cormorants, and the concept of manipulating underwater habitat to protect fish from avian predators appears to be a relatively recent idea (Mott & Boyd 1995). As the habitat available to fish in inland waters in the UK is at its lowest level in the winter months, when aquatic vegetation dies back, and at a time when water temperatures and hence fish swimming speeds are the least, but cormorant numbers are the highest, such a technique may be particularly applicable at this time. It has therefore been suggested that artificial fish refuges have the potential to reduce fish losses caused by cormorants feeding at inland fisheries (McKay et al. 1999; Russell, Dare, McKay & Ives 2003). This investigation tested the hypothesis that the presence of artificial refuges in an experimental pond would reduce cormorant foraging success, fish losses
and the number of foraging cormorants compared with an adjacent, identical control pond. Materials and methods Study site
Trials were conducted in two identical, adjacent, drainable ponds located at a disused water treatment works near Reading, southern England. The ponds measured 45.0 m · 27.2 m (i.e. 0.12 ha each), with vertical sides and a uniform water depth of 1.35 m. Cormorants also used a number of neighbouring ponds and rivers. The trial ponds were drained prior to the onset of the trials to ensure they were free of aquatic plants and any other items that might have provided cover for fish. Experimental design
The trials involved a cross-over, paired-pond, experimental design, with artificial fish refuges placed in one of the ponds, whilst the second pond had no refuges and acted as a control. Two sequential trials were conducted in 2003 and 2004, with the refuges moved between ponds between trials, i.e. one cross-over experiment per year. Switching the refuges between ponds controlled for any potential bias in pond selection by cormorants independent of refuge presence/absence. Trials were completed between January and March each year, coinciding with the period of peak inland cormorant occupancy in England and Wales (Hughes et al. 1999). The timing of the trials was broadly similar in the 2 years, although the duration of individual trials varied between 24 and 35 days, dependent on the level of cormorant activity at the site (2003: trial 1: 18 January–10 February, trial 2: 14 February–20 March; 2004: trial 3: 10 January–3 February, trial 4: 7 February–8 March). The refuges were made from a number of individual cage units, each measuring 2 m · 2 m · 1.2 m high, and incorporating overhead shading, internal structure and an overlay of light-gauge stock fencing (Rylock light fencing L8/80/15) around all four sides and the top. The internal structure was made up of 4–6 small conifers (Picea abies L.), dependent on their size, with each tree weighted with a brick. The stock fencing was composed of differing mesh dimensions reducing progressively from 15 cm · 15 cm square mesh on one edge of the roll to rectangular mesh measuring 15 cm · 6.5 cm on the other edge. Twelve refuge cage units were deployed in each trial, grouped in two
2008 The Authors. Journal compilation 2008 Blackwell Publishing Ltd.
USE OF ARTIFICIAL REFUGES TO REDUCE FISH LOSSES TO CORMORANTS
discrete blocks of six units at fixed locations within the pond. In total, the refuges represented about 3.5% of the water volume in the refuge pond. The upper surface of the refuge was approximately 15 cm below the water surface. At the start of each trial, the ponds were filled and stocked with roach, Rutilus rutilus (L.), and smaller numbers of perch, Perca fluviatilis L. and carp, Cyprinus carpio (L.). The fish were predominantly sourced from wild populations, although some carp were sourced from a fish farm. These species are widely distributed in the UK and are common in the diet of cormorants feeding at freshwater sites (Russell, Dare, Eaton & Armstrong 1996). Prior to stocking, all fish were anaesthetised (2-phenoxyethanol, 0.1 mL L)1), measured (fork length to the nearest mm) and inspected for damage consistent with cormorant attack (Ransom & Beveridge 1983; Carss 1990). Any damaged fish were rejected. Individual fish mass was calculated using species-specific length/mass regressions derived from large samples of fish, from a range of locations in England and Wales, used for disease testing (Environment Agency, personal communication). Although equivalent numbers and sizes of fish were allocated to the refuge and control ponds within each trial, there were some differences in the proportions of different species and in the sizes of the fish used between trials because of variability in the available stock fish. The mean lengths of the fish, in individual trials, ranged from 11.0 to 13.9 cm for roach, 10.4– 12.6 cm for perch and 8.2–14.5 cm for carp, with minimum and maximum recorded values (all species) of 4.6–21.3 cm. The stocking density of fish in the trials varied from 259 to 459 kg ha)1; these values are consistent with those occurring in recreational stillwater fisheries in England and Wales (Environment Agency 2001). At the end of each trial, the ponds were drained and all remaining fish recovered, measured and inspected for damage. Bird monitoring
During each trial, refuge and control ponds were monitored for cormorant activity and foraging. Monitoring was conducted from a hide situated equidistant between the two ponds that were P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ns not significant (Mann–Whitney U-test; except bouts successful and cormorant visits – FisherÕs exact test). Data from complete bouts only. A complete bout was one where the arrival, departure and all intervening dives were recorded. à Total prey includes fish assumed swallowed underwater (from behavioural cues) in addition to fish consumed at surface. Table 2. Comparison of the observed number of cormorant visits to the control and refuge ponds in each trial, and the difference in visits to the refuge pond relative to the control pond Pearson chi-squared test
Observed number of cormorant visits Year
Trial
Control
Refuge
% change
v2
2003 2003 2004 2004
1 2 3 4
43 44 77 217
12 3 25 56
)72.1 )93.2 )67.5 )74.2
17.47 35.77 26.51 94.95
Mean
P < < < <
0.001 0.001 0.001 0.001
)76.8
Estimates of the number of fish lost in each pond, based on prey capture rates from focal bird monitoring and estimated total numbers of cormorant visits, were consistently lower than actual losses (apart from the refuge pond in trial 3). The total fish (observed and underwater) caught by focal birds was 83 and 15 in the control and refuge ponds, respectively, during trial 1; with comparative values of 70/10, 151/23 and 162/17 for trials 2, 3 and 4 respectively. These values represented 49%/17%, 16%/11%, 45%/127% and 82%/ 62% of the actual numbers of fish lost in trials 1–4 respectively. Environmental data and other predators
between refuge and individual species, i.e. refuge presence did not reduce losses in one species more than any other. In 2004, there was no effect of fish species on levels of stock loss. Prey consumption per cormorant visit was also compared given the marked difference in the total number of cormorant visits between control and refuge ponds. The mass of fish consumed per cormorant visit, based on estimated total cormorant visits during the entire trial and the biomass of the missing fish, ranged from 138 to 530 g per cormorant visit (mean 314 g) in the control pond and between 51 and 195 g per cormorant visit (mean 102 g) in the refuge pond. The estimated mean mass of fish consumed per cormorant visit was 67% lower in the refuge pond than the control pond (Table 4).
In trial 3, water turbidity was significantly lower in the refuge pond than the control pond (Mann–Whitney U-test: U43,21 = 294, P = 0.024. No other significant differences were detected between the refuge and control ponds for any abiotic factor (ambient light, air temperature, water temperature and water turbidity) that might have influenced cormorant foraging success (P > 0.05) (Table 6). Herons were observed to visit the site during some trials and gulls (Larus spp.) also regularly used the site. In 2003, eight heron visits were recorded during trial 1 and five during trial 2, but no prey capture was observed. In 2004, no herons were recorded at the ponds during trial 3, but six heron visits were recorded during trial 4. A total of four fish was observed being taken by herons during these visits (two each from the
2008 The Authors. Journal compilation 2008 Blackwell Publishing Ltd.
193
194
I. RUSSELL ET AL.
Table 3. Number and percentage of fish ÔlostÕ in the control and refuge ponds in each trial Fish ÔlostÕ – control pond
Fish ÔlostÕ – refuge pond
Year (trial)
Fish species
n
%
n
%
2003 (1)
Roach Perch Carp Total Roach Perch Carp Total Roach Perch Carp Total Roach Perch Carp Total
622 129 274 1025 821 125 412 1358 972 52 98 1122 1194 118 142 1454
42.9 100 100 55.3 89.3 94.0 99.5 92.6 96.2 68.4 87.5 93.7 99.5 96.7 98.6 99.2
61 51 80 192 18 48 120 186 0 32 13 45 60 49 15 124
4.2 39.5 29.1 10.5 2.0 36.6 28.9 12.7 0 42.7 11.5 3.7 5.0 40.5 10.5 8.5
2003 (2)
2004 (3)
2004 (4)
Table 4. Comparison of the relative loss of fish and the estimated mass of fish consumed per cormorant visit between control and refuge ponds for each trial
Total mass of fish lost (as % of stocked mass)
Year 2003 2003 2004 2004
Estimated mass of fish consumed per cormorant visit (g)
% % Trial Control Refuge change* Control Refuge change* 1 2 3 4
Mean
58.4 91.9 94.2 99.8
6.7 8.5 4.6 9.7
)51.7 )83.4 )89.6 )90.1
250.9 529.5 336.9 137.7
)79
109.8 195.0 50.6 52.5
)56.2 )63.2 )85.0 )61.9 )67
*Refuge pond compared with the control pond.
control and refuge ponds). Despite being regular visitors, gulls were not observed to predate on fish from the ponds. No other avian or mammalian (e.g. otter and mink) fish predators were observed at the site. Discussion The behavioural indicators predicted to change if cormorant foraging efficiency was reduced in the presence of fish refuges were readily apparent during the paired-pond trials. Foraging success was consistently lower in the refuge ponds than the control ponds. The differences in foraging parameters between
Table 5. Results of generalised linear model (logit link) testing for an effect of refuge, pond and fish species on levels of lost stock, for trials conducted in (a) 2003 (trials 1 and 2) and (b) 2004 (trials 3 and 4)
Parameter (a) 2003 Pond Refuge Species Refuge · species Residual Total (b) 2004 Pond Refuge Species Refuge · species Residual Total
d.f.
Deviance
Mean deviance
Deviance ratio
P
1 1 2 2
70.88 2990.91 1168.58 24.18
70.88 2990.91 584.29 12.09
1.69 71.27 13.92 0.29
0.250
