Characterizing patterns of animal movement is a major aim in population ecology, and yet doing so at an appropriate spatial scale remains a major challenge. Estimating the frequency and distances of movements is of particular importance when species are implicated in the transmission of zoonotic diseases.
European badgers (Meles meles) are classically viewed as exhibiting limited dispersal, and yet their movements bring them into conflict with farmers due to their potential to spread bovine tuberculosis in parts of their range. Considerable uncertainty surrounds the movement potential of badgers, and this may be related to the spatial scale of previous empirical studies. We conducted a large-scale mark–recapture study (755 km2; 2008–2012; 1935 capture events; 963 badgers) to investigate movement patterns in badgers, and undertook a comparative meta-analysis using published data from 15 European populations.
The dispersal movement (>1 km) kernel followed an inverse power-law function, with a substantial ‘tail’ indicating the occurrence of rare long-distance dispersal attempts during the study period. The mean recorded distance from this distribution was 2·6 km, the 95 percentile was 7·3 km and the longest recorded was 22·1 km. Dispersal frequency distributions were significantly different between genders; males dispersed more frequently than females, but females made proportionally more long-distance dispersal attempts than males.
We used a subsampling approach to demonstrate that the appropriate minimum spatial scale to characterize badger movements in our study population was 80 km2, substantially larger than many previous badger studies. Furthermore, the meta-analysis indicated a significant association between maximum movement distance and study area size, while controlling for population density. Maximum long-distance movements were often only recorded by chance beyond the boundaries of study areas.
These findings suggest that the tail of the badger movement distribution is currently underestimated. The implications of this for understanding the spatial ecology of badger populations and for the design of disease intervention strategies are potentially significant.
Animal movement is one of the most important phenomena in nature (Holyoak et al. 2008; Nathan et al. 2008). Knowledge of animal movement affects our understanding of fundamental ecological processes – for example genetic diversity, population viability and range expansion – and has a major impact on how many wildlife management and conservation issues are addressed (e.g. attempts to control wildlife disease or landscape management for migrating species; Cullingham et al. 2008; Clobert et al. 2012). In general, ecological investigations tend to concentrate on common population-level events, while rarer events within populations are often ignored or assumed to be unimportant (Trakhtenbrot et al. 2005; Holyoak et al. 2008). Long-distance dispersal attempts (LDDs) are, by their nature, infrequent but can be extremely influential (Trakhtenbrot et al. 2005). For example, LDDs determine the velocity of species spread across a landscape (Trakhtenbrot et al. 2005), which have implications for the ability of species to track their climatic tolerance under a rapidly changing climate (Schloss, Nunez & Lawler 2012). Despite substantial advances in our theoretical understanding of dispersal (Clobert et al. 2012), and decades of empirical studies across a broad range of taxa, estimating long-distance movements generally and LDDs in particular remains a major challenge.
In general, LDDs are important because they can connect disparate populations, allowing for genetic connectivity, the colonization of vacant habitat across changing landscapes and maintain global persistence in the face of local extinctions (Baguette 2003; Trakhtenbrot et al. 2005; Ronce 2007; Schloss, Nunez & Lawler 2012). Over longer temporal scales, LDDs can also be a major driver of evolutionary speciation (Ronce 2007 and references therein).
The occurrence of LDDs in animal or plant populations that harbour infectious disease may also result in the rapid spread of disease (Cullingham et al. 2008; Mundt et al. 2009). While lasting effects of dispersal on genetic diversity require successful reproduction (Trakhtenbrot et al. 2005), there is no such absolute requirement for the spread of disease, providing there is a susceptible population and the possibility of a transmission event occurring. Zoonotic disease transmission to domestic animals and humans is currently a major issue in many parts of the world (Pech et al. 2010). In disease control programmes and field trials, the movement of infectious animals into an area under disease control is often minimized through the use of barriers (cordon sanitaire; Sleeman et al. 2009; Pech et al. 2010). These barriers can be physical, such as rivers (e.g. Sleeman et al. 2009), or biological, such as buffer zones where the host species is vaccinated (e.g. Pech et al. 2010). When designing biological barriers to immigrating disease hosts, an understanding of the movement ability of the host, including the ability of the host to pass through neighbouring territories, is required. Often, the species dispersal kernel may inform an appropriate width for such a buffer (Pech et al. 2010).
Badger (Meles meles Linnaeus) movements are of intense interest from pure and applied disease ecology perspectives. Despite their importance, there is little known about the frequency LDDs made by badgers, or whether they differ amongst groups within populations (e.g. sex biases). Like other philopatric mammals, dispersal in the badger may be important in avoiding inbreeding depression and may be modulated by competition or facilitation amongst conspecifics (Greenwood 1980; Macdonald et al. 2008). The frequency and length of dispersal movements may be effected by density dependence (DD; Matthysen 2005), which can be a positive or negative process. A positive DD relates to the movement of animals from high densities to lower densities, with the consequence of reduced competition, increased chance of survival and reproductive output. Negative DD suggests that high densities result in lower dispersal propensity due to the groups being socially constrained (packed) by aggressive neighbours, or where there are benefits to individuals to aggregate at higher densities (Le Galliard, Ferriere & Dieckmann 2005; Matthysen 2005). The length of dispersals can also be affected by the socio-spatial distribution of territories and the distribution of resources (mates and feeding patches), which may affect groups (e.g. genders, age classes) within populations differentially (Ronce 2007).
Badgers are an important wildlife reservoir of Mycobacterium bovis, the causative agent of bovine tuberculosis (bTB) in cattle, in Britain and Ireland (Griffin et al. 2005; Bourne et al. 2007) and a bTB spill-over species in other countries (Spain and France; Gortázar et al. 2012). Badger movements and dispersal attempts may facilitate the spread and maintenance of bTB within badger populations (Rogers et al. 1998; Vicente et al. 2007; Riordan et al. 2011) and also the transmission to, and from, cattle (Donnelly et al. 2006). Population reduction through culling has been studied in both Britain and Ireland as a means to reduce intraspecific and interspecific transmission of bTB amongst badger and cattle populations (Griffin et al. 2005; Bourne et al. 2007). Research has found that culling can disrupt the territorial behaviour of badgers (increasing movement) that evade capture and neighbouring non-culled populations (Tuyttens et al. 2000a). It has been hypothesized that this perturbation of badger populations can increase the risk of disease spread from badgers to cattle herds (Donnelly et al. 2006; Macdonald, Riordan & Mathews 2006; Carter et al. 2007). In high-density badger populations, there is evidence of increased bTB risk to cattle herds up to 2 km from a core badger removal area, for up to 2 years post-cull (Donnelly et al. 2006; Jenkins, Woodroffe & Donnelly 2010). Due to these findings, some proposed strategies to tackle bTB in badgers and cattle include badger culling in core areas and vaccinating badgers in a ring of 2 km in width around the culled area (Smith, McDonald & Wilkinson 2012). Other studies have used culling buffers of 3 km or more to reduce inward badger migration (Griffin et al. 2005; Sleeman et al. 2009). Currently, there are no estimates of the proportion of badgers that could move distances >2 or 3 km as an empirically derived dispersal kernel has yet to be described. While the efficacy of such buffer zones will depend on many factors (such as culling efficacy, starting density, natural barriers to movement), the dispersal kernel would allow one to estimate an appropriate minimum distance required to ensure site independence to a defined acceptable threshold (Pech et al. 2010).
Here, we use the largest spatial-scale mark–recapture study of European badgers hitherto undertaken to investigate movement patterns in a medium-density population. We used a repeat subsampling technique to estimate an appropriate minimum spatial scale for badger movement studies in medium-density populations. We fitted two dispersal kernels to estimate the movement probabilities at various distances. Comparisons between dispersal kernels were used to test: (i) whether badger dispersal is better characterized by a fat-tailed distribution over an exponentially bounded distribution and (ii) whether there was a gender difference between dispersal distributions.
We hypothesized that there would be a relationship between study area size and maximum distance moved by badgers reported in the literature and that this relationship may be confounded by estimated local density. Thus, we reviewed and modelled published data from studied badger populations across Europe to assess whether there is a relationship between study area size and LDDs, while controlling for density. We discuss the implications of the findings for our understanding of badger ecology and programmes aimed at controlling the spread and maintenance of bTB within badger populations.
Materials and methods
The study area was located in north-western Co. Kilkenny, Republic of Ireland (52°40′ N; 7°24′ W), and extended over 755 km2. The eastern boundary of the study area was delineated by the River Nore; a large river (~30 m width at Kilkenny City; Mean Annual Flow: 27·1 m3 s−1) that may represent an impediment (but not a barrier) to badger movement, with the exception of bridging points (Sleeman et al. 2009). Other boundaries were made up of smaller rivers or roads (non-badger impediments). The land is low-lying at 60–180 m ASL, and soils are predominantly rich and well drained. Approximately 75% of Co. Kilkenny is farmed (improved grasslands and tillage) and divided by an extensive hedgerow network, while 9·8% of land cover is forested. The mean population density is 46 persons km−2.
Estimated mean badger density during the study period (based on mark–recapture models) was approximately 1 badger km−2 (true population densities may be slightly higher if ‘untrappable’ badgers are present; Byrne et al. 2012). Currently, part of the badger population outside of the study area in the Republic of Ireland is culled (see Byrne et al. 2013a,b), though no culling has taken place in the area for 2 years prior to the start of this study (Byrne et al. 2012). Approximately one-third of this area (253 km2) was part of a reference area in the Four Area Project (a large-scale TB-related experimental project), where culling in response to herd breakdowns was limited during the years 1997–2002 (0·06 km−2 year−1; Griffin et al. 2005). Badger populations can be disrupted for a period of time after culling, for example 3 years for a population of 4–8 badger km−2 in Britain (Tuyttens et al. 2000b). Given the lower densities in Ireland, we expected that a 2-year no-cull period would allow the population to recover to approximately pre-removal densities. However, it is possible that the population structure was still recovering during the initial period of the study.
An extensive mark–release–recapture (MRR) programme was undertaken at the study site between September 2009 and June 2012. All badger setts (complex subterranean burrow systems) known from surveys were visited during six separate ‘sessions’, each taking 20–24 weeks to complete, where an assessment was made of sett activity. Setts were considered ‘active’ using combinations of multiple cues [presence of fresh spoil, bedding, etc. and the absence of vegetation growth within the sett entrance (Byrne et al. 2012, 2013a)]. Each active sett was trapped for eight nights of capture effort (Monday night–Thursday night over two sequential weeks) using a combination of stopped body restraints (minimum closure: 28 cm) and steel wire mesh cage-traps (see Byrne et al. 2012). Traps were checked daily, as early as possible, and always before 12 pm to limit stress for any captured animals. The capture histories of badgers marked and released during an additional training period between June 2008–August 2009 were also included in the analysis (n = 122 marked badgers during this period; further details in Byrne et al. 2012).
Captured badgers were anaesthetized with ketamine hydrochloride (0·1 ml kg−1) and medetomidine (0·1 ml kg−1, Domitor®; Pfizer, Dublin, Ireland) administered by intramuscular injection by an onsite veterinarian (Murphy et al. 2010). When first captured, each badger was implanted with an identifying passive transponder and tattooed with a unique number in the inguinal region. Badgers were placed in a wooden box with a ‘flap’ doorway beside the sett of capture during recovery from the anaesthetic. This box was dark and provided protection for the animal during the recovery phase. Marked dead badgers [recoveries; e.g. due to road traffic accidents (RTAs)] inside or outside the study area were recorded.
Movements were measured as the Euclidean distance (D) between recaptures (or encounter points when using recoveries). True movement distances are likely to be substantially greater as badgers typically move in tortuous routes (e.g. Loureiro et al. 2007). Also, there is a low temporal resolution within this data set, and therefore, the methodology is insensitive to detect temporary/transient movements. The data collection has increased extent, and sample size, to the detriment of temporal resolution. However, this does not invalidate our inferences as (i) the underlying process of badger dispersal may take long periods of time (Roper, Ostler & Conradt 2003), (ii) we make inferences at the population scale (the scale at which the data were collected) (Cushman 2010), and (iii) one of the movement models (inverse power function) is robust against marking frequency and effort and can be used to make reliable proportional dispersal estimates (Fric & Konvicka 2007). There is the potential for badgers to make frequent short-term forays, and this could affect disease spread and other ecological attributes at smaller temporal and spatial scales (Macdonald et al. 2008). The movements recorded therefore represent minimum distance travelled.
Work on badgers (trapping, marking and vaccinating) was approved by the University College Dublin (UCD) Animal Research Ethics Committee (AREC) and met appropriate guidelines for animal field studies (Sikes et al. 2011). The capture of badgers was conducted under licences (1876 Cruelty to Animals Act) issued by the Irish Department of Health & Children. The capturing methods conformed to national legislation for the humane trapping of wildlife [Wildlife Act, 1976, Regulations 2003 (S.l. 620 of 2003), Republic of Ireland]. Capture via stopped body restraints or cages result in very low rates of injury to badgers (Woodroffe et al. 2005a; Murphy et al. 2009). Throughout the study, on-site veterinarians monitored the health of captured animals. Standard capturing protocol was employed during this study, where traps were laid by experienced field staff in a manner that would maximize the probability of badger capture (e.g. at active sett openings).
Effect of study scale on movement metrics
We used a subsampling technique with a circle sampling frame to investigate the effect of study scale on metrics of badger movements (Franzén & Nilsson 2007). All subsampling was repeated 1000 times at regular circular area sizes varying from 0·5 km2 to 177 km2. Circular sample frames were increased in size by the addition of 500 m to the circle's radius. The subsamples were randomly located, using a random-point generator in arcgis 9.2 (ESRI, Redlands, CA, USA). Only sampling frames containing a movement and completely contained within the study area were used to derive movement statistics. All movement distances >0 km (i.e. not recaptures at the same sett) were included in these statistics. Similarly, only movements completed within sample frames were used to generate movement statistics. We calculated the mean, median and maximum movement lengths per subsampled study area. In addition, we calculated movement statistics for the whole study area (755 km2). Finally, we used all movement data available, including data from marked badgers recovered outside the study area (n = 3) to assess the potential largest scale movements within the region.
The movement data from the subsampling study were modelled in a regression model relating movement distance (D) to the subsample area size (loge-transformed) and number of movements recorded per subsample. Separate models were constructed using all data, and for maximum D only. Regression models were assessed for linearity, normality of residuals, multicollinearity and heteroscedasticity (Dohoo, Martin & Stryhn 2009). Asymptotes were estimated from the predicted values from these models and related to sample frame area sizes.
Dispersal kernels measure the frequency distribution of movements within a population by fitting mathematical forms (kernels) to the empirical data (Nathan et al. 2012). Derivations from these forms allow for the estimation of the expected frequency of movement distances within a population (Baguette 2003; Fric & Konvicka 2007). We use a broad definition of dispersal to generate dispersal kernels; dispersal being the ‘movement of an individual away from its source’ (Nathan et al. 2003), in our case from the sett of capture. However, we restricted the dispersal kernel analysis to only movements >1 km (i.e. we invoked a threshold; Calsbeek 2009). Movements <1 km are more likely representative of within territory movements, as the mean and median nearest-neighbour distances between main setts within the study area were 841 m and 802 m, respectively (for all setts including non-main setts: mean = 367 m; median = 268 m). These movements were considered within ‘patch’ movements (sensu Baguette 2003), meaning our approach is similar to those taken during metapopulation dispersal studies (e.g. Baguette 2003; Calsbeek 2009).
We investigated the dispersal kernels of badgers in this population using two functions: inverse power function (IPF) and negative exponential function (NEF). Over twenty different movement kernel forms have been described (Hui et al. 2012; Nathan et al. 2012), but these two forms have been popular and used in a number of different systems (Baguette 2003; Fric & Konvicka 2007; Hui et al. 2012; Nathan et al. 2012). Linear models (IPF vs. NEF) were compared using the Akaike Information Criterion (AICc) following Hui et al. (2012). We considered the best-fitting dispersal kernel to be fat-tailed if (i) the probability of long-distance movements exceeded the predictions of a negative exponential distribution (Paradis, Baillie & Sutherland 2002), and (ii) exponent β (n) value from the IPF was 0 < β < 2 (Hui et al. 2012). We assessed whether there were sex-related differences in frequency of badger movements >1 km in our study population by comparing IPF and NEF dispersal kernels for both sexes. Separate models were run for both kernel types and the estimated βs, and their 95% confidence intervals (CI) compared to see whether there was an overlap as a heuristic measure of significant dispersal difference. A comparison of regression slopes was undertaken using analysis of covariance (ancova) with length as the covariate and gender as a factor variable. Finally, a two-sample Kolmogorov–Smirnov (KS) test was performed on the kernels generated from data from female and male badgers (Hui et al. 2012). The critical value for rejecting the null hypothesis (that the kernels derive from the same distribution) being P < 0·05.
Investigating the relationship between density, study scale and LDDs
In total, 1935 capture events were recorded with 963 individual badgers identified during the study period. Badgers within the study population exhibited sett fidelity, with a slight majority of recapture events (52%) being at the same sett as the previous capture. Of the badgers that moved, the movement frequency distribution was highly skewed with 43% of movements (n = 204) being ≤1 km.
Using all movements within the study area (755 km2; presented in Fig. 1), the mean movement distance was 1·67 km; the inclusion of recoveries (n = 3) of marked badgers from outside the study area increased the mean to 1·72 km. There was no difference in the median movement length when badgers recovered from outside the study area were excluded or included in the analysis (both medians = 1·19 km). The greater mean length including recoveries from outside the study area was due to one exceptionally long movement of 22·08 km.
Overall, there were 270 movements >1 km recorded during the study, which were used to derive dispersal kernels. Of these, the mean and 95 percentile attempted dispersal distance were 2·61 km and 7·28 km, respectively. The distribution was highly leptokurtic (kurtosis = 23·99; variance = 4·35). The kernel for these movements was best represented by an IPF (Fig. 2). The IPF model had a better fit and smaller AIC (IPF: R2=0·87; AIC = 199·04) than the NEF model (NEF: F1,268 = 380·56; P <0·0001; R2=0·59; AIC = 554·54). The β of the IPF regression model was <2, indicating that the expected dispersal distance is unbounded. The NEF severely underestimated the fraction of movements at distances 1–2 km (∆prop = −0·29) and overestimated the fraction of movements at distances between 2 km and 10 km (mean ∆prop = 0·024; Fig. 2). The NEF underestimated the probability of LDD attempts at distances ≥10 km in comparison with the IPF model. The IPF estimated population probabilities of movement (fraction) of 0·011, 0·005 and 0·003 for dispersal distances of 10 km, 15 km and 20 km respectively, over the period of the study (Table 1).
Table 1. Results of fitting inverse power functions (IPF) and negative exponential functions (NEF) to badger (Meles meles) movement data. β is the slope of the fitted linear regression; all regression slopes were significant at P < 0·001. Prop represents the model-estimated probability of movement at 10 km, 15 km, and 20 km over the study period.
β (95% CI)
Male (n = 184)
(−0·398 to −0·311)
(−1·947 to −1·741)
Female (n = 86)
(−0·293 to −0·205)
(−1·599 to −1·362)
All (n = 270)
(−0·329 to −0·268)
(−1·754 to −1·598)
There was a higher frequency of male movements ≥1 km (68%; n = 184) than female movements (32%; n = 86). IPF models indicated shallower estimated slopes of decline (β) for female badgers than male badgers (Table 1; Fig. 3). The 95% CI values did not overlap when kernel models were estimated separately (upper 95% CI for female: −1·599; lower 95% CI for male: −1·741). An ancova indicated that this difference between slopes was significant (P <0·001). Similar results were found with NEF models. The Kolmogorov–Smirnov test suggested that the movement distributions were significantly different between the sexes (Df = 0·176; P = 0·039). The 95 percentile dispersal distance for females was 9·44 km, whereas for males the 95 percentile was 4·37 km. These results indicate at movement distances ≥1 km (i) male badgers move more frequently than female badgers, and (ii) females make comparatively more long-distance movements, as a proportion of movements recorded, than male badgers.
Movement distances (D) increased significantly with increasing log-area size (β = 0·197; t =12·83; P <0·001) and were also positively associated with the number of movements (β = 0·001; t =3·85; P <0·001) detected per sample. Log-area was positively associated with the maximum distance (max. D) moved (β = 1·987; t =13·73; P <0·001), but not the number of movements (β = −0·005; t =−1·76; P =0·107) detected per sample. Using circle sampling frames of increasing size, the mean and maximum movement distances increased logarithmically before approaching an asymptote at spatial scales >80 km2 in size (Fig. 4a,b; model predicted asymptote sample areas size for mean movements: 132·7 km2; maximum movement: >176·7 km2). The median movement distance increased logarithmically up to 20 km2 and then increased very gradually thereafter up to 80 km2 (Fig. 4a). Many of the longest distance displacements would not have been recorded at scales below this (80 km2) spatial extent (Fig. 4b).
Multipopulation analysis: scale, density and LDDs
In total, data from 15 badger populations were gathered from the literature (including data from the present study; Table 2). The data met the assumptions of the linear model with linearity between outcome and dependent variables (lowess curve), normally distributed residuals and no evidence of heteroscedasticity (all models Cook–Weisberg test: P > 0·42). VIF was very low (1·16–1·17), despite there being a moderate negative correlation between dependent variables (r =−0·38). Overall, the model using mean density as a predictor significantly explained 71% of the variation in the data set (n = 15; F2,12 = 14·79; P <0·001; r2=0·71). Controlling for mean estimated density, there remained a significant positive relationship between study area size and max D (β = 0·02; P =0·003). Mean density was a marginally significant predictor and was negatively associated with max D (β = −0·22; P =0·044). Similar model results were found when minimum density (area: β = 0·02; P =0·003; min. density: β = −0·36; P =0·054) and maximum density (area: β = 0·02; P =0·003; Max. density: β = −0·17; P =0·042) were used as dependent variables in separate models.
Table 2. Badger movements (D) recorded during mark–recapture (MRR) and tracking studies (T) from 14 populations. Tracking data allow for fine temporal resolution of movements. MRR movements are of coarser temporal scale, where movements may have occurred at times usually >30 days. MMR are the minimum distance moved, as they are measured as straight lines between capture points.
Max. D (km)
Density (Adults km−2)
Study area (km2)
Reference (Data S1)
Sample size = number tracked. aFound dead inside study area. bEstimated from data presented in paper. cFound dead outside study area. dUsing coloured-bait returns.
Figure 5 demonstrates the predicted relationship between maximum D and study area size, while adjusting for estimated mean population density (marginal mean effect). Studies far below the model predicted regression line have shorter reported maximum D than would be expected given study area size and density [Oxford, Gloucestershire 2, Brighton and Offaly (see Table 2)]. These studies did not include chance recoveries of animals found outside of the study areas. Studies above the regression line tended to include chance recoveries from outside of the study area (Cork, Gloucestershire 1, Bristol and Kilkenny, but not Białowieża).
European badgers are territorial across much of their geographical range, with many movements occurring within the boundaries of their social group's territory (e.g. Rogers et al. 1998). The population studied here exhibited this general trend, with the majority of badgers being recaptured at the same sett where they had been captured previously. When badgers were encountered at a different location, 43% represented movements of ≤1 km in length (which are likely within group movements in this population, but could represent intergroup movements in higher-density populations). These findings are broadly consistent with previous research on badger populations at a range of densities. However, the recording of long-distance movements within this population, and a quantification of the frequency of such movements, is novel and can be attributed to the unprecedented scale of this study.
We found that there was a strong relationship between metrics of badger movement (mean, median and maximum D) and the spatial scale of the subsampled study area within our study. The size of the subsampled area had a particular impact on the distribution of distances moved, fundamentally truncating the movement distribution. According to our data, in order to adequately characterize badger movement distributions, study areas of ≥80 km2 may represent appropriate scales for similar density populations. Currently, many studies of badger movements in Britain and Ireland are undertaken at scales of <20 km2 (Table 2). In continental Europe, the spatial scale of badger movement studies has been often >80 km2. Nonetheless, for logistical reasons, these continental studies can be compromised by small sample sizes (mean number of badgers tracked: 12·5; range: 7–24; data from Table 2). The probability of recording LDDs with such small sample sizes is low.
A meta-analysis of the available data in the literature on badger movements indicated that there is a significant relationship between the greatest movement distances recorded (max. D) and the spatial scale of the study (Table 2). Though this relationship is confounded by the estimated density of the populations under study, there is a tendency (P =0·042–0·054) for longer distance movements to be recorded in lower-density populations. This pattern may result from an interaction between population ‘packing’, increasing the resistance to movement between groups (higher ‘social viscosity’) due to population density, and also due to socio-spatial distribution where lower-density populations have larger distances to travel between groups (Ronce 2007). Often where long-distance movements have been recorded in higher-density populations, they are a result of chance findings (recovery) of marked badgers outside the study area (Table 2). In low-density populations, badgers are capable of moving large distances in relatively short periods of time (maximum 6·9–17·5 km night−1; see references in Table 2); but typically, these distances are travelled tortuously and almost exclusively within their home range (Loureiro et al. 2007). It should be noted that due to the nature of the data collection, many movements made at shorter temporal scales may have been missed during the present study. However, at the spatial scale of the study, the relative frequency of movements across the distribution should not be significantly biased, especially estimates from the IPF which is invariant to marking frequency (Fric & Konvicka 2007). During our study, we recorded a movement by a female badger of 22·1 km; to the best of our knowledge, this is the longest displacement recorded for a badger in the British Isles.
Sex-biased dispersal is common in the animal kingdom (Greenwood 1980). In birds, for example, the tendency is for females to disperse more frequently than males. In mammals, generally the opposite is true (i.e. higher frequency of male dispersal) (Greenwood 1980), and there is also a tendency for males to travel further (Lawson-Handley & Perrin 2007). In mustelids, clear trends have not emerged with dispersal frequency and distances being found to be male, female or non-biased amongst species (e.g. Blundell et al. 2002; Dalerum et al. 2007). There has been variable evidence of sex-biased dispersal in badgers specifically in the literature (Roper, Ostler & Conradt 2003). Some studies found evidence of male-biased dispersal (e.g. Rogers et al. 1998), while other data suggest female-biased dispersal (e.g. Woodroffe, Macdonald & da Silva 1993).
The literature on badger dispersal has been biased towards very high-density populations (Frantz et al. 2009). An exception to this used genetic methods to compare a high-density (up to 25 km−2) English population with a lower-density (~1 km−2) Swiss population (Frantz et al. 2009). That study reported badgers dispersing greater distances in the lower-density Swiss population than the higher-density English population. Furthermore, dispersal was female-biased in the lower-density Swiss population whereas no consistent trend was found for a sex bias in the English data set. In our study population, female badgers had higher probability of moving longer distances than male badgers, though there was higher frequency of male movements recorded. This may suggest that sex-biased movement patterns in badgers is a scale-dependent phenomenon (biased towards males at intermediate scales and females at larger scales), as found in other animals systems (e.g. common vole Microtus arvalis, Gauffre et al. 2009). This hypothesis would help to explain the apparent discordancy within the literature, but further research is needed to test this (or alternate) hypothesis.
In the dispersal kernel analyses, we limited the data set to badger movements ≥1 km in length. These dispersals may represent either permanent or temporary dispersal movements away from a home territory, with both being potentially biologically meaningful. Temporary visits to adjacent territories have been shown to result in successful breeding (Dugdale et al. 2007) and have been correlated with tuberculosis disease status (Rogers et al. 1998; Vicente et al. 2007). Permanent movements can result in greater likelihood of breeding success for dispersing badgers (Woodroffe, Macdonald & da Silva 1993; Macdonald et al. 2008), and long-distance dispersals are known to contribute to the spread and maintenance of bTB in other model systems (e.g. brushtail possum Trichosurus vulpecula; Ramsey & Efford 2010).
Our data have implications for disease transmission, as long-distance movement made by individuals could result in the transmission of disease across landscapes. For example, dispersing badgers can accumulate more bite wounds (Macdonald et al. 2008) than badgers that stay within their social group, and this is a risk factor for bTB (Murphy et al. 2010). Previous work has found multiple bTB strains within setts, and some strains being different to those of the local cattle population (Olea-Popelka et al. 2005; Woodroffe et al. 2005b; Kelly, McGrath & More 2010) with strains geographically clustering at large scales (>6 km in Ireland; Olea-Popelka et al. 2005; Kelly, McGrath & More 2010). This led to speculation that long-distance badger movements could have been involved in the spatial spread of bTB strains across landscapes (Olea-Popelka et al. 2005), but until now, there were no data to support large-scale movements interconnecting badger populations (Fig. 1). Simulation modelling suggests that increasing the frequency or length of dispersal attempts can have effects on the spread and persistence of bTB in badger populations (e.g. Hardstaff et al. 2012). In lower-density populations, infection may be seeded, or disease prevalence maintained, in relatively isolated populations from dispersing badgers (Hardstaff et al. 2012). Currently, the test status of the badgers within this population is unknown, thus the epidemiological significance of badger LDDs is still speculative. However, this speculation must be considered against a backdrop of high bTB badger prevalence in Ireland (36%; Murphy et al. 2010) and parts of Britain (35–53%; Carter et al. 2012).
When designing vaccination or culling field trials with different treatments, it is important to ensure that the study sites are spatially independent (Bourne et al. 2007; Pech et al. 2010). Conversely, it is desirable for study sites to be as geographically similar as possible (Bourne et al. 2007). Given our findings, we would expect that 5% of badger dispersal attempts in this population would be at distances of >7·3 km over a 4-year study. This would indicate that, in the absence of physical movement barriers, buffers of ≥7·3 km may be needed to restrict inward dispersal and maintain site independence with a high degree of confidence. Despite this, the probability threshold chosen would depend on many other (including socio-economic) factors (Pech et al. 2010).
Our study highlights the difficulty in studying the ‘tail’ of the dispersal distribution of mammals – even in a species that exhibits natal philopatry and general low dispersal probabilities (Woodroffe, Macdonald & da Silva 1993). Future studies of badger dispersal will require large scales and long time periods if an Eulerian approach is used (i.e. the use of marked animals; Nathan et al. 2003). Alternatively, for Lagrangian methods (i.e. the use of tracked animals; Nathan et al. 2003), there is a need to increase sample size and temporal scales, but also the ability to track animals once they move outside the boundary of a study site (Hebblewhite & Haydon 2010). Development of GPS/GSM technology may improve such tracking studies as costs decrease permitting more animals to be tracked over greater spatial scales (Hebblewhite & Haydon 2010). Until then, our results highlight the importance of spatial scale in the study of a medium-sized mammal movements, especially one which is a host for an economically significant disease of domestic animals.
We would like to acknowledge all field staff and Drs. M. De Jong (Wageningen University), L. Corner and E. Gormley (UCD) who designed the vaccine trial associated with this mark–recapture study and Dr. D. Murphy for the kind use of the badger image. A.W.B. was supported by a Teagasc Walsh Fellowship (www.teagasc.ie). The vaccine trial was funded by the Department of Agriculture (www.agriculture.gov.ie).