Introduction

Social organisation of group-living vertebrates may vary in relation to extrinsic (environmental) and intrinsic (population-related) factors such as resource availability, habitat quality, season, climate, density and age structure. The pattern of social interactions in these species is therefore likely to show some variability between different populations. Numerous studies have shown that different individuals fulfil different roles in a social hierarchy. For example, dominant individuals can play a more important role in territorial defence than subdominants and males more than females (Schradin 2004; Le Roux et al. 2008; Rosell et al. 2008). However, it is also increasingly recognised that there may be considerable variation in behaviour between individuals, which extends beyond social group structures or hierarchical roles (Böhm et al. 2009). Such heterogeneities in individual social behaviour can have a considerable impact on population ecology and survival, including gene and information flow, resource utilisation and the spread of infectious pathogens that transmit through direct contact between individuals (Andreasen and Christiansen 1989; McCallum et al. 2001; Newman 2002; Meyers et al. 2005; Whitehead 2009). However, the contact patterns of many free-ranging wild vertebrate populations are still poorly quantified.

Epidemiological models on the direct transmission of pathogens often assume homogeneous mixing of individuals, either across the whole population or within its constituent social groups. Social organisation has also been shown to vary temporally in relation to changes in activity connected with breeding and dispersal (White and Harris 1994; Ji et al. 2005; Böhm et al. 2008). Even though epidemics can be highly sensitive to such patterns of mixing (Lloyd-Smith et al. 2005), spatial and temporal heterogeneities in a social organisation are rarely included in host–disease models. This is because data on contact rates in wild populations are scarce, largely because they are difficult to measure in the field. Advances in the measurement of direct interactions among free-living animals have become possible recently through the use of proximity loggers. Unlike traditional methods such as direct observation (Cowan 1987a, b; Mitani et al. 2002; Mollema et al. 2006), recording co-occurrence of individuals during sightings or censuses (Vonhof et al. 2004; Lusseau et al. 2006), and radio-telemetry (White and Harris 1994; White et al. 1995, 2003; Böhm et al. 2008), proximity loggers provide continuous and unbiased high-resolution data on the close contacts between free-ranging animals (Ji et al. 2005; Prange et al. 2006; Böhm et al. 2009).

Here we use proximity loggers to investigate the social contact behaviour of European wild rabbits (Oryctolagus cuniculus) in two sites with apparent differences in habitat quality in a temperate region of Australia. Unlike in Southern Europe where the rabbit is a keystone species which provides an important food source for several endangered predator species (Delibes-Mateos et al. 2007, 2008; Moreno et al. 2007), rabbits have become one of the most damaging vertebrate pests in Australia since their arrival in 1859 due to their devastating impacts on agriculture and biodiversity (Williams et al. 1995; Edwards et al. 2004; McLeod 2004). Since the introduction of myxomatosis in the 1950s, biological control using disease has formed a mainstay of Australian rabbit population control. In response to the declining effectiveness of myxomatosis, a second disease agent, rabbit haemorrhagic disease virus (RHDV), was introduced to Australia in the 1990s to aid the existing rabbit control (Cooke and Fenner 2002). RHDV, which presents a conservation problem in Southern Europe due to its severe impacts on the native rabbit populations (Moreno et al. 2007), was at first highly effective in reducing rabbit numbers also in Australia (Mutze et al. 1998). However, since the promising initial reports, the effectiveness of the virus has been inconsistent and mathematical simulations of the rabbit–RHDV system based on homogeneous mixing (Barlow and Kean 1998; Barlow et al. 2002) have failed to explain the variability in RHDV spread and persistence at different locations.

As one of the main routes of RHDV spread is through oral and nasal transmission during direct contact (Cooke and Fenner 2002), social organisation of rabbits may be an important factor contributing towards the observed variations in RHDV effectiveness. Rabbits form highly distinct breeding groups that typically consist of up to three males and nine females that maintain strong social bonds (Surridge et al. 1999). Various studies in different countries using direct observation and radio-tracking have suggested that interactions among rabbits within a social group occur commonly and individuals from neighbouring social groups also interact, although less frequently than group members (Cowan 1987a; Gibb 1993; Williams et al. 1995). However, the availability of proximity loggers means that information on social interactions of rabbits that occur both above ground as well as inside warren systems can be collected at a level of detail previously not possible. Here, we quantify intra- and inter-group contact rates and determine the extent of spatial and temporal heterogeneities in the social organisation of free-ranging wild rabbits. Specifically, we predict that there will be heterogeneities in contact behaviour (1) between different populations living in habitats with different characteristics likely to affect rabbit social organisation, (2) between sexes and groups over time according to breeding patterns, and (3) between individuals within and between groups.

Materials and methods

Study sites

The fieldwork was carried out in the Central Tablelands of New South Wales, between Orange and Bathurst (33°24′40″ S, 149°22′00″ E). We used two sites, Oaky Creek and Valpine, which are less than 3 km apart but different in terms of habitat quality and resource availability. Both sites are at an altitude of 700–950 m and have an annual rainfall of approximately 500 mm during the study year. Within both study sites, rabbit breeding commenced in February after an increase in rainfall following the dry summer months and continued through the rest of the study year (M. Marsh, unpublished data).

Oaky Creek is an unmanaged site used infrequently for sheep grazing. The vegetation consists of patches of trees, shrubs and unpalatable grasses. The study site was a small part (c. 0.05 km2) of the 200-ha paddock. The majority of rabbits in Oaky Creek lived under the numerous bramble (Rubus fruticosa spp.) bushes that extended over large areas. Valpine is a more intensively managed site, with more uniform, abundant pasture throughout the year, used for sheep and cattle production. The study area in Valpine covered c. 0.05 km2 and has rabbit burrows which were distributed in four parallel gullies c. 50 m apart. Neither site had a history of substantial rabbit control prior to the study. In Oaky Creek, RHDV has been detected on various occasions, but its impacts have been restricted to single warren systems rather than the whole site. In Valpine, RHDV epidemics have been more successful in reducing rabbit numbers in the past (Glen Saunders, personal communication).

Proximity loggers

We used proximity loggers (Sirtrack Ltd., Havelock, NZ) to record direct contacts among rabbits. These loggers have previously been used on other mammal species (Ji et al. 2005; Prange et al. 2006; Swain and Bishop-Hurley 2007; Böhm et al. 2009), and the accuracy of the loggers has been confirmed and quantified in both laboratory and field settings (Prange et al. 2006; Böhm et al. 2009).

The data logger units, which weighed less than 20 g (2% of the body weight of our smallest collared rabbit), were attached to the rabbits using a neck collar. The loggers transmit unique identification codes via a UHF transceiver within 1.5-s intervals while simultaneously “listening” for other loggers within a pre-defined detection range. Once another logger is detected, data are recorded until the contact is broken for the length of a pre-defined “separation time” (30 s in this study). Logger ID, date, start time and duration of the recorded contact of the transmitting unit are then stored in the logger memory.

Prior to use in the field, all loggers were tested in a laboratory and their detection range was set to 50–100 cm (equivalent to power setting 32). Since contact distances are reduced by around 50% in the field compared with laboratory tests (Böhm et al. 2009), this setting ensured that we would be recording only close contacts whilst minimising the likelihood of missing contacts completely. Each proximity logger emits a VHF signal transmitting at 150 MHz, which allows their location in the field. The logger battery life was 5–6 months for most collars.

Rabbit trapping and collaring

Trapping was carried out in both sites once every season throughout the study year (summer, December–February; autumn, March–May; winter, June–August; spring, September–November). At each site, a total of 100 traps was used during each trapping session. Traps were placed at least 5 m apart from each other near every active warren, and the number of traps near a warren was increased with warren size up to a maximum of 20 traps per warren. We aimed to conduct the trapping sessions over five consecutive days. Before every trapping session, pre-baiting was conducted on three occasions: 5, 3 and 2 days before the first trap night.

Rabbits were trapped in wire cages baited with diced carrot. The traps were set at dusk and checked the next morning. New captures within each trapping session were sexed, weighed and inspected for presence of fleas, abnormalities and clinical signs of myxomatosis. Blood samples were taken from a random selection of captures during each trapping session for the detection of RHDV outbreaks (see “Electronic Supplementary Material”, Fig. S1). The reproductive status of females was also recorded. A pair of metal ear tags containing a unique identification code was attached to the ears of all newly captured rabbits. ID numbers of all recaptures were noted for the capture–mark–recapture analysis (see “Electronic Supplementary Material”, Appendix S1).

Proximity loggers were fitted on a random selection of those adult rabbits caught that weighed more than 1 kg. When a collared rabbit died, the proximity logger was recovered from the field and fitted on a new rabbit during the next trapping session. Altogether, 126 rabbits were fitted with proximity loggers at least once during the study. Proximity data collection was carried out in two separate data periods: December 2006–April 2007 (summer and autumn; data period 1) and September 2007–November 2007 (spring; data period 2) (Table 1).

Table 1 Total number of proximity loggers on male and female rabbits in Oaky Creek and Valpine during data period 1 (Dec 06–April 07) and data period 2 (Sept 07–Nov 07)
Full size table

Radio-tracking and defining social groups

Radio-tracking on foot using hand-held Yagi antennae was carried out during most weeks of the study year to monitor the survival of collared rabbits and to collect data on their space use. Radio-tracking effort was distributed evenly across day and night, and locations were recorded using a hand-held GPS. To minimise autocorrelation in the radio-tracking data, location fixes for a rabbit were separated by at least 1 h and a maximum of four fixes was collected per 24 h. This protocol, designed to obtain a representative sample of movement over the periods of analysis (White and Garrott 1990), was in line with previous studies on rabbit social organisation in the area (White et al. 2003). The 1-h time interval is sufficient to avoid temporal autocorrelation, being well in excess of the time taken by an average individual rabbit to traverse its home range and using a maximum of four fixes per 24 h minimises any potential influence of specific time periods on overall home range estimates.

Rabbit warrens within the study sites did not have clear boundaries on the ground. Consequently, a priori allocation of rabbits into social groups on the basis of capture site was not considered to be reliable. Instead, following the approach of White et al. (2003) for the same study area, we based our initial allocation of rabbits into social groups on evidence from radio-tracking data. Based on our radio-tracking data, we identified three categories of rabbits. Firstly, we had 12 rabbits whose 95% minimum convex polygons (MCP) home ranges (estimated from inactive day-time fixes) did not overlap with any others. These rabbits were defined as being outside the social group structure and were excluded from the subsequent analysis. This left us with 145 out of an original 157 rabbits. For 137 of these remaining rabbits, we were able to define asymptotic home ranges based on 95% MCP estimated from inactive daytime fixes. These rabbits were very faithful to specific warren clusters and could therefore be allocated confidently to specific social groups on the basis of home range overlap patterns. Finally, we had eight rabbits (one in Oaky Creek during data period 1, five in Valpine during data period 1, and one in each of Oaky Creek and Valpine during data period 2) for which we had insufficient inactive fixes to provide reliable 95% MCP home ranges and hence could not be allocated to specific groups confidently based on radio-tracking alone. The majority of our radio-tracking data and field observations indicated clearly that social groupings were present, so it was important to determine the group to which these rabbits belonged. To do this, we used cluster analysis based on the proximity data. We created a similarity matrix from the presence and absence of contacts between each pair of rabbits using the Bray–Curtis coefficient (Zuur et al. 2007). We then used non-metric multidimensional scaling to confirm the exclusion of rabbits with no overlapping minimum convex polygons from the social groups (12 rabbits altogether throughout the study as mentioned above). For the cluster analysis, a dissimilarity value of 0.87 best represented the social group allocation that we had made for the 137 rabbits with sufficient radio-tracking data (see “Electronic Supplementary Material”, Fig. S2). We therefore used average linkage cluster analysis based on this cutoff value to confirm the allocation of the eight individuals with scarce radio-tracking data to social groups.

Data analysis

Home range overlap between social groups

Home range estimates were made in ArcView 3.2 using the ‘Animal Movement’ extension (Hooge and Eichenlaub 2000). To investigate space use of rabbit social groups, all radio-tracking data (including night-time fixes) recorded for rabbits belonging to the same social group were combined and 95% minimum convex polygons were calculated. This method makes no assumptions about the distribution of fixes and provides a more conservative estimate of the total area used than 100% minimum convex polygons (Boitani and Fuller 2000). The potential for interaction between individuals from different social groups was estimated by calculating their home range overlap as a percentage of the total area covered by their combined 95% minimum convex polygon home ranges.

Proximity data sorting

Prior to proximity data analysis, we removed all contacts recorded during trapping periods. The beginning of a trapping session was defined as sunset in the evening when traps were first opened, and the end of trapping session was defined as the average release time on the last trapping day (12 h after last trap opening). We also removed the first 12 h of data recorded after the end of each trapping session.

Throughout the study, several collared rabbits were lost to predation, other causes of mortality and emigration. In these circumstances, the date when a rabbit was last known alive within the study area (as determined by regular radio-tracking) was taken as the last date of data recording. This way, the amount of data was maximised without risking an accumulation of false contacts. In the event of logger battery failure (usually after 5–6 months of data recording), the last recorded contact was taken as the last day of battery life. During data period 1, most logger batteries had been exhausted by mid May 2007, so all data recorded after April 2007 were excluded from the analysis.

Because proximity loggers tend to record extended-duration contacts as multiple events (Prange et al. 2006), we combined the “broken” contacts that occurred within less than the pre-defined separation time (30 s). Of all recorded contacts, 37% lasted for 1 s, which is typical for proximity logger data (Ji et al. 2005; Prange et al. 2006; Böhm et al. 2009). Following the protocol from Böhm et al. (2009), all 1-s contacts were omitted from the analysis. Finally, in order to reduce spatial correlation in the analysis, contact data were divided into intra-group contacts (occurring among rabbits from the same social group) and inter-group contacts (occurring among rabbits from different social groups).

Intra- and inter-group contact measures

To standardise the contact measures between individuals, mean daily intra-group contact frequency C Freq and daily intra-group contact duration C Dur (in seconds) were calculated for each rabbit, based on the total number of contacts and the total contact duration divided by the number of days for which the proximity logger was attached. Each daily contact frequency and duration estimate was then divided by the number of individuals available for contact within the rabbit’s social group at any one time (i.e. the number of individuals within a social group wearing proximity loggers on that day). Therefore, intra-group C Freq and C Dur represent the average daily contact frequency and duration for a specific rabbit with any other collared individual within its social group on any one day. Thus, they are measures of contact behaviour per collar available rather than total measures. Prior to statistical analysis, intra-group C Freq was cubic root-transformed and C Dur was quadratic root-transformed to gain normality and homogeneity of variances.

Daily inter-group contact frequency and duration were calculated similarly, with the exception of dividing the daily inter-group contact frequency and duration estimates by the number of rabbits available for contact outside the individual’s social group at any one day. Because inter-group contacts were infrequent, only the intra-group contacts were analysed statistically.

Statistical analysis

The analysis of intra-group contacts was performed in the ‘R’ statistical programming environment (R Development Core Team 2009). Pearson’s correlations were performed on the transformed C Freq and C Dur in order to assess the consistency of data logging. Linear mixed-effects (LME) modelling was then applied to investigate the differences in C Freq and C Dur within each data period using the ‘lnme’ package (Pinheiro et al. 2008) in R. LME allows the analysis of hierarchically structured data with both fixed and random coefficients and multiple error terms (Zuur et al. 2007) and was suitable for our analysis of daily contact frequency and duration that were repeatedly sampled from the same rabbits and social groups. The random error structure was fitted using restricted log-likelihood (REML) for the C Freq and C Dur for each data period. It included a random intercept for rabbit ID for all models except C Freq model for data period 1, which included both rabbit ID and social group as random factors. The fixed factors included in the model selection were sex, time of day (daytime after sunrise and before sunset), site and month. The fixed components of the selected model were optimised using maximum likelihood (Burnham and Anderson 2002), while the results of the most optimal model (sensu Zuur et al. 2007) were presented with REML (Zuur et al. 2007). Inspection of correlation in residuals justified the exclusion of autocorrelation structures in the models.

Results

Home range overlap between social groups

The mean home range overlap between the social groups was 50% for Oaky Creek and 45% for Valpine during data period 1 (Fig. 1). The group home ranges overlapped less during data period 2 (10% in Oaky Creek and 5% in Valpine).

Fig. 1
figure 1

Overlap of home ranges between rabbit social groups in a Oaky Creek and b Valpine. Solid lines show group home ranges for data period 1 and dotted lines for data period 2

Full size image

Proximity logger returns

Due to rabbit population fluctuations as well as losses and additions of collared rabbits in the field, the percentage of adult rabbits collared varied between 29% and 87% in Oaky Creek and between 21% and 68% in Valpine (Table 2; see “Electronic Supplementary Material”, Fig. S1). The proximity loggers were recovered from 91% and 96% of collared rabbits in Oaky Creek and 93% and 96% of collared rabbits in Valpine for data periods 1 and 2, respectively. Three loggers had recorded no data, all of which were on rabbits that did not belong to a social group. The other nine transient rabbits had recorded highly infrequent contacts and were excluded from the analysis of intra- and inter-group contact behaviour. Four loggers (two pairs) ran out of memory space during data period 1 in Valpine, all of which were recovered and downloaded within a week from them becoming full. Altogether, the proximity loggers recorded 110,221 intra-group contacts representing a total duration of 3,946 h, and 73 (38 min) inter-group contacts were recorded.

Table 2 Percentage of the total population and adult population included in the contact data analysis in Oaky Creek and Valpine within each season during each data period. No proximity data were collected between May and August (Dec–Feb, summer; Mar–May, autumn; Sept–Nov, spring)
Full size table

Intra-group contacts

All rabbits included in the analysis made contacts with rabbits in their social groups. We recorded within-individual variation in contact rate and duration over the data periods (Fig. 2a–d). Mean intra-group C Freq was 1.54 ± 0.23 (SE) contacts/day, and mean intra-group C Dur was 202 ± 38 s/day (Table 3). However, both measures were positively skewed as indicated by the large differences between the mean and median estimates. The intra-group C Freq ranged from 0.004 to 12.0 contacts/day and C Dur from 0.01 s to 3.5 h/day. The frequency of high and low intra-group C Freq estimates in the rabbit populations also varied between the study months (Fig. 3). Intra-group C Freq and C Dur decreased with increasing seasonal rabbit abundance within the study sites, although because both measures are adjusted for the availability of collars a decrease in individual contact rates does not necessarily represent a decline in the total contact rate across the population. Both intra-group contact measures were higher in Valpine than in Oaky Creek for all seasons (Fig. 4; see “Electronic Supplementary Material”, Fig. S1 and Table S1, for further information on rabbit abundance).

Fig. 2
figure 2

Log duration (in seconds) of each contact made by selected individual male rabbits at the study sites. Each dot represents the log duration of single contact. The changes in the number of contactable collars within the individual’s social group (black line) and outside its social group (grey line) over the data period are also shown. a Male rabbit (ID 30), Oaky Creek, data period 1, a total of 72 contacts, of which two were inter-group contacts lasting for 9 s in total. b Male rabbit (ID 8), Oaky Creek, data period 1, a total of 317 contacts, of which none were inter-group contacts. c Male rabbit (ID35), Oaky Creek, data period 1, a total of 1,470 contacts in total, of which none were inter-group contacts. d Male rabbit (ID 27), Oaky Creek, data period 1, a total of 3,980 contacts, of which two were inter-group contacts lasting for 4 s in total

Full size image
Table 3 Mean (standard error; SE) and median (semi-interquartile range; SIQR) C Freq and C Dur (in seconds) for intra- and (b) inter-group contacts in Oaky Creek and Valpine for the two data periods
Full size table
Fig. 3
figure 3

Frequency histograms for C Freq estimates for all rabbits in Oaky Creek (in black) and Valpine (in grey) for a December 2006, b January 2007, c February 2007, d March 2007, e April 2007, f September 2007, g October 2007 and h November 2007

Full size image
Fig. 4
figure 4

Relationship between a mean C Freq (±SE) and b mean C Dur (±SE) and abundance of adult rabbits for Oaky Creek (in black) and Valpine (in grey). The variables were calculated for each study season (summer, autumn, spring). See “Electronic Supplementary Material” for further details on abundance estimates

Full size image

Significant positive correlations were found between intra-group C Freq and C Dur for both data periods (data period 1: r = 0.956, d.f. = 512, P < 0.001; data period 2: r = 0.949, d.f. = 200, P < 0.001; Fig. 5). The optimised fixed structures in the LME models based on cube root-transformed intra-group C Freq and C Dur were similar for both data periods (Tables 4 and 5). Sex was not a significant factor and was therefore not included in the most optimal structures for any of the models. Time of day was also not included in the most optimal model for C Freq (data period 1) and C Dur (data period 2). Month was a significant factor for both data periods for C Freq (data period 1 LME: F Month = 2.926, d.f. = 4, P < 0.05; data period 2 LME: F Month = 15.566, d.f. = 2, P < 0.001) and for C Dur for data period 2 (LME: F Month = 8.727; d.f. = 2, P < 0.001). In addition, for both data periods, there was a significant site–month interaction for C Freq (data period 1 LME: F Month:Site = 5.249, d.f. = 4; P < 0.001; data period 2 LME: F Month:Site = 7.710, d.f. = 2; P < 0.001) and C Dur (data period 1 LME: F Month:Site = 4.746, d.f. = 4; P < 0.001; data period 2 LME: F Month:Site = 4.362, d.f. = 2; P < 0.05). C Dur was significantly higher during daytime than night-time during data period 1 (LME: F Time of day = 10.955, d.f. = 1; P < 0.001). For data period 2, C Freq was significantly higher during night-time (LME: F Time of day = 8.035, d.f. = 1; P < 0.01).

Fig. 5
figure 5

Mean C Freq and C Dur estimates for each rabbit in Oaky Creek (in black) and Valpine (in grey) during a data period 1 and b data period 2. The means were calculated over the entire data periods

Full size image
Table 4 Table 4 Random and fixed effects for the linear mixed-effects models for differences in the intra-group daily contact frequency (C Freq) and duration (C Dur, in seconds) for data period 1 and data period 2. Factors considered are site (Oaky Creek, Valpine), month (Dec, Jan, Feb, Mar, Apr (data period 1); Sept, Oct, Nov (data period 2)) and time of day (night-time, daytime). For both models, sex (male, female) and the three-way interaction (site–month–time of day) were left out during the model selection procedure. Additionally, time of day was not in the optimal fixed structure for C Freq for data period 1 and C Dur for data period 2Anova tables indicating the significance of fixed effects for the linear mixed-effects models for differences in the intra-group daily contact frequency (CFreq) and duration (CDur, in seconds) for (a) data period 1 and (b) data period 2. Factors considered are: site (Oaky Creek, Valpine), month (Dec, Jan, Feb, Mar, Apr (data period 1); Sept, Oct, Nov (data period 2)), and Time of day (night-time, daytime). For both models, sex (male, female) and the 3-way interaction (site-month-time of day) were left out during the model selection procedure. Additionally, time of day was not in the optimal fixed structure for CFreq for data period 1 and CDur for data period 2
Full size table
Table 5 Random and fixed effects with coefficient values from the linear mixed-effects models for differences in the intra-group daily contact frequency (CFreq) and duration (CDur, in seconds) for (a) data period 1 and (b) data period 2. Factors considered are: site (Oaky Creek, Valpine), month (Dec, Jan, Feb, Mar, Apr (data period 1); Sept, Oct, Nov (data period 2)), and time of day (night-time, daytime). For both models, sex (male, female) and the 3-way interaction (site-month-time of day) were left out during the model selection procedure. Additionally, time of day was not in the optimal fixed structure for CFreq for data period 1 and CDur for data period 2. Results are shown relative to reference category (site = Oaky Creek; month = December (data period 1), September (data period 2); time of day = night-time; site-month = Oaky Creek:December (data period 1), Oaky Creek:September (data period 2))
Full size table

Inter-group contacts

The mean inter-group C Freq was 0.0007 ± 0.0004 (SE) (median = 0) contacts per day and mean C Dur was 0.007 s ± 0.004 (SE) (median = 0) (Table 3). Most inter-group contacts were short in duration (<1 min), and female–male contacts were the most frequent within both sites and data periods (Fig. 2a, data period 2 not shown). During data period 1, inter-group contacts were more common and longer in total duration in Valpine (N = 42, total duration = 588 s) than in Oaky Creek (data period 1, N = 20, total duration = 203 s) (Fig. 6). Fourteen rabbits in Oaky Creek and fifteen in Valpine were involved in these interactions, and the majority of the contacts occurred at night-time (74% in Oaky Creek, 61% of total duration; 87% in Valpine, 52% of total duration). During data period 2, only four inter-group contacts (total duration = 39 s) were recorded in Oaky Creek and seven (total duration = 91 s) in Valpine (data not shown).

Fig. 6
figure 6

Inter-group contact profiles over time, showing the cumulative duration (in seconds) of contacts in a Oaky Creek and b Valpine during data period 1. Each contact is shown for female–male (in white) and male–male (in black) and female–female (in grey) pairs separately. Each segment within a bar represents a different pair of individual rabbits. The data are summarised per week

Full size image

Discussion

Our use of proximity loggers in this study has enabled the first quantification of intra- and inter-group contact frequency of free-ranging wild rabbits within and between individuals and populations. These data have highlighted spatial and temporal heterogeneities in the social organisation of rabbits between populations and between and within social groups in the same population.

Intra-group contacts

Intra-group interactions among rabbits fitted with proximity loggers were less frequent than we expected based on previous research. Even though we recorded a large number of intra-group interactions overall and all rabbits within a social group made contacts with at least some of the other group members, contacts among rabbits sharing warrens, food resources and social associates were relatively infrequent when standardised to take into account the number of data recording days and collared rabbits within any given day in the social groups. On average, a rabbit made only 1.54 ± 0.23 (SE) (median = 0.54) intra-group contacts per day with the other rabbits in its social group which were fitted with a proximity logger, lasting for 202 ± 38 s (SE) (median = 29 s) (Table 3).

Our findings also demonstrated the presence of substantial heterogeneities in the daily intra-group contact rate and duration within and between individuals (Figs. 2a–d and 3). Even though most rabbits made relatively few contacts with collared rabbits in their social groups over time (for example, Fig. 2a), some made contacts with each other extremely frequently (for example, Fig. 2d). Rabbits with low average daily intra-group contact frequency and/or duration (min C Freq = 0.004 contacts/day, min C Dur = 0.01 s/day) may be socially displaced individuals living on the periphery of their social groups. Although these individuals may still be sharing resources and associates with other members of the social group, they are less involved in social interactions within the group. These rabbits may correspond to “surface dwellers”, which have previously been identified in several rabbit populations (King et al. 1984; Gibb 1993; White et al. 2003). The unusually high average daily contact frequency and duration (max C Freq = 12.0 contacts/day, max C Dur = 3.5 h/day) are most likely to occur between breeding pairs which typically exist within rabbit warren systems (Gibb 1993). These individuals formed a minority of the rabbits in this study and were more common in Valpine than in Oaky Creek (Fig. 5).

In addition to the heterogeneities in intra-group contact rates within and between individuals, we recorded significant diurnal and seasonal variations in the average daily intra-group contact rates (Table 5). For example, within both sites, C Freq was significantly lower during the early summer months (December and January) than during autumn. As the start of the breeding season (February) coincided with an increase in the frequency of high daily contact rates in the collared rabbit populations (Fig. 3; also see Fig. 2a–d for changes in contact rates for individuals), the temporal variation in contact behaviour is likely to be related to increased activity during breeding. Temporal heterogeneity in rabbit activity has also been recorded in various radio-tracking studies (Gibb 1993; Moseby et al. 2005). We also recorded significant differences in the average daily contact frequency and duration between sites over time (Table 3; Fig. 4), which may be driven by site-specific factors such as resource availability, habitat quality, population density and social group sex and/or age structure.

Inter-group contacts

Various radio-tracking studies have shown a regular occurrence of home range overlap between rabbits from neighbouring social groups, which suggests potential for frequent inter-group interactions (Gibb 1993; White et al. 2003). Even though our study also showed a high percentage of home range overlap between neighbouring social groups especially during data period 1 (Fig. 1), inter-group contacts were extremely infrequent despite the close vicinity of social groups and small size of the study areas (Table 3). On average, only 0.0007 ± 0.0004 (SE) (median = 0) inter-group contacts were made per day, lasting for 0.007 ± 0.004 s (SE) (median = 0). The highly uncommon occurrence of inter-group contacts suggests that wild rabbits may avoid close contact with individuals from different social groups. Cowan (1987a) also recorded rare occurrence of inter-group interactions during direct observations at daytime; however, our study is the first to quantify this conclusively without potential biases caused by limited sampling periods or disturbance during observations.

The inter-group contacts that did occur revealed further heterogeneities in the contact behaviour of rabbits. For example, only a small proportion of the collared rabbits made inter-group contacts, and intersexual encounters between social groups, presumably sexual of nature (Cowan 1987a), were more common and lasted for longer than intrasexual interactions (Fig. 6). The rarity of inter-group interactions between male–male and female–female pairs indicates a low occurrence of territorial encounters, thus providing empirical data to confirm a suggestion made by a previous study based on patterns of home range overlaps (White et al. 2003). Furthermore, inter-group contacts were more common in Valpine than in Oaky Creek, which may be due to less distinct social groups or more intense competition for mates and warrens.

Implications for disease transmission in rabbit populations

The outcome of an epidemic spreading directly though a contact network relies on both relatively frequent local (within-group) and occasional global (between-group) population mixing (Ball et al. 1997). Consequently, the social organisation of a host population can have important implications for the impact of infectious disease which is transmitted wholly or in part via direct contacts between individuals. Our quantification of intra- and inter-group contact rate (for the given the sampling intensity) among free-ranging wild rabbits can be directly applied to aid the parameterisation of mathematical models on the direct transmission of infectious diseases such as RHDV in rabbit populations. This is likely to improve the predictive power of epidemiological models on the rabbit–RHDV system, which traditionally have assumed homogeneous mixing of individuals (Barlow and Kean 1998; Barlow et al. 2002). However, direct transmission of a pathogen not only depends on contact rates but also requires one of the interacting individuals to be infected and the other susceptible, and the probability of infection resulting from such “appropriate” contacts also affects the outcome of the interaction (McCallum et al. 2001). A complete quantification of direct pathogen transmission based on empirical data in rabbits or any other wildlife host therefore remains to be achieved.

The variations in contact structure we have observed could contribute towards the observed patchy RHDV effectiveness in Australia and make the management of RHDV in southern Europe more problematic. The presence of heterogeneity in daily intra- and inter-group contact behaviour may have a crucial role in disease dynamics and could ultimately determine the outcome of an epidemic (Lloyd-Smith et al. 2005). Additionally, the higher past effectiveness of RHDV in Valpine may be influenced by the less variable intra-group contact rates over time and the more frequent interactions between rabbit social groups observed in this study, as both may create more opportunities for direct disease spread. We also detected a slight decline in daily contact frequency and duration of rabbits with increasing rabbit abundance (Fig. 4). This is contrary to the assumptions of many disease models which assume density-dependent contact rates, but empirical data have shown similar patterns in other social mammals (Böhm et al. 2009). This suggests that the social structure of rabbits may be more stable at high densities and also has implications for the transmission dynamics of diseases based on direct contacts. The drivers for the differences in contact behaviour need to be further investigated, and even a relatively infrequent mixing of more distant individuals in a contact network may be enough for an epidemic to spread through a population (Kleczkowski and Grenfell 1999).

Since it was not justifiable to fit proximity collars on young rabbits for welfare reasons, this study was limited to quantifying the contact patterns of sub-adult and adult rabbits (>1 kg) only. Because kittens and juvenile rabbits may behave differently to mature rabbits, the intra- and inter-group contact rates reported here may represent a biased sample of the contact behaviour of rabbit populations. For example, even though young kittens are unlikely to move far from their home warrens (Kunkele and Von Holst 1996), they may be important in the intra-group contact behaviour. Juvenile rabbits may interact with individuals from neighbouring social groups more frequently than mature rabbits due to less established social ranking and dispersal from natal territories (Kunkele and Von Holst 1996). Such age-specific heterogeneities may also affect the rate of RHDV transmission in contact networks. Finally, as our study only measured direct contacts among collared rabbits, we cannot draw conclusions on the role of indirect transmission via insect vectors, latrines and carcasses in RHDV effectiveness despite their likely importance in the spread and persistence of the virus (Cooke and Fenner 2002).

Conclusions

Our study provides the first quantification of intra- and inter-group contact rates among wild rabbit. Intra-group contacts were rarer than we expected based on the previous assumptions on the highly sociable nature of rabbits. Inter-group interactions were extremely infrequent and brief despite the considerable home range overlap between the social groups. We also found considerable spatial and temporal heterogeneities in the contact patterns of wild rabbits within and between individuals and evidence for a decline in contact rates with increasing density, which is contrary to the assumptions of many disease models. Data such as these can be used to aid the parameterisation of mathematical models on the direct transmission of diseases such as RHDV in rabbit populations. All else being equal, such heterogeneities in the social behaviour of rabbits are likely to lead to differences in the probability of disease transmission per unit time between individuals and may be a contributing factor in the variable effectiveness of RHDV in temperate regions of Australia. Future work should be targeted at identifying the ecological drivers of these heterogeneities.