Hen harriers on a Scottish grouse moor: multiple factors predict breeding density and productivity

Authors


Summary

  1. Predator density and productivity can be influenced by ecological and anthropogenic factors. Given the controversy over predators impacting on prey of economic and conservation importance, it is essential that these relationships are understood to provide a sound scientific basis for their conservation and management.
  2. We compared numbers and productivity of hen harriers Circus cyaneus, a protected specialist predator of conservation importance, in relation to a change in generalist predator management at Langholm, a moor managed for red grouse Lagopus lagopus scoticus shooting in southern Scotland between 1992 and 2007. During 1992–1999, the moor was managed for grouse and keepers legally controlled predators, thereafter keepering ceased.
  3. Following full protection being given to nesting harriers, their numbers increased from two breeding females in 1992 to 20 in 1997, when predation by harriers limited numbers of grouse available for shooting. After grouse management stopped in 1999, carrion crows Corvus corone and red foxes Vulpes vulpes increased and numbers of female harriers dropped to below five from 2002 onwards.
  4. Numbers of breeding harriers were negatively correlated with meadow pipit Anthus pratensis, crow, and July grouse abundance during the keepered period and positively with spring grouse abundance. Harrier clutch size was positively correlated with vole abundance.
  5. Harrier clutch survival and productivity were higher when the moor was keepered. Predation by foxes was the main cause of harrier breeding failure.
  6. Synthesis and applications. We consider this study to be the first that quantifies how control of generalist predators as part of grouse moor management can benefit harrier productivity. This adds to the importance of finding ways to ensure that grouse moors are managed for harriers, but are still economically viable. If techniques can be devised and put in place to reduce the impact of harriers on grouse, then the control of generalist predators may be viewed as a more acceptable component of conservation management for ground-nesting birds.

Introduction

Global increases in human activity have led to increased contact between expanding human populations and wildlife (Woodroffe, Thirgood & Rabinowitz 2005). Competition between predators and people for shared limited resources has often led to human–wildlife conflict (Graham, Beckerman & Thirgood 2005). Controversy arises when the resources concerned are of economic value and the predators involved are often of high public profile and legally protected (e.g. Thirgood et al. 2000a). Predation on game birds by raptors is an increasingly common source of conflict, especially in Europe, where hunting is economically important (Valkama et al. 2005; Park et al. 2008). Despite a growing understanding of predator–prey dynamics, there have been few attempts to integrate this into management measures to help mitigate this conservation conflict (Thirgood & Redpath 2008; Redpath et al. 2012).

In the United Kingdom, there is a well-documented, long-standing conflict of interest between the conservation of the hen harrier Circus cyaneus (L.) and the management of heather Calluna vulgaris (L.)-dominated moorland for red grouse Lagopus lagopus scoticus (Latham) shooting (Thirgood et al. 2000a; Thirgood & Redpath 2008). The hen harrier is on the UK's red list of birds of high conservation concern and on Annex 1 of the European Union Birds Directive. Despite raptors in general being protected by law in the UK since 1954, the principal threat to successfully breeding harriers is their widespread illegal killing on grouse moors to conserve grouse stocks (Bibby & Etheridge 1993; Etheridge, Summers & Green 1997). Consequently, harriers are being restricted in their distribution (Anderson et al. 2009) and are declining in areas associated with grouse shooting (Sim et al. 2007). The concern of grouse moor managers, supported by research, is that harriers, if not controlled, may increase to densities whereby they reduce numbers of grouse available to hunt and cause economic loss (Thirgood et al. 2000b,c). Well-managed grouse shoots provide quantifiable conservation benefits by helping maintain globally restricted heather moorland habitats (Robertson, Park & Barton 2001) and their associated bird communities (Tharme et al. 2001; Fletcher et al. 2010), as well as providing economic benefits to rural communities (McGilvray 1995). This continued conservation conflict has increased the social, economic and ecological importance of studies to assess factors determining the population dynamics of harriers and their impacts on grouse. Solutions to the problem have, to date, been difficult to achieve but stakeholders on both sides of the debate have recently expressed their desires to collectively work towards finding answers (Thirgood & Redpath 2008; Sotherton, Tapper & Smith 2009; Thompson et al. 2009).

In the UK, the abundance of breeding harriers varies in relation to prey availability, particularly meadow pipits Anthus pratensis (L.) and field voles Microtus arvensis (L.) (Redpath & Thirgood 1999). It is suggested that managed grouse moors have the potential to provide benefits to breeding harriers by providing more prey, particularly red grouse (Redpath et al. 2002a), and providing protection from carrion crows Corvus corone (L.) and hooded crows Corvus cornix (L.), which pre-date harrier clutches elsewhere in Scotland (Picozzi 1978; Amar & Burthe 2001), and from the red fox Vulpes vulpes (L.), which may pre-date chicks or adults (Watson 1977). These generalist predators are routinely legally killed by gamekeepers on grouse moors to maximize grouse stocks. However, the evidence for foxes limiting harrier densities and breeding success is conflicting (Potts 1998; Green & Etheridge 1999; Whitfield, Fielding & Whitehead 2008). Nevertheless, we hypothesized that harriers, in the absence of illegal persecution, would breed more successfully on grouse moors where their hunting opportunities are greater and their natural predators are fewer than on moors not managed for grouse. Here, we test that hypothesis on a managed grouse moor in south-west Scotland. The study began in 1992, specifically to assess the impact of raptor predation on grouse (Redpath & Thirgood 1999). Throughout the study, all raptors were rigorously protected and allowed to breed in compliance with the law, whilst other legally controllable generalist predators, notably foxes and crows, were killed as before. In 1999, grouse moor management ceased and generalist predators were no longer routinely controlled (Baines et al. 2008). Thus, changes in numbers of generalist predators and prey abundance, as well as annual variations in weather, could be related to changes in the numbers and productivity of breeding harriers.

Materials and methods

Study area

The study took place at Langholm (55°10′N, 2°55′S), a privately owned moor on undulating ground from 150 to 568 m a.s.l, which receives 1500 mm of rain annually situated within the Southern Upland hills of south-west Scotland. The moor totalling 76 km2 comprised 59% blanket mire and wet heath on deep peat soils, 16% dry heath on shallow peat and species-poor acidic grasslands (Central Environmental Heritage Surveys 2003). Approximately 40 km2 was heather dominated and managed for driven grouse shooting by five gamekeepers who managed the area by rotational burning of heather and legally controlling predators of grouse, particularly foxes, stoats Mustela erminea (L.), weasels Mustela nivalis (L.) and crows. Grouse moor management ceased at the end of 1999 after increasing harrier numbers had reduced autumn grouse densities below that required for driven shooting (Thirgood et al. 2000b,c). As a result, routine control of generalist predators, including fox and crows, ceased.

Harrier nest monitoring

Harriers can be polygynous, with one male mating with several females. Their nests were located by watching birds displaying and later watching for males passing food to incubating females (Picozzi 1978). Between 1992 and 2007, 133 nests were monitored. The intensity of monitoring was reduced in 2000–2001 and again from 2004 to 2007 relative to that in 1992–1999, but in all years, nests were visited at least three times: during incubation to measure clutch size, within a week after hatching to measure clutch survival and to record the number of chicks hatched and shortly before fledging to record how many chicks had survived to fledging. In 2001, due to a foot and mouth disease outbreak, access restrictions prevented access to the moor until June. Therefore, it is possible that early nesting failures were missed in this year.

The stage of breeding failure, whether of clutches or of broods, was recorded, as well as the likely cause. The following criteria were used to assign causes of failure: fox predation of well-grown chicks or adult females was typified by birds having either puncture wounds in the torso at patterns and depths consistent with those of fox jaws (Corbet & Harris 1991) or chick feathers were present in the nest that had been chewed off rather than plucked. Predation at the clutch phase was assigned to fox if the incubating female had been killed with the above characteristics or if harrier or dummy wax-filled eggs (2002 only) showed teeth marks consistent with those of fox or where fresh fox scats were found in close proximity to the nest and, on one occasion, where the nest site also smelled strongly of fox. Evidence from predation incidents were inspected in situ, before removal. Clutches were considered to have been deserted when the eggs remained intact within the nest and the female was still alive. When nest contents disappeared and the predator remained unidentified, human interference could not be ruled out.

Breeding productivity was expressed as the number of fledglings per female. Mayfield's (1975) method for analysing nesting success was not used as it was considered that all breeding attempts were identified by recording the presence of territorial birds. With the exception of 2001, it was considered unlikely that breeding attempts failed prior to discovery. Annual productivity data for Langholm were compared with regional estimates for the Southern Uplands and with national estimates for Scotland extracted from Scottish Raptor Study Group Annual Reports between 2003 and 2007 (http://www.scottishraptorstudygroup.org/srms.html).

Harrier prey, crow indices and weather

During the harrier nestling period, meadow pipits and red grouse chicks, together with small mammals, formed the majority of prey items delivered to nests (Redpath & Thirgood 1997). The availability of these three food items was assessed annually between 1992 and 2007. Small mammals were trapped in late March on 10 lines of 50 unbaited snap traps set for two nights, giving a total of 1000 trap nights per year. Traps were checked and reset daily and the number of field voles wood mice Apodemus sylvaticus (L.), common shrews Sorex araneus (L.) and pygmy shrews Sorex minutus (L.) caught were recorded (Redpath, Thirgood & Redpath 1995). Voles comprised 68% of all small mammals caught, and hence annual vole index, expressed as the number caught per 100 trap nights, is used instead of small mammals.

Meadow pipit and carrion crow abundances were measured in fifteen 1-km2 Ordnance Survey grid squares by walking two parallel 1-km transects once in either May or June (Thirgood, Leckie & Redpath 1995). Annual indices of pipit and crow abundance were derived from Generalized Linear Models (GLMs) adjusted for observer and count square (see Baines et al. 2008). Red grouse were counted twice within ten 50-ha blocks using pointer dogs, first in spring to assess pair densities and again in July to estimate pre-shooting densities and breeding success (see Thirgood et al. 2000b). Naturally occurring prey of harriers was supplemented in 1998 and 1999, when dead white rats and day-old poultry chicks were provided in the vicinity of harrier nests on one part of the moor to determine whether feeding harriers could reduce the numbers of red grouse killed (Redpath, Thirgood & Leckie 2001). As an index of fox abundance, fox scats were collected annually from 2002 along three routes each of 10 km in length and corresponding to a different historical gamekeeper's beat. A clear-up round was conducted in late March, with subsequent monthly collections in April, May and June (Baker & Harris 2006). Small between-year differences in observation period were corrected for by expressing the annual scat index as scats km−1 per 10 days.

Weather can impact harrier productivity, either in May during egg laying and incubation (Whitfield, Fielding & Whitehead 2008) or in June, particularly during the first 20 days after chicks hatch (Redpath et al. 2002b). To consider any effects of weather on productivity, we used mean monthly temperatures for May and June for Eskdalemuir weather station, situated 30 km west of Langholm.

Statistical analysis

Rates of changes in the number of breeding harriers were examined within an increasing phase (1992–1997) and a declining phase (1998–2003) by linear regression. The numbers of breeding harriers, their clutch size, chick survival and breeding productivity were compared with the annual abundance of voles, pipits, grouse and crows. The degree of intercorrelation between these covariates was tested on annual mean values using Pearson correlations. Fox indices were not collected before 2002, so whether or not the moor was keepered, and hence generalist predators controlled, was used as a surrogate of fox abundance. To do this, we defined ‘period’ as a factor with two levels, 1992–1999, when generalist predators were legally controlled by gamekeepers, and 2000–2007 when they were not. In 1992, both harrier nests failed as a result of suspected human interference and were excluded from analyses of productivity.

Harrier abundance was analysed by Poisson regression in GLMs, with the number of breeding male or female harriers in turn as the response variable, ‘period’ as a factor with two levels (1992–1999, 2000–2007) and the abundance of small mammals, pipits, grouse and crows, loge-transformed where necessary to normalize their distributions, each in turn as independent variables. Relationships between annual harrier breeding data [clutch survival, chick survival and productivity (chicks reared per female)] and prey and predator indices and weather were also analysed using GLMs. Clutch and chick survival were both modelled using binomial regressions (binomial distributions with logit links), with clutches hatched and chicks fledged as the response variables and the total number of clutches laid and the number of chicks hatched as the binomial denominators. Productivity was analysed by Poisson regressions (Poisson distribution, log link adjusted for overdispersion), with the number of chicks reared as the response variable, loge breeding females as an offset and either period as a factor for analyses involving the Langholm data only or ‘site’ as a factor for comparisons between Langholm and regional and national estimates. Interactions between ‘period’ and harrier prey, crow abundance and weather variables were tested in turn to determine whether relationships between either harrier abundance or productivity and prey and predator indices or weather differed between keepering periods. Differences in harrier measures of productivity between the keepered and unkeepered periods were analysed using GLMMs, with ‘year’ as a random effect and ‘period’ as a factor. Data from 1998 to 1999, when some harrier nests were provided with supplementary food, were treated in the same manner as data from those years without feeding. Statistical analyses were conducted in GenStat version 14 (GenStat, Rothamsted, UK).

Results

Harrier numbers in relation to prey indices and crows

The number of breeding female harriers showed two distinct phases: an increase of 16 ± 3% SE per annum (F1,4 = 30·51, < 0·001) from two in 1992 to a peak of 20 in 1997 and then a decline at 15 ± 1% SE per annum (F1,5 = 129·71, < 0·001) to only two again in 2003, before stabilizing at two to four females thereafter (Fig. 1). The mean number of breeding females was higher when the moor was keepered (1992–1999) than when it was unkeepered (2000–2007), with an average of 11 and four females, respectively.

Figure 1.

The abundance of breeding female (filled symbols) and male (open symbols) hen harriers (triangles = keepered years, squares = unkeepered years) from 1992 to 2007.

There were four peaks in small mammal abundance: 1994, 1997, 2001 and 2004 (Fig. 2a). From a sample of 587 small mammals trapped, 68% were field voles, 20% common shrews, 8% wood mice and 4% pygmy shrews. Numbers of field voles caught were positively correlated with the total number of all small mammals caught (r13 = 0·85, < 0·001). Pipit abundance showed a year-on-year decline between 1993 and 1998, before increasing again to a peak in 2003 (Fig. 2b). The abundance of grouse declined in spring, but particularly post-breeding in July (Fig. 2c). Densities reached a minimum in 2003, before showing small annual increases in July thereafter. Crows were few when the moor was keepered, but increased after 1999, when routine killing of crows on the moor largely stopped (Fig. 3a). From 2000 onwards, crows were four times more numerous than in the period when gamekeepers were present. Average fox scat abundance within transect increased between the first year of collection in 2002 and 2005, then levelled off (Fig. 3b).

Figure 2.

The abundance of (a) voles, (b) meadow pipits and (c) red grouse in spring (open symbols) and July (filled symbols; triangles = keepered years, squares = unkeepered years) from 1992 to 2007.

Figure 3.

The abundance of (a) carrion crows from 1992 to 2007 and (b) fox scats from 2002 to 2007 (triangles = keepered years, squares = unkeepered years).

None of the harrier prey abundance indices were intercorrelated (Table 1), but crow abundance was negatively correlated with grouse abundance and positively with vole abundance. Grouse abundance in spring and July was closely correlated (r12 = 0·96, < 0·001). The annual abundance of male and female harriers was closely correlated (r14 = 0·95, < 0·001). Harrier abundance over the whole study period was negatively correlated with both pipit (Table 2, Fig. 4) and crow abundance, the latter significant for males, whilst female harrier abundance was positively correlated with grouse density in spring and negatively correlated with post-breeding grouse density in the keepered period alone. Harrier and vole abundances were unrelated.

Table 1. The inter-relationship between the annual abundance indices of three harrier prey species and carrion crow abundance
 VoleMeadow pipitRed grouseCarrion crow
  1. Values are Pearson correlation coefficients, n = 16.

  2. *< 0·05, **< 0·01.

Vole−0·06−0·200·51 *
Meadow pipit −0·180·11
Red grouse (July)  −0·67 **
Table 2. Relationships between the annual number of breeding female and male hen harriers and the abundance of their main foods: voles, meadow pipits, red grouse pre-breeding in spring and post-breeding in July and a predator; carrion crow, during the whole study period (n = 16 years: 1992–2007) or separately during keepered (k) (n = 8 years: 1992–1999) and unkeepered (u) periods (n = 8 years: 2000–2007) when there was a significant interaction between period
 Interaction with periodParameter estimate
F 1,12 P Slope ± SE F d.f. P
Female harriers
Vole1·330·270·04 ± 0·270·021,120·89
Meadow pipit2·810·12−1·25 ± 0·448·051,120·01
Red grouse (spring)2·610·140·91 ± 0·3010·141,120·008
Red grouse (July)5·950·03(k) −0·05 ± 0·0211·501,60·02
(u) 0·05 ± 0·060·861,60·39
Carrion crow2·530·14−0·51 ± 0·254·081,120·06
Male harriers
Vole0·870·37−0·06 ± 0·210·081,120·79
Meadow pipit3·230·10−0·92 ± 0·375·911,120·03
Red grouse (spring)3·840·05(k) −0·05 ± 0·041·691,50·25
(u) 0·08 ± 0·043·541,50·12
Red grouse (July)7·700·02(k) −0·09 ± 0·0224·801,60·002
(u) 0·05 ± 0·060·701,60·43
Carrion crow2·180·17−0·44 ± 0·204·951,120·04
Figure 4.

The relationship between the number of breeding female hen harriers and (a) meadow pipit and (b) vole abundance (triangles = keepered years, squares = unkeepered) from 1992 to 2007.

Harrier productivity

Overall, 133 breeding attempts were monitored: 94 in the keepered period (mean 13 per annum, range 5–20) and 39 in the keepered period (mean 5 per annum, range 2–9). Of these, clutch size was recorded for 125, with 111 considered to be first clutches and 14 to be replacement clutches following loss of the first clutch. Female breeding status was known for 119 of the clutches: 74 were considered to be from monogamous females, 42 from bigamous females and three from polygynous females. Clutch size did not differ between first and replacement clutches (F1,101 = 0·16, = 0·69) or between females of differing breeding status (F2, 101 = 2·62, = 0·08), so data were pooled for further analysis of clutch size. Clutch size averaged 5·0 (±0·1 SE) eggs and was positively correlated with annual vole abundance (F1,13 = 7·69, = 0·016). Clutch survival was positively related to grouse abundance, but chick survival was not correlated with any measure of food availability or with crows (Table 3). Productivity varied between years from 3·6 chicks reared per female in 1995 to 0 in 2007 (Fig. 5) and was positively correlated with grouse densities in July (Fig. 6). Weather was not related to harrier breeding success and did not differ between the keepered and unkeepered periods.

Table 3. Relationships between clutch survival, chick survival and overall productivity of hen harriers (chicks reared per female; n = 15 years, 1993–2007) in relation to three prey groups: voles, meadow pipits and red grouse (in July), a predator; carrion crow and mean daily temperature in May and June
 Clutch survivalChick survivalChicks reared/female
Slope ± SE F 1,13 P Slope ± SE F 1,13 P Slope ± SE F 1,13 P
  1. Values shown are mean slopes (±SE) from Generalized Linear Models. Relationships did not differ between the keepered and unkeepered periods.

Vole−0·19 ± 0·340·310·59−0·13 ± 0·310·180·680·01 ± 0·140·000·98
Meadow pipit−0·90 ± 0·552·660·13−0·67 ± 0·671·010·33−0·39 ± 0·301·790·20
Red grouse0·97 ± 0·2613·840·0030·33 ± 0·410·640·440·43 ± 0·168·250·01
Carrion crow−0·66 ± 0·392·840·12−0·29 ± 0·350·680·42−0·27 ± 0·192·110·17
May temperature0·16 ± 0·170·860·350·15 ± 0·190·810·43−0·10 ± 0·091·110·31
June temperature−0·19 ± 0·330·320·57−0·53 ± 0·372·060·18−0·28 ± 0·192·360·15
Figure 5.

Mean annual productivity (+SE) (chicks reared per female) of hen harriers at Langholm between 1993 and 2007.

Figure 6.

The relationship between annual densities of red grouse in July (birds km−2) and hen harrier productivity (chicks fledged per female; triangles = keepered years, squares = unkeepered years) between 1992 and 2007.

Harrier clutch size did not differ between the keepered and unkeepered periods. Clutch survival and overall productivity (chicks reared per female) were higher when the moor was keepered and generalist predators were controlled, with 2·8 chicks fledged per female, equating to 80% of breeding attempts successfully rearing one of more chicks, compared with 1·4 fledglings, or only 38% of breeding attempts, when the moor was unkeepered and generalist predators were not controlled (Table 4). There was no between-period difference in either chick survival or the chicks fledged per successfully breeding female.

Table 4. Measures of productivity of hen harriers (means ± 1SE) when the moor was keepered (1992–1999) and when the moor was unkeepered (2000–2007)
 KeeperedUnkeepered
n Mean ± SE n Mean ± SE inline image P
  1. n = the number of breeding attempts upon which the estimates are based.

Clutch size895·00 ± 0·16364·94 ± 0·120·080·78
Clutch survival (%)9287 ± 33969 ± 75·370·02
Chick survival (%)7969 ± 53349 ± 92·870·12
Chicks fledged per female882·77 ± 0·18341·44 ± 0·3112·300·003
Successful attempts (%)9280 ± 43938 ± 719·70<0·001
Chicks fledged per successful female743·30 ± 0·14153·27 ± 0·280·100·76

During 2003–2007, annual productivity was lower at Langholm (1·0 chicks per female) than in the Southern Uplands region (1·7) or in Scotland overall (1·9) (Table 5). Clutch survival at Langholm and in the wider Southern Uplands was lower than the Scottish estimate, whilst brood sizes from successful Langholm nests were lower than those in the remaining Southern Uplands.

Table 5. A comparison of annual measures of productivity of hen harriers at Langholm with those from the Southern Uplands and Scotland as a whole for the period 2003–2007
 LangholmSouthern UplandsScotland F P
n Mean ± SE n Mean ± SE n Mean ± SE
  1. n = the number of years upon which the estimates are based.

Clutch survival (%)568 ± 9371 ± 1380 ± 15·180·036
Chicks fledged per female51·0 ± 0·451·7 ± 0·251·9 ± 0·1188·11<0·001
Successful attempts (%)542 ± 17550 ± 4566 ± 110·460·002
Chicks fledged per successful female42·4 ± 0·553·4 ± 0·252·8 ± 0·1165·94<0·001

Of 130 breeding attempts at Langholm of known fate, 41 (32%) failed to fledge any chicks, with the likely causes of breeding failure assigned to 22 breeding attempts (Table 6). Of these, a total of 12 (55%) were attributed to predation by fox: seven clutches (in which three adult females were also killed) and five broods. No predation by fox was recorded before 1999, the last year of keepering, but in that year, three clutches, one brood and two adult females were pre-dated by fox. One failure at the brood stage was attributed to predation by stoat. A further seven failures were attributed to predation (two of clutches and five of broods), but with the predator unknown. Two clutch desertions followed after cameras were set at nests in 2002, but each desertion incident was also preceded by ravens Corvus corax predating eggs prior to clutch completion. A third desertion of a clutch immediately followed the insertion of a wax-filled dummy egg into the nest.

Table 6. A breakdown of the outcome of 130 breeding attempts of known fate by hen harriers during keepered (1993–1999) and unkeepered phases (2000–2007) and a breakdown of likely causes of breeding failure
 KeeperedUnkeepered
n (%)n (%)
  1. Note that all losses to foxes during the keepered phase, including those of adult females, took place in 1999 the last year of keepering.

Attempts of known fate9139
Successful (fledge ≥ 1 chick)74 (81%)15 (38%)
Failed17 (19%)24 (62%)
Failed with eggs11 (65%)12 (50%)
Deserted 0  3
Fox3 (+2 females)4 (+1 female)
Stoat 0 0
Unknown predator 0 2
Unknown 8 4
Failed with chicks6 (35%)12 (50%)
Fox 1 4
Stoat 1 0
Unknown predator 1 4
Unknown 3 4

Discussion

This study supports the hypothesis that in the absence of illegal persecution, grouse moor management can benefit harriers by legally reducing generalist predators that may pre-date their clutches, chicks or even adults themselves. Historically at Langholm, harrier numbers were limited through illegal killing by grouse keepers (Redpath & Thirgood 1997). After agreement was reached at Langholm in 1991 to protect harriers, thus complying with the law, their numbers increased. Cessation of gamekeepering in 1999, and with it the routine legal control of generalist predators such as foxes and crows, was associated with a halving of harrier productivity. Thus, in the absence of gamekeepers, predation, especially by foxes, appeared to limit harrier productivity and density. This view is, however, contentious and contrasts with findings elsewhere in Scotland and in Wales that showed no difference in harrier nesting success in relation to foxes (Green & Etheridge 1999; Whitfield, Fielding & Whitehead 2008). However, in the Welsh study, with declining grouse stocks and few gamekeepers, it is likely that the low levels of fox control were ineffective relative to that on grouse moors managed for driven shooting. We conclude that potential benefits to harriers provided by grouse moor management are only rarely realized and in practice come nowhere near to compensating for the impacts on harrier productivity and adult survival from continued illegal persecution of nests and breeding adults (Etheridge, Summers & Green 1997). Nevertheless, we consider this study to be the first that quantifies how gamekeeping as part of grouse moor management can benefit harrier productivity if harriers are left to breed undisturbed.

Productivity of 1·4 fledglings per female during the unkeepered period remained higher than on other Scottish grouse moors (0·8), equal to that in coniferous forests (1·4) (Etheridge, Summers & Green 1997) and in theory, sufficiently high to maintain breeding numbers. That numbers declined so rapidly, suggests that either some additional factor at Langholm may have changed to render the moor less attractive to recruiting harriers, thus causing individuals to recruit elsewhere, or that overwinter survival rates had fallen. Habitat quality at Langholm, in terms of availability of tall heather, did not diminish (Baines et al. 2008), and there were inconsistent patterns in the abundance of the three main prey species, with grouse declining, pipits increasing and voles continuing to exhibit 4-year cyclical fluctuations in abundance. That productivity of successful females did not differ between periods suggests that variations in prey abundance did not influence productivity. New et al. (2011) predicted that changes in harrier numbers at Langholm were driven by recruitment rather than by productivity, with recruitment in turn driven chiefly by vole abundance. However, their data were limited to the 1992–1999 period only, when predators were managed. Our longer time series, including a further two vole cycles and the period when predators were not managed, showed no relationship between voles and either harrier numbers or productivity. It is possible, but unproven, that rates of illegal persecution of harriers away from Langholm may have increased following publication of earlier studies that confirmed the views of grouse moor managers that predation of grouse by harriers limited grouse populations in autumn and reduced shooting bags (Thirgood et al. 2000b,c). Numbers of territorial pairs of harriers increased nationally in Scotland between 1988/1989 and 1994 (Sim et al. 2001), but then between 1998 and 2004, whilst the national population remained stable, numbers in the Southern Uplands, including Langholm, showed a 38% decline (Sim et al. 2007), followed by a further 22% decline in the Southern Uplands and an overall 20% decline in Scotland between 2004 and 2010 (Hayhow et al., unpublished). Lower clutch survival and a lower proportion of successful breeding attempts in the Southern Uplands suggest that the situation described at Langholm may be more widespread in the Southern Uplands and possibly other parts of the harrier range on the Scottish mainland. The possible links between reductions in grouse moor management, increases in generalist predators and lower breeding success of ground-nesting birds, including harriers, merit further investigation.

Control of generalist predators, together with diversionary feeding and habitat restoration, resumed at Langholm in 2008 as part of the Langholm Moor Demonstration Project. The project aims to determine whether sustainable grouse shooting can be restored in the presence of breeding harriers. To date, whilst the number of breeding harriers has remained low, varying only between one and three nests per annum, productivity has increased to a 5-year average of 3·8 chicks fledged per female (Ludwig & Baines 2013). These provisional data strongly support our hypothesis that legal control of generalist predators on grouse moors can benefit breeding harriers.

Grouse moors can bring conservation benefits not only to harriers, but also to other ground-nesting birds (Tharme et al. 2001; Fletcher et al. 2010), including wading birds, which also declined at Langholm following cessation of grouse moor management (Baines et al. 2008). Hence, it would appear that if management for grouse was to stop, it could exacerbate the current declines of several upland birds (Sim et al. 2005). This has important repercussions for the conflict resolution process between those who wish to manage moors for grouse and those who wish to conserve hen harriers (Thirgood et al. 2000a; Thirgood & Redpath 2008).

Dialogue between hunting and conservation stakeholder groups is a key way forward and has already generated a range of views (e.g. Thirgood & Redpath 2008; Thompson et al. 2009; Redpath et al. 2010, 2012). Several conflict-reducing techniques have been considered. These range from restoring natural assemblages of predators in an attempt to naturally reduce harrier densities, introducing a ‘harrier quota’ principle, whereby surplus harriers or their broods above a set density would be removed (Potts 1998), possibly resulting in an increase in range and numbers of hen harriers on British grouse moors but maintaining their density at levels below which they can significantly impact upon grouse shooting (Redpath et al. 2010), and diversionary feeding of nesting harriers with carrion to try and reduce the rate at which they provision grouse chicks (Redpath, Thirgood & Leckie 2001). Successful implementation of these measures on grouse moors may diminish the perceived necessity by many grouse moor managers to illegally kill harriers to protect grouse stocks and render the legal control of generalist predators by gamekeepers a more acceptable technique for conserving ground-nesting birds including harriers.

Acknowledgements

We thank Buccleuch Estates for support throughout the study. This work was only possible with considerable assistance in the field from Arjun Amar, Andy Dowse, Ann-Marie MacMaster, Jenny Owen and Daffyd Roberts. The first phase of the work between 1992 and 1999 was funded by The Game Conservancy Trust, The Game Conservancy Scottish Research Trust, Centre for Hydrology and Ecology, Scottish Natural Heritage, Royal Society for the Protection of Birds and the Joint Nature Conservation Committee. Ecological monitoring from 2000 to 2007 was funded by GCT and SNH. We are grateful to Nicholas Aebischer for statistical advice. Steve Redpath supported our use of the data, and both Steve Redpath and Des Thompson commented on an earlier draft of this manuscript. Above all, we are indebted to the late Simon Thirgood for the enthusiasm and encouragement he gave to this study. Finally, we are grateful for the constructive comments of three anonymous reviewers who helped improve earlier versions of this paper.

Ancillary