• hen harrier;
  • peregrine falcon;
  • population limitation;
  • predation;
  • red grouse


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. We assessed the impact of predation by hen harriers and peregrine falcons on a red grouse population in southern Scotland during 1992–98. Grouse density in April, July and October declined during this time, coincident with an increase in the numbers of breeding harriers and peregrines.

2. Winter losses of grouse between October and April averaged 33% and were density-dependent. Raptors were the cause of about 70% of winter mortality and they killed about 30% of the grouse present in October. We were unable to determine whether winter mortality in raptors was additive to other losses.

3. Summer losses of adult grouse between April and July averaged 30% and were density-dependent. Raptors were the cause of more than 90% of the early summer mortality of adult grouse. Summer losses of grouse chicks between May and July averaged 45% and were not density-dependent. Harriers killed about 28% of grouse chicks by late July and about 37% by the end of August. Summer raptor predation on adult grouse and chicks appeared to be largely additive to other losses and we estimated that it reduced autumn grouse densities by about 50%.

4. A model combining the estimated reduction in autumn grouse density caused by raptors with the observed density dependence in winter loss predicted that, in the absence of raptors for 2 years, grouse density in spring would be 1·9 times greater, and grouse density in autumn 3·9 times greater, than in the presence of raptors. The model suggested that raptor predation prevented the grouse population from increasing and was thus a limiting factor.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Generalist predators take a variety of prey and can be sustained at high densities by one prey species whilst preying heavily on another (Andersson & Erlinge 1977). Theoretical models predict that generalist predators can suppress the densities of prey populations (Hanski, Hansson & Henttonen 1991; Hanski & Korpimaki 1995; Turchin & Hanski 1997) and empirical studies of a range of vertebrate predator–prey systems support this view (Erlinge et al. 1983; Marcström, Engren & Kenward 1988; Lindstrom et al. 1994; Messier 1994; Krebs et al. 1995; Reid, Krebs & Kenney 1995; Tapper, Potts & Brockless 1996; Boveng et al. 1998; Korpimaki & Norrdahl 1998).

Red grouse (Lagopus lagopus scoticus) fall prey to a range of generalist predators and there has been debate regarding the role of predation in limiting grouse populations. Early studies on grouse populations at high density showed that peregrine falcons (Falco peregrinus) and red foxes (Vulpes vulpes) killed non-territorial grouse during the winter and thus did not reduce breeding density (Jenkins, Watson & Millar 1963, 1964; Watson 1985). In contrast, Hudson (1992) found no evidence for a surplus of non-territorial grouse during winter and suggested that winter predation by peregrines and foxes reduced breeding densities. Picozzi (1978) examined prey brought to hen harrier (Circus cyaneus) nests in summer and estimated that harriers removed only 7% of chicks from a high density grouse population. Recent studies indicate that the proportion of chicks removed by harriers is greater when grouse are at low density than at high density and that this mortality is additive to other losses (Redpath 1991).

The impact of predation on grouse populations has, in the past, been confounded by vigorous control of predators by gamekeepers on grouse moors. Some gamekeepers now let harriers and peregrines attain natural densities and this has allowed us to study the effect of predation by these raptors on grouse populations. In an earlier paper we described the numerical and functional responses of these generalist predators to their prey (Redpath & Thirgood 1999). Harriers and peregrines occur at high densities on grouse moors where meadow pipits (Anthus pratensis), field voles (Microtus agrestis) and pigeons (Columbia livia) are abundant. If grouse densities on these moors fall to approximately 12 pairs km-2, models suggest that raptor predation can prevent grouse densities from increasing and dampen population cycles.

Our aim in the study presented here is to examine whether raptor predation can limit grouse populations. The clearest way to assess the impact of predation is through experiments comparing prey dynamics where predators are present with dynamics in areas where predators are removed. However, for various reasons this option was not open to us, and instead we investigated grouse demography in areas with different natural levels of predation. Here we first describe the patterns of predation observed in a grouse population where raptors were allowed to breed freely. We then use these data to investigate to what extent predation is additive to other mortality and whether it reduces breeding density and breeding production. Finally, we develop a simple model to investigate the longer-term effects of predation on the grouse population.


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Study area

The study was conducted during 1992–98 on Langholm moor in southern Scotland. The co-ordinates for Langholm are 55°10′N and 2°55′W. Foxes and corvids were killed by gamekeepers but raptors were protected throughout the study. Harriers increased from two to 20 breeding females during the period 1992–98 and peregrines increased from three to six pairs. Peregrines and female harriers were seen on the moor throughout the year but male harriers generally wintered elsewhere. Red grouse were resident on the moor throughout the year.

Grouse counts and carcass searches

Grouse densities were estimated in April, July and October from 1992 to 1998 on 12 areas of 1·0 km × 0·5 km. Transects were walked through each area at 150-m intervals with a pointer dog quartering to 100 m either side of the transect flushing all grouse encountered. 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 the first and last search each winter and at 100-m intervals in the intervening months. Observers scanned for carcasses and searched with binoculars at 100-m intervals. Grouse remains were classified as a carcass only if bones, soft tissue or primary feathers were present. Field signs were used to determine cause of death. Comparison of grouse killed by known predators showed that it was possible to distinguish between grouse killed by raptors and by mammals but not between grouse killed by peregrines and by harriers (Thirgood et al. 1998).

Demographic variables and analysis

The counts and carcass searches produced the following demographic variables:

1. winter loss: the difference between October and April grouse counts on each area;

2. winter kills: the number of carcasses found on each area between October and March;

3. winter movement: the difference between winter loss and winter kills on each area assumed to represent net losses or gains resulting from movement of grouse;

4. summer loss: the difference between April and July counts of adult grouse on each area.

Analysis of variation in demographic variables was by ancova on log-transformed data with area as factor and time as covariate using procedure Multivariate General Linear Hypothesis (MGLH) in Systat (v. 6·0·1).

We tested for density dependence in winter loss across sites by comparing the numbers of grouse on each counting area in October with the numbers in the following April. We used a log linear model with variance of the April count at a given site proportional to the mean of the counts, augmented to include a random site effect. The April (A) and October (O) counts are related by:

  • image

where gi is a random effect for the ith site, at is an effect for year t, and eit is a random effect with zero mean and variance proportional to giat

  • image

. A test for density dependence tests the null hypothesis of unit slope (b = 1). A random site effect is appropriate for testing density dependence across sites with years, and also allows for the lack of statistical independence arising from the repeated use of the same sites in different years. This is an example of a generalized linear mixed model (GLMM) (Schall 1991) which can be fitted using the procedure GLMM in Genstat (v. 5·4.). Similar models were used to test for density dependence in summer loss and grouse chick loss.

Radio-tagging and survival analysis

We radio-tagged 130 grouse in October 1994, 135 grouse in September 1995 and 43 grouse in March 1996, and monitored survival from October 1994 to September 1996. Grouse were captured at night in hand nets after dazzling them 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 survival and breeding success (Thirgood et al. 1995). We located radio-tagged grouse weekly and determined causes of mortality as above. We calculated survival rates for winter (1 October−31 March) and summer (1 April−30 September) using the Kaplan–Meier method generalized to a staggered entry design (Pollock et al. 1989). Differences in survival rate were tested using a two-tailed z-test with Bonferroni corrections for multiple testing (Pollock et al. 1989).

Territorial status

We used radio-tagged grouse to investigate the effects of territorial status during October to December on survival during January to March. We defined male grouse as territorial if > 50% of radio-locations from 1 October to 31 December were within an area of 25 ha, and female grouse were defined as territorial if the bird was paired on > 50% of radio-locations during this same period. We included only those grouse for which we had more than 10 radio-locations during the period October to December.


Compensation of predation occurs if grouse lost to predators are replaced by immigrants from elsewhere. The amount of any compensation will be influenced by the timing of mortality and movement. We used the two cases of all mortality first and all movement first to estimate the extreme values of compensation between which the true value should lie (Anderson & Burnham 1976). The number of grouse lost to the population was expressed as k-values, the differences between log population densities in October and April.

In the first case, movement occurs before predation. For each counting site in each winter 1992–96 we calculate:

  • image
  • image

In the second case, predation occurs before movement. For each counting site in each winter 1992–96 we calculate:

  • image
  • image

We then plot k losses to movement against k losses to predation for each case to estimate the extreme values of compensation between which the true value should lie.


We collected 123 grouse killed in winter where the carcass included the whole caeca. The numbers of Trichostrongylus tenuis were estimated using established techniques (Hudson 1986). Infection intensity is expressed as geometric mean worms per bird.

Grouse chick losses

Grouse chick losses were estimated by comparing brood size in late May and early June to brood size in July. Brood size in May 1995 and 1996 was estimated from the broods of radio-tagged females at hatch. Brood size in June in 1993–96 was estimated by working a pointer dog on transects through count areas. We included females without chicks as broods of zero. We used estimates of brood size in May and June and chick density in July to estimate chick density in May and June as:

  • image
  • image

Harrier predation

The numbers of grouse chicks killed by harriers during 1993–96 was estimated from observations at harrier nests during the nestling period, from observations of harrier territories during the incubation and postfledging periods, and from the numbers of breeding harriers in each year. Twenty-six harrier nests were observed from hides for 2678 h over the 4-year period. Prey were identified to species where possible, but where this was not possible they were classified as passerines, gamebirds or waders, small mammals and lagomorphs. Of 2101 prey items seen delivered to nests, 92% were identified to type and 73% to species. Provisioning rates during incubation were estimated by watching seven male harriers during 1994–96 for an average of 20·2 ± 6·0 h. Provisioning rates during the postfledging period were estimated by watching six harrier territories during 1994–95 for an average of 11·0 ± 2·5 h. We used these provisioning rates together with male diet during the nestling period to estimate the number of grouse chicks taken during incubation and postfledging. Previous work demonstrated that diet composition during the nestling period was similar to that during the incubation and postfledging periods (Redpath & Thirgood 1997).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Population trends

Grouse densities in April, July and October declined during 1992–98 (Fig. 1; April: F1,55 = 17·4, P < 0·001; July: F1,67 = 21·4, P < 0·001; October: F1,67 = 22·2, P < 0·001).


Figure 1. Grouse densities at Langholm in (a) April, (b) July, and (c) October during the period 1992–98. Values are means ± SE for twelve 0·5 km2 counting sites.

Download figure to PowerPoint

Winter predation

Patterns of winter predation. There was no within-area trend through time in winter loss, winter kills or winter movement (Table 1; winter loss: F1,35 = 1·09, P = 0·30; winter kills: F1,35 = 0·12, P = 0·73; winter movement: F1,35 = 0·75, P = 0·39). There were significant area differences in winter kills and winter movement but not winter loss (winter kills: F11,35 = 14·33, P < 0·001; winter movement: F11,35 = 2·64, P = 0·01; winter loss: F11,35 = 1·71, P = 0·11). April grouse numbers were, on average, 33% lower than numbers the previous October, the numbers of grouse found dead represented, on average, 43% of the numbers of grouse counted in October, and winter movement, in this case net gains through immigration, averaged 10% of the numbers counted in the previous October.

Table 1. . Demographic data for the grouse population at Langholm 1992–96. Values are mean ± standard error of twelve 0·5 km−2 counting areas. All percentages relate to October density except summer loss which relates to April density.
October density24·8 ± 3·323·3 ± 3·022·9 ± 3·220·7 ± 2·8
April density14·4 ± 2·117·8 ± 2·315·3 ± 1·713·6 ± 1·8
July density12·3 ± 1·411·7 ± 1·69·4 ± 1·09·0 ± 1·2
Winter loss (%)10·4 ± 2·2 (41·9)5·4 ± 1·4 (23·2)7·7 ± 2·0 (33·6)7·1 ± 1·4 (34·3)
Winter kills (%)10·8 ± 1·7 (43·5)9·2 ± 2·2 (39·5)9·2 ± 1·7 (40·2)10·3 ± 2·4 (49·8)
 Raptor kills6·4 ± 1·16·6 ± 1·76·5 ± 1·47·9 ± 2·3
 Mammal kills3·2 ± 0·61·9 ± 0·52·1 ± 0·41·7 ± 0·5
 Other deaths1·2 ± 0·40·7 ± 0·20·6 ± 0·30·7 ± 0·3
Winter movement0·4 ± 2·33·8 ± 2·01·5 ± 1·73·2 ± 1·7
Summer loss2·2 ± 1·26·2 ± 1·65·8 ± 1·24·6 ± 1·0

We tested for density dependence in winter loss across areas by comparing grouse numbers on each count area in October with numbers in April (Fig. 2). The log linear model applied to these data had a slope of 0·45 (SE = 0·098) which was significantly less than unity, slope 1·0 (t10 = −5·61, P < 0·001), thus giving strong evidence for spatial density dependence in winter loss.


Figure 2. Density dependence in (a) winter loss of adult grouse, (b) summer loss of adult grouse, and (c) summer loss of grouse chicks at Langholm during the period 1992–96.

Download figure to PowerPoint

A total of 474 grouse carcasses were found on the count areas during the 1992–96 winters (Table 1). Raptor predation was the most important cause of winter mortality and was responsible for 70% of the grouse found dead and for 30% of the grouse counted alive in October. There was no temporal trend in the numbers of grouse found dead which had been killed by raptors (F1,35 = 1·31, P = 0·26).

Winter survival of radio-tagged grouse was significantly higher in 1994–95 than in 1995–96 (1994–95: survival = 0·61, 95% CI = 0·52–0·69; 1995–96: survival = 0·44, 95% CI = 0·37–0·50; Z = 3·16, P < 0·01). This difference was largely a result of increased losses of radio-tagged grouse to raptors in 1995–96 compared to 1994–95 (Table 2).

Table 2. . Causes of mortality of radio-tagged grouse at Langholm
 Total% of total% of deaths
Live 1 October122
Dead 1April4738·5
Raptor kill3427·972·3
Mammal kill75·714·9
Other deaths64·912·8
Live 1 October153
Dead 1 April8958·2
Raptor kill7347·782·0
Mammal kill106·511·2
Other deaths63·96·7
Live 1 April73
Dead 1 October3345·2
Raptor kill2230·166·7
Mammal kill34·19·1
Other deaths34·19·1
Live 1 April93
Dead 1 October3840·9
Raptor kill3638·794·7
Mammal kill11·12·6
Other deaths000

Compensation of winter predation. All radio-tagged male grouse alive on 1 January 1995 and 1996, respectively, were territorial during the previous October to December and of these 64% survived until April (Table 3). Of female grouse alive on 1 January, 80% were territorial in the previous October to December and survival until April of territorial females (79%) was similar to non-territorial females (69%) (G-test; G1= 0·67, P > 0·25). This demonstrates that most grouse alive in January were territorial and that some of these birds subsequently died.

Table 3. . Territorial status in October to December and survival of male and female radio-tagged grouse during January to March during 1995 and 1996
Territorial statusSurvival19951996
Territorial maleSurvive3334
Non-territorial maleSurvive00
Territorial femaleSurvive3421
Non-territorial femaleSurvive56

Local compensation of winter predation might occur if grouse lost to predators are replaced by immigrants but the strength of compensation will be influenced by the relative timing of the two processes. Figure 3a plots k losses to movement against k losses to predation, assuming that movement occurs before predation. The line for total additivity assumes that losses resulting from predation do not influence movement and has zero slope. The line for total compensation assumes that losses to predation result in identical numbers gained through movement and has a slope of −1·00. The regression line through the observed data lies at approximately 55% compensation. Figure 3b plots k losses to movement against k losses to predation, assuming that predation occurred before movement. The lines for total additivity and total compensation are drawn as before. The regression line through the observed data lies at approximately 95% compensation. This analysis suggests that between 55% to 95% of winter predation on the count sites could be locally compensated by movement of grouse.


Figure 3. Plots of k losses as a result of kills against k losses resulting from movement during winter. In (a) movement occurs before kills; in (b) kills occur before movement.

Download figure to PowerPoint

Compensation of winter predation might occur if increased grouse densities in the absence of raptors resulted in higher rates of predation by mammals such as foxes. Raptors killed 30% and mammals killed 10% of the grouse population overwinter (Table 1) and there was no evidence that mammal predation of grouse was density-dependent. Assuming a constant rate of mammal predation, 10% of grouse which survived in the absence of raptors would be killed by mammals.

Compensation of winter predation might occur if increased grouse densities in the absence of raptors resulted in reduced survival or breeding success because of the parasite T. tenuis. Worm counts on grouse killed by predators indicated that parasite burdens were low in comparison to the infection levels of 2000+ worms per bird that reduce survival and productivity (young male: n = 43, geometric mean = 70 worms per bird; young female: n = 23, geometric mean = 57 worms per bird; old male: n = 30, geometric mean = 164 worms per bird; old female: n = 27, geometric mean = 196 worms per bird).

Summer predation

Patterns of summer predation on adult grouse. Rates of summer loss of adult grouse did not differ between years (Table 1; F1,35 = 0·53, P = 0·67). However, there were significant between-area differences in summer loss (F11,35 = 3·10, P = 0·006). On average, 30% of adult grouse disappeared from each count area between April and July. We did not search for carcasses in summer and cannot distinguish between mortality and movements of grouse.

We tested for density dependence in summer loss across areas by comparing the number of adult grouse on each counting area in April with the number in July (Fig. 2). The log linear model applied to these data had a slope of 0·67 (SE = 0·097) which was significantly less than unity, slope 1·0 (t10 = −3·40, P < 0·01), thus giving strong evidence for spatial density dependence in summer loss.

Summer survival rates of radio-tagged grouse were similar in 1995 and 1996 (1995: survival = 0·56, 95% CI = 0·44–0·67; 1996: survival = 0·61, 95% CI = 0·50–0·71; Z = 0·63, P > 0·05). Raptor predation was the cause of most summer mortality of radio-tagged grouse (Table 2). Raptors were responsible for more than 90% of deaths in April and May (18/21 in 1995 and 31/32 in 1996).

Compensation of summer predation on adult grouse. Compensation of summer predation on adult grouse might occur if non-territorial birds remained in the population to replace territorial birds that were killed. In 1995 and 1996, all radio-tagged female grouse laid a clutch of eggs, suggesting that these females were paired and territorial. Furthermore, 99% of female grouse counted in April 1993–96 were paired with males. Together these observations indicate that virtually all females attempt to breed and thus were unlikely to be replaced.

Compensation of raptor predation of adult grouse in summer might occur if increased grouse densities in the absence of raptors resulted in increased predation by mammals. Raptors killed 30% and mammals killed 3% of the April grouse population by June (Table 2). Assuming a constant rate of predation, 3% of the grouse that survived in the absence of raptors would be killed by mammals.

Patterns of summer predation on grouse chicks. Grouse chick losses were estimated by comparing the size of broods in the first week of June and in the third week of July in 1993–96 (Table 4). The median hatch date of radio-tagged hens was 28 May in 1995 (quartiles 25 May−3 June, N = 27) and 30 May in 1996 (Quartiles 25 May−2 June, N = 33), thus the period monitored was from week one to week seven. Grouse chick losses during this time ranged from 10 to 27%, corresponding to 2·3–8·3 chicks per 0·5 km2.

Table 4. . Estimates of chick loss at Langholm during the period 1993–96. Data from June brood counts and July grouse counts. Values are means ± SE ± sample size)
June brood size4·0 ± 0·45·6 ± 0·36·6 ± 0·35·9 ± 0·3
July brood size3·6 ± 0·34·4 ± 0·34·8 ± 0·34·5 ± 0·3
Chick loss %10·121·027·024·4
June chicks per 0·5 km222·931·630·724·8
July chicks per 0·5 km220·6 ± 3·024·9 ± 4·522·4 ± 3·518·8 ± 2·5
Chick loss per 0·5 km22·36·68·36·0

These estimates of chick losses were based on June brood size, by which time some chick mortality had already occurred. We recalculated grouse chick losses for 1995 and 1996 using the mean brood size at hatch of radio-tagged hens in late May for comparison to mean size of broods during grouse counts in the third week of July (Table 5). Chick loss from week zero to week seven averaged 45% in 1995–96, corresponding to 16·8 chicks per 0·5 km2.

Table 5. . Estimates of grouse chick loss between the last week in May and the third week in July and the numbers of grouse chicks killed by harriers per 0·5 km2 and as a percentage of May chick density
Per 0·5 km219951996
May brood size8·4 ± 0·48·5 ± 0·4
July brood size4·8 ± 0·304·5 ± 0·3
Chick loss percentage42·647·4
May chicks39·135·7
July chicks22·4 ± 3·518·8 ± 2·5
Chick loss to17 July16·716·9
Chicks killed by harriers to17 July (%)11·2 (28·6)9·7 (27·2)
Unexplained chick loss to 17 July (%)5·5 (14·1)7·2 (20·2)
Chick killed by harriers to dispersal (%)13·6 (34·8)14·0 (39·2)

We tested for density dependence in chick losses by plotting July chick density against July hen density for each counting area in each year (Fig. 2). The log linear model applied to these data had a slope of 1·14 (SE = 0·088) which was not significantly greater than unity, slope 1·0 (t10 = 1·59, P = 0·14), therefore, on this basis, there was no evidence for density-dependent chick loss.

The numbers of grouse chicks killed by harriers between grouse hatch in May and harrier dispersal in August for 1993–96 was estimated during incubation, nestling and postfledging periods (Table 6). Dividing these figures by the area of heather moorland indicated that harriers killed 4·3–14·2 grouse chicks per 0·5 km2 of moorland. We also estimated the numbers of chicks killed by harriers by 17 July by counting the number of days that harriers were present during incubation, nestling and postfledging periods and multiplying these values by the relevant provisioning rates. Estimates of grouse chick predation by harriers to 17 July ranged from 3·1 to 11·2 chicks per 0·5 km2 heather moorland (Table 6).

Table 6. . Estimates of grouse chick losses to harriers at Langholm from harrier incubation to dispersal
Successful males4667
Successful females49812
Early nestling159·6 ± 30·3554·4 ± 76·6502·3 ± 99·1432·2 ± 125·5
Late nestling84·0 ± 9·9228·7 ± 58·0226·8 ± 74·0219·9 ± 55·2
Post fledging82·8338·1 ± 117·2390·3496·3
Total loss357·91176·41129·01166·2
Loss per 0·5 km24·314·213·614·0
Loss before 17 July257·2917·4931·3804·6
Loss per 0·5 km23·111·011·29·7
Loss after 17 July100·7259·0197·7361·6
Loss per 0·5 km21·23·12·44·4

Estimates of grouse chick losses from late May to mid-July during 1995 and 1996 suggest that 45% of chicks disappeared during this time (Table 5). Harriers were estimated to have killed 28% of the chicks present at hatch by 17 July (Table 6). There were thus large losses of chicks which could not be attributed to harrier predation. Estimates of the number of chicks killed by harriers should also include predation during the period 17 July to harrier dispersal. Including predation during this period suggested that 35% (1995) and 39% (1996) of the grouse chicks present at hatch were killed by harriers by the time of harrier dispersal (Table 5).

Compensation of summer predation on grouse chicks. Differences between estimates of grouse chick losses (45%) and estimates of grouse chicks killed by harriers (28%) suggest that unexplained losses of chicks by 17 July reduced chick densities by 17%. Studies of individual chick survival demonstrate that most of these losses occur in the first week after hatch (Redpath & Thirgood 1997). These findings suggest that grouse chick mortality is high prior to harrier predation and that harrier predation may be largely additive to other mortality. A conservative estimate of compensation assumes that the same proportion of the grouse chicks killed by harriers would have died from other causes, as in the grouse chick population at large. This suggests that 17% of the grouse chicks killed by harriers to 17 July would have died from other causes in the absence of harriers.

Effect of predation on grouse numbers

Raptor predation was clearly an important source of grouse mortality. To what extent did this predation reduce grouse density? While we were unable to determine the level of compensation of raptor predation in winter, most raptor predation in summer appeared additive to other losses. In the sections below we estimate the effects of summer raptor predation on autumn grouse density and on subsequent spring density.

Did raptor predation reduce autumn grouse density?. Grouse density in April averaged 14·5 grouse per 0·5 km2 in 1995 and 1996 when data collection was most complete (Fig. 4). Spring mortality reduced density to 10·0 grouse per 0·5 km2 at the end of May. Chick density at hatch in May was 37·4 chicks per 0·5 km2 of which 6·4 were lost to unknown causes and 10·5 were killed by harriers, leaving 20·6 chicks per 0·5 km2 by 17 July. A further 3·4 chicks were killed by harriers by August, leaving 17·3 chicks, which combined with the surviving adults gave an estimated August density with raptors present of 26·5 grouse per 0·5 km2.


Figure 4. Schematic view of grouse demography at Langholm during the period 1995–96. Figures refer to grouse per 0·5 km2. Figures in italics refer to potential grouse chicks lost when adults were killed in early summer.

Download figure to PowerPoint

Estimation of August grouse density without raptors must take into account direct losses of grouse chicks and adult grouse to raptors in early summer (Fig. 4). Adult mortality also resulted in the indirect loss of chicks and these were estimated on the basis that productivity of dead grouse would be similar to survivors. Compensatory increases in other losses of adults and chicks in the absence of raptors were estimated as above. The estimated August density without raptors was 53·4 grouse per 0·5 km2 suggesting that raptor predation reduced August density by about 50% (Table 7).

Table 7. . Estimate of August grouse density in the absence of summer raptor predation based on data for 1995 and 1996. All figures denote grouse per 0·5 km2
Observed August grouse density with raptors26·5
Direct loss
 Adults killed by raptors between April and May+3·9
 Compensatory increase in other deaths of adults−0·1
 Chicks killed by harriers between May and July+10·5
 Compensatory increase in unexplained losses of chicks−1·8
 Chicks killed by harriers between July and August+3·4
Indirect loss as result of loss of adults
 Chicks killed by harriers between May and July+3·9
 Compensatory increase in unexplained losses of chicks−0·7
 Chicks killed by harriers between July and August+1·4
 Chicks which survive to August+6·4
August grouse density in absence of summer raptors53·4

Did raptor predation reduce spring grouse density? We developed a model which used the estimated reduction in autumn grouse density caused by raptor predation in summer together with the observed density dependence in winter loss to predict grouse densities in the absence of raptor predation in summer. The model assumed that October density (O) in year t was proportional to the previous April density (A) as:

  • image

R0 was the reproductive rate which was estimated as October density divided by April density. R0 with raptors was estimated from the observed densities in 1992–96, and equalled 1·5. R0 without raptors was estimated by doubling the observed October densities, on the basis that raptors reduced October density by 50%, and equalled 3·0. We used a non-linear equation to describe the density-dependent relationship between April density in year t + 1 and October density in year t as:

  • image

where a is the proportionate change at low density and b is the density-dependent parameter. We fitted the equation to the observed October and April grouse densities in 1992–96 by non-linear Poisson regression which gave estimates of a = 0·97 (SE = 0·17) and b = 0·0172 (SE = 0·0093). We applied the model with the observed April density in year t to predict April and October densities in years t + 1 and t + 2. We used 15 grouse per 0·5 km2 as a starting point which was the mean April density in 1992–96. Monte Carlo simulation was used to estimate 95% confidence limits on the model predictions based on the observed pattern of variation in R0 and the estimated overdispersion based on the Poisson distribution.

In the presence of raptor predation, the model predicts that grouse density in years t + 1 and t + 2 remains similar to year t (Table 8). In the absence of raptor predation, the model predicts that April density in year t + 2 would be 1·9 times greater, and October density 3·9 times greater, than with raptors present (Table 8).

Table 8. . Model prediction of long-term effects of raptor predation on grouse density in April (A) and October (O) at Langholm. All figures refer to grouse per 0·5 km2
With raptor predation:
At = 1595% C.L.
Ot = 22·519–27
At+1 = 15·713–20
Ot+1 = 23·518–32
At+2 = 16·112–21
Ot+2 = 24·217–34
Without raptor predation:
At = 1595% C.L.
Ot = 4538–54
At+1 = 24·520–30
Ot+1 = 73·456–96
At+2 = 31·326–37
Ot+2 = 93·872–120


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Effect of predation on autumn density

Raptor predation in summer appeared to reduce autumn grouse density in two ways: directly through the predation of adults and chicks, and indirectly through the loss of potential chicks resulting from the deaths of breeding adults. Harriers killed about 37% of the chicks that hatched and most of these losses occurred after an early period of high chick mortality. Previous work indicates that grouse chick mortality as a result of starvation and chilling can be high in the first 2 weeks of life (Erikstad 1985; Hudson 1986). The occurrence of chick losses prior to harrier predation, and the lack of density dependence in chick losses, suggests that harrier predation of grouse chicks at Langholm was mostly additive, supporting earlier work by Redpath (1991). Both studies, however, were conducted on moors where corvids and foxes were killed by gamekeepers and, under natural conditions, losses of chicks to harriers may be partially compensated by increased losses to other predators. Predator removal experiments with ptarmigan and voles suggest that removal of one group of predators may be partially compensated by increased predation by other species in the same guild (Parker 1984; Korpimaki & Norrdahl 1998).

Predation on adult grouse during April and May appeared to reduce breeding densities by about 30%. This predation was almost entirely attributed to raptors, but whether harriers or peregrines were responsible could not be determined. It occurred at a time when virtually all female grouse were paired and there appeared to be little replacement of breeding females by non-territorial birds. Mammalian predator control was intense during spring throughout the study and fox predation on breeding grouse was low. Such levels of control, and the lack of density dependence in fox predation (Leckie et al. 1998), suggests that relatively little raptor predation could have been compensated by increased losses to foxes. Under natural conditions, with an intact guild of predators, it is possible that some raptor predation may have been offset by increased losses to other predators, as described above. Parasite burdens were low throughout the study and it seemed unlikely that they strongly influenced grouse survival or productivity. However, if grouse densities increased in the absence of raptors, parasite burdens might increase and influence grouse dynamics, although the extent to which this might occur is unknown (Hudson, Newborn & Dobson 1992a).

The combined effect of direct and indirect losses to raptors during summer appeared to reduce autumn grouse density at Langholm by about 50%. This remains a tentative conclusion as it is dependent upon certain assumptions regarding compensation which we were unable to test experimentally. If summer predation is such an important demographic event, however, why was it not identified in previous red grouse studies? Summer losses were low during the early work on grouse in NE Scotland, presumably because predator densities were also low (Jenkins et al. 1963, 1964). Fox predation on breeding females was high during recent studies in NE Scotland but was not thought to have caused the observed decline in density (Moss et al. 1990, Moss, Watson & Parr 1996). The numbers of grouse corpses recovered during a recent study in Highland Scotland peaked during March and April and it was suggested that these late winter losses to foxes and raptors reduced breeding density (Hudson 1992; Hudson, Newborn & Robertson 1997).

Effect of predation on spring density

If raptor predation reduced autumn density of grouse by 50% within a single year, what effect did this have on spring and autumn density in subsequent years? We used our estimates of summer loss combined with the observed density dependence in winter loss in a simple model to predict spring and autumn density in the absence of summer raptor predation. Although there were uncertainties in the model, it predicted that within 2 years of removing raptors, grouse density in spring would increase by 1·9 times and density in autumn by 3·9 times. The model did not incorporate any delayed density dependence, thus its usefulness in making longer-term predictions of equilibrium densities in the absence of raptors was limited. However, it did suggest that both spring and autumn densities of grouse would increase in the absence of raptors, and it made quantitative predictions that could be tested experimentally.

There has been considerable debate regarding the role of winter predation in reducing breeding densities of grouse in spring (Watson 1985; Hudson 1992). Although we were able to document the extent and cause of winter mortality, we were unable to determine the extent to which it reduced spring density. Predation was not restricted to non-territorial grouse, and during the latter half of the winter there were few surplus birds to replace territorial birds that were killed. However, count areas gained birds through immigration, and these birds presumably locally compensated some of the losses to raptors. We did not know where these birds came from or whether they were territorial before moving. At the scale of the count areas, movement of grouse could have locally compensated between 55% and 95% of the winter mortality. At the scale of the grouse population, however, the extent of compensation was unknown, because we did not know the extent to which birds increased their breeding success through movement. Future studies will need to work at spatial scales large enough to incorporate the dispersal distance of grouse to answer this question.

Generalist predators and population dynamics

The main finding of this study was that predation in summer by harriers and peregrines appeared to reduce autumn densities of grouse by about 50%. Modelling indicated that this predation also prevented spring and autumn densities from increasing in subsequent years, although these increases would not be boundless because of density dependence in winter loss. Predictions of increased density in the absence of raptors were supported by comparison of shooting bag records at Langholm with two nearby moors where raptors continued to be controlled (Thirgood et al. 2000a). The numbers of grouse shot at Langholm and the nearby moors cycled in synchrony with a 6-year periodicity during the 1970s and 1980s when raptors were uncommon. Grouse bags at Langholm declined year on year from 1990 to 1998, mirroring the increase in raptor numbers and decline in grouse density, whilst bags on the nearby moors increased to cyclic peaks in 1996. Raptor predation appeared to hold the Langholm grouse population at low density and prevent cycling. Generalist predators were predicted to dampen red grouse cycles in the model of Hudson, Dobson & Newborn (1992b), but in this model the mechanism was through selective removal of heavily parasitized grouse reducing the effect of the parasite on grouse reproduction that is thought to generate the population cycles (Hudson, Dobson & Newborn 1998).

Theoretical studies suggest that generalist predators may be able to stabilize fluctuations in their prey if predator densities are unrelated to prey densities and predation is density dependent (Hanski et al. 1991; Hanski & Korpimaki 1995; Turchin & Hanski 1997). The raptor–grouse system described here appears to fit these models fairly well. Harriers and peregrines respond functionally but not numerically to the abundance of grouse, and high densities of raptors occur on moors where small prey are abundant (Redpath & Thirgood 1999). If grouse populations fall to low densities on these moors, density-dependent raptor predation can suppress recovery. This pattern is reminiscent of the supposed effects of generalist predators in suppressing vole cycles in southern Fennoscandia (Angelstam, Lindstrom & Widen 1984; Erlinge 1987; Korpimaki & Norrdahl 1991, 1998). In the raptor–grouse system, however, regional effects of predation are confounded by illegal control of raptors (Etheridge, Summers & Green 1997; Thirgood et al. 2000b).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank the Buccleuch Estates for allowing this work to take place on their moor. We are grateful to M. Avery, I. Bainbridge, C. Galbraith, P. Hudson, R. Moss, D. Potts, P. Robertson, D. Thompson, A. Watson and, especially, I. Newton for advice. Comments from S. Albon, R. Green, K. Lessels, K. Norrdahl and an anonymous referee greatly improved this paper. G. Buchanan, S. Campbell, C. Coles, C. Cronin, E. Donnelly, M. Denny, R. Foster, I. Graham, C. Hill, J. Jansen, J. Johnson, K. Laurenson, F. Leckie, P. Lindley, A. Martin, J. Martinez, R. May, D. Newborn, D. Parish, C. Redpath, A. Smith, T. Smith, A. Tharme and A. Walton helped with fieldwork. Briadh, Islay, Jura and especially Jet worked harder than any of us and were great companions on the hill. This research was funded by Buccleuch Estates, Game Conservancy Trust, Institute of Terrestrial Ecology, Joint Nature Conservation Committee, Royal Society for the Protection of Birds, Scottish Natural Heritage, Westerhall Estates and Natural Environment Research Council (Grant GR9/1507).


  1. Top of page
  2. Summary
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Anderson, D.R. & Burnham, K.P. (1976) Population Ecology of the Mallard VI. The Effect of Exploitation on Survival. US Fish and Wildlife Resource Publication, No. 128, 1 66.
  • Andersson, M. & Erlinge, S. (1977) Influence of predation on rodent populations. Oikos, 29, 591 597.
  • Angelstam, P., Lindstrom, E., Widen, P. (1984) Role of predation in short-term fluctuations of birds and mammals in Fennoscandia. Oecologia, 62, 199 208.
  • Boveng, P.L., Hiruki, L.M., Schwartz, M.K., Bengtson, J.L. (1998) Population growth of Antarctic fur seals: limitation by a top predator, the leopard seal? Ecology, 79, 2863 2877.
  • Erikstad, K.E. (1985) Growth and survival of willow grouse chicks in relation to home range, brood movements and habitat. Ornis Scandinavica, 16, 181 190.
  • Erlinge, S. (1987) Predation and noncyclicity in a microtine population in southern Sweden. Oikos, 50, 347 352.
  • Erlinge, S., Goransson, G., Hansson, L., Hogstedt, G., Liberg, O., Von Nilsson, I., Schantz, T., Sylven, M. (1983) Predation as a regulating factor on small rodent populations in southern Sweden. Oikos, 40, 36 52.
  • Etheridge, B., Summers, R.W., Green, R. (1997) The effects of illegal killing and destruction of nests on the population dynamics of hen harriers in Scotland. Journal of Applied Ecology, 34, 1081 1106.
  • Hanski, I. & Korpimaki, E. (1995) Microtine rodent dynamics in northern Europe: parameterized models for the predator–prey interaction. Ecology, 76, 840 850.
  • Hanski, I., Hansson, L., Henttonen, H. (1991) Specialist predators, generalist predators and the microtine cycle. Journal of Animal Ecology, 60, 353 367.
  • Hudson, P.J. (1986) Red Grouse: The Biology and Management of a Wild Gamebird. Game Conservancy Trust, Fordingbridge, UK.
  • Hudson, P.J. (1992) Grouse in Space and Time: The Population Biology of a Managed Gamebird. Game Conservancy Trust, Fordingbridge, UK.
  • Hudson, P.J., Dobson, A.P., Newborn, D. (1992b) Do parasites make prey vulnerable to predation? Journal of Animal Ecology, 61, 681 692.
  • Hudson, P.J., Newborn, D., Dobson, A.P. (1992a) Regulation and stability of a free-living host–parasite system: Trichostrongylus tenuis in red grouse. I. Monitoring and parasite reductions. Journal of Animal Ecology, 61, 477 486.
  • Hudson, P.J., Newborn, D., Robertson, P.A. (1997) Geographical and seasonal patterns of mortality in red grouse populations. Wildlife Biology, 3, 79 87.
  • Hudson, P.J., Dobson, A.P., Newborn, D. (1998) Prevention of population cycles by parasite removal. Science, 282, 2256 2258.
  • Jenkins, D., Watson, A., Millar, G. (1963) Population studies on red grouse in north-east Scotland. Journal of Animal Ecology, 32, 317 376.
  • Jenkins, D., Watson, A., Millar, G. (1964) Predation and red grouse populations. Journal of Applied Ecology, 1, 183 195.
  • Korpimaki, E. & Norrdahl, K. (1991) Do breeding nomadic avian predators dampen population fluctuations of small mammals? Oikos, 62, 195 208.
  • Korpimaki, E. & Norrdahl, K. (1998) Experimental reduction of predators reverses the crash phase of small-rodent cycles. Ecology, 79, 2448 2455.
  • Krebs, C.J., Boutin, S., Boonstra, R., Sinclair, A.R.E., Smith, J.N.M., Dale, M.R.T., Martin, K., Turkington, R. (1995) Impact of food and predation on the snowshoe hare cycle. Science, 269, 1112 1115.
  • Leckie, F.M., Thirgood, S.J., May, R., Redpath, S.M. (1998) Variation in the diet of red foxes on moorland in relation to prey abundance. Ecography, 21, 599 604.
  • Lindstrom, E., Andren, H., Angelstam, P., Cederlund, G., Hornfeldt, B., Laderberg, L., Lemnell, P., Martinsson, B., Skold, K., Swenson, J. (1994) Disease reveals the predator: sarcoptic mange, red fox predation and prey populations. Ecology, 75, 1042 1049.
  • Marcström, V., Engren, E., Kenward, R.E. (1988) The impact of predation on boreal tetraonids during vole cycles—an experimental study. Journal of Animal Ecology, 57, 859 872.
  • Messier, F. (1994) Ungulate population models with predation: a case study with the North American moose. Ecology, 75, 478 488.
  • Moss, R., Trenholm, I., Watson, A., Parr, R. (1990) Parasitism, predation and survival of hen red grouse in spring. Journal of Animal Ecology, 59, 631 642.
  • Moss, R., Watson, A., Parr, R. (1996) Experimental prevention of a population cycle in red grouse. Ecology, 77, 1512 1530.
  • Parker, H. (1984) Effect of corvid removal on reproduction of willow ptarmigan. Journal of Wildlife Management, 48, 1197 1205.
  • Picozzi, N. (1978) Dispersion, breeding and prey of the hen harrier in Glen Dye, Kincardineshire. Ibis, 120, 489 509.
  • Pollock, K.H., Winterstein, S.R., Bunck, C.M., Curtis, P.D. (1989) Survival analysis in telemetry studies. Journal of Wildlife Management, 53, 7 15.
  • Redpath, S.M. (1991) The impact of hen harriers on red grouse breeding success. Journal of Applied Ecology, 28, 659 671.
  • Redpath, S. & Thirgood, S. (1997) Birds of Prey and Red Grouse. HMSO, London.
  • Redpath, S.M. & Thirgood, S.J. (1999) Numerical and functional responses in generalist predators: hen harriers and peregrines on Scottish grouse moors. Journal of Animal Ecology, 68, 879 892.
  • Reid, D.G., Krebs, C.J., Kenney, A. (1995) Limitation of collared lemming population growth at low densities by predation mortality. Oikos, 73, 387 398.
  • Schall, R. (1991) Estimation in generalised linear models with random effects. Biometrika, 78, 719 727.
  • Tapper, S.C., Potts, G.R., Brockless, M.H. (1996) The effect of experimental reductions in predator presence on the breeding success and population density of grey partridges. Journal of Applied Ecology, 33, 968 979.
  • Thirgood, S.J., Redpath, S.M., Hudson, P.J., Hurley, M.H., Aebischer, N.J. (1995) Effects of necklace radio transmitters on survival and breeding success of red grouse. Wildlife Biology, 1, 121 126.
  • Thirgood, S.J., Redpath, S.M., Hudson, P.J., Donnelly, E. (1998) Estimating the cause and rate of mortality in red grouse. Wildlife Biology, 4, 65 71.
  • Thirgood, S.J., Redpath, S.M., Haydon, D.T., Rothery, P., Newton, I., Hudson, P.J. (2000a) Habitat loss and raptor predation: disentangling long- and short-term causes of red grouse declines. Proceedings of the Royal Society, Series B, 267, 651 656.
  • Thirgood, S.J., Redpath, S.M., Newton, I., Hudson, P.J. (2000b) Raptors and red grouse: conservation conflicts and management solutions. Conservation Biology, 14, 95 104.
  • Turchin, P. & Hanski, I. (1997) An empirically based model for latitudinal gradient in vole population dynamics. American Naturalist, 149, 842 874.
  • Watson, A. (1985) Social class, socially-induced loss, recruitment and breeding in red grouse. Oecologia, 67, 493 498.

Received 10 February 1999;revisionreceived 16 November 1999