The management and sustainable harvest of wildlife populations should be based on a thorough understanding of population density, age and sex structure, and the demographic processes. Live capturing of wild animals to collect data on density and demography, to mark individuals and fit radio-tags or GPS-collars is one of the most widely used tools in wildlife research (e.g. Krebs et al. 1995, Korpimäki et al. 2003, Cavanagh et al. 2004, Flowerdew et al. 2004). As with all sampling methods, a fundamental assumption of live capture is that the trapped population is a representative sample of the target population (Conroy & Nichols 1996, Greenwood 1996, Crawley 2002). Any particular capture method may, however, give a biased sample due to differences in capture probability among size, age or sex classes, and may be influenced by body condition, suggesting that a sampling protocol should combine different sampling methods (Conroy & Nichols 1996, Greenwood 1996, Laves & Loeb 2006).

Sampling biases may have profound implications for the estimation of population size, age structure, body condition, disease screening or indeed the estimation of any life history parameter, and may misguide management, conservation or control strategies (Baker et al. 2001, Elphick 2008). The analysis of trapping data to estimate density or survival recognises the problem of non-random sampling, and methods that recognise individual, time and behavioural heterogeneity in capture probability are widely available (White & Burnham 1999, Amstrup et al. 2005). However, although trapping data are regularly used to estimate, for example, body condition, disease prevalence, dispersal and is also used to test ecological hypotheses, parameterize models, assess habitat quality and population viability, trapping bias tends to be ignored (Baker et al. 2001, Pearson et al. 2003).

Mountain hares Lepus timidus are widespread throughout their northern Palaearctic distribution, and although generally occurring at low densities in Fennoscandia and the European Alps, hares can be locally abundant with densities in excess of 200 km⁻² (Angerbjörn & Flux 1995, Mitchel-Jones et al. 1999). Mountain hares are an important prey species, and as a widespread and at times locally abundant herbivore they can play an important role in shaping vegetation and predator assemblages. They are an important game species throughout much of their distribution and are designated as a species of conservation concern in some areas (Mitchel-Jones et al. 1999).

In this article, we use live-capture data collected during our long-term studies of the role of parasites, nutrition and management on the demography and population dynamics of mountain hares in Scotland to assess differences in body condition and relative numbers of individuals caught using two different methods of capturing hares: ‘live trapping’ using baited cage traps and ‘long-netting’ where hares are driven into nets.

Methods

Results

In total, we captured 190 individual hares in 2002 and 2005 using traps and long nets on the five different study areas in Central Scotland (see Table 1). After controlling for body size, method of capture and age were the only terms retained in the model (in order of removal from full model: Method*Age: χ²₁ = 0.19, P = 0.67; Method*Sex: χ²₁ = 1.35, P = 0.24; Sex*Age: χ²₁ = 3.17, P = 0.08; Sex: χ²₁ < 0.01, P = 0.99), and they had a significant effect on body mass; on average, individuals caught in long nets were significantly heavier than individuals caught in cage traps (estimate = 0.13 kg; SE = 0.05, t = −2.58), and juveniles were significantly lighter than adults (estimate = −0.37 kg; SE = 0.06, t = −6.45). Age, sex, method and the age*method interaction were retained in the model (in order of removal from full model: Age*Sex: χ²₁ = 0.01, P = 0.76; Sex*Method: χ²₁ = 0.15, P = 0.70). Age, sex and the age*method interaction had significant effects on the proportion of females and juveniles caught, indicating that adults were more likely to be caught than juveniles (logit estimate_juveniles = -0.99, SE = 0.26, z = 3.85, P < 0.001), females were more likely to be caught than were males (logit estimate_males = −0.53, SE = 0.15, z = −3.50, P < 0.001) and juveniles were less likely to be caught in nets than in traps (logit estimate_traps = 0.78, SE = 0.32, z = 3.3, P = 0.02).

Discussion

Live-capture of small to medium-sized mammals is a commonly used method in wildlife research. Inference from and effective use of live-capture data assumes that all individuals in the target population are equally likely to be captured. Whilst methods for estimating survival and density from capture-recapture data have long recognised individual differences in capture probability and a suite of methods, which incorporate individual heterogeneity into the analysis, have been developed (White & Burnham 1999, Amstrup et al. 2005); most other applications of live capture generally ignore heterogeneity or capture bias. Using data at first capture from two fundamentally different methods of capturing mountain hares, we present evidence that body condition at first capture differs significantly between capture method and age, and that the number of individuals caught at first capture differs between age, sex and with the interaction between age and method.

Mountain hares accumulate body fat over the winter, thereby attaining maximum fat reserves and body condition in February, coinciding with the onset of the breeding season, and they have least body fat in August at the end of the breeding season (Merwe & Racey 1991). Mountain hares, in particular females, preferentially graze on grasses, only switching to bulk browsing of more woody species, such as heather during winter when grazing is limited (Hulbert et al. 2001). The dietary change from grazing to browsing occurs at the end of the summer, when grass production declines, and it is associated with the digestive constraints related to a lower quality diet (Hulbert et al. 2001, Iason & Van Wieren 2006). The switch to a lower quality diet and minimum body condition at the end of the breeding season is associated with a peak in hare mortality at the onset of the winter, which particularly affects juveniles (Iason 1989).

Mountain hares and snowshoe hares Lepus americanus both forage to minimise risk, balancing the needs to feed against the risk of exposure to predation (Boutin 1984, Hulbert et al. 1996). Snowshoe hares in poorer body condition are more likely to be predated than hares in better condition through state-dependent risk taking where individuals are likely to forage in a more risk-prone manner with worsening body condition (Murray 2002). The data we used were collected during September-November which coincides with the period of greatest nutritional stress in mountain hares. The lower body condition of individuals caught in traps may reflect increased risk taking and willingness to enter baited traps, possibly to benefit from the bait. Not surprisingly, juveniles caught in early winter were, even after controlling for body size, lighter than were adults, and potentially less risk averse, accounting for the higher proportion of juveniles caught in traps. Although an alternative hypothesis to account for the high proportion of juveniles caught in traps compared to long nets may be the use of borrows by juveniles to escape while being chased during long-netting (Flux 1970). The energetic and nutritional costs of breeding means that females also attain the lowest body condition in early winter, however, although a greater proportion of females were caught in traps, we found no statistical evidence that females are more likely to be caught in traps than males.

The bias towards females in the trapped population that we report is consistent with the results of other studies. Thus, Flux (1970) reports consistently more females than males in samples of shot (males = 46.6%) and snared (males = 45.3%) mountain hares, but he also reports appreciable seasonal variation in the sex bias with the proportion of males in the samples declining over the course of the breeding season with decreasing mating behaviour from 55% during the mating season to 35% after the mating season, but with an approximately 1:1 sex ratio in samples shot in early winter. The proportion of juveniles in samples of shot mountain hares ranged from 21.7% in August to 55.3% in December (Flux 1970). Sex and age bias has also been reported for brown hares Lepus europaeus; a greater proportion of females (N = 32) than of males (N = 10) was caught in long nets compared to cage traps, and a greater proportion of juveniles was caught in non-baited traps (40%) (but where the bottom of the trap was lined with straw) than in long nets (17%; Smith et al. 2004). The consistent pattern in the proportion of different age and sex groups caught by live trapping with baited and non-baited traps and long-nets is intriguing and highlights the importance of understanding method bias in wildlife trapping studies, but different methodologies, lack of metadata and problems of ensuring a non-biased age and sex population baseline structure make direct comparison between these studies impossible.

Statistical inference is particularly challenging in ecology. Statistical populations are not the same as biological populations; even defining a study population may be impossible and whilst statistical methods can deal with random error, systematic error or bias, is more problematic and sampling regimes should aim to minimise bias (Greenwood 1996, Krebs 1999). In this article, we show that live-capture of mountain hares using cage traps results in a sample of individuals biased towards poor body condition which, as well as resulting in low estimates of population body condition, over-samples juveniles and females, at least when trapping in early winter when juveniles and females are likely in poor body condition and less risk adverse. When using capture-recapture data for density estimation these biases can and should be investigated, and, if necessary, allowed for in the analysis by using an appropriate model that accounts for heterogeneity in capture probability, inclusion as covariates or a stratified analysis (Amstrup et al. 2005). In other applications of live-capture data, the way to manage bias is less clear and investigators should aim at minimising the bias by using more than one method of capture and apply caution when interpreting the results from possibly biased samples of individuals.