Changes in breeding success and abundance of ground-nesting moorland birds in relation to the experimental deployment of legal predator control


*Correspondence author. E-mail:


1. An 8-year-field experiment on moorland in northern England manipulated the abundance of legally controllable predators whilst maintaining consistent habitat conditions. Subsequent changes in both the breeding success and abundance of five ground-nesting bird species were monitored: lapwing Vanellus vanellus, golden plover Pluvialis apricaria, curlew Numenius arquata, red grouse Lagopus lagopus scoticus and meadow pipit Anthus pratensis and the abundance only of snipe Gallinago gallinago and skylark Alauda arvensis.

2. Control of fox Vulpes vulpes, carrion crow Corvus corone, stoat Mustela ermina and weasel Mustela nivalis reduced the abundance of fox (−43%) and crow (−78%); no changes were detected in already low stoat or weasel abundances.

3. Reductions in foxes and crows led to an average threefold increase in breeding success of lapwing, golden plover, curlew, red grouse and meadow pipit.

4. Predator control led to subsequent increases in breeding numbers (≥14% per annum) of lapwing, curlew, golden plover and red grouse, all of which declined in the absence of predator control (≥17% per annum).

5.Synthesis and applications. Controlling predators is a potentially important management tool for conserving a range of threatened species. Considerable sums of public monies are currently spent on habitat improvement for conservation and some of these public funds should be used to underpin habitat works with predator removal.


Native vertebrate predators can limit the population size of their prey (Messier & Crête 1985; Sinclair et al. 1998). Culling such predators is widely used for protecting domestic stock (Allen & Sparkes 2001) or enhancing game abundance (Potts 1986). The benefits of predator control (PC) for conservation are receiving growing recognition (Gibbons et al. 2007). However, such management can be contentious if predators and prey are valued differently by different stakeholders (Redpath et al. 2004) and the effects of PC must be properly quantified.

Amongst birds, ground-nesting species are particularly susceptible to predation by avian and mammalian predators and negative impacts of predation have been recorded for gamebirds and waterfowl on incubating adults, eggs and chicks (Marcström, Kenward & Engren 1988; Newton 1993). The effects of predators on populations of other ground-nesting species such as waders (Charadriiformes) are less well studied (Hill 1988; Bolton et al. 2007), although predators are known to have a substantial impact on their breeding success (Baines 1990; Grant et al. 1999).

Generalist predators such as foxes and carrion crows have increased markedly in the United Kingdom in recent decades (Tapper 1992; Gregory & Marchant 1996), associated with increased food availability and declines in the number of gamekeepers (Hudson 1995; Fuller & Gough 1999). The impact of increasing predator numbers on prey populations may be exacerbated when coupled with habitat deterioration, which can increase a prey species’ susceptibility to predation (Baines 1990; Evans 2004). There is, however, a need to quantify the benefits of PC on target species in isolation from variation in habitat quality.

Heather Calluna vulgaris (L.) dominated moorland in the uplands of the United Kingdom is of high international conservation importance for a range of bird species (Thompson et al. 1995). Breeding lapwing Vanellus vanellus (L.) and curlew Numenius arquata (L.) have shown widespread population declines in these areas over the last 20 years (Sim et al. 2005). PC is a well-established management technique across large areas of heather moorland managed for red grouse Lagopus lagopus scoticus (Latham) shooting, thus providing an ideal system in which to test whether PC benefits other ground-nesting birds. The predators controlled are primarily foxes Vulpes vulpes (L.) and carrion crows Corvus corone (L.), which combined with rotational strip burning of heather and management of grouse parasites, maximizes the number of red grouse (Hudson & Newborn 1995). In addition to legal PC, illegal killing of birds of prey still occurs on some moors (Etheridge, Summers & Green 1997).

We conducted a large-scale field experiment where foxes, crows, stoats Mustela ermina (L.) and weasels M. nivalis (L.) were controlled to examine responses in bird breeding success and breeding numbers.

Materials and methods

Study sites

Four study plots located in Northumberland, United Kingdom [plot A (centre 55°17′N 2°9′W), plot B (55°19′N 2°18′W), plot C (55°12′N 2°04′W) and plot D (55°15′N 2°25′W)] were 9·3–14·4 km2 in size and consisted of moorland and marginal farmland between altitudes of 220–470 m a.s.l. Each study plot was 6·1–6·9 km from its nearest neighbour. Plots consisted of similar mosaics of habitats of which heather-dominated heath and heath/acid grassland mixtures were commonest at the higher altitudes, with grasses and rushes more extensive on the lower slopes. All plots were grazed by sheep Ovis aries (L.), typically at summer densities of 1·0–1·5 ewes ha−1. Negligible levels of PC occurred either on the plots prior to the experiment or on areas adjoining the plots during the experiment.

Experimental design

The experiment consisted of two treatments, year-round PC and no predator control (NPC). The experiment commenced in 2000 when all plots were monitored for a base-line breeding season with NPC. In September 2000, treatments were implemented at the plot scale using a paired plot approach. Plots A and B formed the first pair, with treatments initially randomized within the pair then switched halfway through the experiment (in September 2004). Plots C and D formed the second pair within which treatments were randomized; the plots then remained under the same treatment in all subsequent years (Fig. 1). Data on vegetation and the abundance and breeding success of birds were collected annually from spring 2000 to 2008. Access restrictions during a foot and mouth disease outbreak limited data collection in 2001 and prevented PC activities on plot C (2001 = NPC for analyses).

Figure 1.

 Design of study: plot-years with no predator control (white) and plot-years with predator control (grey). Access restrictions in 2001 meant that predator control activities were suspended on plot C.

Habitat monitoring

As the aim of the study was to manipulate the levels of predators only, grazing and heather burning were maintained at similar levels throughout the experimental period. Typical strip burning of heather was undertaken by project gamekeepers and the mean fire perimeter and area burnt as a proportion of area of heather of burnable age (Hudson & Newborn 1995) were calculated for each plot at the end of the permissible burning period (mid-April) in each year.

To monitor potential changes in habitat structure and composition during the experiment, 1-km transects were surveyed each year between late June and mid-July (plots A and B, = 20; plots C and D, = 15). At 20-m intervals along each transect, the dominant plant species and their height were recorded within 25 × 25 cm quadrats. The mean and variation (standard deviation) in height were calculated for heather and for all other species combined in each plot-year. The proportions of quadrats with heather and Molinia caerulea (L.) dominating were calculated.

Predator control methods and predator monitoring

Predators were culled year round by two full-time gamekeepers using legally sanctioned techniques (see Appendix S1). When a plot was subjected to PC, these activities occurred within the plot and within a 2-km buffer around each plot. Legally protected predators [11 raptor species, badger Meles meles (L.) and otter Lutra lutra (L.)] that occurred in the study plots were not killed or disturbed.

Fox abundance was monitored monthly during the bird breeding season (March to June) in each year, by collecting scats from set routes (18·2–26·0 km on each plot) that followed fence lines, walls and river banks following an initial clear-up visit in February (units: mean scats km−1, Webbon, Baker & Harris 2004). The abundance of stoats and weasels was monitored in April and September in each year. Fifty tracking tunnels were deployed for 2 weeks in each plot and the proportion of tunnels with footprints was recorded (King & Edgar 1977). Annual abundance indices for carrion crows were calculated by recording all sightings during four timed transect counts (mean ± SE 58·8 ± 2·1 hours and 136·3 ± 4·0 km for each plot-year) per plot between April and June (units: birds seen per hour). Magpie Pica pica (L.), jackdaw Corvus monedula (L.) and rook C. frugilegus (L.) were infrequently recorded during these counts and are not considered further. The number of each raptor species was recorded whilst monitoring carrion crow abundance and combined into an index of raptor abundance (birds seen per hour).

Monitoring abundance of ground-nesting birds

Birds monitored comprised four species of waders [lapwing, golden plover Pluvialis apricaria (L.), curlew, snipe Gallinago gallinago (L.)], a gamebird (red grouse) and two passerines [meadow pipit Anthus pratensis (L.) and skylark Alauda arvensis (L.)]. These species have been designated as species of conservation concern within the United Kingdom owing to population declines or internationally important breeding or non-breeding populations (Eaton et al. 2009), and were sufficiently numerous to allow meaningful analyses of annual breeding success (all species except snipe and skylark) or changes in abundance (all species).

Abundance of wader species

Study plots were divided into blocks (on average 1·5 km2) that could be surveyed during periods of peak activity: 04:15–08:00 or 16:00–20:00 Greenwich Mean Time (GMT) (Reed et al. 1985). Observers walked to within 100 m of every point within a block marking all sightings on 1 : 10 000 scale maps (Brown & Shepherd 1993). Each block was surveyed twice between mid-April and early June (mean interval between surveys of 22 days). Periods of heavy rainfall and winds exceeding force 5 on the Beaufort scale were avoided.

Bird registrations from the two surveys were combined to estimate the annual number of breeding pairs for curlew, golden plover and lapwing within each plot (Brown & Shepherd 1993). The number of territorial pairs from these visits was corroborated on subsequent visits when monitoring breeding success to maximize the reliability of the abundance data (Grant et al. 2000; Pearce-Higgins & Yalden 2005). Within plot A was an enclosure of 0·04 km2 surrounded by a 1-cm mesh fence, through which mammalian predators could not pass. Within the enclosure the vegetation was kept short, creating a favoured nesting location for lapwings. Once hatched, the chicks could not pass through the fence and a large proportion died from starvation or exposure. No such areas occurred on the other plots therefore data on breeding success of pairs nesting within the fence was excluded from analyses. However, as young fledging from outside the enclosure may in later years have bred within it, we included pairs from within the enclosure when calculating the total numbers of lapwing pairs on plot A.

For snipe, the mean number of drumming males recorded on the two surveys was used as a surrogate for annual nesting density (Green 1985). Drumming activity by snipe varies with weather conditions (Hoodless, Inglis & Baines 2006), but as the proportion of surveys in each wind and rain category did not differ between treatments any bias is unlikely.

Abundance of passerine and gamebird species

Surveys along line-transects of 1-km length (plots A and B, = 20; plots C and D, = 15) allowed the calculation of an annual index of breeding numbers of meadow pipit and skylark. These surveys were conducted twice from late May to early July (on average 14 days apart) between 04:00 and 08:00 GMT. As the aim was to examine breeding numbers only, subsequent analyses were restricted to numbers of displaying males to avoid the inclusion of recently fledged young. To minimize observer bias we utilized only sightings of displaying males within 25 and 200 m of transects for meadow pipits and skylarks, respectively, although observer difference is thought to be small in these species (Buchannan, Pearce-Higgins & Grant 2006). The proportion of surveys in each weather category did not differ between treatments so any bias is unlikely. The mean number of displaying birds per hour was calculated for each transect and the annual mean across transects within each plot used in the analyses.

Red grouse annual breeding numbers across the whole study plot were recorded by undertaking surveys using pointing dogs from mid-March to -April within the same blocks used for wader surveys. On each plot, the total numbers of grouse flushed were used.

Monitoring breeding success of ground-nesting birds

Wader breeding success

No attempts were made to locate and monitor wader nests as visiting a nest may have increased the risk of location by predators, and the impact on breeding success may have been disproportionate between PC treatments. Breeding success was therefore determined from behavioural observations of adults and direct observations of broods. Golden plover and curlew are typically silent and cryptic during incubation but become extremely vocal once chicks have hatched (Yalden & Yalden 1989; Grant et al. 2000). Lapwings, in particular females, display greater vigilance once eggs hatch, producing alarm and contact calls distinct from other calls when chicks are present. For each sighting trained observers assessed behavioural cues and adults were assumed to have chicks if the following were recorded: (i) intense and persistent alarming; (ii) reluctance to leave the location; (iii) flying towards/circling the observer; (iv) short flushing distances and/or distraction display. Adults that showed none of these behaviours were assumed not to have chicks. The use of similar behavioural surrogates has been validated with marked broods for lapwing, for which five visits spread across the breeding season can provide an accurate estimate of the number of pairs fledging broods (M. Bolton, Unpublished data).

In early May, following the first round of wader surveys, provisional maps of pair locations were derived. Attempts were then made to relocate pairs at approximately weekly intervals. Although pairs were not individually recognizable, sightings were related to previous visits primarily not only using proximity of locations but also ensuring that the behaviour of the pair was consistent with expected progression through the breeding cycle. Using location to identify pairs was facilitated by the low density of golden plover and curlew with 59% and 58% of pairs, respectively, nesting >500-m apart and with nearest neighbour distances being 650 ± 30 m for curlew and 738 ± 93 m for golden plover, with no difference between PC treatments. Pairs in close proximity were visited in succession on the same day. Lapwing pairs nested in closer proximity than the other wader species (56% of pairs nested <200 m from nearest neighbour). Due to the closer nesting proximity, lapwing breeding areas were also observed from vehicles, particularly during incubation period, to accurately detemine the number of breeding pairs.

A wader pair was assumed to have successfully fledged a brood if behaviour indicating the presence of chicks was recorded for a minimum of 3 weeks. Although these time intervals were less than the typical fledging period for these species, chick mortality close to fledging is likely to be very low as mortality rates decline markedly with chick age in wader species (Galbraith 1988; Grant 1991; Pearce-Higgins & Yalden 2002). If pairs were thought to have failed, the area was still revisited on at least two further occasions, with the original assessment reclassified in only 3% of pairs. Chicks that had fledged or were close to fledging were also seen in 5% of curlew, 29% of golden plover and 28% of lapwing pairs estimated to have fledged young across all plot-years. Pairs thought to have entered the study plot during chick rearing were not included in analyses.

Annual breeding success for waders was calculated as the proportion of pairs of each species in each plot-year that fledged young. It was not possible to monitor breeding success of snipe owing to their cryptic nature at all stages of the breeding cycle. Of the 28 plot-years between 2000 and 2007 (excluding 2001), breeding success data were available in 28, 26 and 24 plot-years for curlew, golden plover and lapwing respectively (missing data were plot-years when no breeding pairs were found).

Breeding success of passerine and gamebird species

Meadow pipit and skylark nests were found when incubating or brooding adults were inadvertently flushed whilst undertaking other fieldwork. The position of nests was marked inconspicuously using sheep’s wool attached to vegetation. To minimize disturbance, nests were revisited either 8 days after first finding if nests contained eggs or 5 days if with chicks, and subsequently every 5 days. Too few skylark nests were found to allow meaningful analysis (mean 1·6 nests per plot-year, with zero nests in 6 plot-years). To account for the duration of time that nests were under observation the Mayfield (1975) method was used following Aebischer (1999). There were no differences in daily survival rates between the incubation and nestling periods, so data from both periods were combined to estimate the probability of breeding attempts surviving from egg-laying to fledging. Pointing dogs were used from mid-July to mid-August to monitor annual breeding success (young to adult ratio) of red grouse (Jenkins, Watson & Miller 1963).


Within a generalized linear model framework, the experimental design meant that the following structural terms were fitted as fixed effects in all models: plot nested within plot-pair, year and plot-pair × year interaction. The effect of PC was tested by adding treatment to the model. In all cases plot-pair × treatment interactions were tested, but none were significant. Binomial, Poisson or Normal error structures and the canonical link functions were specified as appropriate. The logarithm of transect length, observation time or number of adult red grouse were added as offsets to the Poisson models (Crawley 2002) in the analyses of scats km−1, crows or raptors hour−1 and young to adult ratio for red grouse breeding success respectively. In these analyses, we assumed that any effect of PC would become apparent in the year of treatment, so no lag was considered when PC ceased on plot A in September 2004.

When investigating the changes in numbers of breeding birds the logarithmic ratio of change in numbers between consecutive years was calculated as the response variable for each species. In species with zero pairs breeding in any plot-year (lapwing and golden plover), 0·5 was added to all numbers of pairs before the logarithmic ratio of change was calculated. If PC was effective in the predicted direction, an increase in breeding numbers could result from increased productivity in previous years or from higher settlement rates in the current year, linked to a perceived lower abundance of predators (Fontaine & Martin 2006). For all species, we therefore examined changes in population size in relation to PC in the current year to cover the predator perception hypothesis, and in previous years to examine the effects of productivity. When examining effects of PC in previous years we used a 1-year lag relating to the typical recruitment age for lapwing, golden plover, snipe, red grouse, meadow pipit and skylark (Jenkins et al. 1963; Hotker 1988; Thompson et al. 1994; Pearce-Higgins & Yalden 2003; Delius 2008). Although curlew may begin breeding at 2 years, Grant et al. (1999) consider 3 to be a more typical age, and we therefore use a 3-year lag for curlew. Red grouse populations are known to exhibit delayed density dependence (e.g. Hudson, Dobson & Newborn 1998), therefore, the previous spring grouse density (log-transformed) was included as a covariate to control for potential density-dependent effects. As data were missing in 2001 we considered logarithmic change from 2000 to 2002, annualized by dividing by two. To assess the relationship between year-to-year changes and productivity for each species, models above were repeated, replacing PC treatment with the breeding success using the approriate lag. Standard errors from log-scales have been approximately backtransformed to normal scale using Taylor series linearization (Seber 1982). All statistical analysis was undertaken using genstat 11.1 (Lawes Agricultural Trust 2008).


Habitat composition

No significant treatment-related differences in habitat variables were detected (Table 1), indicating that habitat can be excluded as a possible confounding factor within the experiment.

Table 1.   Habitat variables collected in April (burns) and June–July (all others)
 No predator control mean (95% CI)Predator control mean (95% CI)d.f. = 1.31
  1. Mean (95% CI) adjusted for effect of plot and year (see text for details), indicating no significant differences between predator control treatments.

Annual burning of Calluna vulgaris
 Proportion burnt0·02 (0·01–0·03)0·02 (0·01–0·04)0·050·8
 Mean perimeter of fires (m)288 (238–338)272 (224–320)0·170·7
Vegetation composition and structure
 Proportion of quadrats with
  Calluna vulgaris dominant0·23 (0·21–0·25)0·22 (0·19–0·24)0·240·6
  Molinea caerula dominant0·14 (0·12–0·17)0·13 (0·11–0·16)0·200·7
 Calluna vulgaris
  Mean height (cm)33·1 (31·9–34·3)33·3 (31·9–34·7)0·030·9
  Variation in height (cm)27·2 (24·8–29·6)26·3 (23·7–28·9)0·230·6
 Other species
  Mean height (cm)25·5 (23·9–27·1)26·8 (24·9–28·7)0·860·4
  Variation in height (cm)39·5 (34·0–45·0)45·6 (39·2–52·0)1·560·2

Predator abundances

The numbers of foxes, carrion crows and weasels that were culled varied between years (Fig. 2). In plot-years with PC, the abundance of foxes and carrion crows were 1·8 and 4·6 times lower, respectively, than in plot-years with NPC (fox PC = 0·29 ± 0·04 scats km−1, NPC = 0·51 ± 0·04 scats km−1: χ21 = 9·24, = 0·002; carrion crow PC = 0·46 ± 0·05 birds hour−1, NPC = 2·11 ± 0·21 birds hour−1: χ21 = 94·19, < 0·001). Stoat abundance indices were too low for analysis across all plot-years (range 0–2% tunnels with prints). Weasel abundance indices did not vary between treatments in the spring (PC = 8 ± 2% tunnels with prints, NPC = 8 ± 2% tunnels with prints: χ21 = 0·08, = 0·78), despite almost twofold higher abundance in the autumn on plot-years with PC (PC = 31 ± 5% tunnels with prints, NPC = 17 ± 2% tunnels with prints: χ21 = 5·16, = 0·023). Significantly more raptors were recorded on plots under the PC treatment when considering all species combined (PC = 0·22 ± 0·03 birds hour−1, NPC = 0·14 ± 0·02 birds hour−1: χ21 = 10·87, < 0·001). The 345 raptor sightings comprised buzzard Buteo buteo (L.) (40%), kestrel Falco tinnunculus (L.) (22%), merlin Falco columbarius (L.) (13%), peregrine Falco peregrinus (Tunst.) (10%), short-eared owl Asio flammeus (Pont.) (7%), goshawk Accipiter gentilis (L.) (4%) and hen harrier Circus cyaneus (L.) (3%).

Figure 2.

 Spring predator abundance indices (left axis and dots) for each of the main predator species or groups combined with the annual total number of predators culled (right axis and bars) for each of the study plots. Plot-years with predator control denoted in grey.

Bird breeding success

For lapwing, golden plover and curlew, the percentage of pairs fledging young, after controlling for plot and year effects, was 3·5 times greater with PC than without PC (Fig. 3: lapwing PC = 57 ± 4%, NPC = 19 ± 7%, χ21 = 11·37, <0·001; golden plover PC = 75 ± 8%, NPC = 18 ± 8%, χ21 = 8·19, = 0·004; curlew PC = 51 ± 9%, NPC = 15 ± 6%, χ21 = 5·50, = 0·019).

Figure 3.

 Breeding success (left axis, bars) and numbers of breeding pairs (right axis, dots) for (a) lapwing, (b) golden plover and (c) curlew on each of the study plots. Plot-years with predator control (grey) and with no predator control (white).

Red grouse breeding success in terms of the ratio of young to adult birds was 3·2 times greater with PC than without PC (PC = 1·93 ± 0·22; NPC = 0·60 ± 0·09 χ21 = 36·87, < 0·001, mean adult birds per plot-year = 78·0, range = 14–199). The percentage of meadow pipit pairs fledging young was twice as high in the presence of PC as without it (PC = 52 ± 3%; NPC = 28 ± 2%χ21 = 23·1, < 0·001, mean nests per plot-year = 15·9, range = 8–29).

Changes in numbers of birds

When considering PC treatment in the previous year (Fig. 4), year-to-year changes in the numbers of breeding lapwing were positively related to PC (PC = +66 ± 30% per year, NPC = −36 ± 10% per year: F1,27 = 11·60, = 0·006). The pattern was similar, but the effect was smaller, for changes in numbers of breeding golden plover (PC = +36 ± 31% per year, NPC = −29 ± 15% per year: F1,27 = 3·15, = 0·10) and curlew considering a 3-year lag (PR = +14 ± 11% per year, NPR = −17 ± 5% per year: F1,27 = 5·89, = 0·030). Breeding success in the previous year explained a significant amount of variation in the year-to-year changes in the numbers of breeding lapwing (F1,23 = 11·84, = 0·011), and golden plover (F1,25 = 8·94, = 0·015) but not in curlew (with 3-year lag F1,19 = 0·01, = 0·91), with PC having no additional effect in any species (all > 0·2). No significant effect of PC was found for changes in numbers of breeding snipe (PC = +2 ± 12% per year, NPC = +3 ± 13% per year F1,27 = 0·00, = 0·98).

Figure 4.

 Mean (SE) percentage annual change in breeding numbers between successive years, in relation to predator removal treatment in previous year (or 3 years earlier for curlew; top) and current year (bottom) for lapwing (L), golden plover (GP), curlew (CU), snipe (SN), red grouse (RG), meadow pipit (MP) and skylark (S). Plot-years with predator control (grey) and with no predator control (white). Significance: n.s. > 0·05, *< 0·05, **< 0·01.

When examining changes in wader abundance in relation to PC treatment in the current year, as a measure of whether waders preferrentially nested in areas of lower predator abundance, there was no significant effect of treatment for any of the four wader species (> 0·2, Fig. 4).

Changes in numbers of breeding red grouse showed significant positive effects of PC in both the previous year (PC = +46 ± 17% per year, NPC = −22 ± 8% per year: F1,27 = 17·32, = 0·002) and in the current year (PC = +47 ± 17% per year, NPC = −26 ± 9% per year: F1,27 = 12·90, = 0·005), with breeding success in the previous year explaining a significant amount of variation in the year-to-year changes (F1,27 = 11·10, = 0·007). No significant effect of PC in either the previous or current year (Fig. 4) was detected on changes in the breeding population indices of meadow pipit (both = 0·6) or skylark (= 0·9 and = 0·7). Breeding success in the previous year also did not explain variation in the year-to-year changes (= 0·7).


Predation can reduce breeding success in ground-nesting bird species (Baines 1990; Grant et al. 1999). It is important to understand whether a reduction in predator abundance can manifest itself in improvements in breeding success and breeding numbers if PC is to be recognized as a conservation tool. Habitat structure and composition and predator pressure are known to interact to influence avian predation rates (Baines 1990; van der Wal & Palmer 2008). Unquantified habitat change may confound the results of PC experiments (Bolton et al. 2007). In this study, habitat composition and structure did not differ between PC treatments.

Predator abundances

The numbers of predators culled did not consistently decrease as the numbers of years of PC progressed within this study (Fig. 2). This probably reflects immigration of predators from neighbouring unkeepered areas. Predator abundance indices indicated that keepering effort significantly reduced foxes and carrion crows in line with other PC experiments (Tapper, Potts & Brockless 1996), but no reduction in the already low spring abundance of weasel and stoats was detected. Weasel indices in the autumn were almost twice as high in plot-years with PC, although this could not be quantitatively corrected for fluctuating vole densities (Graham 2002). Although the present study did not attempt to elucidate the effects of individual predator species on ground-nesting birds, control of a suite of predator species in an area by gamekeepers has been found to be more effective than the control of single predators (Holt et al. 2008), presumably because compensatory predation by unsuppressed species can occur (Dion, Hobson & Larivière 1999).

Changes in breeding success and abundance of ground-nesting birds

An average threefold improvement in breeding success for lapwing, golden plover, curlew, red grouse and meadow pipit in the presence of PC was consistent with a meta-analysis of 20 studies with a global distribution, which found in 12 out of 14 bird species hatching success was improved and the post-breeding numbers were significantly higher in the presence of PC (Côté & Sutherland 1997). Previous studies on waders have not been able to detect an effect on breeding success (Parr 1993) or have only detected an effect of predator removal on sites with high initial predator densities (Bolton et al. 2007). The index of breeding success used in the present study (proportion of pairs fledging at least one young) does not incorportate any potential difference in brood sizes at fledging between PC treatments. If the presence of PC were to result in larger broods from successful attempts then the effect of PC on productivity recorded here may be an underestimate.

To maintain a stable breeding population in lapwing and curlew, it is estimated that 0·87–0·97 and 0·48–0·62 fledglings per pair per annum, respectively, are required (Peach, Thompson & Coulson 1994; Grant et al. 1999). Without PC 19% and 15% of lapwing and curlew pairs in this study produced young. Hence, each successful pair would need to produce 4·6 fledglings for lapwing and 3·2 fledglings for curlew annually merely to maintain breeding numbers. This is unrealistic in species that typically lay four eggs and raise one brood per season. Thus, without PC, the low breeding success recorded would result in subsequent declines in numbers of breeding lapwing and curlew.

Predator control resulted in increases in numbers of breeding lapwing, golden plover, curlew and red grouse, but not in snipe, meadow pipit or skylark. These findings mirror those of extensive surveys across 122 estates where golden plover, curlew, lapwing and redshank Tringa totanus (L.) were up to five times more abundant on moorland managed for red grouse, whereas meadow pipit, skylark, whinchat Saxicola rubetra (L.) and carrion crow were less abundant (Tharme et al. 2001). Hence, moorland management on grouse moors provided a plausible explanation for the higher densities of some bird species in the study of Tharme et al. (2001), although the potential confounding effect of habitat could not be disentangled from PC.

Detecting a link between increased breeding success and increased breeding numbers will depend partly on the degree of philopatry and site fidelity of breeding adults between years (Côté & Sutherland 1997; Bolton et al. 2007; Holt et al. 2008). In a mark-recapture study, 39% of lapwings were recorded breeding >10 km from where they were originally marked (Thompson et al. 1994). Therefore, many individuals may return to breed outside the study plots. For consistency within the analyses we also assumed that all young return to breed in the same year (first year for lapwing (Thompson et al. 1994) and golden plover (Pearce-Higgins & Yalden 2003) and third year for curlew (Grant et al. 1999). This lag effect makes it difficult to detect an effect on abundance from the improved breeding success due to PC. Previous studies on the impact of PC on birds have shown inconsistent results when looking at the effect of PC on breeding population sizes. Although in 10 out of 14 studies PC had a positive effect, no statistically significant effect of PC was detected across all studies on breeding population size (Côté & Sutherland 1997). Where studies have detected an effect of PC on bird abundance, the species generally have relatively high philopatry and adult site fidelity as found in gamebirds (Marcström et al. 1988; Tapper et al. 1996).

Management implications

This study shows that controlling predators can improve breeding success and, in some species, abundance amongst birds of conservation concern. On upland moorland in the United Kingdom these benefits are often a by-product of management undertaken to maximize red grouse numbers for shooting. Within the United Kingdom, Special Protection Areas have been designated that support high densities of breeding birds of conservation concern. Of the upland Special Protection Areas designated in England, 74% by area is managed as grouse moor (Tapper 2005). If PC were to cease across such areas then we would predict considerable population declines in a range of birds.

Although this study focussed solely on the benefits of legal control of foxes, crows and mustelids, additional sources of predation are likely to have occurred from raptors. Within this study there were no more than one pair of peregrine and no hen harriers nesting on any study plots with PC (maximum 0·08 pairs km−2). This density is eightfold less than that recorded on the grouse moor at Langholm in South West Scotland (Thirgood et al. 2000), where predation on grouse by raptors led to the cessation of grouse moor management. Accordingly, from the present study, it could be misleading to speculate on the extent to which benefits of legal PC can be shown in the presence of raptors. Despite PC, increased raptor abundance was associated with only 1 plot-year when red grouse densities were sufficiently high to merit driven grouse shooting and hence derive any economic return from employing gamekeepers.

Where the aims of conservation can be met in conjunction with other primarily privately funded land uses, such as gamebird hunting, it seems self-evident that they should be strongly encouraged (Oldfield et al. 2003). In many countries, funds are available for landowners to conduct management for conservation, such as European agri-environment schemes. At present these schemes typically encourage sympathetic habitat management (Whittingham 2007). However, for groups such as waders, the negative impacts of predation may be inadequately addressed by current agri-environment scheme prescriptions (van der Wal & Palmer 2008). We suggest that PC be considered as a general tool for conserving a range of bird species across a range of habitats and that it should be incorporated into agri-environment schemes particularly when targeted at threatened wader populations.


We thank J. C. Coulson, R. E. Green, X. Lambin, I. Newton and R. Moss for critical discussions at all stages of the study’s development and progress. Three anonymous referees provided helpful comments. The commitment of C. Jones, his game keepering team and all field staff is gratefully acknowledged. We are indebted to the landowners (Ministry of Defence, M. Edgar, Viscount Devonport and the Duke of Northumberland) and their tenant farmers. This work was funded by monies donated to the Game & Wildlife Conservation Trust by the Uplands Funding Appeal and the Sir James Knott Trust.