Do habitat characteristics influence predation on red grouse?

Authors

  • Simon J. Thirgood,

    Corresponding author
    1. Game Conservancy Trust, Newtonmore, Inverness-shire PH20 1BE, UK;
    2. Centre for Conservation Science, University of Stirling, Stirling FK9 4LA, UK; and
      Simon J. Thirgood, Centre for Conservation Science, University of Stirling, Stirling FK9 4LA, UK (e-mail s.j.thirgood@stir.ac.uk).
    Search for more papers by this author
  • Stephen M. Redpath,

    1. Centre for Ecology and Hydrology, Banchory, Kincardineshire AB31 4BY, UK
    Search for more papers by this author
  • Steve Campbell,

    1. Game Conservancy Trust, Newtonmore, Inverness-shire PH20 1BE, UK;
    Search for more papers by this author
    • §

      Present address: Steve Campbell, Scottish Agricultural Science Agency, East Craigs, Edinburgh EH12 8NJ, UK.

  • Adam Smith

    1. Game Conservancy Trust, Newtonmore, Inverness-shire PH20 1BE, UK;
    Search for more papers by this author

Simon J. Thirgood, Centre for Conservation Science, University of Stirling, Stirling FK9 4LA, UK (e-mail s.j.thirgood@stir.ac.uk).

Summary

  • 1Predation is not only an important ecological process in the population dynamics of red grouse Lagopuslagopusscoticus, but also has conservation implications for their predators such as hen harrier Circus cyaneus and peregrine falcon Falcoperegrinus. It has been suggested that habitat management might reduce the susceptibility of grouse to predation and thus reduce conflicts between grouse management and raptor conservation.
  • 2We investigated whether habitat characteristics influenced predation on red grouse on a managed moor near Langholm in southern Scotland during 1992–96. We combined demographic studies of the grouse population with radio-telemetry of individual grouse to assess the influence of habitat on mortality rates. Systematic observations of hen harriers were also used to assess the effect of habitat characteristics on their encounter rates and strike success with grouse and other prey.
  • 3There was no evidence that habitat characteristics directly influenced grouse mortality rates at the scale of the grouse population. However, grouse densities were higher and overwinter losses of grouse were lower on areas with greater cover of heather Callunavulgaris. The most likely explanation for the observed pattern of winter loss was that grouse dispersed into areas with more heather, to some extent locally compensating for losses to predators.
  • 4Individual radio-tagged grouse that survived the winter had more blaeberry Vacciniummyrtillus in their home ranges than grouse that were killed in winter by predators. There was, however, no effect of heather cover, vegetation height or vegetation density on the likelihood of individual grouse survival.
  • 5Hen harriers were more likely to encounter grouse broods in a mixture of heather and grass than expected from the observed distributions of grouse broods on moorland. However, having encountered a grouse brood, there was no effect of habitat type or vegetation height on the strike success of the harriers.
  • 6We conclude that the direct effects of habitat on the susceptibility of red grouse to predation are limited under current predator control regimes on managed moorland. Habitat management aimed at mitigating conflicts between raptor conservation and grouse management should focus on reducing the availability of passerine and small mammal prey for hen harriers and thus reducing harrier abundance on grouse moors. Further research is required, however, to assess the consequences of such management for biodiversity.

Introduction

Predation is an important demographic process in many vertebrate populations, and its role in population regulation has received considerable theoretical and empirical support (Krebs 2001). The risk of predation has a major influence on animal behaviour and it has been shown that individuals may modify foraging and reproductive strategies as a result (Lima 1998). Research on a variety of taxa has shown that individual animals may reduce the risk of predation by avoiding areas with high densities of predators (Hileman & Brodie 1994; Kennedy et al. 1994; Flowers & Graves 1997; Durant 2000) or avoiding habitats where they are susceptible to attack (Ferguson, Bergerud & Ferguson 1988; Brown et al. 1988; Lister & Aguayo 1992; Suhonen, Norrdahl & Korpimaki 1994; Cowlishaw 1997; Norrdahl & Korpimaki 1998). Despite the potential importance of the interaction between habitat and predation there have been relatively few studies that have quantified this relationship by demonstrating that mortality rates are higher in certain habitats than others. A notable exception to this paucity of data is the large volume of research on the effects of vegetation characteristics on avian nest predation (see recent reviews in Greenwood et al. 1995; Newton 1998; Campbell et al. 2002).

Red grouse Lagopus lagopus scoticus (Latham) are economically important gamebirds that are widely distributed in the heather Calluna vulgaris (L.) moorlands of the UK. Grouse are reliant on heather for food and there are regional associations between heather cover and grouse abundance (Brown & Stillman 1993; Stillman & Brown 1994), although these habitat effects are less pronounced within the restricted range of managed grouse moors (Smith et al. 2001). Grouse populations are maintained at high densities for sport shooting by habitat management and the control of parasites and predators such as red fox Vulpes vulpes (L.) and carrion crow Corvus corone (L.) (Thirgood et al. 2000a). Predator control is often extended to protected raptors such as hen harrier Circus cyaneus (L.) and peregrine Falco peregrinus (L.), and this illegal killing has a detrimental effect on the distribution and abundance of these species (Etheridge, Summers & Green 1997; Scottish Raptor Study Groups 1997). Grouse managers kill raptors because they believe that raptor predation reduces grouse shooting bags. Recent research in southern Scotland tends to support these views: predation by harriers and peregrines limited grouse populations at low density, suppressed population cycles and reduced shooting bags (Thirgood et al. 2000b,c). However, the relationships between red grouse and their avian predators are complex and, in the case of harriers, are indirectly influenced by habitat. Hen harriers breed at high density on grouse moors where meadow pipits Anthus pratensis (L.) and field voles Microtus agrestis (L.) are abundant, and these species are most common on moorland where there is a mosaic of grass and heather (Redpath & Thirgood 1999; Smith et al. 2001; Redpath, Thirgood & Clarke 2002). There is evidence from southern Scotland and elsewhere in the British uplands that there has been widespread loss of heather as a result of grazing by sheep (Thompson et al. 1995; Fuller & Gough 1999; Thirgood et al. 2000c). One possible solution, albeit long-term, to the conflict between harriers and grouse, may be to reduce sheep grazing in the uplands, thus restoring heather and reducing the densities of pipits, voles and harriers (Thirgood et al. 2000a; Smith et al. 2001). Restoring heather cover should also have a direct positive effect on grouse numbers, at least to the extent of increasing the availability of suitable habitat (Thompson et al. 1995).

Could habitat management also reduce the susceptibility of red grouse to predation? Whilst predation may be the proximate cause of mortality, the ultimate reason could lie with heavily grazed heather moorland that provides little cover for grouse. We have previously demonstrated that grouse nests are found in vegetation that is taller, denser and with greater cover than that found at random on moorland, but this nest site selection provides no benefit in terms of increased nest success (Campbell et al. 2002). This paradox might be explained by the legal control of foxes and corvids on grouse moors and the resultant high level of nest success irrespective of nest site characteristics. However, our earlier work has demonstrated that raptor predation on adult grouse and grouse chicks can be high when raptors are not killed (Thirgood et al. 2000b). If habitat management could reduce the susceptibility of grouse to predators it might reduce the perceived need for grouse managers to kill raptors illegally. In this study we therefore addressed the following questions. First, can habitat characteristics explain any of the observed demographic patterns at the grouse population level? Secondly, do differences in habitat selection at the individual level influence the probability of grouse survival? Finally, are individual grouse more detectable or accessible to raptors in certain habitats?

Methods

The study was conducted during 1992–96 on approximately 100 km2 of moorland near Langholm in southern Scotland (55°10′N and 2°55′S). Detailed descriptions of Langholm moor are given in Redpath & Thirgood (1997) and Thirgood et al. (2000b). In the context of the current study, the key points are that Langholm moor consisted of a mosaic of heather and grass (mainly Molinia caerulea (L.) and Eriophorum vaginatum (L.)), red grouse occurred at varying densities throughout the area, and hen harriers and peregrines were protected during the study whilst foxes and corvids were killed by gamekeepers.

GROUSE DEMOGRAPHY AND HABITAT

Grouse densities were estimated in April, July and October from 1992 to 1996 on 12 areas of 1·0 × 0·5 km. Three transects were walked through each area at 150-m intervals, with a pointing dog quartering to either side of the transect flushing all grouse encountered. Grouse were classified as adults or chicks during the counts. Winter mortality was estimated by monthly searches of the count areas for grouse carcasses between October and April during 1992–96. Transects were walked at 50-m intervals during October and April and at 100-m intervals in the intervening months. Grouse remains were classified as a carcass if bones, tissue or primary feathers were found, and field signs were used to distinguish between grouse killed by raptors and by mammals. Previous work indicated that virtually all grouse carcasses were found during the searches (Thirgood et al. 1998). These data produced the following demographic variables (for details see Thirgood et al. 2000b): (i) winter loss, the difference between October and April grouse counts; (ii) winter kills, the number of carcasses found between October and April; (iii) winter movement, the difference between winter loss and winter kills assumed to represent movement of grouse; (iv) summer loss, the difference between April and July counts of adult grouse; and (v) chick production, the ratio of chicks to adult females in July. The habitat characteristics of the count areas were assessed during October 1996 using methods described in Smith et al. (2001). Briefly, two parallel 1-km transects were placed 200 m apart through each count area and a 2 × 2-m quadrat was placed on each transect at 50-m intervals, giving a total of 40 quadrats per count area. In each quadrat we estimated the percentage cover of heather, blaeberry Vaccinium myrtillus (L.), bracken Pteridium aquilinum (L.), rushes Juncus spp. and grass. We measured the height of the vegetation in each quadrat with a metre ruler and used a checkerboard to estimate vegetation density.

The effect of habitat characteristics on the average grouse density on each count area in October, April and July from 1992 to 1996 was tested using multiple linear regression with forward stepwise selection. Grouse density, vegetation height and vegetation density were log-transformed and vegetation cover was arcsine-transformed before analysis. Previous analyses have shown that winter loss and summer loss on our count areas were density-dependent (Thirgood et al. 2000b). To investigate the effect of habitat characteristics on patterns of loss on the count areas, we therefore included grouse density from the previous count date in the multiple linear regression models with forward stepwise selection. We adopted a similar approach to test for the effects of habitat on chick production.

INDIVIDUAL SURVIVAL AND HABITAT

We radio-tagged 130 grouse in October 1994 and 135 grouse in September 1995 and monitored their survival from October 1994 to September 1996. Grouse were captured at night in hand nets after dazzling with lights and they were then fitted with necklace radio-tags weighing 15 g. Previous research detected no significant effect of these radio-tags on grouse survival (Thirgood et al. 1995). We located radio-tagged grouse weekly and determined causes of mortality as above. We used the mean x and y coordinates from all radio-tracking locations for individual grouse between October and March to estimate the central point of each bird’s winter home range. Individuals for which there were less than five fixes were excluded, as were birds whose radios failed or who disappeared. In the field, the central point was marked in each home range and four 100-m transects were marked along cardinal axes. Transect length was determined from an average grouse winter home range size of 4 ha (S.J. Thirgood & S.M. Redpath, unpublished data). Five 2 × 2-m quadrats were placed on each transect at 20-m intervals, giving a total of 20 quadrats per grouse home range. The percentage cover of vegetation and the average height and density of vegetation in each quadrat were estimated using the techniques described in Smith et al. (2001).

We divided the radio-tagged grouse into birds that survived the winter (n = 80) and those that died as a result of winter predation (n = 82). We used logistic regression to examine survival in relation to the habitat variables, transformed as above. These analyses were initially conducted separately for each habitat variable and then together in a multiple logistic regression model with forward stepwise selection that also included variables for year, sex and age of the grouse.

HARRIER HUNTING AND HABITAT

Observations of hunting hen harriers were made in three winters (1992–93, 1993–94 and 1994–95) and three summers (1994–96) during 1360 hours at Langholm (Redpath & Thirgood 1997). In each season we selected from six to 10 areas of approximately 2 km2 for systematic observations. Observation areas were selected on the basis that they provided a good vantage point and were reasonably accessible. Raptor watches were conducted for periods of 3 h, during which time the entire area was scanned with binoculars at 2-min intervals and the presence and behaviour of all raptors within the area was recorded. When harriers were observed during these scans they were watched continually until they disappeared from sight.

To test whether habitat type influenced the detection rate of grouse broods, we compared the distribution of observed interactions between harriers and grouse broods in four habitat types with the distribution of grouse broods in early June. The four habitat types were: (i) burnt or young heather; (ii) building or mature heather; (iii) mixed stands of heather and grass; and (iv) stands of grass or rushes. The distribution of grouse broods between these habitat types was estimated in June 1994–96 by searching with a pointing dog on transects through the grouse count areas described above. More than 90% of the harrier–grouse brood interactions were observed in the area represented by four of the 12 count areas, thus we used only grouse distribution data collected in these areas. We used the tests for habitat selection described in Manly, McDonald & Thomas (1993). We first calculated the χL2 statistic to test for evidence of selection and then calculated Bonferroni confidence limits around the proportion of attacks seen in each habitat to test whether the observed distribution of grouse broods fell outside these limits.

To test whether harriers were more successful in catching prey in certain habitat types than others we recorded the number of strikes that harriers made at all prey types and how many of these strikes resulted in prey capture. We recorded the habitat in which the strikes took place in the same four categories as described above. Where possible we estimated the vegetation height at the site of the strike in 10-cm bands. We compared the success rate of strikes against all prey in different habitat types and different vegetation heights using G-tests. The same approach was used to investigate the effect of habitat on harrier strike success against grouse broods, combining data from the current study with similar data collected by Redpath (1992) on three grouse moors in northern Scotland.

Results

GROUSE DEMOGRAPHY AND HABITAT

Grouse density was higher in October, April and July on count areas that contained more heather (Table 1 and Table 2). Neither vegetation height nor vegetation density explained any significant variation in grouse density on the count sites. We tested the effects of habitat on winter losses by entering April grouse density into a multiple regression model, with October grouse density, heather cover and the height and density of the vegetation as explanatory variables. Having removed the effects of October density, heather cover explained an additional 11·4% of the variation in April density (Table 3). Winter losses were thus lower on areas with more heather. Was this due to lower predation or movement of grouse? After removing the effects of October density, the habitat variables had no effect on the number of winter kills found on each count area (Table 3). It was not possible to test directly whether movement of grouse was influenced by the habitat variables because our measure of movement incorporated October grouse density and therefore was not statistically independent. However, as winter loss was determined by winter kills and movement, and winter kills were not influenced by habitat, the most likely explanation was that grouse moved into areas with more heather. We conducted a similar analysis on the effect of habitat on summer loss of adult grouse. Having removed the effect of April density, the inclusion of heather cover and vegetation height into the multiple regression model explained an additional 8·5% of the variation in July density (Table 3). As we had no data on summer kills we could not distinguish between the effects of habitat on predation and on movement of grouse. We tested the effects of habitat on summer chick production by entering July chick density into a multiple regression model with July hen density and the habitat characteristics as explanatory variables. The habitat variables had no effect on chick production (Table 3).

Table 1.  Mean grouse density during 1992–96 in April, July and October and habitat characteristics of 12 counting areas at Langholm. Density measured from counts with pointing dogs expressed as grouse per 0·5 km2. Sites arranged in order of increasing April density
AreaApril densityJuly densityOctober densityHeather cover (%)Vegetation height (cm)Vegetation density index
LLL 4·5 6·0 8·011·618·512·5
LLH 5·815·411·4 5·122·510·2
LSL 8·512·018·450·615·015·6
LSB11·030·823·048·616·013·7
LMM16·026·428·460·926·3 9·2
LCL16·830·625·260·922·9 9·9
LCF17·343·225·443·915·614·5
LMB18·535·024·256·818·310·1
LDH19·051·228·863·612·116·7
LDL19·349·230·045·815·312·9
LRF21·339·432·852·813·417·4
LRR25·545·835·053·115·814·2
Table 2.  Multiple regression analyses of the effects of measured habitat variables on mean grouse density in 12 counting areas at Langholm during 1992–96
VariableCoefficient (SE)R2F(1,10)P
April grouse density
Constant1·54 (0·26)
Heather cover2·19 (0·49)0·6619·700·001
July grouse density
Constant2·26 (0·35)
Heather cover2·23 (0·66)0·5311·500·007
October grouse density
Constant2·19 (0·25)
Heather cover1·82 (0·47)0·6015·070·003
Table 3.  Multiple regression analyses on the effects of measured habitat variables on demographic parameters in 12 grouse counting areas at Langholm during 1992–96
VariableCoefficient (SE)R2TF(2,45)P
April grouse density
Constant 0·53 (0·21)
October grouse density 0·48 (0·09)0·62 5·380·001
Heather cover 1·28 (0·30)0·73 4·320·001
Full model0·73 61·950·001
Winter kills
Constant–10·27 (3·53)
October grouse density 6·76 (1·16)0·43 5·830·001
Full model0·43 34·010·001
July grouse density
Constant 2·08 (0·77)
April grouse density 0·43 (0·13)0·57 3·310·002
Heather cover 0·96 (0·39)0·62 2·480·017
Vegetation height –0·50 (0·24)0·66–2·140·038
Full model0·66 27·950·001
July chick density
Constant 1·30 (0·12)
July hen density 1·10 (0·07)0·8214·480·001
Full model0·82212·40·001

INDIVIDUAL SURVIVAL AND HABITAT

We compared the habitat characteristics of the home ranges of 80 grouse that survived the winter and 82 grouse that died as a result of winter predation. We used logistic regression to examine survival in relation to each individual habitat variable and then together in a model that also included variables for year, sex and age of grouse (Table 4). Of the measured habitat variables only blaeberry cover had a significant (positive) effect on the probability of survival. There was no effect of heather cover, vegetation height or vegetation density on survival.

Table 4.  Logistic regression analysis of adult red grouse survival in winter in relation to the vegetation characteristics in their home ranges. The analysis was initially conducted individually for each habitat variable and then in a full model with forward stepwise selection including variables for year, age and sex. Figures show the mean values for each vegetation characteristic and estimates of regression coefficients, standard errors, t-ratio and significance in single models and t-ratio and significance in full model
VariableSurvive (n = 80)Die (n = 82)Single model coefficientSETPFull model
TP
Heather (Calluna)50·953·1−0·0080·007−1·09NS  0·92NS
Heather (Erica) 2·2 1·9  0·0820·065  1·26NS  0·31NS
Blaeberry 3·6 2·1  0·1220·043  2·76< 0·01  2·89< 0·01
Grass36·632·9  0·0150·008  1·95NS  1·18NS
Rushes 1·8 2·0−0·0260·044−0·59NS−1·41NS
Bracken 0·4 0·4  0·0380·101  0·37NS−0·90NS
Height20·419·5  0·0570·030  1·87NS  1·50NS
Density 0–10 cm14·414·3  0·0130·054  0·25NS−1·46NS
Density 10–20 cm 6·9 6·5  0·0520·043  1·19NS−0·80NS
Density 20–30 cm 2·1 1·7  0·1340·074  1·81NS  1·08NS

HARRIER HUNTING AND HABITAT

Interactions between harriers and grouse broods were more likely to be observed in a mixture of heather and grass than expected from the distribution of grouse broods observed at random (Table 5; χL2 = 49·3, d.f. = 3, P < 0·001). Only 6% of 65 broods were found in a mixture of heather and grass when searching randomly with dogs, whereas 36% of 28 interactions between harriers and grouse broods were observed in this habitat. We observed 201 strikes by harriers at prey, of which 36% resulted in capture (Table 6). Strikes at prey in a mixture of heather and grass tended to be more successful than strikes at prey in either pure heather or pure grass, but these differences were not statistically significant (G = 6·6, d.f. = 3, P < 0·1). We were able to estimate the height of the vegetation at the site of the strike in 182 cases. There was no effect of vegetation height on the success of strikes (G = 0·4, d.f. = 3, P > 0·9). We observed 31 interactions between harriers and grouse chicks, of which 32% resulted in capture. We combined these data with observations by Redpath (1992) of 42 strikes by harriers at grouse chicks, of which 46% resulted in capture, to examine the effects of habitat on harrier strike success (Table 6). There was no effect of habitat type (G = 1·73, d.f. = 3, P > 0·5) or vegetation height (G = 1·2, d.f. = 1, P < 0·3) on the success of strikes by harriers at grouse chicks.

Table 5.  The proportions of randomly located grouse broods in four habitat categories at Langholm during 1992–96 compared with the proportions of observed interactions between harriers and grouse broods in these habitat categories. Confidence limits on the interactions calculated following Manly, McDonald & Thomas (1993)
HabitatYoung heatherMature heatherHeather/grassGrass
  1. on interactions.

Broods (n = 65)0·250·680·060·02
Interactions (n = 28)0·110·500·360·04
95% confidence limits0·00–0·260·26–0·740·13–0·590·00–0·12
Table 6.  The number and success of strikes at prey by adult hen harriers at Langholm during 1992–96 classified by habitat type and vegetation height. Data for strikes at grouse broods include observations from Redpath (1992)
HabitatStrikes at all preyStrikes at grouse broods
Number% successNumber% success
Young heather 10401747
Mature heather107353033
Heather/grass 46502348
Grass 3824 333
Height
0–10 cm 1942
10–20 cm 30371947
20–30 cm 9935
30+ cm 34384430

Discussion

HABITAT AND PREDATION

There was no evidence at the scale of the grouse population that predation was influenced by any of the measured habitat variables. However, our demographic data indicated that habitat was an important factor influencing the dynamics of the grouse population. Grouse densities were highest on areas with the most heather, and the amount of heather had a positive effect on the persistence of grouse on the count areas, with the highest losses in both winter and summer occurring on areas with the least heather. This was not due to higher levels of predation in areas with less heather as, in winter at least, the number of kills found on the count areas was not related to habitat. The most likely explanation for the observed pattern of winter loss was that grouse dispersed into areas with more heather, to some extent replacing birds that had been killed by predators. In this context, therefore, moorland habitat may have an important influence on local levels of compensation of winter mortality. At the scale of the grouse population, however, the extent of compensation of winter mortality was unknown, because we did not know the extent to which grouse increased their breeding success through dispersal (Thirgood et al. 2000b). Our understanding of the dynamics of red grouse populations in winter remains poor and future studies could profitably focus on the interaction between predation, dispersal, breeding success and habitat characteristics.

The finding that predation in winter was not directly influenced by habitat at the scale of the grouse population was supported by the lack of clear habitat differences in the home ranges of individual grouse that survived the winter compared with those that were killed by predators. Although increased blaeberry cover in winter home ranges was associated with a greater probability of grouse survival, the difference in blaeberry cover between the two groups was very small and probably has little biological significance. There was no effect on individual grouse survival of heather cover or the height or density of vegetation in the home range. At both individual and population scales, therefore, there is little consistent evidence that habitat characteristics have any measurable effect on predation on grouse in winter. Winter predation on grouse during the course of this study was predominantly due to raptors (Redpath & Thirgood 1997; Thirgood et al. 2000b). Much of the winter raptor predation was thought to be due to peregrines, although we were unable to distinguish with certainty between the kills of different raptor species (Redpath & Thirgood 1997; Thirgood et al. 1998). Habitat characteristics may have little effect on grouse mortality due to peregrines because this raptor tends to strike prey whilst airborne (Ratcliffe 1993).

Hen harriers more frequently encountered grouse chicks in a mixture of heather and grass than expected from the observed distribution of grouse chicks on the moor. This may have been due to the increased detectability of grouse chicks in this habitat or it may have resulted from habitat selection by hunting harriers. Redpath (1992) has previously demonstrated that when large prey such as grouse chicks are rare, harriers hunt heather moorland in accordance with distribution of meadow pipits. Meadow pipits tend to occur at higher density on areas of moorland where there is a mosaic of heather and grass (Redpath & Thirgood 1997; Smith et al. 2001) and this may result in harriers encountering grouse chicks more frequently in these habitats. Harriers tended to be more successful at catching prey in a mixture of heather and grass than in pure heather or pure grass habitats, although these differences were not statistically significant. When analysis was restricted to attacks on grouse chicks, however, there was no effect of habitat type and vegetation height on strike success. These findings need cautious interpretation because habitat characteristics were assessed from vantage points rather than at the strike site. In some heterogeneous habitats, a harrier may appear to be hunting over tall vegetation whereas in reality the vegetation at the strike site may be short. A number of other studies have shown that raptors spend more time hunting over low cover habitats (Wakely 1978; Baker & Brooks 1981; Bechard 1982; Preston 1990) and African marsh harriers Circus ranivorus are least successful at catching rodents in tall vegetation despite the high densities of rodents which occur in these habitats (Simmons 2000).

MANAGEMENT IMPLICATIONS

In summary, we found no evidence that predation on adult grouse was directly influenced by any measured habitat variable. Habitat had a more subtle effect on grouse dynamics, with birds appearing to disperse into heather-dominated areas, possibly compensating locally for some of the density-dependent loss to predators. The interaction of predation, dispersal, breeding success and habitat requires further investigation and may be important in understanding red grouse population dynamics. There was some evidence that interactions between hen harriers and grouse chicks were more likely to occur in mosaics of heather and grass but it was impossible to determine whether this was due to increased detectability of chicks or harrier foraging patterns. Once detected, however, there was no evidence that strike success on grouse chicks was influenced by habitat. Overall, it appears that the direct effects of habitat on the susceptibility of red grouse to predation are limited. Taken in conjunction with the finding that red grouse nest site selection appears to provide little benefit in terms of increased nest success (Campbell et al. 2002), these results suggest that the primary effect of habitat on predation on grouse is indirect, through its effects on alternative prey abundance for generalist predators such as hen harriers (Redpath & Thirgood 1999; Smith et al. 2001).

What are the implications of these results for moorland management? Habitat management has been proclaimed as the long-term solution to the conflict between raptor conservation and red grouse management (Raptor Working Group 2000). Our results suggest that there is only limited scope for habitat management to reduce the susceptibility of red grouse to predation and hence increase the compatibility of grouse management with raptor conservation. Not withstanding the regional association between heather cover and grouse abundance (Brown & Stillman 1993; Stillman & Brown 1994), these results suggest that the primary aim of habitat management to reduce conflicts with raptors on grouse moors should be to reduce the density of passerine and small mammal prey for harriers and hence reduce harrier abundance and predation on grouse (Redpath & Thirgood 1999; Thirgood et al. 2000b; Smith et al. 2001). The potential benefits of this management strategy for red grouse, however, must be balanced against the potential costs to avian biodiversity in the uplands resulting from the production of a heather monoculture (Thompson et al. 1995; Thirgood et al. 2000a).

Alternative approaches to reducing conflicts between hen harriers and red grouse involve increased levels of intervention. Recent experiments demonstrate that the provision of supplementary food to harriers during the breeding season reduced provisioning rates of grouse chicks to harrier nests sevenfold (Redpath, Thirgood & Leckie 2001). Grouse managers have been reluctant to adopt the technique, however, largely due to concerns over the perceived effects of feeding on harrier numbers. Similarly, proposals to limit directly harrier densities on grouse moors, either through translocation or through the destruction of eggs, have been controversial with both grouse managers and conservationists (Potts 1998; Watson & Thirgood 2001). Central to the debate about raptor conservation on grouse moors is the more fundamental question of whether or not grouse moor management benefits biodiversity (Thirgood et al. 2000a). This remains an unanswered question for the future, with recently published research suggesting that grouse management benefits some breeding bird species but is to the detriment of others (Tharme et al. 2001).

Acknowledgements

We thank the Buccleuch Estates for permission to work at Langholm and for their help over the years. We are grateful to a large number of colleagues for assistance but here we particularly single out Ian Newton and Fiona Leckie for special thanks. Nicholas Aebischer, John Connelly, Pekke Helle, Steve Ormerod, Kirsty Park and Vidar Selas improved earlier versions of the manuscript with their comments. This research was funded by the Buccleuch Estates, Game Conservancy Trust, Centre for Ecology and Hydrology, Joint Nature Conservation Committee, Royal Society for the Protection of Birds, Scottish Natural Heritage and Westerhall Estates. This paper is dedicated to the late Gareth Lewis in recognition of his contribution to conservation in Scotland.

Ancillary