Scottish Natural Heritage Commissioned Report No Analysis of capercaillie brood count data: Long term analysis

Transcription

1 Scottish Natural Heritage Commissioned Report No. 435 Analysis of capercaillie brood count data: Long term analysis

2 COMMISSIONED REPORT Commissioned Report No. 435 Analysis of capercaillie brood count data: Long term analysis For further information on this report please contact: Susan Haysom Scottish Natural Heritage Great Glen House INVERNESS IV3 8NW Telephone: This report should be quoted as: Baines, D., Aebischer, N., Brown M. & Macleod, A. (2011). Analysis of capercaillie brood count data: Long term analysis. Scottish Natural Heritage Commissioned Report No.435. This report, or any part of it, should not be reproduced without the permission of Scottish Natural Heritage. This permission will not be withheld unreasonably. The views expressed by the author(s) of this report should not be taken as the views and policies of Scottish Natural Heritage. This report is from a partnership project with partners: Forestry Commission Scotland, the Game and Wildlife Conservation Trust, the Royal Society for the Protection of Birds and Scottish Natural Heritage. Scottish Natural Heritage 2011.

3 COMMISSIONED REPORT Summary Analysis of capercaillie brood count data: Long term analysis Commissioned Report No. 435 (ibids n o 7822) Contractor: The Game & Wildlife Conservation Trust Year of publication: 2011 BACKGROUND Capercaillie Tetrao urogallus numbers have declined in Scotland since the mid 1970s, (Moss, 1994; Catt et al., 1998), but the last survey in suggested that, although capercaillie remain seriously threatened, population size may now have stabilised at about 2000 birds (Eaton et al., 2007). The rapid decline has been linked with poor breeding success (Moss et al., 2000) associated with both changes in weather patterns (Moss et al., 2001) and with increases in generalist predators that may predate eggs and chicks (Baines et al., 2004; Summers et al., 2004). MAIN FINDINGS Capercaillie breeding success and indices of hen density measured annually between 1991 and 2009 showed a significant decline over time. Declines in breeding success were associated with proportionally fewer hens rearing chicks as opposed to a reduction in actual brood size. Breeding success averaged 0.6 chicks per hen and did not differ between forest types. Birds bred less well in forests in Perthshire towards the southern edge of their current range, than they did in forests in Strathspey. Perthshire, together with Argyll and Moray, had the highest declines in indices of hen density and only in the Strathspey sub-population were densities considered stable. Breeding success ( chicks per hen and broods per hen ) was strongly influenced by weather and was higher in years when there was a larger increase in temperature in April (APRWARM), when temperatures at chick hatch time (HATCHTEMP) were higher and when April on the whole was cooler (APRTEMP). Two weather variables; APRTEMP and HATCHRAIN, increased over time. The mean pine marten Martes martes index had increased 3.7-fold since 1995, and the fox Vulpes vulpes index by 2.7-fold. Those of carrion crow Corvus corone and raptors, chiefly buzzards Buteo buteo, showed no change. When simultaneously considering the effects of both weather and predator explanatory variables, together with region and forest type on breeding success, we found that chicks per hen and broods per hen varied negatively with mean April temperature (APRTEMP) and negatively with both marten and crow indices of abundance. Broods per hen was higher in years when the temperature rose more in April (APRWARM) and when temperatures at chick hatching (HATCHTEMP) were higher, whilst brood size was negatively associated with rainfall at chick hatching (HATCHRAIN) and in forests with more crows. High indices of foxes were significantly related to the decline in hen density indices. Increases in mammalian predators and subsequent increased predation, together with changes in weather, and the continued presence of crows, provide i

4 an alternative hypothesis to that of climate change alone in explaining the reductions in capercaillie breeding success and the population decline in Scottish forests. In the absence of mechanisms to deal with the effects of climate change on capercaillie, it may be possible that the decline in capercaillie can be halted and even reversed by continued improvements in habitat management and by restoration of predator control in remaining capercaillie strongholds. For further information on this project contact: Susan Haysom, Scottish Natural Heritage, Great Glen House, Inverness, IV3 8NW Tel: For further information on the SNH Research & Technical Support Programme contact: DSU (Policy & Advice Directorate), Scottish Natural Heritage, Great Glen House, Inverness, IV3 8NW. Tel: or ii

5 Table of Contents Page BACKGROUND...1 METHODS... 3 RESULTS DISCUSSION REFERENCES TABLES FIGURES Table 1. Location and characteristics of the 26 study forests Table 2. Locations of the nearest weather stations to each forest Table 3. Mean values of capercaillie breeding success across three forest types Table 4. Mean values of capercaillie breeding success across six regions Table 5. Mean decline rates of capercaillie hens Table 6. Relationship between capercaillie breeding success, indices of hen density and weather variables Table 7. Trends in capercaillie breeding success and weather 1989 to 2009 Table 8. Forest specific indices of predator abundance in 1995 and 2009 Table 9. Mean predator indices in 1995 and 2009 Table 10. Correlation coefficients between predator indices in 1995 and 2009 and between weather variables Table 11. Relationship between capercaillie breeding success, weather and predators Table 12. Relationship between indices of capercaillie hen density, weather and predators Figure 1. Location of the 26 study forests Figure 2. Mean annual capercaillie breeding success across forests from 1991 to 2009 Figure 3. Change in mean annual capercaillie index of density between 1991 and 2009 Figure 4. Trends in hen and cock numbers in 16 forests surveyed annually iii

6 ACKNOWLEDGEMENTS We would like to thank the numerous land managers and their staff for access to their forests. Capercaillie brood counts were also conducted by Mark Andrew, Paul Baker, Lois Canham, Mick Canham, Norman Cobley, Kathy Fletcher, Isla Graham, Rupert Hawley, Andrew Hoodless, David Howarth, David Lambie, Fiona Leckie, Robert Moss, Raymond Parr, Adam Smith, Philip Warren and staff at RSPB Abernethy & Craigmore. John Woods assisted with the collection of predator data. The weather data were kindly provided to Melanie Brown by the UK Meteorological Office supplied through Natural Environmental Research Council Data Centres as part of her undergraduate dissertation thesis. The lek data were kindly provided by the Capercaillie Project Officers funded through the Capercaillie EU LIFE Project assisted by numerous gamekeepers, foresters and volunteers. Ron Summers (RSPB), Kenny Kortland (FCS), Susan Haysom and Megan Davies (SNH) provided helpful comments on the draft report. Stewart A Hara of Forest Research undertook the DNA analysis of mammalian scats. The study was funded by Scottish Natural Heritage, with Susan Haysom as the nominated officer, and by the Game & Wildlife Conservation Trust. iv

7 BACKGROUND Capercaillie Tetrao urogallus numbers have declined in Scotland since the mid 1970s, (Moss, 1994; Catt et al., 1998; Wilkinson et al., 2002), but the last survey in suggested that, although capercaillie remain seriously threatened, population size may have stabilised at about 2000 birds (Eaton et al., 2007). The rapid decline has been linked with poor breeding success (Moss et al., 2000) associated with changes in weather patterns (Moss et al., 2001), increases in generalist predators that may predate eggs and chicks (Baines et al., 2004; Summers et al., 2004a) and detrimental changes in silvicultural practices (Moss, 1994). Simultaneous to this, mortality of full-grown birds through flying into deer fences, has become a significant source of mortality (Catt et al., 1994; Baines & Summers, 1997; Baines & Andrew, 2003). That the population appears to have now stabilised is probably attributable to considerable funding being dedicated to removing or marking deer fences and thus reducing deaths following collisions. Previous analyses have identified several factors that influence between-year and betweenforest variations in capercaillie breeding success in Scotland. A comparison of 14 forests showed that birds bred more successfully where predators, particularly foxes Vulpes vulpes and crows Corvus corone, were fewer, and where bilberry Vaccinium myrtillus, a preferred habitat component whose leaves and berries are a key part of the diet (Storch, 1993; 1994; Summers et al., 2004b), was more plentiful (Baines et al., 2004). Recognition of the importance of bilberry to capercaillie has resulted in substantial collaborative habitat management work designed to increase the amount of bilberry cover, especially within plantation forests in Scotland. Associated improvements in predator control, disturbance management and removing forest fences through the LIFE Nature Project "Urgent Conservation Management for Scottish Capercaillie", the Species Action Framework and other initiatives have sought to deliver progress towards the UKBAP targets of increasing population size in Scotland to 5000 birds and increasing range from 40 to 45 occupied 10- km squares by Meeting these targets has not proved possible and meeting such optimistic targets in the future will almost certainly rely on improving breeding success. The principal cause of poor breeding success, causing large between-year variations, relates to weather at or around chick hatching time in June, both here in Scotland (Moss, 1985; Summers et al. 2004a) and elsewhere in the birds range (Slagsvold & Grasaas, 1979). This relationship with annual weather has been further complicated by climate change, with a delay in natural warming during April thought to affect the timing of plant growth in spring, essential to gravid hens, and the availability of invertebrates needed by growing chicks in June (Moss et al., 2001; Wegge & Rolstad, in press). Even when annual weather appears suitable for successful breeding, productivity is often only modest and varies markedly between forests in relation to habitat quality and indices of fox and crow abundance (Baines et al., 2004). In Scandinavia, pine martens Martes martes have been shown to reduce capercaillie breeding success through predating both clutches and chicks (Marcstrom et al., 1988; Kastdalen & Wegge, 1989; Kurki et al., 1997), but were not linked to low breeding success in Scotland (Baines et al., 2004). In Scotland, pine martens were historically persecuted by man to protect gamebirds, but legal protection within the last 25 years has allowed martens and other predators to recover much of their former range and abundance (Tapper, 1999).There is now evidence from one of the forests used in this study (Forest A) that martens have become more numerous since the 1995 survey (Summers et al., 2004a). Here, where crows and foxes are controlled, a recent study of capercaillie nest outcomes showed that of 20 nests, pine martens predated 33-57%, depending on the interpretation of the data (Summers et al., 2009). Although increasing in numbers and range the pine marten, also a UK BAP priority species, is still rare in Scotland with only an estimated 3,500 adults in Scotland, representing at least 95% of the British population (Birks, 2002). 1

8 Whilst the Summers and co-workers study (2009) indicates that martens can be significant predators of capercaillie clutches, it is not known to what extent this effect occurs elsewhere within the bird s range in Scotland. It is likely that similar increases in marten abundance have occurred in other forests since the mid 1990s and it has been suggested that this may have resulted in increased levels of predation on capercaillie. In an attempt to determine what factors influence annual breeding success of capercaillie, this report considers the results from annual surveys of capercaillie breeding success in Scottish forests between 1991 and 2009 (e.g. MacLeod, 2007; 2008) in relation to annual weather variables and findings from two surveys of predators in 1995 and 2009 in forests used by breeding capercaillie. 2

9 METHODS Capercaillie breeding success and density Measures of capercaillie breeding success and indices of adult densities were obtained in July and August from a total of 26 forests in Scotland (Table 1) (Figure 1). Data were obtained from six regions between 1989 and 2009: Strathspey (8 forests), Aberdeenshire (Deeside / Donside (6)), Perthshire (6), Moray (2), Easter Ross (3) and Argyll (1 forest). The forests surveyed were classified into three types following Baines et al., (2004); (1) Open canopy as in native pinewoods, (2) mature Scots pine Pinus sylvestris plantation, canopy sufficiently open for some dwarf shrubs, and (3) mixed species plantation with closed canopy, often with some clear-felled areas and restocked ground. The number of forests surveyed in each year varied from only two in 1989 and 1990 to 20 in 2009, but in most years count data were available from forests providing a total area searched over all forests each year that ranged from 31 to 78 km 2. Counts of hens and accompanying chicks were conducted using pointing dogs to locate birds. Counts began at one site (Forest L) in the first week of July, but at all other sites they began in mid-july and all sites were completed by the end of August. The mean area searched per forest was 3.9 km 2 but search areas varied between forests from 0.6 to 11.0 km 2. The parts of a forest favoured for searching were those more open areas where dogs could be readily used and were perceived to be good breeding habitat. Not all hens may breed in their first spring, but as there was no way of distinguishing between hens that did not breed and hens that bred but failed to rear chicks, all hens seen were included in the estimates of breeding success. Three measures of reproductive success were calculated each year in each forest: the proportion of hens with at least one chick was broods per hen, the number of chicks per hen that had at least one chick was brood size and the overall measure of breeding success was the number of chicks per hen, which was the total chicks divided by total hens for each forest. The number of hens encountered in each forest in relation to the area searched in each year was used as an index of hen density. Typically, the same part of a forest was searched in each year that the forest was surveyed. Where this was not the case, such as at Forest A, where the count method also changed over time, the data were removed from analyses relating to hen density. Annual estimates of breeding success at Forest A were however retained as the number of hens encountered remained a high proportion of those estimated to be present and hence representative of the forest as a whole and well above the threshold level of 10 hens which formed an initial constraint in the selection of study forests (Baines et al., 2004). When, such as at Forest G, the area searched differed between years, the change in effort was taken into account when calculating indices of density. A similar index was calculated for the number of cocks seen. Indices of hens and cock density were in turn compared with the total numbers of cocks observed at leks (Picozzi et al., 1992) in 16 forests surveyed between 2002 and 2009, the years when lek data were available. Weather data Weather data were obtained for the nearest weather station to each forest for the period 1989 to 2009 (Table 2) from the UK Meteorological Office, Three weather variables; temperature (the mean of daily maximum and daily minimum temperatures, o C), rainfall (mean daily rainfall in mm) and rain-days (the number of days with rain) were summarised into 10 day periods for April, May and June (1-10, and (or for May)). Weather data were restricted to these months as they coincide with the timing of hens coming into breeding condition in April, egg laying in late April and early May and chick hatch 3

10 in late-may and early June. Counts of hens and their chicks to determine breeding success began at the end of the first week of July, so July weather was not considered. The combination of weather variables and periods considered provided 27 potential explanatory weather variables. Spurious correlations with weather become increasingly likely as more meteorological data are included. A previous analysis of the effects of weather on capercaillie breeding success by Moss et al. (2001) found seven weather variables, each of 10-day periods, to be significantly related to capercaillie breeding success. These were three successive 10 day measures of mean daily April temperature, temperature in late May, temperature in early June, the number of rain-days in late-may and number of rain-days in early June. For our analyses, we further reduced these to four weather variables; mean April temperature (APRTEMP), mean temperature at or around the peak time of capercaillie chick hatching in the last eleven days of May and the first ten days of June (HATCHTEMP), the total number of rain-days in the same period (HATCHRAIN) and, following Moss et al. (2001), we constructed an index of April warming (APRWARM) calculated as half the difference between the temperature in the first half of April and that in the second half of April:- (T2 T1) / 2. T1 = mean temperature in first 10 days of April, T2 = last 10 days. Predator surveys In 1995, pine marten, fox, carrion crow and raptor indices were obtained from 14 forests where measures of capercaillie breeding success were also collected (Baines et al., 2004). In 2009, 11 of these 14 forests were re-surveyed. These forests had predator indices from 1995, together with data on capercaillie breeding success from the 1990s and more recently. These combinations of data allowed changes in measures of activity and distribution of pine martens and other predators to be related to changes in measures of abundance and breeding success of capercaillie. Of the 14 forests from 1995, three (Forests C, E and R) were not surveyed for predators in 2009 because brood counts in them were largely discontinued in the early 1990s and the latter forest has since been clear-felled. The 11 forests retained for the 2009 survey included three (Forests J, K and U) where capercaillie were considered to be locally extinct. These 11 forests were supplemented with five additional forests (Forests B, N, W, X and Y), where brood counts have been conducted in more recent years, bringing the total number of forests surveyed in 2009 to 16. These 16 forests showed a range of forest types and exhibited a geographical range that encompassed several Special Protection Areas which featured capercaillie as a qualifying interest and include both core and peripheral parts of the perceived pine marten distribution. For each survey, approximately 10 km of unsurfaced vehicle tracks within each forest were searched for mammal scats. Wherever possible the same 1995 routes were used in The tracks were walked five times, initially during a clear-up round in the second half of April to count and remove all scats, then twice in May (middle and end of the month) and twice in June (middle and end of the month). The observer simultaneously scanned both sides of vehicular tracks for scats at a slow walking pace. The location of all scats was recorded as a 10-figure grid reference using a GPS. All scats were initially classified as fox, marten or other in the field when characteristic elements of their morphology had not been altered by handling. Prior to handling, a digital photograph was taken of the scat in situ with a ruler placed adjacent for scale. All scats, other than those from dogs, were collected. In each survey year, two different observers collected scats in the field; observers differed between 1995 and In 2009, to try to reduce the observer error, a secondary check of identification was performed on the 4

11 collected scats by the more experienced of the two field observers (AM), whose correct classification during a previous survey had been verified by DNA analysis to be 88% (R. Trout, pers. comm.). In % (414) of the mammalian scats collected were DNA tested to verify their originator. Of the 305 scats that provided sufficient DNA, 77% had been correctly identified as either fox or marten prior to DNA testing. Of those misidentified, there was a significant tendency to over-identify scats as fox, thus under-identifying marten scats. Conversion rates of x 0.49 for fox and x 1.30 for marten were calculated 1, but this process had not been conducted in Subsequent analyses of changes in mammalian predator indices over time consider scenarios when conversion rates were either applied or not applied to both the 1995 and 2009 scat data and when rates were applied in 2009 only, but not in Approximately 5 km of the same tracks used for scat surveys were also used as transects along which carrion crows Corvus corone, C. cornix and raptors were counted. Bird counts were conducted just after dawn (i.e. within 2 hours of sunrise), twice monthly in May and June and immediately preceded the scat collection round on the same day. Sightings and calls were given a GPS position. DATA ANALYSIS All analyses were conducted using GenStat version 12 (GenStat, 2009). Analysis 1: Capercaillie breeding success and hen density Data: 26 forests surveyed between 1991 and 2009 Type of model fitted: Generalised linear mixed model (GLMM) adjusted for over-dispersion Dependent variable: 1.1) Chicks per hen, i.e. number of chicks in each forest in each year (Poisson error - natural log of number of hens as the offset); 1.2) Brood size, i.e. number of chicks in each forest in each year minus the hens with no chicks (Poisson error - natural log of number of broods as the offset); 1.3) Broods per hen, i.e. number of broods in each forest in each year divided by the number of hens (binomial error with a logit link); 1.4) Hen density, i.e. number of hens in each forest in each year (Poisson error - natural log of area searched as offset). Fixed effect explanatory variables (for each model): Forest regions (6); forest type (3); year (as continuous and factor in separate models) Random effect explanatory variables (for each model): Forest The differences in breeding success and indices of hen density between forests in six Scottish regions, three forest types and between years were considered. Year was included both as a factor and then as a continuous variable in separate models leading to a total of eight models fitted. These analyses were conducted on all 26 forests surveyed between 1991 and 2009, but not when only two forests were surveyed. Significance of model terms was tested (Type III sum of squares) using Wald tests. Analysis 2: Relating hen and cock densities to lek counts Data: 16 forests surveyed between 2002 and For details see SNH Commissioned report 415 5

12 Type of model fitted: Generalised linear model (GLM) with Poisson error, log link, natural log of the area as the offset, adjusted for over-dispersion Dependent variable: 2.1) Hen density; 2.2) Cock density. Explanatory variables (for each model): Forest (16); count of leking cocks For the years when lek data were available, the annual indices of hen and cock abundance from brood surveys in 16 forests were compared with the total number of cocks attending leks in spring in the same years. This was done using GLMs with Poisson error, with hens or cocks from the brood surveys as the dependent variable and forest and the number of leking cocks as explanatory variables (Poisson distribution, log link, adjusted for over-dispersion) with log e area as an offset. Significance of model terms was tested (Type III sum of squares) using likelihood-ratio tests. Analysis 3: Relating hen densities to cock densities Data: 16 forests surveyed between 2002 and 2009 Type of model fitted: GLM with Poisson error, log link, natural log of the area as the offset, adjusted for over-dispersion Dependent variable: Hen density Explanatory variables: Forest (16); cock density; forest*cock density Hens on brood counts were related to cocks on the same counts in a GLM with cocks as a continuous explanatory variable, forest as a factor and an interaction of cocks*forest. Analysis 4: Trends in the hen and cock densities Data: 16 forests surveyed between 2002 and 2009 Type of model fitted: GLM with Poisson error, log link, natural log of the area as the offset, adjusted for over-dispersion Dependent variable: 4.1) Hen density; 4.2) Cock density; 4.3) Number of cocks attending leks. Explanatory variables (for each model): Region (6); forest type (3); year (as continuous); year*region; year*forest type Trends in numbers of cocks and hens observed on brood counts and numbers of cocks attending leks over time ( ) were considered using GLMs with Poisson error, log link, adjusted for over-dispersion, with each bird count (cocks or hens on brood counts, or cocks on leks) in turn as the dependent variable, log e area as an offset, region and forest type as factors and year as a continuous variable and interactions of year*region and year*forest type. Analysis 5: Weather effects and its interaction with forest type and region Data: 25 forests surveyed between 1989 and 2009 Type of model fitted: GLMM adjusted for over-dispersion Dependent variable: 5.1) Chicks per hen (Poisson error - natural log of number of hens as the offset); 5.2) Brood size (Poisson error - natural log of number of broods as the offset); 5.3) Broods per hen (binomial error with a logit link); 5.4) Hen density (Poisson error - natural log of area searched as offset). 6

13 Fixed effect explanatory variables (for each model): Forest type (3); region (5); APRWARM; APRTEMP; HATCHRAIN; HATCHTEMP; region*aprwarm; region*aprtemp; region*hatchrain; region*hatchtemp; Random effect explanatory variables (for each model): Forest This stage of the analysis looked at the effects of weather on breeding success and how weather interacted with forest type and region. Weather data were available for each year, in each region other than Argyll (Forest V) and were matched to breeding data collected from a total of 25 forests (i.e. 26 forests in analysis 1 minus Forest V) between 1989 and To assess the effects of explanatory forest type, region, and weather on breeding success and indices of density, GLMM were used. Variations in chicks per hen between forests and years were considered by setting the number of chicks seen in each forest in each year as the dependent variable and forest type, region, year and weather measures as fixed effects and forest as a random effect in Poisson regressions (Poisson distribution, log link, adjusted for over-dispersion), with the natural logarithm of the number of hens set as an offset. Brood size was analysed in the same way, but excluding hens with no chicks and setting the natural logarithm of the number of broods as an offset. Broods per hen was modelled using logistic regression (binomial distribution, logit link), categorising hens as successful if they had one or more chicks and unsuccessful if they had no chicks. Differences in indices of hen density were modelled using hens seen in each forest in each year as the dependent variable in Poisson regressions with the natural logarithm of the area searched in each forest as the offset. Analysis proceeded by fitting models that included interactions between region and the weather variables, with subsequent models reduced by removal of non-significant weather variables. Main effects were not allowed to be removed before any interactions that they were included in. Region and forest type were retained in all models as we considered them to be structural components of the design. Analysis 6: Trends in breeding success and hen density with weather variables as explanatory variables Data: 25 forests surveyed between 1991 and 2009 Type of model fitted: Generalised linear mixed model (GLMM) adjusted for over-dispersion Dependent variable: 6.1) Chicks per hen (Poisson error - natural log of number of hens as the offset); 6.2) Brood size (Poisson error - natural log of number of broods as the offset); 6.3) Broods per hen (binomial error with a logit link); 6.4) Hen density (Poisson error - natural log of area searched as offset). Fixed effect explanatory variables (for each model): year (as continuous); weather variables from analysis 5 as appropriate; forest region (6); forest type (3) Random effect explanatory variables (for each model): Forest Trends in the three measures of breeding success and hen density between 1991 and 2009 were considered by entering year as a continuous variable in the GLMM s described for analysis 4. Analysis 7: Trends in weather variables Data: 25 forests surveyed between 1991 and 2009 Type of model fitted: GLM with normal error and no adjustment for over-dispersion Dependent variable: 7.1) APRWARM; 7

14 7.2) APRTEMP; 7.3) HATCHRAIN; 7.4) HATCHTEMP. Explanatory variables (for each model): Weather station (5); year (as continuous) Trends in weather variables over time were considered in GLMs (all normal error, identity link function and not adjusted for over-dispersion) with each variable in turn as the dependent variable, weather station as a factor and year as a continuous explanatory variable to explore the long term trends in the weather variables. Analysis 8: Predator indices Data: 14 forests surveyed in 1995 and 16 in 2009 Type of model fitted: Generalised linear model (GLM) with Poisson error, log link, natural log of transect length * time interval as offset, adjusted for over-dispersion Dependent variable: 8.1) Fox index; 8.2) Marten index; 8.3) Crow index; 8.4) Raptor index. Explanatory variables (for each model): Forest (19); year (as factor) (2) Subsequent analyses addressed whether indices of mammal and predatory bird activity have changed in the same suite of forests since 1995 using generalised linear models (GLM) with Poisson error, log link, adjusted for over-dispersion, with the number of mammal (pine marten or fox) scats or bird (carrion crow or raptor) sightings as the dependent variable, log e (transect length*time interval) as an offset and forest and year as factors. Analyses were conducted first with no misidentification correction rates for mammalian predator indices in either year, then with rates applied in both years and finally with correction rates applied in 2009 only, and not in Analysis 9: Correlations between predator indices and weather variables Relationships between individual predator indices (fox, marten, carrion crow and raptors) were compared by Pearson correlations on log e (index + 1) transformed predator data. Relationships between the four weather variables were also compared by Pearson correlations. Analysis 10: Considering the effects of weather and predator simultaneously on breeding success and hen density Data: 14 forests surveyed between 1991 and 1995 and 16 between 2005 and 2009 Type of model fitted: Generalised linear mixed model (GLMM) adjusted for over-dispersion Dependent variable: 10.1) Chicks per hen (Poisson error natural log of number of hens as the offset); 10.2) Brood size (Poisson error natural log of number of broods as the offset); 10.3) Broods per hen (binomial error with a logit link); 10.4) Hen density (Poisson error natural log of area searched as offset). Fixed effect explanatory variables (for each model): Forest type (3); forest region (5); APRWARM; APRTEMP; HATCHRAIN; HATCHTEMP; Predator indices; Random effect explanatory variables (for each model): Forest; period within forest; year within period within forest 8

15 This stage of the analysis simultaneously considered the effects of weather and predators on breeding success. Data on capercaillie breeding success were split into two periods; and , which related to the collection of the predator indices in 1995 and 2009 respectively. The selection of the range of years within these periods was a compromise between obtaining a sufficiently robust sample of capercaillie breeding success based on enough hens by combining years, keeping the groups of years small enough to be representative of the year when predator indices were collected, and maximising the period between surveys in order to maximise the independence between the two periods. This reduced the data to information from 19 forests. Only one index value per predator was available per period and forest so the analysis became multi-level and the GLMMs described for analysis 5 were modified accordingly by considering forest, period within forest and year within period within forest as the three levels of variation. By nesting, it is not assumed that the period effect is the same across all forests. The approach adopted allows the inherent variation in the forest*period interaction to manifest itself in the test. Full models involving all weather variables, together with all four predator variables, were produced. Minimal models were obtained from this by backwards stepwise elimination of non-significant predator and weather variables from the current model, stopping when all remaining variables were significant at P < 0.1. The minimal models were checked by forward stepwise addition of predator and weather variables, one at a time in relation to their significance within the current model until no further predator or weather variables were significant at P < 0.1. The forward and backward stepwise procedures produced identical models for each measure of capercaillie breeding success and hen density. 9

16 RESULTS Capercaillie breeding success & density From analysis 1: Chicks per hen averaged 0.6 between 1991 and 2009 and did not differ between forest types (Table 3). Highest chicks per hen was found in Strathspey (0.86) and lowest in Perthshire (0.37), but there was no overall difference between the six regions considered (Table 4). Chicks per hen varied seven-fold between years, from 0.2 to 1.4 (χ 2 18 = 76.04, P < 0.001) (Fig. 2a). There was a significant trend for chicks per hen to decline across the period of study at a rate of -3.4% (SE: 1.2) per annum (χ 2 1 = 9.34, P = 0.002) (line fitted on a logarithmic scale from the GLMM Poisson regression). This represented an approximate mean decline of one chick per hen every 30 years, or 0.6 chicks per hen during the period of study. The decline in chicks per hen was consistent among regions and forest types. From analysis 1: Brood size did not differ between regions (table 4) and averaged 2.0 chicks. There was a significant forest type*year interaction (χ 2 34 = 83.55, P < 0.001), with brood sizes in native pinewoods and pine plantations showing less annual variation than those in mixed plantations. There was however no difference in brood size between forest types over time (χ 2 2 = 0.59, P = 0.75), and brood size across all three forest types showed no linear trend through time (-1.1 % (SE: 0.07 per annum) (Fig 2b). From analysis 1: Broods per hen averaged 0.32 and did not differ between regions (table 4) or forest type (table 3). Broods per hen varied six-fold between years and ranged from 0.09 to 0.57 (χ 2 18 = 94.50, P < 0.001). There was a linear decline in broods per hen over time at a rate of -4.0% (SE: 1.5) per annum (χ 2 1 = 7.63, P = 0.006) (Fig. 2c). From analysis 1: Over the study period, hen density indices averaged 1.4 km -2 and showed no difference either between regions (table 4) or between forest types (table 3). Hen densities differed between years (χ 2 18 = 89.37, P<0.001) and showed a significant decline over time at a mean rate of -6.5 % (SE: 0.7) per annum (χ 2 1 = 63.79, P < (Fig. 3). The decline rate differed between regions (χ 2 5= , P < 0.001) and was highest to the south of the range in Argyll and Perthshire (Table 5), only in Strathspey did hen densities show no change. From analyses 2 and 3: In the 16 forests where both brood surveys and counts of males attending leks were conducted each year from 2002 to 2009 indices of hen density were significantly correlated with both the number of cocks seen on brood surveys in summer (F 1,95 = 6.70, P = from analysis 3) (Fig. 4) and with the number of cocks attending leks in the same spring (F 1,95 = 8.53, P = from analysis 2). However, the number of cocks on brood counts (a mean of four cocks per forest) was not related to the numbers of cocks seen attending leks (F 1,95 = 2.02, P = 0.16 from analysis 2). From analysis 4: For the sub-set of data , hens exhibited a significant overall decline rate (-8.8% + 2.6%: F 1,111 = 10.69, P = 0.001), which although was consistent between forest types, differed between regions (F 5,111 = 2.38, P = 0.04) with significant declines in all regions other than Strathspey. A similar rate of decline was found for cocks (- 6.8% + 3.1%, F 1,111 = 4.47, P = 0.037), which again, whilst being consistent between forest types, differed between regions (F 5,111 = 3.05, P = 0.01) with steepest declines in Perthshire and Morayshire. In contrast, counts of cocks attending leks showed no trend over time (- 2.3% + 2.3, F 1,111 = 0.92, P = 0.34) or between forest types and regions. 10

17 Effects of weather on capercaillie From analysis 5: Analyses of chicks per hen and broods per hen found no interaction between region, forest type and any of the weather variables. Both chicks per hen and broods per hen were significantly related to APRWARM, HATCHTEMP (both positive) and APRTEMP (negative) (Table 6). In other words, chicks per hen and broods per hen were higher in years when there was a larger temperature rise in April, when temperature in late May / early June was higher and when April temperature was cooler. Brood size was not significantly related to any of the four weather variables. The index of hen density was positively related to APRWARM, which means that more hens were found when there had been a larger temperature rise in April. The index of hen density was negatively related to HATCHRAIN, so fewer hens were found when hatch time was wetter. From analysis 6: All three measures of breeding success and hen density showed a significant decline between 1991 and 2009 (table 7). This decline was consistent between regions and forest types. From analysis 7: Although predictably there were significant temperature differences in April and at hatch time between weather stations, there were no region*year interactions for any of the four weather variables indicating that in all cases weather patterns were consistent across all regions considered. Chicks per hen and broods per hen were both negatively correlated with annual variations in APRTEMP (analysis 5). There was a significant trend for increasing April temperature over time (Table 7). Two of the variables (APRWARM and HATCHTEMP) were positively associated with annual variation in chicks per hen and broods per hen (analysis 5). APRWARM was also positively correlated with hen density (analysis 5). Neither APRWARM nor HATCHTEMP showed a trend with time. HATCHRAIN, which was not significantly correlated with any measure of breeding success but was with the hen density (analysis 5), showed a significant increase over time. Accordingly, of the weather variables, APRTEMP showed the most likely change over time which could readily be predicted to explain the decline in capercaillie breeding success observed. Effects of predators and weather on capercaillie From analysis 8: In the 1995 survey, the number of scats found on the initial clear-up round (x) was a good predictor of the cumulative number of scats found on the subsequent four survey rounds (y) (n = 14, pine marten: y = x, r 2 = 0.76, P < 0.001, fox: y = x, r 2 = 0.91, P < 0.001). In 2009, the relationship between scats on the clear-up round and on subsequent rounds was weaker; it was still significant for martens (n = 16, y = x, r 2 = 0.42, P = 0.004)), but not for fox (r 2 = 0.17, P = 0.06). From analysis 8: Predator indices in each forest are given in Table 8. Signs of martens were found in all but two of the forests (88%) in 2009, compared to only eight out of 14 (57%) in 1995, indicating a possible spread in range as well as a likely increase in abundance. Predator indices from the 1995 survey of 14 forests and the 2009 survey of 16 forests are compared in Table 9a. However when the 11 forests with predator surveys in both years are considered; martens were found in eight forests in 1995 (73%), compared to nine in 2009 (82%). Statistical tests are based on data from the sub-set of forests that were surveyed in both years (n = 11). There was a 3.7-fold increase in marten scats across forests (F 1,11 = 8.39, P = 0.016) and a 2.7-fold increase in fox scats (F 1,11 = 14.34, P = 0.004) (Table 9b). These relationships remained the same irrespective of whether corrected or uncorrected mammal indices were used. If corrected values were applied to the 2009 indices, but not to the 1995 indices, then the level of change for martens increased (F 1,10 = 11.29, P = 0.008), but that of fox decreased and became non-significant (F 1,10 = 1.18, P = 0.31). However, applying the correction factor to the 2009 indices but not to the 1995 indices makes the 11

18 assumption that no correction was necessary in 1995, i.e. an assumption that no misidentification occurred in Since using the scat correction factor or not made no difference to the results, and one was not available for 1995, the scat correction factor was not applied in any subsequent analyses. There were no changes in carrion crow or raptor (chiefly buzzard Buteo buteo) sightings between surveys (F 1,11 = 0.01, P = 0.91 and F 1,11 = 0.85, P = 0.38 respectively). However with a mean of only one raptor sighted per 10 km of transect, this index is relatively weak. From analysis 9: In the 1995 survey, we found no significant correlation between the predator indices (Table 10a). However in 2009, several of the indices were inter-correlated. The fox index was positively correlated with that of crows, whilst marten was negatively correlated with both crows and raptors (Table 10b). Raptors were positively correlated with crows. There was a weak, but significant positive correlation between APRTEMP and HATCHTEMP (Table 10c). From analysis 10: After having controlled for region and forest type, chicks per hen varied negatively with APRTEMP and with marten and crow indices (Table 11). That is, chicks per hen was higher in years when April was cooler and in forests with lower marten and crow indices. From analysis 10: Broods per hen had a significant negative relationship with APRTEMP, marten and crow and a significant positive relationship with APRWARM and HATCHTEMP (Table 11). That is, broods per hen was higher in years when April was cooler, in forests with lower marten and crow indices, in years when there was a larger temperature rise in April and when temperatures were higher at or around chick hatching. From analysis 10: Brood size had a significant negative relationship with HATCHRAIN and crow (Table 11). That is, brood size was higher in years of low rainfall and in forests with fewer crows. From analysis 10: Indices of hen density varied between years (χ 2 11 = 63.84, P < 0.001) and showed a decline over the period of the study (χ 2 1 = 46.25, P < 0.001) (Figure 3). Hen density had a significant negative relationship with the fox index (Table 11). That is, hen density was lower where there were more foxes. No weather variable significantly explained variations in hen densities when predators were also considered. 12

19 DISCUSSION Analysis of capercaillie brood count data collected annually between 1991 and 2009 showed a temporal decline in two measures of breeding success ( chicks per hen and broods per hen ) and in indices of hen densities. Breeding success did not differ between semi-natural and two types of plantation forests, but was more than twice as low at the edge of the current range in Perthshire as it was in the core region of Strathspey. High population declines, as measured by changes in hen densities, were also found in Perthshire, but also on other forests at the edge of the range in Argyll and Moray. Hen indices were relatively stable in Strathspey, where it is considered from lek data that approximately 60% of Scottish capercaillie now remains (K. Kortland pers. comm.). These findings are consistent with those of Moss et al. (2000), who state that the on-going decline in capercaillie in Scotland has been primarily due to lower breeding success. The average chicks per hen observed in this study is 0.6 combining all forests and all years, is the lowest recorded by all of 16 previous studies, which had a median of 1.6 chicks per hen (range ), summarised by Borchtchevski (1993). Despite the data on reproductive success and density indices being widely based from 26 Scottish forests, they may not necessarily be strictly representative of the Scottish capercaillie population. Forests selected for survey tended to be those where most capercaillie were known to occur, or if the areas surveyed formed part of larger forests, then parts favoured by breeding hens, usually more open parts of the forest, were surveyed. However counts of hens during brood surveys were significantly correlated with counts of males displaying at leks in the same forests in the same years, suggesting that trends from brood counts were broadly representative of general regional and national trends reported from winter transects counts in national surveys during , (Catt et al., 1998; Wilkinson et al., 2002) and including a more recent stabilisation of numbers in (Eaton et al., 2007). Continued declines at the edge of the range are, however, apparent not only from our brood surveys, but also from counts of males at leks (RSPB, unpublished data). This indicates that trends in population size can be predicted from brood surveys and that reduced breeding success is the likely demographic mechanism for the declines observed. Furthermore, as hen survival rates tend to be lower than those of cocks (Wegge et al., 1987; Moss et al., 2000), reductions in numbers of hens encountered on brood counts might be used as an early warning of impending declines in numbers of leking cocks. A variety of mechanisms to account for reduced breeding success were reviewed by Moss (1994 and thereafter). These included deterioration in forest habitat following changes in silvicultural practices, reductions in the quality of chick rearing habitats following overgrazing by red deer Cervus elaphus (Baines et al., 1994), increased predation (Baines et al., 2004; Summers et al., 2004a) and climate change (Moss et al., 2001). In 1995, habitat data including measures of forest structure and ground vegetation were collected from 14 forests included in this analysis. At that time, higher capercaillie breeding success was found in forests with more bilberry cover (Baines et al., 2004). No recent equivalent habitat data have been collected across the same suite of forests to enable a comparison of habitat quality through time. Both the study by Baines and co-workers (2004) and others have shown a close positive association between capercaillie and bilberry (e.g. Storch, 1993; 1994; Summers et al. 2004b) and have been instrumental in promoting management operations such as selective thinning and deer reduction to favour bilberry in forests used by capercaillie. Indeed, this management formed a key part of the recent EU LIFE Nature Project: Urgent Conservation Management for Scottish Capercaillie. Accordingly, there has been considerable expenditure aimed at improving conditions for breeding capercaillie in many of our study forests within the last 10 years. Consequently, we surmise that it is likely that forests are now managed better for capercaillie than they were in the 1990s and that overall forest habitat quality is likely to be improving rather than the converse (K. Kortland, pers. comm.). 13

20 It is well established that poor weather in the form of lower temperatures or higher rainfall at or just after hatch time in June is the most important factor determining annual variations in capercaillie breeding success (Slagsvold & Grasaas, 1979; Moss, 1985), and also that of the related black grouse Tetrao tetrix (Summers et al., 2004). Although this study found a trend for increased rainfall in late May or early June over time, we found that when the effect of the predator indices were controlled for, low temperature at hatch time was a better predictor of breeding success in capercaillie rather than rainfall. With the effect of predators controlled for, the temperature around hatching was a predictor for broods per hen, but not chicks per hen or brood size. The rainfall around hatching was a predictor of brood size, but not chicks per hen or broods per hen. Moss et al. (2001) found that hens reared more chicks when temperatures rose more in early April possibly stimulating plant growth and improving hen nutrition and thus the viability of their chicks. Furthermore, over the period , there had been a progressive cooling in mid-april temperatures relative to the rest of the month. He concluded that a climatic change involving protracted spring warming could have been a major cause of the capercaillie decline in Scotland. Repeating these types of analyses with data from more forests over the period 1989 to 2009, we confirmed not only a positive effect of temperature at hatch time, but also a positive effect of April warming. Both these weather variables acted significantly upon the proportion of hens that reared broods ( broods per hen ), but not on the number of chicks in those broods (brood size). Unlike Moss et al. (2001) we did not find a trend through time for April warming. Instead, we showed a significant trend for increased mean April temperature over time. Hatch in Tetraonids and tits Parus spp. appears timed to coincide with the peak availability of Lepidopteran larvae (Perrins, 1991; Baines et al., 1996), preferred tetraonid chick prey items (Kastdalen & Wegge, 1985; Picozzi et al., 1999). Larvae are able to respond to earlier onset of plant growth, but grouse and other birds are probably less flexible, although dates of onset of breeding have advanced (McCleery & Perrins, 1998). Consequently, increasing April temperature, as observed in this study, could bring about a mismatch in the timing of capercaillie chick hatch and the availability of their larval prey (Baines et al. 1996; Wegge & Rolstad, in press) thus creating more adverse conditions for chicks. Marten and fox indices, both recognised as predators of capercaillie or their eggs and chicks (Marcstrom et al., 1988; Kastdalen & Wegge, 1989; Wegge & Storaas, 1990; Kurki et al., 1997; Summers et al., 2009) increased 3.7-fold and 2.7-fold respectively between 1995 and Indices of both marten and crow were significant negative explanatory variables, together with weather, when accounting for variation in chicks per hen and broods per hen (the crow index also explained some of the variation in brood size), whilst fox indices were similarly negatively linked to changes in indices of hen density. This suggests a further or revised hypothesis that may account for the decline of capercaillie in Scotland: that of increased predation and climate change, additional to that of climate change alone. We suggest that impacts of predation may not be mutually exclusive to those of climate change alone, but instead they could work in parallel to them and exacerbate poor breeding success and subsequent decline rates. Marten and fox indices were solely derived from scat collections along forest tracks. Their use in deriving either estimates of abundance or population size can be limited and open to different interpretation (Webbon et al., 2004). Scat decay rates are likely to differ according to diet, as well as seasonal and habitat differences in deposition (Davison et al., 2002) and weather (Laing et al., 2003). Many of these potential biases were overcome by standardising survey timing and duration to one season, one substrate type within forests and one broad geographic area. Accordingly, we consider that by comparing 1995 and 2009 data, we compared changes in abundance as opposed to changes in activity, but see Birks et al., (2004). However, the precise nature of the relationship between predator numbers and scat density is unknown, is unlikely to be collinear and hence reported magnitudes of increase in 14

21 predator indices may not equate to the same levels in abundance. An increase in the marten index across our range of forests was perhaps predictable from a similar magnitude of increase already reported from one of the study forests (Forest A) (Summers et al., 2004; Summers et al., 2009). Here, where crows and foxes were controlled to benefit breeding capercaillie, a study of capercaillie nest outcomes showed that of 20 nests, pine martens predated 33-57%, depending on the interpretation of the data (Summers et al., 2009). Their study indicates that martens can be significant predators of capercaillie clutches, but did not consider predation of chicks. As such, these data are similar to those from northern England, where stoats Mustela erminea show a similar impact on the survival of black grouse clutches and, together with weather, limit breeding success and attainment of BAP targets (Baines et al., 2007). Our study confirms that increases in marten indices have occurred in other forests over the same time period and that martens, together with crows and weather, are both linked to lower breeding success across the 19 forests where predator indices and capercaillie breeding data were collected. This increase in marten index probably reflects increases in overall abundance and possibly re-colonisation of parts of their former range in the Scottish Highlands following their legal protection. Perhaps less predictable was the doubling of the fox index relative to 1995, particularly when many of the sample forests had participated in the Capercaillie LIFE Project and had received money to improve their levels of fox control. The increase in the fox index is no longer significant if correction indices are applied in 2009, but not in This assumes all scats were correctly identified in This is unlikely, but if so, would make any significant relationship of fox with capercaillie invalid. Despite both marten and fox indices increasing between surveys, their indices were not correlated. This does not readily fit with current understanding of intra-guild predator relationships, where typically there is a negative relationship between the abundance of larger predators such as the fox and that of mesopredators such as martens (Lindstrom et al., 1995; Smedshaug et al., 1999) and stoats (Warren & Baines, 2004). This is not always the case, however, and Kurki et al. (1998) found no such negative relationship whilst Summers et al. (2004) showed an increase in the marten index whilst fox cubs were being controlled, but adult fox numbers were maintained. The collective evidence base suggests that increasing fox and marten abundance in Scottish forests, together with the continued presence of crows and changes in weather, is significantly linked with reductions in capercaillie breeding success, subsequent population size and possibly breeding range. Numbers and distribution of several predator species in Scotland, including both marten and fox, but also crows and some raptors, have increased markedly in Scotland over the last 50 years (Hewson, 1984; Gibbons et al., 1994; Stone et al., 1997). These increases corresponded with a decrease in the number of gamekeepers employed in Scotland (Hudson 1992). Hitherto, predator numbers had been limited, legally or illegally, by gamekeepers, whose objective has been to conserve gamebirds, particularly red grouse Lagopus lagopus scoticus, but also capercaillie, for sport shooting (Redpath & Thirgood, 1997; Tapper, 1999; Whitfield et al., 2004). Recent declines in red grouse and driven grouse shooting (Barnes, 1987; McGilvray, 1995) are often associated with reductions in gamekeeper numbers and predator management. This in turn has probably improved the conservation status of the now protected pine marten, but has also resulted in increases in foxes and possibly crows. Reaching the UKBAP Capercaillie Species Action Plan population size and range targets may ultimately depend upon improving breeding success through continued reductions in predator abundance. It is of interest to note that Moss et al. (2001), in the absence of data on changes in predator indices, concluded that climate change was sufficient to account for the fall in breeding success at Forest L which, at the time, was effectively keepered. Whilst most probably correct at the time, our data have shown a three-fold increase in the fox index at Forest L, the arrival of pine martens and there has been a reduction in gamekeepers from three to one. This illustrates the rapidity of some of the on-going changes in predator communities 15

22 and their management in Scottish forests. Whilst climate control is outside of our immediate direct influence, it may be possible that the decline in capercaillie can be halted and even reversed by continued improvements in habitat management and by restoration of predator control in remaining capercaillie strongholds. The link between a rare predator and a rare prey species such as pine marten and capercaillie is likely to be a contentious issue. Better understanding of factors influencing marten numbers will be called for, including habitat requirements, prey availability and their relationship with fox numbers. Aspects of this will be considered in a forthcoming study by the RSPB at their Abernethy Forest Reserve (J. Wilson pers. comm.). However, it is also likely that there will be calls for culls of martens to protect wildlife and the evidence to support such an option needs to be robust and carefully considered. The data presented here are statistical associations from which we hypothesise that martens may be impacting upon capercaillie. Carefully designed predator removal experiments in select forests or other management possibilities involving aversion or diversionary feeding, may need to be instigated for compelling evidence of cause and effect. 16

23 REFERENCES Baines, D. & Summers, R.W Assessment of bird collisions with deer fences in Scottish forests. Journal of Applied Ecology, 34, Baines, D. & Andrew, M Marking of deer fences to reduce frequency of collisions by woodland grouse. Biological Conservation, 110, Baines, D., Sage, R.B. & Baines, M.M The implications of red deer grazing to ground vegetation and invertebrate communities of Scottish native pinewoods. Journal of Applied Ecology, 31, Baines, D., Wilson, I.A. & Beeley, G Timing of breeding in black grouse Tetrao tetrix and capercaillie Tetrao urogallus and distribution of insect food for the chicks. Ibis, 138, Baines, D., Moss, R. & Dugan, D Capercaillie breeding success in relation to forest habitat and predator abundance. Journal of Applied Ecology, 41, Baines, D., Warren, P. & Richardson, M. (2007). Variations in the vital rates of black grouse in the United Kingdom: consequences for recovery programmes. Wildlife Biology, 13, Barnes, R Long-term declines of red grouse in Scotland. Journal of Applied Ecology, 24, Birks, J The Pine Marten. The Mammal Society, London. 27 pp. Birks, J., Messenger, J., Braithwaite, T., Davison, A., Brookes, R. & Strachan, C Are scat surveys a reliable method for assessing distribution and population status of pine martens? Pages , in: Harrison, D.J., Fuller, A.K. & Proulx, G. (eds.), Martens and Fishers (Martes) in Human Altered Environments. Springer, London, UK. Borchtchevski, V.G Population biology of the capercaillie, principles of the structural organisation. Central Laboratory of the Management and Hunting of Nature Reserves, Moscow, Russia. Catt, D.C., Baines, D., Piccozzi, N., Moss, R. & Summers, R.W Abundance and distribution of capercaillie Tetrao urogallus in Scotland Biological Conservation, 85, Davison, A., Birks, J.D.S., Brookes, R.C., Braithwaite, T.C. & Messenger, J.E On the origin of faeces: morphological versus molecular methods for surveying rare carnivores from their scats. Journal of Zoology, 257, Eaton, M.A., Marshall, K.B. & Gregory, R.D Status of capercaillie Tetrao urogallus in Scotland during winter 2003/04. Bird Study, 54, GenStat GenStat 12 th Edition. Rothamsted. Gibbons, D., Gates, S., Green, R.E., Fuller, R.J. & Fuller, R.M Buzzards Buteo buteo and ravens Corvus corax in the uplands of Britain: limits to distribution and abundance. Ibis, 137, S75-S84. 17

24 Hewson, R Changes in the number of foxes (Vulpes vulpes) in Scotland. Journal of Zoology, London, 204, Hudson, P Grouse in Space and Time: The Population Ecology of a Managed Gamebird. Game Conservancy Ltd, Fordingbridge, Hampshire. Kastdalen, L. & Wegge, P Animal food in capercaillie and black grouse chicks in south-east Norway a preliminary report. Proceedings of the International Grouse Symposium, 3, Kastdalen, L. & Wegge, P Why and when do capercaillie chicks die preliminary results based on radio-instrumented broods in south-east Norway? Proceedings of the International Grouse Symposium, 4, Kurki, S., Helle, P., Linden, H. & Nikula, A Breeding success of black grouse and capercaillie in relation to mammalian predator densities on two spatial scales. Oikos, 79, Kurki, S., Nikula, A., Helle, P. & Linden, H Abundances of red fox and pine marten in relation to the composition of boreal forest landscapes. Journal of Animal Ecology, 67, Laing, S.E., Buckland, S.T., Burns, R.W., Lambie, D. & Amphlett, A Dung and nest surveys: estimating decay rates. Journal of Applied Ecology, 40, Lindstrom, E.R., Brainerd, S.M., Heldin, J.O. & Overskaug, K Pine marten red fox interactions: a case of intraguild predation? Annales Zoologici Fennici, 32, MacLeod, A., Moss, R. & Baines, D The productivity of breeding capercaillie Tetrao urogallus at sites across their Scottish range Unpublished report to Scottish Natural Heritage Project No. FO3AC301a. MacLeod, A., Canham, L. & Baines, D The productivity of breeding capercaillie Tetrao urogallus at sites across their Scottish range Unpublished report to Scottish Natural Heritage Project No. FO3AC301a. Marcstrom, V., Kenward, R.E. & Engren, E The impacts of predation on boreal tetraonids during vole cycles: an experimental study. Journal of Animal Ecology, 57, McCleery, R.H. & Perrins, C.M Temperature and egg laying trends. Nature, 391, McGilvray, J An economic study of grouse moors. Game Conservancy Ltd, Fordingbridge, Hampshire. Moss, R Rain, breeding success and distribution of capercaillie Tetrao urogallus and black grouse Tetrao tetrix in Scotland. Ibis, 128, Moss, R Decline of capercaillie (Tetrao urogallus) in Scotland. Gibier Faune Sauvage, 11, Moss, R., Picozzi, N., Summers, R. & Baines, D Capercaillie Tetrao urogallus in Scotland demography of a declining population. Ibis, 142,

25 Moss, R., Oswald, J. & Baines, D Climate change and breeding success: decline of the capercaillie in Scotland. Journal of Animal Ecology, 70, Perrins, C.M Tits and their caterpillar food supply. Ibis, 133, Picozzi, N., Catt, D.C. & Moss, R Evaluation of capercaillie habitat. Journal of Applied Ecology, 29, Picozzi, N., Moss, R. & Kortland, K Diet and survival of capercaillie chicks in Scotland. Wildlife Biology, 5, Redpath, S.M. & Thirgood, S.J Numerical and functional responses in generalist predators: hen harriers and peregrines, on Scottish grouse moors. Journal of Animal Ecology, 68, Slagsvold, T. & Grasaas, T Autumn population size of the capercaillie in relation to weather. Ornis Scandinavica, 10, Smedshaug,.A., Selaes, V., Lund, S.E. & Sonerud, G.A The effects of a natural reduction of fox Vulpes vulpes on small game hunting bags in Norway. Wildlife Biology, 5, Stone, B.H., Sears, J., Cranswick, P.A., Gregory, R.D., Gibbons, D.W., Rehfisch, M.M., Aebischer, N.J. & Reid, J.B Population estimates of birds in Britain and the United Kingdom. British Birds, 90, Storch, I Habitat selection of capercaillie in summer and autumn: is bilberry important? Oecologia, 95, Storch, I Habitat and survival of capercaillie Tetrao urogallus nests and broods in the Bavarian Alps. Biological Conservation, 70, Summers, R.W., Green, R.E., Proctor, R., Dugan, D., Lambie, D., Moncrieff, R., Moss, R. & Baines, D. 2004a. An experimental study of the effects of predation on the breeding productivity of capercaillie and black grouse. Journal of Applied Ecology, 41, Summers, R.W., Proctor, R., Thornton, M. & Avery, G. 2004b. Habitat selection and diet of capercaillie Tetrao urogallus in Abernethy Forest, Strathspey, Scotland. Bird Study, 51, Summers, R.W., Willi, J. & Selvidge, J Capercaillie Tetrao urogallus nest loss and attendance at Abernethy Forest, Scotland. Wildlife Biology, 15, Tapper, S A question of balance: Game animals and their role in the British countryside. The Game Conservancy Trust, Fordingbridge, Hampshire. UK Meteorological Office. MIDAS Land Surface Stations data (1853-current), [Internet]. British Atmospheric Data Centre, Available from Warren, P. & Baines, D Black grouse in northern England: stemming the decline. British Birds 97: Webbon, C.C., Baker, P.J. & Harris, S Faecal density counts for monitoring changes in red fox numbers in rural Britain. Journal of Applied Ecology, 41,

26 Wegge, P. & Rolstad, J. in press. Timing of breeding and subsequent hatch in boreal forest grouse. Wegge, P., Larsen, B.B., Gjerde, I., Kastdalen, L., Rolstad, J. & Storaas, T Natural mortality and predation of adult capercaillie in south-east Norway. Proceedings of the International Grouse Symposium 4: Whitfield, D.P., Fielding, A.H., McLeod, D.R.A. & Haworth, P.F The effects of persecution on age of breeding and territory occupation in golden eagles in Scotland. Biological Conservation, 118, Wilkinson, N.I., Langston, R.H.W., Gregory, R.D., Gibbons, D.W. & Marquiss, M Capercailie Tetrao urogallus abundance and habitat use in Scotland, in winter Bird Study, 49,

27 TABLES Table 1. Location and characteristics of the 26 forest areas surveyed for capercaillie breeding success over the period , and for predator indices in either 1995 and / or Forest type: 1, open canopy as in native pinewoods; 2, mature Scots pine plantation, canopy sufficiently open for some dwarf shrubs; 3, mixed species plantation with closed canopy, often with some clear-felled areas and restocked ground. Capercaillie Forest Region Forest type Predator survey count years A Strathspey , B Perthshire C Strathspey , D Perthshire E Aberdeenshire , F Easter Ross G Strathspey , H Moray , , I Perthshire , J Perthshire , , 2009 K Perthshire , , 2009 L Aberdeenshire , M Aberdeenshire N Strathspey O Strathspey , P Aberdeenshire Q Strathspey , , R Perthshire S Strathspey , , T Moray U Aberdeenshire , , 2009 V Argyll W Easter Ross X Easter Ross Y Aberdeenshire Z Strathspey ,

28 Table 2. Locations of nearest weather stations to each of the forests surveyed for capercaillie Region Weather Station Forests Years covered Aberdeenshire Balmoral E Braemar E 2001 P Aboyne L M U Y Moray Nairn All in region Perthshire Ardtalnaig B D I Strathallan J K R Easter Ross Nairn All in region Strathspey Aviemore All in region

29 Table 3. Mean values of capercaillie breeding success and indices of hen density (+ SE ) across three forest types between 1991 and 2009 predicted from GLMM output adjusted for region, forest type and year. Wald statistic (χ 2 ) tests the null hypothesis of no difference between forest types. Forests Chicks per hen Broods size Broods per hen Hens km -2 Native pinewoods Scots pine plantations Mixed plantations χ 2 2 =2.48 χ 2 2 = 0.74 χ 2 2 = 3.09 χ 2 2 = 2.18 P = 0.33 P = 0.50 P = 0.25 P =

30 Table 4. Mean values of capercaillie breeding success and indices of hen density (+ SE) across six Scottish regions between 1991 and 2009 predicted from GLMM output adjusted for region, forest type and year. Wald statistic (χ 2 ) tests the null hypothesis of no difference between regions. Forests Chicks per hen Broods size Broods per hen Hens km -2 Strathspey Aberdeenshire Perthshire Moray Easter Ross Argyll χ 2 5 = 7.17 χ 2 5 = 2.01 χ 2 5 = 3.91 χ 2 5 = 3.03 P = 0.27 P = 0.10 P = 0.58 P =

31 Table 5. Mean percentage decline rates of capercaillie (hens km -2 annum -1 ) (+ SE) for each of six Scottish regions between 1991 and Wald statistic (χ 2 ) tests the null hypothesis of no difference between regions. Region Percentage annual decline + SE Strathspey Aberdeenshire Perthshire Moray Easter Ross Argyll χ 2 5 = P <

32 Table 6. Relationship between capercaillie breeding success, an index of hen density (hens km -2 ) and weather variables Parameter estimates are mean slopes (SE) from GLMMs adjusted for region and forest type. Dependent Parameter variable Fixed term estimate χ 2 1 P Chicks/hen APRTEMP (0.061) APRWARM (0.030) HATCHTEMP (0.056) HATCHRAIN (0.043) Brood size APRTEMP (0.029) APRWARM (0.014) HATCHTEMP (0.023) HATCHRAIN (0.018) Broods/hen APRTEMP (0.073) APRWARM (0.036) HATCHTEMP (0.071) HATCHRAIN (0.053) Hen density APRTEMP (0.042) APRWARM (0.021) HATCHTEMP (0.044) HATCHRAIN (0.028)

33 Table 7. Trends in capercaillie breeding success and indices of hen density (mean annual % change) and weather measurements ( ). Parameter estimates for capercaillie breeding success and weather variables are mean slopes (+ SE) from GLMMs involving region and forest type, and GLMs including weather station and year respectively. Slope + SE χ 2 1 P Capercaillie Chicks per hen -3.4 % Broods per hen -4.0 % Mean brood size -1.1 % Density (hens km -2 ) -6.5 % <0.001 Weather Slope + SE F 1,101 P APRTEMP <0.001 APRWARM HATCHTEMP HATCHRAIN

34 Table 8. Indices of mammalian and avian predators from forest transects walked between April and June in 1995 (from Baines et al. 2004) and 2009 in the forest areas where capercaillie brood counts were conducted. Values for fox and pine marten are mean scats 10 km -1 day and exclude the clearance round. Those for crows and raptors are mean observations 10 km -1 visit -1. Fox Marten Crow Raptor Forest A B C E G H I J K L N Q R S U W X Y Z

35 Table 9a. Predator indices, (means + 1SE) from 14 forests used by breeding capercaillie in 1995 and 16 forests in Mammal indices are scats 10 km -1 day -1 x 100 and exclude scats from the clear-up round. Bird indices are sightings 10 km -1 visit (n = 14 forests) 2009 (n = 16 forests) Predator Forests with sign Abundance Forests with sign Abundance Index Index Fox 13 (93%) (100%) Pine marten 8 (57%) (88%) Carrion crow 13 (93%) (69%) Raptors 12 (86%) (56%) Table 9b. Predator indices, (means + 1 SE) from 11 forests used by breeding capercaillie surveyed in both 1995 and Mammal indices are scats 10 km -1 day -1 x 100 and exclude scats from the clear-up round. Bird indices are sightings 10 km -1 visit Predator Forests with sign Abundance Forests with sign Abundance Index Index Fox 11 (100%) (100%) Pine marten 8 (73%) (82%) Carrion crow 10 (91%) (73%) Raptors 9 (82%) (45%)

36 Table 10. Correlations between a) predator indices (log e (index + 1) in 1995 (14 forests), b) predator indices in 2009 (16 forests) and c) weather variables (n = 21). Values are Pearson correlation coefficients. Significant relationships are given in bold, * P < 0.05, ** P < a) 1995 Fox - Marten Fox Marten Crow Raptor Crow Raptor b) 2009 Fox Marten Crow Raptor Fox - Marten Crow 0.50 * * - Raptor * 0.52 * - c) Weather variables HATCHRAIN HATCHTEMP APRTEMP APRWARM HATCHRAIN - HATCHTEMP APRTEMP * - APRWARM

37 Table 11. Relationship between capercaillie breeding success and indices of hen density with weather variables and predator indices for the years and Parameter estimates are mean slopes (SE) from GLMMs including region and forest type. Dependent Parameter variable Fixed term estimate χ 2 1 Prob_ Chicks/hen APRTEMP (0.083) APRWARM (0.069) HATCHTEMP (0.076) Marten (0.002) Crow (0.057) Brood size HATCHRAIN (0.021) Crow (0.022) Raptor (0.027) Broods/hen APRTEMP (0.101) APRWARM (0.085) HATCHTEMP (0.100) Marten (0.002) Crow (0.079) Hens km -2 Crow (0.108) Fox (0.009)

38 FIGURES Figure 1. Locations of the 26 forests in which capercaillie were surveyed within the years