Breeding success of the hen harrier Circus cyaneus in relation to the distribution of grouse moors and the red fox Vulpes vulpes

Authors

  • R.E. Green,

    Corresponding author
      *Present address and correspondence: R. E. Green, Royal Society for the Protection of Birds and Conservation Biology Group, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK (e-mail R.Green@zoo.cam.ac.uk).
    Search for more papers by this author
  • B. Etheridge

    1. Royal Society for the Protection of Birds, Dunedin House, 25 Ravelston Terrace, Edinburgh EH4 3TP, UK; and Royal Society for the Protection of Birds, Etive House, Beechwood Park, Inverness, IV2 3BW, UK
    Search for more papers by this author

*Present address and correspondence: R. E. Green, Royal Society for the Protection of Birds and Conservation Biology Group, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK (e-mail R.Green@zoo.cam.ac.uk).

Summary

1. A recent analysis showed that the survival and breeding success of female hen harriers were much lower on grouse moors than on other upland land management classes in Scotland, excluding the Orkney islands: a difference attributed to persecution of hen harriers by humans on grouse moors. However, this study did not look for possible beneficial effects on hen harrier breeding success of the control by moorland gamekeepers of other predators, particularly red foxes.

2. These and other recently published data were re-examined to see whether there was regional variation in hen harrier nest success associated with the presence or absence of foxes and the control of foxes and other predators by moorland gamekeepers.

3. The main finding of the previous analysis, that nest success was much lower on grouse moors than on moorland that was not managed for grouse or in young conifer forests, was even more striking when comparisons were made among land management classes within the same region.

4. The mean number of broods fledged per female hen harrier per year on two grouse moors at which gamekeepers prevented human interference with harrier nests and continued to control other predators was compared with the average for moorland in Scotland. Harrier productivity on these moors was much higher than the average for grouse moors, but similar to the average for moorland not managed for grouse shooting.

5. Nest success of hen harriers breeding on moorland not managed for grouse averaged 19% higher in regions where there were, on average, more grouse moors in the vicinity, but in young conifer forests success was 34% lower in regions with more grouse moors nearby. Neither of these differences was statistically significant.

6. Nest success within a particular land management class was not significantly different inside and outside the range of the red fox.

7. Even a generous assessment of the magnitude of a supposed beneficial effect of the control of foxes and other predators by moorland gamekeepers on hen harrier nest success indicated that its effect on the population trend of hen harriers would be small relative to the large negative effect of persecution of harriers on grouse moors.

Introduction

Persecution by humans was shown recently to have a considerable impact on the demography of the hen harrier Circus cyaneus (L.), a scarce bird of prey that breeds mainly on heather Calluna vulgaris (L.) Hull moors in Scotland (Etheridge, Summers & Green 1997). Both the breeding success and survival of females were much lower on moorland managed for the sport shooting of red grouse Lagopus lagopus scoticus (Latham) than on unmanaged moorland or young forestry plantations. Persecution of hen harriers by humans was the most likely cause of this difference and occurred on most grouse moors, in spite of legal protection. Persecution arises because harriers reduce the number of young reared per female grouse and, at least where they nest in abundance, reduce the density of grouse available for shooting in August (Redpath 1991; Redpath & Thirgood 1997).

Persecution of hen harriers by humans appears to occur on a sufficient scale to limit the total population in Scotland (Etheridge, Summers & Green 1997; Potts 1998). However, previous analyses of breeding success ignored a possible counterbalancing effect of the control of other predators by grouse moor gamekeepers on hen harrier population growth. The killing of red fox Vulpes vulpes (L.), carrion/hooded crow Corvus corone L. and several other predators of ground-nesting birds and their eggs is carried out on a large scale for the purpose of game management, and is permitted by law, although illegal methods are sometimes employed (RSPB/NCC 1991; RSPB 1996). It is widely believed that this predator control benefits other bird and mammal species that breed on moorland, but there has been no rigorous examination of this idea (Hudson 1992). It has even been claimed that hen harriers are net beneficiaries of predator control and other management by moorland gamekeepers, because control of predator numbers improves harrier breeding success and managed moors provide high densities of red grouse as prey (Budiansky 1995). The recent analysis by Etheridge, Summers & Green (1997) showed that any such benefit is outweighed by the negative effects of persecution for hen harriers that nest on grouse moors. However, the net effect of grouse moor management on the hen harrier population as a whole might still be positive if the control measures depress populations of other predators in nearby areas where harriers are not persecuted. In these areas harrier breeding success might be enhanced by a reduction of predation on harrier nests and perhaps also of the negative effects of other predators on harrier food supplies. If this is the case, then the population of hen harriers, and perhaps of other upland birds, might decline if management of moorland for grouse shooting was to cease, even if heather-dominated open habitats were to be maintained for some other reason. In this paper we assess this possibility by examining the relationship between hen harrier breeding success and the proximity of the nest to areas managed for grouse shooting and by comparing breeding success inside and outside the geographical range of the red fox.

Methods

Hen harrier breeding success on moorland managed for grouse shooting but without raptor persecution

We wanted to obtain estimates of hen harrier breeding success on moors managed for red grouse shooting by patch burning of heather and predator control where harrier breeding success was not influenced by human interference. However, Etheridge, Summers & Green (1997) found records of destruction of nests by humans on a high proportion of the estates with grouse moors and no examples of grouse moor estates for which a large sample of failed nests was available for analysis where there were no records of human interference (their Fig. 2). Grouse moor estates without recorded instances of failed nests with signs of interference by humans tended to be those for which there was only a small sample of cases available. Hence, we were unable to identify a subset of grouse moors from the database of Etheridge, Summers & Green (1997) where we could be confident that nest destruction by humans was not occurring. Redpath & Thirgood (1997) studied hen harriers at six moors in Scotland where information available to staff of the Royal Society for the Protection of Birds suggested that there was little or no persecution by humans because of a deliberate policy of the landowners and their staff. On four of these moors (moors A, B, D and E) very few grouse were shot in any year from 1980 onwards (see Fig. 2·5 of Redpath & Thirgood 1997) and we believe that there might not have been as intensive a programme of predator control as is usual on grouse moors. However, at Langholm (Dumfries & Galloway) and moor C (Grampian), driven grouse shooting continued and there was effective control of predators other than raptors. We therefore used data on the number of breeding female harriers and the number of broods they fledged per year for Langholm and moor C taken from Fig. 3.1 of Redpath & Thirgood (1997) and calculated the mean number of broods fledged female–1 year–1.

We compared these estimates with the equivalent value calculated for grouse moors and unmanaged moors in Scotland from Table 8 of Etheridge, Summers & Green (1997). We multiplied the Mayfield (1975) estimate of the proportion of nesting attempts that were successful by the mean number of attempts female–1 year–1. We believe that the estimates derived from the data of both Redpath & Thirgood (1997) and Etheridge, Summers & Green (1997) are unbiased and therefore directly comparable. Hen harriers were studied sufficiently intensively at Langholm and moor C that it is unlikely that females which attempted to breed went undetected. The same cannot be said of all the study areas from which the data in Table 8 of Etheridge, Summers & Green (1997) were obtained. Hence, for these data, a crude ratio of successful nests to numbers of known breeding female-years would overestimate the proportion successful, because the breeding attempts of some females almost certainly failed before they were detected by the fieldworkers (Green 1989). However, Etheridge, Summers & Green (1997) used the Mayfield method and a technique for calculating the mean number of attempts female–1 year–1 to correct this bias.

Proportion of hen harrier nests that were successful at other sites

We used the hen harrier nest records described by Etheridge, Summers & Green (1997), which derive from Scotland, excluding Orkney, in the period 1988–95. Each nest was attributed to one of three land management classes (LMC): grouse moors, moorland not managed for grouse shooting (referred to henceforth as unmanaged moors) and young conifer forests, as described by Etheridge, Summers & Green (1997), who also give details of fieldwork methods and the calculation of derived statistics.

Etheridge, Summers & Green (1997) analysed the nest data separately for four large regions within which there was considerable variation in climate, the proximity of hen harrier nests to grouse moors and other factors. For example, one of their regions included both fox-free islands and mainland areas with foxes present. We used smaller, more uniform regions that were all entirely within or outside the geographical range of the fox. The data were subdivided into nine regions (Fig. 1) selected to maximize the number of within-region comparisons that could be made between LMC. We did not calculate nest success for region–LMC combinations in which there were data from five or fewer nests at either the egg or the nestling stage. The seven mainland regions were continuous areas delimited by straight north–south and west–east lines on the National Grid. Region 8 was the fox-free island of Mull. Region 9 combined the islands of Islay, Jura, Arran and Bute, but most (93%) of the nests recorded were on Islay. Too few nests were recorded on Jura, Arran or Bute for island-specific estimates of nest success to be made for any LMC. These data were combined with those for Islay because foxes were absent on all four islands.

Figure 1.

Map of Scotland showing the regions used in the analysis. Regions 8 and 9 are outside the geographical range of the red fox. The scale bar represents 100 km.

The statistics on hen harrier breeding success considered in this paper are the estimated proportion of nests that survived from the beginning to the end of a particular phase of breeding without failing completely. We refer to these as egg-stage, nestling-stage and overall nest success, where the last proportion is the product of the first two. We calculated standard errors of the proportions of nests successful and pairwise Z-tests of the significance of differences by the method of Hensler (1985). We tested the null hypothesis of equality of mean overall nest success among regions using the Hensler SEs and the Games & Howell Method (Sokal & Rohlf 1981). We took the n used in the calculation of the denominator degrees of freedom to be whichever number of nests used in the calculation of egg-stage or nestling-stage success was the smaller. We also analysed the clutch size and the number of fledglings in broods where at least one fledged. We performed correlation analyses and one-way anovas to test for variation among regions.

Proximity to grouse moors

For each hen harrier nest site, we calculated a score, the grouse moor index (GMI), intended to be a smoothed density of grouse moors in the vicinity. The index, described by Gibbons et al. (1995), is the sum of the reciprocals of distances from a nest to each of a sample of grouse moors whose locations are shown on a map in Hudson (1992).

Status of the red fox

Hen harrier nests were classified as being inside (mainland and Skye) or outside (other islands) the Scottish range of the red fox according to Corbet & Harris (1991). The boundaries of regions used to divide up the nest data were chosen so that all of each region was either within or outside the range of the fox.

One of our aims was to compare the nest success of hen harriers in areas with and without foxes, but our only data for fox-free areas came from Islay, Jura, Arran, Bute and Mull (Fig. 1). Therefore, we considered expanding the geographical coverage of the study beyond that of Etheridge, Summers & Green (1997) by including harrier nest data for the Isle of Man and Orkney, both of which are shown as being outside the range of the fox by Corbet & Harris (1991). However, it is now known that foxes are present in considerable numbers on the Isle of Man, apparently having been introduced in 1986 (Macdonald & Halliwell 1994). The hen harrier nest records we have access to are for 1990–96. As Macdonald & Halliwell (1994) estimated a population of 200–400 foxes (0·3–0·7 foxes km–2) in 1990 and the nest records include cases of fox predation on harrier nests (P. Lindley, personal communication), it would clearly be inappropriate to use the data for the Isle of Man to expand our sample of fox-free islands.

For reasons that are not fully understood, hen harriers in Orkney show a higher incidence of polygyny than in the rest of Britain, more than half of males having more than one mate in most years (Balfour & Cadbury 1975; Picozzi 1978, 1983). They nest on moorland that is not managed for grouse shooting. Mean productivity (fledglings female–1 year–1), clutch size and nest success are lower in Orkney than on moorland not managed for grouse shooting elsewhere in Scotland (Balfour & Cadbury 1975; Picozzi 1984; Bibby & Etheridge 1993; Etheridge, Summers & Green 1997). This might partly be because polygynous males can only supply enough food for some of their mates and offspring, although even monogamous pairs have rather low breeding success (Picozzi 1984). We think that the unexplained difference in the incidence of polygyny complicates the comparison of breeding success in Orkney with other parts of Scotland and would tend to lead to the underestimation of the possible impact of fox predation, because the absence of foxes is confounded with the high level of polygyny and its ecological cause. We therefore decided not to include data from Orkney in the analysis.

Results

Hen harrier breeding success on moorland managed for grouse shooting without raptor persecution

The proportion of breeding females that reared a brood to fledging was higher at Langholm during the period when harriers were protected there than beforehand (Redpath & Thirgood 1997; Table 1) and higher during the protected period at both Langholm and moor C than the equivalent estimate for grouse moors in general for 1988–95 from Etheridge, Summers & Green (1997).

Table 1.  Mean number of hen harrier broods fledged female–1 year–1 at two grouse moors in Scotland (Langholm and moor C) where persecution was prevented (= protected) but control of other predators and grouse shooting continued. At Langholm results are also shown for the period immediately before the policy of preventing persecution was adopted (= unprotected). The data are from Fig. 3.1 of Redpath & Thirgood (1997). Also shown are estimates of the same statistic for hen harriers on grouse moors and unmanaged moors at many sites in Scotland (excluding Orkney; see text). LMC = land management class
Site(s)Persecution status/LMCYearsNumber of female-yearsBroods fledgedBroods female–1 year–1
LangholmUnprotected1986–891110·091
LangholmProtected1990–9645370·822
CProtected1989–9628160·571
BothProtected1989–9673530·726
ScotlandGrouse moors1988–95– – 0·251
ScotlandOther moors1988–95– – 0·751

Neither at Langholm nor at moor C was the proportion of females that were successful much different from the mean for 1988–95 from Etheridge, Summers & Green (1997) for moorland not managed for grouse shooting (Table 1). We excluded four of the moors for which data are given by Redpath & Thirgood (1997) because we were not confident that control of predators other than raptors was maintained at the level that is usual on grouse moors (see the Methods). It can be argued that it was incorrect for us to do this because these sites had been classified as grouse moors on other criteria. However, the mean number of broods reared female–1 year–1 for these moors (A, B, D, and E) combined, during the period when raptors were not persecuted, was 0·695 (range 0·545–0·778), which is similar to the value for Langholm and moor C (0·726) and the mean for moors not managed for grouse shooting (0·751) from Etheridge, Summers & Green (1997). Hence, our conclusion that breeding success was similar on grouse moors with control of predators other than raptors and moorland not managed for grouse shooting was unaffected by the exclusion of data for these moors.

Differences in nest success among land management classes

The proportions of nests that survived the egg and nestling stages and overall nest success are shown in Table 2. Differences among years and among LMC in the aggregated data were examined by Etheridge, Summers & Green (1997) and are not considered further here. There were too many region–LMC combinations with no data or sparse data for us to use a two-way anova to separate the effects of these two factors, so we focused first on differences among LMC within regions. Overall nest success was lower on grouse moors than unmanaged moorland within all five regions where a comparison could be made. The difference was large (70% lower on grouse moors) and significant (matched-pairs t-test, t4 = 21·32, P < 0·001). Overall nest success was also lower on grouse moors than in young conifer forests by an average of 39% in the three regions where a comparison could be made, but this difference was not significant (t2 = 1·86, P > 0·2). There was no significant difference in overall success between unmanaged moorland and young conifer forests (t6 = 1·73, P > 0·1), although success was higher in unmanaged moorland in six of seven regions. A similar pattern of differences among LMC was found for the components of nest success as for overall success. Within regions, nest success at both the egg and nestling stages was significantly lower on grouse moors than unmanaged moors (egg-stage t4 = 3·72, P < 0·05; nestling-stage t5 = 3·14, P < 0·05) and egg-stage success was higher on unmanaged moors than in young conifer forests (t7 = 2·40, P < 0·05).

Table 2.  Egg-stage, nestling-stage and overall nest success (proportion successful) of hen harriers in Scotland in 1988–95 in relation to land management class (LMC) and region (see Fig. 1).The mean value of the grouse moor index (GMI) is shown for all nests within an LMC and region. No value is shown for region–LMC combinations with a sample size of 5 or fewer. Regions 8 and 9 are outside the geographical range of the red fox
Grouse moors Other moors Young conifer forests
EggNestlingOverall EggNestlingOverall EggNestlingOverall
RegionMean
GMI
NSSENNSSENNSSEMean
GMI
NSSENNSSENNSSEMean
GMI
NSSENNSSENNSSE
  • *

    Approximate value taken to be half of the 95% binomial confidence interval of a proportion zero and n= 7 from Diem (1962).

  • NS, nest success; N, nests.

110100·7170·169110·3050·137100·2190·11310170·6970·095410·9120·042620·6350·0929510·6300·119240·8750·068310·5510·113
212110·2080·087210·5590·16390·1160·06110410·7220·118200·6980·095320·5040·10810680·1550·1028
312500·3500·054890·6740·063660·2360·04211630·7800·137140·9270·070160·7220·139
46980·4040·164120·7470·15490·3020·1406890·2220·118130·7030·175130·1560·094
513000·3280·057750·5420·069670·1780·03913640·7370·080490·8580·044860·6320·07612480·3400·122180·6370·144160·2170·094
66110·6820·107290·8470·070350·5770·1036390·8030·102270·9210·054360·7390·103
78270·3380·0431370·5590·057840·1890·0317260·8130·16890·7000·144130·5700·1685770·5540·189110·5550·146150·3070·135
85220·4600·20680·8320·108160·3830·1805050·3700·21261·0000·205*70·3700·229
95990·8110·17095640·7560·067760·8670·0391120·6560·0655220·7140·120310·8100·064580·5770·108

Differences in nest success among regions

Games & Howell tests did not reveal any significant variation among regions in overall nest success at the α = 0·05 level within any of the LMC. After pooling data across regions, overall nest success was not significantly different inside and outside the range of the fox on unmanaged moorland (Fig. 2; with fox, 0·588, SE = 0·046; without fox, 0·621, SE = 0·062; Z = 0·43, P > 0·6), nor in young conifer forests (with fox, 0·379, SE = 0·061; without fox, 0·520, SE = 0·111; Z = 1·11, P > 0·2). Nor were there any significant differences in egg-stage or nestling-stage success between nests inside and outside the range of the fox. There were too few nest records for grouse moors on fox-free islands for a comparison to be made for that LMC.

Figure 2.

Proportion of nests that fledged at least one young in relation to the grouse moor index (see text) for hen harrier nests on (a) grouse moors, (b) unmanaged moors and (c) young conifer forests. The open symbols represent regions outside the geographical range of the red fox.

Regions differed markedly in the mean GMI value at hen harrier nest sites. Those with high GMI (regions 1, 2, 3 and 5) occurred to the north and east of those with low GMI (regions 4, 6, 7, 8 and 9; Table 2 and Fig. 1). It should be noted that the GMI values are means for the hen harrier nest sites that were studied and not unbiased estimates of the density of grouse moors within the whole region. In spite of the lack of significant variation in overall nest success among regions, we tested for correlations across regions and within LMC between mean nest success and the mean GMI value for nests in each LMC and region. There was no significant correlation across regions between overall nest success and mean GMI within any of the LMC, the most significant correlation being a positive coefficient for unmanaged moorland (r7 = 0·478, P = 0·19; Fig. 2). The correlations for the other two LMC were negative. Exclusion of the fox-free islands did not alter this conclusion and resulted in only a small increase in the correlation for unmanaged moorland (r5 = 0·548, P = 0·20).

Differences in clutch size among land management classes

Clutch size was higher on grouse moors than on unmanaged moorland in all five regions where a comparison could be made (Table 3). The mean within-region difference was 0·43 eggs and was significant (matched-pairs t-test, t4 = 3·88, P < 0·02). Clutch size also tended to be larger on grouse moors than in young conifer forests, but this difference was not significant (t3 = 1·76, P = 0·17). There was no significant difference in clutch size between unmanaged moorland and young conifer forests (t7 = 0·10, P > 0·9).

Table 3.  Mean clutch size and mean number of fledglings from successful broods of hen harriers in Scotland in 1988–95 in relation to land management class (LMC) and region (see Fig. 1)
Grouse moorsOther moorsYoung conifer forests
RegionClutch sizeSENestsBrood sizeSEBroodsClutch sizeSENestsBrood sizeSEBroodsClutch sizeSENestsBrood sizeSEBroods
15·100·23104·970·12393·400·16584·910·19213·190·2327
25·910·25113·400·5155·560·20163·440·20256·000·587
35·550·14583·380·15485·080·29123·470·2915
44·750·3183·290·6174·880·3583·910·2911
55·420·19453·270·18445·020·14423·310·14774·820·33113·000·4412
64·730·18223·530·17304·630·22242·790·2033
75·490·10833·780·18514·670·3393·000·52104·430·3773·400·3410
84·880·3083·290·30145·200·3753·290·427
92·600·5154·660·09592·990·12944·440·21272·820·1745

Differences in clutch size among regions

One-way anova revealed significant variation among regions in clutch size on unmanaged moorland (F8,206 = 2·41, P = 0·017) and almost significant variation for young conifer forests (F7,102 = 2·05, P = 0·056). There was no significant correlation across regions between regional mean clutch size and mean GMI within any of the LMC, but there were positive correlations in all three LMC and that for unmanaged moorland approached significance (r7 = 0·623, P = 0·073; Fig. 3). A regression was carried out of mean clutch size, specific to each region and LMC, on mean GMI, LMC being included in the regression model as a factor. This analysis revealed a significant within-LMC positive relationship between clutch size and GMI (t18 = 2·32, P = 0·032), although the effect of LMC was not significant (F2,18 = 1·85, P = 0·185).

Figure 3.

Mean clutch size in relation to the grouse moor index (see text) for hen harrier nests on (a) grouse moors, (b) unmanaged moors and (c) young conifer forests. The open symbols represent regions outside the geographical range of the red fox.

Differences in successful brood size among land management classes

Matched-pairs t-tests revealed no significant within-region differences among LMC in the mean number of fledglings from successful broods (P > 0·5; Table 3).

Differences in successful brood size among regions

One-way anova revealed significant variation among regions in the mean size of successful broods on grouse moors (F6,151 = 2·45, P = 0·028), but not for the other two LMC (P > 0·15). There was no significant correlation across regions between regional mean successful brood size and mean GMI within any of the LMC (P > 0·3; Fig. 4).

Figure 4.

Mean number of young fledged from successful nests in relation to the grouse moor index (see text) for hen harrier nests on (a) grouse moors, (b) unmanaged moors and (c) young conifer forests. The open symbols represent regions outside the geographical range of the red fox.

Discussion

Differences in nest success among land management classes

The present analysis of the same data as that used by Etheridge, Summers & Green (1997) supports their conclusion that nest success and productivity of hen harriers were low on grouse moors, high on unmanaged moors and intermediate in young conifer forests. Although the previous analysis showed that the difference in nest success between LMC remained highly significant when geographical region was also included as a factor, the regions used were much larger than in the present study. Hence, it could be argued that some scope remained in the earlier analysis for some of the difference between grouse moors and the other LMC to be caused, not by nest destruction by humans associated with grouse moor management, but by unmeasured within-region habitat or weather differences correlated with the distribution of grouse moors or because one of the large regions included some fox-free islands with few grouse moors. These possibilities are rendered less plausible by two findings of the present study: (i) nest success was lower on grouse moors than on the other LMC within all of the smaller regions used here in which a comparison could be made, and (ii) there was little evidence of marked variation in nest success among regions within LMC.

The inference by Etheridge, Summers & Green (1997) that human interference was the cause of the lower nest success on grouse moors than on other LMC is supported by the ninefold increase in the proportion of female hen harriers that fledged a brood at Langholm when gamekeepers prevented the persecution of hen harriers (Redpath & Thirgood 1997; Table 1).

Nest success within and outside the geographic range of the red fox

No significant differences in nest success were found between fox-free islands and areas within the geographical range of the fox. On moorland that was not managed for grouse shooting, nest success was very similar in regions with and without foxes. Although nest success tended to be lower in young conifer forests in regions with foxes than in those with none, the difference did not approach significance and was small relative to differences among regions within the range of the fox for this LMC. Our decision to exclude data from the Orkney Islands (fox-free) because of a difference in the predominant mating system could be criticized as being somewhat arbitrary. However, the inclusion of data from Orkney would have indicated lower nest success in fox-free areas than within the range of the fox, a result unlikely to be explicable in terms of cause and effect. Hence, although the fox is known to be a predator of hen harrier eggs and chicks (Watson 1977), our data give no indication that it has a large influence on hen harrier nest success and productivity.

Our analyses do not allow for the possibility of density-dependence in hen harrier nest success. If hen harrier nest success tends to decline with increasing population density, perhaps because of competition for food or predation, then it might be that, at any given density, nest success would be higher where foxes are absent. If harriers reached a higher equilibrium population density in fox-free areas, the mean nest success in areas with and without foxes might then be about the same. However, although hen harriers are abundant on some fox-free islands (Islay, Jura, Arran, Bute and Orkney) they are scarce or absent on others (Lewis, Harris, Uists, Coll, Tiree, Rum and Mull; Gibbons, Reid & Chapman 1993). Moreover, in our fox-free regions, nest success was considerably lower in region 8 (Mull) than in region 9 (mainly Islay; Table 2), even though harriers were more abundant in region 9 than in region 8 (Gibbons, Reid & Chapman 1993). Hence, these observations provide no support for the hypothesis that density-dependence of hen harrier nest success masked a real effect of the absence of foxes, although they cannot entirely exclude it.

Influence of proximity to grouse moors on hen harrier nest success

The success of hen harrier nests located outside grouse moors was not significantly higher in regions in which such nests were, on average, close to grouse moors than in those where they were remote from grouse moors. Indeed, there was no significant variation in nest success among regions once the differences in success among LMC were allowed for, even though there were large differences among regions in the mean GMI value for hen harrier nest sites that were not on grouse moors. There was a non-significant positive correlation across regions between mean nest success and mean GMI value for moorland not managed for grouse shooting, but the correlation for young conifer forests was in the opposite direction. Hence, the suggestion that predator control on grouse moors causes a significant improvement in the nesting success of hen harriers breeding on unmanaged moors or young conifer forests nearby is not supported by these analyses.

As we pointed out above, density-dependence might conceal a real effect of proximity to grouse moors if variation among regions in GMI was confounded with variation in harrier abundance. However, Redpath & Thirgood (1997) found no indication of a density-dependent decline in breeding success at Langholm during 1990–96, when the number of breeding females increased from one to 14. Nor did they find evidence of significant density-dependence of breeding success across their larger sample of moors. This suggests that density-dependent breeding success because of competition for food, mates or territories is unlikely, at least within the range of population densities in our study areas. Density-dependent effects of nest predation on harrier breeding success are not excluded by Redpath & Thirgood's (1997) results because there was control by gamekeepers of foxes and crows in their study areas. It might be that density-dependent nest predation only occurs where predators are not controlled, as has been observed for the grey partridge Perdix perdix (Potts 1986). However, comparison of hen harrier densities (Gibbons, Reid & Chapman 1993) between regions with high and low nest success on unmanaged moors (Fig. 1 and Table 2) does not suggest that nest success is density-dependent in the absence of predator control.

We think it likely that the GMI (Gibbons et al. 1995) is only a crude measure of the influence of moorland gamekeepers on predator populations in the upland habitats outside grouse moors for at least two reasons. First, the index is based on the mapped distribution of a sample of moors contributing to an extensive survey of a grouse disease in 1983 and takes no account of regional variation in the proportion of moors that contributed, the size of moors or changes in the distribution of grouse moors between 1983 and 1988–95 when the harrier nest data were collected. However, there were reasonably high and significant correlations between the number of mapped grouse moors upon which the GMI calculation was based (from Fig. 33·1 of Hudson 1992) and the estimated total number (r12 = 0·661, P = 0·01) and total area (r12 = 0·642, P = 0·013) of all upland estates with grouse shooting (Table 22 of Hudson 1992) in each of 14 regions of Scotland defined in Fig. 3·1 of Hudson (1992). There is no reason to think that the distribution of grouse moors changed markedly during the 1980s. Our second reason for thinking that the index is crude is that the number of gamekeepers per unit area of grouse moor, the methods they use for predator control and other factors also influence the impact of grouse moor management upon predator populations. It is obvious that these factors vary among regions. We consider that these weaknesses are unlikely to have concealed a true relationship between harrier nest success and the prevalence of grouse moor management across the regions used in our study, because regions fell into two distinct groups on the basis of the mean GMI of hen harrier nests within them. Reference to the data on grouse shooting presented by Ratcliffe (1997; his Fig. 16) and Hudson (1992) indicates that our regions are unlikely to have been placed into the wrong group by the method we used.

It can be argued that the results indicate that proximity to grouse moors improved hen harrier nest success on unmanaged moorland, but that small sample sizes prevent the effect from attaining statistical significance. An indication of the scale of such an effect can be obtained by comparing nest success in regions 4, 6 and 7, where the harrier nests on unmanaged moorland and in young conifer forests were, on average, remote from grouse moors, with the success of nests in regions 1, 2, 3 and 5, where typical nests that were not on grouse moors were close to them. We excluded the data from fox-free islands (regions 8 and 9) to avoid this complicating factor. Nest success on moorland not managed for grouse shooting in the low GMI regions was 19% lower than in the high GMI regions. However, in young conifer forests nest success was 34% higher in the low GMI than in the high GMI regions. The mean number of young fledged per successful brood was very similar in the two groups of regions, so differences in the mean number of young fledged per breeding attempt were very similar to those for nest success. The mean number of young per attempt was 19% lower for low than high GMI regions on unmanaged moorland and 36% higher for low than high GMI regions for young conifer forests.

Implications of the differences in clutch size among regions and land management classes

Even if there was actually a positive relationship across regions between the success of hen harrier nests that are not on grouse moors and their mean proximity to grouse moors, this might not be due to the direct effects of predator control by moorland gamekeepers on the success of harrier nests. Hen harrier clutch size was higher on grouse moors than in other LMC within the same region and was positively correlated, within LMC, with the average proximity of hen harrier nests to grouse moors. The clutch size of raptors tends to be positively correlated with food supply (Newton 1979) and supplementary feeding has been shown to increase clutch size in several raptors (Newton 1998), including Circus ranivorus, a close relative of the hen harrier (Simmons 1994). The higher clutch size of hen harriers on grouse moors and on other LMC in regions where grouse moors are abundant may well arise because the food supply of hen harriers is higher on grouse moors and on other LMC in regions where there are many grouse moors. Food supply can also influence nest success, so any tendency for nest success to be positively associated with mean GMI might be due to food supply rather than direct effects of predator control on the risk of predation of harrier nests. An association between plentiful food for hen harriers and the presence of grouse moors could result from a positive effect of predator control and vegetation management on the abundance of red grouse, and perhaps other harrier prey. Selås (1998) found that, during a temporary depression of fox populations in Norway by disease, the density of breeding pairs of goshawk Accipiter gentilis (L.) increased, probably because of an increase in numbers of their main prey, woodland grouse. Alternatively, grouse moors may occur in places whose climate, soils and topography cause them to have high populations of harrier prey regardless of the effects of grouse moor management. A reliable assessment of the relative importance of these possibilities could only be obtained from large-scale field experiments.

Implications for hen harrier population trends

The nest success of hen harriers on unmanaged moors was lower, although not significantly so, in regions where they were, on average, remote from grouse moors. Although this might not be a real effect of predator control (see above), we wished to assess its potential to counterbalance the negative effect of direct persecution of harriers on grouse moors. We therefore used the survival and productivity estimates reported by Etheridge, Summers & Green (1997) to calculate the change in harrier population multiplication rate λ that would be expected if harrier productivity fell by the amount indicated by the comparison of low GMI and high GMI regions described above. To do this, we calculated the mean number of young fledged per attempt on unmanaged moorland in the low GMI regions (4, 6 and 7) and then multiplied by the mean number of nesting attempts female–1 year–1 on unmanaged moorland from Table 8 of Etheridge, Summers & Green (1997) to obtain a productivity estimate of 2·13 fledglings female–1 year–1 for the low GMI regions. This estimate is 11% lower than the productivity for unmanaged moorland calculated from all of the data. We assumed that, if any beneficial effect of predator control by moorland gamekeepers to hen harriers nesting outside grouse moors was removed, then all harriers nesting on unmanaged moorland would have this productivity. We also assumed that the productivity of harriers nesting in young conifer forests in these hypothetical circumstances would be 11% lower than the estimate for this LMC from all of the data. This results in a generous assessment of the potential benefit from predator control because nest success of harriers breeding in young conifer forests in low GMI regions was actually 36% higher than for high GMI regions. We then calculated λ by the methods used by Etheridge, Summers & Green (1997) in compiling their Table 17 and making all the same assumptions about survival rates, fledgling sex ratio and distribution of breeding females among LMC. In the absence of persecution, grouse moors were assumed to have the same productivity and female survival as unmanaged moorland.

We compared the recent λ estimates with those for hypothetical scenarios in which the supposed benefit from predator control to harriers nesting close to grouse moors was maintained or lost and/or the direct effects of persecution on productivity and the survival of adult females continued or ceased (Table 4). It can be seen that the presence or absence of the supposed benefit to harriers of predator control by moorland gamekeepers had a much smaller effect on λ than the continuation or cessation of persecution.

Table 4.  Adult female population multiplication rate λ of hen harriers in Scotland as calculated by Etheridge, Summers & Green (1997) and under various hypothetical scenarios in which a supposed benefit of predator control on grouse moors is lost and/or persecution of hen harriers on grouse moors ceases. The upper group of estimates for individual land management classes (LMC) shows what would be expected for a closed population living entirely in one land class. The lower group show estimates for all LMC combined assuming that dispersal maintains a stable distribution of breeding females among LMC. See text for details of the scenarios
LMCScenarioλ
Grouse moorNo change0·542
Unmanaged moorNo change1·190
Unmanaged moorLoss of benefit1·146
Conifer forestsNo change1·021
Conifer forestsLoss of benefit0·994
All classesNo change0·940
All classesPersecution continues and benefit lost0·915
All classesPersecution ceases and benefit continues1·127
All classesPersecution ceases and benefit lost1·090

In conclusion, no clear evidence exists for a beneficial effect of the control of foxes and other predators by moorland gamekeepers on hen harrier nest success. Even a generous assessment of the magnitude of such an effect, consistent with the available data, indicates that its effect on hen harrier population growth would be small relative to that of persecution of harriers on grouse moors.

Acknowledgements

This study was only made possible by the efforts of members of the Highland, North-east Scotland, Argyll, Central Scotland, Tayside & Fife, South Strathclyde, Dumfries & Galloway and Lothian & Borders Raptor Study Groups. We thank them and several individual fieldworkers for their enormous contribution. We are grateful to landowners who allowed access to their properties. We thank P. Lindley for advice about hen harriers on the Isle of Man and M.I. Avery, I.P. Bainbridge, I. Newton, S.M. Redpath, S.C. Tapper, G.M. Williams and an anonymous referee for constructive criticisms.

Received 15 October 1998; revision received 25 February 1999

Ancillary