General enquiries on this form should be made...

General Enquiries on the form should be made to:Defra, Strategic Evidence and AnalysisE-mail: [email protected]

Evidence Project Final Report

EVID4 Evidence Project Final Report (Rev. 10/14) of 59

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 Evidence 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 websiteAn Evidence Project Final Report must be completed for all projects.

This form is in Word format and the boxes may be expanded, 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 WM0323

2. Project title

Refining approaches to the surveillance of wild boar presence, abundance and environmental impact

3. Contractororganisation(s)

National Wildlife Management Centre

Animal and Plant Health Agency (APHA)

54. Total Defra project costs £ 280,263(agreed fixed price)

5. Project: start date................ 07/02/2011

end date................. 30/06/2014


6. It is Defra’s intention to publish this form. Please confirm your agreement to do so.................................................................................YES x NO (a) When preparing Evidence Project Final Reports 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 Evidence Project Final Report 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.

1. A number of separate free-living and self-sustaining wild boar populations have become established in England as a result of farm escapes and illegal releases. These have the potential to grow and spread and there is a need for methods to estimate presence, local densities and the environmental impact of wild boar to support future management.

2. This study builds on the results of project WM0318 which produced initial tools to estimate local densities and impact of wild boar. The objectives of the current project were to refine these surveillance methods to monitor the presence and density of wild boar; to assess impact of wild boar on key components of biodiversity; to explore the relationships between density and environmental impact derived from rooting and to produce best-practice principles and protocols for future monitoring.3. The work was carried out in five study sites, three in Gloucestershire (Penyard/Chase in Ross-on-Wye, Serridge in the North of the Forest of Dean and Oakenhill in the South of the Forest of Dean) and two in East Sussex (Beckley/Bixley and Brede), previously surveyed under project WM0318. Serridge and Oakenhill are part of the main woodland complex of the Forest of Dean

4. The first objective was to refine methods for detecting wild boar presence to help assess range expansion. Five methods were developed or refined to detect wild boar presence in an area and/or to estimate the relative effort to monitor the species’ range expansion: (i) large-scale mapping of wild boar sightings, (ii) bait stations with camera traps, (iii) camera grids and activity signs on transects, (iv) use of attractants; and (v) modelling the effort required to detect wild boar at low density. Large-scale mapping of sightings showed that wild boar range increased between 2004 and 2014 and suggested this method is useful to monitor the long-term expansion of the species at the national level. Bait station with camera traps in 20 woods around the Forest of Dean indicated that no range expansion had occurred in the last three years and showed that this method could be used at a local scale to monitor the rate of expansion. Camera grids and activity signs on transects, based on current densities of wild boar, showed that a minimum of 2-4 camera traps/100 ha, left for 9 days, or a minimum of 1-7 transects/100 ha were required to have a > 90%


probability of detecting wild boar presence in relatively small woods (200-400 hectares). A putative site attractant, based on birch wood tar, was found effective in modifying wild boar behaviour: a pilot trial with the attractant applied to stakes confirmed that wild boar spent more time rubbing against treated than non-treated stakes. When the trial was applied to trees, this behavioural change was confirmed: wild boar left signs on trees treated with this attractant more often than on control trees. These findings suggested that this compound, used in conjunction with bait stations and camera traps is a reliable method to confirm presence of wild boar in an area. The modelling of the effort required to detect a single wild boar in a large woodland (55 km2) indicated that circa 15 camera traps/100 ha should be deployed for 10 days to have a 90% probability of detecting a single animal.

5. The second objective was to refine methods for quantifying wild boar density. Five methods were developed or refined to assess wild boar population trends or density: (i) Passive Activity Index (PAI) based on camera trap grids and activity signs on transects, (ii) density estimates based on camera trap grids, (iii) distance sampling through thermal imaging, (iv) spatially-explicit individual-based model simulations to investigate the accuracy and precision of virtual camera traps and virtual distance sampling in estimating pre-determined wild boar densities and (v) monitoring of road traffic accidents (RTA). Trends in PAIs calculated for each study site in winter 2011-2012 and 2012-2013, based on both camera traps and activity signs on transects indicated that wild boar populations were stable or increasing in numbers. In most instances, no significant differences were found in PAIs between years, due to the wide variation surrounding these estimates. Density based on camera traps in winter 2012-2013 was estimated as 4.5-7.0 wild boar/100 ha per site. Trends in densities based on camera grids suggested populations were stable in one study site and increasing in all the others. The distance sampling method using thermal imaging was calibrated in an Italian area with a high density of 15-41 wild boar/100 ha and then used in the Forest of Dean in winter 2012-2013. The resulting density was 8.7 wild boar/100 ha. The computer simulation model of densities was based on pre-determined numbers of wild boar groups, variable between 50 and 200: virtual camera traps or virtual distance sampling were used to estimate these known densities of wild boar. The model found that both camera traps and distance sampling estimates may underestimate the known density of wild boar by 18-30%. Thus the maximum densities measured in the field using camera traps and distance sampling might have been underestimated by 18-30%. The model also suggested that camera trap estimates are relatively precise and not affected by population size, although highly sensitive to group size: if the estimate of average group size is accurate, then the population size estimate can also be accurate. Density estimates based on distance sampling have wider confidence intervals but do not change with group size or population size. We concluded that both camera trap grids and distance sampling can be used to assess wild boar density. Between 2009 and 2013, the number of vehicle collisions with wild boar in the Forest of Dean increased in parallel with the increase in density recorded through PAIs or density estimates since 2008. Howeever, traffic flow did not change. This suggested that the number of RTAs could be employed as an indicator of wild boar population trends. 6. The third objective was to assess the large-scale impact of wild boar rooting in the five study sites and the local scale impact of wild boar on plant and invertebrate species numbers. Large-scale impact was studied by quantifying rooting activity on permanent plots twice a year for two years. The results indicated that fresh rooting by wild boar in English woodlands, at current local densities, affected between <1 –3%, and exceptionally ~8% of a wood, depending on season and year. The current study confirmed that wild boar preferred to root in broadleaves stands rather than in other habitat types. In sites where the wild boar density has increased in recent years, the proportion of woodland rooted increased from < 1% recorded in winter 2009-2010 during project WM0318 to 2.1-4.4% recorded in winter 2012-2013. The percentage of woodland where rooting occurred in the five English study sites surveyed for this project was relatively small compared to that recorded in studies conducted in other European countries (e.g. 4% in Poland, 12% in Sweden and between 27 and 57% in Switzerland). The local scale impact


was derived by assessing the number of plant species and insects caught in pan traps against the amount of rooting recorded along five independent transects. Possibly due to the relative low level of rooting recorded, no significant correlation was found between the numbers of flower-visiting insects and the total numbers of plant species sampled, the numbers and proportion of species in flower and the amount of rooting. This highlighted the complexity of assessing impact of wild boar on plant and invertebrate species and reflected the different spatial and temporal scales at which the insect traps operated compared with the distribution pattern of rooting. Whilst there seem to be no obvious impact on plant and invertebrates at current wild boar densities, this might change if wild boar local densities increase

7. The fourth objective was to explore relationships between wild boar density and environmental impacts. No correlation was found between the density of wild boar calculated from camera traps and the proportion of plots rooted, suggesting the impact due to rooting is unlikely to be related to density in any simple way, and is probably complicated by the effects of other factors such as precipitation and availability of different food sources.

8. The fifth objective was to provide best practice principles for the methods developed during this project. The resulting operational document contains a list of methods and instructions to assess large-scale impact of wild boar and to monitor presence, density and population trends and the relative effort required to implement these methods. This will form the core of a best-practice document that Defra will provide to stakeholders to facilitate the regional management of wild boar.

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 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 Exchange).

Introduction and Policy Rationale

Defra has the responsibility to facilitate the regional management of wild boar by providing local communities with advice on methods to control human-wild boar conflicts. The control of wildlife populations requires information on their abundance and distribution so that control efforts can be deployed knowledgeably and efficiently. The approach used will be determined by the information required (e.g. estimates of presence, relative abundance, density or absolute abundance) and by the resources available (Mayle et al. 1999). Defra and the Forestry Commission also have the responsibility to meet the British Government’s objectives for biodiversity and sustainable forest management under UN Resolution 65/161 Decade of Biodiversity Resolution, and Resolution 61/193 the International Year of Forests, 2011. In continental Europe wild boar populations are associated with significant human-wildlife conflicts such as damage to crops, reductions in the abundance of plant and animal species, spread of diseases, damage to livestock production and vehicle collisions (Apollonio et al. 2010). Trends in wild boar population increase have been observed consistently throughout Europe, even in countries characterised by harsher winters than England (Massei et al. submitted). In the UK recently established wild boar populations are still localised and are likely to be in the initial phase


of population increase and expansion (Holland et al. 2007). Current numbers of wild boar in England are low, but they have the potential for rapid increase. Management of this expanding population and its impacts requires methods to assess wild boar presence, abundance and environmental impact.Worldwide, several methods have been used to monitor the economic and environmental impact of wild boar (Massei et al. 2011). In project WM0318 we developed methods for the detection and quantification of wild boar presence, range expansion and abundance as well as methods to measure both the large-scale impact of wild boar and the impact of this species on specific plant and invertebrate communities within woodlands. The surveillance component of this earlier project identified a suite of candidate methods and developed a cost-effective staged approach to apply these methods. However, the generally low densities at which wild boar occurred in England limited the implementation of some of these methods. The current study builds on these results to refine and expand the tools to monitor growth, expansion and impacts of wild boar.

Objectives 1. Refine cost-effective methods for detecting wild boar and quantifying range expansion. 2. Refine cost-effective methods to quantify wild boar density and abundance.3. Quantify impacts of wild boar on key biodiversity components of the environment.4. Explore relationships between wild boar density/site usage and environmental impacts.5. Produce best-practice principles on the field deployment of methods developed during this project and on the analysis and interpretation of the resultant data.

This project was carried out as collaboration between the National Wildlife Management Centre (Animal and Plant Health Agency, APHA, formerly Food and Environment Research Agency) and Forest Research. Forest Research was responsible for work concerning distance sampling and thermal imaging (Objective 2, Section 2c) and for the study on quantification of wild boar impact on ground flora and on pollinator communities (Objective 3, Sections 3a and 3b).

Study areaThe study was carried out at five sites, four previously used in project WM0318, where wild boar populations are well-established: Beckley/Bixley and Brede High Woods in East Sussex (the Sussex Weald), Penyard and Chase Woods (Ross-on-Wye), Serridge (north Forest of Dean) and Oakenhill (south Forest of Dean). As in one of the sites originally used (Ruardean, north Forest of Dean) camera traps were stolen on a few occasions, the site was moved to the neighbouring area called Serridge. A summary of the fieldwork carried out for this study is provided in Appendix 1.

Objective 1. Refine cost-effective methods for detecting wild boar and quantifying range expansion. Project WM0318 developed a staged approach for detecting wild boar presence in an area where the species had not been recorded previously. This approach proceeded from initial anecdotal evidence of wild boar in an area, derived from road traffic accidents or sightings, through to confirmation of the species presence and assessment of relative local density of wild boar. Nevertheless, the project did not determine the optimum effort, in terms of number of camera traps or transects, required to detect wild boar presence with a known level of confidence, in a new area.The first aim was to update the wild boar distribution in England based on sightings. The second aim was to repeat the method based on single bait stations and camera traps in woods surrounding the Forest of Dean to assess wild boar spread and to examine the relationship between increasing survey effort and likelihood of detecting wild boar presence. This relationship was used to determine the effort required to be 90% or 95% certain that wild boar occurred in a wood. As newly colonised areas tend to have relatively low densities of animals, we also established the minimum effort required to confirm wild boar presence in a large wood where the species occurs at extremely low density.


MethodsCompared to project WM0318, when only baited camera traps were used to detect wild boar presence in woodlands at the edge of the Forest of Dean, we doubled the effort to detect wild boar presence in woods around the Forest of Dean by using baited camera trap stations as well as activity signs on transects and attractants. Data from camera grids and from activity signs on transects were used to quantify the minimum effort required to detect the presence of wild boar in areas with the different wild boar densities. In addition, we used computer simulations to determine the probability of detecting a single wild boar with an increasing density of camera traps placed in a woodland.

1a . Mapping of wild boar sightingsData on wild boar sightings were provided by the Wildlife Management and Licensing Service of Natural England (hereafter NE). Data were obtained by collating ad hoc reports from the public, other agencies and the media, including sightings, reports of damage or rooting activity and reports of illegal releases and escapes. The location of each report was recorded to the nearest 1km UK Ordnance Survey national grid square and data were presented as number of 5 x 5 km squares where wild boar presence had been recorded. Data collected from 1980 to 2004, as part of a government consultation on management of feral wild boar in England (Defra 2005) were compared to data collected in the last decade, up to March 2014.

1b. Bait stations with camera traps to detect wild boar presence.Bait stations with camera traps were set up in the same 20 woods surrounding the core area of the Forest of Dean, initially surveyed in winter 2010 under project WM0318, (Fig. 1). The original protocol stated that the 20 woods would be re-surveyed every winter, in November-December 2011, 2012 and 2013, by placing one bait station and two camera traps in each wood for two weeks. To maximise the likelihood of detecting wild boar presence, bait stations were placed either on sites that were most likely to be visited by wild boar (such as mature oak or chestnut woods) or where wild boar had previously been recorded. The bait used was maize (circa 7-8 kg per bait station), replaced after one week. When placing camera traps in each wood, 1-2 hours were spent walking on tracks in the wood and recording ad hoc wild boar activity signs (rooting or trails). At the end of week 1 and week 2, the amount of maize eaten at each bait station was recorded as well as the number of non-target species consuming maize. In winter 2010, wild boar activity signs or photos had been recorded in just four of the 20 woods surveyed. In November-December 2011 activity signs were again found in four woods, and camera-traps recorded the presence of wild boar in two of these woods, where signs had also been found in December 2010. Because of the apparent lack of range expansion recorded between 2010 and 2011 the next survey was carried out two years later, in November 2013, sampling the same 20 woods and using the same methods described above. Cameras were removed after two weeks and the proportion of woods with ascertained presence of wild boar was calculated. Fisher’s exact test was used to evaluate between-year bait consumption by comparing the proportion of bait stations with > 90% bait consumed at the end of the two weeks.


Fig. 1 Map of the 20 woods (blue circles) surveyed for detecting wild boar expansion through baited camera traps. Eight woods (4 old ones in pink triangles and 4 new ones

in yellow squares) were selected for detecting wild boar presence through activity signs and site attractants.

1c. Camera trap grids and activity signs on transects to detect wild boar

presence.Camera trap grids and wild boar trails on transects were used in the five study sites in winter 2011-2012 and 2012-2013 to determine the minimum effort, in terms of number of transects to be surveyed and camera traps to be employed, to detect the presence of wild boar in an area.For each wood, forest rides and pre-existing tracks (hereafter referred to as forest tracks) were mapped using Ordnance SurveyTM MastermapTM data series and ArcMap 9.3 GIS software (ESRI, California). In each wood, 200 m x 1m transects were located along forest tracks to obtain a density of one transect every 10 ha of wood (equivalent of 10 transects /100 ha), resulting in 20-37 transects per wood. Transects were surveyed in November 2011 (winter 2011-2012) and in November 2012 (winter 2012-2013). The start point of each transect was randomly placed on forest tracks using Hawth’s Analysis Tools for ArcGIS. On each transect, the number of wild boar trails that crossed the transect and the number of areas with wild boar rooting were recorded.In parallel, motion-activated cameras (Reconyx HC Hyperfire 500, RECONYX, Inc. 3828 Creekside Lane Holmen, WI, US www.reconyx.com) were placed in each of the five study sites a grid pattern. In project WM0318, following Rowcliffe et al. (2008), cameras were placed at a density of one every 11-12 ha. In the current project, the density of cameras was doubled and cameras were placed approximately every 6 hectares (i.e. ~ 16 camera traps/ 100 ha). As a minimum of 250 camera nights per site, based on > 20 camera traps per site is recommended by the literature on density estimated using camera traps (Rovero and Marshall 2009), 30-47 evenly distributed camera traps were placed in each of the 180-280 ha study sites and left in situ for nine nights. Monitoring was carried out during January-March 2012 and 2013 in the five study sites. As 23% of the camera traps placed in Serridge in 2012 malfunctioned, the site was re-surveyed in April 2012. The method developed by Rowcliffe et al. (2008) assumes that, if the survey is completed within a few weeks, immigration and emigration in the studied population can be regarded as negligible and spatial behaviour is unlikely to change. Therefore, if the number of camera traps required to monitor a site exceeds the number available, camera traps can be moved at regular intervals to cover each site. The current project aimed at completing the winter surveys in all five sites between January and February of 2012 and 2013. Thus, starting from the northernmost part of each study site, groups of camera traps were left for nine days and then moved to the centre and the southernmost part of the site so that each survey was completed in 18-27 days. As fully randomized placement could result in cameras being positioned in areas of no visibility, cameras were positioned in areas of relatively higher visibility within 25 m of the grid points. The number of wild boar visits per camera per 9 days was then calculated for each site. One visit was defined as >1 photos of wild boar until there was a lapse of at least 10 minutes between consecutive photos: photos of wild boar taken > 10 minutes apart were counted as a new independent visit as


preliminary observations with ear-tagged animals indicated the same animals rarely return to the same area after 10 minutes. Bootstrapping with replacement was used to derive the probability of detecting wild boar presence in a wood in relation to the density of camera traps or transects per 100 ha. Bootstrapping was carried out by randomly selecting a set number of transects or camera traps (from the original transects and cameras data set for each site) and assessing whether wild boar had been detected in those. This process was reiterated 10,000 times and the probability of detection of wild boar was derived as the proportion of those iterations where wild boar had been detected.Following the results of this process, the site that required the greatest effort, in terms of number of transects/ 100 ha that had to be surveyed to give a > 90% probability of detecting wild boar presence, was chosen as the most conservative approach. This effort was then applied to calculate the number of transects required to detect the presence of wild boar in 8 new woods around the Forest of Dean, where it was assumed that wild boar density would be low.As these 8 woods were widely used by the public that in the past had interfered with the camera traps, only the method of activity signs on transects to detect wild boar presence was tested.The woods were selected as follows: 4 woods (varying between 68 and 148 ha in size) were chosen among the 20 woods mentioned above, 3 where wild boar presence had been confirmed through activity signs or baited camera traps (wood no. 8, 18 and 20) and 1 where wild boar had been sighted by forest rangers or dog walkers in previous years (wood no. 17). An additional four woods of similar size (41 to 157 ha each) were selected (wood no. 21, 22, 23 and 24), each located at least 2.5 km from the edge of the Forest of Dean (Fig. 1). In all the new woods, located on Forestry Commission land, wild boar had been sighted occasionally by Forestry Commission rangers.In these 8 woods, the same density of 200m transects used in each of the 5 study sites (one 200 m transect every ~10 ha) was mapped. The proportion of transects to detect wild boar presence with > 90% confidence was calculated for each wood using the detection probability for the site with the lowest density of wild boar. In November 2013, transects in the 8 woods were surveyed for signs of wild boar activity (rooting and trails).

1d. Simulation of effort to detect wild boar at low density As the detectability functions obtained in each study site could not be tested against known densities, we used a bespoke simulation model designed in R (v3.0.2) to determine the likelihood of detecting the presence of a single wild boar in a site where camera traps were deployed. A single animal represented the absolute minimum number of wild boar in an area. Cameras were placed randomly inside a virtual habitat representing the Forest of Dean (c. 55 km2) and camera density was varied between 1 and 28 per 100 ha. We chose to exceed the density of camera traps that had been actually used for the study in 1c to see whether increasing this density (to a reasonable extent, subjectively set at 28 camera traps /100 ha) significantly affected the probability of detecting a single wild boar in a large (5500 ha) wood. For each camera trap density tested, cameras were left in place for 10 days and the model was run 100 times. The “virtual” wild boar moved according to speed and activity rhythm recorded when radiotracking animals during project WM0408 and was placed in a random location of the study area at each simulation. At the end of the 100 simulations, the number of simulations where the single individual was detected at least once was used to obtain the probability of detection within 10 days of survey for each camera trap density.

1e. Attractants to detect wild boar presence in new areas Although the original experimental protocol had included using bait trails, this technique was discarded for the following reasons: 1. bait trails are most attractive to wild boar when the natural food is scarce, thus bait trails could be used only in some seasons; 2. the relatively high density of deer and badgers, recorded in this project as well in project WM0408, consistently feeding at bait stations and reducing the amount of bait available for wild boar; 3. the significant increase of maize price in recent years that would result in this method being relatively expensive to implement on a large scale.As an alternative to bait trails, a non-food putative attractant (hereafter referred to as “attractant”)


was tested as described below. In May 2013 a pilot trial was conducted to test whether a commercially available wild boar attractant (Buchenholzteer, based on birch wood tar, commercialised by Bush Wear, Stirling, Scotland) increased the likelihood that wild boar would visit the area. Eight sites were selected in the Forest of Dean and Penyard woods, were wild boar had been regularly observed. At each site, two locations were chosen, 200 m apart from each other: each location had a 1m x 6 x 6 cm stake planted in the ground for ~ 30 cm, ~ 4 kg maize placed in a plastic pipe tied to a tree about 2 m from the stake and one camera trap (Reconyx HC Hyperfire 500) placed at 1.20 m from the ground and overlooking the stake. At each site, one stake was treated with the attractant (a single brush stroke) and the other was brushed with water and served as control. The cameras and the stakes were removed after 14 days. As some locations were not visited, the trial was repeated in August 2013 and 12 new sites were used.For each site, the number of wild boar visits were recorded and assigned to one of the following behavioural categories: 1. “sniffing”, 2. “Scratching” and 3. “walking”. One visit was defined as >1 photos of wild boar until there was a lapse of at least 10 minutes between consecutive photos: photos of wild boar taken > 10 minutes later were counted as a new visit. “Sniffing” was defined as a wild boar extending its neck and snout within 20 cm from the stake; “scratching” was defined as a wild boar rubbing its body against the stake and “walking” was assigned to all the visits where sniffing or scratching had not been observed. Data from May and August 2013 were pooled for the analyses and a Chi-square test was used to test whether sniffing and scratching were directed more towards treated than control stakes.As this pilot test suggested the attractant was effective in modifying wild boar behaviour (see Results), this compound was used in the 8 woods around the Forest of Dean in December 2013 (described in 1c), once the survey of activity signs on transects in these woods had been completed, to test a potentially simple method to detect wild boar presence and quantify range expansion. In each of the 8 woods ten pairs of trees, 2-5 m away from each other, were selected. The distance between the closest pairs of trees within a wood varied between 120 and 300 m. For each pair, one tree was treated with the attractant and the other tree was sprayed with water. To confirm wild boar presence through camera traps in each wood, a single pair of trees per wood (randomly selected out of the 10 pairs treated with either the putative attractant or water) was also chosen: a bait station with maize in pipes and two camera traps were placed only next to this pair of trees but not in the proximity of the other nine pairs of trees in each wood. Maize at the trees with camera traps was replaced at the bait station after 7 days. Treated and control trees were examined 1, 2 and 4 weeks after treatment with the attractant, and the presence of wild boar hair, rooting around the tree or tusk marks was recorded. A Chi square test was used to test the effectiveness of the putative site attractant by comparing the number of treated and control trees with wild boar activity signs. Results1a . Mapping of wild boar sightingsData on wild boar sightings collected up to 2004 and at up to 2014 suggest that wild boar have spread in the last decade (Fig.2). For instance, the number of 5 x 5 km squares where wild boar was recorded in Kent/East Sussex rose from 7 in 2002 to 10 in 2010; in parallel, this number rose from 4 to 9 in West Dorset and from 5 to 8 in Gloucestershire (Wilson 2014).

1b. Bait stations with camera traps to detect wild boar presence.Out of the 20 woods surveyed around the Forest of Dean to detect wild boar presence through bait stations with camera traps, activity signs were found in two woods, and camera traps recorded the presence of wild boar in two other woods in November-December 2011 (Table 1). In November 2013, pictures of wild boar were recorded in the same two woods where they had been found in November-December 2011, activity signs were recorded in one wood where they had previously been recorded and in a new wood (Table 1). In 2011, a single wild boar visit was recorded in wood no. 6 after 12 days and two boar visits were recorded in wood no. 20 after 8 and 9 days. In 2013 a single wild boar visit was recorded in wood no. 6 after 9 days and three boar visits were recorded in wood no. 20 after 2, 3 and 13 days. Compared to the 2010surveys (project WM0318), wild boar presence in 2011 and 2013 was still recorded in 4 out of the 20 woods surveyed (Table 1).


Fig. 2. Distribution of reports of free-ranging wild boar in England from 1980 to 2004 (left) and to March 2014 (right). Left map: black dots indicate that animals were still present at the end of 2004, green dots show new releases/escapes since 2003, pale grey dots show areas where animals were believed to be no longer present in 2004 (Source :Defra 2005). Right map: black dots indicate wild boar were still present; light grey dots show animals no longer present and dark grey is ‘unknown’ (Source: courtesy of C. Wilson 2014).

In 2011 and in 2013 badgers and deer (fallow deer, roe deer, muntjac) were observed feeding on maize at 16 out of 20 and 17 out of 20 bait stations respectively. The proportion of bait eaten after two weeks differed between 2011 and 2013 (Fisher’s two-tailed probability test P= 0.0103): in 2011, maize had been completely consumed (<10% left) after 2 weeks in 15 out of 20 bait stations, whilst in 2013 maize had been completely consumed only in 6 out of 20 bait stations. This difference might be explained by the fact that 2013, unlike 2011, was characterised by widespread availability of acorns which would have been consumed by boar (J. Coats pers. obs.).

1c. Camera trap grids and activity signs on transects to detect wild boar presence.In total, 10 surveys (1 per site per year in 5 study sites for 2 years) were carried out. The detection functions established to estimate the survey effort (in terms of number of transects or camera traps/100 ha) required to detect wild boar presence in a site suggested that this effort varied between sites and years, reflecting the different densities of wild boar in each site and time (Fig.3 and Fig. 4). In 9 out of the 10 surveys, the number of transects/100 ha required to detect wild boar presence with > 90% confidence varied between 1 and 7 transects/100 ha. In one survey (Brede in 2011) the maximum number of transects surveyed per 100 ha (i.e. 10 transects/100 ha) resulted in an 88.6% probability of detecting wild boar. With the exception of Brede 2011, the number of transects to detect wild boar presence with > 95% confidence varied between 1 and 9 /100 ha.

_________________________________________________________

Site codeDistance from FoD edge (km)

Boar signs (BS) or photos (Pic) 2010 2011 2013

1. 15 No No No 2. 10.7 No No No3. 13 No No No


____________________________________________________________ Table 1. Woods surveyed in winter 2010 (project WM0318), 2011 and 2013 to assess range expansion of wild boar. * = wood used to assess wild boar presence through activity signs on transects and site attractants. ? = possible presence of wild boar from relatively old activity signs.

Fig. 3. Detectability functions for wild boar derived from activity signs (rooting and trails) recorded on transects against the density of transects surveyed in each study site in winter 2011-2012 and 2012-2013.

In total, the number of camera-nights (defined as number of camera traps x number of nights) per study site per survey varied between 270 and 378. The minimum number of camera traps/100 ha required to detect wild boar presence with >90% confidence varied between 2 and 9 (Fig.4). If the site that had the minimum density of wild boar recorded during the study (Oakenhill in 2011) was excluded, the minimum number of camera traps/100 ha required to detect wild boar presence with >90% confidence varied between 2 and 4. The minimum number of camera traps/100 ha required to detect wild boar presence with >95% confidence varied between 3 and 13 or between 3 and 5 if Oakenhill 2011 was excluded (Fig. 4).

Fig. 4. Detectability functions for wild boar derived from pictures recorded by camera traps against the density of camera traps used in each study site in winter 2011-2012 and 2012-2013.

For each of the 8 woods around the Forest of Dean, the number of transects, with a density of 10/100 ha, was calculated and reported on a map. Due to the results obtained in Brede in 2011, assuming a worst-case scenario (lowest density) we planned to survey all transects in these 8 woods. If the presence of wild boar was confirmed beyond doubt (fresh rooting, presence of boar prints) on at least one transect before all transects were surveyed within a wood, the surveyor moved to another wood. The results indicated that wild boar occurred in at least 7 of the 8 woods surveyed (Table 2), including all the new woods and three of the old woods, thus confirming the results obtained using single bait points and camera traps. All 8 woods were surveyed by the same operator in 3 days. In wood no. 17, very old rooting (at least 3 month old) was observed and the presence of wild boar could not be confirmed.

"Old" or Wood ID number N. N. Type of activity


"new" transects surveyed

transects with boar signs signs

Wild boar present?

New 21 2 2 Rooting and prints YesNew 22 2 2 Rooting and prints YesNew 23 9 3 Rooting YesNew 24 1 1 Rooting and prints YesOld 20 5 2 Rooting and prints YesOld 8 2 2 Rooting and prints YesOld 17 7 1 Very old rooting ?Old 18 5 1 Rooting Yes

Table 2. Wild boar presence detected by monitoring activity signs along 200m transects in 8 woods around the Forest of Dean. ”Old” refers to a wood already sampled during project WM0318 and “new” to a wood were wild boar presence had been reported by rangers.

1d. Simulation of effort to detect wild boar at low density The results of the simulation estimated the probability of detecting a single wild boar in a 55 km2 study site (equivalent to the area covered by the Forest of Dean) as a function of camera trap density (Fig. 5). Although some variability was observed around the detection probability curve, the model showed that about 15 camera traps/100 ha resulted in a 90% chance of detecting a single wild boar, whilst 19-20 camera traps/100 ha resulted in a detectability of 95%. For a 99% chance, the number of camera traps required increased to 28/100 ha.

Fig. 5. Probability of detecting the presence of a single wild boar in relation to the number of camera traps deployed in a 55 km2 (5500 ha) woodland.

1e. Attractants to detect wild boar presence in new areas The results of the trials in May and August 2013 indicated that 149 wild boar visits were recorded around 12 out of 20 control stakes and 15 out of 20 treated stakes. The wild boar behaviour of sniffing or scratching against stakes was directed significantly more towards treated than towards control stakes (χ2= 7.258, df=1, P =0.003) (Fig. 6).

Fig. 6. Number of wild boar visits at 20 sites where stakes were treated with an attractant (Treated)


or with water (Control) in May and in August 2013. The behaviour of wild boar during a visit, observed through camera traps, was classified as either walking past the stake or sniffing/ scratching against the stake.

The results of the study on attractant, carried out in December 2013 to detect wild boar presence in the woods surrounding the Forest of Dean confirmed that wild boar occurred in 7 out of 8 woods. The only wood where wild boar presence could not be confirmed was wood n. 17, that had signs of old rooting and that was discarded from further analyses. At the end of the trial, 4 weeks after the attractant had been applied to the trees, 33 treated trees out of 63 available in the 7 woods where wild boar presence had been confirmed through activity signs on transects (excluding wood n. 17 and excluding tree number 1 that also had bait) had bite marks, mud or hair or rooting around the tree and no control tree had any sign of wild boar activity (χ2= 57.72, df=1, P < 0.001). If only the activity signs that wild boar left directly on the trees were taken into account, at the end of the trial 17 treated trees in 6 woods had bite marks, mud or hair whilst no control tree had any sign of wild boar activity (χ2= 24.52, df=1, P < 0.001). In four woods (n. 18, 19, 21 and 23) pictures of wild boar were recorded by camera traps and in one wood (n. 21) pictures of a wild boar rubbing against a treated tree were obtained (Fig. 7).

Fig. 7. Mud left by wild boar on a tree treated with the attractant (left) and wild boar rubbing against treated tree (right).

The number of activity signs left by wild boar on or around trees treated with the attractant increased with time (Fig.8). By the end of Week 1, only 3 woods and a single treated tree per wood had signs of wild boar activity; by the end of Week 2, all 7 woods had activity signs on or around 1- 4 treated trees per wood and by the end of Week 4 the number of treated trees with wild boar activity signs was between 3 and 6 per wood (out of 10 available in each wood). Although non-target species (deer, badgers, squirrels) were occasionally observed near treated trees, no obvious pattern emerged to suggest the compound was also attractive to these species. Four weeks after application on the trees, the compound was still visible and its smell detectable by the human nose.


Fig. 8. Number of woods with activity signs of wild boar recorded on or around trees treated with the attractant and total number of treated trees with wild boar activity signs. Wild boar presence was detected in 7 out of the 8 woods surveyed. All 7 woods showed signs of wild boar by week 2.

General DiscussionThe distribution maps of wild boar suggest that in the last 10 years this species has spread in England as the number of 5 x 5 km squares where wild boar was recorded increased or even doubled in Kent/East Sussex, West Dorset and Gloucestershire (Wilson 2014). At a local level, the results of the surveys in 20 woods surrounding the Forest of Dean suggest little expansion since the winter of 2010-2011. This is a shorter period than that used for investigating national change, so direct comparisons should be treated with caution. The results of the two methods employed for detecting wild boar presence in study sites and years characterised by wild boar densities that varied between 0.7 and 7 animals/100 ha (see Objective 2) suggested the minimum effort required to obtain a >90% probability of detecting wild boar was at least 9 camera traps/100 ha; if the site that had the minimum density of wild boar recorded during the study was excluded, the minimum number of camera traps required to detect wild boar presence with >90% confidence was 4/100 ha. In the same five study sites, the minimum the number of transects/100 ha required to detect wild boar presence with > 90% confidence was at least 7 transects/ 100 ha in 9 out of the 10 surveys. In one survey, the maximum number of transects surveyed per 100 ha (i.e. 10 transects/100 ha) resulted in an 88.6% probability of detecting wild boar.These findings suggest that in areas where the density of wild boar is assumed to be low, such as recently colonised sites or woodlands with no previous record of wild boar where RTAs or wild boar sightings occurred, camera traps might be marginally better than activity signs on transects to detect wild boar. Unlike activity signs on transects, that can be relatively old and for untrained staff more difficult to distinguish from signs left by other wildlife, camera traps provide unequivocal proof of wild boar presence. On the other hand, the method of activity signs on transects is relatively quicker and less expensive than that based on camera traps that must be left in place for several days and may be stolen (Table 3). The simulation of the effort required to detect a single wild boar in a large wood (55 km2) indicated that circa 15 and 20 camera traps/100 ha should be deployed to have respectively a 90% and 95% probability of detecting a single wild boar. Future research, initially carried out through a similar exercise, should establish whether the effort required to detect a single wild boar through camera traps is affected by the area of the wood. A putative site attractant, based on birch wood tar, was found effective in changing boar behaviour as wild boar were observed sniffing the attractant and rubbing against the treated stakes more times that the control stakes. When the trial was applied to trees, this behavioural change was confirmed: wild boar left signs on trees treated with this attractant more often than on control trees. The number of activity signs such as rooting around the tree, tusk marks, hair and mud left on the tree was significantly higher on treated than on control trees and the effect of the compound used persisted for at least 4 weeks without re-treating the trees. The use of the attractant had at least two advantages over bait stations: 1. the specific compound modified the behaviour of wild boar but not of other species and 2. the attractiveness persisted for at least 4 weeks without re-treating the trees, unlike the bait that had to be replenished. These results suggest that this compound could be used to improve the probability of detecting the presence of wild boar in a woodland and possibly also to increase the efficiency of trapping or attracting wild boar to areas where vaccines and contraceptives are delivered in wild-boar specific devices such as the Boar-Operated-System (Massei et al. 2010, Campbell et al. 2011), although this would need to be confirmed by further study. We suggest that bait stations with camera traps (rather than bait trails that employ significantly larger amounts of bait), used in conjunction with attractants may represent the less expensive method to confirm unequivocally the presence of wild boar in an area, particularly where the species occurs at low density. The advantages and limitations of the individual methods developed through this project and


through project WM0318, as well as the effort associated to each method, are summarised in Table 3. Different methods can be used in any combination, such as site attractants and camera to confirm wild boar presence. The study established the effort required for detecting the presence of wild boar in an area in relation to the number of camera traps deployed or transects surveyed in a study site, at various densities of wild boar. In addition the study estimated the minimum number of camera traps that must be deployed in large woodland (55 km2) to detect the presence of a single wild boar and the probability of detection associated with increasing densities of camera traps. Future research should establish whether using site attractants would increase the probability of detecting wild boar in an area and hence decrease the effort (in terms of camera traps deployed) to assess the species’ presence in an area.These tools could be used by stakeholders to monitor presence of wild boar in a new area, to quantify with reasonable confidence the spread of the species at local scale or to assess whether wild boar are still present after an eradication campaign. In parallel, the mapping of sightings provides the historical background against which trends of wild boar in England can be assessed in future years.

Method Advantages Limitations Cost and staff time

Single camera trap at bait point

Provide direct evidence of wild boar presence in the area

Cost of camera traps Camera traps stolen or malfunction Bait must be replaced once/ week Bait consumed by non-target Camera must be left for > 2 weeks Experience to operate camera trap Require minimum of 3 site visits Might not detect wild boar at very low

density or when availability of natural food is high

Time to place and retrieve 1 camera and to replace bait: 2 hr per visit

Camera trap: £ 200-500

Bait: £10

Multiple camera traps

Provide direct evidence of wild boar presence in the area

No bait required

Cost of multiple camera traps Camera traps stolen or malfunction Camera traps must be left for > 2 weeks Experience to operate camera traps Require minimum of 2 site visits

Time to place and retrieve 12 cameras: 8 hrCamera trap: £ 200-500 each

Activity signs on transects

No equipment required

Require single site visit

Provide indirect evidence of boar presence in the area

Experience to recognise wild boar signs Age of activity signs difficult to assess Best used in winter than in summer

Time: 8 hr for 14 200m transects in ~ 150 ha wood.

Site attractant

Likely to attract boar throughout the year

No experience required to apply attractant

Provide direct evidence of boar presence in the area

Age of signs easy to assess if no signs present before the attractant is applied

Must be left for a minimum of 2 weeks Require minimum of 2 site visits Experience to recognise wild boar signs

(e.g. hair, tusk mark) Tree growth may be affected by

frequent rubbing by boar If used on stakes, stake may be

displaced by boar rubbing

Time : > 8 hr to include 1st and 2nd visit to a wood where attractant is placed on ~ 10 trees

Attractant: £ 12.50


Can be used on stakes or trees

Table 3. Advantages and limitation of methods to detect presence of wild boar in a new area or where recent wild boar sighting or road traffic accidents have occurred. ‘Time’ refers to the time estimated to complete the survey in a single wood of ~150 ha and assumes an operator takes 30 minutes to drive to the wood, thus an hour driving for each site visit.

Objective 2. Refine cost-effective methods to quantify wild boar density and abundance.

The second objective was to use empirical and theoretical approaches for refining the methods developed during project WM0318 to quantify wild boar relative density or population trends. Thus objective 2 was divided into 5 parts as follows:

2a. Activity signs on transects 2b. Camera traps 2c. Distance sampling through thermal imagining2d. Simulation of density assessment through camera traps and distance sampling2e. Road Traffic Accidents (RTA)

Activity signs and camera trap surveys were carried out in the five study sites mentioned under Objective 1, distance sampling was used in the Forest of Dean and calibrated in an Italian study site with a high density of wild boar and the method of RTA was used on data collected in the Forest of Dean. For all the methods described below, a closed wild boar population was assumed, i.e. the amount of immigration and emigration of animals in the study site was assumed to be negligible for the time taken to complete each survey. Methods, results and a brief discussion for each method are treated separately and the overall results reviewed in the general discussion.

2a. Activity signs on transects

MethodsThe transects described under Objective 1 and used to detect wild boar presence in the five study sites were also used to quantify wild boar rooting signs and trails and to derive indices of wild boar population trends. No independence was assumed for data collected within and between transects within a site. The proportion of transects with rooting signs and with trails and the number of trails per transects was calculated for each site and each winter. A residual maximum likelihood analysis (REML) was used to derive an activity index for each site. A REML analysis, using year as a fixed effect, was employed to compare the activity index between years within sites.

The predicted number of boar trails or signs of rooting xij for year i and transect j was calculated as follows: xij = μ + Si + Tj + εij

where Si is a fixed effect for year, Tj is a random effect for transects and εij accounts for residual variability within year and transect.

Following Engeman (2005) and Engeman et al. (2002) a Passive Activity Index (PAI) for each winter i and each site was then derived as:

where ti is the number of transects within year i. Bootstrapping (Efron, 2000) was used to estimate the uncertainty associated with each PAI by re-sampling 10,000 times the data from boar trails or signs of rooting on transects at random. Thus, for each site and each season, a mean PAI and a standard error were obtained from the bootstrapped data.


A REML analysis was used to compare PAIs between years within sites. Chi-square tests with Yates’ correction were used to test between-year differences in the proportion of transects with rooting or with wild boar trails separately for each site. A Spearman’ rank correlation coefficient was used to test the association between the proportion of transects with trails and the proportion of transects with rooting signs in all five sites across two years.

Results and DiscussionThe trend in the proportion of transects with boar trails or rooting suggested that these signs increased in four out of five sites, with the exception of Penyard/Chase (Fig. 9). However, significant differences were only found in the proportion of transects with boar trails between years in Oakenhill (χ² = 5.01, P = 0.024) and Penyard/Chase (χ² = 4.96, P = 0.025). (Fig. 9). There were no differences (at the 5% significance level) in PAIs calculated separately on number of trails and on number of rooting signs between years in any of the sites (Wald statistics) with the exception of the number of trails per transect in Oakenhill (F = 6.92, d.f.= 1,20, P= 0.016) (Table 4 and 5).The results also showed a close association between the proportion of transects with trails and the proportion of transects with rooting signs (t= 5.45, df=8. P<0.001, Correlation coefficient Rho=0.89). The main limitations of using trails as an index of population trends are that i) trails are more visible in the wet season than in summer and ii) several animals can use same trail, thus resulting in an underestimate of number of individual trails. On the other hand, the area covered by rooting activity can be extremely variable in size (ranging from ~ 50 cm2 to many tens of square metres) and rooting per se is strongly dependent on food availability. Despite these limitations, these results suggest that both wild boar trails and signs of rooting could be used as the simplest way to monitor wild boar population trends by calculating the proportion of transects where rooting or trails were recorded.

Fig. 9. Proportion of 200 m transects with wild boar trails (left) and rooting (right) in five study sites in England in winter 2011-‘12 and winter 2012-‘13. In each site the density of transects was 10/100 ha.

The increasing trend in the proportion of transects with wild boar trails and rooting signs between 2011 and 2012, coupled with the increase in number of trails per transect suggested an increase in number and spread of wild boar in all the sites studied except for Penyard/Chase (Table 4 and Table 5). ________________________________________________________________________

Winter 2011-2012 Winter 2012-2013 Site Number of PAI SE PAI SE P transects Beckley/Bixley 28 0.04 0.15 0.36 0.15 0.142Brede 26 0.04 0.06 0.15 0.06 0.185Oakenhill 21 0.05 0.12 0.48 0.12 0.016Penyard/Chase 20 1.00 0.22 0.55 0.22 0.165Serridge 37 1.11 0.31 1.78 0.31 0.104_________________________________________________________________________Table 4. Winter Passive Activity Index (PAI) and standard error calculated using the number of


wild boar trails on transects (200 x 1 m) in 5 study sites. In each site the density of transects was 10/100 ha.______________________________________________________________________

Winter 2011 Winter 2013 Site Number of PAI SE PAI SE P transects Beckley/Bixley 28 0.14 0.15 0.21 0.15 0.745Brede 26 0.08 0.07 0.15 0.07 0.490Oakenhill 21 0.52 0.33 1.05 0.33 0.199Penyard/Chase 20 1.80 0.47 0.85 0.47 0.081Serridge 37 2.22 0.33 2.11 0.33 0.742 _________________________________________________________________________Table 5. Winter Passive Activity Index (PAI) and standard error calculated using the number of wild boar rooting on transects (200 x 1 m) in 5 study sites. In each site the density of transects was 10/100 ha.

When PAIs were compared in each site between years, significant differences resulted only in the PAI calculated on the number of trails in Oakenhill (Table 4). This was probably due to the relatively high variability surrounding the estimated PAIs obtained from wild boar trails and rooting signs, particularly for Beckley/Bixley, Brede and Oakenhill in winter 2011-2012.

2b. Density estimates by camera traps

MethodsThe aim of this part of the study was to refine the use camera trap surveys to obtain PAIs as well as relative densities of wild boar and to compare these indexes and densities within sites between years. Camera trap surveys were carried out in January-February 2012 (winter 2011-2012) and January-February 2013 (winter 2012-2013) by placing 16 camera traps/100 ha, as explained under Objective 1.Chi-square tests with Yates’ correction for each site were used to test between-year differences in the proportion of camera traps with wild boar visits (defined as in Objective 1). A PAI on the number of wild boar visits per 9 days was then calculated for each site as follows:

where ci is the number of cameras within winter i and xij is the predicted number of

visits for camera traps j in winter i and can be written as:xij = μ + Si + Cj + εij

where Si is a fixed effect for winter and Cj is a random effect for cameras and εij accounts for residual variability within winter and camera traps.REML analyses were carried out on data obtained from camera trap surveys: winter was entered as a fixed effect, to investigate potential differences in PAIs between winters, for each site. PAIs within sites between years were compared by Z-tests.A density estimator D was calculated (after Rowcliffe et al. 2008) for each study site and each survey separately, based on the number of wild boar visits per 9 days as follows:

where y/t = number of visits y per unit time t r and θ= radius and angle of the camera’s detection area v = speed of movements.

D was then multiplied by mean group size to obtain the density of wild boar/100 ha in each study site (Rowcliffe et al. 2008). Independent estimates of group size were obtained in three of the sites


used for this study, by using bait stations and camera traps. Group size was calculated in October 2011 (n= 26 observations) by placing 18 camera traps in Penyard/Chase, Oakenhill and Ruardean and in October-November 2012 (n=39 observations) by placing 20 camera traps in the same locations used the previous year. To minimize potential double counting, individual groups or animals were identified by a number of features including ratio of females to piglets and physical traits such as body size and coat colour.

Group size in October 2011 was 2.50 (SD 1.70) and in October-November 2012 was 3.74 (SD 2.91). The speed of movements, (v) = 0.274 (SD 0.052) km/hr was obtained from wild boar (n=7) equipped with GPS collars which were programmed to record fixes and activity every 15 minutes in the Penyard-Chase site (Quy et al. 2014). Although this was a relatively small sample size, it was the only one available to us on wild boar in England.The radius (r) of a camera trap was 18.288 m and the angle (θ) of the camera trap’s detection area = 40 degrees (= 0.698 radians), Bootstrapping (Efron, 2000) was used to estimate the uncertainty associated with the density estimates by re-sampling 10,000 times the data from camera trap pictures at random and by estimating the corresponding wild boar density. Then, for each site and each season, a mean density and a standard error were obtained from the bootstrapped data. A REML analysis was used to compare densities between years within sites. Bootstrapping was also used to determine how increasing the number of camera traps reduced the variation around the estimated mean.

Results and DiscussionThe proportion of camera traps with wild boar pictures between years differed significantly in Oakenhill (χ² = 6.058, P = 0.014) but not in all the other study sites (P>0.05) (Fig. 10). The PAI obtained from the camera trap surveys suggested a general trend for wild boar populations to increase from 2011 to 2012, although the difference was significant only for Oakenhill (Table 6).

Fig. 10. Proportion of camera traps with wild boar pictures in five study sites in England in winter 2011-2012 and winter 2012-2013. In each site the density of camera traps was 16/100 ha.

Site Month N. of cameras PAI SE P

________________________________________________________________________Beckley-Bixley Jan-Mar 12 42 0.43 0.15 0.460 Jan-Mar 13 38 0.61 0.17

Brede Jan-Mar 12 34 0.51 0.17 0.199 Jan-Mar 13 35 0.74 0.17

Oakenhill Jan-Mar 12 35 0.14 0.25 0.031 Jan-Mar 13 35 0.94 0.25


Penyard Jan-Mar 12 33 1.09 0.40 0.540 Jan-Mar 13 32 0.75 0.41

Serridge Jan-Mar 12 27 0.33 0.19 0.255 Jan-Mar 13 29 0.65 0.18_______________________________________________________________________Table 6. Passive Activity Index (PAI) and standard error calculated using number of wild boar visits per camera trap in five study sites in winter 2011-2012 and 2012-2013. P refers to comparisons between years. In each site the density of camera traps was 16/100 ha. The resulting number of camera traps reported in the table excludes those that did not function.

The mean densities of adult wild boar in different sites varied between 0.7 and 5.4 animals/ 100 ha in winter 2011-2012 (or between 1.7 and 5.4 animals/ 100 ha if Oakenhill was excluded) and between 4.5 and 7 animals/100 ha in winter 2012-2013 (Table 7). Between years, wild boar densities showed a general trend to increase in all 5 sites, although the difference was only significant for Oakenhill and nearly significant in Brede (Table 7). _______________________________________________________________________Site Month N. of N. wild boar/ SE P

Cameras 100 ha ________________________________________________________________________Beckley-Bixley Jan-Mar 12 42 2.14 0.61 0.14 Jan-Mar 13 38 4.48 1.45

Brede Jan- Mar 12 34 2.62 0.68 0.06 Jan-Mar 13 35 5.54 1.40

Oakenhill Jan- Mar 12 35 0.71 0.25 0.01 Jan-Mar 13 35 6.99 0.25

Penyard Jan- Mar 12 33 5.41 1.64 0.95 Jan-Mar 13 32 5.63 3.45

Serridge Jan- Mar 12 27 1.66 0.52 0.07 Jan-Mar 13 29 4.90 1.72_______________________________________________________________________Table 7. Estimated number of wild boar/100 ha and standard error in five English study sites in winter 2011-2012 and in winter 2012-2013. Numbers are derived from camera trap surveys and from independent field estimates of mean group size, speed of movement. P refers to comparisons between years. In each site the density of camera traps was 16 camera traps/ 100 ha. The resulting number of camera traps reported in the table excludes those that did not function.

The relationship between the number of camera traps employed to estimate wild boar densities and the 95% confidence intervals around each estimated density across all sites and years (Fig. 11) suggested that between 6 and 15 camera traps/100 ha would be sufficient to calculate wild boar density. The results indicated that increasing the sampling effort, in terms of number of camera traps, would only marginally increase the precision of the density estimate, at least for wild boar densities similar to those recorded in this study. For field applications, a minimum number of 15 camera traps/100 ha, left in place for at least 9 days, is thus recommended to assess density of wild boar with relatively small confidence intervals.

2c. Distance sampling through thermal imagining Thermal imaging cameras capture radiant thermal energy of wavelengths 8-12μm, presenting the energy received as a monochrome image. These cameras are particularly suited to observation of warm-bodied animals at night which typically have surface temperatures several degrees above


ambient, and which do not resort to concealment during the hours of darkness. In an ideal case, a test of a density estimation method should involve a trial where a known number of animals exist. However, since the method is intended for application in extensive areas of forest and woodland, the establishment of a known population of wild animals in such an environment is impractical. Instead our work focussed on identifying the precision achievable in relation to the effort involved and on determining sources of bias can arise in practice. For distance sampling, four assumptions need to be met to eliminate biased results (Buckland et al 2001):

1) All animals on the transect line must be detected 2) Animals are detected prior to any avoidance movement3) Distances are measured without bias4) The location of transect lines should have no relationship to the distribution of animals

through the survey area Compliance with the first assumption is relatively straightforward. Animals are most likely to be detected on the transect because surveyors concentrate most observation effort on the area ahead of them. The use of thermal imaging is intended to help with compliance with the second assumption because it facilitates detection of animals either before they become aware of the observer or before they need to take flight. Previous studies have indicated that the majority of animals are stationary or walking when first observed, not running (Gill et al 1997).


Fig. 11. Wild boar density estimates (in blue) and 95% confidence intervals (in green) calculated for each study site and year in relation to the number of camera traps/100 ha used.

The suggested approach to achieve compliance with the fourth assumption is to ensure transects are located at random across the survey area (Buckland et al 2001). However while this is usually relatively easy for marine or airborne surveys it is impractical for nocturnal terrestrial surveys. Instead transects need to follow existing paths. Simulation studies however have revealed that this will not lead to biased results if a sufficient density of transects can be placed across the survey area (Gill et al 1997), and if animals do not avoid the transect routes.The purpose of this part of the project was to develop and test distance sampling through thermal imaging as a method for estimating wild boar abundance in UK woodlands, and to assess the strength of biases that may arise in a practical application of the method. For the reasons stated above, we concentrated on testing how detection rates and distances may be affected by the choice of methods or equipment, and whether wild boar movements may lead to bias.Thermal imagers are designed with differing sensors and lenses which affect image resolution and detection range. Some of the more expensive models are equipped with cooled detector arrays which improves capacity to resolve small temperature differences. Further, some models have been equipped with coaxial laser rangefinders for distance measurement. Without this, distances need to be estimated from apparent body size in the viewfinder. Since the choice of thermal imager may therefore have an effect on results, this study compared detection rates, distance estimates and population estimates obtained from when imagers with different focal length lenses and different methods of distance estimation (laser vs body size) were used.

MethodsEstimates of densityEstimates of feral boar density were obtained from Beckley wood in 2009 and for Penyard wood in 2010 under project WM0318. The numbers of boar in these sites were much lower than expected, possibly because of a recent increase in culling by local landowners. A high sampling effort was adopted in these two woods in an effort to achieve an adequate number of observations, but the precision of the density estimates obtained was nonetheless rather poor. As a result, further trials were carried out in Italy in 2011 and 2012, in an area of relatively high density of wild boar and in collaboration with colleagues from ISPRA (Instituto Superiore per la Protezione e la Ricerca Ambientale).During the course of this project, the small population in the Forest of Dean increased rapidly, to the extent that that a survey of wild boar numbers was needed by the forest managers. A survey was therefore carried out in 2013 and the results have been included in this report. The areas surveyed varied in size and landscape composition (Table 8). However, all sites included a substantial proportion of broadleaved woodland and most included an element of coniferous woodland and open fields, either in or adjacent to the woodland area. Fields were surveyed at the same time as woodlands. The sampling intensity routinely used for these surveys is circa 2.5 km of transects/km2 of area surveyed in woodlands and 1 km of transects/km2 of area surveyed in open fields. As the density of wild boar in the English study sites was relatively low, the sampling intensity was increased (Table 8).

Site Year Transect length Sampling intensity


Table 8. Size of surveyed areas and survey effort.

The survey methods adopted were similar to those developed for deer (Gill et al 1997; Mayle et al 1999). Observations were made at night from paths or roads (used as ‘transect routes’) that traverse the study area, either from a vehicle or on foot, as appropriate. When a group of animals was detected in the thermal imager, the species, group size, distances and compass bearing from the transect were recordedAt one site, (Alto Merse) two observers worked together to compare thermal imagers and methods of distance estimation. The two thermal imagers (a Flir Thermacam TM B640® with a 12˚ (x4) lens and a Pilkington Thorn Optronics with a 5.7˚ (x9) lens differed in field of view and image magnification. When a group of animals was detected simultaneously by both observers distance to the group was estimated using a co-axially mounted laser rangefinder as well as from apparent body size measured in the viewfinder of the thermal imager (Mayle et al 1999).During the survey, an attempt was made to identify all animals to species level. Identification, however, becomes increasingly difficult with increasing distance to the animals, so during analysis, unidentified animals were assumed to be either boar or deer in the same ratio as those identified. The distances to each group detected were used to estimate the detection function, by fitting a curve to the frequency distribution of detection distances using ‘Distance’ software. Density (D) is then estimated from :

D = [E(S).n/L]/2ESWwhere E(s) is the mean group size; n/L is the number of groups encountered per unit transect length (also referred to as “encounter rate”) and ESW is the effective strip width, the definite integral of the detection function between the transect (0) and (w), the maximum perpendicular distance of animals detected.

To obtain density estimates using distance sampling, it is recommended that observations on 50+ groups (comprising one or more wild boar) are used to fit a detection function (Buckland et al 2001). In view of the fact that few observations were obtained from the two smallest UK sites, data from these sites were pooled to fit a detection function for the UK sites, after testing to ensure the frequency distribution of detections did not differ between sites. As detectability of ungulates varies with vegetation density, fields and forest were sampled separately and detection functions and densities estimated for each in turn.

Investigations of wild boar movementsThe GPS data on movements of 11 wild boar, collected during project WM0408, were used to investigate evidence of avoidance of transects, a potential source of bias in distance sampling, and selection for concealment. Selection ratios, Si = Ln (Ui/Ai), were calculated where Ui = the proportion of all GPS fixes in habitat i, and Ai = the proportion of study area in habitat i. Habitats with relative use greater than availability yield values of Si > 0; and conversely Si < 0 for those with use less than availability. Diurnal activity was investigated by calculating the speed moved between successive pairs of fixes, provided they were obtained less than 100 minutes apart. This was to ensure that the estimated speed could be associated with a particular hour of the day.

ResultsThe results of the analysis of 24hr activity pattern reveal substantial differences (F=16.79; p<0.001 df= 23,1497; speed+1 log transformed; Fig 12) and suggest about 12 hours of activity at night followed by 12 hours of rest in the day for both summer and winter. There was greater activity in winter than summer with a peak in activity before dawn in winter and a short increase after dawn in summer.The analysis of habitat use in relation to transects indicated that boar have a tendency to avoid transects, up to a distance of about 20m during day but not at night time.


For groups detected by both thermal imagers there was a significant relationship between distances obtained by rangefinder and body length (Fig. 13) (F=113.1; df= 3,138; R2=70.9%).

Fig. 12 Mean speed (m/min) between successive fixes for each hour. Allocation to hour was based on the median between successive fixes, excluding pairs of fixes > 100 min apart. W = Winter (Dec21- Mar 16); S=Summer (Jun 21 – Sep 24). Data derived from 11 wild boar radio-tracked during project WM0408.

Fig. 13. Regression obtained comparing detection distances derived from a laser rangefinder (DetRf) with distances estimated from body length (DetB). Green triangles indicate observations made in forests, red indicate observations made across fields. This regression indicated no significant difference between deer and wild boar and therefore both groups of species have been included.

When two thermal imagers were used together the majority of groups of animals (138 of 194 or 71%) were detected by both observers. However the mean detection distance of groups detected only by the thermacam, with a wide angle lens (45m) was significantly less than the mean of those detected only by the Pilkington, with a narrow angle lens (148m; t=4.28; df=44; p<0.001) The frequency distribution of perpendicular distances to each group of wild boar revealed a greater proportion of groups detected at further distances across fields than in woodland (Fig. 14). There was some indication of a more extended distribution of detections from the two UK sites than Alto Merse, possibly due to more open woodland conditions, however the difference is not significant (χ2 = 6.295; 4df; p>0.1). There was no indication of avoidance behaviour in any of the sites, which is usually seen as a peak in observations some distance away from the transect (Figure 14). The detection functions fitted to the distance data indicate only a minor difference between distances estimated by range finder versus body length (Figure 14).The estimated densities in the UK sites were all rather low ranging from 2.2-8.7/ 100 ha, with a


relatively large coefficient of variation in 2009 and 2010. This is a reflection of the small numbers of observations from which they were derived (Table 9). In the Forest of Dean, a density of 8.7 wild boar/ 100 ha (95% Confidence Intervals 5.3-14.4) were estimated in March 2013: this resulted in an estimate of 535 wild boar (95% C.I = 325-885 animals).

Fig. 14. Frequency distribution of perpendicular distances (m) to groups of wild boar for each study site (left) and detection function curves fitted to distance data for wild boar in Alto Merse (right).

Population density estimates for wild boar in Alto Merse are summarised in Table 10. The numbers of both wild boar and deer detected differed substantially in different parts of the study area (Fig. 15). A significantly higher density of wild boar was detected in forests in the north part of Alto Merse than in the south in both 2011 and 2012 and a higher density in 2012 than in 2011. There was no significant difference in the densities found in the fields between 2011 and 2012. Overall the change in numbers suggest an increase of about 101% between 2011 and 2012. Densities of wild boar were significantly higher inside the reserve than outside. These estimates were based on 884 observations of which only 46 (5.2%) could not be distinguished between boar or deer with confidence, suggesting that any misallocation between these two groups is likely to be minor.

No. of groups

No. nights of survey

Transect length sampled(km)

Encounter rate(no. of groups/km)

Mean group size(no. of boar)

ESW (m)

Mean density(no. boar/ 100 ha)

Mean Est. Pop. Size

CV%

Penyard 24 14.5 81.8 0.269 3.3

82.4

5.4 13 42.6

Beckley 6 7 30.8 0.195 1.8 2.2 10 60.1

Dean 37 11 166.6 0.222 3.6 46.0 8.7 535 26.0

Table 9. Results of density estimates in the UK sites. ESW= Effective Strip Width; CV= Coefficient of variation.

Habitat Method Year Effort (m)

No. of transectssampled

Width(m)

No.groupsseen

ESW (m)

Encounter rate(no. of groups/km)

Meangroupsize

Mean density (no. boar/ 100 ha)

CV%

Forest Laser2011 40.2 20 181 26 47 0.65 2.2 14.8 35.7

2012 30.4 19 181 38 47 1.25 3.1 41.2 30.0

Field Laser 2011 22.1 14 239 45 151 2.04 4.5 30.6 40.72012 11.1 17 239 28 151 2.52 4.1 34.4 38.3

Forest Body2011 40.2 20 150 25 45 0.62 2.1 14.5 37.3

2012 30.5 19 150 36 45 1.18 2.8 37.5 32.5

Field Body2011 22.1 14 354 45 150 2.04 4.5 30.6 41.3

2012 11.1 17 354 28 150 2.52 4.1 34.4 38.9


Table 10. Density estimates obtained for wild boar in Alto Merse calculated from distances obtained from laser rangefinder and body length. ESW= Effective Strip Width; CV= Coefficient of variation.

Fig. 15 Map of Alto Merse, Siena province, indicating locations of groups of ungulates

DiscussionNo differences in population density were found between the two thermal imagers used. Comparisons between thermal imagers with different focal length lenses indicated that use of a long focal length lens resulted in fewer detections of animals close to the transect, whereas use of a shorter focal length lens resulted in fewer detections of more distant animals. In addition, some groups of animals were not detected by both types of thermal imagers. Using an imager with a long focal length lens in dense vegetation requires rapid sweeps by the observer, which is difficult in a moving vehicle and clearly resulted in some groups being missed. With distance sampling, the failure to detect animals close to the transect will cause a larger under-estimation of density than failure to detect more distant animals. These results suggest that a long focal length lens is better suited to open environments whereas a shorter focal length lens would improve detections in woodland. Due to these findings, using an imager with a zoom lens to adjust focal length to the visibility afforded by the vegetation would offer most versatility. There was no indication of significant avoidance of the transect from 11 GPS collared boar in one site nor was there any indication of avoidance of either observer or transect from the frequency distribution of perpendicular distances to each group. Both observations and GPS data confirmed that wild boar were more active at night and more likely to use fields or other openings at night where they are more readily detected. A substantial proportion of observations in Alto Merse were of boar using fields, from which they were absent in daylight. The results indicated that incorporating open areas when assessing density of wild boar on a site would ensure that a portion of the population is more easily detected and lead to an improved estimate of overall numbers. The data obtained on densities in all the study sites indicated a high degree of spatial and temporal variation in wild boar densities. Wild boar densities between 2.2 and 8.7/ 100 ha were obtained for the English sites and 14.8-40.7 / 100 ha for the site in Italy. For the English sites, these densities were likely to represent a reasonable range of maximum and minimum densities of wild boar in English woodlands. The precision of the density estimates obtained was rather poor, with a coefficient of variation ranging from 30% for Alto Merse to 60% for Beckley. The reason for the poor precision in the two UK sites is largely because few animals were present in either site. Indeed, the coefficient of variation associated with a higher wild boar density in the Forest of Dean in 2013 was lower (26%) than that recorded in previous years for UK wild boar populations. A common outcome of thermal imaging surveys is that they reveal substantial spatial variation in numbers of ungulates. This was very evident in the results obtained for wild boar in both Italy and UK, where high numbers were encountered in some parts of the study area in contrast to very few


Key Green – Wild Boar Red – Fallow Deer Blue – Roe Deer Purple – Red Deer Brown – unidentified Deer Grey – Unidentified ungulate Shaded – Reserve area

or none in other parts. This variation suggested that information on the spatial distribution of wild boar should be taken into account when planning a thermal imaging survey. Precision of the overall estimate will be improved by partitioning the area, where possible, to delineate areas of high from low abundance. This project found that distance sampling coupled with thermal imaging will return useful results for wild boar management. The variability in wild boar densities will sometimes yield estimates with a wider confidence interval than sought by management, particularly in small or low-density populations. Thermal imaging may prove to be a relatively expensive approach for sites where few animals occur or where there is low certainty about presence. In the latter instances, surveys based on activity signs or camera traps may provide a useful alternative as indicators of relative wild boar density. On the other hand, distance sampling for larger populations of wild boar will return relatively accurate estimates, thus suggesting this method could be appropriate for these populations.

2d. Modelling accuracy and precision of wild boar density estimated from distance sampling and camera trappingThe aim of this part of the project was to 1. develop a simulation model to investigate the accuracy and precision of wild boar density estimated from distance sampling and camera trapping and 2. investigate the effect of wild boar population size on density estimates obtained from distance sampling and camera trapping.

MethodsModel descriptionWe built an individual-based spatially explicit model using R . The model process is illustrated in Appendix 2. Briefly, a known number of wild boar groups were initially randomly assigned to cells in a grid representing the habitat being monitored (25 x 25 cells each representing a 300m x 300m grid in real life, for a c. 55 km2 study site which equates with the size of the Forest of Dean). The choice of grid cell size was guided by the maximum observable distance of transect used in situ when distance sampling, 150 m on both sides of the line transect. The model was that of a closed population, meaning that there was no birth, death, emigration or immigration during the study period; this was a realistic assumption due to the small survey time-frame of 15 days. Each group was assigned a size (i.e. number of individuals within it), which was randomly sampled from a normal distribution of mean group size (see Table 12) and standard deviation of 0.5 and then rounded to the nearest integer.At each time-step, wild boar groups were allowed to move or stay where they were; the Cartesian coordinates of their position in the grid was recorded at every time-step. The movement behaviour was randomly selected out of nine possible choices: step in the direction of one of the eight surrounding grids or remain stationary. If the direction randomly chosen took the animal outside the study site, another direction was randomly selected. The size of the step was determined by the average observed speed for wild boar during the day and during the night . Initially, wild boar groups were allowed to move in the habitat without being monitored; after 5 days, monitoring began. Camera traps were placed randomly with a density of one in every 6 ha (baseline density) and their Cartesian coordinates in the grid recorded. They were then allowed to continuously record sightings for 10 days. During that time, distance sampling using line transects was also undertaken. At each time-step, the occurrence and location of a transect survey was determined randomly. Baseline transect target number was set at 3 x 1km lines per 100 ha ; once the target number of transects was reached, distance sampling stopped.Because of the social nature of wild boar, when camera trapping, pictures within 10 minutes of each other are considered of the same family group. To reproduce this in the model, we set the time-step to 10 minutes, thus allowing a maximum of one picture per camera being taken within 10 minutes of monitoring. However, the chance of triggering a camera will be dependent on the number of individuals within a group to an extent. Therefore, at each time-step, for each group, we randomly assigned a unique position for each individual within 5 metres of the recorded position of the group. We then checked whether any individual in the group was within the detection area of


each camera, mimicking individual chances to trigger cameras.

ParameterEstimate

Group size in winter 2.5, 3.74

Number of groups 50, 100, 200

Single transect length (m) 800

Transect density 3km per 100 ha

Average winter day time speed (m/min) 1

Average winter night time speed (m/min) 4

Camera traps density 1 per 6ha

Camera radius r (m) 18.3

Camera angle (rad)θ 0.698

Grid size (Forest of Dean; 100 ha) 56

Cell size (m) 300 x 300

x1 (days) 5

x2 (days) 10

k 3Table 11. Parameters used in the wild boar model baseline simulations.

For the transects we estimated that a 10 minute time-step corresponded to 800m being surveyed . During distance sampling, each time a group was detected, all individuals within it were recorded as seen at the same distance from the observer . To replicate real distance sampling conditions, moreover, the chance of a group being seen was the highest (100%) if it was on the transect line, and decreased to circa 0% at a distance of 150m. The actual decrease in detection probability followed a second degree polynomial function, which we obtained by fitting a regression through observed detectability data from the Alta Merse site (Fig. 14).At the end of the 15-day timeframe, we estimated density from the camera traps record and transects record. For the former, we used the REM, described in Rowcliffe et al. ; values used in the formula are given in Table 11 and we used non-parametric bootstrapping to estimate the variance around estimated densities for each run. For the latter, we used the package ‘unmarked’ in R, which was developed for the statistical analysis of data from surveys of unmarked animals, included those collected from distance sampling (Royle et al. 2004, Fiske and Chandler 2011). Specifically, we used the function ‘distsamp’ in ‘unmarked’ which generates parameter estimates using the hierarchical multinomial-Poisson model of Royle et al. and tested three models (‘half-norm’, ‘exponential’, and ‘uniform’). Those models were then compared using the Akaike Information Criterion for small datasets , choosing the best model from which to extract density estimates and their standard errors. To investigate the sensitivity of density estimation to population size, we ran simulations with small, medium and large wild boar populations (Table 11). We also investigated the impact of group size on density estimates by using two different mean group sizes; they corresponded to the lowest mean observed group size in the Forest of Dean (2011 estimates) and the highest mean group size observed (2012 estimates) (Table 11). Each simulation was run 100 times. We then calculated the mean estimated density across 100 simulations and the 95% confidence interval, for all population size and group size combinations. We also examined the relationship between survey effort (expressed as number of survey days for


camera trapping or number of 1 km transects per 100 ha for distance sampling) and estimation of population; we thus ran simulations where camera trapping lasted 5 and 20 days and where number of transect was 1.5 and 6 per 100 ha for a population of 100 groups and mean group size 2.5. We assumed that between 5 and 20 days the wild boar population could be regarded as a closed one.

Results and DiscussionIn order to implement an average of 1 camera per 6 ha, 938 cameras were placed inside our study site, which corresponded to an average of 2.3 cameras per grid cell; the average daily trapping rate (total number of photographs taken by all the cameras per day) across 100 simulations was 0.027, 0.053, 0.107 for 50, 100 and 200 groups of mean size 2.5 respectively, and 0.033, 0.066, 0.128 for groups of mean size 3.74. As far as distance sampling was concerned, 223 unique transects were conducted for each baseline simulation, yielding a total of 178 km being surveyed. On average circa 26% of the population was detected each time, regardless of the number of groups, or their sizes. The best model selected using AICc for each simulation was ‘half-norm’ for 88% of all simulations; the ‘exponential’ model accounted for the rest.We found that both camera trapping and distance sampling underestimated modelled density, regardless of the number of groups present, or their average size (Table 12; Figure 16). The amount of bias in camera trapping results appeared to be dependent on mean group size, while being consistent across population size (i.e. number of groups) with the estimated density for smaller groups being less accurate than that of larger groups for all three population sizes (Table 12). This could reflect the fact that the probability of a group triggering a camera varied with group size but that this relationship was non-linear, a possibility not taken into account in the REM formulation. For distance sampling, there was no pattern in the underestimation of density. While the average bias varied between -18% and -30%, it was not clearly linked to population size or mean group size (Table 12; Figure 16). While neither camera trapping nor distance sampling accurately estimated density, the uncertainty surrounding mean density estimates was much larger for distance sampling than for camera trapping (Figure 16; Table 12).

No. of groups

Group size

True density (wild boar/100 ha)

Mean estimated density across simulations [CI]

Average difference from true density (%)

RSE across simulations (%)

Camera trapping

Distance sampling

Camera trapping

Distance sampling

Camera trapping

Distance sampling

502.5 2.22

1.35 [1.04 – 1.66]

1.82[0.86 – 2.77]

-39.19 -18.02 11.63 26.92

3.74 3.332.46[1.91 – 3.00]

2.59[1.59 – 3.40]

-26.13 -22.22 11.38 19.69

1002.5 4.43

2.70[2.26 – 3.14]

3.13[2.13 – 4.12]

-39.05 -29.35 8.15 16.29

3.74 6.654.85[4.00 – 5.61]

4.62[3.41 – 5.83]

-27.07 -30.53 8.04 13.42

200

2.5 8.865.34[4.72 – 5.96]

6.37[4.93 – 7.81]

-39.74 -28.10 5.99 11.46

3.74 13.309.50[8.44 – 10.57]

10.06[8.06 – 12.06]

-28.57 -24.36 5.68 10.14

Table 12. Mean density estimates yielded by 100 simulations. CI stands for the 95% Confidence Intervals. RSE is the relative standard error, which is expressed as a percentage and calculated as RSE = 100 * SE / Mean, where SE is standard error.

Moreover, overall, mean densities estimated from distance sampling tended to be closer to the true density in the habitat. For both camera trapping and distance sampling, precision around mean density estimates increased with population size.


The results indicated that mean density estimates were not very sensitive to the amount of survey effort put into camera trapping or distance sampling. For both monitoring methods, whether the effort was halved or doubled, the mean difference between true density and mean estimated densities were very similar to the results of the baseline simulations (CT 5 days: -40%; CT 20 days: -41%; DS 1.5 transects/100 ha : -21%; DS 6 transects/100 ha: -27%). Increasing or decreasing survey effort only had an impact on the size of the uncertainty surrounding mean density estimates for both methods (i.e. the precision). This suggested that increasing the number of days for which cameras are in place, or increasing the number of transects performed, may improve the precision of field-based density estimates. The model indicated that both camera trap and distance sampling may underestimate wild boar density. Camera trap density estimates were fairly precise and were not sensitive to population size, but were highly sensitive to mean group size whilst distance sampling density estimates were more uncertain but not sensitive to group size or population size, at least at the densities modelled here.This study is the first to use a simulation-only approach to compare and investigate density estimation using camera trapping and distance sampling. Others have evaluated one or the other method against field-based data and other monitoring approaches . For example Parment er et a l. (2003) compared the output of distance sampling against count of rodents obtained from a controlled experiment and found that distance sampling can produce estimates with small or large biases and consistently poor precision. Smart et al. (2004) built a simulation model to compare the accuracy of distance sampling with faecal standing crop and faecal accumulation rate for monitoring woodland deer species in the UK. They found that overall precision of density estimates increased with survey effort and population density, and that accuracy of distance sampling did not depend on population density, which seems to be consistent with the findings of the current study , although the latter suggested that distance sampling using a thermal imager should not be used for small wild boar populations.The results presented here are thus important for the management of wild boar; as populations have the potential to increase quickly , accurate estimates of densities are needed to monitor population trends and to quantify the impact of management options on population size. Such knowledge can be used to design efficient control and management strategies to protect the natural environment, crops, and livestock against the impact of wild boar. The results of this study suggest that, theoretically, the most precise density estimate can be obtained from camera trapping. However, this is dependent on the availability of a robust estimate of mean group size; this should be measured just before the camera trap census and in winter when piglet numbers are lowest, in order to use the most appropriate value for the camera trap density estimate method proposed by Rowcliffe et al. (2008). The model used in this study needs to be validated experimentally, as the bias we found is likely to partly stem from model design despite our best efforts. However, in theory, it could be used to adjust field-based density estimates to reflect actual densities of wild populations, especially because the direction of the bias is consistent across simulations. On average, the model suggested that mean camera trap density estimates were biased by c. 33.9%, although it varied between group sizes (2.5 group size: 40.6%; 3.74 group size: 27.2%). Distance sampling underestimated densities by circa 25.7%. As a result, the best monitoring method to employ will depend on the information available to managers. If robust data are available on the mean group size of the target population, then camera trapping will yield the least uncertain estimates of density. In the absence of these data, distance sampling should be used to estimate wild boar density, provided managers are aware of the large uncertainty surrounding those estimates.

The model carries some limitations, which could be addressed through further code development. For example, habitat was assumed to be homogeneous throughout the study site. This was unlikely to be true and may have introduced some bias in the results that were not due to the monitoring method but to model design. However, introducing different types of habitats would have required habitat-specific data on the ecology of the wild boar (e.g. habitat use) and the habitat-specific accuracy of distance sampling methods and camera traps , which were not available at the time of


model development. Moreover, cameras were placed randomly in grid cells, ignoring the fact that in the field, placement is often non-random, with cameras facing tracks, or at least not placed in areas where they have no chance of recording wild boar, and with cameras placed closely together more likely to capture the same animals. The assumption of random camera placement made in our model meant that every camera had an equal chance of capturing animals, which might have introduced a bias in our results. This could be addressed by introducing an error term, i.e. stochasticity, in the probability that each camera captures individuals within their range. Alternatively, accounting for the impact of non-random placement of cameras based on true landscape would be possible by introducing bias specific to habitat type; however, data on the relationship between cameras triggering and habitat structure are not available.

Fig. 16. Spread of the density estimates yielded by 100 simulations of camera trap monitoring (CT) and distance sampling (DS) when there are (a, b) 50, (c, d) 100, and (e, f) 200 groups in the population; two average group sizes are shown: 2.5 (a, c and e), and 3.74 (b, d and f). The black horizontal line in each panel represents the modelled density.


Another big assumption made in the model was that the movement of individuals was random. In reality, movement is likely to be habitat and season-dependent. This should not be a problem for the camera trap estimates, as the method developed by Rowcliffe et al. (2008) is based on the same assumptions. In fact, the latter was found to not be sensitive to non-random and non-independent movement of animals , thus the assumption made in the model developed for this study did not significantly impact our results. Contingent on external validation, the results of this study could be used to directly inform wild boar management in England. Moreover, this model can potentially be used to inform the management of any wild population for which marking individuals is either not possible or not cost-effective. In summary, the model indicated that, despite biases, camera trapping and distance sampling seem to perform reasonably well for estimating the density of wild boar populations. By quantifying, the difference between true density and estimated densities with both methods, this study filled an important gap in ecological research and presented the first step towards improving the robustness of these field-based estimates of density.

2e. RTAs and detection of wild boar presence from transects and camera trapsMethodsInformation on RTAs in the Forest of Dean from March 2008 to March 2013 was provided by the Forestry Commission. In addition, data on traffic flow in the Forest of Dean between 2008 and 2013 were provided by Gloucester Highways (between 2008 and 2010: Transport Monitoring Team, Gloucestershire Highways, [email protected]) and by Gloucester County Council (between 2011 and 2013). Data provided by Gloucestershire Highways were expressed as proportional increase or decrease of traffic flow on Gloucestershire roads in one year compared to the previous year. Data from Gloucester County Council were expressed as number of vehicles/24hr in different parts of the Forest of Dean.

Results and DiscussionThe data on traffic flow indicated that in Gloucestershire traffic flow decreased by 1-2% between 2008 and 2010 (Gloucestershire Highways data). The data on number of vehicles/24hr monitored in three locations of the Forest of Dean also decreased from 8793 (SD 2812) in 2011 to 8506 (SD 1424) in 2013. Thus the traffic flow remained relatively unchanged between 2008 and 2013. In contrast, the number of RTAs increased (Fig 17). Although the relatively small sample size (n = 5 years) prevented statistical analyses, the trend observed in the number of RTAs indicated that this number steadily increased (Fig. 17). As this trend reflected the same trend in density estimates, and the traffic flow does not vary significantly, the number of RTAs could be used as indicators of wild boar population trends in an area.

Fig. 17. Number of Road Traffic Accidents (RTAs) involving wild boar in the Forest of Dean between 2009 and 2013.


General discussionSeveral methods have been developed and applied to determine the absolute or relative density of wild boar. These methods include catch per effort based on harvesting (Geisser and Reyer 2005, Acevedo et al. 2006), pellet counts (Hone and Martin1998, Acevedo et al. 2007), capture-mark-recapture based on marked animals (Hebeisen et al. 2008) or faecal DNA (Ebert et al. 2012) and distance sampling (Focardi et al. 2002, Franzetti et al. 2012). Methods based on camera traps, recently employed to estimate densities of various species (e.g. Rowcliffe et al. 2008, Rovero and Marshall 2009, Royle et al.2009, Gardner et al. 2010, Gerber et al. 2010), have not been used for wild boar. One option to compare density estimates with actual densities of wild boar would be to use one or more methods to estimate density and then remove animals from the area to calibrate the method against the known number of animals (Andersen 1953). To our knowledge, no study has carried out such comparison on wild boar, even where whole populations of feral pig populations have been removed (e.g. Parkes et al. 2010). On the other hand, the fact that different methods achieve similar conclusions when applied to the same population, increases the confidence in the accuracy of these methods (e.g. Engeman et al. 2001).This study showed that the two methods used to estimate densities of wild boar in five English study sites (each comprising a woodland or part of a woodland), namely camera traps surveys and distance sampling provided similar results. These results were also confirmed by the theoretical approach based on simulation modelling that examined precision and accuracy of these methods, although the simulation concluded that both approaches may underestimate wild boar numbers. In terms of practical applications, thermal imaging and activity sign surveys are restricted to winter due to better visibility (thermal imaging) or to the persistence of activity signs which are more detectable in winter than in summer. Camera trap surveys are less likely to be affected by season and could be used any time of the year: although camera trapping may be constrained by seasonal factors, such as growth of dense understorey vegetation, camera traps can be moved, within a set radius, to maximise detection of wild boar. Thermal imaging and line transects have been used to assess ungulates’ numbers in several studies (e.g. Focardi et al. 2002, Haroldson et al. 2003, Franzetti and Focardi 2006) and in areas with relatively high densities of animals. For instance Focardi et al. (2002) used this method in a Mediterranean area and obtained densities of 10.6 wild boar/100 ha with relative low coefficients of variation (24%). Similarly, distance sampling carried out in an Italian site for the current project suggested even higher densities of 14.8-34.4 wild boar/100 ha, with coefficients of variation of 35-39%. The densities of wild boar estimated in the current study fall within the range of those reported in continental Europe where they range from 2 – 7 animals/100 ha (Smiet et al. 1979; Ickes 2001; Pihal et al. 2010) to 10-11/km² (Focardi et al. 2002, Hebeisen et al. 2007) with extremes of 66 animals/100 ha (Franzetti et al. 2012). The densities recorded in the current study are also similar to those observed for populations subjected to hunting pressure in the Bialowieza Primeval Forest in Poland, with densities of 0.7-5.1 wild boar/100 ha (Jedrzejewska et al.1994), in coastal regions of California with densities of 0.7-3.8 wild pigs/100 ha (Sweitzer et al. 2000) and in Germany with densities of 4.5- 5.0 wild boar/100 ha (Ebert et al. 2012).Although no specific data on four of the sites used in this project are available for winter 2014, the latest distance sampling estimate carried out in the Forest of Dean in March 2014 found that wild boar numbers increased from an estimated population of 535 (i.e. 8.7 wild boar/100 ha, 95% c.i. 5.3-14.4) in March 2013 to 819 animals in March 2014 (i.e. 12.3 wild boar/100 ha, 95% c.i. 7.6-19.9) despite a cull of circa 130 animals in the same period (Gill 2014). It is possible that such increase will also lead to spread to new areas if the population is not effectively controlled. On the other hand, densities in the Weald are still relatively low, possibly due to the intense culling occurring in these areas.The results of the current study suggested that, at current densities, both camera trap surveys and distance sampling could be employed for estimating local wild boar numbers. Estimating the absolute density would require knowledge of the true number of animals in an area, which could be possible only if all wild boar were removed after the above methods were applied (Hounsome et al. 2005). In the majority of cases this might not be possible or indeed desirable and it is likely to be expensive. However, if the eradication of a localized population of wild boar became necessary, for


instance following a disease outbreak, camera trap surveys (in any season) and distance sampling (in winter) might be applied prior to eradication. Alternatively, these methods could also be applied before and after a large portion of the wild boar population is removed. Keuling et al. (2013) suggest that wild population growth can be reduced only if > 65 % of the summer population is harvested. If this proportion was culled in an English woodland, estimating density before and after known numbers of wild boar have been removed would allow managers to determine the accuracy of the method used.This project also indicated that activity sign surveys, mapping of sightings and monitoring of RTAs can be employed as methods to monitor population trends. These indices could also be used to trigger intervention once a set threshold, such as the number of RTAs or the incidence and severity of crop damage in an area have been reached.In summary, the results of this study suggested that local densities of wild boar in England are now comparable to those of other populations in mainland Europe and that a further increase in numbers like the one observed in the Forest of Dean, might be expected if population growth is not prevented. Wild boar in Europe have increased dramatically in the last 30 years (Massei et al. submitted), despite intensive culling. In parallel, the environmental and economic impact of this species, such as crop damage and vehicle collisions has also grown (Schley et al. 2008, Liberg et al. 2010, Imesh-Bebie’ et al 2010). This study refined the tools to monitor wild boar population numbers, spread and large-scale in England: whilst populations are still isolated, landowners in England have the opportunity, unlike most European countries, to use these tools to quantify the impact of intervention on wild boar populations. For these reasons the monitoring of density of wild boar should be maintained in key woodlands, such as the Forest of Dean where wild boar populations are growing fast. In these areas, other methods such as faecal genotyping, developed recently for wild boar population estimates (Ebert et ala. 2012), could be tested and carried out in parallel to camera traps and distance sampling to increase the range of tools available in different contexts and seasons. To calibrate these methods, future research should apply them to areas where a large proportion of the wild boar population is removed in a relatively short timeframe (months) so that the effect of natality and mortality can be minimized. Future studies should also investigate the relationships between wild boar densities and the effort required to decrease population size, prevent population growth or reduce wild boar impact.

Objective 3. Quantify impacts of wild boar on key biodiversity components of the environment.The results of project WM0318 indicated that wild boar rooting activity varies throughout the year but that woodland rides may be hotspots for rooting activity. The aims of this part of the project were to develop methods to assess the impact of wild boar on key plants and invertebrates on woodland rides.The study site used for developing methods to evaluate potential impacts on ground flora and invertebrates had to be re-located from Beckley Woods in Kent to the Forest of Dean, Gloucestershire, during autumn 2012, following management work in Beckley that rendered the rides unsuitable for further vegetation surveys and insect sampling. The Forest of Dean was an equally suitable site, with a resident boar population and an extensive network of open, grassy rides, but the additional time and costs required to travel to and from the Forest meant that some of the more intensive and time-consuming sampling procedures (butterfly walks and timed insect counts) had to be abandoned in favour of stand-alone trapping methods. As a result, statistical analyses were restricted by the level of replication that could be achieved and by interference with the insect traps by people and boar. However, sufficient data were collected to evaluate methodologies and to make recommendations on best practice for future research.

3a. Vegetation assessment

MethodsFive 100 m long transects containing patches that were rooted and not rooted by boar (Fig. 18 and


Fig. 19) were located and marked out along suitable grassy rides during February-March 2013. Transect 1 was located at Crabtree Hill in the New Beechenhusrt Inclosure of the Forest of Dean at OS grid ref. SO633138. Transects 2-5 were located in Saintlow Inclosure near the Forest of Dean arboretum, at OS grid refs. SO630121, SO628118, SO632114 and SO628113, respectively. The distance between transects varied between 100 and 2000 m: for vegetation and invertebrate assessment, transects were regarded as independent from each other. The ends of the transects were marked using spray paint on nearby trees, as the study area had high visitor numbers, particularly through the summer months, and a low visibility method of marking was required to minimise the likelihood of markers being removed or damaged by members of the public.

Fig. 18. Fresh rooting by wild boar on the edge of a grassy ride in the Forest of Dean, March 2013.

Each transect was divided into 50 2m x 2 m contiguous quadrats and, in April, the amount of rooting in each quadrat was recorded using the scoring system developed for WM0318 (Table 13).

Score Amount of rooting (% area)

0 01 <32 4-103 11-254 26-505 51-756 76-100

Table 13. Scoring system using to record the amount of rooting in each quadrat.

Within each transect, three pairs of quadrats were selected for detailed vegetation assessments. One of each pair was set in a recently rooted area, the other in a non-rooted area. Ideally, paired quadrats were adjacent to each other, but similarity of vegetation community was given priority. The position of each selected quadrat was marked on a small scale map showing the full transect. (5 transects x 2 x 3 quadrats = 30 quadrats in total)Monthly assessments of vegetation were carried out in May, June, July and August, but were not continued into September, because after August plants had largely finished flowering and the data indicated that the numbers of species identified had peaked. At each assessment, the whole transect was walked and a cumulative list of all plant species present was compiled and flowering status noted. For each of the 50 quadrats along the transect, the amount of rooting was scored as shown in Table 13. For each quadrat, all plant species were identified, and the percentage area covered by each species was scored using the system in Table 13.The vegetation data were analysed to investigate whether there were differences between rooted and non-rooted quadrats in the number of plant species present, proportions of insect- and wind-


pollinated plant species, proportions of annual, biennial and perennial plant species, and the number of plant species in flower. As this study focused on potential effects upon pollinating insects, grasses, sedges, rushes, trees and shrubs or ferns recorded during assessments were excluded from the final analysis.Differences between mean numbers of plant species in rooted and non-rooted quadrats were investigated using ANOVA. Differences in proportions of annual and perennial plants, insect- and wind-pollinated plants, and number of species in flower were investigated using a generalised linear model. A generalised linear mixed model was used to test for the effects of month, rooting and any interaction upon proportions of annual and perennial plants, proportions of insect and wind pollinated plants, and the number of species in flower. A binomial error distribution was used with a logit link function, and transect was included as a fixed effect.

ResultsThe amount of rooting by boar varied between transects and was highest in transects 2 and 4 (Table 14). The amount and pattern of rooting remained much the same over the summer period.

rooting >25% *

1 1.08 22 2.06 143 0.92 64 3.08 255 1.36 6

* i.e. a rooting score ≥ 4

Table 14. Mean rooting score for each transect in April 2013 and the number of quadrats on each transect that had moderate or extensive rooting, i.e. a rooting score ≥ 4.

A total of 143 plant taxa were recorded on the transects during the study. The highest number of species were recorded on transect 4 (= 89 species including trees, shrubs, grasses, sedges, rushes and ferns). This transect occupied an area where very shallow soils had formed over a former go-kart track and were reminiscent of a typical calcareous grassland flora with plants such as Pilosella officinarum, Lotus corniculatus and Linum catharticum. Alongside these were plants more typical of woodland such as Allium ursinum, Digitalis purpurea and Hyacinthoides non-scripta. Source of

variationd.f. s.s. m.s. v.r. F pr.

Transect 4 645.80 161.45 15.28 <0.001Rooting 1 0.30 0.3 0.03 0.868Transect.Rooting 4 95.53 23.88 2.26 0.099Residual 20 211.33 10.57Total 29 952.97

Table 15. ANOVA on the total number of plant species per quadrat in rooted and non-rooted areas.

In contrast, transect 5, located along the edge of a wide, regularly mown ornamental ride contained the lowest number of species (= 48 including grasses, sedges, rushes, trees, shrubs, and ferns). Species lists for each transect are given in Appendix 3. The variation amongst transects in the total numbers of plant species recorded over the whole season was reflected in significant differences between transects in the mean number of plant species recorded per quadrat (p<0.001, Table 15). There was no significant difference however, in the mean number of plant species recorded in rooted and non-rooted quadrats (p=0.868, Table 15) which ranged from 8.7 to 27.0 in transect 4 and 5, respectively (Table 16).

Transect Mean number of plant speciesper quadrat (± SD)

Rooted Non-rooted

1 13.7 (2.1) 10.7 (3.8)2 14.3 (2.9) 14.7 (2.8)3 14.7 (5.1) 16.3 (1.2)4 21.0 (5.3) 27.0 (1.7)5 12.7 (3.2) 8.7 (1.2)Overall mean 15.3 15.5

Table 16. Mean (SD) number of plant species per quadrat in rooted and non-rooted areas of the five transects.

Regression analysis indicated that overall there was a significant difference in the proportion of annual and perennial plants between rooted and non-rooted quadrats (P = 0.002, Table 17), but that this arose because of significant differences between transects (P < 0.001, Table 18) rather than between rooted/non-rooted quadrats within transects (P = 0.929, Table 18). There were no significant differences in the proportions of insect- and wind-pollinated plants between transects or between rooted and non-rooted quadrats (P = 0.198) (Tables 19 and 20).


Source of variation

d.f. deviance Mean deviance

Deviance ratio

Approx χ² pr

Regression 5 18.1 3.7623 3.76 0.002Residual 24 12.20 0.5083Total 29 31.01 1.0693

Table 17. Summary of regression analysis of the proportion of perennial plant species in rooted and non-rooted quadrats. “Approx. chi pr.” refers to Chi-squared probability.

Parameter estimate s.e. t P Antilog of estimate

Constant 2.598 0.482 5.39 <0.001 13.44Transect 2 -1.810 0.518 -3.49 <0.001 0.1636Transect 3 -0.878 0.547 -1.60 0.109 0.4157Transect 4 -1.364 0.506 -2.70 0.007 0.2557Transect 5 -1.512 0.546 -2.77 0.006 0.2204Rooting 0.021 0.237 0.09 0.929 1.021

Table 18. Parameter estimates from the regression analysis of the proportion of perennial plant species in rooted and non-rooted quadrats.

Analysis of the flowering data found that the number of species in flower was lowest in May and highest in July. However, there was no significant difference in the proportion of species in flower or not in flower between rooted and non-rooted areas (P = 0.290).

Source of variation

d.f. deviance Mean deviance

Deviance ratio

Approx χ² pr

Regression 5 7.32 1.4634 1.46 0.198Residual 24 16.08 0.6700Total 29 23.40 0.8068

Table 19. Summary of regression analysis of proportions of insect-pollinated plant species in rooted and non-rooted quadrats. “Approx. χ² pr.” refers to Chi-squared probability.

Parameter estimate s.e. t P Antilog of estimate

Constant 2.925 0.515 5.68 <.001 18.63Transect 2 -0.715 0.566 -1.26 0.207 0.4894Transect 3 -0.292 0.595 -0.49 0.623 0.7465Transect 4 0.036 0.580 0.06 0.951 1.036


Transect 5 -0.788 0.588 -1.34 0.180 0.4546Rooting -0.512 0.336 -1.52 0.127 0.5993

Table 20. Estimates of parameters for regression analysis of proportions of insect-pollinated plant species in rooted and non-rooted quadrats.

3b. Insect samplingMethodsThe occurrence of flower-visiting insects on the transects, and in rooted and non-rooted areas nearby, was assessed using yellow pan traps (Fig. 19). These were made to a standardised design and were set at 1.3 m above the ground, and contained 250 ml of saturated salt solution to trap and partly preserve the insects captured. Two traps were located along each transect, one in a well-rooted quadrat and the other in a non-rooted quadrat, at least 50 m apart. Two further traps were placed outside the transect, but no more than 100 m away, one in a rooted area and the other in a non-rooted area. (5 transects x 4 traps = 20 traps in total).The insect traps were set up and primed on 30th April and were emptied and re-set every week until the end of June (8 x collections). The samples were returned to Alice Holt, where the insects were sorted and identified. Total numbers of individuals in five key groups representing pollinators and other flower visitors were recorded: i.e. bees (Hymenoptera: Apocrita), hoverflies (Diptera: Syrphidae); sawflies (Hymenoptera: Symphyta), soldier beetles (Coleoptera: Cantharidae) and weevils (Coleoptera: Curculionidae).Differences in the mean numbers of insects caught per trap between rooted and non-rooted areas were tested for significance using ANOVA, applied to untransformed or square-root transformed numbers of insects caught in each trap over the whole of the sampling period. Relationships between insect captures and rooting at the transect level were investigated by regressing total numbers of insects caught per transect against the mean rooting scores for the transects.Several of the traps were damaged or broken during the sampling period, on more than one occasion, mainly by people, but in some instances by boar or other animals. This resulted in the loss of 22 (14%) of the insect samples, but as similar numbers of samples were lost from rooted and non-rooted quadrats (10 and 12 samples, respectively), it is unlikely that the missing samples influenced the comparison of rooted and non-rooted areas. Consequently, the dataset was not adjusted for missing data.

Fig. 19. Yellow pan trap set at 1.3 m above the ground on a non-rooted section of transect 3 (left) and on a rooted section of transect 3 (right) in May 2013.

ResultsAt the trap level, higher numbers of pollinators and flower-visiting insects were caught in non-


rooted patches compared with rooted patches, but differences were not significant (Table 21).of insects. Placing traps lower down, within the vegetation, might have caught insects that were more closely associated with the plants and flowers in the quadrats, but this was not an option because traps on the ground were more likely to be damaged by people and animals. Boar and deer, in particular, appeared to be attracted to the salt in the traps, and placing traps on the ground made the salt more accessible and the trap more likely to suffer interference.Trapping is an indirect method for relating insect numbers to plant abundance and diversity, and only by averaging across a large number of traps and vegetation quadrats are relationships between insect numbers and plant communities likely to be obtained. Direct counts of insects in rooted and non-rooted quadrats, or on particular flowers within the quadrats, are far more likely to reveal relationships between insects and patch vegetation, but such counts are extremely time consuming and they could not be included in the current study once the work was relocated to the Forest of Dean. This would be the method of choice however, for future work.

Objective 4. Explore relationships between wild boar density and environmental impacts

One of the objectives of project WM0318 was to compare wild boar density estimates with the frequency of impacts at woodland (study site) scale. However, wild boar densities observed in the field were low and impact, measured as percentage of study site rooted, was modest. During the current study, the effort to detect wild boar large-scale impact through rooting was doubled to increase the precision and accuracy of impact assessment and to analyse the relationship between impact, habitat type and local density derived from camera traps. In addition, the long-term effects of large-scale impact were analysed by measuring rooting rates at each site twice per year for two years. The objectives of this part of the study were to (i) refine the method developed during project WM0318 to assess the distribution, extent and timing of rooting in English woodlands, (ii) determine the minimum frequency of wild boar rooting in each woodland and (iii) relate the spatial and temporal aspects of rooting to habitat type and to wild boar densities.

MethodsThe method to estimate the large scale impact of wild boar was refined by doubling the

density (compared to project WM0318) of permanent plots per site in the same five study sites described under Objective 1. The distance between each adjacent site was measured by the distance of the closest point of each wood edge to the other: Penyard/Chase- Serridge: 5.7 km; Serridge-Oakenhill: 7.2 km; Beckley/Bixley-Brede: 3.2 km. These areas were regarded as independent sites, as preliminary data on radiotracking, showed that wild boar monthly movements in England were generally restricted to individual woods (Quy et al. 2014).In total, 863 permanent circular plots (10 m radius, plot area = 314 m2) were established and geo-referenced using a Garmin eTrex Venture HC (R) unit; the centre of the plot was marked with red tape to facilitate surveying. Plots were uniformly distributed along a grid overlapped to the study area and placed at approximately one every 1.5-2 ha with a density of 50-66 plots/100 ha. Plots were assigned to the following habitat types: 1. broadleaved woodland and broadleaved plantation, 2. conifer plantation, 3. mixed woodland and 4. recently felled and scrub.Plots were surveyed for rooting signs twice per year, in summer (July-August 2011 and June-July 2012) and in winter (January-March 2012 and January-February 2013), to quantify temporal and spatial patterns of rooting. On each visit, rooting was assessed by recording the extent of the area covered by rooting and the age of rooting (fresh, intermediate, old). Age of rooting was determined by a modification of Killian’s (2009) categories using the same method described in project WM0318. A sketch map was drawn for each rooted plot, to enable subsequent rooting to be compared. Only fresh (< 1 week old) and intermediate (< 1 month old) rooting (referred to as “fresh rooting”) was included in the assessment of impact as old rooting was difficult to quantify. The effort required to visit and record rooting in each plot was quantified in terms of number of plots visited per day.

Generalized Linear Model (GLM) was used to compare the proportion of plots rooted in different habitat types within each site. As the number of plots rooted was found to be relatively


small, data from two consecutive years were pooled by season to obtain the proportion of plots rooted in different habitat types within each site in summer and in winter.The response (rooting / no rooting) at the plot level was binary, so a logit link function was used in the GLM. As the proportions of rooted plots was relatively low, the p-value of the test for differences between habitats was derived using a permutation test. When differences in rooting between habitat types within each wood and season were found statistically significant (i.e. P< 0.05), pairwise comparisons were carried out to compare individual habitat types.

ResultsOut of the 863 plots established in the five study sites, 313 (36.3%) were rooted at least once and 138 (16.0%) were rooted at least twice or more during the 2-year study (Table 23). The proportion of plots freshly rooted in each survey varied between 1.1 and 34.8% (Fig. 20). The proportion of plots rooted between seasons differed in the following sites: Brede (χ2 = 10.80, d.f. = 3, P =0.013) with fewer plots rooted in the summer 2011 than in the other seasons; Oakenhill (χ2 = 30.66, d.f. = 3, P =0.001) and Serridge (χ2 = 11.41, d.f. = 3, P =0.011). In both Oakenhill and Serridge a larger proportions of plots rooted was recorded in the last two seasons compared with the first two seasons, probably reflecting local increases in wild boar densities as established from camera trap surveys.

Site Total n. of plots

N. and % plots rooted at least once in 2 years

N. and % plots rooted once in 2 years

N. and % of plots rooted twice

N. and % of plots rooted 3 times

N. and % of plots rooted 4 times

Beckley/Bixley 186 16 (8.6) 13 (7.0) 3 (1.6) 0 (0.0) 0 (0.0)Brede 169 34 (20.1) 26 (15.4) 7 (4.1) 1 (0.6) 0 (0.0)Oakenhill 135 79 (51.9) 42 (31.1) 18 (13.3) 5 (3.7) 5 (3.7)Penyard/Chase 134 49 (36.6) 28 (20.9) 11 (8.2) 8 (6.0) 2 (1.5)Serridge 239 144 (60.3) 66 (27.6) 42 (17.6) 28 (11.7) 8 (3.3)total 863 313 (36.3) 175 (20.3) 81 (9.4) 42 (4.9) 15 (1.7)

Table 23. Number and proportion (in brackets) of circular plots (area=314 m2) where wild boar rooting occurred during the 2-year study. All plots were surveyed 4 times, in summer 2011, winter 2011-2012, summer 2012 and winter 2012-2013. In each site, the density of plots was 50-66/100 ha.

The percentage of site where rooting occurred, derived from the percentage of plot area rooted per site and season, varied between 0.06% and 8% (Fig. 21).The frequency of habitat types varied between sites. Conifer plots were less numerous in the Weald compared to the Forest of Dean, where they comprised >40% of the plots (Fig. 22). Significant differences in habitat use, assessed by the proportion of plots rooted in each habitat, occurred in Penyard/Chase in summer (χ2 = 23.19, d.f. = 3, P =0.001) and winter (χ2 = 8.79, d.f. = 3, P = 0.041) and in Serridge in summer (χ2 = 33.69, d.f. = 3, P =0.001) and winter (χ2 =28.02, d.f. = 3, P =0.001). Higher proportions of plots were rooted in broadleaves than in other habitat types in both sites (Fig. 22).


Fig. 20. Proportion of plots freshly rooted by wild boar per site and season.

Fig. 21. Proportion of area freshly rooted by wild boar in each plot, derived by the ratio between area rooted per plot and total number of plots in each site and season. In each site, the density of plots was 50-66 /100 ha.

Fig. 22. Percentage of habitat type in each site, derived from the percentage of plots in each habitat, compared to the percentage of plots where rooting was recorded in winter and summer.

There was no correlation between density of wild boar calculated from camera traps and the proportion of plots rooted (Spearman’s rho =0.394; P= 0.26; n =10; Fig. 23).


Fig. 23. Relationship between the proportion of plots freshly rooted and wild boar densities estimated from camera traps.

General discussionThis study refined the method to quantify spatial and temporal patterns of wild boar rooting in England. Circular permanent plots are easy to survey and can be used repeatedly to monitor the extent of rooting, the association between rooting and habitat type and whether rooting re-occurs on the same plot at different times of the year. By extrapolating the rooting recorded to the area covered by a woodland, we established that rooting by wild boar in English woodlands, at current local densities, affected between <1 –3%, and exceptionally ~8% of a wood per annum, depending on season. In agreement with project WM0318, the current study confirmed that wild boar preferred to root in broadleaves stands rather than in other habitat types. This study also suggested that in areas where the wild boar density has increased in recent years, such as Oakenhill and Serridge, the estimated proportion of site rooted increased from < 1% recorded in winter 2009-2010 during project WM0318 to 2.1-4.4% recorded in winter 2012-2013.The percentage area of woodland affected by seasonal rooting in this study is relatively small compared to that recorded for wild boar/feral pigs in other countries. For instance, in Australia, Hone (1988a) found ~2.7% of the study area rooted (ranging between 0.1 and 7.4 % depending on sites and months), in Poland rooting generally did not exceeded ~4% of the study area, although peaks of 10% were recorded in particular seasons (Jezierski and Myrcha 1975), in Sweden Welander (2000) found that rooting affected a total of ~12% of the study area throughout the three years of study, with large variations between years ranging from 1 to 6% of area rooted per year. The six-fold increase in rooted area between years was dependent on food availability, particularly a high abundance of acorns and hazelnuts that led to increased rooting in oak woodlands. In Hawaii, Ralph and Maxwell (1984) and Cooray and Mueller-Dombois (1981) reported ~10% and 14-38% of the study area rooted respectively. Bratton (in Howe et al. 1981) stated that as much as 80% of the surface area of mesic northern hardwoods (USA) was rooted annually, with sites rooted as many as 3-7 times during the growing season (Howe et al. 1981). In Switzerland, Risch et al. (2010) found that the range of forest soil disturbance by wild boar rooting varied between 27 and 54%.Although some studies suggested that the extent of rooting and population density are correlated (Belden and Pelton 1975, Andersson and Stone 1993), others found that the relationship is not linear (Hone 1988a, 1988b, 1995). This seems to be the case in the current project where the extent of rooting was not correlated with the estimated density of wild boar on five sites for two years. The study thus confirmed the preliminary findings of project WM0318 and suggested that the extent of rooting is not correlated with population size.In most study sites surveyed during this project, there were no apparent seasonal (summer versus winter) patterns in rooting activity, though this might be a reflection of the relatively small number of plots rooted in each site during each survey. In addition, the fact that mast production was high in winter 2012-2013 might have affected the extent of rooting. Temporal patterns of rooting have been found in most wild boar studies (Bratton 1975; Dardaillion 1987; Kotanen 1995; Focardi et al. 2000; Welander 2000; Baubet et al. 2003). Patterns of rooting vary according to a combination of food availability and accessibility of below-ground resources. Schley and Roper (2003)


reviewed wild boar diet studies and concluded that the availability of mast (acorns, beechnuts and sweet chestnuts) determines the consumption of other plant components. When energy-rich food such as mast becomes scarce, wild boar must rely on other, less preferred sources : in this case, the proportion of soil disturbed by rooting will vary in relation to the availability of alternative food sources. If the latter are underground, such as roots, rhizomes, invertebrates, etc. the amount of rooting in these areas will increase; if wild boar can feed on crops such as maize, or wild fruits, or invertebrates such as gastropods, the proportion of soil disturbed by rooting will decrease. (Massei et al. 1996, Barrios-Garcia and Ballari 2012). The scale, intensity, frequency and timing of rooting are also influenced by precipitation, habitat type and availability of crops (Genov 1981, Hone 1988a, Kotanen 1995, Welander 2000, Hone 2002). Feeding opportunistically on crops and on other species of plants, vertebrates and invertebrates, wild boar have a direct impact on the flora and fauna of woodlands and on other habitats such as pastures (Howe et al. 1981, Singer et al. 1984, Kotanen 1995, Schley and Roper 2003, Barrios-Garcia and Ballari 2012, Bueno and Jimenez 2014). Whilst rooting at landscape scale may be useful to monitor the extent of disturbance to forest soil by wild boar, it does not necessarily explain the reason for rooting nor the factors affecting this activity. Should wild boar local densities increase, understanding these factors might be the key to focus intervention for the most vulnerable sites. By combining the use of camera trap grid with plot surveys, future research could assess habitat use in relation to different activities of wild boar. In addition, expanding the method of permanent plot surveys to farmlands, could be used to clarify spatial and temporal patterns of rooting in agricultural and woodland landscape. In parallel, repeated surveys over a number of years and seasons, in conjunction with monitoring wild boar density trends as well as diet analysis, will clarify likely impacts and areas of possible concern. In conclusion, this study confirmed that relatively low level of disturbance occurred at most sites, although the overall level of disturbance due to rooting has increased, at least in Gloucestershire. The study also provided and tested a tool that can be used to quantify the spatial and temporal extent of rooting in English woodlands and to assess the impact that feral wild boar have on an environment where they were last present several centuries ago.

Objective 5. Produce best-practice principles on the field deployment of methods developed during this project and on analysis and interpretation of the resultant data.

The following section provides a list of protocols, developed through the current project, to assess large-scale impact of wild boar and to monitor presence, density and population trends. A brief section also provides the estimated effort required to implement these methods in woodlands or parts of woodlands of 180-400 ha that have wild boar densities comparable to those estimated in this study, i.e. between 0.7 and 7 wild boar/100 ha. A decision tree derived from project WM0318 and modified by using the results of the current study is provided at the end of this section, to guide choices of methods for different contexts.

1. Monitor wild boar presence through activity signs on transects or camera traps

Wild boar presence can be assessed through activity signs that include wallows, mud and tusk marks on trees, pellet groups, rooting, tracks and trails. In low density areas, the majority of these signs are too infrequent to be used to determine wild boar presence. Rooting may provide an indication of wild boar occurrence in an area but can vary significantly between seasons: thus the lack of rooting does not necessarily mean that wild boar do not occur in an area. Wild boar trails on forest roads and tracks can also be used as indicators of wild boar presence, although trails are easier to detect during the wet season.

Similarly, camera traps, increasingly employed in wildlife management and conservation, can be used to detect wild boar presence. The results of surveys carried out in five English woodland sites with free-living populations of wild boar, suggest that a density of at least 7 transects/100ha (each


transect 200x1 m) should be used to have > 90% probability of detecting wild boar. Alternatively, a density of at least 9 camera traps/100 ha, left in place at least 9 days, should be used to have > 90% probability of detecting wild boar. It is important to highlight that these results were obtained for woods of 180-400 ha and for densities of wild boar that varied between 0.7 and 7 animals/100 ha.For larger woodlands that cover an area of about 5000 ha, a minimum of 15 camera traps/100 ha, left in place for 10 days, should be used to have a 90% probability of detecting a single wild boar and 20 camera traps/100 ha should be employed to have a 95% chance of detecting a single wild boar.

Adding bait such as maize will increase the probability of attracting wild boar to a site. Ideally, bait should be placed in plastic tubes secured to a tree, with holes that allow the maize to drop out if shaken by wild boar. Plastic tubes will reduce the probability of non-target species (deer, badgers) feeding on the bait.

Studies in English woodlands established that adding the commercially available wild boar attractant called “Buchenholzteer” (based on birch wood tar) to a tree or to a stake, will increase the likelihood of detecting wild boar presence. When this attractant was used on trees, within three weeks wild boar left hair, mud and tusk marks on the tree trunks and rooting was observed around these trees but not on neighbouring trees that had not been treated with the attractant.

2. Wild boar population density indices obtained from activity signs and camera traps Wild boar activity signs, such as trails and rooting can be used for calculating indices of abundance and for monitoring population trends over time. Transects 200x1 m are established at random along forest rides and paths at a density of 10 transects/100 ha. Wild boar trails (or area rooted) are counted and converted into an index of density (PAI= Passive Activity Index) as follows:

PAI = sum of total number of trails (or area rooted) observed/ number of transectsIn addition, it is possible to estimate the proportion of transects with at least 1 wild boar trail.This second number, considered along the mapping of trails on transects, provides an estimate of the spread of wild boar in an area.

Example:Surveyed 12 transects and found the following number of trails:

Transect number 1 2 3 4 5 6 7 8 9 1

0 11 12

N. of boar trails 2 0 0 0 1 3 0 0 1 1 0 2

PAI= Total N. of trails/ Tot N transects= 10/12= 0.83

The proportion of transects with trails can be calculated as :% transects with trails= N transects with trails/ Tot N transects= 6/12= 0.50.

Camera traps, employed at a density of 15 camera traps/100 ha can also be used to calculate PAIs as follows:

PAI= Total N. of wild boar visits / Tot N of camera traps


A “visit” is defined as 1 or more photos of wild boar until there was a lapse of at least 10 minutes between consecutive photos: photos of wild boar taken > 10 minutes apart are counted as a new independent visit as preliminary observations with ear-tagged animals indicated the same animals rarely return to the same area within 10 minutes.

PAI can be used in the same wood to compare density indices between years and monitor wild boar population trends. Statistical comparisons in wild boar density between sites or between years within a site require estimate of variability (such as Standard Deviation and Standard Error) that can be obtained by using basic statistical packages.

PLEASE NOTE: the method of activity signs is effective in winter or during the wet season. In hot weather wild boar trails are much more difficult to detect, due to the relative lack of precipitations. Conversely, camera traps can be used in any part of the year and have the advantage of providing unequivocal proof of wild boar presence but are more expensive (compared to activity signs surveys) and may be stolen.

3. Wild boar density obtained from camera traps

Wild boar density for an isolated population can be estimated as a function of wild boar speed, camera trap features, number of wild boar visits recorded by the camera traps and group size. If the population is not isolated, the method described below can still be used provided the survey is completed within 3-4 weeks, ideally at least 2 months before the peak of births, so that it is possible to assume the effects of immigration and emigration on the population size are negligible.

Group size of wild boar varies throughout the year and must be calculated 2-3 weeks before camera traps are used to quantify wild boar density. Ideally, group size should be based on at least 20-30 independent groups of wild boar (please note: single animals are regarded as “groups”).

The number of camera traps to deploy to assess group size will depend on the local density of animals but should be at least 15, placed in 15 sites located as far as possible from each other to decrease probability of wild boar visiting more than one location. Bait (8-10 kg of maize) should be used at each site, ideally placed in plastic pipes with holes and secured to trees. Camera traps should be left in place until 20-30 independent groups of wild boar have been identified. When calculating group size, all the double counts (groups recognizable and already counted once) must be discarded so that each group is counted only once. If in doubt, discard the group.

Example: 15 camera traps resulted in the following number of wild boar visiting the locations:

Camera trap number

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Mean group size= Tot. N of boar/Tot N of groups

Group size= number

5, 1, 4

0 6, 3

0 1, 2, 8

0 1, 1

4, 1, 4

2 0 6, 4

8, 1, 2

1 2 6 3.18


of boar per group

Mean group size is calculated by the ratio between the total number of boar observed (in the example = 70) and the number of groups (in the example = 22).

Density estimateOnce group size is estimated, place camera traps in a grid with a density of about 15 camera traps/100ha. Camera traps should be left in place for at least 9-10 days.The literature on ungulate density, estimated by using camera traps, recommends a minimum of 250 camera nights per site, based on > 20 camera traps per site (Rovero and Marshall 2009). N. camera nights= n. cameras x n of nights.

Example 25 cameras for 10 nights = 250 camera nightsFor woods that cover <150 ha, managers might achieve a minimum of 250 camera nights by either increasing the number of camera traps to 20, left in place for 2 weeks (=20 x 14 =280 camera nights) or 15 camera traps left in place for 18 days (=15 x 18 =270 camera nights)

PLEASE NOTE: camera traps used when assessing density should not have bait.

The following formula is used to calculate density D of wild boar groups (Rowcliffe et al. 2008):

where y/t = number of visits y per unit time t r and θ= radius (expressed in km) and angle of the camera’s detection area (measured in radians) v = speed of movements.

Multiplying D by group size results in the density of wild boar/100 ha.From radiotracking data, we calculated wild boar speed as 0.274 km/hr, equivalent to 3.56 km per days assuming wild boar are active 13 hours per day.One visit is defined as >1 photos of wild boar until there is a lapse of at least 10 minutes between consecutive photos. Photos of wild boar taken at intervals greater than 10 minutes can be considered independent as indicated by observations with individually marked animals.

Example for calculating wild boar density in a wood:

N. of camera visits= 33N. of camera trap days=315wild boar speed = 0.274 km/hr, i.e. 3.56 km per day (as wild boar are active ~ 13 hours/ day)

v=3.56

r= 0.01829

θ= 0.698 radians

y/t = 33/315= 0.1048

group size= 3.74


D/100 ha= [0.1048 x 3.14/ (3.56 x 0.01829 x 2.698)] x 3.74=7 wild boar/100 ha4. Quantify large scale impact of wild boar on woodlands

This method has been developed to monitor the proportion of woodland affected by wild boar rooting.Permanent plots of 10 m of radius (314 m2) are used and placed in woods at a density of approx. density of 50-66 plots/100 ha. All plots within a wood must be surveyed within 4-6 weeks.For each plot, estimate the area that has been rooted.Example:

YES NO Stop

Density required?

YES NO Stop

Index of abundance (RTAs, activity signs)Camera trap surveyDistance sampling survey

Decision tree of methods to monitor presence and local abundance of wild boar.

ReferencesAcevedo P., Escudero M. A., Muńoz R. and Gortázar C. (2006). Factors affecting wild boar

abundance across an environmental gradient in Spain. Acta Theriol. 51: 327-336.Acevedo P., Vicente J., Höfle U., Cassinello J., Ruiz-Fons F. and Gortázar C. (2007). Estimation of

European wild boar relative abundance and aggregation: a novel method in epidemiological risk assessment. Epidemiol. Infect. 135: 519-527.

Andersen J. (1953). Analysis of the Danish roe deer population based on the extermination of the total stock. Danish Rev. Game Biol. 2: 127-155.

Anderson S. and Stone C.P. (1993). Snaring to control feral pigs Sus scrofa in a remote Hawaiian rain forest. Biol. Cons. 63: 195–201. Apollonio M., Andersen R. and Putman R. (2010). European Ungulates and their Management in

the 21st Century, Cambridge University Press, Cambridge, pp. 1-604.Barrios-Garcia M. N. and Ballari S. A. (2012). Impact of wild boar (Sus scrofa) in its introduced

and native range: a review. Biol. Invas. 14: 2283-2300.Baubet E., Ropert-Coudert Y. and Brandt S. (2003). Seasonal and annual variations in earthworm

consumption by wild boar (Sus scrofa scrofa L.). Wildl. Res. 30:179-186 .Belden R. C. and Pelton M. R. (1975). European wild hog rooting in the mountains of east

Tennessee. In Proc. Ann. Conf. Southeastern Ass. Game Fish Comm 29: 665-671. Bieber C. and Ruf T. (2005). Population dynamics in wild boar Sus scrofa: ecology, elasticity of

growth rate and implications for the management of pulsed resource consumers. J. Appl. Ecol. 42:1203-1213.

Bratton S.P. (1975). The effect of wild boar on the ecology of the gray beech forest. Ecology 56:1356-1366.

Buckland S. T., Anderson D. R., Burnham K. P., Laake J. L., Borchers D. L. and Thomas L. (2001). Introduction to Distance Sampling: Estimating Abundance of Biological Populations Oxford University Press.

Bueno C. G. and Jiménez J. J. (2014). Livestock grazing activities and wild boar rooting affect alpine earthworm communities in the Central Pyrenees (Spain). Appl. Soil Ecol. 83: 71-78.

Burnham K.P. and Anderson D.R. (2002). Model selection and multi-model inference: a practical information-theoretic approach Springer-Verlag, New York.

Campbell T.A., D. B Long and Massei G. (2011). Efficacy of the Boar-Operated-System to deliver baits to feral swine. Prevent. Vet. Med. 98:243-249.


http://www.amazon.com/s/ref=ntt_athr_dp_sr_6?_encoding=UTF8&field-author=Len%20Thomas&search-alias=books&sort=relevancerank

http://www.amazon.com/s/ref=ntt_athr_dp_sr_5?_encoding=UTF8&field-author=D.%20L.%20Borchers&search-alias=books&sort=relevancerank

http://www.amazon.com/s/ref=ntt_athr_dp_sr_4?_encoding=UTF8&field-author=J.%20L.%20Laake&search-alias=books&sort=relevancerank

http://www.amazon.com/s/ref=ntt_athr_dp_sr_3?_encoding=UTF8&field-author=K.%20P.%20Burnham&search-alias=books&sort=relevancerank

http://www.amazon.com/s/ref=ntt_athr_dp_sr_2?_encoding=UTF8&field-author=D.%20R.%20Anderson&search-alias=books&sort=relevancerank

http://www.amazon.com/s/ref=ntt_athr_dp_sr_1?_encoding=UTF8&field-author=S.%20T.%20Buckland&search-alias=books&sort=relevancerank

Cooray R. G. and Mueller-Dombois D. (1981). Feral pig activity.In: Island ecosystems: biological organization in selected Hawaiian communities. Hutchinson Ross, Stroudsburg, PA: 309-317.

Dardaillon M. (1987). Seasonal feeding habits of the wild boar in a Mediterranean wetland the Camargue (Southern France). Acta Theriol. 32: 389-401.

Defra 2005. Feral wild boar in England: a consultation by the Department for Environment, Food and Rural Affairs. Defra Publications, London.

Ebert C., Knauer F., Spielberger B., Thiele B. and Hohmann U. (2012). Estimating wild boar Sus scrofa population size using faecal DNA and capture-recapture modelling. Wildl Biol 2: 142-152.

Efron B. (2000). The Bootstrap and Modern Statistics. J Am Stat Assoc 95: 1293-1296.Engeman R.M., Constantin B.U., Nelson M., Woolard J. and Bourassa J. (2001). Monitoring

changes in feral swine population and spatial distribution of activity. Environ. Cons. 28:235-240.Engeman R., Pipas M., Gruver K., Bourassa J. and Allen L. (2002). Plot placement when using a

passive tracking index to simultaneously monitor multiple species of animals. Wildl. Res. 29:85-90.

Engeman R.M,. Smith H.T., Schwiff S.A., Constantin B,. Woolard J., Nelson M. & Griffin D. (2003). Prevalence and economic value of feral swine damage to native habitat in three Florida state parks. Biol. Cons. 30: 319-324.

Engeman R. (2005) Indexing principles and a widely applicable paradigm for indexing animal populations. Wildl. Res. 32: 202–210.

Fiske I. and Chandler R. (2011). An R package for fitting hierarchical models of wildlife occurrence and abundance. J. Stat. Softw. 43: 1-23.

Focardi S., Isotti R. and Tinelli A. (2002). Line transect estimates of ungulate populations in a Mediterranean forest. J Wildl. Manage. 66:48-58.

Franzetti B. and Focardi S. (2006). La stima di popolazione di ungulati mediante distance 367 sampling e termocamera a infrarossi. Ist. Naz. Fauna Selvatica, Documenti Tecnici, 26:1-88.

Franzetti B., Ronchi F., Marini F., Scacco M., Calmanti R., Calabrese A., Aragno P., Montanaro P. and Focardi S. (2012). Nocturnal line transect sampling of wild boar (Sus scrofa) in a Mediterranean forest: long-term comparison with capture–mark–resight population estimates. Eur J Wildl. Res. 58: 385-402.Gardner B., Reppucci J., Lucherini M. and Royle A.J. (2010). Spatially explicit inference for

open populations: estimating demographic parameters from camera-trap studies. Ecology 91: 3376-3383.

Geisser H. and Reyer H. U. (2005). The influence of food and temperature on population density of wild boar Sus scrofa in the Thurgau (Switzerland). J. Zool. 267: 89-96.

Genov P. (1981). Significance of natural biocenoses and agrocenoses as the source of food for wild boar (Sus scrofa L.). Ekol Pol 29 :117-136.

Gerber B., Karpanty S.M., Crawford C., Kotschwar M. and Randrianantenaina J. (2010). An assessment of carnivore relative abundance and density in the eastern rainforests of Madagascar using remotely-triggered camera traps. Oryx 44: 219-222.

Gill R. and Brandt G. (2010). Thermal imaging to estimate wild boar density. Forest Research, Wrecclesham, Surrey.

Gill R. (2014). Wild Boar and Deer in the Forest of Dean. Forest Research, Forestry Commission. http://www.forestry.gov.uk/pdf/WildBoarDeerCensusDean2014.pdf/$FILE/WildBoarDeerCensusDean2014.pdf (Accessed 8 September 2014).

Gill R., Thomas M.L. and Stocker D. (1997). The use of portable thermal imaging for estimating deer population density in forest habitats. J. Appl. Ecol. 34:1273-1286.

Haroldson B.S., Wiggers E.P., Beringer J., Hansen L.P. and McAninch J.B. (2003). Evaluation of aerial thermal imaging for detecting white-tailed deer in a deciduous forest environment. Wildl. Soc. Bull. 31:1188-97.

Hebeisen, C. and Bshary, R. (2007). Population size, density ans dynamics, and social organization of wild boar (Sus scrofa) in the Basin of Geneva.

Hebeisen, C., Fattebert, J., Baubet, E. and Fischer, C. (2008). Estimating wild boar (Sus scrofa) abundance and density using capture–resights in Canton of Geneva, Switzerland. Eur J Wildl. Res. 54: 391-401.


http://www.esajournals.org/doi/pdf/10.1890/09-0804.1

http://www.esajournals.org/doi/pdf/10.1890/09-0804.1

Holland E.P., Burrow J.F., Dytham C. and Aegerter J.N. (2009). Modelling with uncertainty: Introducing a probabilistic framework to predict animal population dynamics. Ecol Model 220 : 1203-1217.

Hone J. (1988a) Evaluation of methods for ground survey of feral pigs and their signs. Acta Theriol. 33: 451-465.

Hone J. (1988b) Feral pig rooting in a mountain forest and woodland: Distribution abundance and relationships with environmental variables. Austr. J. Ecol. 13: 393-400.

Hone J. (1995). Spatial and temporal aspects of vertebrate pest damage with emphasis on feral pigs. J. Appl. Ecol. 32: 311-319.

Hone J. (2002). Feral pigs in Namadgi National Park Australia: dynamics impacts and management. Biol. Cons. 105: 231-242.

Hone J. and Martin W. (1998). A study of dung decay and plot size for surveying feral pigs using dung counts. Wildl. Res. 25: 255-260.

Hounsome T.D., Young R.P., Davison J., Yarnell R.W., Trewby .ID., Garnett B.T., Delahay R.J. and Wilson G.J. (2005). An evaluation of distance sampling to estimate badger ( Meles meles ) abundance. J. Zool. 266: 81-87.

Howe T., Singer J.S. and Ackerman B.B. (1981). Forage relationships of European wild boar invading northern hardwood forest. J. Wildl. Manage. 45: 748-754.

Ickes K., Dewalt S.J. and Appanah S. (2001). Effects of native pigs (Sus scrofa) on woody understorey vegetation in a Malaysian lowland rain forest. J. Tropic. Ecol. 17: 191-206.

Imesh-Bebie’ N., Gander H. and Schnidrig-Petrig R. (2010). Ungulates and their management in Switzerland, in European Ungulates and their Management in the 21st Century, ed. by Apollonio M, Andersen R and Putman R. Cambridge University Press, Cambridge.

Jędrzejewska B., Okarma H., Jędrzejewski W. and Miłkowski L. (1994). Effect of exploitation and protection on forest structure, ungulate density and wolf predation in Białowieża Primeval Forest, Poland. J. Appl. Ecol. 31: 664-676.

Jezierski W. and Myrcha A. (1975). Food requirements of a wild boar population. Pol. Ecol. Stud. 1: 61-83.

Keuling O., Baubet E., Duscher A., Ebert C., Fischer C., Monaco A., Podgórski T., Prevot C., Ronnenberg K., Sodeikat G., Stier N. and Thurfjell, H. (2013). Mortality rates of wild boar Sus scrofa L. in central Europe. Eur J Wildl. Res. 59: 805-814.

Killian G. (2009). Development Implementation and Evaluation of Management Tools to Reduce Feral Swine Damage at Avon Park. AFR Florida Annual Report Year 1 August 15..

Kotanen P.M. (1995). Responses of vegetation to a changing regime of disturbance: effects of feral pigs in a Californian coastal prairie. Ecography 18: 190-199.Liberg O., Bergstrom R., Kindberg J. and Von Essen H. (2010). Ungulates and their management

in Sweden, in European Ungulates and their Management in the 21st Century, ed. by Apollonio M, Andersen R and Putman R, Cambridge University Press, Cambridge, pp. 37-70.

Massei G., Genov P.V. and Staines B.W. (1996). Diet, food availability and reproduction of wild boar in a Mediterranean coastal area. Acta Theriol. 41 :307-320.

Massei G., Coats J., Quy R., Storer K. and Cowan D. P. (2010). The BOS (Boar-Operated-System): a novel method to deliver baits to wild boar. J. Wildl. Manage. 74:333-336.

Massei G., Roy S. and Bunting R. (2011). Too many hogs? A review of methods to mitigate impact by wild boar and feral pigs. Human-Wildl. Interact. 5 :79-99.

Massei G., Kindberg J., Licoppe A., Šprem N., Kamler J., Baubet E., Hohmann U., Monaco A., Ozoliņš J., Cellina S., Podgórski T., Fonseca C., Markov N., Pokorny B., Rosell C. and Náhlik A. (submitted). Wild boar populations up, numbers of hunters down? A review of trends and implications for Europe. Pest Manage. Sci.

Mayle B., Peace A. And Gill, R. (1999). How Many Deer? A guide to estimating deer population size. Forestry Commission field book 18. Forestry Commission, Edinburgh. 96p.

Parkes J.P., Ramsey D.S., Macdonald N., Walker K., McKnight S., Cohen B.S. and Morrison S.A. (2010). Rapid eradication of feral pigs from Santa Cruz Island, California. Biol. Cons 143: 634-641.

Parmenter, R.R. Yates T.L., Anderson D.R., Burnham K.P., Dunnum J. L., Franklin A. B.,


http://link.springer.com/search?facet-author=%22Tomasz+Podg%C3%B3rski%22

http://journals.cambridge.org/abstract_S0952836905006606

http://journals.cambridge.org/abstract_S0952836905006606

Friggens M. T., Lubow B. C., Miller M., Olson G.S., Parmenter C. A., Pollard J., Rexstad E., Shenk T. M., Stanley T.R. and White G.C. (2003). Small-mammal density estimation: a field comparison of grid-based vs. web-based density estimators. Ecol. monogr. 73: 1-26.

Pihal R., Kamler J. and Homolka M. (2010). Comparison of four methods for estimation of wild boar population density in forest environment. 8th International Symposium on Wild boar and Other Suids York, UK1-4 September.

Quy R. J., Massei G., Lambert M.S., Coats J., Miller L.A. and Cowan D.P. (2014). Effects of a GnRH vaccine on the movement and activity of free-living wild boar (Sus scrofa). Wildl. Res. 41: 185-193.

Ralph C. J. and Maxwell B.D. (1984). Relative effects of human and feral hog disturbance on a wet forest in Hawaii. Biol. Cons. 30: 291-303.

Risch A. C., Wirthner S., Busse M. D., Page-Dumroese D. S. and Schütz M. (2010). Grubbing by wild boars (Sus scrofa L.) and its impact on hardwood forest soil carbon dioxide emissions in Switzerland. Oecologia 164: 773-784.

Rovero F. and Marshall A. R. (2009). Camera trapping photographic rate as an index of density in forest ungulates. J. Appl. Ecol. 46: 1011-1017.

Rowcliffe J. M., Field J., Turvey S. T. and Carbone C. (2008). Estimating animal density using camera traps without the need for individual recognition. J. Appl. Ecol. 45: 1228-1236.

Royle J. A., Dawson D. K. and Bates S. (2004). Modeling abundance effects in distance sampling. Ecology 85: 1591-1597.

Royle J.A., Nichols J.D., Karanth K.U. and Gopalaswamy A.M. (2009). A hierarchical model for estimating density in camera-trap studies. J. Appl. Ecol. 46:118-127.

Schley L. and Roper T.J. (2003). Diet of wild boar Sus scrofa in Western Europe, with particular reference to consumption of agricultural crops. Mamm. Rev. 33 :43-56.

Schley L., Dufrêne M., Krier A. and Frantz A.C. (2008). Patterns of crop damage by wild boar (Sus scrofa) in Luxembourg over a 10-year period. Eur. J. Wildl. Res. 54 :589-599.

Singer F.J., Swank W.T. and Clebsh E.E.C. (1984) Effects of wild pig rooting in a deciduous forest. J. Wildl. Manage. 48: 464-473.

Smiet A.C., Fulk G.W. and Lathiya S.B. (1979). Wild Boar Ecology in Thatta District: A Preliminary Study. Pakistan J. Zool.11 : 295-302.

Sweitzer R.A., Van Vuren D., Gardner I.A., Boyce W.M and Waithman J.D. (2000). Estimating sizes of wild pig populations in the North and Central Coast regions of California. J. Wildl. Manage. 64:531 543.

Smart J. C., Ward A. I. and White, P. C. (2004). Monitoring woodland deer populations in the UK: an imprecise science. Mamm. Rev. 34: 99-114.

Welander J. (2000). Spatial and temporal dynamics of wild boar (Sus scrofa) rooting in a mosaic landscape. J Zool Lond 252 :263-271.

Wilson C.J. (2005). Feral wild boar in England: status, impact and management.Defra, London, UK.

Wilson CJ. (2014). The Establishment and Distribution of Feral Wild Boar (Sus scrofa L.) in England. Wildl. Biol. Pract. 10 :1-6.

Wegge P. and Storaas T. (2009). Sampling tiger ungulate prey by the distance method: lessons learned in Bardia National Park, Nepal. Anim. Conserv. 12: 78-84.


Appendix 1. Summary of fieldwork carried out during the study.Several parts of the study were carried out at five main study sites in England: 1. Beckley/Bixley in Sussex Weald, 2. Brede High Woods in East Sussex, 3. Penyard and Chase Woods (Ross-on-Wye), 4. Serridge (north Forest of Dean) and 5. Oakenhill (south Forest of Dean). Study sites 1-2 are individual woodlands, study sites 4 and 5 constitute parts of the Forest of Dean woodland complex. Additional study sites (specified in the tables) were used to complete parts of Objectives 1, 2 and 3.

Objective 1. Refine cost-effective methods for detecting wild boar and quantifying range expansion.

Site name Date Aim Method20 woodlands around the

Nov-Dec 2011Nov 2013

Bait stations with camera traps to

One bait station and 2 camera traps and maize (as bait) placed in each


20 sites in the Forest of Dean and in Penyard woods where boar activity signs occurred

May and Aug 2013

Test putative site-attractants to detect wild boar presence in new areas.

At each site, 2 stakes placed 200 m apart; one stake treated with putative attractant and one with water. Maize was placed next to the stake. Camera traps recorded wild boar behaviour for 2 weeks.

8 woodlands around the Forest of Dean

Dec 2013 Test putative site-attractants to detect wild boar presence in new areas.

In each woodland, 10 pairs of trees selected, one treated with the putative attractant and one used as control. Activity signs on or around all trees recorded 1, 2 and 4 weeks after the trees were treated with attractant.

Objective 2. Refine cost-effective methods to quantify wild boar density and abundance.

Site name Date Aim MethodAll 5 study sites

Winter 2011-‘12Winter 2012-‘13

Quantify wild boar population trends from activity signs on transects.

At each site, 10 transects/100 ha surveyed for wild boar trails and rooting. For ecah site, results expressed as Passive Activity Index

All 5 study sites

Winter 2011-‘12Winter 2012-‘13

Quantify wild boar population trends and estimate density using camera traps.

At each site, 16 camera traps/100 ha left for 9 days and moved North-Southwards to complete a site survey in 18-27 days. For ecah site, results expressed as Passive Activity Index and as a density of wild boar/100 ha.

Beckley/Bixley (2009)Penyard (2010)Forest of Dean (2013)Alto Merse (Italy, 2011 and 2012)

Distance sampling and thermal imaging to estimate wild boar densities.

For ecah study site, the number of wild boar groups observed from transects and the distance of the animals from the transects were recorded. Detection functions were calculated and the results used to estimate wild boar population size.

Objective 3. Quantify impacts of wild boar on key biodiversity components of the environment.


Site name Date Aim MethodForest of Dean Feb-Aug 2013 Assess impact of

wild boar rooting on vegetation.

Five 100 m transects with rooted and non- rooted areas located along grassy rides. Three pairs of 2m x 2m quadrats for each transect selected for monthly assessments of rooting and vegetation.

Forest of Dean Apr-Jun 2013 Assess impact of wild boar rooting on insects.

Five 100 m transects as above, with 2 pan traps per transect (one in rooted and one in non-rooted areas) and another 2 pan traps within 100 m from each transect, one in rooted and one in non-rooted areas. Insects collected weekly.

Objective 4. Explore relationships between wild boar density and environmental impacts

Site name Date Aim MethodAll 5 study sites

Summer 2011Winter 2011-‘12Summer 2012Winter 2012-‘13

Explore relationship between wild boar density and large-scale environmental impact.

At each site, permanent circular plots (10m radius) surveyed in summer and winter to assess % of area rooted. The % of plots with rooting was regressed against wild boar densities estimated by camera traps. Data on large-scale rooting in summer and winter were used to quantify temporal and spatial patterns of rooting.

Appendix 2. Conceptual flowchart of the model used to simulate density estimates through camera traps and distance sampling. Values for k, x1 and x2 are given in Table 12.


Appendix 3. Plant species identified on each transect

TransectSpecies 1 2 3 4 5


Agrostis canina XAgrostis capillaris XAgrostis sp. X XAlchemilla sp. XAllium ursinum XAnemone nemorosa XAngelica sylvestris X X X XAnthoxanthum odoratum X X XAphanes arvensis X XArabis hirsuta XArrhenatherum elatius X X XBellis perennis X X X XBetula sp. X XBlechnum spicant X XBrachypodium sylvaticum X X X XCapsella bursa-pastoris XCardamine flexuosa X X X XCardamine pratensis XCarex divulsa X X XCarex otrubae XCarex pendula XCarex spicata XCarex sylvatica X X XCentaurea nigra XCentaurium erythraea X XCerastium fontanum X X X XCerastium holostea XCircea lutetiana X X XCirsium arvense X XCirsium palustre XConopodium majus XCrataegus monogyna X XCrepis capillaris X X X XCynosurus cristatus X X X XDactylis glomerata X X X X XDeschampsia cespitosa X X X XDeschampsia flexuosa XDigitalis purpurea X X X XEpilobium montanum XEpilobium roseum XEpilobium sp. XEpilobium tetragonum XEuphorbia amygdaloides XEuphrasia sp. X XFagus sylvatica XFestuca rubra X XFragraria vesca XGalium aparine X X XGalium mollugo X X X XGalium saxatile XGeranium dissectum XGeranium molle X X XGeranium robertianum X X X XGeum urbanum X XGlechoma hederacea X XGnaphalium sp. XHolcus lanatus X X X X X


Hyacinthoides non-scripta X X X X XHypericum humifusum XHypericum perforatum X X XHypericum pulchrum XHypochaeris radicata X X XIlex aquifolium XJuncus acutiflorus XJuncus articulatus X X XJuncus conglomerata XJuncus effusus X X XJuncus inflexus XJuncus tenuis X X X XLapsana communis X X X XLathyrus pratensis X X XLinum catharticum X X X XLolium perenne X X X XLotus corniculatus X XLotus pedunculatus X X X X XLuzula sp. XLuzula sylvatica XLychnis flos-cuculi XLysimachia nemorum X XLysimachia nummularia X XMatricaria matricarioides XMedicago lupulina X X XMentha aquatica XMentha arvensis XMentha sp. XMentha suaveolens X XMyosotis arvensis XOdontites vernus X X X XOriganum vulgare XOxalis acetosa XPicea abies XPilosella officinarum X XPlantago lanceolata X X X X XPlantago major X X X X XPoa annua X XPoa trivialis X XPolygala serpyllifolia XPolygala vulgaris XPotentialla reptans XPotentilla anserina X X X XPotentilla erecta X XPotentilla reptans X X X XPotentilla sterilis XPrunella vulgaris X X X X XPteridium aquilinum X X X XPulicaria dysenterica X XQuercus sp. X XRanunculus acris X X XRanunculus ficaria XRanunculus repens X X X X XRubus fruticosus agg. X X X X XRubus idaeus XRumex acetosa XRumex acetosella X


Rumex crispus X X XRumex sp. X X XSagina procumbens XSalix sp. XScrophularia sp. XSenecio jacobea X XSenecio sp. X XSilene dioica XSorbus aucuparia XStachys sylvatica X XStellaria graminea X XStellaria holostea X X X X XTaraxacum agg. X X X X XTorilis japonica X X X X XTorillis japonicaTrifolium dubium X X X XTrifolium pratense X X X XTrifolium repens X X X X XTrifolium sp. XTrifolium/MedicagoUlex europaeus XUrtica dioica X X X XVeronica becabunga XVeronica chamaedrys X X X XVeronica officinalis XVeronica serpyllifolia X X XVicia sativa X X XViola riviniana X XViola sp. X X


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.


General enquiries on this form should be made...

Documents

Transcript of General enquiries on this form should be made...