Does supplementary feeding reduce predation of red grouse by hen harriers?

Transcription

1 Ecology , Blackwell Oxford, JPE Journal British December Ecological of UK Science 2001 Applied Ltd Society, Ecology2001 PRIORITY CONTRIBUTION Supplementary S.M. Redpath, S.J. feeding Thirgood of hen & harriers F.M. Leckie Does supplementary feeding reduce predation of red grouse by hen harriers? STEPHEN M. REDPATH*, SIMON J. THIRGOOD and FIONA M. LECKIE* *CEH Banchory, Hill of Brathens, Banchory, Aberdeenshire AB31 4BW, UK; and Game Conservancy Trust, ICAPB, University of Edinburgh, West Mains Road, Edinburgh EH9 3JT, UK Summary 1. Hen harriers Circus cyaneus can reduce the numbers of red grouse Lagopus lagopus scoticus available for shooting. We conducted a supplementary feeding experiment on Langholm Moor, UK, in 1998 and 1999 to determine whether feeding hen harriers could reduce the numbers of red grouse killed. The experiment was done at two distinct stages of the breeding cycle: prior to incubation (spring experiment) and after hatching (summer experiment). In spring, Langholm Moor was divided into two areas, one with food and one without. In summer a number of birds were provided with food in both areas. 2. Providing harriers with food in spring had no significant effect on the breeding density of males or females, although feeding was associated with an increase in density on one area in one year. In addition, over the 2 years of the experiment, there was no evidence that feeding led to more chicks returning to breed in subsequent years. Fed harriers had larger clutches but did not lay earlier than unfed birds. 3. A minimum of 78% of the radio-tagged grouse that were killed during spring were killed by raptors. The mortality rates of adult grouse did not differ between the two areas or between the two years despite the availability of supplementary food and the large differences in harrier breeding density between areas. We infer that other raptors were responsible for much of the predation of adult grouse. 4. During the nestling period, female harriers took supplementary food at a higher rate than males. Females that were fed during the spring took more supplementary food in summer than those fed only during summer. Fed birds did not deliver more food overall to nests than those not provided with food. 5. Both male and female harriers at nests where supplementary food was available caught grouse chicks at a lower rate than harriers at nests not provided with food. For both years combined, fed harriers delivered on average 0 5 grouse chicks to their nests per 100 h, compared with 3 7 grouse chicks delivered to nests without supplementary food. 6. We estimated that feeding all harriers at Langholm would cost approximately per annum. In both 1998 and 1999, the numbers of grouse chicks lost were 10 times higher than expected from harrier predation rates. Some other, unknown, factor had a strong influence on grouse chick survival in these years. Feeding some of the breeding harriers did not lead to an increase in grouse density at Langholm. 7. The results suggest that supplementary feeding may provide a useful tool in reducing the number of grouse chicks taken by harriers. Further experiments are now necessary to see under what conditions this reduced predation will lead to increases in grouse density. Key-words: Circus cyaneus, gamebirds, heather moorland, Lagopus l. scoticus, raptors, wildlife management. Ecology (2001) 38, Ecological Society Correspondence: Steve Redpath, CEH Banchory, Hill of Brathens, Banchory, Aberdeenshire AB31 4BW, UK (fax ; S.Redpath@ceh.ac.uk). Present address: Centre for Conservation Science, Stirling University, Stirling FK9 4LA, UK.

2 1158 S.M. Redpath, S.J. Thirgood & F.M. Leckie Introduction Heather moorland is an internationally important habitat and its maintenance is viewed as a conservation priority (Anonymous 1995; Thompson et al. 1995). Within this habitat red grouse Lagopus lagopus scoticus Lath. management represents an important form of land use, the aim of which is to maximize the number of grouse available for shooting, through heather burning and the control of parasites and predators (Hudson 1992; Hudson & Newborn 1995). As part of the predator control, raptors have traditionally been killed, as have red foxes Vulpes vulpes L. and corvids. Despite the fact that raptors have received full legal protection since 1954, their persecution on grouse moors is still widespread (Etheridge, Summers & Green 1997; Scottish Raptor Study Groups 1997; Green & Etheridge 1999). Recent research strongly suggested that predation during spring and summer by high densities of generalist predators, particularly hen harriers Circus cyaneus L., was able to limit grouse populations at low density and to reduce shooting bags (Redpath & Thirgood 1997; Redpath & Thirgood 1999; Thirgood et al. 2000a,c). These findings have highlighted the conflict between those who wish to manage grouse and those who wish to conserve raptors (Thirgood et al. 2000b). One possible short-term solution to this conflict is to provide harriers with supplementary food throughout the breeding season, thus reducing predation rates on grouse. The effects of feeding hen harriers have not previously been investigated, although Simmons (1994) showed that African marsh harriers Circus ranivorus Daudin took supplementary food in spring. In many raptors, breeding density is correlated with food abundance (Newton 1998). Such correlations exist for harriers in Scotland and elsewhere (Hamerstrom 1979; Korpimaki 1985; Redpath & Thirgood 1999; Redpath, Thirgood & Clarke 2002). It is possible, therefore, that additional food for harriers may increase their breeding density, thus reducing the possible benefits of any reduced predation on grouse chicks. In addition, supplementary feeding in other raptors tends to advance laying, increase clutch size (Dijkstra et al. 1980; Newton & Marquiss 1981; Aparicio 1984; Korpimaki 1987; Korpimaki & Wiehn 1998) and improve fledging success (Korpimaki 1987; Gehlbach & Roberts 1997; Wiehn & Korpimaki 1997). We therefore predicted that harriers provided with supplementary food would show improved breeding performance. We examined the effectiveness of supplementary feeding as a management tool in reducing predation by hen harriers on adult grouse during the spring and on grouse chicks during the summer. We examined the effect of feeding on harrier breeding density and success, and we considered the financial costs and benefits of such a feeding programme. Methods This study was based on Langholm Moor, located at N and 2 55 S, in south-west Scotland (Redpath & Thirgood 1997, 1999). The feeding experiment was conducted in 1998 and 1999, but data from years before ( ) and after (2000) the experiment were also included. Throughout the paper we have divided the study into spring and summer sections. Spring refers to the period between harriers returning to their breeding sites up to clutch completion (March April), and summer refers to the period between hatching and chick dispersal (May August). SPRING FEEDING EXPERIMENT On the basis of the distribution of heather-dominant vegetation (Thirgood et al. 2000a) and harrier nests in 1997, Langholm Moor was divided into two areas, A and B (Fig. 1). During spring, harriers were provided with food (treatment) in area A in 1998 and in area B in Harriers on the other area were not fed (control). No food was provided in Males were located as they returned to their breeding sites in March and feeding perches were erected in their territories in the areas where males were displaying and prospecting for nests. Perches consisted of standard 1 5-m high 10-cm diameter fence posts with a 30-cm section of post nailed across the top, to form a T-shape. Supplementary food, primarily dead white rats and day-old poultry chicks, was placed daily on the perches from late March until the start of incubation. Any food remaining the next day was removed and disposed of. Because feeding attracted potential harrier egg predators (see later) we did not continue feeding during the incubation period. We noted the number of males establishing territories on each area in spring, defined as birds consistently seen displaying and defending one area of moorland. We also recorded the number of females breeding (i.e. laying at least one egg) with each male. Despite the fact that juvenile (brown) males were often seen displaying in the spring, there was only one instance of a juvenile male breeding. The remaining non-breeding juvenile males were excluded from all analyses. In some adult (grey) males held territories throughout spring and summer but failed to breed. As some of these males were provided with food during 1998 and 1999, these birds have been included in analyses of the effect of feeding on the numbers of females breeding per male. Nests were generally found during egg laying or early incubation, and both laying date (first egg) and clutch size were recorded. Where possible female harriers were aged as first year birds or older, based on wingtags (colour-coded tags had been fitted to nestlings in previous years) or eye colour (first year birds have brown eyes). Harrier breeding status was recorded as monogamous or polygynous (two or more females) for males, and as monogamous and either primary or secondary for females. Polygynous females were

3 1159 Supplementary feeding of hen harriers Langholm Area B Area A 5 km Fig. 1. Location of Langholm Moor in Scotland and the division of the study area into two experimental areas. Symbols indicate the distribution of harrier nests in 1998 ( ) and 1999 ( ). Nests joined by lines indicate polygynous males. Harriers in area A were provided with food during spring 1998, and harriers in area B were provided with food in spring classified according to laying date, with the earliest laying female being considered the primary female. In 1998 there was one tertiary female classified for analyses of breeding success as a secondary female. The abundance of small mammals and adult grouse was estimated in all years using standard techniques. Small mammal abundance was estimated from 10 lines of 50 snap traps set for two nights in March, and grouse were counted on km 2 areas in late March. Briefly, a trained dog quartered the ground in front of the observer and flushed the grouse (for techniques see Redpath & Thirgood 1997). In both 1998 and 1999, some grouse were fitted with 15-g necklace radio-transmitters ( Biotrack Ltd, Wareham, UK): in area A, 51 grouse in 1998 and 48 in 1999; in area B, 49 grouse in 1998 and 43 in Birds were caught at night in hand-held nets after dazzling them with a strong light. All birds were relocated once before 1 April and their survival was monitored weekly until the end of June. From these data we calculated weekly survival rates on the basis of the spring season starting on 1 April (Thirgood et al. 2000c). Survival data were compared within experimental years and also with similar data collected in years before the experiment. Raptor and mammalian predators could be distinguished from field signs (Thirgood et al. 1998). SUMMER FEEDING EXPERIMENT Birds at a number of nests from each area were provided with supplementary food from hatch to chick dispersal, partly because harrier breeding density and laying date varied between treatment and control areas (see the Results) and also to avoid the effects of the two experiments being confounded. Nests were allocated randomly to experimental or control (not fed) groups, within the constraints that both groups showed similar median laying dates and breeding system. Supplementary food was provided to nine nests in 1998 and five nests in No food was provided to six nests in 1998 and to five nests in No food was provided in the years before or after the experiment. Once harrier chicks had hatched, perches were placed on average 9 ± 1 m (mean ± SE) away from the nests. Food was placed on perches each morning, the amount varying according to the food requirements of the chicks. Estimates of maximum food requirements were based on previous observations of food deliveries to young harriers at Langholm (S.M. Redpath & S.J. Thirgood, unpublished data). As in the spring, any food remaining the next day was removed. Harriers were provided with food from hatching for 60 days, the estimated time to dispersal (Redpath & Thirgood 1997). To examine what the harriers were eating, we set up hides 5 7 m from each nest. We aimed to spend a minimum of two 6-h watches per nest per week, for 5 weeks from hatching. In 1998 and 1999 all nests were watched, with the exception of one nest in 1999 with no additional food, which was judged to be too close to a public road. Nests were also watched in (Redpath & Thirgood 1999). No nests were watched in

4 1160 S.M. Redpath, S.J. Thirgood & F.M. Leckie 1997 or During each watch we recorded the time watched and the identity and quantity of food brought to the nests by males and females. After fledging, some wing-tagged broods (12 unfed and six fed) were observed every 2 3 days to record dispersal date, taken as the day after the last day on which the birds were seen on the moor. In May, radio-receivers were used to locate the nests of grouse carrying transmitters, so that clutch size and the number of chicks hatching could be recorded. In early June and late July, grouse broods were located with trained pointing dogs on transects through count areas and the number of chicks in each brood was counted. As there was approximately 41 5 km 2 of suitable grouse habitat on the estate (Redpath & Thirgood 1997), we were able to estimate the number of grouse taken by harriers per 0 5 km 2. We thus obtained estimates of the number of grouse chicks available to harriers and the number present after the main period of harrier predation (Thirgood et al. 2000c). Based on the watches at harrier nests we estimated the number of grouse chicks taken by harriers as follows: grouse taken = grouse delivered nest 1 h 1 no. nests 15 X (Redpath & Thirgood 1997), where 15 = number of hours available for hunting per day (Watson 1977) and X = the number of days from hatching. Two calculations were made, for the period from hatching to fledging (X = 42) and hatching to dispersal (X = 60). The percentage of grouse chicks in the pellets of harriers during the first 4 weeks was not different from the percentage during weeks 5 6 and from weeks 7 up to dispersal (Redpath & Thirgood 1997). This suggests that predation rates on grouse chicks do not change as harrier chicks get older. STATISTICAL ANALYSES Analyses were done in Minitab, version 13 (Minitab Inc. 2000) and SAS version 6.12 (SAS Institute 1990). To test for the effects of supplementary feeding on harrier breeding parameters and on the provisioning rate of grouse chicks to harrier nests, we incorporated data from years before and after the experiment. Unless otherwise stated, we used generalized linear mixed models (GLMM) with both year and area included in all models as random effects. Parameters that had nonnormal error structures were analysed with a Poisson error structure and a log-link function; otherwise models incorporated a normal error function and an identity function. Extra dispersion in all the models was corrected for by dividing the deviance by the residual degrees of freedom. The models were implemented using a GLIMMIX macro in SAS (Littell et al. 1996). Denominator degrees of freedom were calculated in SAS using Satterthwaite s formula (Littell et al. 1996). Models were constructed using a backward elimination procedure in a SAS type III analysis, dropping the least significant term in the subsequent model, until only terms significant at the 10% level remained. A general linear model (GLM) was used to test for differences in the numbers of breeding harriers in areas A and B in relation to feeding. Numbers of males have increased over the course of the study and are correlated with field vole Microtus agrestis L. abundance (Redpath, Thirgood & Clarke 2002). To examine the effect of supplementary feeding on the numbers of harriers, year and vole numbers in areas A and B were included as covariates, with area (A and B) and supplementary feeding as fixed effects. We used an autocorrelation function to test for serial correlation among the residuals of the fitted models. As sample size was small, we also conducted a randomization test, by reiterating the GLM for the 72 possible combinations of feeding in the two areas in separate years. For other models, the moor was divided up into 10 nesting areas of approximately 200 ha (five each in areas A and B). These areas included the sites of all the nesting attempts and nesting occurred in each area in at least one year. Nesting areas were included in GLMM models as random effects, to control for spatial effects of nest location on harrier breeding performance and rates of grouse chick provisioning. Over the 8 years, six females failed and relaid a clutch. Clutch size and laying date were taken for the first clutch. Models examining variation in provisioning of prey to harrier nests incorporated the log of the hours watched as an offset function along with female status and whether or not birds were fed in spring and summer as fixed effects. For models of male provisioning, a male identification number was included as a random effect. For models of grouse provisioning the density of grouse chicks (log) in any one year was also included. When comparing adult grouse survival in relation to spring feeding, we used the Kaplin Meier product limit method and tested for end-point differences in survival using a two-tailed z-test statistic (Redpath & Thirgood 1997). Sample sizes for testing the effects of feeding on adult grouse survival, harrier breeding success and harrier chick dispersal were small and therefore their power was rather low. Throughout the text, means ±1 SE. are presented and analyses other than those indicated above are noted in the text. All tests were two-tailed. Results EFFECTS OF SPRING FEEDING ON HARRIER BREEDING NUMBERS During spring, 193 kg of food was put out in area A on eight territories (eight males, 13 females) in 1998 and 63 kg of food put out in area B on five territories (five males, three females) in Less food was put out in 1999 partly because there were fewer females but also because less food was removed in total. In 1998, 91% of the food had disappeared by the next day, whereas in 1999 only 44% of the food had disappeared. A similar proportion of food was seen being taken in both years

5 1161 Supplementary feeding of hen harriers Harrier numbers Harrier numbers Field voles per 100 trap nights Area A Area B Fig. 2. Changes in the number of field voles caught per 100 trap nights in areas A and B combined and the number of male ( ) and female ( ) harriers breeding in area A and B from 1992 to Arrows indicate year of feeding. (4% in 1998, 3% in 1999), and of this harriers took a similar proportion (15% in 1998, 16% in 1999). The remainder of the food seen being taken (85% in 1998, 86% in 1999) was removed by corvids (ravens Corvus corax L., carrion crows Corvus corone L. and rooks Corvus frugilegus L.). In 1998, the first item seen taken by harriers was on 4 April, whereas in 1999 the first item seen taken was on 4 May. In the 85 recorded male territories established during , 97 females attempted to breed. In the last 3 years, 11 grey males failed to attract a female (four in A and seven in B). In all years area A had more breeding male harriers than area B (area A 5 7 ± 0 8, area B 2 6 ± 0 6). During the years of the supplementary feeding experiment harrier numbers increased in the area where food was provided (A) in 1998, but not in 1999 (Fig. 2). Both area (males, F 1,13 = 38 2, P < 0 001; females, F 1,13 = 25 3, P < 0 001) and vole abundance (males, F 1,14 = 9 89, P = 0 008; females, F 1,13 = 8 2, P = 0 013) had a strong statistical effect on the numbers of breeding harriers but, controlling for these, supplementary feeding in spring had no detectable effect on breeding harrier numbers (males, F 1,13 = 1 20, P = 0 29; females, F 1,13 = 2 35, P = 0 15). There was no statistically significant interaction between feeding and area (GLM males, F 1,12 = 0 89, P = 0 36; females, F 1,12 = 2 92, P = 0 11). We found no significant serial correlation among the residuals of the fitted models (males r = 0 03, females r = 0 15, n = 18, P > 0 1), suggesting that harrier numbers in one year could be considered independent from those in the previous year. In the randomization test the F-values for the effect of feeding on breeding numbers were not unusual for either males or females, with 28% of the randomized values being higher for males and 14% being higher for females. When we considered the effect of supplementary feeding on the numbers of females breeding per male, there was no evidence for a statistically significant effect of feeding, either including (F 1,68 = 0 66, P = 0 42) or excluding (F 1,57 = 0 74, P = 0 39) males that failed to attract females. We concluded that there was little evidence for an effect of supplementary feeding on breeding harrier numbers. In years before and after the feeding experiment, an average of 44% of 55 breeding females that could be aged were first years (range 80% in 1993 to 0% in 1995). During these years, there was no relationship between the percentage of first year birds and vole abundance (F 1,4 = 0 6, P = 0 5). In 1998, of 16 females of known age, four (25%) were first years, which all bred in area A. In 1999, three (25%) of 12 females of known age were first years, two of which bred in area A. There was thus no apparent increase in the proportion of young females breeding in the population during years when birds were fed. Of birds tagged as chicks in 1998 and 1999, six females returned to breed in the next year. Three of these were from nests provided with food and three were from control nests. No chicks reared in 1998 bred in No young males reared in 1998 and 1999 bred during 1999 or Thus feeding did not lead to an increase in young females entering the breeding population in 1999 or 2000, nor did feeding lead to an increase in the return of chicks reared by fed birds. EFFECTS OF SPRING FEEDING ON HARRIER LAYING DATE AND CLUTCH SIZE Laying date varied considerably between years (GLM F 7,79 = 6 7, P < 0 001) and was on average 9 days earlier in area A than B (F 1,79 = 17 1, P < 0 001). Controlling for area and year as random effects in the GLMM model, and female age, female status and field vole abundance (log) as fixed effects, we tested whether feeding in spring could account for any variation in laying date, but we found no effect (F 1,64 = 0 88, P = 0 35). The model estimate for effect of feeding was 0 03 ± In the areas where food was provided, median laying date varied from 21 April in 1998 to 13 May in 1999, a difference of 23 days. This was similar to the difference in dates when harriers were first seen to take food from the perches. This suggests that the use of food in the spring may be a consequence of female condition, rather than the availability of supplementary food.

6 1162 S.M. Redpath, S.J. Thirgood & F.M. Leckie Clutch size Clutch size varied from two to seven eggs, with an overall mean of 5 0 ± 0 1. Clutch sizes were larger in years when field voles were abundant (GLMM F 1,6 = 17 8, P = 0 006) and harriers that were provided with food laid more eggs (Fig. 3; F 1,74 = 4 5, P = 0 04). As there was a strong negative relationship between clutch size and laying date (linear regression F 1,91 = 12 8, P = 0 001), we included laying date in the mixed model. After controlling for laying date, the effects of spring feeding on clutch size were still significant (F 1,73 = 4 1, P = 0 05). So, feeding during the spring had no apparent effect on laying date but led to larger harrier clutches. EFFECTS OF SPRING FEEDING ON ADULT GROUSE SURVIVAL Fig. 3. Relationship between hen harrier clutch size and supplementary feeding in the two experimental years (1998 and 1999). Data presented as means ±1 SE with sample size above. Black bars indicate birds not provided with food in spring and open bars those birds with food provided in spring. Raptors were the main source of mortality for adult grouse during April and May in 1998 and 1999, killing a minimum of 18% of the grouse overall and a minimum of 78% of those grouse which died (Table 1). Survival rates of radio-tagged grouse did not differ significantly between areas A and B in either 1998 or 1999 (Table 2; 1998, Z = 0 48, P > 0 5; 1999, Z = 0 36, P > 0 5). Similarly, survival rate (area A, Z = 0 92, P > 0 1; area B, Z = 0, P = 1 0) did not differ between years. We compared survival rates of all radio-tagged grouse at Langholm in 1998 and 1999 with survival during the spring of 1995 and 1996 when no supplementary feeding was conducted (Redpath & Thirgood 1997). Spring survival rates in 1998 and 1999 tended to be higher, but not significantly (Table 2; 1995 vs. 1998, Z = 1 04, 10 3 P > 0 1, P > 0 1; 1996 vs. 1998, Z = 1 72, P > 0 1; 1996 vs. 1999, Z = 1 18, P > 0 1). EFFECTS OF SUMMER FEEDING ON HARRIER BREEDING SUCCESS Of the 17 female harriers that laid clutches in 1998, three failed at incubation or during hatch. A mammalian predator (probably a stoat) ate one clutch, another clutch disappeared and one young brood apparently starved despite both adults being present. Two of these nests were in area A and one (which relaid) was in area B. Of the 13 females that laid in 1999, four failed due to fox predation in area A during incubation. Two of these females were killed on the nest by foxes and one of the other two relaid. There was no significant effect of spring feeding on the number of harrier chicks that hatched (F 1,29 = 0 06, P = 0 81). Controlling for female age and status did not improve the level of significance (P = 0 95), and excluding nests that failed during incubation had little effect (F 1,52 = 0 22, P = 0 64). Of those nests where young were successfully hatched, there was a tendency for birds fed during the nestling period to fledge more chicks in 1998 and 1999 (3 4 young for fed broods vs. 2 4 young for broods not fed). However, this difference was not significant, when controlling for clutch size and female status (F 1,51 = 1 02, P = 0 32). EFFECTS OF SUMMER FEEDING ON HARRIER CHICK DISPERSAL In 1998 and 1999 we obtained data on estimated dispersal dates for six nests provided with food and compared these to dates from four nests without food in these years and eight nests without food in previous years. Median dispersal dates did not differ (Mann Whitney U = 101, P = 0 24) and were 60 days for unfed broods (range 59 64) and 62 days for fed broods (range 50 66). EFFECTS OF SUMMER FEEDING ON PROVISIONING OF SUPPLEMENTARY FOOD During summer 1998, 6499 food items were put on perches next to nine harrier nests, of which 85% had Table 1. Number of grouse radio-tagged in areas A and B in 1998 and 1999, and numbers killed by raptors, mammals and other causes. Number lost indicates cases where contact was lost with the birds Area A Area B Area A Area B Totals (%) Number radio-tagged Number killed before 1 June (24) Number killed by raptors (18) Number killed by mammals (1) Number killed by unknown (4) Number lost (3)

7 1163 Supplementary feeding of hen harriers Table 2. Survival rates of radio-tagged grouse on treatment and control areas at Langholm Moor, Scotland, during spring 1998 and 1999, in comparison with spring 1995 and Survival rates (S) and 95% confidence intervals (CI) calculated using the Kaplin Meier method with staggered entry. *Areas provided with supplementary food during the spring Year Area n S 95% CI 1995 Total Total Area A* Area B Total Area A Area B* Total Supplementary items 100 h Male Female Year Fig. 4. Mean provisioning rates (±1 SE) of supplementary food by male and female hen harriers in 1998 and Numbers indicate sample sizes. disappeared by the next day. Harriers took 92% of the food seen being taken, while crows and lesser blackbacked gulls Larus fuscus L. took the remaining 8%. In 1999, 4069 items were put out at five harrier nests, of which 69% had disappeared by the next day. Harriers took 83% of the food seen taken with the remaining 17% taken by lesser black-backed gulls. Over the 2 years 2007 h were spent in hides (83 h per nest with food and 84 h per nest without food). We only saw supplementary prey items delivered to nests where food was provided nearby. Over the 2 years, females provisioned 40 supplementary items per 100 h compared with four supplementary items by their males (Fig. 4; t-test, t = 6 3, 19 d.f., P < 0 001). There appeared to be an effect of mating system on provisioning of supplementary food by males. The two bigamous males in 1998 brought in 34 and 11 supplementary items per 100 h, respectively. In contrast, the remaining five monogamous males in 1998 delivered on average 0 6 items in 100 h (range 0 2), and the five monogamous males in 1999 delivered on average 0 2 items in 100 h (range 0 1). For females, the use of supplementary food varied according to whether or not they were fed during the spring. For both years combined, females fed in spring delivered more (GLM controlling for year F 1,11 = 9 01, 5 5 P = 0 012) supplementary items (50 3 ± h 1 ) compared with females not fed in spring (23 3 ± h 1 ). EFFECTS OF SUMMER FEEDING ON PROVISIONING OF NATURAL PREY The availability of supplementary food had a large effect on the provisioning rates of natural prey by females (fed 30 ± 6 vs. unfed 8 ± 2 items 100 h 1, F 1,43 = 18 0, P = ). For males, the provisioning rates of natural prey were slightly higher at those nests where no food was provided (fed 62 ± 4 vs. unfed 51 ± 7). However, differences in provisioning rates between males were not significant (F 1,40 = 2 3, P = 0 13). So the rate at which females delivered wild prey was affected by feeding, but this was not the case for males. Interestingly, although provisioning rates of all prey (including supplementary items) were slightly higher at nests where supplementary food was provided (fed 97 ± 8, unfed 81 ± 7), this difference was not significant once the effects of area, year and female status were controlled (F 1,43 = 2 18, P = 0 15). Despite the fact that fed females provisioned food at a higher rate (F 1,39 = 6 9, P = 0 01), fed males did not (F 1,40 = 1 2, P = 0 28). This suggested that the additional food acted as a substitute rather than a supplement to natural prey. EFFECTS OF SUMMER FEEDING ON PROVISIONING OF GROUSE CHICKS In 1998, 23 grouse chicks were seen delivered to the 15 harrier nests in 1088 h of observation. Of these, only two (9%) were delivered to the nine nests where food was provided. In 1999, 16 grouse chicks were seen delivered to nine harrier nests in 919 h of observation, of which five (31%) were delivered to the five nests where food was provided. There was a significant effect of providing supplementary food during the summer on the rate at which both male and female harriers delivered grouse chicks to their nests (Fig. 5; males, F 1,31 = 4 8, P = 0 03; females, F 1,35 = 6 10, P = 0 02; both combined F 1,41 = 11 8, P = 0 001). Over both years combined, harriers at nests where food was provided delivered 0 5 grouse chicks every 100 h, or 4 5 chicks from hatching to dispersal. In contrast, harriers at nests where no food was provided delivered 3 7 grouse chicks every 100 h, or 33 3 chicks from hatching to dispersal. Other factors that were significant in the models were spring feeding for females (F 1,36 = 6 03, P = 0 02), status of females (F 2,39 = 6 49, P = 0 004) and grouse abundance for males (F 1,4 = 8 6, P = 0 04) and both sexes combined (F 1,4 = 23 5, P = 0 008). In 1998 and 1999 harriers delivered grouse chicks to their nest at a lower rate than during In years prior to the experiment males delivered grouse chicks at h 1, compared with 2 3 in 1998 and 1999 (data for nests without supplementary food). For females the rate dropped from 6 7 to 1 9 grouse. These differences

8 1164 S.M. Redpath, S.J. Thirgood & F.M. Leckie Grouse chicks 100 h Total Male Female Total Male Female 1998 Fed Unfed 1999 Fed Unfed Fig. 5. Mean rate (±1 SE) at which harriers delivered grouse chicks to their young in 1998 and Data shown separately for males and females at nests with and without food. spring numbers have declined from km 2 to km 2. Feeding harriers in 1998 and 1999 did not lead to an increase in grouse density. The average losses of grouse chicks from hatch to late July increased from 45% in to 62% in (Table 3). However, this increased loss was not due to increased predation by harriers. Harriers took only 6 5% of the available grouse chicks in the years of the feeding trial, compared with 28% in The reduction in grouse chick losses to harriers in the last 2 years was partly as a result of a lower density of grouse chicks on the moor. In 1998, we estimated that harriers took 187 ± 80 grouse chicks, or 2 2 ± km 2. Yet observed chick losses from hatching to late July were 21 7 chicks 0 5 km 2, or 10 times higher than the expected value. Similarly in 1999, we estimated that harriers took 107 ± 53 grouse chicks, or 1 3 ± km 2. Yet observed losses were 12 7 chicks 0 5 km 2, or 10 times higher than expected had harriers been the sole mortality agent. Grouse 0 5 km Year were significant (t-tests, male, t = 2 6, 19 d.f., P = 0 018; female, t = 4 24, 25 d.f., P < 0 001). CHANGES IN GROUSE DENSITY AND GROUSE BREEDING SUCCESS Autumn Spring Fig. 6. Change in spring and autumn grouse density on Langholm Moor, Scotland, from 1993 to Data as means (±1 SE) from km 2 count areas within areas A and B. From 1993 to 2000 there was a significant decline in the number of grouse counted in autumn, from km 2 to km 2 (Fig. 6) and in the numbers counted in spring (Thirgood et al. 2000c). Since 1997, COSTS OF SUPPLEMENTARY FEEDING Costs of feeding will vary depending on the number and location of harriers requiring feeding, the number of additional staff employed, the type of food used and the transport costs of getting to the harriers each day. The figures provided here are a rough estimate of the costs associated with the technique. Had all harriers been fed, the total weight of food needed would have been 1020 kg in 1998 and 639 kg in If rats (160 g) only were used, the cost ( 0 30 each) would have been 1912, or 127 per harrier, in 1998, and 1198, or 120 per harrier, in Had chicks (40 g) only been used, the cost ( each) in 1998 would have been 637 ( 42 per harrier) and in 1999 would have been 399 ( 40 per harrier). These costs could potentially be reduced through the provision of other prey such as locally caught rabbits. Transport costs varied between years, as most nests with food were reasonably close to public roads in 1998, but not in In 1998, we estimated costs at 875. In 1999 we hired an all-terrain vehicle for 2000 and covered an estimated 6000 miles, equivalent to In total therefore travel costs were 3500 in The main cost of supplementary feeding of hen harriers will lie in wages. Feeding all the harriers on Table 3. Changes in grouse brood size and chick density per 0 5 km 2 from May to July in comparison with the number of grouse chicks estimated to have been taken by harriers from hatching to fledging. Data from two periods are given: , when no feeding of harriers occurred, and when some harriers were provided with food Year Grouse brood size at hatch May chicks July chicks Difference May July 0 5 km km km 2 (%) Grouse brood size in July Harrier nests (%) Taken by harriers 0 5 km ± (42) (8) 11 2 (29) ± (48) (12) 9 7 (27) ± (67) (15) 2 2 (7) ± (56) (10) 1 3 (6)

9 1165 Supplementary feeding of hen harriers Table 4. Estimates of the financial cost of feeding all breeding harriers at Langholm Moor, Scotland, in 1998 and 1999 Cost Years combined Food Transport Miscellaneous Wages Total Total per harrier nest Langholm from March to August would be equivalent to one full-time job, plus additional casual help. An approximate cost of such staff-time for 5 months would be Had all the harriers been fed at Langholm, we estimated that the cost per annum would have been between and (Table 4). Discussion Providing harriers with supplementary food greatly reduced the rate at which grouse chicks were delivered to harrier nests. Over both years combined, harriers at nests where food was provided delivered one grouse chick to their nest every 200 h, whereas harriers without supplementary food delivered one grouse chick every 27 h. This alone suggests that supplementary feeding provides a useful tool in reducing the number of grouse chicks taken by hen harriers. SPRING FEEDING EXPERIMENT The two areas of Langholm Moor used for the experiment varied considerably in the numbers of harriers that bred there, with area A consistently attracting more birds. Redpath & Thirgood (1997) attributed this pattern to differences in elevation, with birds breeding at higher density and laying earlier on low-lying heather areas. Although feeding was associated with an increase in harrier density in one year, overall we found no statistical evidence that feeding harriers in spring led to increases in breeding density. After controlling for the effects of year and vole abundance, there was no apparent increase in breeding males in areas where food was provided. Nor did these fed males breed with more females than unfed males. In addition, we found no evidence that chicks from fed broods were more likely to return to breed in subsequent years. While several studies have found a strong correlation between raptor density and food supply, few studies of raptors have examined the effect of artificial feeding on density, although this issue has been addressed in a variety of other species (Newton 1998). In hen harriers, breeding density is correlated with food supply, both in Scotland (Redpath & Thirgood 1999; Redpath, Thirgood & Clarke 2002) and elsewhere (Hamerstrom 1979; Korpimaki 1985). However, the finding that feeding did not increase male density is perhaps unsurprising, given that feeding was only directed at males that had already established a territory. Although these males fed their females with the supplementary food, birds did not take this extra food until females were approaching egg laying, when their food requirements were increased. In a supplementary feeding experiment with sparrowhawks Accipiter nisus L., Newton & Marquiss (1981) found that all fed birds laid clutches whereas 27% of unfed birds failed to lay eggs. Similarly, in a further feeding experiment with harriers in Orkney, Amar & Redpath (in press) found that feeding led to increases in the number of females that laid a clutch. This did not appear to be the case in the present study, as apparently all females seen consistently on territory in spring went on to breed. The two other studies occurred in areas where food was considered scarce, so a difference in natural prey availability may account for the difference in breeding attempts. Several studies have found an effect of supplementary feeding on clutch size and laying date (Dijkstra et al. 1980; Newton & Marquiss 1981; Aparicio 1984; Korpimaki 1987; Simmons 1994; Korpimaki & Wiehn 1998). In addition, Redpath, Thirgood & Clarke (2002) found a close correlation between vole abundance and clutch size, so a relationship between supplementary feeding and clutch size was expected. In our study, fed females laid larger clutches but laying date did not appear to be influenced by feeding. The timing of food removal suggested that supplementary items were taken only when females were approaching laying date, rather than the food causing a change in laying date. Adult grouse survival was slightly higher in 1998 and 1999 than it was in 1995 and 1996, although the differences were not statistically significant. This pattern may have reflected the decrease in grouse density and an associated decline in density-dependent mortality (Thirgood et al. 2000c). Despite the fact that there were large differences in harrier breeding density between the areas and that some birds were fed in each year, there was no significant difference in adult grouse survival between the two areas in either year, or between the 2 years in either area. Overall, the results indicated that providing harriers with food in spring did not greatly improve adult grouse survival. This could either be because feeding did not stop harriers killing grouse and they were killing grouse over the whole moor, or because harriers were killing few adult grouse anyway

10 1166 S.M. Redpath, S.J. Thirgood & F.M. Leckie at this time of year. We know little about the ranging behaviour of harriers in spring, but females appeared to spend most of their time on territory, being fed by their males with small prey for 2 3 weeks prior to laying (Redpath & Thirgood 1997). In addition, pellet analysis suggested that voles, not grouse, were the principal food of harriers during this period (Redpath, Thirgood & Clarke 2002). Given that most spring mortality was due to raptor predation, the most likely explanation for the apparent lack of an effect of spring feeding on adult grouse mortality rates is that much of the mortality at this time of year may have been due to other predators, particularly raptors such as peregrines Falco peregrinus Tunstall and goshawks Accipiter gentiles L. SUMMER FEEDING EXPERIMENT Providing harriers with supplementary food during the harrier nestling period had a marked effect on harrier provisioning. Females delivered much more supplementary food than males and decreased their delivery rate of natural prey. In addition, females used considerably more of the supplementary food during summer when they were provided with food during the spring. Males did not alter their overall delivery rates when supplementary food was provided, and it was only the bigamous males (in 1998) that delivered any of the supplementary food. Two other studies have examined the effects of supplementary feeding on parental provisioning rates in raptors (Wiebe & Bortolotti 1994; Wiehn & Korpimaki 1997). Both found that provisioning rates were reduced at nests with supplementary food. However, only Wiehn & Korpimaki (1997) distinguished between male and female deliveries, finding that only female kestrels Falco tinnunculus L. reduced their provisioning rates in response to feeding. They suggested that male parental effort was inflexible within a season and was fixed at a level that maximized lifetime reproductive success. This idea was supported by the finding that male hunting effort in kestrels was not influenced by brood size manipulations and males were therefore not sensitive to brood demands (Tolonen & Korpimaki 1996). The results of the present study support the idea that male parental effort is fixed. However, while provisioning rate might not be flexible, our results suggest that males are able to alter the types of prey they deliver to nests. Notably, fed males reduced the rate at which they delivered grouse chicks to their nests. Grouse chicks were seen delivered to only 29% of nests where food was provided, compared with 70% of nests where no food was provided. Overall, sevenfold more grouse chicks were delivered to nests where no supplementary food was available. So, providing harriers with food at Langholm clearly reduced the number of grouse chicks taken by harriers. Three other factors appeared to be important in determining the rate at which harriers took grouse chicks. First, the density of grouse chicks themselves, with fewer chicks being taken in later years as grouse density declined (Redpath & Thirgood 1999). Secondly, the breeding status of the birds, with females mated with bigamous males taking more grouse chicks than monogamous ones. Thirdly, the provision of supplementary food in the spring reduced the number of grouse chicks taken by female harriers. From a management point of view, these relationships are important. They suggest that in order to maximize the benefits of supplementary feeding: (i) birds should be provided with food during the pre-laying period; (ii) bigamous birds should have priority over monogamous ones; and (iii) the greatest reductions in the numbers of grouse chicks removed by harriers are likely at high grouse chick densities. In both years, delivery rates of grouse chicks were low, even at nests where no food was provided, and this reflected the scarcity of grouse on the moor. In July 2000 there were only 19 grouse km 2, compared with 73 grouse km 2 in Harriers were estimated to have taken 7% of available grouse chicks from hatching to fledging in 1998 and 6% during the same period in This compared with predation rates of up to 29% in previous years. Yet despite the reduced predation by harriers, losses of grouse chicks were higher than in previous years (67% in 1998, 56% in 1999, compared with 42% in 1995 and 48% in 1996). We infer that some other factor was having a strong influence on grouse chick survival in these 2 years, but whether this was related to weather, food quality, parasites or other predators was unknown because we did not focus our studies on grouse chick survival and we have no data on changes in predator or parasite abundance. Brood sizes at hatching were comparable between areas during the years of the experiment, suggesting that neither female condition nor clutch predation was responsible for the higher chick losses. The loss of two adult harriers and two further clutches to foxes suggested that there may have been more foxes on the moor in 1999, which may in turn have reduced grouse brood size. Also, the summer of 1998 was exceptionally wet and cold and there was a severe outbreak of heather beetle over Langholm Moor during 1999 and 2000, which may have contributed to grouse chick losses. A further possible explanation for the high grouse chick losses is that feeding increased the abundance of corvids and gulls on the moor, which subsequently preyed on grouse chicks. Crows, in particular, are thought to be important predators of grouse chicks (Hudson & Newborn 1995). At Langholm, grouse brood size at hatch and therefore clutch size was not reduced and we saw no removal of grouse chicks by these predators. However, increased predation by corvids and gulls cannot be excluded as a possible reason for poor grouse breeding success and should be tested in any future trial of supplementary feeding. Previously, data have suggested that most chick predation by harriers was additive and not compensated for by other sources of chick mortality in June and July

11 1167 Supplementary feeding of hen harriers (Redpath 1991; Thirgood et al. 2000c). In the 2 years of the experiment, however, unexplained chick mortality was considerably higher than it had been in other years, suggesting that there may be variation between years in the amount of potential compensation. The only way to test the extent of compensation is through an experiment, whereby harrier predation in one area is reduced and grouse chick survival in this area is compared with survival in a similar area with harrier predation. Other studies have found that food supplementation at raptor nests increased the number of fledglings and decreased nestling mortality ( Wiebe & Bortolotti 1994; Gehlbach & Roberts 1997; Wiehn & Korpimaki 1997). Although not statistically significant, there was a tendency for harrier breeding success to be improved by summer feeding, with fed broods rearing on average one more young than unfed broods. One possible reason why the results were not more marked was that overall provisioning rates were not significantly higher for fed broods, so food intake rate was similar for both groups of birds. As for the number of breeding attempts, it is likely that the impact of feeding on harrier fledging success will be greater in areas where natural food is scarce. Nestling survival may also be improved through the increased nest defence of the adult female as she is able to spend more time near the nest and less time hunting (Ward & Kennedy 1996). Unlike some studies of other raptors involving supplementary food, we found no evidence that feeding increased the time that young harriers spent on territory before dispersing (Kenward, Marcstrom & Karlbom 1993; Frumkin 1994). However, in contrast to these studies, which fed the young until they dispersed, we only put food out up to a previously determined dispersal date (Redpath & Thirgood 1997). We found no evidence that summer feeding increased the probability that young birds would return to breed at Langholm. We estimated that the costs of this feeding experiment were in the region of per annum, though this amount will vary according to a variety of factors (see the Results). It is unclear who would pay for supplementary feeding, but it is worth noting the costs of feeding relative to the costs of grouse management. Buccleuch Estates, who own Langholm Moor, have estimated that the annual cost of maintaining grouse moor management at Langholm was (Redpath & Thirgood 1997). This compares to approximately 900 per harrier nest, or an increase in the cost of approximately 1% per nest. In conclusion, providing harriers with food during the nestling period had a clear impact on harrier provisioning. Both male and female harriers reduced the rate at which they took grouse chicks when provided with supplementary food. Despite this, there was no clear improvement in grouse breeding success compared with previous years, and grouse density declined over the course of the experiment. Our experiment was conducted over a 2-year period on one grouse moor. Over such a relatively short time it is difficult to quantify the long-term impact of feeding on harrier survival and recruitment. Similarly, our experiment was not designed to test the effect of feeding on annual changes in grouse density. To address these issues properly, future experiments will require a longer time span of 5 10 years. Measuring the impact on grouse density would also require increased replication, where all birds were fed throughout the spring and summer, paired with control areas where no birds were fed. Ideally such areas would be geographically isolated to prevent control birds hunting in the experimental areas. Our results imply that the benefits of supplementary feeding will be greater when improvements in grouse chick survival are not compensated for by increased losses to other factors. Future experiments should specifically address the extent to which compensation can occur. Acknowledgements We are grateful to Buccleuch Estates for allowing us to conduct this work on their land, and in particular to Brian Mitchell and the late Gareth Lewis for all their help. Thanks to Ian Bainbridge, Colin Galbraith, Ian Newton and Des Thompson for constructive comments throughout the project, and to Nicholas Aebischer, Steve Albon, Arjun Amar, Dave Elston, Rhys Green, Pete Hudson, Mick Marquiss, Dick Potts and Steve Tapper for their guidance and helpful comments on the first draft. The work could not have been completed without the efforts of Lucy Bellini, Karen Bouwman, Steve Campbell, Eric Donnelly, Chris Gall, Kerry Lock, Mark Mainwaring, Andrew Walton and a number of enthusiastic volunteers. We also thank Brian Etheridge, Mike Henry, Malcolm Henderson and especially John and Bettina Halliday for their help. The work was led by the SNH Moorland Working Group and funded by Scottish Natural Heritage, Buccleuch Estates, The Game Conservancy Trust, The Game Conservancy Scottish Research Trust, The Centre for Ecology and Hydrology, The Royal Society for the Protection of Birds and The Scottish Landowners Federation. References Amar, A. & Redpath, S.M. (in press) Determining the cause of the hen harrier decline in the Orkney Islands: an experimental test of two hypotheses. Animal Conservation, in press. Anonymous (1995) Biodiversity: The UK Steering Group Report. HMSO, London, UK. Aparicio, J.M. (1984) The seasonal decline in clutch size: an experiment with supplementary food in the kestrel Falco tinnunculus. Oikos, 71, Dijkstra, C., Vuursteen, L., Daan, S. & Masman, D. (1980) Clutch size and laying date in the kestrel Falco tinnunculus: effect of supplementary food. Ibis, 124, Etheridge, B., Summers, R.W. & Green, R.E. (1997) The effects of illegal killing and destruction of nests by humans