Differences in trappability of European badgers Meles meles in three populations in England

Authors


*Present address and correspondence: F.A.M. Tuyttens, Department of Mechanisation, Labour, Buildings, Animal Welfare & Environmental Protection, Van Gansberghelaan 115, 9820 Merelbeke, Belgium (fax 32 9252 42 34; e-mail rvl.cigr@pophost.eunet.be).

Summary

1. Many ecological studies on the European badger Meles meles, as well as certain programmes to control bovine tuberculosis, would benefit from a greater understanding of the factors that influence the probability of capturing this animal in cage-traps. We therefore investigated some of the factors that could explain differences in trappability between three badger populations in England: the high-density protected populations of Wytham Woods and Woodchester Park, and the low-density culled population of North Nibley.

2. Trappability (the percentage of all individuals known alive that were actually captured) did not differ between sexes or adult age classes, but significant differences were found between cubs and adults, study areas, seasons and years, and various interactions between these variables.

3. Circumstantial evidence suggests that the culling of badgers in North Nibley may have resulted in a decrease of adult trappability in the following year.

4. Adult badgers at Wytham Woods and Woodchester Park were significantly more likely to be trapped zero times (‘trap-shy’) or all three times (‘trap-happy’) in 1996 than predicted by the estimated capture probabilities under the assumption of equal trappability.

5. Wytham Woods differed from the other study areas in that trappability of its badgers was positively related to their body weight and its adult badgers were more likely to be trapped than cubs. These differences could be a consequence of differences in trapping procedures that were followed at Wytham (no prebaiting and fewer traps per badger).

6. Trappability of badgers was not associated with social group size. Although it is difficult to determine precisely the movement and tuberculosis status of badgers based on mark–recapture data, our analyses did not suggest that either variable affected the likelihood of being trapped.

7. Studies that compare demographic, biometric and epidemiological parameters based on data collected from badgers captured at different times or places ought to account for the observed differences in trappability.

Introduction

Understanding the factors that determine the success of capturing European badgers Meles meles L. in cage-traps is important for two reasons. First, capture–mark–recapture studies have formed the basis of many research projects on the behaviour and ecology of this species (Kruuk & Parish 1985; Harris & Cresswell 1987; Cheeseman et al. 1988; Woodroffe & Macdonald 1995; Rogers et al. 1997). Secondly, the strategies of the UK government to control bovine tuberculosis (TB; caused by Mycobacterium bovis) in cattle have, until recently, relied on cage-trapping and shooting badgers where they were believed to maintain a reservoir of infection (Dunnet, Jones & McInerney 1986; Krebs et al. 1997). Research into badger trappability may improve the efficiency of research programmes and TB control operations.

This study aimed to investigate some of the factors that may influence badger trappability by comparing the capture probability among three badger populations in England that are currently subjected to frequent live-trapping: Wytham Woods, Woodchester Park and North Nibley. The latter study area has been subjected to repeated badger removal operations (BROs) by the UK Ministry of Agriculture, Fisheries and Food (MAFF) to control TB in cattle. The other two populations, however, have been studied while free from BROs for at least a decade. This provided us with the opportunity to evaluate the hypothesis that removal operations affect trappability of surviving or neighbouring badgers. We also investigated whether capture probability was related to season, year, age, sex, population density, variations in trapping protocol, body weight, social group size and, as far as these could be determined correctly from trapping data, movement and TB status.

Methods

Study sites and trapping procedures

Woodchester Park

Woodchester Park (National Grid Reference SW81,01) lies in the Cotswold Escarpment in Gloucestershire, south-west England (Cheeseman & Mallinson 1981). Capture–mark–recapture studies began there in the mid-1970s. Analyses were restricted to 24 contiguous social groups (occupying an area of 7·87 km2) that, since the early 1980s, have not been subjected to BROs nor, to our knowledge, any other form of persecution by humans. Trapping took place in spring (May–June), summer (July–September), autumn (October–November) and winter (December–February), but was suspended for the months of March and April when cubs are dependant on the mother. Badgers were captured following standard protocols (Cheeseman et al. 1987; Rogers et al. 1997) using cage-traps, which were prebaited with peanuts for 7–10 days and then set for 2 nights. Traps were positioned near badger runs in the vicinity of active setts, with more traps set than the expected number of badgers present. Badgers were anaesthetized (0·2 ml kg–1 ketamine hydrochloride) and marked with a permanent tattoo on initial capture (Cheeseman & Harris 1982). A variety of biometric data, including weight, sex and age (if year of birth was known), was recorded. The TB status of captured badgers was diagnosed by ELISA (Goodger et al. 1994) and by bacterial culture of faeces, urine, tracheal aspirate and pus from wounds (Pritchard et al. 1986). In order to avoid recaptures, badgers trapped on the first night were not released until the following day.

North Nibley

North Nibley (National Grid Reference ST74,96) is also located in Gloucestershire and lies approximately 8 km south-west of Woodchester Park between the Severn valley and the Cotswold Escarpment. Details of the habitat are given by Tuyttens et al. (1999). Although the spatial organization of the Nibley badger population is in flux due to its exposure to repeated BROs, there are about 22 social groups in the 13·4-km2 part of the study area where live-trapping takes place. Since the establishment of the Nibley study in March 1995, there have been two BROs in the area. The first BRO (September 1995) was a ‘live test’ strategy (Krebs et al. 1997) during which 27 badgers from six social groups were killed by MAFF. During the second BRO, an ‘interim’ strategy (Krebs et al. 1997), two more badgers were killed in June 1996.

Trapping for the purposes of research followed the same procedures as at Woodchester, although there were only three trappings per year (no trapping in winter). In the summer of 1995 there was no trapping apart from the BRO carried out by MAFF. During this BRO, badgers from seven social groups were trapped for 4 consecutive days and subsequently released, badgers from another six social groups were trapped over a period of up to 6 weeks and killed, while no attempts were made to trap badgers from the other groups in the study area that summer. Otherwise trapping procedures and collection of clinical and biometric data were similar to Woodchester Park.

Wytham Woods

Wytham Woods (National Grid Reference SP46,08) is situated some 10 km north-west of Oxford city at the eastern extremity of the Cotswold escarpment. Geological and vegetational details are given by Kruuk (1978) and Hofer (1988). Since 1993 there have been 20 social groups in the 6-km2 area. From 1987 onwards badgers were trapped at least four times annually at about the same time and following similar procedures as at Woodchester. However, there was no prebaiting period and traps were set for 3 consecutive nights. Badgers were released on the same day and, due to shortage of traps, setts may not have been saturated with traps. The TB status of badgers was not tested in Wytham.

Estimation of population size and trappability

In order to ensure comparability between study areas, we restricted analyses to the spring, summer and autumn trappings between 1995 and the end of 1997. Trappability (pi) was defined as the percentage of all individuals known alive that were actually captured during a particular trapping event i:

pi = 100 (ti/Ti)(eqn 1)

where Ti is a closed subsection of the population consisting of badgers that were known, post hoc, to have been alive and within the study area during trapping i irrespective of whether or not these badgers were actually trapped during that trapping. By counting the actual number of badgers trapped during trapping i (ni) and the number of these badgers that were included in the closed subpopulation (ti), the closed subpopulation method estimates population size as (Chapman 1951; Tuyttens et al. 1999):

Ni = [(Ti + 1) (ni + 1)/(ti + 1)]– 1(eqn 2)

For example, in Nibley, Ti included some radio-collared badgers that were known to have been alive and within the boundaries of the study area during the time of trapping. As explained in Tuyttens et al. (1999) it is also legitimate to include in the closed subpopulation of the ith trapping occasion any cub trapped (or found dead) later in the same year. However, adult badgers (which are more likely to move between groups than are cubs) should have been trapped at least once before and once after the ith trapping occasion to be included in Ti. Adult population size at the time of the first trapping was estimated using the arguable assumption that all badgers trapped in the summer of 1995 should be included in the closed subpopulation of the spring 1995 trapping. The estimates of adult N and p were most reliable for the trappings in 1996 as the adult closed subpopulation tended to be smallest for the trappings at the beginning and end of the study period. In fact, population size could not be estimated for the last trapping (autumn 1997). This method assumes that there is no temporary emigration across study area boundaries, and that all animals in the population have the same probability of capture.

For comparison, we also estimated the minimum number alive (MNA) (Krebs 1966), which we defined as:

MNAi = ni + Titi(eqn 3)

Data analyses

Using the closed subpopulation method we estimated population size and trappability during each trapping (spring 1995–summer 1997) in all three study areas for each age and sex class separately. As the Nibley study had only recently been established, badgers there could only be aged as cubs (< 1 years) and adults (≥ 1 years) in 1995. Further differentiation of yearlings (1–2 years) and 2 year olds was not possible until 1996 and 1997, respectively. Because the adult population was believed to be younger at Nibley than at the other study areas, we first tested whether trappability of adult badgers was related to age. For every adult badger included in the closed subpopulation Ti we tested whether being captured or not was associated with adult age (if known), using mixed effects logistic regression models (Littell et al. 1996).

We then tested whether the likelihood of being trapped was associated with various variables using a mixed effects logistic regression model. These analyses were restricted again to badgers included in Ti. Variables included in this model were (if available) age class (cubs vs. adults), sex, year (1995, 1996, 1997), season (spring, summer, autumn), study area (Wytham, Woodchester, Nibley) and population density. Interactions between these variables were included in the model if they were found to have a significant effect.

We also compared between study areas the number of individuals in each age/sex class that were trapped zero, one, two or three times in 1996, as an alternative measure of trappability. These analyses were restricted to those badgers that were known to be alive during all three trappings in 1996, on the basis that they had been trapped at least once in both 1995 and 1997. These observed frequencies were compared with the expected number of badgers that should have been trapped zero, one, two and three times if all animals had the same probability of being trapped, as estimated by the closed subpopulation method. This allowed us to verify whether the capture probabilities had been consistently over- or underestimated by the closed subpopulation method. Trappability would seem to have been overestimated by this method if the expected number of badgers trapped frequently (two or three times a year) exceeded the observed number and if the observed number of badgers not trapped (or trapped only once) that year exceeded the expected number. We performed the same analyses for cubs. However, because by definition cubs included in all three 1996 closed subpopulations had to be trapped at least once in that year, we could compare only the numbers of cubs trapped one, two and three times. Cubs known to be alive that evaded all three 1996 trappings could not be counted. Adult badgers that had been trapped zero times in 1996 we called ‘trap-shy’, while badgers that had been trapped all three times were called ‘trap-happy’. We used logistic regression analyses to test whether trap-happiness vs. trap-shyness in 1996 was associated with study area, sex, adult age and social group size.

We used another mixed effects logistic regression model to test for an association between trappability and body weight, social group size, TB status and movement status, while controlling for age/sex class, year and study area. These analyses were restricted to those badgers that had been trapped during two consecutive spring trappings. Trappability was measured as the number of times (range 0–2) an animal had been captured during the period between these two consecutive spring trappings. Age class, body weight and body condition were measured at the time of the first spring trapping. Animals that had been trapped in the second spring at the sett of a different social group than where they had been trapped the previous spring were classified as ‘movers’, and other animals as ‘non-movers’. Badgers that had a positive serological or culture test for TB during one or both of the spring trappings were assumed to be ‘TB-positive’, all other badgers were classified as ‘TB-negative’. Social group size was estimated as the number of different cubs and adults trapped at each group during any of the three trappings of that year divided by the probability that a badger had been trapped at any of these trappings (as estimated from the calculated p values given in Table 1).

Table 1.  Statistics used to estimate population size (N) and trappability (p with exact 95% confidence intervals) by the closed subpopulation model for each trapping, study area and age class that could be separated using 1995–97 mark–recapture data only. n is the number of animals trapped, T is the number of animals included in the closed subpopulation, and t is the number of T animals that were trapped during a trapping occasion. No comparable estimates could be given for summer 1995 in Nibley because of the badger removal operation (see main text)
Woodchester ParkWytham WoodsNorth Nibley
(years)nSeason
T
Age
t
Np(95% CI)nTtNp(95% CI)nTtNp(95% CI)
1995
Spring05247366877(64–86)2438194850(36–64)2723154165(47–80)
> 01271529221061(53–68)1021628319951(44–59)3528156454(37–69)
Summer04344404791(81–96)2030134643(28–59)      
> 01491068319078(70–85)1281117718469(61–77)      
Autumn03337304181(68–90)1231113435(22–51)13961967(40–86)
> 0861276317350(41–58)541364915036(29–44)162074435(19–54)
1996
Spring03122183882(65–92)2134116332(20–47)2225163464(46–79)
12327213078(62–89)2030183360(44–75)81171364(39–83)
> 01201349417170(62–77)15215710822169(62–75)2229115738(24–54)
> 1971077314268(59–76)1011077514470(61–78)91843722(10–41)
Summer01522152268(50–83)201756229(14–50)2321163076(58–89)
12224192879(63–90)1628143250(34–66)41141136(17–61)
> 01051147515966(57–74)1141528520456(48–63)2328125343(28–59)
> 183905613362(53–71)751036012958(49–67)91471850(29–71)
Autumn01518112461(41–78)121491964(42–82)1324122650(33–67)
182262927(14–45)122493238(22–55)1111119(2–28)
> 0321062513424(17–32)851376119045(37–53)1301303(1–12)
> 124841910523(15–32)60924412548(38–57)0120120(0–95)
1997
Spring04461308949(38–61)1623310113(5–28)1927173063(46–78)
1121081580(55–93)111381862(39–81)2016122775(54–89)
21516112269(48–85)141582653(32–73)6831538(16–65)
> 080874914256(46–65)1151237518861(53–69)4238236961(46–74)
> 168774112753(43–63)77965712959(50–68)1620112955(36–73)
> 251613010349(38–61)63814910460(50–70)786975(47–91)
Summer04749376276(63–85)16925622(7–48)2419172789(74–97)
1103310100(29–100)11641667(36–88)101042340(19–65)
215752071(42–90)13532060(28–85)3511120(5–52)
> 078301912163(47–77)112644217066(54–76)252386835(20–53)
> 168271611359(42–75)82483212267(54–78)111343331(14–54)
> 25320119455(36–73)69432910267(54–79)6312733(9–71)

Results

Effect of adult age

Table 1 shows N^ and p for each trapping and age class for all three study areas. In Wytham and Woodchester there was no association between adult age and trappability (F5,1751 = 1·34, P = 0·246) (Fig. 1). In Nibley there was also no difference in trappability between yearlings and older badgers in 1996 (F1,70 = 0·31, P = 0·580), nor between yearlings, 2 year olds and older badgers in 1997 (F2,43 = 0·57, P = 0·571). We therefore combined adult age classes for subsequent analyses.

Figure 1.

Effect of age on badger trappability (ṕ with the exact 95% confidence intervals) for each trapping between 1995 and 1997 in Woodchester Park (WP) and Wytham Woods (WW).

Effects of sex, year, season, cubs–adults, population density and study area

Table 2 shows that trappability of badgers differed between age classes (cubs vs. adults), years, seasons and study areas, but not between the sexes. The effect of population density depended on the study area. There were also significant differences between study areas depending on year, season and age class, and between seasons depending on year and age class.

Table 2.  Test statistics and probabilities from a mixed effects logistic regression to partition the variation associated with badger trappability
Sourced.f.FP
Age class (cubs vs. adults)1, 275918·220·0001
Sex1, 27590·090·7699
Year2, 275916·570·0001
Season2, 275917·410·0001
Study area2, 27597·590·0005
Population density1, 27592·480·1156
Year × season3, 27595·220·0014
Year × study area4, 27595·810·0001
Season × study area3, 27593·860·0090
Population density × study area2, 27593·920·0200
Age class × study area2, 275925·500·0001
Age class × season2, 275915·760·0001

Combining trappability estimates for all trappings from 1995 until the end of 1997, trappability was on average 15% and 29% higher for cubs than adults in Woodchester and Nibley, respectively (Table 1). These differences were most pronounced in autumn (1995 and 1996). In Wytham, however, the opposite held as trappability was on average 21% lower for cubs than for adults, and this difference was most pronounced in spring and summer. Combined, these results indicate that adults were least trappable in autumn while cub trappability seemed less affected by season.

With the exception of autumn 1996, cub trappability was substantially less in Wytham than in both other study areas (Table 1). Although overall adult trappability was lowest in Nibley and highest in Woodchester, these differences were not clear-cut and were complicated by season, year and population density. For example, adult trappability was lower in Wytham than Woodchester until autumn 1996 but higher afterwards. Adult trappability was very low in Nibley during all three trappings in 1996, while in 1995 it was at similar levels to Wytham. Cub trappability seemed to be inversely related to population density in Wytham and Woodchester, but not in Nibley.

Statistical analyses of the alternative measure of trappability (numbers of times badgers known alive had been trapped during the three trappings in 1996) confirmed these differences in trappability between study areas and age classes in 1996 as described above. As again there were no differences for either cubs (

image

 = 1·36, P = 0·51) or adults (

image

 = 5·52, P = 0·14) we pooled both sex classes. Cubs were trapped more frequently at Woodchester than at Wytham (

image

 = 9·09, P = 0·01), but there were no significant differences between Nibley and either Woodchester (

image

 = 1·25, P = 0·54) or Wytham (combining cubs trapped two and three times,

image

 = 2·48, P = 0·12) (Fig. 2a). Adult badgers were trapped less frequently at Nibley than at Woodchester (

image

 = 11·17, P = 0·01) and at Wytham (

image

 = 13·47, P = 0·004), but there were no statistically significant differences between Woodchester and Wytham (

image

 = 2·29, P = 0·51) (Fig. 2b).

Figure 2.

Percentage of (a) cubs and (b) adults included in the 1996 closed subpopulation that were trapped (zero), one, two and three times in 1996 in North Nibley (NN), Woodchester Park (WP) and Wytham Woods (WW).

Observed and expected (under the assumption that each animal has equal probability of being captured as estimated by the closed subpopulation method) frequency distributions were not significantly different for cubs in any of the three study areas [Nibley:

image

 = 0·39, P = 0·82; Woodchester:

image

 = 0·30, P = 0·58 (combining cubs trapped two and three times); Wytham:

image

 = 0·21, P = 0·65 (combining cubs trapped one and two times)] (Fig. 3a). There were also no significant differences between observed and predicted frequency distributions for adults in Nibley (combining adults trapped two and three times,

image

 = 1·77, P = 0·41) (Fig. 3b). However, these differences were statistically significant for both Woodchester (

image

 = 14·32, P = 0·003) and Wytham (

image

 = 16·65, P < 0·001). Therefore, in both these areas the assumption of homogeneous trappability was unrealistic because then the number of adult badgers that were never (trap-shy) and always (trap-happy) trapped in 1996 was underestimated (Fig. 3b). There was no evidence from the logistic regression model of a relation between adult age, group size or sex and trap-shyness vs. trap-happiness. However, there were relatively more trap-shy than trap-happy badgers in Nibley compared with Wytham (

image

 = 10·98, P < 0·001) and Woodchester (

image

 = 8·02, P = 0·005).

Figure 3.

Frequency distributions of the observed and predicted (under the assumption of equal trappability) number of times (a) cubs and (b) adult badgers were trapped in North Nibley (NN), Woodchester Park (WP) and Wytham Woods (WW) in 1996. As only badgers included in the closed subpopulation of all three trappings in 1996 are considered, the observed and predicted number of cubs that were not trapped in 1996 could not be derived.

Effect of the bro in nibley

We investigated whether the BRO in Nibley in the summer of 1995 affected capture probability by comparing trappability of badgers in spring 1995 (before the current BRO), 1996 (9 months after the BRO) and 1997 (21 months after the BRO) between cubs and adults and between North Nibley and the undisturbed study areas. Unfortunately, sample sizes were too small to fit a three-way interaction between age × study area × year using the above mixed effects logistic regression model. However, inspection of Table 1 does not allow us to reject the hypothesis that the BRO might have negatively affected badger trappability in Nibley. Before the BRO, adult trappability in Nibley in spring was intermediate between that of Woodchester and Wytham. However, adult trappability in Nibley was 16% lower in spring 1996 compared with the previous year, while in Woodchester and Wytham adult trappability had increased by 9% and 18%, respectively (Fig. 4b). In the spring of 1997, adult trappability in Nibley was similar again to that of the undisturbed populations. Furthermore, such a reduction in trappability in the first year postBRO was not observed for cubs (which had not experienced the BRO) (Fig. 4a).

Figure 4.

Trappability (ṕ with the exact 95% confidence intervals) of (a) cubs and (b) adult badgers from North Nibley (NN), Woodchester Park (WP) and Wytham Woods (WW) in spring 1995, 1996 and 1997.

Effects of body weight, social group size, movement status and tb status

Adjusting for study area, age class (cubs vs. adults), sex and year, the number of times a badger was trapped between two subsequent spring trappings was not associated with social group size, social group size × area, movement status or TB status. However, there was an effect of body weight depending on study area (F2,69 = 4·78, P = 0·011). In Wytham, the heavier a badger in spring the more likely it was to be trapped in summer and autumn (t69 = 2·32, P = 0·024), whereas there was no such relation in Woodchester (t69 = –0·78, P = 0·441) and Nibley (t69 = 0·17, P = 0·862).

Discussion

Trappability of badgers differed among the three study areas. These differences depended on effects of year, season, population density and age class (cubs vs. adults). For example, in Nibley and Woodchester trappability was higher for cubs than adults, especially in autumn. In Wytham, however, adults were more likely to be trapped than cubs and this difference was least pronounced in autumn. There were also seasonal, yearly and age-related (cubs vs. adults) differences in trappability independent of the study area effect. It should be borne in mind that our analyses did not include all possible factors that might influence badger trappability. Food availability and local weather conditions, for example, are likely to affect trappability as well. As we studied natural populations, all these potential influences could not be controlled as in a scientific experiment. Consequently, competing hypotheses to explain these differences are not easily disentangled and remain speculative.

For example, the potential negative effect of badger culling operations on trappability of the surrounding/surviving population needs to be investigated further. Five lines of circumstantial evidence suggest, but do not prove, that the apparent reduction in adult trappability in Nibley in 1996 could have been a direct consequence of the BRO that took place there in summer 1995. First, in spring 1996 (9 months postBRO) trappability of adults was much lower than the previous year, a trend that was not observed for cubs (which had not experienced the BRO). Secondly, the opposite trend was observed in the undisturbed populations of Woodchester Park and Wytham Woods, where trappability of adult badgers was higher in spring 1996 than 1995 (Fig. 4). Unfortunately, only spring capture probabilities could be compared to test the effect of the BRO, because comparable preBRO trappability estimates were not available for summer or autumn trappings. Thirdly, experience from radio-tracking, sett watching and video monitoring had already suggested that badgers in Nibley, and particularly in the aftermath of the BRO, were more shy and vigilant compared with badgers from the other study areas (Tuyttens 1999). Fourthly, the single adult badger from the removed groups that was known to have survived the culling operation in Nibley was trapped only once out of five subsequent trappings during which it was known to be alive, whereas adult badgers from neighbouring groups known to be alive were trapped on average 2·3 times (n = 12, SD = 0·965) more often during the same five trapping occasions. Finally, weather conditions (dry and warm) during the spring 1996 trapping in Nibley were believed to favour high trapping success.

As there was little opportunity for avoidance learning, the apparent reduction in adult trappability postBRO could more easily be explained by artificial selection for trap-shyness. It is conceivable that trap-happy badgers are more likely to be killed during BROs than trap-shy badgers. The significant population density by study area effect on trappability suggests that the short-term reduction in adult trappability observed in Nibley could also have been an indirect consequence of the lowered density.

Indeed, population density effects may explain differences in trappability between Nibley, Woodchester and Wytham, perhaps through the association with food competition. Because the Nibley population had been subjected to several BROs before the start of our study, the population is believed to be well below its carrying capacity, unlike both other study areas. Consequently, badgers in Nibley are heavier and in better body condition than in the high-density areas (Tuyttens 1999). In autumn, when food competition is less intense than in late spring or summer (Brown 1993), trappability tends to be lowest. The lowest ever trapping success was recorded for the autumn 1996 trapping in Nibley, during which only 3% of the adult badgers were captured (Table 1). It is probably no coincidence that such an extremely low trapping success occurred in the disturbed population of Nibley, in the autumn, and at a time when it had been very wet during, and shortly before, the trapping. Rainfall causes earthworms, the staple diet of badgers in our study areas, to surface. We speculate that it is more difficult to lure badgers to enter traps baited with peanuts at such times when per capita food availability is suspected to be very high. Presumably, food competition is more intense in high-density populations such as at Woodchester and Wytham.

It is intriguing that in Wytham, where badger density is highest, cubs were less trappable than in Woodchester or Nibley. These differences could be related to the different trapping protocol that is followed at Wytham: there is no prebaiting period and proportionally fewer traps are put down per badger. Perhaps trappability of cubs was relatively low because they were outcompeted by adults for access to peanuts in traps. Alternatively, cubs may not have had time to learn to enter traps because there was no prebaiting period during which they could familiarize themselves with this unusual source of food. The former hypothesis predicts that in high-density populations heavy adult badgers outcompete cubs and lighter badgers for access to the bait in the traps, particularly if relatively few trapping cages have been set and when competition for food is intense. Arguments in favour of this hypothesis are: (i) in Wytham trappability of cubs seemed to be negatively associated with population density while this was not the case for adults nor for the other study areas; (ii) heavy badgers were more likely to be trapped during subsequent trappings than lighter badgers in Wytham but not in the other study areas. It is also possible that adult badgers emerge earlier from their underground setts compared with cubs and that this difference was greater, or affected trappability more, in Wytham than in the other study areas. Further experiments are needed to test these hypotheses. Indeed, some findings were more in agreement with the no prebaiting hypothesis. For example, even in Wytham it was not unusual to find that peanut baits in trapping cages had not been eaten. Also, the number of badgers living in a social group did not affect trappability in any of the three study areas.

We also failed to find an association between trappability and whether or not a badger was detected to make an extra-group movement or was infected with M. bovis. However, these results should be interpreted with caution, because of the difficulty of determining the movement and TB status of badgers that are captured infrequently. We suspect that it was more likely that TB-infected badgers were wrongly classified as ‘non-infected’ than the other way round, because the specificity of the ELISA test is greater than its sensitivity (Clifton-Hadley, Sayers & Stock 1995) and because there is evidence that infectious badgers can have negative culture results because excretion of bacilli is intermittent (Clifton-Hadley, Wilesmith & Stuart 1993). Cheeseman & Mallinson (1981) have reported that heavily infected badgers show behavioural changes and seem to be less fearful. It might therefore be expected that the trappability of such heavily infected badgers would also be altered. Given its applied importance for programmes to control M. bovis this possibility needs to be investigated further.

The main implication of our analyses for demographic studies of badgers that rely on capture–mark–recapture, concerns our finding that trappability can differ considerably between, as well as within, populations of badgers. This study has shown that variations in trappability between populations may differ, for example, between years, seasons, age classes and according to density. It has also shown that variations in trappability within a population may differ, for example between years, seasons and age classes. Although not proven by this study, we have argued that other factors, such as human disturbance, differences in trapping procedures, per capita food availability and weather conditions, are likely to cause additional variation in trappability. Studies based on mark–recapture should use appropriate population estimation methods that account for and incorporate such variation in trappability when comparing aspects of population demography between study areas, seasons or years.

In order to illustrate the effect of accounting for differences in trappability, we compared population estimates of cubs and adults in the three study areas based on the closed subpopulation method, the MNA and the actual count of the number of badgers trapped during each trapping occasion. Figure 5 shows that there are considerable and inconsistent differences between these three estimates of population size depending on the age class of the badger, study area and time. The actual number of badgers trapped during a trapping occasion is clearly not a reliable method to describe accurately trends in population size because of variations in trappability. The MNA technique partially corrects the negative bias of the former method by using later trapping periods to allow back-calculation of the existence of animals ‘missed’ during early trapping periods. As expected (Hilborn, Redfield & Krebs 1976), the MNA approaches the closed subpopulation method estimate when capture probability is generally high (e.g. cubs in Woodchester). However, the MNA is much lower than the closed subpopulation estimate when capture probability tends to be low (e.g. Wytham cubs or Nibley adults), especially in the last few trapping periods. These results therefore support earlier criticisms of the MNA that the varying and unknown extent of its negative bias may invalidate comparisons between studies in space as well as time (Nichols & Pollock 1983; Nichols 1986; Montgomery 1987; Hallett et al. 1991).

Figure 5.

The number of cubs and adult badgers trapped (n), the minimum number alive (MNA), and the closed subpopulation population estimate (N●) in the three study areas for all but the last trapping between 1995 and the end of 1997.

The closed subpopulation model is a reliable method for estimating population size and trappability if the underlying assumptions are not violated, and if the actual number of badgers trapped (ni) as well as the number of badgers included in the closed subsection of the population (Ti) are sufficiently large (Tuyttens et al. 1999). The assumption that was most difficult to fulfil is that every animal in the population should have the same probability of capture. Even when cubs and adults were treated separately, there were indications that there were heterogeneities in trappability among badgers in Wytham and Woodchester. There were more adult badgers that had been trapped zero times (trap-shy) and all three times (trap-happy) in 1996 than expected under the assumption of equal trappability (Fig. 3b). Although never trapped in 1996, the former badgers were not truly trap-shy, because they had been trapped in 1995 and 1997. The proportion of unmarked animals encountered in road traffic accidents or mark–resight trials, however, suggests that the proportion of ‘truly trap-shy’ badgers is low (Rogers et al. 1997; Tuyttens et al. 1999). Nevertheless, the closed subpopulation estimates of population size could be negatively biased (and estimates of trappability positively biased) because of the violation of the equal trappability assumption (Nichols & Pollock 1983).

It is obvious that other demographic parameters, such as cub : adult ratio, fecundity, mortality and dispersal derived from capture–mark–recapture data, may also be biased if differences in trappability are not adequately accounted for. Our results show that many variables that are potentially difficult to adjust for in comparative studies can influence the likelihood of trapping certain types of animals such that the captured sample is not representative of the true population. For example, we have shown that the likelihood of trapping cubs differed between study areas and seasons, and in Wytham trappability was affected by body weight while in the other populations there was no such association. Estimates of disease prevalence may also be biased if such differences in trappability are not estimated.

All capture–mark–recapture models to estimate population size are most reliable when capture probability is high. Unfortunately, few scientific studies on trapping badgers have been published (but see Pigozzi 1988). It is likely that capture probability of cubs in Wytham could be improved by adopting the trapping procedures followed at Woodchester and Nibley. However, this would require more trapping cages, time and labour. Clearly, more research is needed to determine, for example, the optimal duration of the prebaiting period, or the optimal number of traps set per badger. Prolonged prebaiting may influence badger behaviour and invalidate certain research objectives. Furthermore, trapping procedures that work well at Woodchester may be unsuccessful elsewhere (C. Kollinsky, personal communication). The welfare implications of live-trapping badgers and of keeping them overnight also need to be considered. A greater understanding of the factors that influence badger trappability would also benefit the British and Irish governments that have largely or totally (in the case of the British) relied on cage-trapping for the removal of badgers that are suspected of having spread bovine tuberculosis to cattle. The possibility, highlighted but not proven by this study, that the trappability of badgers is reduced following culling operations could be an important consideration in the formulation of TB control strategies.

Acknowledgements

We thank the landowners who allowed us access to their property in order for the field work to be carried out. We also thank P. Johnson for statistical advice, and the staff at CSL and the many volunteers and students at WildCRU for their expertise and help in the field. We are indebted to the staff in the bacteriology section of VLA for testing the TB status of the badgers. S. Gillgan, J. Howell, J. Nichols and an anonymous referee kindly commented on an earlier draft. The badger study at Wytham was funded by the PTES, and at Woodchester Park and North Nibley by MAFF.

Received 24 November 1998; revision received 27 August 1999

Ancillary