General enquiries on this form should be made to:Defra, Science Directorate, Management Support and Finance Team,Telephone No. 020 7238 1612E-mail: research.competitions@defra.gsi.gov.uk

SID 5 Research Project Final Report

SID 5 (2/05) Page 1 of 36

NoteIn line with the Freedom of Information Act 2000, Defra aims to place the results of its completed research projects in the public domain wherever possible. The SID 5 (Research Project Final Report) is designed to capture the information on the results and outputs of Defra-funded research in a format that is easily publishable through the Defra website. A SID 5 must be completed for all projects.

A SID 5A form must be completed where a project is paid on a monthly basis or against quarterly invoices. No SID 5A is required where payments are made at milestone points. When a SID 5A is required, no SID 5 form will be accepted without the accompanying SID 5A.

This form is in Word format and the boxes may be expanded or reduced, as appropriate.

ACCESS TO INFORMATIONThe information collected on this form will be stored electronically and may be sent to any part of Defra, or to individual researchers or organisations outside Defra for the purposes of reviewing the project. Defra may also disclose the information to any outside organisation acting as an agent authorised by Defra to process final research reports on its behalf. Defra intends to publish this form on its website, unless there are strong reasons not to, which fully comply with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.Defra may be required to release information, including personal data and commercial information, on request under the Environmental Information Regulations or the Freedom of Information Act 2000. However, Defra will not permit any unwarranted breach of confidentiality or act in contravention of its obligations under the Data Protection Act 1998. Defra or its appointed agents may use the name, address or other details on your form to contact you in connection with occasional customer research aimed at improving the processes through which Defra works with its contractors.

Project identification

1. Defra Project code ZF0531

2. Project title

The Ecological Consequences of Removing Badgers from the Ecosystem

3. Contractororganisation(s)

Central Science LaboratorySand HuttonYorkYO41 1LZ                         

54. Total Defra project costs £ 1,846,627

5. Project: start date................ 01 February 1999

end date................. 31 March 2007

SID 5 (2/05) Page 2 of 36

6. It is Defra’s intention to publish this form. Please confirm your agreement to do so...................................................................................YES NO (a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They

should be written in a clear and concise manner and represent a full account of the research project which someone not closely associated with the project can follow.Defra recognises that in a small minority of cases there may be information, such as intellectual property or commercially confidential data, used in or generated by the research project, which should not be disclosed. In these cases, such information should be detailed in a separate annex (not to be published) so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report without including references to any sensitive or confidential data, the information should be included and section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No" answer.In all cases, reasons for withholding information must be fully in line with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.

(b) If you have answered NO, please explain why the Final report should not be released into public domain

Executive Summary7. The executive summary must not exceed 2 sides in total of A4 and should be understandable to the

intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together with any other significant events and options for new work.One of the main recommendations of the Krebs review (1997) was the implementation of a badger Meles meles culling experiment, designed to test the efficacy of badger culling as a means of controlling the incidence of bovine tuberculosis (TB) herd infections. This experiment was implemented as the Randomised Badger Culling Trial (RBCT). The aim of the RBCT was to measure and compare the incidence of cattle infections in a series of 100km2 study areas, each with one of three experimental ‘treatments’: no badger culling; reactive badger culling around infected farms; proactive removal of a substantial proportion of resident badgers. These treatments were arranged in ten experimental triplets, situated in TB hotspots in southern England, with the experiment planned to run for several years. It is known that the removal of native predators affects the dynamics of the food webs of which they are a part. It seemed highly likely that the sustained population reduction of an abundant carnivore such as the badger over large areas of land could have the potential to affect the dynamics of the ecosystem. It was deemed important to monitor this. Additionally, the RBCT provided a rare scientific opportunity to study ecosystem response to a predator removal operation, using a replicated and randomised experimental design on a large scale. Defra contracted the Central Science Laboratory (CSL) to carry out an assessment of the ecological consequences of badger removal during the RBCT. The aim of the contract was to monitor the populations of species that may be affected by badger removal, and where possible determine the underlying ecological processes driving any observed population responses.

Main objectives, as set in original CSG7

1. To identify suitable sites, species and monitoring techniques to allow an assessment of the ecological consequences of removing badgers from an ecosystem.

2. To conduct monitoring of predator and prey abundance, predator diet and prey fecundity over a five year period in treatment (badger removal) and experimental control (no badger removal) areas, concentrating on changes in abundance of other species of agricultural or conservation concern.

3. To provide a replicated experimental assessment of the ecological consequences of badger removal.

4. To construct a model of food web dynamics based on data from both treatment and control areas.

The methods fell into two main categories: population monitoring and complimentary research. Work was carried out in 4 to 5 of the RBCT triplets (depending on the species), in the proactive cull areas (experimental treatment) and no-cull areas (experimental control). Where possible, population monitoring and complimentary research was initiated prior to the onset of culling, and continued for several years

SID 5 (2/05) Page 3 of 36

while badger numbers were maintained at a low level.

Population monitoringMonitoring was carried out to identify any changes in population size of the selected species that occurred in response to badger culling. Monitoring of fox, hare and rabbit populations was carried out using nocturnal spotlight surveys from point counts distributed throughout the study areas. A distance sampling approach was used. Standardised spotlight surveys for hedgehogs were carried out, on a large sample of rural pasture fields and on amenity grassland areas in and around villages. Maximum counts of singing males were used to monitor populations of skylarks and meadow pipits.

Complimentary researchA wide-ranging program of complimentary research was implemented to investigate the mechanisms underlying any observed population responses. These included studies of fox diet using scat analysis, hedgehog ranging behaviour, source and rate of ground nesting bird nest failure, impacts of livestock on ground nesting bird abundance and nest success, use of badger setts by breeding foxes, and food web modelling.

Main Results

1. Fox populations increased in proactive cull areas relative to no-cull controls. This population increase took place mainly over the first three years from the onset of badger culling, and then stabilised at a higher level while badger densities continued to be depressed through culling. Scat analysis revealed no systematic change in fox dietary habits in relation to badger culling, therefore we concluded that the population increase was not a response to increasing food resources brought about by badger culling. The likeliest explanation for this effect is a reduction in competition for space, in particular breeding den availability, caused by badger removal.

2. Hedgehogs were only very rarely encountered in rural sites, but were found relatively frequently in amenity areas. Population density increased by over 100% over the course of the RBCT in amenity sites in proactive cull areas, while declining slightly in no-cull control sites. No similar increase was detected in rural sites. This provided strong evidence that badger predation restricts hedgehog populations, and that amenity areas near villages act as spatial ‘refugia’ for hedgehogs, where predation risk from badgers is lower than in rural areas. These results, combined with the radiotracking findings suggest that hedgehogs respond to badgers through large-scale movements away from areas of high predation risk, but do not demonstrate strong predator avoidance behaviour when moving within their home range.

3. The abundance of skylarks and meadow pipits remained relatively constant in proactive badger culling areas, but fell in areas of no badger removal. However, this pattern was only statistically significant for meadow pipits. There was also a tendency for higher survival rates of artificial nests in badger removal areas. One possible explanation for this is that the removal of badgers, and hence a degree of predation enabled populations of skylarks and meadow pipits to remain constant while other factors served to suppress populations in non-removal areas. However, by chance the proactive cull sites had larger areas of prime habitat for these species. Hence, an equally valid interpretation is that regional reduction in meadow pipit and skylark abundance due to unmeasured environmental factors could potentially have caused a contraction in range towards these core areas, resulting in the observed patterns. Therefore these results should be treated with caution.

4. There was evidence of a reduction in hare populations in proactive badger culling areas relative to no-cull control areas. This trend continued for the duration of the RBCT. However, there was no statistically significant effect of badger removal. The highly variable nature of hare population densities in our study areas, both in space and through time meant that only large changes in relative density could have been detected. However, we tentatively suggest that there may have been a ‘knock-on’ effect of badger removal on hare numbers. This would be consistent with the observed increase in fox densities in proactive areas relative to experimental control areas.

Main conclusions

Badger culling as carried out in the RBCT appears to have consequences for other species. This is consistent with a range of studies from around the world where predators and carnivores have been shown to have an important role in ecosystem functioning. This should be taken into account in any consideration of culling as an option for the management of TB in cattle. Foxes, hares and hedgehogs are important species for different reasons. Foxes are important predators in the ecosystem, and are of management relevance in the context of disease (rabies) control and livestock predation. Increasing populations in response to badger culling is likely to be of interest to a range of parties, and should therefore be a material consideration. Hedgehog populations are limited by badger predation, as indicated

SID 5 (2/05) Page 4 of 36

by previous research. However despite high badger densities, they continue to exist in reasonable numbers in refugia provided by amenity areas. The general lack of hedgehogs in rural areas in all our study areas, both culled and unculled, and the extent to which these refuges are isolated from each other is notable given that hedgehogs have recently been added to the UK Biodiversity Action Plan (BAP) Priority Species list. Hares have become a species of conservation concern, with population declines over several decades associated with habitat loss and changing farmland management practices. Hares are also a UK BAP Priority Species. The species plan aims to substantially increase hare populations over the next few years through positive habitat management encouraged through the Countryside Stewardship Scheme. There is some evidence from this study that indicates that widespread badger culling could have negative effects on hare populations.

Options for future research Badger removal has a range of ecological consequences. From an applied perspective, the data generated here could be utilised to model the ecosystem effects of variations in badger populations, whether natural or through management. From a scientific point of view, continuation of the monitoring programs while badgers recolonise the proactive cull areas would represent an experimental reversal. The population response to this recolonisation would give us a greater understanding about the nature of the relationship between badgers and these species.

These surveys cover large areas, and span several years and therefore provide valuable data on regional trends in certain important species. Therefore by interrogating the data we could investigate the factors that determine presence, distribution, and population trend of several key species.

Project Report to Defra8. As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with

details of the outputs of the research project for internal purposes; to meet the terms of the contract; and to allow Defra to publish details of the outputs to meet Environmental Information Regulation or Freedom of Information obligations. This short report to Defra does not preclude contractors from also seeking to publish a full, formal scientific report/paper in an appropriate scientific or other journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms. The report to Defra should include: the scientific objectives as set out in the contract; the extent to which the objectives set out in the contract have been met; details of methods used and the results obtained, including statistical analysis (if appropriate); a discussion of the results and their reliability; the main implications of the findings; possible future work; and any action resulting from the research (e.g. IP, Knowledge Transfer).

Main report – outline

This was a very large, complex project spanning several years. In order to monitor the range of species of interest, and study the necessary underlying ecological processes, we had to employ many methodologies. Therefore even though they are summarised briefly here, as are the associated results, the report is extensive with a large number of sections. For the sake of clarity the rationale, methods, results and discussion are presented together for each separate section. The population monitoring is covered in Section 2, followed by the complimentary research in Section 3. Owing to the large number of Tables and Figures, they are presented separately in Appendix 1. References cited in the text are presented in Appendix 2. Three PhD projects were completed based on different aspects of this project. The reader is referred to these documents at the appropriate points, should they wish to investigate the topic in more detail. The project began in 1999, and fieldwork was completed on the population monitoring of foxes, hares, rabbits, and hedgehogs in 2006. Other aspects of the work were carried out over more limited timescales within this range, according to the aims or logistic considerations. The time over which work was carried are specified at the appropriate points in the text.

Scientific objectives

1. To identify suitable sites, species and monitoring techniques to allow an assessment of the ecological consequences of removing badgers from an ecosystem.

SID 5 (2/05) Page 5 of 36

2. To conduct monitoring of predator and prey abundance, predator diet and prey fecundity over a five-year period in treatment (badger removal) and experimental control (no badger removal) areas, concentrating on changes in abundance of other species of agricultural or conservation concern.3. To provide a replicated experimental assessment of the ecological consequences of badger removal.4. To construct a model of food web dynamics based on data from both treatment and control areas

Extent to which objectives were met

1. Surveys were carried out 4 to 5 (depending on the study) RBCT triplets. A suite of monitoring methods were developed, generating an extremely valuable dataset which allowed us to assess the impacts of badger removal.2. The populations of several important species, thought to be related to the badger in the food web, were monitored in proactive badger cull areas, and no-cull control areas. Monitoring began prior to the onset of culling, and for several years after culling began.3. Working in the RBCT framework, using the combination of population monitoring programs and complimentary research, allowed us to provide a robust experimental assessment of the ecological consequences of badger removal.4. A food web model predicting the population responses of a range of species was developed. The mass-balance model was based on trophic interactions, and predicted the response of species’ populations given different efficacy of badger culling over different time scales.

FINAL REPORT

1. INTRODUCTION

1.1 Context

The project was carried out in the experimental framework provided by the RBCT. The need to monitor the effects of large scale removal are manifold. It is known that the removal of native predators affects the dynamics of the food webs of which they are a part. It seemed highly likely that the sustained population reduction of an abundant carnivore such as the badger over large areas of land could have the potential to effect the dynamics of the ecosystem of which it is a part. Possible effects might include: a reduction in competition for space or food resources with other species such as the fox Vulpes vulpes, reduction in predation pressure on prey species such as invertebrates, a reduction in both competition for food and predation pressure for the hedgehog Erinaceus europaeus. It is unlikely however, that these factors would operate in isolation from one another. For example, an increase in fox numbers due to reduced competition may result in increased predation pressure on the preferred prey of foxes, e.g. small mammals or lagomorphs, which could in turn reduce the number of prey available to small carnivores such as stoats Mustela ermina and weasels Mustela nivalis. This is just one possible scenario, and a large list of ecosystem responses was possible. The RBCT also provided an extremely rare scientific opportunity to study the ecosystem responses to a predator removal operation, using a replicated and randomised experimental design on a large scale.

1.2 Species selection

At the start of the project, decisions were required on the species to monitor. A short list of candidates were drawn up based on their; economic, conservation or management relevance; likely relationship with the badger; our ability to monitor them. The target species are listed below, and the rationale for their selection. 1.2.1 Foxes

Badgers and foxes are sympatric carnivores of similar size, with overlapping dietary preferences & similar denning behavior. Therefore the potential for competition certainly exists. However, despite eating a comparable range of foods, badgers and foxes specialise on different food resources, and evidence of aggressive encounters between the two species over food resources are rarely seen. However, when aggression between them is seen, the badger, being stronger and heavier, is the dominant competitor (Macdonald et al. 2004). This apparent lack of competition for food resources, suggests that badgers and foxes are more likely to compete along the niche axes of time and space. Both badgers and foxes are mainly crepuscular and nocturnal in their activity patterns, with similar habitat requirements. Therefore, there is considerable potential for competition for temporal and spatial resources. This is particularly true if the competitive interaction takes the form of interference competition, which can occur even when the level of the resource is not limiting. There is considerable evidence of overlap between badgers and foxes in the utilisation of these resources, with reports of foxes using unoccupied and even active badger setts as breeding dens, and fox cubs being predated by badgers (Macdonald et al. 2004;

SID 5 (2/05) Page 6 of 36

Neal & Cheeseman 1996). Additionally, given the potential importance of the fox as a potential disease vector and predator, it was important to monitor.

1.2.2. Hedgehogs

In the UK, the badger is an important predator of the hedgehog Erinaceus europaeus (Doncaster 1992), and in addition, the two species are potential competitors for the same invertebrate prey, particularly earthworms Lumbricus terrestris and beetles (Scarabeidae and Carabidae) (Neal & Cheeseman 1996; Reeve, 1994). Such a system is called Intraguild Predation (IGP). Previous experimental work has shown that badgers may have a regulatory effect on hedgehog populations (Doncaster 1992, 1994). The fact that badgers and hedgehogs have common prey types means that in addition to being in a predator-prey relationship, they are potentially in competition for resources. Predation and competition frequently interact to produce indirect effects that can have a strong influence on animal behaviour and population dynamics (Sih et al. 1985). Asymmetrical intraguild predation (IGP) is such an interaction, where predator and prey are also potential competitors for a shared food resource (Polis, Myers & Holt 1989). The impact of IGP on prey populations is more complex than predation and competition alone, as the act of predation reduces future competition for food resources as well as providing direct energetic benefits to the predator (Polis et al. 1989), and therefore can have strongly negative impacts on the prey species. The RBCT provided an experimental framework within which to investigate both the spatial relationship between badger abundance and hedgehog occurrence, and the effects of the experimental reduction in badger density.

1.2.3. Ground nesting birds

Although badgers are known to eat birds and their eggs (Neal & Cheeseman 1996), their potential impact on avian populations is not clear. There is anecdotal evidence from game managers that badgers have periodically been responsible for heavy losses of game birds, particularly eggs and chicks (e.g. Andersen 1955). There has been a dramatic decline and range contraction of farmland birds from the mid-1970s onwards has been well documented (see Fuller et al. 1995). There has been some popular speculation that in recent years in particular, predation by badgers may have contributed to the continuing decline in ground nesting bird populations (e.g. Butler 2003), although no supporting scientific evidence has been presented. Public concern over the potential role of the badger has culminated in questions being asked in the UK Parliament. A number of possible ecosystem responses to badger removal could have an impact on bird populations. If badgers do directly predate the nests of ground nesting birds, then their removal could precipitate increasing populations. Alternatively, if it were to lead to an increase in bird predators such as foxes or hedgehogs, this could increase predation pressure and a subsequent negative impact on populations. The species selected were skylarks Alauda arvensis and meadow pipits Anthus pratensis, as they are known to occur in badger diet, and were relatively abundant in the areas within which this project took place.

1.2.4. Hares

Hares have declined in numbers over the last few decades, to the extent that they are now of conservation concern and have a Biodiversity Action Plan. Additionally, their populations are sensitive to predation by foxes (Reynolds & Tapper 1995). Hence they may be subject to ‘knock-on’ effects of badger removal through any population response of foxes.

1.2.5. Rabbits

Rabbits are an ecologically and economically important species. They are an important prey for a range of carnivore species. Should the fox numbers respond to badger culling by increasing, the potential then exists for rabbit populations to be depressed, which may have implications for the prey base of species such as stoats Mustela erminea and weasels mustela nivalis. They are also considered pests in agriculture, therefore any increases in their populations in response to badger removal would be of considerable interest in the farming community.

1.3 Experimental design: population monitoring and complimentary research

A large program of population monitoring and complimentary research was implemented. Work was carried out in 4 to 5 of the RBCT triplets (depending on the species), in the proactive cull areas (experimental treatment) and no-cull areas (experimental control). Where possible, population monitoring and complimentary research was initiated prior to the onset of culling, and continued for several years while badger numbers were maintained at a low level throughout the duration of the RBCT.

Population monitoring

SID 5 (2/05) Page 7 of 36

Monitoring was carried out to identify any changes in population size of the selected species that occurred in response to badger culling. Surveys of fox, hare and rabbit populations were carried out using nocturnal spotlight surveys from around 60 point count locations distributed throughout each study area. A distance sampling approach was used to estimate density. Standardised spotlight surveys for hedgehogs were carried out, on a large sample of rural pasture fields and on amenity grassland areas in and around villages. Maximum counts of singing males were used to monitor populations of skylarks and meadow pipits.

Complimentary research

A wide-ranging program of complimentary research was implemented to investigate the mechanisms underlying any observed population responses. These included studies of fox diet using scat analysis, hedgehog ranging behaviour, source and rate of ground nesting bird nest failure, impacts of livestock on ground nesting bird abundance and nest success, use of badger setts by breeding foxes, and food web modelling.

2. METHODS

2.1 Monitoring the effects of badger removal on foxes, brown hares and rabbits

2.1.1 General

This section reports on the multi-species monitoring we carried out for nocturnal mammals, using vehicle-based spotlight surveys. Foxes Vulpes vulpes, hares Lepus Europaeus and rabbits Oryctolagus cuniculus were surveyed / studied simultaneously. The data collection and methods of estimating density for these species were therefore similar. Thus, the methods used apply to all three mammal species. However, different analytical approaches were taken for each species, therefore these are presented, followed by the results, for each species in turn.

The primary method used for monitoring the abundance of foxes, hares, and rabbits was spotlight counts at night using the distance sampling methodology. Distance sampling is a widely used method for estimating animal population density (Buckland et al. 1993). Direct observation data from spotlighting surveys can be used to estimate density, by applying distance sampling methodology to the collection and analysis of the data. Distance sampling is a method employed to correct count surveys, where only a certain fraction of the animals present are actually seen (because they are small and/or difficult to see, or spend some of their in cover etc.). The key to distance sampling is the declining frequency of animal observations at increasing perpendicular distances from the surveyors at point or line transects. These data are used to produce a model of detectability, or ‘detection function’, which estimates the probability of detecting animals at a given distance from the observer. By combining the number of actual sightings during a survey with this probability of seeing animals, an estimate of the true number of animals present is made. Initial surveys were carried out in the RBCT areas prior to the start of badger culling, therefore the timing of the start of work in each area was dependent on when they were recruited into the RBCT. Hence, surveying began in 2002 in RBCT triplet I in 2002, which was recruited rather later than the others due to the FMD outbreak. Therefore, surveying took place in 2000 to 2006 inclusive in the proactive and control areas of RBCT regions E, G & H, and from 2002 to 2006 inclusive in region I. A full description of the detailed survey methodology and data analysis is given in Trewby et al (2007), Davidson et al (in prep), and Trewby et al (submitted). However, the following is a brief summary of the methods:

2.1.2 Population monitoring

2.1.2.1 Data Collection – spotlight surveys

Surveying was carried out using spotlights from the roof of modified Land Rovers. A point count approach was taken, as it was considered too unsafe to carry out driven, line transects with an observer on the roof. Spotlight counts of foxes, hares and rabbits were carried out from approximately 60 points transects in each study area, providing adequate coverage of each study area and a sufficient sample with which to estimate spatial variance in foxes using distance (Buckland et al. 2001). Points were positioned along tracks and minor roads at least 500 m apart, to reduce the likelihood of double-counting the same individual that may flushed from one point to another by the surveyors. Points were only included if there was an unobstructed line of sight for 40 m and for a greater than 90° angle of the visibility plane. The ‘angle of view’ for each point was measured at the beginning of every survey year and was included as a measure of sampling effort. The first point on a transect was located at a given distance from permanent landmarks, such as road junctions, so that the same locations could be repeatedly surveyed. Accurate tripmeters installed in the vehicles were used to navigate to these points, and between points. This design allowed each point to be re-surveyed in consecutive years of the study. Surveys were carried out between January and May in each year, vegetation was low, maximising the chance of detecting animals. Each study area was surveyed at the same time of year in each year of the study, to reduce any seasonal behaviour effects that may affect the detectability of animals. Spotlight counts began at 21:00h GMT, or an hour

SID 5 (2/05) Page 8 of 36

after sunset, whichever was later, and were completed by 03:00h GMT. Each area was surveyed between 3-7 nights over the same 2-3 week period each year. Survey teams consisted of a driver and an observer. Surveying was suspended if weather conditions reduced visibility to less than 200m. Points were surveyed from the roof of a stationary Land Rover, using a red-filtered spotlight to sweep the visible portion of surrounding land. Distances to animals seen were measured using laser rangefinders. For each point, the time, observer, precipitation (dry, drizzle, rain, heavy rain, sleet/snow), cloud cover (0-8) and wind speed (Beaufort scale) were recorded, as these factors may affect the detectability of animals.

2.1.2.2. Distance analyses to estimate density of nocturnal mammals

Buckland et al. (2001) recommend 60-80 sightings to produce a reliable detection function, and hence probability of seeing animals. Numbers of hare and fox sightings in individual study areas in a single year were generally lower than this. It was necessary, therefore, to pool the distance data to produce the detection function model, and hence estimate the probability of detecting foxes. This is standard practice in distance sampling surveys. As the primary aim of the study was to monitor change in fox density between years, within each year the data were pooled across different study areas by treatment, (giving sample sizes between 45-101 observations), and used to model the detection function. The analyses were then stratified by study area to provide a density estimate for each study area in every year that it was surveyed. Density estimates for any given study area were therefore fully independent between years. Multi-covariate distance sampling was used to analyse the data, using DISTANCE 5.0 software (Thomas et al. 2005). Study area was included as a covariate as this allows for different detectability by study area, which may be caused by different geographical features between study areas. Observer, time of night, weather and wind speed were also included as covariates in the detection functions. An alternative method of producing density estimates using distance was also investigated. Distance data were pooled across years for each study area and then stratified by year, producing density estimates for each study area in every year that it was surveyed. Density estimates are given in Tables 1, 2 & 3 for foxes, hares and rabbits respectively, with % coefficient of variation (%CV). % CV is a measure of noise in the data and hence the precision of the estimates, calculated by dividing the standard error by the estimate itself, and multiplying by 100.

2.1.2.3. Data analyses to investigate the effects of badger removal on fox populations

Variance in fox densities was analysed by fitting a REML model to density estimates for each trial area in each year. The assumptions of normality and homogeneity of fox density estimates were checked by visual assessment of a histogram of residual values, and by plotting residual values against fitted values. The assumptions held, and no outliers were identified. Annual observations were treated as repeated measures of each area by modelling errors from sequential observations with a first order autoregressive structure. The main effects, treatment, triplet, trial area and year and the interactions between treatment, triplet and year, were entered as fixed, categorical terms. Treatment had two levels: culled (a proactive treatment area after the initiation of badger culling) or not culled (treatment areas before the initiation of badger culling, and control areas with no culling). The significance of explanatory terms was assessed by their Wald statistics, tested against a chi-square distribution. Only significant interactions were included in the final model. The analysis was weighted by using the inverse of the standard error squared for each density estimate; hence density estimates with greater precision were given more weight.

The main effects ‘treatment’, ‘cull status’, ‘region’ and the interaction between ‘treatment’ and ‘cull status’ were entered as fixed terms into the model. To determine if any post-culling effect in cull areas increased linearly over time, the main effect ‘years since initial cull’, was entered as a continuous variable, along with the corresponding interaction with ‘treatment’ in a multiple linear regression. The model structure was identical to that outlined above. The model was also run with ‘years since initial cull’ as a categorical variable to investigate the effect of badger culling in each year of the experiment. The R2 values from these two models were compared and used to determine which model best explained variation in fox density. The population growth rate ( r), of fox populations in the areas was calculated as loge(Nt+1/Nt), where Nt is the density of foxes at time t, and Nt+1 is the density of foxes in the following year. The estimated population growth rate for each area by year combination was plotted against fox density in year Nt and a multiple linear regression was used to investigate the relationship between r and fox density to test for density dependence. ‘Treatment’ and ‘density’ and their interaction were entered as fixed terms to determine if there was a treatment effect on this relationship. ‘Treatment’ was classified as a binary variable dependent on whether an area had been culled or not. ‘Region’ was included in the model to account for spatial variation in r and its interaction with ‘treatment’ and ‘density’ was also tested. All analyses were conducted in Genstat 8.1 (Lawes Agricultural Trust, Rothamstead, UK).

2.1.2.4. Data analyses to investigate the effects of badger removal on hare populations

The analyses carried out on the hare data were identical to those carried out on the fox data, and therefore for the sake of clarity, are not repeated here. However, due to the susceptibility of hare populations to variations in weather at key times of year, the influence of weather was also investigated. Weather data was collated, where

SID 5 (2/05) Page 9 of 36

available from the British Atmospheric Data Centre. Met Office weather stations were identified as close as possible to the centre of the survey areas. The weather stations were located between 2km and 10km from the centre of each 100km² area. The weather variables chosen for the analysis were annual rainfall and mean minimum January and July air temperature, chosen because it has been found that these variables have an association with hare density (Smith, Vaughan & Harris 2005). A multiple linear regression was used to investigate the relationship between hare population growth rate and hare density in order to test for evidence for density dependence. Factors including treatment, region, weather variables and fox density for the previous year were tested separately to investigate their effect on population growth.

2.1.2.5. Data analyses to investigate the effects of badger removal on rabbit populations

An identical set of analyses were carried out on the rabbit dataset as for hares (Section 2.1.2.4.)

2.1.3. Complimentary research on fox diet

Full details of this study are given in Trewby (2007). We studied both fox and badger diet during the period that we monitored fox populations to try to establish whether any observed changes in fox populations occurred as a result of changing resource availability in response to badger removal. We used scat analysis to attempt to determine whether (1) trophic niche overlap between foxes and badgers occurs; (2) fox niche breadth of both species alters in response to badger culling; and (3) the occurrence of food items in fox diet is affected by badger culling.

2.1.3.1 Scat collection

The diet study ran for four years from 2000 to 2003 and both badger and fox scats were collected as part of a survey, which aimed to quantify fox faecal accumulation rates to estimate fox density. Scats were collected from Wiltshire (triplet E) in 2000, 2002 and 2003, and in 2002 and 2003 for the Cotswolds (triplet I) and Staffordshire/Derbyshire (triplet G) regions. Hence there are no pre-cull data available for the Staffordshire/Derbyshire region, which could therefore only be considered as ‘control’ data. Reactive study areas were dropped from the RBCT at the end of 2002 and thus no scats were collected from these areas in 2003. As a result of a foot and mouth epidemic, no faeces were collected in 2001, which corresponds with cull year 1 for Wiltshire and Staffordshire/Derbyshire. Badger and fox scats were collected in a standardised manner. Scats were collected from a three metre wide strip along a transect that followed linear features within a randomly selected 1 km2 square. Eight 1 km2 squares were surveyed in each 100km2 study area and each square was surveyed twice, with 2 to 5 weeks between each survey. Surveys were carried out between January and April in each year and each study area was surveyed at the same time of year throughout the study. Linear features were defined as: (a) hedgerows and fence lines, (b) roadside verges, (c) edges of woodland, (d) footpaths, bridleways and other tracks (e) edges of rivers, canals, ponds, dykes, ditches, and lakes and an average of 4.5 km of linear features were surveyed per 1km square.

2.1.3.2. Identification of food components in scats

The procedures used to analyse faecal components were based on those recommended by Reynolds and Aebischer (1991), Cavallini and Volpi (1995) and Ciucci et al. (1996). Faeces were dried to a constant dry weight and their weights recorded. In order to separate faecal components into macro and micro-fractions for identification and quantification, faeces were washed through a stack of two brass Endecott sieves with mesh sizes of 500 microns and 150 microns. Macro-fractions were analysed by examining the sample under a dissecting microscope (10X). Components of the macro-fractions were then identified using a reference collection and appropriate keys according to the Phylum of the component. Hair remains were identified using the methods and keys of Teernik (1991) and Day (1966). Where possible, feather remains were identified using the keys provided by Day (1966) and Brom (1986). Invertebrates were identified from the fragments that occurred in the macro-fraction using the keys in Chinery (1993). Fruit remains were identified from the seeds and skin fragments that were found in the faeces, while grain husks were usually undigested and thus easily identifiable. Grass and other plant material were also identified from the macro-fraction. Micro-fractions were examined using a dissecting microscope (30X). Food remains in the micro-fraction were either earthworm chaetae or barbules from feathers.

2.1.3.3. Quantification of fox and badger diet

Badger and fox diet was quantified by calculating the frequency of occurrence of a food item and by estimating the percentage of food biomass consumed. The frequency of occurrence of food items is expressed as the percentage of the total number of faeces from which a particular kind of food item was recovered. The dry weights of food remains were converted into the percentage of food biomass consumed using published conversion factors established from feeding trials (e.g. Reynolds & Aebischer 1991). Invertebrates were converted into the percentage of biomass consumed by counting the number of individuals in a scat and

SID 5 (2/05) Page 10 of 36

multiplying by their fresh weight. The total number of earthworm chaetae in a scat was estimated by counting the number of chaetae in a sub-sample and extrapolating up to estimate the total number of chaetae (Palphramand et al. 2007). This can then be transformed into the percentage of biomass consumed (Reynolds and Aebischer 1991; Palphramand et al. 2007). Bootstrap simulations were used to estimate 95% confidence intervals around the estimated values of fresh biomass for each food item ingested (Reynolds & Aebischer 1991).

Levins’ measure of standardised niche breadth was calculated in order to assess the degree of specialisation of fox and badger feeding habits (Krebs, 1989). Values of this index can range between 0 (minimum niche breadth) and 1 (maximum niche breadth). The trophic niche overlap of the two species was calculated using Pianka’s measure of niche overlap (Krebs 1989). This measure ranges from 0 (no shared resources) to 1 (complete overlap). Trophic niche overlap was calculated using both the frequency of occurrence of food items and the percentage of biomass ingested. Both measures were calculated for the total sample of scats for both species and fox niche breadth was also calculated separately for samples from cull areas after badgers had been culled.

2.1.3.4. Data analyses to investigate the effect of badger removal on fox diet

To test for significant differences in the frequency of occurrence of the 11 main food types between badgers and foxes tests were used (Reynolds & Aebischer, 1991). The effect of badger culling on the occurrence of food items in fox diet was investigated using generalised linear models (GLM’S). The response variable was the presence or absence of a food item from a population of scats. The model had binomial distribution and a logit link function. The total number of scats from each study area for each year of the survey was the sampling unit used in these analyses. The main effects ‘region’, ‘cull status’ and ‘treatment’ and the interaction between ‘cull status’ and ‘treatment’ were entered as fixed terms into the model. The variable ‘cull status’ was initially included as a binary variable identifying samples as pre- or post-cull. ‘Cull status’ was also included as a continuous variable to investigate a treatment effect over time. Testing the effect of the interaction between treatment and cull status allowed the effect of badger culling on the occurrence of food items to be investigated. Spatial variation in the occurrence of food items in fox diet, at the regional level was taken into account by including the term ‘region’ in the GLM. The effect of ‘study area’ on the occurrence of food items in fox diet was also analysed in a further GLM, with the data being analysed separately for each region.

2.1.4. Complimentary research - the use of badger setts as fox breeding dens.

2.1.4.1. General

Foxes rear young in variety of underground burrows called earths (Reynolds & Tapper, 1995). Earths range in size from enlarged rabbit burrows to unoccupied, and occasionally occupied badger setts (Lloyd, 1980). Foxes use more than one earth to raise a litter (Reynolds et al. 1993). Vixens will also move a litter between sites, if the earth is disturbed in any way. Therefore, in any one fox territory it may be beneficial for the vixen to have more than one potential earth in which to wean cubs. Despite the need for a breeding earth, foxes prefer burrows dug by other mammals rather than digging their own (Weber 1982; Lloyd 1981). Foxes are not adapted for digging and therefore find using burrows dug by other species a more cost effective method of obtaining a den (Lloyd, 1980). Reynolds and Tapper (1995), found that 10 out of 15 litters were found at some stage to be occupying badger setts. Foxes also prefer breeding dens that have more than four entrances (Meia & Weber, 1992) making badger setts ideal fox breeding dens. Therefore, badger setts may be an important resource to breeding vixens, given that they are often large, with deep burrows that may have many branching tunnels (Neal & Cheeseman, 1996). A reduction in competition for den sites due to badger removal could therefore have an effect on fox productivity.

2.1.4.2. Data collection

Badger setts were surveyed for fox breeding signs in 4 of the 10 Randomised Culling Trial areas each covering a 100km2. These areas were H ,E, I and G. All three treatment areas were surveyed in each triplet. These were the areas where we carried out population monitoring.

In three of the four trial areas, the areas proactive areas had already been subject to culling. Only in area I was surveying first carried out prior to the onset of culling, enabling the collection of pre and post culling data. Badger setts were randomly selected from the DEFRA badger sett database. Each sett surveyed was chosen to be independent i.e. no two setts surveyed to be within the same vixen territory. Therefore setts were selected to be at least 1km apart. Once the survey setts were selected. Approximately 20 setts per area were surveyed (minus a small number in certain areas where landowner permission was refused). A total of 237 setts were surveyed across the twelve treatment areas in 2002. In 2003 and 2004, fieldwork was discontinued in the reactive areas, consistent with the cessation of work in this project and the RBCT itself in the reactive treatment. Data collection took place in the first 3 weeks of June in each year. This period was chose chosen for the survey because it is the period immediately after fox cubs are weaned, and therefore the chances of disturbing a vixen with cubs was minimal. Furthermore, fox breeding signs such as scats and carcass remains will have accumulated and be

SID 5 (2/05) Page 11 of 36

easier to find before they decayed. The same setts were surveyed each year and the proportion of badger setts used by foxes between each of the treatments compared in order to detect any changes once badgers have been removed. At each sett the following information was recorded: number of entrances; habitat type; presence of badger footprints, recent excavations, bedding and compacted tunnel floor; number of latrines and fresh badger droppings; number of fox scats; presence of carcass remains; location of holes used by foxes in a sett; presence of fox signs and presence or absence of fox breeding. If small fox scats were found at a sett it was assumed that these were deposited by cubs, and that the sett was being used by a vixen to rear cubs. Other indicators of fox breeding at a burrow include fresh food debris in and around the earth, and fox smell.

2.2 Monitoring the impact of badger removal on hedgehog populations

2.2.1. General

RBCT framework allowed us to investigate the influence of badgers on hedgehog populations in two main ways: 1. Spatial analysis - using the RBCT data on sett numbers and distribution to study the relationship between badger numbers and hedgehog distribution and numbers. 2. Experimental evidence – the effect of removing badgers on hedgehog populations, with experimental controls. Full and detailed descriptions of these analyses are given in (Young 2006) and (Young et al, 2006). A summary of the analyses and main findings are presented here. The detailed statistical analyses were carried out on the hedgehog data collected from 2000 – 2004, although the population monitoring data is presented up to 2006 and it is very clear that trends observed in the first four years continued through the course of the RBCT (Figure 3).

2.2.2 Population monitoring

2.2.2.1 Hedgehog surveys

Hedgehog surveying was carried out in 5 RBCT triplets: A, E, G, H and I. Surveys were carried out in the proactive and control area only of triplets A and E, and in all three treatment areas of G, H and I. A slightly different subset of study areas was used in the two main analyses. Hence, all areas surveyed prior to any badger culling (i.e. the undisturbed situation) were used in the analysis to look at the spatial relationship between badger and hedgehog distribution and abundance. In contrast, only proactive and matched control areas were used in the analysis of the effect of badger removal on hedgehog populations. Surveying began prior to the onset of badger culling. Surveys were repeated annually while badger numbers remained depressed, from 2000 – 2006 inclusive. Each field was surveyed over three separate visits between mid June and mid September and between the hours of 23:00 and 03:00. Amenity grassland and pasture fields were surveyed for hedgehogs by two fieldworkers using red-filtered spotlights (1.2 million candle power, Optronics, Oklahoma, USA). Each field was systematically searched from the perimeter inwards. All hedgehogs observed at a site were captured and examined to determine weight, sex and age class (adults > 1 year, juveniles < 1 year). Hedgehogs were uniquely marked by attaching eight heat-shrink plastic tubes (RS Components, Northamptonshire, UK) over individual spines. A portable soldering iron was used as the source of heat to shrink the tube onto the upper half of the spine, in order to avoid burning the skin. By applying different coloured heat-shrink tubes in discrete groups in various positions on the dorsal coat of spines, individuals could be uniquely marked. Hedgehogs were released within 10 minutes of capture. The total number of individual hedgehogs caught at each site over the three repeat visits provided an estimate of relative abundance and was divided by the area of the field to calculate relative density. Hedgehogs tend to forage on short grassland areas, where invertebrates are easily accessible. This includes amenity grassland areas (e.g. playing fields, parks etc) in suburban habitats and short grazed pasture fields in rural habitats (Doncaster 1992). The survey was therefore stratified by habitat type so that approximately three amenity grassland fields and nine pasture fields were surveyed per 100 km2 study site. Typically there were between five and seven villages within each 100km2 study site from which three villages were selected at random, ensuring each village was a minimum of 1.5 km from the edge of the study site. One amenity grassland field was surveyed per village. Amenity grassland fields tended to be situated on the edge of villages or within approximately 100 m of the village boundary. Three pasture fields were selected randomly from all pasture fields available within a 1.5 km radius of each village. A total of 23 amenity grassland fields and 82 pasture fields were sampled.

2.2.2.2. Badger activity surveys

During the hedgehog survey, the number of badgers observed in each field was recorded. In addition, a daytime survey of each field was also conducted to record signs of activity, including droppings, setts, runs and other field signs (e.g. tracks and hair caught on fences), in order to establish whether badgers were active in a field. Data on the location of badger setts and latrines across the study sites were provided from DEFRA badger surveys. These surveys were carried out no more than one year previous to the hedgehog surveys. DEFRA badger surveys were carried out by teams of trained, professional surveyors according to standard field protocols (Independent Scientific Group 1998), to identify setts and activity for the RBCT. Sett density was treated as an

SID 5 (2/05) Page 12 of 36

index of badger abundance that provided a broad picture of how badger abundance changed across relatively large scales rather than an accurate measure of badger numbers. The badger activity data from DEFRA were analysed in a GIS (ArcView 3.2, Esri, California, USA). The number of active main and subsidiary setts around each field was counted at two different spatial scales. These two scales were chosen to reflect (1) the average diameter of a badger home range of 500 m in a high badger density area (Rogers et al. 1997) and (2) the average dispersal distance of hedgehogs in rural areas of 2 km (Doncaster et al. 2001). The area of land surveyed by DEFRA within these two radii was also measured. Indices of badger density were calculated at these two spatial scales by dividing either (a) the number of main setts or (b) the number of all setts, by the area of land surveyed. Sett density at these two spatial scales was then named ‘local sett density’ (within 500 m radius) and ‘regional sett density’ (within 2 km radius). The proximity of each field to badger activity was calculated from the DEFRA survey data as the distance from the edge of the field to the nearest badger activity (i.e. sett or latrine).

2.2.2.3. Data analysis to examine spatial variations in hedgehog abundance in relation to badger distribution and abundance

The dependency in hedgehog occurrence (presence/absence) to habitat type, indices of badger abundance, proximity to badger activity and distance to nearest suburban habitat, was analysed by logistic regression in a generalised linear model (GLM). The variables retained in this procedure were then entered into a generalised linear mixed model (GLMM). Village nested in study site nested in region were entered as random effects in the model. The relationship between hedgehog numbers in amenity grassland fields and sett density in the surrounding area was examined using a GLM, including a measure of the size of fields (effort). A linear regression was conducted to investigate the relationship between mean hedgehog density in amenity grassland habitat and badger sett density at the study site scale. All statistical analyses were conducted using GenStat 6.2 (Lawes Agricultural Trust, Rothamstead, UK).

2.2.2.4. Data analysis to examine the effects of removing badgers on the hedgehog populations: Evidence for regulation of hedgehog populations by badgers

This experiment was based on a repeated measures nested analysis of variance design. The response of hedgehog abundance to badger culling was investigated through the interaction of fixed factors ‘time’, ‘treatment’ and ‘triplet’, with repeated measures random factor ‘field’ nested within each combination of triplet and treatment. The fixed factor treatment of badger abundance had two levels, high density (no-cull control) and reduced density (proactive) and time had five levels (0 indicates before the initial cull, and 1, 2, 3 and 4 years after culling began). The effect of interest was the interaction between treatment and time, which reflects the response of hedgehogs to badger removal.

The count of individual hedgehogs in each field over three visits for a given year was treated as the response variable (hedgehog abundance). An Iterative Reweighted Restricted Maximum Likelihood (IRREML) procedure was used to fit a generalised linear mixed model (GLMM) to the response variable. The area of amenity and pasture grassland fields was log-transformed and entered as an offset into the IRREML model, to take account of the variability in field size (i.e. survey effort). The main effects ‘treatment’, ‘time’ and ‘triplet’ and their interactions were entered as fixed terms into the model. ‘Field’ nested within triplet and treatment was entered as a random term. Testing the effect of the interaction of treatment and time allowed the effect of badger culling on hedgehog abundance to be investigated. A residual maximum likelihood (REML) generalised linear mixed model, was used to investigate the response of hedgehog occurrence to the manipulation in badger density. A REML linear mixed model was used to investigate the effect of the interaction of treatment and time on the weight of hedgehogs. Sex, age and subject (to take account of repeated measures on the same individual hedgehog) nested in field nested within triplet, were included as random terms. The growth rate, i.e. the per capita growth rate or instantaneous growth rate (r), of hedgehog populations in amenity grassland fields was calculated as loge

[(Nt+1)/Nt], where Nt is the density of hedgehogs (hedgehogs ha-1) at time t, and Nt+1 is the density of hedgehogs in the following year. A REML linear mixed model was used to investigate the relationship between growth rates of hedgehog populations in amenity grassland fields and hedgehog density (Nt) to test for density dependence. Observations of badgers in individual fields and treatment areas were too sparse to use as a reliable index of badger abundance and therefore it was not possible to carry out a full analysis of the effects of badger numbers on population growth rates. Consequently, the number of badgers observed in fields (in both amenity and pasture grassland) was averaged across all treatment areas for both proactive and control treatments and this was used as an index of badger abundance for a given year. Hedgehog population growth rate was calculated (as given by the equation above) from annual changes in mean density of hedgehogs in fields (in both habitat types) in proactive and control treatments. Linear regression was used to investigate the relationship between hedgehog population growth rate and mean badger relative abundance.

2.2.3. Complimentary research on hedgehog movement behaviour

2.2.3.1. General

SID 5 (2/05) Page 13 of 36

The mechanisms underlying the regulation of hedgehog populations is not fully understood. Therefore we carried out a study of hedgehog ranging behaviour, using radiotracking. In this study, working in the framework of the RBCT in areas where badgers were culled and where they were not culled, we investigated whether hedgehogs modified their habitat selection or their use of cover in response to predation risk by badgers. A full and detailed description of the methods, results and interpretation of this study is given in Young (2006).

The risk of predation is thought to strongly influence a range of behavioural traits in prey (Lima & Dill 1990). For example, when selecting habitats for foraging, animals are often required to trade-off the availability of food resources against exposure to predation risk (e.g. Sih et al, 1985; Gilliam & Fraser 1987). Animals have evolved a number of behaviours to reduce the risk of predation during foraging. Firstly, they may spend less time in risky habitats or avoid them entirely and use spatial refuges from predation (Jeffries & Lawton 1984). Animals often have to accept lower energetic returns in order to forage in safer habitats, as the latter are frequently the poorest in terms of their foraging profitability (Lima & Dill 1990). Secondly, animals in risky habitats may adopt different behaviours to minimise the risk of mortality, for example by forming groups to dilute predation risk or by increasing vigilance. Such strategies may have an energetic cost for prey through increased intraspecific competition for food or the loss of foraging opportunities (Banks 2001). In systems with intraguild predation (IGP), prey species compete with their predators for a shared food resource (Polis et al 1989). Consequently, both species tend to have similar foraging behaviours and habitat preferences, which increases the potential for encounters between predator and prey. Furthermore, in productive habitats, predators can achieve high densities by exploiting abundant shared food resources, and thus exert strong predation pressure on prey species even as prey abundance declines (Holt & Polis 1997). Therefore, in IGP systems, the most profitable foraging habitats for prey species are frequently the most dangerous, which makes habitat selection a key behavioural decision for intraguild (IG) prey. As discussed above, hedgehogs are intraguild prey of badgers. In summary, non-lethal effects of predation pressure are thought to have population level consequences and therefore large impacts on ecological systems.

Therefore, in this study, we estimated habitat selection by foraging hedgehogs in badger culling and control areas of the RBCT and tested whether it changed in response to a reduction of badger abundance. We hypothesised that if habitat use by hedgehogs is determined by the risk of predation by badgers, hedgehogs would use suburban habitats less and increase their use of preferred rural habitats in culling areas compared to controls. We also hypothesised that if hedgehogs modify their movement during foraging according to predation risk by badgers, then hedgehogs would forage further from cover after badger culling compared to controls. Additionally, we hypothesised that the size of home ranges and core activity areas of hedgehogs would change after badger culling in relation to controls. Finally, we monitored any dispersal events that occurred during the study to evaluate whether hedgehogs were capable of dispersing sufficient distances to move between suburban habitat patches.

2.2.3.2. Study design

Fieldwork was conducted in the Cotswolds triplet (triplet I) of the RBCT during 2002 and 2003. Data were collected in three study sites in the proactive area and four study sites in treatment areas that received no culling (reactive and control areas) which thus acted as controls. The aim of the study was to capture four animals (two adult males and two adult females) in each study site both pre and post badger culling, i.e. different subject animals were studied after the cull compared to before. The experimental design was based on a nested analysis of variance design. The response of habitat selection by hedgehogs to badger culling was investigated by the interaction of fixed factors ‘time’ with ‘treatment’ and ‘sex’, and the random factor ‘study site’ nested within treatment. The fixed factor treatment had two levels, high badger density (control) and low badger density (proactive cull), time had two levels (‘0’ indicates before the badger cull and ‘1’ after the cull) and sex has two levels (male and female). The effect of interest was the interaction between treatment and time which reflects the response of hedgehogs to badger removal.

2.2.3.3. Radiotracking

Hedgehogs caught on amenity grassland fields were fitted with radio transmitters (TW-3, 9g on acrylic mount, Biotrack, Dorset, UK). The transmitter was positioned longitudinally along the vertebrae to allow the antenna wire to trail behind the animal. Animals were released back at the site of capture within 20 minutes. Hedgehogs were relocated, using a radio receiver (TR-4, Telonics Inc., Mesa, Arizona, USA) and antenna (H-Adcock, Telonics Inc., Mesa, Arizona, USA), three times per week at night in order to identify location and habitat. Fixes on individuals were collected at approximately hourly intervals. Radiotracking began a minimum of 1 hour after dusk and ceased no later than 1 hour before dawn, to minimise the effect of the location of nest sites on the samples of locations collected. If possible, hedgehogs were relocated visually. The positions of animals were recorded on a field map with information on time, habitat type and behaviour using visual fixes or triangulation. Every two weeks, attempts were made to recapture all radiotagged hedgehogs so that the transmitter could be checked and the individual animal examined and weighed. Hedgehogs were radio tracked until a minimum of thirty fixes were

SID 5 (2/05) Page 14 of 36

obtained per individual, following recommendations by Seaman et al. (1999) for home range studies using kernel estimates, which took approximately 70-90 days.

2.2.3.4. Home range estimation and habitat mapping

A total of 47 adult hedgehogs were tracked during the study, of which sufficient fixes (approximately 30) were collected for 26 animals (15 females and 11 males; 12 individuals before badger culling and 14 after). The mean number of fixes per animal was 33.4 fixes ( 1.78 SE). The locations of all visual and radio fixes were mapped in a GIS (ArcView 3.2, Esri, California, USA). The Animal Movement extension of ArcView 3.2 was used to calculate two different home range indices for each animal to estimate parameters of animal movement and habitat availability: 1. A 100% minimum convex polygon (MCP) (Kenward 1987) home range estimator was calculated to encompass all the recorded movements of hedgehogs. 2. A fixed kernel estimator (with probability densities ranging between 5-95% at 5% intervals) (KHR) (Worton 1989) was calculated to estimate the core activity areas of active hedgehogs at night. All habitat types in the study sites were mapped and digitised in the GIS. Five broad habitat categories were used in the study: (1) amenity grassland; (2) suburban; (3) pasture; (4) arable; and (5) hedgerow and woodland. Polygons of home range indices and the locations of fixes were overlaid on the habitat data to calculate the composition of habitat in each home range and core activity area and to assign a habitat type to each fix. Patches of suburban habitat, woodland and hedgerows were considered as cover and were digitised in the GIS. The distances between the locations of foraging animals in amenity grassland and rural habitats, and the nearest cover were measured using the Nearest Features v. 3.6c extension of ArcView 3.2. If animals were located in a patch of habitat that represented cover, e.g. in a hedgerow, the distance was 0m.

2.2.3.5. Data analysis

Habitat selection by active hedgehogs at night was investigated at three spatial scales:(1) A comparison of home range habitat composition to habitat availability in the surrounding area. Habitat availability was defined for each individual by drawing a radius equal to the diameter of an average home range (374 m for males; 296 m for females) around the point of initial capture and calculating the habitat composition within this circle. (2) A comparison of core activity area habitat composition to habitat availability in the surrounding area (the latter calculated as above). (3) A comparison of the habitat associated with animal locations (fixes) to home-range habitat composition.

Compositional analysis (Aebischer, Robertson & Kenward 1993) was used to investigate habitat selection, by comparing habitat use with habitat availability. The difference between the habitat use and availability was calculated for each animal. A Wilk’s lambda test was used to determine whether habitat use was significantly non-random, using a chi-square test statistic to calculate the probability value. A matrix of the mean log-ratio differences of all possible pairs of habitat types across all hedgehogs was constructed and habitats ranked according to preference. A Student’s t-test revealed whether differences between pairs of mean log ratio differences within the matrices were significant. A residual maximum likelihood (REML) linear mixed model and multivariate analysis of variance (MANOVA) was used to investigate any changes in mean log ratio difference scores of habitat selection in response to badger culling. A maximum likelihood (REML) linear mixed model was used to investigate whether hedgehogs foraged further from cover in response to badger culling. Treatment, time and sex and their interactions were entered as fixed effects into the model. Subject nested in study site was entered as a random effect. A REML linear mixed model was also used to determine if the size of home ranges (100% MCP and 95% KHR) and core activity areas (65% KHR) changed in response to badger culling.

2.3. Monitoring the impact of badger removal on ground nesting birds

2.3.1. General

A full and detailed description of this work is given in Hounsome (2005). The main approaches to assessing the effect of badgers on ground nesting bird populations were: 1. A literature review and meta-analysis of studies of badger diet, to quantify the potential of badgers to affect bird populations. 2. The monitoring of bird abundance and nest survival before and after badger culling, within the framework of the RBCT. 3. The potentially confounding effect of livestock on the abundance and nest success of ground nesting birds was investigated, through experimental manipulation of livestock density and analysis of national datasets.

2.3.2. Monitoring the impact of badger removal on numbers of skylarks and meadow pipits, and on nest predation rates

2.3.2.1. Review and meta-analysis of birds in the badgers diet

SID 5 (2/05) Page 15 of 36

A full an detailed description of this aspect of the study is given in Hounsome & Delahay (2005). This literature review collated published and unpublished information on the occurrence of birds in the diet of the Eurasian badger, as an essential first stage in the process of understanding any potential impact of badgers on bird populations.

A total of 110 published studies were reviewed, 102 of which contained either reports of predation of birds by badgers, or recorded the presence of bird remains in badger faeces or stomach contents. This extensive literature search was not specifically targeted at those studies that reported bird remains, therefore it is hoped that the sample is both exhaustive and unbiased. Unfortunately very few studies reported the type of remains found and, as a consequence, comparisons of egg, nestling and adult bird predation were not possible. Some studies only recorded presence or absence and in others it was un-clear as to whether bird remains were absent or simply not recorded. These studies were not included in the analysis. As a result data on the percentage frequency of occurrence of dietary items were taken from 89 of the original 110 published sources. Where the percentage frequency of occurrence of bird remains was not directly quoted in a study (n = 9 cases), values were estimated from published graphs. Reynolds & Aebischer (1991) suggested that for comparative purposes percentage frequency of occurrence is a valid measure to use. As comparability between studies was the main requirement, percentage frequency of occurrence was therefore used.

2.3.2.2. Skylark and meadow pipit abundance surveys

The following hypotheses were tested: 1.The removal of badgers has no significant effect on the abundance of ground nesting birds. 2. The removal of badgers has no significant effect on the predation rate of ground nesting bird’s nests. 3. The removal of badgers has no significant effect on the source of nest predation.

Two RBCT triplets were chosen selected based on the timing of the badger culling operations, as it was essential that monitoring of ground nesting birds be carried out for a full breeding season both before and after badger removal. The two selected triplets were triplet H (Devon/Somerset border), and triplet G, the Staffordshire/Derbyshire border. The selection of the study areas for this study was severely restricted by timescale of this PhD study in relation to the allocation and recruitment of culling areas. Unfortunately these were not ideal areas for this particular study, in that the treatment (badger culling) areas H2 and G2 contained large areas of moorland (Exmoor National Park and Westmoorland), an important habitat for the two species in question. This undermined the ability of the present experiment to compare treatment effect over space i.e. the comparison in any one year between proactive and control, as it seemed likely that these different habitats would support different densities of skylark and meadow pipit and different predator assemblages. We attempted to take account for this where possible in the analyses and interpretation of results. Skylark and meadow pipit were chosen as study species as they have been recorded in the diet of the badger (Hounsome & Delahay 2005) and were relatively widespread and abundant in the two selected regions. Other ground nesters such as lapwing (Vanellus vanellus) also recorded in the badgers diet, and curlew (Numenius arquata) were present in the regions, but in such limited numbers as to make them unsuitable as study species. In March 2000, seven 0.25 km² sample squares were selected in each of the three treatment areas in the two triplets, giving a total of 42 sample squares. Squares were selected to be as distant as possible from any of the other sample squares in the treatment area (approx. 5 km), as their independence from each other in relation to predators, was deemed more important than their random selection within the area. In 2002, squares in the reactive areas were dropped from the experiment, in order to concentrate efforts in proactive and control treatments. This meant the number of sample squares was reduced from 42 to 28. Four further squares were dropped from the study due to removal of landowner permission. The data provided by pre-cull surveys in reactive areas, were used in the analyses as pre-cull, control data, as no culling of badgers had taken place.

In order to quantify the effect of a badger removal operation on the abundance of skylarks and meadow pipits, territory holding males were counted and mapped in each of the sample squares during the pre-removal breeding season (2000) and for the following four years. The maximum number recorded on any one visit was taken as the minimum population for that sample square, for that year. Owing to the foot and mouth disease outbreak in the UK, no monitoring could be undertaken in 2001. Post-removal monitoring therefore began 18 months after the initial cull. Field area was calculated using a GIS (ArcView version 3.2, Environmental Systems Research Institute Inc., California, USA). Population density derived from dividing the number of singing males by the total area of fields 2.3.2.3. Quantification of the rate and source of nest predation

A full description of the data collection, analysis and results is given in Hounsome (2005). As mentioned above, the effects of predator removal on the abundance of prey species can be masked by compensatory predation. Artificial nests have been used extensively in studies of avian breeding biology to assess the rate and source of predation of the real nests they mimic. The use of artificial nests in the present study was to enable robust analyses of the effect of removing badgers, as finding an equal number of real nests was beyond the scope of

SID 5 (2/05) Page 16 of 36

this study. A full review of the advantages and disadvantages of using artificial nests is given in Chapter 2 the PhD study of which this study was a part (Hounsome 2005).

Artificial nests comprising one canary egg and one plasticine egg were placed in the 0.25km 2 sample squares and monitored. This was carried out both pre-and post-cull. Two rounds of deployment were carried out in each. Seven nests were randomly placed in each square, in each round, using randomly generated coordinates. These were then visited six weeks later and their fate recorded. During the post-cull season (2002) there were two deployment rounds of seven nests in each square, but each nest was visited for four consecutive days after deployment and again ten days later (day 14). Nests were also checked four weeks later (six weeks since deployment) to enable comparison with pre-cull data. Nest status (survived or failed) was recorded, and in the case of a failure, a description of the remains was made, and number of eggs present and their condition was recorded, and any remains collected. An attempt was then made to determine the likely source based on evidence such as teeth or beak marks in plasticine eggs, or whether canary eggs were crushed, broken or holed. In general, allocations of predator source from field evidence are fraught with inherent bias and as a result the source of predation for the majority of nests was recorded as ‘unsure’. The source of predation for those nests that had disappeared completely was also recorded as ‘unsure’. Occasionally plasticine eggs were obviously chewed by larger mammals and in the case of fields without livestock this was then recorded as ‘mammal’. In fields with livestock however, these had to be recorded as ‘unsure’. Livestock were allocated as the cause of failure when nests had obviously been trampled or defecated on. Where silage cutting had been carried out failures were allocated to a sixth category ‘machinery’. The latter category is likely to be under represented, as cattle often aftermath graze post-harvested silage fields. The nest deployment experimental protocol is illustrated diagrammatically in Figure 13.

The overall data collection timetable is given in Table 4.

2.3.2.4. Data analysis - review and meta-analysis of birds in the badger diet

Approximate geographic co-ordinates (latitude and longitude) were obtained for 86 studies. A linear regression was used to investigate the effect of geographic position on the log-transformed percentage frequency of occurrence of birds in the badger’s diet. In order to test the hypothesis that predation on birds was more likely where badger diet was less specialised, the frequency of occurrence of bird remains was related to the overall diversity of badger diet in each of the 81 studies for which appropriate data were available. Standardised Levins indices (BA) (Krebs 1989) were calculated as a measure of dietary diversity for each study. The index produces values between 0 and 1, where a higher value indicates a more diverse diet.

2.3.2.5. Data analysis to investigate the effect of badger removal on the abundance of skylarks and meadow pipits

The data were analysed with a generalised linear model (GLM) that used a Poisson distribution, with an estimated dispersion parameter to test the response variable, maximum count of singing birds. A log link function was included and the log of field area was included as an offset. Triplet area was entered as a factor. Habitat data (from the Land Cover Map 2000) and livestock data (DEFRA) were entered into the model as covariates. An interaction term for the two explanatory variables, treatment and time, was also included in each analysis. The analysis was carried out using GenStat for Windows Version 7 (VSN International LTD, Hemel Hempstead, UK).Four separate analyses were carried out, for both skylarks and meadow pipits. Firstly, data from all sample squares (including the four containing moorland) were analysed simply as control or proactive, pre-cull or post-cull, with the three years post-cull data being pooled. This analysis was repeated, but excluding the squares containing moorland. The effect of time after culling was then investigated using the same model but with time since cull entered as a quantitative variable. This was also carried out with and without data from moorland squares.

2.3.2.6 Data analysis to investigate the rate and source of nest predation

Binary logistic regression (SPSS Inc. Chicago Rel. 10.0.7) was used to analyse nest survival data from the six-week checks carried out pre and post-cull, and from the 14-day post-cull check. Data points indicated that a nest had either survived or not survived. Information on the survival times over the initial four-day and 14-day post-cull checks were analysed using survival analysis (SPSS Inc. Chicago Rel. 10.0.7). A nest was classified as ‘censored’ if it remained undamaged at the end of four days or 14 days respectively. These data were analysed within a Cox regression in which differences between groups are parameterised as hazard ratios, combining information from both censored and uncensored observations. Habitat and livestock data were entered as covariates. Statistical inference was based on Wald tests. All tests were carried out both with and without data from moorland sample squares. The effect of time of year was investigated by comparing predation rates between deployment rounds.

SID 5 (2/05) Page 17 of 36

Data on the likely sources of predation were combined into six categories: bird, livestock, machinery, mammal (all mammals excluding rodents), rodent, and unsure. Predation of plasticine eggs by birds and small mammals often produces very clear teeth and beak marks, which then enables confident assignment of at least one predator (Møller 1989). Other classifications were based on highly likely sources. Mammals other than rodents were assigned when either of the eggs had been chewed. Flattened eggs in fields where livestock were present were assigned to the livestock category and nests that had failed in harvested fields were assigned to the machinery category. In most cases it was not clear enough to assign predation events to any of the categories above and in such cases predation source was recorded as ‘unsure’. It is arguable that in the absence of video cameras, all nest predation events should have been recorded as unsure. Therefore, the results based on predation source presented here, should be interpreted with caution. These data were placed in frequency tables and then analysed using Chi-squared goodness of fit tests. A global test of both pre and post-cull, proactive and control areas, was initially carried out. The effect of treatment was then analysed for pre and post-cull data separately. The effects of treatment and time for each category, was then investigated using binary logistic regression.

2.3.3. Complimentary research on nest success of ground nesting birds – the effects of livestock density

2.3.3.1. General

Full details of this study can be found in Hounsome (2005). During the initial stages of the experiment to investigate the effect of badger removal on nests of ground nesting birds it became apparent that direct damage by livestock was a source of considerable nest loss. When fieldwork was suspended across the country due to the FMD epidemic in 2001, we were allowed access to an area of pastural farmland adjacent to our research station. We took this opportunity to run a experiment to investigate the impact of livestock on ground nests. We expanded this to analyse national datasets of skylark and meadow pipit abundance, and livestock densities.

Declines in the abundance and distribution of farmland birds during this period, particularly ground-nesting species associated with grassland, have been well documented (e.g. Siriwardena et al. 1998; Chamberlain et al. 2000). The potential impacts of recent changes in farming practice on the success of ground nests are likely to be multifaceted and suggestions vary from indirect effects resulting from habitat modification, to the direct effects of trampling by livestock. For example, the shorter swards created by heavy grazing could deter certain species of birds from attempting to nest, thereby affecting breeding densities of some birds, or could attract others (e.g. Pearce-Higgins & Yalden 2004). The reduction in cover that results from heavy grazing has been suggested as a cause of higher predation rates on nests, especially by visually cued predators such as Corvids (Rands 1988; Baines 1990). Nest trampling by livestock has also been observed (Beintema & Muskens 1987; Baines 1990; Shrubb 1990; Wilson et al. 1997; Hart et al. 2002). In some studies, the impact of trampling was considered negligible (Baines 1990; Hart et al. 2002) while in others the effect was significant (Bareiss et al. 1986), and increased with stocking density (Beintema & Muskens 1987). Shrubb (1990) used the British Trust for Ornithology’s (BTO) Nest Record Scheme (NRS) data, to show that the percentage of nests lost to trampling was positively correlated with overall densities of sheep and cattle on English and Welsh grassland. The literature therefore suggests that both direct and indirect effects of livestock may influence the survival rates of ground nesting bird nests, although their relative importance remains un-clear.

2.3.3.2. Small-scale experiment - the effect of grazing regimes on ground nest survival rates.

Firstly, it was necessary to test whether there was a significant difference in nest survival times between grazed and un-grazed fields. A significant negative effect of cattle on the survival of nests, would invite a further test to determine whether it was habitat modification due to grazing, or the livestock themselves that were responsible. Hence in the second test, livestock were excluded from a previously grazed plot, and nest survival compared to a plot with livestock present. Consequently sward length and other aspects of the grazed habitat were, at least initially, the same in both treatments, the only difference being the presence or absence of livestock. A rejection of the null hypothesis (of no difference between treatments) in this case would indicate that it was the livestock themselves that had a significant impact on survival times and not the habitat they created. To strengthen this inference, a third hypothesis could be tested. If livestock were directly responsible for shorter nest survival times in grazed plots, then the survival times of nests in the un-grazed (silage) and recently grazed but livestock excluded plots, would not be significantly different (i.e. the failure levels of nests in both treatments would simply reflect the ‘normal’ background level of nest failure irrespective of sward length).

The experiment was carried out in June and July of 2001, on a farm in Gloucestershire, U.K. where sheep and beef cattle were grazed. The livestock used in the present study were a herd of 19 cattle, of mixed breed and a variety of sex and age, and a flock of 15 Suffolk sheep. Cattle and sheep were grazed in separate fields. Initially, three 1-hectare plots were marked out (50 m x 200 m) in three separate fields, each of which was subject to a different treatment. One field (containing cattle plot 1) held cattle at a stocking density of 3.2 cattle ha -1, the second field (containing sheep plot 1) held sheep at a density of 10.5 sheep ha -1, the third field (containing silage plot 1) had no livestock but was in use for silage production and therefore served as an un-grazed/cattle absent,

SID 5 (2/05) Page 18 of 36

experimental control. After three weeks an electric fence was erected to exclude cattle from cattle plot 1 for the remaining three weeks of the study. At the same time a further plot (cattle plot 2), was constructed adjacent to the original in the cattle field, and the cattle were allowed free access for the remainder of the study. Unfortunately as a result of husbandry practices it was not possible to carry out the same level of exclusion for the sheep, which could only be excluded for five days, before being reintroduced. A new silage plot was also marked out (silage plot 2) to replace silage plot 1 (Table 5).

Each artificial nest consisted of a clutch of two eggs. Eighteen nests (three of each possible combination) were placed randomly, in each treatment plot, weekly, for five weeks (Table 5). Each nest was visited once a day to determine its status and was deemed to have failed if it had been moved or eaten. In each case an attempt was made to determine the likely source of each nest failure using field evidence. The number of eggs present and their condition were recorded, and any remains collected. Two time-lapse video cameras were placed over-looking randomly chosen nests and once a failure event had occurred they were re-deployed to another randomly chosen nest. Nests were monitored for 12 days, approximately the incubation time of passerines (such as the Skylark Alauda arvensis and if they survived this period they were removed. Failure events were recorded in hours and days. To examine the effect of stocking density nests were placed at a similar density to the experiment described above in two different stocking densities of cattle, 2.71 cattle ha -1 (26 nests) and 13.75 cattle a-1 (11 nests). The fate and survival time of each nest was recorded as before.

2.3.3.3. Data analysis – small scale experiment

Data were analysed using survival analysis in SPSS (SPSS Inc. Chicago Rel. 10.0.7). The time to “failure” of each nest was recorded. A nest was classified as ‘censored’ if it remained undamaged at the end of 12 days. These data were analysed within a Cox regression in which differences between groups are parameterised as hazard ratios, combining information from both censored and uncensored observations. Statistical inference was based on Wald tests. The effect of any re-growth of grass in the recently grazed treatment was investigated using the above method, with a time dependent covariate. In addition, the effect of video cameras on nest survival rates was tested.

2.3.3.4. Data analysis - regional and national scale, effects of livestock on ground nesting bird densities

Data collected on abundance and nest survival within the RBCT (Section 2.3) framework were used here. Additionally, June Census data (DEFRA) at the scale of the 5 x 5 km based on the OS national grid were used to calculate stocking densities for cattle and sheep for. Stocking densities on grassland for 1990 – 2003 were calculated by dividing the total number of either cattle or sheep by the total area of grassland within each 5 x 5 km square, to provide a broad scale indication of the grazing intensity. The Land Cover Map 2000 data were used to provide habitat data, which is important in determining ground nesting bird densities, and could therefore be controlled for in the analyses. National abundance data for meadow pipits and skylarks were derived from the BTO Breeding Bird Survey (BBS). Nest success data for meadow pipits and skylarks at the national scale were taken from the datasets of the Nest Record Scheme (NRS) run by the BTO.

2.3.3.5. Data analysis – regional and national abundance and nest success

The regional analysis was based on the RBCT areas that we worked in, described in Section 2.3. Estimates of abundance and nest success were derived as per section 2.3. A GLM procedure was used to partition the variation in the count data associated with stocking density and habitat composition. The variables included were: proportion of each broad habitat that made up sample squares as derived from LCM2000, and the cattle and sheep density for the 5 x 5 km area the sample square fell within. This allowed the impact of livestock density on the abundance of both species of bird to be quantified directly. The national analyses, abundance of meadow pipits and skylarks were derived using the data derived from the BBS. Maximum counts for both species from all BBS squares surveyed from 1994 – 2003 were used as an indication of abundance, and entered separately into the analysis as response variables. Habitat effects were controlled for using LCM2000 data for each BBS square. The livestock density was calculated as above and represented the density of sheep and cattle for the 5 x 5 km square, which encompassed the BBS square. A GLM was used to investigate the data. The variables included were: proportions in each BBS square of the broad habitat categories (BH) derived from LCM2000, and the cattle and sheep density in that area. The error in the response variable (Maximum count) was modelled within a Poisson distribution, and a log link function was applied for analysis. Where the residual mean deviance was considered high (>2) inference was based on an F-test statistic. The procedure was carried out for meadow pipits and skylarks separately.

A binary logistic regression was used to investigate the effect of stocking density on the survival of skylark and meadow pipit nests. Very few of the nests in the NRS database had been revisited enough times to allow a formal survival analysis to be carried out, and it was felt that selecting only these nests would have greatly reduced the sample size available for the analysis and therefore the power to detect any effect. Only nests that

SID 5 (2/05) Page 19 of 36

had been recorded as failed were entered as such, all other nests were entered as “other” which includes successful nests and those of unknown fate. The NRS data included a habitat code, which best described the location of each nest. These broad habitat categories were entered as a fixed factor, explanatory variable. For each nest the sheep and cattle density for the relevant 5 x 5 km square for that year was calculated and also entered as explanatory variables. Statistical inference was based on Wald statistics.

2.4. Food web modelling

2.4.1. Methods & rationale

The main aim of food web modelling aspect of this project was to develop a multi-species model of the trophic dynamics of the major species of interest in the system under study, in the presence and absence of badgers. The mass-balance model ‘Ecopath’ was used (http://www.ecopath.org/index.php?name=About). This model has been used extensively in fisheries multi-species modelling, but has recently been used by terrestrial ecologists in the analysis of the food web e.g. Hodges, Krebs and Sinclair, 1999). In order to apply mass balance models to the UK terrestrial food web parameter data are required on density, biomass, productivity and diets of the major species. For the food webs in which badgers are involved, these data are available from the published literature and surveys or can be estimated. Once the model is balanced with respect to production and consumption of biomass, the sister model Ecosim (http://www.ecopath.org/index.php?name=About) can be used to assess changes in the biomass of species following the removal of other species. An important aspect to consider is that Ecopath is a food web, or trophic model and can only explore the likely changes in species abundance relating to changes in consumption rates of different species. Its value in this current context is to assess the community consequences of an increase in potential prey biomass due to the removal of a predator, the badger. It can not be used to inform us about other kinds of competitive interactions between badgers and other species, such as the potential population effects of behavioural interactions.

The list of species comprising the UK food web and which include all taxa thought likely to be most affected either directly or indirectly by the release of the likely badger prey species is shown in Table 6. The abundances and diets of all species used in the model were taken from a variety of sources, including CSL data from the population monitoring aspect of this project. Where data were available on biomass and diet in different habitat types, these were weighted to reflect the habitat composition of the experimental treatment areas (triplets E, G, H & I).

Using Ecopath with Ecosim allows the user to vary the extent that the badger abundance is reduced, to simulate different levels of efficiency in badger removal, which may have varied temporally and spatially in the RBCT. Changes in biomass were predicted using these models across a range of theoretical reductions in badger densities. Only where changes in other species are greater than 5% were they considered

3. Results

3.1. Foxes, brown hares and rabbits

3.1.1. Fox abundance

A full and detailed description of the results is given in (Trewby, 2007 & Trewby et al, In Prep). The main findings are described here. Estimates of fox density ranged from 0.25 foxes/km2 in G3 in 2005 to 5.15 foxes/km2 in E2 in 2000 (Table 1a.). These densities are largely consistent with previously published fox density estimates in UK studies (Corbett & Harris, 1991). Before badger culling was implemented, mean fox density was lower on average in cull areas than in control areas (Figure 1). Also, the pattern of change fox density estimates over the period of the study was different in cull areas than in control areas. Estimated mean fox density increased by 57% in culled areas in the two year period following the initiation of badger culling, whereas it decreased on average by 27% in the control areas (Figure 1). From cull years 3 to 6, while badger culling operations continued, mean fox density in culled areas remained higher, and fairly constant relative to control areas. Therefore the overall pattern was of an increase in fox densities immediately following the reduction in badger numbers, which levelled out after around 3 years, and stayed at this higher level while badger numbers remained depressed due to the ongoing culling. There was a 57% decrease in estimated mean fox density on average in control areas for the remainder of the experiment. All control areas followed this trend, although the degree of change in density estimates varied between years for each control area (Table 1a). This is well within the range that we would consider possible for natural fluctuation in fox populations. Fox densities were significantly affected by badger culling, and this effect varied between triplets (Table 1b). Under control conditions where no culling took place, there was a background decline in fox densities. Controlling for patterns of background temporal and spatial

SID 5 (2/05) Page 20 of 36

change, predicted fox densities in three treatment areas were 1.6–2.3 foxes per km2 higher than in controls, while in triplet I, where there was a less effective initial badger cull fox densities were unchanged (Table 1a)

The foot and mouth epidemic in 2001 prevented the collection of density data to estimate fox density for cull year 1, in three out of the four regions. This meant that for E, G and H, fox population growth rate for the first period is calculated from cull year 0 to cull year 2, and it is only for region I that the initial growth rate was calculated for the period 0-1 (see Table 1a). Therefore, for the regions E, G and H the population growth rate was estimated as the average growth rate of periods 0-1 and 1-2. There was a significant negative relationship between fox density and fox population growth rate (Figure 2), which provides evidence of density dependence in the rate of population growth (F6, 31 = 2.73, p = 0.03). Although there is a significant difference in fox population growth rates between region G and E, the accumulated analysis of variance indicated that there was no overall regional difference. The interaction between density, treatment and region was not significant and there was no effect of treatment or the interaction of treatment and density on fox population growth rates.

3.1.2. Fox diet

3.1.2.1. General

A total of 1559 fox scats and 648 badger scats were analysed. The number of fox scats collected from each study area in each year, ranged from a minimum of 30 in G3 in 2002 to a maximum of 222 in E1 in 2000, with a mean of 86. The equivalent numbers of badger scats ranged from a minimum of four in I3 in 2003 to a maximum of 100 in G1 in 2002, with a mean of 36. 28 food types were identified from fox scats and 24 from badger scats, which were combined into 11 main food groups for analyses (Table 7). Grass and plant material, such as deciduous tree seeds (e.g. Fraxinus excelsior) occurred frequently in scats of both species. However, these were likely to have been ingested incidentally during feeding, and were not considered to be dietary items. It was not possible to distinguish between hare and rabbit remains.

3.1.2.2. Occurrence and biomass

Many of the same food items were common to the diet of foxes and badgers. Indeed, representative items from all of the 11 combined food groups appeared in scats of both species (Table 7). However, there were generally substantial differences in both the frequency of occurrence and in the biomass of these items as recovered from scats of the two species (Table 7). For example, earthworms were the most frequently occurring food group in both species, occurring in 45.5% of fox scats, and in 96.1% of badger scats. However, earthworms comprised only 3.7% of the biomass of fox diet, compared to 85.7% of badger diet. After earthworms, Lagomorpha (rabbits and hares) was the most frequently occurring food group in fox diet, occurring in 42.6% of scats, and this was by far the most important group in terms of biomass, comprising 56.3% of estimated fresh biomass. Muridae and Aves were the next most frequently occurring groups in fox diet, occurring in 35.5% and 26.6% respectively, and represented 10.1% and 14.7% of food biomass respectively. Earthworms were by far the most important item in badger diet in terms of both frequency of occurrence and biomass, with other invertebrates the only other group to occur frequently, in 44.3% of scats. However, in terms of biomass, other invertebrates contributed less than Lagomorpha and Aves, estimated at 6.2% and 4.3% respectively.

3.1.2.3. Niche breadth and overlap

Based on the whole sample of scats, Levin’s measure of niche breadth for foxes was 0.5, while for badgers it was 0.15, indicating that foxes have a broader niche breadth than do badgers. This results from primarily from the fact that earthworms, represent 85.7% of the biomass consumed by badgers and occurred in 96.1% of their scats. These high values of occurrence and biomass are far greater than the values reported for the most important food item in fox diet, which was lagomorphs, representing 56.3% of the biomass consumed and occurring in 42.6% of scats (Table 7). In comparison, food items of secondary importance in fox diet, such as Muridae and Aves, represent a greater percentage of the biomass consumed, 10.1% and 14.7% respectively, than do the secondary food items in badger diet, Lagomorpha (6.2%) and Aves (4.3%). Other food groups, such as Artiodactyla, (mainly Ovis remains) and other mammals (mainly Rattus) also represent a greater percentage of the biomass consumed in fox diet, 6.0% and 8.1%, respectively, than in badger diet, where they represent less than 1% combined (Table 7).

There was no difference in fox niche breadth when calculated from samples collected in cull areas after badger culling (BA 0.53) compared to niche breadth from control areas at the same time (BA 0.52), or from cull areas before badger culling (BA 0.54). Using the frequency of occurrence data from the whole sample of scats to calculate Pianka’s measure of niche overlap indicates that trophic niche overlap between the two species is relatively high (Ojk = 0.67). However, when Pianka’s measure of niche overlap is calculated using the percentage of biomass consumed, overlap is low (Ojk = 0.14).

SID 5 (2/05) Page 21 of 36

Fox diet varied more widely than badger diet both temporally and spatially. However, the results of GLM’s revealed no effects on the occurrence of food items in fox diet in response to badger culling, either when cull status was treated as a binary or continuous variable.

3.1.3. Fox use of badger setts as breeding dens

It became immediately obvious that the proportion of setts in use by foxes for breeding was extremely low. In 2002, only 2 breeding dens were found in the proactive areas, 2 in the control, and one in the reactive. In 2003, one sett was found with signs of breeding foxes in the control and proactive areas. In 2004, 2 breeding dens found in the proactive areas and 2 in the control areas.

3.1.4. Hare abundance

Estimates of hare density ranged from 0 hares/km2 in H2 in 2005, to 10.8 hares/km2 in G2 in 2000 (Table 2.). These are within the published range of hare densities for the UK (Corbett & Harris, 1991). The coefficients of variation (CV) ranged from 24.5% to 107.5%, with an average of 35.3%. The primary reason for the low precision of these estimates was large variations between transects in the number of hares seen. With such large confidence intervals around many of the density estimates, it is clear that only a relatively large treatment effect could be detected.

No consistent relationships were revealed between weather factors and hare density or population growth rates. Before badger culling was implemented, mean hare density was very similar in culled areas and control areas (Figure 3.). The pattern of change in estimated hare densities over the period of the study was apparently different in cull areas to that in control areas. Estimated hare density increased by 113.5% in the control areas in the two year period following the initiation of badger culling, whereas it decreased on average by 34.7% in the culled areas. In cull years three and four, hare density was substantially lower in the cull than the control areas. In year 6, hare density in the culled areas increased and was at a similar level to the hare density in the control areas. Despite the apparent difference in hare population trend between the two treatment groups, the REML analysis revealed no significant effect of treatment, or interaction between treatment and triplet on the density of hares. Therefore despite the suggestion in the raw data that hare numbers declined in proactive badger culling areas compared to control areas, this was not supported by statistical analysis.

There was no relationship between hare density and hare population growth rate, suggesting no evidence of density dependence in the rate of population growth. Hare population growth rates were found to vary significantly between regions, (F = 3.26, df = 3,37 p = 0.034) with Staffordshire having a significantly greater population growth rate. There was no significant relationship between population growth and mean July temperature of the previous year (F=0.66, df=1, p=0.427). There was a significant positive relationship between hare population growth rate and total rainfall for the year previous to the survey (F=15.41, df=1, p<0.001).

3.1.5. Rabbit abundance

Estimates of rabbit density ranged from 4 rabbits/km2 in H2 in 2006, to 132 rabbits/km2 in I3 in 2002 (Table 3.). The coefficients of variation (CV) ranged from 16% to 67%, with an average of 30%. As with hares, the main reason for the low precision of these estimates was large variations between transects in the number of rabbits seen. Temporal variation in rabbit density estimates within study areas was also extremely large. This is consistent with other studies of rabbit populations, which have also shown that within-site variation in rabbit abundance over time is considerable (e.g. Trout et al, 2000). A complex range of interacting factors influence the extent of these variations, and include weather patterns, disease outbreaks and predator numbers. Within the context of this natural ‘noise’, it would have been unlikely to detect a systematic impact on rabbit populations in response to the removal of badgers. The same analyses as applied to the hare dataset were carried out, and indeed no population response was detected.

3.2. Hedgehogs

3.2.1. Spatial variations in hedgehog abundance in relation to badger distribution and abundance

Hedgehogs were extremely scarce in pasture fields, with only six individuals captured in three of 82 fields sampled (4% of fields). The relative density of hedgehogs varied from 0 to 0.79 ha-1 between study sites with a mean of 0.09 0.07 ha-1. Hedgehogs were more abundant in amenity grassland habitat, with a total of 44 individuals observed in 14 of 23 amenity grassland fields (61 %) resulting in a mean relative density of 1.54 0.44 ha-1. Average regional main sett density was 0.81 (range 0.35 – 1.37) and 0.72 (range 0.24 – 1.02) main setts km-2 around amenity grassland and pasture fields respectively. Mean regional density of all setts (i.e. main and subsidiary) was 6.19 badger setts km-2 (range 4.00 - 11.68) around amenity grassland fields and 6.56 badger

SID 5 (2/05) Page 22 of 36

setts km-2 (range 3.60 - 8.26) around pasture fields. There was no difference in regional sett density around fields between habitats taking into account the effect of study site (F1,94 = 0.31, P = 0.579). Nearest badger activity was significantly further away from amenity grassland fields (277.0 36.0 m) than pasture fields (160.3 15.5 m) including the effect of study site (F1,94 = 5.09, P = 0.026). Both backward and forward stepwise selection in a GLM retained habitat type (pasture or amenity grassland) and regional sett density in the optimal model as the best explanatory variables of hedgehog occurrence (F3,101 = 14.73, P < 0.001). There was no interaction between habitat and regional sett density (t101 = 1.24, P = 0.216). As mentioned, there was a strong preference for amenity grassland fields (t101 = -3.17, P = 0.002) with a mean predicted probability of hedgehog occurrence in this habitat of 62% and only 4.4% in pasture fields. There was a negative association between the occurrence of hedgehogs and regional sett density (t101 = -2.30, P = 0.021). Badger activity, local sett density, local main sett density, distance to nearest badger activity and distance to suburban habitat were not selected for inclusion in the model. The relationship between the predicted probability of hedgehog occurrence, estimated from the GLM, and regional sett density in both amenity grassland and pasture habitats is shown in Figure 4. The probability of occurrence was low for pasture habitat even when badger sett density was low. For example, at sett densities of less than 2 setts km-2, the model predicted that there was only a 10% probability of hedgehog occurrence. This declined with increasing sett density and above 7 setts km-2 the probability of occurrence was near to zero. In amenity grassland fields, in areas of low sett density, the model predicted that the vast majority of sites would be occupied. However, the probability of occurrence declined sharply as sett density increased. For example, in high sett density areas of over 10 setts km -2, hedgehog occurrence in amenity grassland fields was predicted to be only 33%. Generalised linear mixed modelling, controlling for the effects of region and study site, also revealed that habitat (Wald statistic = 25.30, df = 1, P <0.001) and regional sett density (Wald statistic = 6.35, df = 1, P = 0.018) were significant explanatory variables of hedgehog occurrence. Hedgehog abundance in amenity grassland fields decreased as regional sett density in the surrounding area increased (F 1,21 = 4.03, P = 0.045). The relationship between the predicted abundance of hedgehogs, estimated from the GLM, and regional sett density is shown in Figure 5a. It is predicted that at very low sett densities, hedgehog density would be greater than 3 hedgehogs ha-1. At high sett densities, above 10 setts km -2, the negative binomial regression model predicted that hedgehog density would fall to less than 0.5 hedgehogs ha -1 (Figure 5a). The relationship between hedgehog abundance and badger sett density in pasture habitat was not investigated due to the lack of fields (n = 3) supporting hedgehogs. There was no relationship between mean hedgehog abundance and sett density at the study site scale (equation: mean hedgehog density = 3.218 - 0.270 sett density, F1,7 = 4.72, P = 0.066). However, the probability value was very close to significance. Badger sett density explained 31.8% of the variation in hedgehog abundance. At very low sett density, hedgehog density was predicted to be greater than 3 hedgehogs ha-1. This is in agreement with the negative binomial regression model. In areas supporting badger sett densities above 11.8 setts km-2, the linear regression model predicted that hedgehogs would be completely absent from amenity grassland fields in suburban habitats (Figure 5b).

3.2.2. Effect of removing badgers on the hedgehog populations: evidence for regulation of hedgehog populations by badgers After the initial badger cull, the proportion of amenity grassland fields that supported hedgehogs in proactive treatment areas increased by 17% over the duration of the field experiment, whereas in controls it declined by 33% (Figure 6a). In the control areas, the proportion of fields with hedgehogs present in amenity grassland habitat varied widely during the experiment, declining from 60% before the cull to 15% in year 2 and recovering to 40% by year 4. Overall, there was a significant effect of the interaction of treatment and time on the occurrence of hedgehogs in amenity grassland fields (Wald statistic = 10.01, P = 0.04). Hedgehogs were very scarce in pastoral grassland habitat in both proactive and control treatment areas throughout the study and were generally found in less than 10% of pasture fields. There were too few observations to carry out statistical analysis to test the effects of badger removal on the occurrence of hedgehogs in pasture fields. However, examination of the raw data (Figure 6b) did not suggest there was any response of hedgehog occurrence in this habitat type. There was a significantly positive effect of the interaction of treatment and time on the number of hedgehogs in amenity grassland fields (Wald statistic = 14.84, df =4, P < 0.01). Hedgehog density increased by over 100% in proactive treatment areas over the duration of the experiment, whereas it decreased slightly in the control areas (Figure 7). In terms of the effects of badger removal on hedgehog abundance by sex and age, only adult females showed a statistically significant effect: there was a significantly positive effect of the interaction of treatment and time on female hedgehog numbers (Wald statistic = 10.55, df =4, P < 0.03).

The relative abundance of badgers varied widely throughout the study, including in the proactive treatment after badger culling started. This allowed an investigation of the relationship between an index of badger abundance at time t+1, as a measure of predation pressure, and the growth rate of hedgehog populations (Figure 8). Growth rate was calculated from changes in mean hedgehog density in the proactive and control treatments, as opposed to mean changes in density in individual fields, so therefore each of the data points in Figure 8 are calculated from repeated measures on hedgehogs. Growth rates of hedgehog populations (y) declined linearly with the index of badger abundance (x, equation: y = 0.64 – 3.30x, F1,6 = 6.42, P = 0.044). The index of badger abundance explained 43.7% of the variation in hedgehog population growth rate. There was no relationship

SID 5 (2/05) Page 23 of 36

between the index of badger abundance at time t and the growth rate of hedgehog populations. This suggests that the effect of badger predation on hedgehog population growth rate occurred without a time-lag.

Visual exploration of the density versus population growth rate data for the two treatment groups (Figure 9) suggests the possibility of an interaction between treatment and log mean density and the existence of two stable states of density. In the control treatment, where badger abundance was not manipulated, hedgehog population growth rate appeared to decline sharply as log mean density increased. This relationship suggested that population growth rate was stable (i.e. zero) at a density of approximately [ln(Nt)] = -0.7 (or 0.5 hedgehogs ha -1

when back-transformed). The slope of the linear regression line was -1.54, which may suggest over-compensating density dependence. In the proactive treatment where badgers were reduced in abundance through culling, hedgehog population growth rate appeared to decline less rapidly with increasing log of mean density. The observed population growth rate in the proactive treatment in year 2 was very low for the given log of mean density. This low growth rate was likely due to the unexpectedly high badger activity in the proactive treatment in year 2, despite previous badger culling (Figure 9). When this data point was excluded the population appeared to reach equilibrium at a higher density of approximately [ln(Nt)] = +0.6 (or 1.8 hedgehogs ha -1 when back-transformed). The slope of the linear regression line was -0.88, which suggests slight under-compensating density dependence. The strongest difference in the density dependent growth rates attributable to badgers is their reduction in hedgehog carrying capacity, represented by the displacement of the control line below the proactive line. The equilibrium population size in proactive treatment areas was approximately three times the density of the equilibrium population size in control areas. However, more replicates would be required to confirm the effect of the interaction of treatment and density on population growth rates.

3.2.3. Hedgehog ranging behaviour and habitat use

Compositional analysis showed there was significant non-random habitat use in hedgehog home ranges when compared to habitat available in the surrounding area (Wilk’s lambda Λ = 0.520; Chi sq = 17.022, df = 4, P = 0.002). A matrix ranked the habitats in the order (>>> indicates a significant difference according to a Student’s t-test): amenity grassland > suburban >>> hedgerow & woodland > pasture > arable. As expected there was significantly greater use of the top two ranking habitats, amenity grassland and suburban, than the rural habitat types. There was also significant non-random habitat use of hedgehog core activity areas from habitat available (Wilk’s lambda Λ = 0.608; Chi sq = 12.94, df = 4, P = 0.012). A matrix ranked the habitats in a similar order: amenity grassland >>> suburban > hedgerow & woodland > arable > pasture. Finally, compositional analysis of habitat use by foraging hedgehogs within the home range also showed a significant non-random use of habitat types in relation to their availability (Wilk’s lambda Λ = 0.248; Chi sq = 36.29, df = 4, P < 0.001). A matrix ranked the use of habitats in the order: amenity grassland >>> suburban > arable >>> hedgerow & woodland > pasture. Hedgehogs again showed the strongest selection for amenity grassland, which was used significantly more than the other habitat types. One notable difference of habitat use by foraging hedgehogs within home ranges, compared to the two previous scales, was that arable habitat was ranked significantly higher than both hedgerow and woodland and pasture.

3.2.4. Effects of badger removal on hedgehog habitat use, and use of cover

There was no effect of badger culling on the log ratio differences between habitat in the home range and the availability of habitat in the study area (log ratio differences; Wald statistic = 0.89, df = 1, P = 0.345; REML). In addition, no effect of the treatment was detected on the log ratio differences between habitat in the core activity area and the availability of habitat in the study area (log ratio differences; Wald statistic = 1.93, df = 1, P = 0.165; REML). Finally, no treatment effect was detected on the log ratio differences between the utilised habitat and the availability of habitat in the home range (log ratio differences; Wald statistic = 0.08, df = 1, P = 0.772; REML). A generalised linear mixed model revealed that there was no effect of the interaction of treatment and time on the proportion of amenity grassland in core activity area. There was some evidence of an effect of time, with a higher proportion of rural habitats in core activity areas in both proactive and control treatments in 2003.

A MANOVA revealed no effects of the interaction of treatment and time on the habitat use (log ratio differences) of hedgehogs both within the home range (log ratio differences; F4,18 = 0.28, P = 0.888; MANOVA) and comparing the proportion of habitat in home ranges (log ratio differences; F4,18 = 1.23, P = 0.333; MANOVA) and core activity areas (log ratio differences; F4,18 = 0.62, P = 0.656; MANOVA) to total habitat availability. However, the raw data suggested that the proportion of amenity grassland habitat (the preferred habitat type identified by the compositional analysis) within core activity areas and home ranges decreased markedly in proactive areas after the badger cull but remained relatively constant in control areas (Figure 10). They also suggested that use of suburban and some individual rural habitats increased after the badger cull. The REML model revealed there was no effect of the interaction of treatment and time on the proportion of amenity grassland in both core activity areas and home ranges. There was an effect of the interaction of treatment and time on the proportion of pasture habitats in core activity areas (Wald statistic = 6.07, df = 1, P < 0.05), which increased in the proactive treatment after the cull but decreased in the control areas. Hedgehogs foraged further away from cover after the badger

SID 5 (2/05) Page 24 of 36

cull in both proactive and control areas (Figure 11) as shown by the effect of time (Wald statistic = 4.6, df = 1, P < 0.05)

3.3. Ground nesting birds

3.3.1. Review and meta analysis

In 93 of the 110 studies reviewed, bird remains were recorded as present in either badger stomach contents or faeces. Combining the 89 studies where the occurrence was quantified, including true zeros, produced a cumulative sample size of 36699 records (31406 faecal samples and 5293 stomach contents). Of these, 2038 samples contained bird remains; thus the overall percentage frequency of occurrence was 5.55%. The mean of the 89 separately reported frequencies of occurrence was 6.94% (median = 4.33%). When data from the UK were analysed separately, the overall frequency of occurrence was 7.97% and the mean of reported frequencies was 7.01% (median = 7.50%). The linear regression model used to investigate associations between the frequency of occurrence of birds in badger diet and latitude and longitude, identified latitude as the only significant predictor in the model. There was a significant positive effect of increasing latitude on the frequency of occurrence of birds in badger diet (F(1,84) = 16.06, P<0.0001, Figure 12). Standardised Levins niche breadth values (Krebs 1989) of between 0.02 and 0.47 (mean = 0.24) were calculated for badger diet in the 81 studies examined. Linear regression suggested there was no significant association between diet niche breadth (BA) and the percentage frequency of occurrence of bird remains, or geographical location.

3.3.2. The effect of badger removal on the abundance of skylarks and meadow pipits

In total, 270 ground nesting bird surveys were completed throughout the course of this study, from 2000-2004. Forty-two sample squares were each surveyed three times in 2000, and 24 squares were surveyed twice in 2002, 2003 and 2004. Maximum counts for each species varied between sites and years. The maximum number of birds recorded in any one square was eight for skylarks, and nine for meadow pipits. The range of densities for each species were: 0 – 20 km-² for skylarks with a mean of 4.2 km-2, and 0 – 16 km-2 for meadow pipits, with a mean of 2.9 km-2 (see Tables 8a and 8b). Due to non-normality in the data, median densities of both skylarks and meadow pipits were calculated. Bird numbers were significantly greater on moorland compared to other habitats (Skylarks: moorland = 7.99 km-2, other = 2.55 km-2, ² = 18.41, P <0.0001, df = 2; Meadow pipits: moorland = 9.93 km-2, other = 0.54 km-2, ² = 31.1, P <0.0001, df = 2). Despite an apparent downward trend of skylark abundance in the control areas with relatively stable numbers in proactive areas (Figure 14a-d), the interaction between treatment and year was not significant, indicating no effect of badger removal on the skylark populations. There was, however, a significant effect of treatment on the abundance of meadow pipits, (including moorland squares Z = 2.2, P = 0.032; excluding moorland squares Z = 2.15, P = 0.036 (Figure 15a and 15b)), and when the data were analysed by year after cull, to investigate any trend over time (including moorland squares Z = 2.44, P = 0.018; excluding moorland squares Z = 2.37, P = 0.022 (Figure 15c and 15d)).

3.3.3. The effect of badger removal on the rate and source of nest predation

In total 853 nests were placed in the treatment areas in 2000 and 2002 (528 pre-cull and 322 post-cull). Overall, only 21 (2.5%) of these remained un-touched after six weeks. In the fourteen-day checks carried out in 2002, 34 (10.6%) nests untouched, with 288 (89.4%) having been disturbed or predated. In 2002 the nests were initially checked daily for four days after which 127 (39.4%) nests remained untouched, with 195 (60.6%) recorded as failed.

The binary logistic regression showed no significant effect of treatment by year on the proportion of nests predated after six weeks (Wald = 0.807, df = 1, P = 0.369). Neither was there a significant treatment effect on the proportion failing after 14 days (Wald = 2.974, df = 1, P = 0.085). The Cox regression survival analysis however, showed a significant effect of treatment on the predation rates of nests that were checked after 14 days (Wald = 4.181, df = 1, P = 0.041), with higher rates of failure in control areas, but was insignificant for those checked after four days (Wald = 2.613, df = 1, P = 0.106). The potential bias from the moorland squares was investigated again by excluding them from the analysis. There was no significant difference in the failure rates between treatments over 14 days (Wald = 0.517, df = 1, P = 0.472), or after four days (Wald = 0.138, df = 1, P = 0.71). This was investigated further by comparing survival rates of artificial nests on moorland compared to non-moorland. There was no significant difference in the survival rates over four days (Wald = 2.281, df = 1, P = 0.131) or 14 days (Wald = 1.765, df = 1, P = 0.184), although in both cases there was a tendency for higher survival times on moorland. Identification of the source of predation from remains in the field was highly subjective and problematic. Of the 832 nests identified as failed, 639 had disappeared completely, leaving 193 (22.3%) with some remains on which to base an assessment of the likely predator (see Table 9). There was a significant difference in the frequencies of causes of nest failure when all were analysed together (² = 158.52, df = 8, P <0.0001) and when comparing proactive and control treatments for pre-cull only (² = 21.05, P = 0.0008, df = 6). However, there was no significant difference in the frequencies of failure categories when post-cull only data were

SID 5 (2/05) Page 25 of 36

analysed (² = 7.39, P = 0.19, df = 6). Binary logistic regression indicated a significant interaction of treatment and time for the ‘un-sure’ category (Wald = 7.34, df = 1, P = 0.007).

3.3.4. The impacts of livestock on ground nesting birds

3.3.4.1. Small scale experiment

In total, 270 artificial nests were deployed during the experiment, although three were never re-found, and a further six were removed from the analysis having been subject to both sheep present and sheep removed (Table 10). The overwhelming majority of the nests (91.9%) failed. Twenty-one nests survived for the full 12 days, and were entered into the analysis as ‘censored’ data. The fate of 13 nests was captured on film (see www.badgerecology.org). Median survival times were used owing to the non-normal distribution of the data.

Artificial nests in grazed fields had significantly lower survival than those in un-grazed fields, while artificial nests in the grazed field had significantly shorter survival times than those in previously grazed fields (Table 11). Survival of artificial nests in previously grazed fields did not differ significantly from those in un-grazed fields (Table 11). The survival of artificial nests in previously grazed fields did not vary over time (i.e. in response to grass re-growth; Wald = 0.055, df = 1, P = 0.815). Neither was there any significant effect of a video camera placement on nest survival (Wald = 2.21, df = 1, P = 0.137. Figure 16 demonstrates the immediate reduction in failure rates of nests after the exclusion of cattle.

Of the thirteen nest failure events captured on video, six of these were the direct result of cattle; two nests were trampled and four were licked/eaten by cattle. It was also clear from the film footage that the cattle actively sought out eggs, and predation was not simply a result of grazing over the nest. The remaining nests were trampled by a sheep (1), trampled by a rabbit (1), eaten by a badger (1), or taken by crows Corvus corone (4). Of the twenty-six nests in the low stocking density field (2.71 cattle ha-1) three nests survived for the full 12 days, whereas all 11 nests in the high-density field (13.75 cattle ha-1) had failed after three days. Survival of nests in low stocking density fields (median = 60 hours) was significantly greater than in high stocking density fields ((median = 12 hours) Wald = 9.93, df = 1, P = 0.002)).

3.3.4.2. Regional scale abundance and nest success

In total 86 values of abundance for both skylarks and meadow pipits were entered into the analysis. The deviance ratio associated with the full model was significant for both of skylarks (F (75,9) = 3.73, P <0.001), and meadow pipits (F(75,9) = 9, P <0.001) indicating that by using the variables entered into the model it was possible to predict the abundance of either species. There was no significant contribution to the model fit from the variable sheep density for either species. Cattle density however, was significant for both Skylarks (Z = -3.28, df = 75, P = 0.002) and meadow pipits (Z = -3.71, df = 75, P <0.001) this shows that a significant proportion of the variance in the abundance of both skylarks and meadow pipits was explained by the density of cattle in the associated 5 x 5 km square. This is consistent with a significant negative impact of cattle on the abundance of both skylarks and meadow pipits as measured in the RBCT areas.

There was a significant difference in the failure rates of nests in grazed land (with or without livestock present) compared to non-grazed land (i.e. arable, silage, and set-aside) (Wald = 34.75, df = 1, P < 0.0001). The direct impact of livestock on the failure rates of nests was quantified by testing the difference between grazed land with livestock present on at least one visit, and grazed land where livestock were absent on all four days. The result of this analysis shows that there was evidence of a significant direct impact of livestock on the failure rates of artificial nests (Wald = 17.88, df = 1, P < 0.0001). Furthermore, in the absence of livestock there was no significant difference in failures rates between fields that had and had not been grazed (Wald = 2.36, df = 1, P = 0.125). These results suggest that direct impacts from livestock were the main cause of increased failure rates in the RBCT areas.

At a national scale, combining all the BBS squares surveyed in England between 1994 and 2003 produced a total sample size of 10531 surveyed squares. The deviance ratio for the model was statistically significant for both meadow pipits (F(19,10531) = 867.77, P <0.001), and skylarks (F(19,10531) = 246.96, P <0.001). Using the full model it was possible to predict the abundance of both species, as with the regional scale analysis. There was a significant negative association between livestock and the abundance of meadow pipits (cattle, Z = -14.96, df = 10531, P <0.001; Sheep, Z = -2.75, df = 10531, P <0.001), and skylarks (cattle, Z = -6.08, df = 10531, P <0.001; sheep, Z = -2.52, df = 10531, P = 0.012). These results indicate that when habitat is taken into account, there is a significant negative association of both sheep and cattle density with the abundance of skylarks and meadow pipits, and furthermore that the effect of cattle density is far greater than that of sheep for both bird species.

There was no significant relationship between livestock density (cattle or sheep) and the failure rates of nests (P always > 0.05).

SID 5 (2/05) Page 26 of 36

3.4. Food web modelling

When badger biomass was allowed to be reduced by 10%, none of the other species changed in biomass by more than 5%. When badgers were reduced by a theoretical 40%, their main prey, the earthworms, A. longa and L. terrestris, and another earthworm predator, the mole showed a response greater than 5% (Figure 17). Complete 100% removal of badgers lead to a response greater than 5% only for stoats, foxes, ‘Medium’ birds, A. longa and L. terrestris, hedgehogs and moles (Figure 18). These responses were further explored in terms of change in the number of individuals and species biomass over long term runs of the model Figures 19 and 20 repsectively. Although the values presented in Figures 19 and 20 are for 100 years after the reduction in badger biomass, to allow the long-term effects to be realised, equilibrium is generally reached for all species after 25 years.

4. Discussion

4.1. Monitoring the effect of badger removal on foxes, brown hares and rabbits

4.1.2. Foxes

4.1.2.1. Fox population response

This is a rare example in wildlife research in which a controlled and replicated removal experiment was used to investigate interspecific competition between two carnivores. By monitoring fox density within the framework of a controlled removal experiment, we were able to provide evidence for the occurrence of interspecific competition between badgers and foxes in our study areas. Following the start of badger culling, mean fox density increased by 57% in culled areas within 24 months, but decreased by 27% in control areas. After this initial response, the mean fox density in culled areas remained constant relative to control areas for the remaining four years of badger culling. The changes in fox densities observed over time in control and cull areas over the duration of this project are comparable to patterns observed elsewhere (e.g. Hewson & Kolb 1980; Heydon & Reynolds 2000; Sidrovich et al. 2006). Studies into factors that may affect fox density have lead to opposing views being reported in the literature. Hewson (1986) and Baker & Harris (2006) suggest that fox density is determined primarily by the availability of resources, while Heydon & Reynolds (2000) conclude that shooting can substantially reduce fox density. It is very likely to be the case that that complex interactions between both resource availability and anthropogenic mortality will determine fox density at a regional scale. It is unlikely that there were systematic biases in the factors that affect fox density, between the 4 control and 4 cull areas in this study. That is, we do not have any reason to believe that after badger culling began, there was any substantial change across all four areas of either of the treatment groups. Hence, the different patterns seen in the two treatment groups is very likely to be due to the reduction in badger numbers in the proactive group. Without an experimental control, the patterns of fox population change viewed in isolation would have been unconvincing. However, because the pattern of change in the culled areas was so markedly different from that in the un-culled countryside (represented by the four matched control areas) we can conclude with some confidence that the fox populations changed in response to badger removal. This illustrates the great value of an experimental approach to such a question.

The density of competitors plays an important role in determining the strength of competitive interactions and the extent of the population responses of the inferior competitor (Glen & Dickman 2005). If interspecific competition between badgers and foxes does exist, then reducing badger density over several years is likely to have decreased the strength of competitive interactions between the two species. This could allow a sustained increase in fox population size due to increased resource availability. This hypothesis is supported by our results, which showed that the initial increases in mean fox density in cull areas were maintained for the duration of the experiment. However, the nature of the interaction between badgers and foxes causing this limitation of the fox population is unclear. Population growth generally occurs through increased survival of adults and juveniles, increased reproductive output, or immigration (Lloyd 1980). However, for this study, we do not have the cohort life history data required to determine which of these were responsible for the observed pattern. It is necessary therefore to consider the aspects of the two species’ ecology that may offer potential explanations for the results.

We carried out a detailed study of fox diet in our study areas from 2000 – 2003, and although there was a great deal of diet information generated, there was no evidence for any systematic change in fox diet after the onset of badger culling. Therefore despite the overlap between the two species in the food types taken, we have no evidence to suggest that competition for food, and an increase in its availability was the reason for the increase in fox densities. The availability of suitable breeding sites for foxes could potentially be an important factor in determining reproductive output. Badger setts are large, well protected structures, and may be a valuable resource to breeding foxes. A fox taking up residence in an empty sett to give birth to and rear cubs may have an

SID 5 (2/05) Page 27 of 36

increased chance of being successful. However, badgers are likely to monopolise these structures to the exclusion of foxes. The large decline in badger numbers due to culling may have caused many badger setts to become available to foxes. A large increase in the availability of this resource may have precipitated an increase in fox reproductive output, through a combination of a greater proportion of female foxes actually giving birth, and greater cub survival. This would in turn quickly lead to an increase in density. It is known that species that are able to adapt their reproductive output to their environment can respond relatively rapidly (Sibly et al. 2005). Studies of fox populations have shown that at densities below carrying capacity, reproductive output increases and at densities above carrying capacity reproductive output is reduced (Hartley et al. 1994; Heydon & Reynolds 2000). Therefore, fox populations may have the potential to respond quickly to changing conditions. This appeared to be the case in this study, where fox populations increased substantially over the two years from the start of badger culling. The fact that the fox populations in culled areas did not increase indefinitely is consistent with there being other density dependent factors operating on the fox population. This is supported by the apparent density dependence in population growth rates which is consistent with studies elsewhere (Heydon & Reynolds 2000; Lindström 1982). Thus, the limiting effect of competition from badgers was reduced, allowing an increase in fox density, which was ultimately constrained by other, density dependent factors. Therefore overall, badgers appear to limit the density of fox populations, and that interspecific competition between the two species has an important role in determining fox population size. The actual mechanism by which badgers limit the fox population remains unclear. To elucidate this, intensive behavioural studies would be required, involving spatial and temporal niche overlap studies, and the construction cohort life tables of sympatric populations under experimental manipulation.

4.1.2.2. Fox diet response

The badger was hypothesised to be the superior competitor, due to its larger size and observed dominance at artificial food sources (Macdonald et al. 2004). Thus, the badger may be expected to monopolise the most profitable food source, while foxes would be forced to feed on alternative food items. If this were indeed the case in reality, we would expect to see a shift in fox diet as a result of badger culling. The fox was found to have a broader trophic niche breadth than badgers, and this did not change in response to badger culling. This implies that fox feeding habits are not affected by competition from badgers and that badgers do not affect the availability of food sources that were consumed by foxes. The results of this study do not provide any evidence that badgers and foxes are interspecific competitors for food resources. Thus, we conclude that the increases in fox density in response to the reduction in badger density are the result of an increase in the availability of another resource, such as den sites, and/or the decrease in some other competitive interaction, such as interference competition.

4.1.2.3. Fox use of badger setts as breeding dens

There were very few signs of foxes using the badger setts identified by Defra staff for the RBCT. This was the case in both proactive and control areas, and did not increase markedly in proactive areas as time went on from the start of culling. This therefore does not provide evidence to support the idea that the increase in fox densities observed in response to badger culling were a result of increasing den site availability. However, we used the Defra database of badger setts as a basis for this work, and thus this was not an exhaustive survey for fox dens. Therefore this result does not rule out the possibility that badger density reduction led to an increase in fox productivity through a reduction in competition for breeding space.

4.2. Hares

The raw data suggested that there was a decline in hare densities in areas where badgers were culled relative to the control areas. However, despite there being a substantial difference in trends in the mean density of hares from the two treatments, this was not a statistically significantly effect. There was considerable variation in the hare data, both across the study areas, and within areas over time, therefore only a very large effect with respect to badger culling could have been detected. It should also be mentioned here that when the hare data was analysed without using repeated measures analysis, there was a significant treatment effect. Therefore we are cautious in interpreting the results. It is possible that removal of badgers did have a negative on hare numbers, and that a larger sample size may have detected this.

Badgers do not readily predate hares, although they may take leverets opportunistically. Hares do not occur frequently in badger diet (Neal & Cheeseman, 1996). Also, there would appear to very little niche overlap, or grounds for interspecific competition between badgers and hares. Therefore, rather than a direct effect of the removal of badgers, any reduction in hare numbers that may have occurred would be a ‘knock-on’ community effect, resulting from the increase in fox densities that occurred after badgers were removed. This is biologically plausible. In contrast to badgers, foxes are important predators of hares, particularly of leverets (Reynolds & Tapper 1999), and that hare populations can be 2.5 – 5 times larger in the absence of foxes. The sustained increase in fox populations observed in culled areas relative to the un-culled areas, has the potential to cause a decrease in hare populations over the period of the study. Despite the lack of statistical support for this, it is worth briefly considering the implications of this effect in the context of the conservation status of the hare. The

SID 5 (2/05) Page 28 of 36

brown hare was one of the first animals to be considered in the UK Biodiversity Action Plan (BAP) programme, the UK’s response to the Convention on Biological Diversity. It was chosen because it was once very common, was widespread and had declined significantly during the post-War period, perhaps declining by as much as 75% since the second world war. The BAP proposed that measures should be taken to improve numbers so that by 2010 our countryside should support at least two million animals in winter. It is generally accepted that the main reason for the decline in hare numbers has been the change in farming practices of the last century. The patchwork landscape of a century ago consisted of a mix of cereals, root crops, and grass with livestock. Small fields allowed hares to move readily between them, grazing on different crops and grass at times of year. For example, long cereals provided cover for hares in summer, and ley grass and pastures provided good grazing conditions. In winter, root crops and winter cereals provided both cover and forage. Modern arable systems make life difficult for hares in fundamental ways. Crops such as oilseed rape and winter cereals provide food in winter but by early summer these crops are too mature to allow grazing. Therefore at this time hares have to rely on field boundaries, tracks and roadsides for foraging. In addition to the difficulties associated with finding sufficient food, this makes hares easier to find for foxes patrolling the field boundaries. On livestock farms, meadows are cut more often than in the past. As well as being subject to high mortality from modern grass cutting mowers, cut meadows provide little cover, and hence hares (leverets in particular) are at greater risk from predation from foxes. Within Defra’s Countryside Stewardship Scheme, there are arable options available to encourage farmers to manage the habitats on their farms in ways that are designed to improve the landscape for the existence of species such as the brown hare. These have been shown to be successful for increasing hare populations (Defra report, MA01010). Although land management issues are the most important factors in determining the status of hare populations, it is clear from the results of this project that any management action that increases fox populations may potentially confound efforts to increase the hare population to reach the targets set by the UK BAP for hares.

4.3. Rabbits

The temporally highly variable nature of the rabbit density estimates meant that no effect of the removal of badgers was detected.

4.4. Hedgehogs

4.4.1. Distribution and abundance of hedgehogs in relation to badgers, and response to badger removal

All the study sites supported badger main sett densities that were higher than the threshold for hedgehog occurrence predicted by a previous study in Oxfordshire of the effect of badgers on hedgehog populations (Micol et al. 1994) of 0.23 main setts km-2. Hedgehogs were at very low density in rural habitats, being found in only 3.7% of pasture fields. This is less than 10% of the distribution recorded in that study (Micol et al. 1994). No historical data on hedgehog populations in the current study sites exist and therefore it is impossible to directly determine if hedgehogs have declined in abundance or were always very scarce. Previous field experiments (Doncaster 1992, 1994) provided strong evidence that badgers can regulate the abundance and distribution of hedgehogs in localised areas of pastoral dominated agricultural habitat. In areas of high badger density, hedgehogs moved from pastoral grassland and used suburban habitats more frequently. Hedgehogs also suffered from higher mortality through predation in high badger density areas than in areas of the same habitat where badgers were less abundant. This work reveals that in large areas in the Midlands and Southwest regions of England, hedgehogs exist at very low densities or are completely absent from pastoral habitats. This is therefore in agreement with the prediction that hedgehogs would be absent in rural habitats supporting high badger numbers.

Hedgehogs were almost completely limited to suburban habitats. Amenity grassland has greater earthworm availability than pasture habitat and consequently is able to support hedgehogs at higher density (Doncaster 1994). It is thought that due to human disturbance badgers tend not to be as active in suburban habitats, which act as a refuge for hedgehogs from badger predation (Micol et al. 1994). Spatial refugia therefore appear to facilitate the coexistence of hedgehogs and badgers at the rural landscape scale. These results are in agreement with a number of previous studies of terrestrial vertebrates (e.g. Mills & Gorman 1997; Durant, 1998, 2000; Sergio et al. 2003) that have identified spatial refugia as a vital mechanism in alleviating the effects of IGP on prey populations.

Both the probability of occurrence and hedgehog abundance in amenity grassland fields declined as regional badger sett density in surrounding areas increased. In areas of high badger sett density, the likelihood of hedgehogs occurring in these sites was predicted to decline towards zero. This was not expected as it was assumed that suburban habitats would be relatively free from the effects of predation. Hedgehogs have been shown to demonstrate predator avoidance behaviour by avoiding localised areas tainted with badger odour (Ward et al. 1996; Ward et al. 1997). However in this project, the presence of badger activity was not related to hedgehog occurrence, which indicates that hedgehogs do not necessarily avoid amenity grassland fields where

SID 5 (2/05) Page 29 of 36

badgers are active. Local sett density and distance to nearest badger activity, which were the best indicators of badger activity in both habitat types were also not related to hedgehog occurrence. This was also unexpected, as it seemed probable that the localised abundance of badgers around a site and the proximity of a site to badger activity would determine predation risk and therefore affect the occurrence of hedgehogs at a particular site. Some amenity grassland fields provide very abundant and accessible food resources (i.e. earthworms), which may be preferentially used by hedgehogs in spite of the presence of badgers. The highly fragmented structure of suburban areas (e.g. gardens), surrounding or near to the amenity grassland fields, provide hedgehogs with more nearby cover which may reduce the risk of predation in this habitat.

It was the index of badger density on the scale of hedgehog dispersal (i.e. 2 km) around amenity grassland fields that was associated with hedgehog occurrence and abundance, rather than local badger density or proximity to badger activity. A likely explanation for this relationship is the effect of high badger density and subsequent predation pressure on the ability of hedgehogs to move between patches of suburban habitats. Hedgehog populations in suburban habitats were spatially discrete and likely to constitute a metapopulation. Hedgehogs have been recorded to disperse up to distances of 3.8 km (Doncaster et al. 2001) and patches of suburban habitat were rarely farther apart than this in the study sites and therefore likely to be connected by dispersal. However, hedgehogs are known to be vulnerable to predation by badgers whilst moving through areas of high badger density (Doncaster 1992). In addition, hedgehogs in rural habitats have been shown to both avoid areas with badger odour through the use of olfactory cues (Ward et al. 1997) and to move further and faster from areas where badgers are abundant (Doncaster 1992). Therefore, areas of high badger density could represent a barrier to movement into and from a site (Doncaster et al. 2001), which may explain the variations in the occupancy of suitable sites by hedgehogs recorded in the current study.

Given such low numbers of hedgehogs in pasture fields, the analyses of the effect of badger removal on hedgehogs were focussed on the spatial refugia represented by the amenity sites. Over the five years covered by these analyses, mean hedgehog density in proactive areas increased by roughly 100% after badger culling began, whereas there was little overall change in control areas. This trend continued to the end of the RBCT. A significant effect of the interaction of treatment and time confirms the hypothesis that badgers restrict the growth of hedgehog populations. The response of hedgehog numbers to badger culling varied by sex, with female hedgehogs increasing in abundance whereas males remained unchanged. Females used open amenity grassland habitats more frequently than males, and also foraged further from cover, and therefore may be more prone to predation and thus have higher rates of mortality than males. There was evidence that badger culling had a positive effect on the occurrence (i.e. presence) of hedgehogs in amenity grassland sites. The model predicted that the occurrence of hedgehogs remained stable in areas where badgers were culled but declined in control areas.

Our results also provided evidence of density dependence in their population growth rates (i.e. growth rate was negatively related to log density), and hence further insight in the mechanism underlying the relationship between badger and hedgehog numbers. When the relationship between log mean density and the rate of hedgehog population growth was investigated separately in control and proactive treatments, there was some suggestion that density dependent growth rates varied according to the magnitude of predation. In control areas, where badgers were unmanipulated, population growth rate appeared to decline rapidly with increasing hedgehog density. The steepness of this slope suggests over-compensating density dependence (i.e. the population is liable to overshoot the equilibrium). In areas where badgers were culled, growth rate appeared to decline less steeply which suggests under-compensating density dependence. These data indicate the existence of two stable states of abundance according to the magnitude of predation. In the control treatment, a lower equilibrium population size (c. 0.5 hedgehogs ha-1) indicated regulation by predators, whereas in areas where badgers were culled, an equilibrium population size more than three times the density was observed (c. 1.8 hedgehogs ha-1). These density dependent relationships fit with type III predation where predators regulate low density prey populations occupying habitat refugia (Sinclair & Krebs 2003). This work has therefore provided evidence that badger predation regulated hedgehog populations in a density dependent rather than depensatory way. In other words, at low hedgehog densities, the per capita predation rate is low but increases as hedgehog density increases, thus regulating the population.

It also important to point out that throughout the duration of the study, very few pasture fields were found to support hedgehog populations and there was no evidence that hedgehogs began to colonise these habitats in response to badger culling. However, badgers were still active in agricultural habitats of proactive areas after culling had taken place, as was observed in the index of badger abundance based on sightings during nocturnal hedgehog surveys. Furthermore, daytime surveys of badger sign revealed that on average badgers were active in 26% of fields in proactive areas after culling, compared to 36% in control areas. This is probably a reflection gaps in landowner participation in the RBCT, and trapping efficiency being less than 100%. As the movements of hedgehogs are likely to be influenced by olfactory cues (Ward et al. 1997), the presence of residual badger activity may have deterred them from colonising agricultural habitats.

SID 5 (2/05) Page 30 of 36

Badgers achieve high population densities in their preferred agricultural habitats, particularly in southern and midlands England (Macdonald & Newman 2002; Rogers et al. 1997). Theory predicts that predators which rely on a persistent primary prey species exert a depensatory (inversely density dependent) effect on the secondary prey population (Sinclair 1989). As one of the factors that determines badger population density is the availability of their primary prey species (i.e. invertebrates), badgers at high density in agricultural habitats are likely to have depensatory effects on hedgehog populations occurring at a range of densities. Badgers are therefore capable of driving hedgehogs to local extinction in these habitats, which are likely to only persist in isolated pockets where badgers occur at a very low density (Micol et al. 1994). The experimental aspect of this study has provided evidence that in suburban micro-habitats, in rural regions of southern and midlands England, badger predation determined the rate of population growth of hedgehogs and regulated their abundance. It also demonstrated the great value of manipulative experiment such as this for identifying factors that determine population growth rate.

Knowledge of the relationship between hedgehog population growth rate and the index of badger abundance could inform the management of hedgehog populations. If hedgehogs were of management concern, badger abundance could be manipulated (e.g. through habitat management) which would lead to a predictable increase or decrease in hedgehog density. Furthermore, knowledge of this relationship provides a tool with which to predict the future dynamics of hedgehog populations according to natural changes in badger density. However, ideally more research is required to investigate the potential existence of alternative stable states in hedgehog populations where their density sufficiently increases to permanently escape predator regulation. One way to do this would be to continue to monitor hedgehog populations after the treatment is reversed, i.e. monitor the response of hedgehog populations as badgers recolonise proactive areas.

4.4.2. Effect of badger on habitat use and ranging behaviour

Hedgehogs exhibited non-random use of habitats at all three spatial scales and as expected the preferred habitat types were amenity grassland and suburban. These habitats are most profitable as they provide both higher food availability and a lower risk of predation than agricultural habitats (Micol et al. 1994). Within the home range and at the core activity area scale, amenity grassland was significantly preferred over all other habitat types, including suburban, and this therefore appears to represent the primary habitat of hedgehogs in the study areas. As pasture grassland habitat is used very frequently by badgers for foraging for earthworms, particularly in spring and autumn (Neal & Cheeseman 1996), they represent a high level of predation risk. Pasture was the least favoured habitat for hedgehogs. There were no treatment effects of badger removal on habitat use by hedgehogs. Nevertheless, this experiment does provide some evidence that the proportions and areas of habitats used by hedgehogs in their core activity areas and home ranges changed in response to the removal of badgers. There was evidence that the area of amenity grassland in the core activity areas of hedgehogs declined in response to the reduction of badger numbers. As the risk of predation declined after the badger cull, hedgehogs appeared to have used the primary habitat less and secondary habitats more. For example, pasture habitat comprised a higher proportion of core activity areas in the proactive areas after the cull whereas it declined in control areas. This provides some support for the a priori prediction that preferred rural habitats would be used more frequently after a reduction in badger abundance compared to controls. Pastoral grassland is likely to be a profitable habitat for foraging hedgehogs if there is a sufficiently low risk of predation and previous studies observed a preference for pasture grassland habitats in low badger density areas. It is noteworthy that there was no effect of badger culling on the proportion of pasture grassland within the home range. This suggests that hedgehogs did not make large shifts in their home range towards rural habitats in response to badger culling, but may have changed the way they used habitats within the home range. There were no other treatment effects detected on the proportion of individual habitat types within core activity areas or home ranges. Although the raw data suggested that the rural habitats in total represented a greater proportion of hedgehogs’ core activity areas in proactive areas after the badger removal operation this was not found to be statistically significant. In summary, the results of this field experiment provide some support for the predictions of IGP that predation risk is an important factor in habitat selection by hedgehogs and that at high badger density risky but productive habitats were not selected for foraging. When badger density was reduced, these previously risky habitats comprised a larger proportion of the core activity areas of hedgehogs. However, evidence for a shift in habitat use in response to the predator removal experiment was not conclusive. Although hedgehogs foraged further from cover in proactive areas after culling, this also occurred in control areas and thus there was no effect of the interaction of treatment and time. Therefore, there was no evidence that hedgehogs foraged further from or closer to cover in response to varying predation risk from badgers.

Three of 26 animals in this study were observed to disperse permanently away from their home range. Two other individuals were observed to move away from their home range but returned after a period of approximately two weeks, which was assumed to be a failed attempt to disperse. Of the five individuals recorded to disperse, four were males and one was female which despite being a small sample, is consistent with the tendency of male biased dispersal in mammals. This number of observations of dispersal events during the study was insufficient to investigate the effect of badgers on the rates and scale of hedgehog dispersal. This is unfortunate, as understanding the effects of predation risk on dispersal for species in fragmented populations is very important as

SID 5 (2/05) Page 31 of 36

their persistence relies on the ability of individual animals to move between sub-populations that occupy favourable habitat patches.

In conclusion, this study suggested that predation risk does not have a strong effect on patterns of movement during foraging and the size of home range and core activity areas of hedgehogs. Therefore the lack of observed, local anti-predator behaviours in the present study, despite a high risk of predation, suggests that hedgehogs are likely to be very susceptible to predation during foraging. Taken together with the survey results, this evidence suggests that hedgehogs respond to badgers through large scale movements away from areas of high predation risk but do not demonstrate strong predator avoidance behaviour when moving within their home range.

4.5. Ground nesting birds

4.5.1. Review and meta-analysis

The literature reviewed here suggests that the national impact of badgers on bird populations is likely to be low in the UK. In addition many studies including the percentage frequency of occurrences calculated here include remains from scavenging in addition to predation. Skoog (1970) estimated that 50% of bird remains found in badger stomachs were from adult birds, and other authors agree (e.g. Leitch & Kruuk 1986; Madsen et al. 2002) have suggested that this often results from consumption of carrion. The wide range of bird species identified in the present review also suggests that scavenging almost certainly plays a contributory role. It is unlikely for example, that badgers would actively predate healthy adult swallows Hirundo rustica.

The analysis identified a trend of increasing frequency of occurrence of birds in badger diet further north. Greater variation in the range of weather conditions in northern latitudes could influence the availability of food resources to badgers at certain times of the year. For example, as earthworms are the single most important food item for badgers across much of England (e.g. Neal & Cheeseman, 1996; Roper 1994), when they are unavailable in winter and dry summers, badgers may target alternative foods including birds. The increased badger predation of corn buntings during a period of drought is consistent with this view (Brickle et al. 2000). During the spring and summer the availability of bird nests and fledglings increases, whereas during the winter it is likely that bird carrion would be most abundant (Leitch & Kruuk 1986). In southern latitudes however, fruits and some invertebrate prey would be available throughout most of the year, possibly negating the need to switch to alternative prey such as birds. Nevertheless, we accept that this is speculation, and other factors such as geographic variations in the biomass of available bird prey, the proportion of ground nesting species and intra-specific variations in productivity, may also play a role.

Badgers have a wide diet and will opportunistically exploit ephemerally abundant food resources. This explains why under certain circumstances they may have the potential to have a local impact on bird nest survival. In addition, badgers will switch their attention between different food sources when conditions dictate (e.g. Neal & Cheeseman 1996; Garnett et al. 2002). Leitch & Kruuk (1986) estimated that birds constituted 1.4% of the total annual biomass (93 – 155 kg) taken each year by UK badgers and concluded that spread over many species the impact on the avifauna would be insignificant. This also assumed that badgers had killed all the birds they ate, which is highly unlikely. Nevertheless even if birds only form a small part of the badger’s diet, an increase in badger abundance would lead to an increase in predation pressure on birds, and vice versa. However, from the reviewed evidence presented here it seems unlikely that badgers have had a significant role to play in the recent national decline in ground nesting bird populations in the UK.

4.5.2. Effect of badgers on ground nesting bird abundance

The figures presented here are consistent with skylark densities reported in the literature, which have a national range of 0.62 km-2 - 30.61 km-2 (Browne et al. 2000). Grazed pasture varies between 3.75 - 5 km-2 while the mean for moorland is 12.95 km-2. However reported densities for meadow pipits are highly variable, and the densities seen in this study are lower than in some (e.g. Tharme et al. 2002), but similar to others (e.g. Fuller et al. 2002).

From 2000 to 2004 there was a marked decrease in the abundance of both skylarks and meadow pipits in the control treatment areas, with the steepest decline 18 months after the initial cull. In contrast, populations in the proactive areas remained fairly constant over the four years of the experiment. However, only the change in meadow pipit abundance was statistically significant in relation to badger culling. The variation in abundance of both species is likely to be controlled by differences in habitat and livestock density (See section 2.3.3.1). The detection of any effect of badger removal is dependent on the explanatory power of these covariates. It is possible that in the case of meadow pipits the LCM2000 habitat data and DEFRA livestock data explained a high proportion of the variance and enabled the effects of badger removal to be quantified. The variance surrounding the abundance estimates of skylarks however may not have been sufficiently explained by these covariates to

SID 5 (2/05) Page 32 of 36

enable the impact of badger removal to be assessed. A theory supported by the fact that removing moorland squares made very little difference to the meadow pipit model, but large differences for skylarks.

It is apparent that populations of both species stayed relatively stable in proactive areas and fell in control areas. This pattern is rather difficult to interpret in the context of this particular experiment. If badgers do limit the population size of skylarks and meadow pipits through predation of nests and young, then we may have anticipated that their removal should have increased breeding abundance in proactive areas, while populations in control areas remained constant. It is possible that factors at a larger scale could have affected these results. As mentioned above, both the proactive treatments contain significant areas of moorland or marginal upland, and such areas naturally support higher densities of both skylarks and meadow pipits compared to lowland pastoral habitats (Browne et al. 2000; Tharme et al. 2001). One explanation for this is that these areas represent the best breeding habitats for both species in the region. If the regional population underwent a reduction in size, as a result of factors other than a badger removal operation, such as low over-winter survival, or habitat change, it may cause a contraction in the range of the species. This would result in the decline in abundance if not abandonment, of the less favoured lowland pastoral habitats, as has been shown here. Examples of this type of range contraction can be found in other studies such as Wilcove & Terborgh (1984) and Gates & Donald (2000). These studies demonstrated a contraction in the range of certain species through the abandonment of peripheral areas, while population densities in the core areas remained unchanged. It is also possible that the removal of a predator such as the badger further increased the suitability of these core areas.

An alternative explanation is that the removal of badgers acted as a buffer from regional conditions, such as poor over winter survival, and thus enabled the maintenance of numbers while in the control areas overall numbers declined. This however is considered unlikely, as over winter survival tends to dictate subsequent breeding numbers, irrespective of the density of birds entering the winter (Perrins & Geer 1980; McCleery & Perrins 1991; Newton et al. 1997). Higher nest success as a result of badger removal may have increased pre-winter abundance but is unlikely to have affected post-winter numbers.

4.5.3. The effect of badgers on the rate and source of nest predation

We accept that in the present study daily survival rates should have been measured before and after the cull, and not just afterwards. It was however, unfeasible to carry out daily visits to all nests, in all squares, in all three treatment areas during the pre-cull period. As a result, the six-week checks failed to demonstrate any difference between treatments over time, owing to high failure rates in both treatments (97.5%). The post-cull results do however show a significantly higher predation rate of nests in the control areas, when measured over 14 days, and suggest similar patterns when measured over four days. This may indicate that badger removal operations lessen the predation pressure on ground nests. We consider this a surprising result given that we demonstrated that bird remains were relatively uncommon in badger diet, and that we would have anticipated compensatory predation by species such as hedgehogs and foxes, which increased in numbers in response to badger removal. Unfortunately we cannot conclude that the lower predation rate is due to badger removal as daily survival data were only gathered post-cull, and the difference shown may be a result of different habitats and predator assemblages. One of the major differences between upland and lowland in the two triplets is the density of livestock (pers. obs.) with more intensive grazing systems operating on lowland sites. Livestock have been shown to adversely effect ground nesting birds either by the trampling of nests (Beintema & Muskens 1987) disturbance of adults (Hart et al. 2002) or by maintaining an unsuitable sward length through grazing (Beintema & Muskens 1987). In addition to the number of livestock, lowland pastoral systems tend to be more intensively farmed than upland enterprises, with greater amounts of fertilizer, herbicide and greater production of silage (Shrubb 2003). These are all factors that have been suggested as contributory factors in the decline of farmland birds, and may also be the reason why skylarks and meadow pipits are maintained in greater numbers on upland sites. Nest failure rates in the present study certainly seemed higher in grazed fields than non-grazed fields (T. Hounsome, pers. comm.). The effect of livestock on ground nesting birds is explored further in Section 3.3.

The source of failure for a nest is fraught with problems, and inherent bias. The only reliable method for the identification of the source of failure is the use of video cameras, but deployment of sufficient video cameras to enable robust statistical analysis was beyond the budget and manpower availability of this study. The allocation of failure source presented here is almost certainly overly conservative, as can be seen by the high proportions that were recorded as ‘un-sure’. Analysis did however show a significant difference between sources of failure in different treatments and over time. Although there was not a significant difference in the frequency of failure sources when just post-cull data were analysed. One explanation for this difference is the allocation of the source of failure. There was a gradual awareness of the potential impact of livestock on failure of artificial nests during 2000, and by 2002 the significance of their impact was known and therefore there may have been a tendency to miss livestock caused failures pre-cull. This in itself underlines the subjective nature of allocating predation source without the use of video evidence.

4.5.4. Small scale experiment

SID 5 (2/05) Page 33 of 36

The results of the experiment showed that artificial nests in all treatments experienced a high level of failure (91.9%), and the presence of livestock played a significant role. Furthermore, the effect of livestock was related to their direct destruction of nests either by trampling or by licking or eating eggs.

The result of the first experiment showed that there was a greater chance of artificial nest failure in the grazed as opposed to the un-grazed treatments, However, the second experiment confirmed that nest failure rates fell significantly as soon as the cattle were removed from the original plot while, at least initially, the sward length would have been the same. This suggests that direct effects from cattle were to blame for the inflated failure rates, and not the indirect effects of the grazed habitat. The result of the third experiment supports this, as the survival times for nests in un-grazed and recently grazed plots were not significantly different which suggests a similar background level of natural predation. It is notable that although the study was carried out in an area of high badger density (>20 km-²), destruction of nests by livestock appeared to be much more frequent than predation by badgers, which was recorded only once. Adjusting the stocking density resulted in an increase in nest failure rates with increased cattle density. Many farmers operate a rotational grazing regime and this often means whole herds are grazed on a single field, and then rotated to the next. This can increase grazing densities to at least those tested here (13-14 ha-1) and in some cases more than ten times higher (200 ha -1 is frequently observed. This small scale study therefore gives us an indication of the importance of changes in livestock stocking practices and density.

4.5.5. Livestock effects on ground nesting bird abundance

Increasing cattle density in the 5 x 5 km squares in and around the RBCT areas was shown to have a significant negative effect on the abundance of both skylarks and meadow pipits. There was not however, any significant relationship with sheep density. The results of the national analysis however, show that the stocking density of both cattle and sheep has a significant negative effect on the abundance of both skylarks and meadow pipits, although the suggested impact of cattle is far greater than that of sheep. The general similarity between the results at the scale of the RBCT and the national scale lends confidence to the assertion that the observed effects are genuine.

4.5.6. Livestock effect on nest survival

Evidence from previous studies has shown that livestock can have a negative effect on nest survival (e.g. Chamberlain & Crick 2003). It seems possible therefore that one of the underlying mechanisms that produces the negative relationship between livestock density and abundance shown described above could be due to lower nest survival in areas of high livestock density. The results of the regional scale analysis in the RBCT framework reflect the findings of the local scale experiment, with artificial nests on grazed land having significantly lower survival rates than those on non-grazed land such as arable and set-aside. Furthermore there was a significant difference in the survival rates of nests on grazed land in the presence of livestock and those on grazed land where livestock were absent, similar to that demonstrated in the small scale experiment. The lack of a significant difference in survival rates of nests on grazed land with livestock absent and those on non-grazed land once again suggests a significant direct impact of livestock themselves, rather than of the grazed habitat and associated increased predation risk.

In contrast to the results from the RBCT areas, analysis of national nest survival data did not demonstrate an association with stocking levels. There are a number of reasons this may have occurred. We have demonstrated that the presence of livestock in the same field significantly affects the probability of a nest surviving. Unfortunately the analysis of national nest data only included estimates of stocking density at a regional scale using data from 5 x 5 km squares. It is therefore possible that this lack of resolution influenced the results.

Chamberlain & Crick (2003) noted that nests in grazed grass had a greater chance of failure when compared to nests in un-grazed grass, a finding similar to that of the present study in the RBCT areas, although they cited predation as the main cause of failure, an assertion that is not supported here. High livestock density may also reduce the number of suitable nesting sites and as such may limit a population in a region, not by direct predation/trampling but by habitat modification (Chamberlain & Crick 2003). It seems possible therefore that the non-significant result presented here for the NRS data may be a result of both the resolution of the data, and the performance of some ground nesters in the presence of livestock.

4.6. Food web modelling

The modelling study predicts that on the basis of trophic interactions alone, even the complete removal of badgers results in a less than 5% change in the majority of species and groups within the UK terrestrial food web. A small number of species were predicted to change in population size given sufficient time and extent of badger

SID 5 (2/05) Page 34 of 36

population reduction: stoats, 5.12%; foxes, 7.06%; the ‘Medium’ birds group, 13.15%; moles, 34.59%; and the earthworms, Lumbricus terrestris, 17.28%, and Allolobophora longa, 17.28%. Clearly the empirical surveys we undertook did not cover all these species, although it is interesting to note that foxes were one of the species where an increase in population size was predicted. In fact, our surveys estimated a larger increase in fox populations in a shorter time than predicted here, which lends support to the theory that the source of competition between foxes and badgers which led to this result was not food resource related. Hedgehogs were also predicted to increase on the basis of the release of prey (in this case, invertebrates) due to badger removal. This agrees with our empirical results in terms of the direction of change. However, we saw a much more marked increase in hedgehog numbers than predicted here, almost certainly due to the reduction in predation pressure precipitated by badger removal. This illustrates the limitations of the approach of only taking into account trophic interactions in the modelling process. The extensive data that we now hold, on a variety of key species means that in the future it should be possible to take a more complete modelling approach to predicting the likely impacts of badger removal.

References to published material9. This section should be used to record links (hypertext links where possible) or references to other

published material generated by, or relating to this project.

SID 5 (2/05) Page 35 of 36

Peer reviewed papers

Hounsome, T.D. & Delahay, R.J. (2005). Birds in the diet of the Eurasian badger Meles meles): a review and meta-analysis. Mammal Review 35: 199 – 209.

Young, R.P., Davison, J. Trewby, I.D. Wilson, G.J., Delahay R.J. & Doncaster C.P. (2006). Spatial variations in the abundance of hedgehogs Erinaceus europaeus in relation to indices of badger Meles meles density and distribution. Journal of Zoology, 269, 349-356

Trewby, I.D., Wilson, G.J., Delahay, R.J., Walker, N., Young, R.P., Davison, J. Cheeseman, C.L., Robertson, P., Gorman, M. & Mcdonald, R.A. (In review) Experimental evidence of competitive release in sympatric carnivores: badger removal increases fox densities. Biology Letters.

Davison, J. Young, R.P..Hounsome, T.D Trewby, I.D. Claridge, M. Yarnell, R.W. Walker, N. Henrys P.A.& Wilson G.J. (In prep) A comparison of distance sampling and scat counts for monitoring fox populations.

PhD Theses

Hounsome, T. D. (2005) The Effects of Badgers (Meles meles) and Livestock on Ground Nesting Birds. PhD. Thesis. University of Aberdeen.

Young, R. P. (2005) Field test of impacts of intraguild predation on a mammalian prey population: the badger Meles meles and hedgehog Erinaceus europaeus. PhD. Thesis. University of Southampton.

Trewby, I. T. (2007) The effects of badger Meles meles removal on fox Vulpes vulpes populations and fox feeding habits. PhD. Thesis. University of Aberdeen.

SID 5 (2/05) Page 36 of 36