European Journal of Wildlife Research

, Volume 58, Issue 3, pp 523–533

Occupancy, colonization and extinction patterns of rabbit populations: implications for Iberian lynx conservation


    • CESAM and Departamento de Biologia da Universidade de Aveiro
  • Joana Cruz
    • Terra eTudo Consultoria e Gestão de Recursos Naturais, Unipessoal Lda.
    • Environment DepartmentUniversity of York
  • Anabela Paula
    • CESAM and Departamento de Biologia da Universidade de Aveiro
  • Catarina Eira
    • CESAM and Departamento de Biologia da Universidade de Aveiro
  • Marisa Capinha
    • CESAM and Departamento de Biologia da Universidade de Aveiro
  • Isabel Ambrósio
    • CESAM and Departamento de Biologia da Universidade de Aveiro
    • Instituto de Investigación en Recursos Cinegéticos (IREC-CSIC-UCLM-JCCM)
  • Catarina Ferreira
    • Instituto de Investigación en Recursos Cinegéticos (IREC-CSIC-UCLM-JCCM)
  • Carlos Fonseca
    • CESAM and Departamento de Biologia da Universidade de Aveiro
Original Paper

DOI: 10.1007/s10344-011-0599-6

Cite this article as:
Sarmento, P., Cruz, J., Paula, A. et al. Eur J Wildl Res (2012) 58: 523. doi:10.1007/s10344-011-0599-6


In Mediterranean ecosystems, rabbits are a key prey species for many predators, such as the Iberian lynx, which is threatened with extinction and has gone extinct locally in several regions of its historical distribution range. One of these regions is Serra da Malcata Nature Reserve, Portugal, which is also currently proposed as a potential site for reintroduction. We intended to investigate annual variation, potential time trends and the effects of management practices on the rabbit population in Serra da Malcata as a model for future potential reintroduction areas. The rabbit population was monitored over 12 years (from 1997 to 2009) by counting latrines along linear transects. These data were used to estimate rabbit occupancy, colonization and extinction patterns using a likelihood-based method including habitat, population and topographic covariate effects. Our results suggest that initial occupancy, when management practices were absent, was driven by the presence of Erica spp. and Cistus ladanifer shrubs and by distance to summits. Site colonization was positively influenced by the presence of edges between shrubs and pastureland and by patterns of rabbit distribution in the previous sampling season. On the other hand, local extinction was negatively influenced by edges. We conclude that the increase in rabbit occupancy and local colonization patterns was clearly associated with management actions (particularly, the creation of pasturelands), although the recovery of the species was noticeably limited by previous patterns of spatial distribution.


Conservation planningLynx pardinusOryctolagus cuniculusOccupancy modelsProgram PRESENCE


The re-establishment of populations requires an understanding of the fundamental conditions of the ecosystem and its adequacy to harbour viable wild populations. In the case of predators such as the critically endangered Iberian lynx (Lynx pardinus) and the Spanish imperial eagle (Aquila adalberti), both the lack of suitable habitat and the low abundance of their main prey, the European wild rabbit (Oryctolagus cuniculus), hamper their natural expansion from the remaining source populations. In fact, the widespread decrease of rabbit abundance in the Iberian Peninsula over the last century is considered as one of the main causes for the Iberian lynx’s current near-extinction status throughout its native range (Guzmán et al. 2004; Sarmento et al. 2009). Rabbit decline, which started in the middle of the past century (Delibes-Mateos et al. 2009), was mainly a consequence of changes in land use (Moreno and Villafuerte 1995; Delibes-Mateos et al. 2010) and of the long-term effect of two viral diseases, myxomatosis and, more recently, rabbit hemorrhagic disease (RHD; Villafuerte et al. 1996). Presently, the wild rabbit is listed as a ‘near-threatened’ species in Portugal (Cabral Almeida et al. 2005) and as a ‘vulnerable’ species in Spain (Villafuerte and Delibes-Mateos 2007) although it is not listed in any protected species legislation in both countries.

The current status of Iberian lynx populations, represented by only two isolated nuclei in Spain (Doñana and Andújar-Cardeña in the eastern Sierra Morena) which count with less than 200 individuals in the wild (Guzmán et al. 2004), turns natural expansion and recolonization of areas included in the former distribution range of the species extremely limited. Furthermore, the areas where lynx populations have persisted since the 1980s roughly match those currently less favourable for rabbits (Real et al. 2009), which could become an obstacle for this predator conservation efforts.

Recently, a reintroduction program was proposed in order to re-establish extinct populations all over the species’ former range (Simón et al. 2009). This program, called IBERLINCE and supported by EU LIFE+ funds, aims to recover lynx populations through habitat and prey management followed by lynx reintroduction. In this context, promoting the recovery of rabbit populations is a critical task (Ferreira and Delibes-Mateos 2011), and the first essential step is to understand the current dynamics of ecosystem conditions that sustain this key prey (Palomares et al. 2001).

The Serra da Malcata Nature Reserve (SMNR), such as other similar areas in the Iberian central system (e.g. Monfrague, Granadilla), is a potential site for Iberian lynx reintroduction in the future (IUCN 2005). This reserve was created in 1988 with the specific purpose of preserving Iberian lynx populations. In the late 1970s, an important population that harboured a total of 30 individuals occupying an area of 200 km2 had been described for this area (Palma 1980). This population was part of the Spanish meta-population of Sierra da Gata–Granadilla–Santa Cruz–Cilleros that included a total of 90 individuals that occupied 1,500 km2 (Rodríguez and Delibes 1992). In order to increase the habitat carrying capacity for the future reintroduction of this carnivore, a long-term project dedicated to the recovery of the rabbit population was established (Table 1). Also, the implementation of a wild rabbit large-scale monitoring program was essential to fully understand the species’ status progression and its response to the management techniques applied. Due to the idiosyncrasies of rabbit population dynamics, consistent methods needed to be applied to provide accurate information on the species’ occupancy, colonization and extinction patterns. Only then could management strategies be progressively adapted to these trends. In this context, occupancy models provide reasonable estimates of population status and trends (Pollock et al. 2002; Bailey et al. 2004).
Table 1

Chronology of management practices carried out in Serra da Malcata for improving habitat suitability for rabbits



Accumulated total managed


Pastureland (ha)

Artificial warrens


Creation of 40 ha of pastureland and 28 artificial warrens




Renewal of 40 ha of pastureland, additional creation of 20 ha of pastureland and 20 new artificial warrens




Renewal of 20 ha of pastureland, additional creation of 30 ha of pastureland and 10 new artificial warrens




Renewal of 60 ha of pastureland




Renewal of 40 ha of pastureland, additional creation of 49 ha of pastureland and 28 new artificial warrens




Renewal of 40 ha of pastureland




Renewal of 40 ha of pastureland



In this paper, we investigated abundance, seasonal variation, potential time trends (occupancy, colonization and extinction) and the effects of management practices in the rabbit population in SMNR using the site occupancy models developed by MacKenzie et al. (2002). We hypothesized that rabbit site colonization is favoured by mosaic landscape where pasturelands are intercepted by shrubs, with effective edges and adequate shelter. A long-term dataset allowed for an in-depth analysis of the factors affecting the rabbit population over the study period (1997–2009). More importantly, because we are focusing on mechanistic processes rather than on the population itself, this analysis potentially provides valuable information to develop an adaptive management program for lynx recovery in other similar areas of its native geographic range proposed for reintroduction.

Study area

Study site

This study was carried out in a 54-km2 area inside the 200-km2 SMNR (Fig. 1) in Portugal, located near the Spanish border (40°08′50″ N–40°19′40″ N and 6°54′10″ W–7°09′14″ W). The climate is Mediterranean, characterized by mild to cold, wet winters and warm, dry summers. This area has experienced significant structural changes since the mid twentieth century (Fernández Alés et al. 1992). Up to 1940, shepherds and farmers managed scrubland vegetation using traditional agricultural techniques, such as slash-and-burn followed by farming, which created a mosaic effect that could sustain a high rabbit density (Moreno and Villafuerte 1995; Delibes et al. 2000). During that period, these patches were surrounded by Mediterranean forest, which also guaranteed a suitable cover habitat for lynx (Palomares 2001). In the 1940s, this region was subjected to intensive farming and, as a result, scrublands and Mediterranean woodlands were nearly eradicated. The abandonment of these practices a decade later together with rural exodus promoted the re-colonization of the landscape matrix by shrubs of mainly resistant species such as Cistus spp. By the 1970s, the original forest and scrubland habitats had been extensively replaced by conifer and eucalyptus stands for timber and wood pulp production. Therefore, today, the dominant vegetation types are dense scrubs of Cistus ladanifer and Erica australis with scattered woodlands of Quercus rotundifolia and Arbutus unedo, located in the steepest slope areas. Commercial plantations of Pinus spp., Eucalyptus globulus and Pseudotsuga menziesii are also present.
Fig. 1

Geographic location of the study area, management interventions and changes in vegetation types during the study period (1997–2009)

Management interventions

Conservation actions included the creation of a matrix of pasturelands inside scrubland patches and the construction of artificial warrens (Table 1; Fig. 1; Sarmento et al. 2009) in order to improve habitat carrying capacity for lynx. Pasturelands, with approximately 1 ha, were created by plowing the soil and sowing with Triticum spp., Festuca spp. and Avena spp. Some of these pastures were frequently renewed following the same protocol (Table 1). Rabbit artificial warrens consisted of pilled tree roots covered by soil and were placed in a grid arrangement in the edges of the pasturelands always near shelter coverage to promote usage by rabbits (Fernández-Olalla et al. 2010). By 2006, these practices were then ceased and scrubland invaded again all pasturelands that had been implemented, except by for ha that was maintained without intervention (Fig. 1).

Materials and methods

Wild rabbit abundance and distribution

Wild rabbit population abundance and distribution were assessed by counting latrines along linear transects (Calvete et al. 2006). Latrines were defined as groups of a minimum of 50 rabbit pellets in an area of approximately 30 cm in diameter (Iborra and Lumaret 1997).

The entire SMNR was divided into 52 2 × 2-km UTM squares, including 18 corresponding to the study area (Fig. 1). Two observers counted rabbit latrines along a 1-km transect defined in each square, within a 6-m-wide band (3 m on each side covered by each observer) along the transect. The respective geographic location of the latrines was registered in a GPS navigator. In each transect, we defined ten random points with a 20-m buffer area (sites) where rabbit presence/absence was assessed by intensive latrine searching. Random points were defined by using X-Tools and Geoprocessing extensions in Arcview 3.2. These polygons were built over 40 m apart in order to avoid buffer juxtaposition. Line transects and sites were sampled from 1997 to 2009 during two sample occasions per year, except for summer 2004 and the whole year of 2008 due to logistic constraints. Annually, sampling occurred during early autumn (rabbit’s annual low-density peak) and during spring (reproductive peak; Gonçalves et al. 2002).

Modelling wild rabbit occupancy, colonization and extinction patterns

Data collected in the field were used to generate rabbit distribution maps by interpolation using the Inverse Distance Weighted Interpolation Method (Kliskey et al. 2000). Spatial interpolation was crucial to provide grid surfaces across the study area representing population nuclei. We attributed the number of latrines per transect (1 km) to the transects’ centroid. This value was optimized through cross-validation by minimizing the root mean square prediction error. We restricted the search neighbourhood to 1,000 m considering the low dispersal rate of rabbits (Cowan 1991).

For quantifying the isolation level of rabbit nuclei, we used the index described by Krauss et al. (2004):
$$ I = \sum {e^{{ - {\text{dij}}}}}{A_j} $$
where Aj is the area (m2) of the nearest nucleus to nucleus i and dij is the Euclidean distance between them (m). The index varies from 0 to 1 and the values close to 0 indicate a high degree of isolation.

A Geographic Information System (GIS) database was constructed for the study area using aerial photographs. This system was updated as the vegetation cover changed over the years. The GIS database was also updated every time habitat interventions were made through their inclusion in the pastureland shapefiles. The latter was designed by GPS and the evolution of the vegetation classes were updated by verification in the field.

We estimated rabbit occupancy (ψ), colonization (δ), extinction (ε) and detection probability (ρ) using a likelihood-based method adapted to imperfect detectability (MacKenzie et al. 2002; MacKenzie and Royle 2005). The arrangement for sampling design was based on Pollock’s robust design (Pollock 1982). In each site (random points), rabbit detections were coded as 1 (detection) and 0 (non-detection). Then, the records were transformed into detection histories for each site (Xi) which were used with a product multinomial likelihood model, to estimate occupancy parameters (MacKenzie et al. 2003), as follows:
$$ {\mathbf{L}}({\psi_{{1}}},\varepsilon, \delta, \rho |{X_{{1}}},{ }.{ }.{ }.{X_n}){ } = {\pi_i}_{{ = {1}}}{\mathbf{Pr}}\left( {{X_i}} \right) $$
where ψ1 is a vector of site occupancy probabilities for the first primary sampling period, ε and δ are matrices of local extinction and colonization and ρ is a matrix of detection probabilities. We define colonization (δt) and local extinction (εt) probabilities as (Mackenzie et al. 2003):
  • δt = the probability that an unoccupied site in season t is occupied by the species in season t + 1; and

  • εt = the probability that a site occupied in season t is unoccupied by the species in season t + 1.

These dynamic processes represent the probabilities of a site transitioning between the occupied and the unoccupied state in consecutive seasons.

We ran analyses in program PRESENCE (http://www.mbr-pwrc.usgov/software.html) using multiple-season models, including covariate effects (Table 2). The study was divided in primary and secondary occasions. For each season, we considered two primary occasions (we assumed that occupancy does not change between them) that consisted in each observer survey. The secondary occasions corresponded to each sampling season (among which the occupancy patterns can change). We used a two-step approach to analyze data. First, we assessed the effect of season and year on detection probabilities while keeping other parameters constant (i.e. ψ [.] ε [.] δ [.] ρ [variable]) (Table 2). Second, we used the best-fitting model for detection probabilities and combined it with a set of a priori models integrating covariates to explain the observed patterns. We excluded potential predator’s occupancy as an additional covariate in the model since a previous study indicated an absence of correlation between this factor and rabbit abundance (Sarmento et al. 2011). Since we started from a situation without pasturelands or significant open areas (Fig. 1), we began by analyzing rabbit initial occupancy. Then, we formulated a set of a priori biological hypotheses that we used to develop specific models that explained variation in wild rabbit occupancy patterns:
Table 2

List of covariates used as predictor variables in the occupancy models for the wild rabbit in Serra da Malcata Nature Reserve (Portugal), 1997–2009






 Erica spp. and C. ladanifer scrubland

Sample covariate

ErCi scrubland

% of areas dominated by dense shrubs of E. australis, E. umbellata and C. ladanifer, inside of the 200-m buffer area


Sample covariate


% of areas lacking dense vegetation cover used for crop production (generally corn and wheat) or with spontaneous herbaceous vegetation, inside the 200-m buffer area


 Number of artificial warrens

Sample covariate

Artificial warrens

Number of artificial warrens inside the 200-m buffer area


 Shannon landscape diversity index

Sample covariate

Shannon index

Measure of relative patch diversity inside a buffer area with a 200-m radius around the site; equals 0 when there is only one patch in the landscape and increases as the number of patch types or proportional distribution of patch types increases

 Patch size

Sample covariate

Patch size

Size (in hectares) of the habitat patch where the site is located

 Edge between scrubland and pastureland

Sample covariate


Linear measure, in m/km2, of the ecotone inside the 200-m buffer


 Rabbit abundance in the previous season

Sample covariate

Abundant previous season

Number of latrines/km in the respective transect during the previous season

 Distance to occupied nuclei

Sample covariate

Dist. nuclei

Distance, in metres, from the sampling point to an occupied nuclei

 Isolation index

Sample covariate

Isolation index

Krauss et al. (2004)



Site covariate




Site covariate



 Distance to summit

Site covariate

Distance summit

Distance from the site to the nearest peak (a place with some significant amount of topographic prominence)

  1. 1.

    Site colonization is positively affected by mosaic landscape where pasturelands are intercepted by shrubs, with an effective edge effect and adequate shelter in the proximity of occupied areas (Moreno and Villafuerte 1995; Carvalho and Gomes 2003; Delibes-Mateos et al. 2009).

  2. 2.

    Extinction is influenced by the degree of isolation of occupied nuclei and by habitat alterations that resulted in the regression of matrix areas.


Occupancy, colonization and extinction probabilities were modelled as a function of two groups of variables: (1) sampling covariates (that could vary between occasions) and (2) site covariates (that do not vary over time; Table 2). The first group included information on patch composition (scrub and pastureland), shelter conditions (number of artificial warrens), landscape variables including suitable habitat patch size (according to Carvalho and Gomes 2003), landscape diversity index and ecotone length and, finally, rabbit distribution patterns in the previous season. Site-specific covariates included topographic variables only (Table 2). Vegetation variables were measured inside a buffer area with a 200-m radius around the site, corresponding to an average rabbit home range area (Lombardi et al. 2007).

All continuous covariates were standardized to z-scores preceding the analyses. The ranking of candidate models was conducted using the Akaike Information Criterion corrected for small sample size (AICc) by calculating their Akaike weights (Burnham and Anderson 2002). Models with ΔAICc values ≤2 from the most parsimonious model were classified as robustly supported.

Akaike’s weights (ω) were used to further interpret the relative importance of each model’s independent variable. Except if a single model had a ωi >0.90, other models were also considered when inferring about the data (Burnham and Anderson 2002). A 90% confidence model set was created by summing all ωi until achieving 0.90. The selected models allowed the calculation of the average estimates of initial occupancy and seasonal estimates of local extinction and colonization probabilities using program MARK (White and Burnham 1999). We used the estimates from program MARK to calculate season-specific (denoted as t) site occupancy probabilities using the formula from MacKenzie et al. (2003):
$$ {\Psi_{\text{t}}} = {\Psi_{{{\text{t}} - {1}}}}\left( {{1} - {\varepsilon_{{{\text{t}} - {1}}}}} \right) + \left( {{1} - {\Psi_{{{\text{t}} - {1}}}}} \right){\delta_{{{\text{t}} - {1}}}} $$


Wild rabbit abundance and distribution

Rabbit abundance (average number of latrines per kilometre) varied substantially across years (Fig. 2), starting at 11.65 latrines/km in October 1997 and peaking at 96.52 latrines/km in October 2005. There were noticeable rabbit abundance declines in 2003 and in 2009 (Fig. 2). For the entire study area, we did not find a significant correlation between rabbit abundance and the availability of pasturelands (r2 = 0.557; P = 0.093), although a positive significant correlation was found between rabbit abundance and pastureland areas inside occupied sites (r2 = 0.931; P = 0.041).
Fig. 2

Changes in wild rabbit relative abundance (number of latrines per kilometre) in Serra da Malcata Nature Reserve, Portugal (1997–2009). Dashed lines indicate time of management interventions (see Table 1)

Initially, the species was mostly concentrated in two isolated nuclei located in two summits covered by scrubland vegetation (Fig. 3). As the management procedures spread throughout the area (Table 1; Fig. 1), particularly in 2000 and 2001, rabbit distribution expanded and started to exhibit a more continuous pattern. In 2003, following the severe species decline, the rabbit’s distribution became fragmented again (Fig. 3). Later, management actions were applied in a wider area (Fig. 1) and intensified (Table 1), and rabbits spread throughout the study area (Fig. 3). In 2009, after habitat management measures were ceased, the rabbit’s geographic distribution range became again more scarce and fragmented (Fig. 3), although with more population nuclei established throughout the study area.
Fig. 3

Patterns of wild rabbit spatial distribution using the Inverse Distance Weight Interpolation method in Serra da Malcata Nature Reserve, Portugal (1997–2009)

Initial occupancy

We did not detect an effect of season or year in detection probabilities (Table 3). So, in subsequent analysis, detection probability was considered a constant parameter with an estimated value of 0.95 ± 0.03 (SE).
Table 3

Model comparison for determining the effects of season and year in the detection probability (ρ) for rabbit in Serra da Malcata Nature Reserve, Portugal, 1997–2009




−2 L

N parameters

ψ (.) ε (.) δ (.) ρ (.)





ψ (.) ε (.) δ (.) ρ (season)





ψ (.) ε (.) δ (.) ρ (year)





No single model emerged as the top-ranking model, i.e. ωi > 0.90, so the averaged model initial occupancy value was chosen as the final estimate (\( \widehat{\psi }\mathop{{}}\limits \)= 0.115 ± 0.005 SE; Table 4). According to our model selection, initial occupancy was best modelled as a function of ErCi scrublands and Dist. summit (Table 4), in a situation where pasturelands were almost absent (Fig. 1). In all selected models, ErCi scrublands had a positive linear trend in the initial occupancy (average \( \widehat{\beta } \)= 20.260; 16.150 to 24.370, 95% CI). The variable Dist. summit exhibits a negative linear trend in the initial occupancy (average \( \widehat{\beta } \)= − 0.038, 0.0051 to 0.025, 95% CI not overlapping 0).
Table 4

Best ranked models (ΔAIC < 2.0) for rabbit occupancy (ψ) in Serra da Malcata Nature Reserve, Portugal, 1997–2009 (cumulative weights ≥0.91)




−2 L

N parameters

\( \widehat{\psi } \)1997

ψ (ErCi scrublands + Dist. summit) ε (eSP + Isol. index) δ (eSP + dist. nuclei) ρ (.)






ψ (ErCi scrublands + Dist. summit) ε (eSP + Isol. index) δ (dist. nuclei) ρ (.)






ψ (ErCi scrublands + Dist. summit) ε (Isol. index) δ (esP + art. warrens + dist. nuclei) ρ (.)






\( \widehat{\psi } \)1997 occupancy probability in 1997

Local extinction and colonization

The model with the greatest support was ψ (ErCi scrubland + Dist. summit) ε (eSP + Isol. index) δ (eSP + Dist. nuclei) ρ (.) (ωi =0.58; Table 4), suggesting that local extinction and colonization patterns were mostly influenced by the presence of edges between shrubs and pastureland and by the patterns of rabbit distribution in the previous sampling season. According to this model, the eSP inside the site had a positive influence on colonization (\( \widehat{\beta } \)=1.150; 0.020 to 1.280, 95% CI not overlapping 0). As expected, the relation between Dist. nuclei and colonization was negative (\( \widehat{\beta } \)=  0.516; 0.187 to 0.845, 95% CI), meaning that the probability of colonization of a given patch decreases with distance to the next rabbit population. Local extinction was significantly influenced by eSP in a negative linear trend (\( \widehat{\beta } \)=  2.981; 3.642 to 2.194, 95% CI) and by Isol. index in the same pattern (\( \widehat{\beta } \)=  3.023; 4.009 to 2.34, 95% CI). The second best-ranked model had lower support but showed a similar pattern (Table 4). Colonization was negatively influenced by Dist. nuclei (\( \widehat{\beta } \)=  3.004; 4.217 to 1.791, 95% CI). Extinction was significantly influenced by Isol. index in a negative linear trend (\( \widehat{\beta } \)=  3.023; 3.642 to 2.194, 95% CI) and by eSP with a similar trend (\( \widehat{\beta } \)=  1.588; 1.843 to 1.333, 95% CI). For the third best-ranked model, eSP and Art. warrens influenced colonization in a positive linear trend (\( \widehat{\beta } \)=1.150; 0.010 to 2.310, 95% CI; \( \widehat{\beta } \)=1.030; 0.010 to 2.070, 95% CI; respectively) and Dist. nuclei in a negative linear trend (\( \widehat{\beta } \)=  2.206; 3.888 to 0.124, 95% CI). Extinction was negatively influenced by Isol. Index (\( \widehat{\beta } \)=  1.902; 2.746 to 1.058, 95% CI).

We estimated site occupancy probabilities for those years subsequent to 1997 from model-averaged estimates of initial occupancy, local extinction and colonization (Fig. 4). Occupancy probabilities were positively correlated with abundance (r2 = 0.677, P = 0.001) and with eSP (r2 = 0.773, P = 0.0001).
Fig. 4

Probability of wild rabbit occupancy, colonization and extinction in Serra da Malcata Nature Reserve, Portugal (1997–2009)


Factors shaping rabbit distribution

We found that vegetation cover and distance to summit determined rabbit initial occupancy. With respect to vegetation, C. ladanifer and Erica spp. dominated the shrub layer in the study area. In southwestern Portugal, Beja et al. (2007) found a positive association between rabbit abundance and C. ladanifer shrubs, which provided protection against predators and an abundant herbaceous understory. However, in SMNR (present study), the herbaceous cover is scarcer mostly due to the dominant shrub-specific characteristics. Erica species are associated with poor acid soils (Cruz and Moreno 2001; Zas and Alonso 2002), which allow for low plant diversity. Also, decades of continuous fires and intensive agriculture followed by abandonment probably led to environmental stress, affecting soil properties and decreasing soil depth and water availability (Bielsa et al. 2005). It is quite clear that the initial situation in 1997 presented low suitability for rabbits since they reached higher densities in ecotones between open and shrub-covered areas, which allow them to feed and avoid predators by moving from one habitat to another at different times of the day (Moreno et al. 1996; Palomares and Delibes 1997). This landscape matrix was almost absent in the initial scenario of the present study when the shrub layer was continuous and food resources were low. The scarce and fragmented distribution of rabbits in 1997 and their association with shrub-covered summits could be an indicator of population stress caused by profound habitat alterations and the incidence of viral diseases. Furthermore, since fleas are an important vector for disease transmission in the European rabbit (García-Bocanegra et al. 2010), the initial association of the species with summit areas could also be related with a lower density of fleas in those areas because they are less wet (Cooke 1995). Our results suggested that the occupancy and local colonization patterns observed over the ensuing years were clearly associated with the conservation actions (habitat management measures) developed to improve habitat carrying capacity for lynx, particularly the creation of pasturelands. Furthermore, local colonization was clearly limited by site distance to occupied areas. In fact, the eSP 95% CI of the regression coefficients barely overlapped 0, which is explained by the areas located farther south that never got to be colonized by rabbits. Although these areas had also been subjected to conservation actions, their distance to occupied sites prevented them from being colonized by rabbits. In this sense, wild rabbit translocations using individuals from high-density nuclei could represent a viable possibility to reduce the distance between rabbit populations in low-density areas (e.g. Rouco et al. 2008). Nevertheless, in order to become a potential donor population (and, hence, a source of rabbits to be translocated elsewhere), active management needs to be implemented to increase rabbit density in areas where rabbits are present. Such a procedure could reduce considerably the negative effects of restocking, related to genetic and epidemiological issues (Delibes-Mateos et al. 2008), by assuring the movement of animals within a very restricted geographic range.

Our results also suggest that rabbits actually present a low dispersion capacity. Considering the initial scenario, the maximum dispersion observed corresponded to about 1,400 m (between a site colonized in 2006 and the nearest occupied site in 1997) and the average annual colonization distance was 211 m (SE = 56; range = 156–367). Our results concur with the findings of Calvete et al. (2006) in northeastern Spain, where rabbit populations are associated with areas presenting interspersed patches of natural vegetation and crops and therefore need to travel short distances between food and shelter patches.

The reduction in rabbit occupancy observed in 2003 (Fig. 3) could not be explained by habitat alterations since all pasturelands had been renewed in the previous year. In this case, this phenomenon could be the result of a RHD outbreak, although it was not possible to obtain data supporting this hypothesis. According to Virgós et al. (2003), while habitat variables can be suitable predictors of rabbit population changes in high-density areas, in low-abundance or fragmented distribution areas other stochastic factors (like disease) should be considered.

We detected a direct effect of artificial warrens in rabbit occupancy. Although rabbits can live aboveground when there is a dense scrub layer (Beja et al.), their abundance and distribution seem to be strongly influenced by soil type (Delibes-Mateos et al. 2009) mainly because they are usually dependent on warrens for breeding and for protection from predators and climatic extremes (Parer and Libke 1985). Therefore, in poor habitat conditions, artificial warrens can in fact be a useful management tool to increase habitat carrying capacity (Rouco et al. 2011). Habitat management, particularly the creation of pasturelands, seems to be the trigger for rabbit recovery. Villafuerte et al. (1996) related the increase of rabbit numbers observed in Doñana National Park with the availability of high-quality food. Our pasturelands, dominated by wheat (Secale cereale), Festuca spp., young shrubs (Baccharis trimera, Halimium alyssoides) and other herbaceous plants, provide a large spectrum of food items that can be used by rabbits (Ferreira and Alves 2009), particularly during summer when food resources become scarcer (Rueda et al. 2008). Additionally, the implementation of these measures seems to be particularly relevant in lower-quality habitats (Moreno and Villafuerte 1995). In this context, rabbit populations may benefit more from habitat management in less productive environments, which is probably the reason that we observed a peak in rabbit abundance in 2006 during the theoretical low phase, most likely a direct result of the management practices that were applied in the previous spring.

Implications for Iberian lynx conservation

In recent years, a significant amount of Iberian lynx conservation projects have been implemented in the Iberian Peninsula (Simón et al. 2009) and they all included the increase of rabbit density as their main goal, using habitat strategies similar to the one we used in this study (Guzmán et al. 2004; Simón et al. 2009). However, most of these projects consider habitat management to be effective for recovering rabbit populations, but only at a small scale. In fact, for an increase in rabbit abundance to be biologically relevant in a given area, rabbit density should be higher than four rabbits/ha during spring and at least one rabbit/ha during autumn (Palomares et al. 2001) in order to support a lynx population. In this context, two case scenarios can arise: (1) the recovery of rabbit in areas were lynx still occurs (Simón et al 2009) and (2) the recovery of rabbit in areas proposed for lynx reintroduction but where the species is not present (Simón et al. 2007). In the first case, habitat management measures should be applied at a micro-scale (home-range scale) in order to sustain lynx reproduction in selective cases. In the second situation, management protocols should contemplate the enabling of the target area to support a viable lynx population. In this case, if the target area is below the biological threshold of rabbit density, it is crucial that rabbit abundance shifts to a level of biological relevance. Because most of habitat management techniques are expensive to apply, habitat management should be efficient in reverting rabbit status (Ferreira and Delibes-Mateos 2011). Our results support this belief, particularly in pre-selected areas for reintroduction that present similar characteristics to our study area (e.g. Sierra de San Pedro, Granadilla; Simón et al. 2007). These areas are quite similar to Malcata in terms of conservation problems, vegetation structure and rabbit density. So, the experience gathered during these 12 years will be important for those areas, too.

It is clear that management interventions based in the improvement of feeding conditions near shelter patches and warrens are critical for increasing rabbit density. Our data, particularly the increase of extinction rates when management practices were ceased, emphasize the need for clearing the patches regularly in order to reduce the scrub density and to promote a continuous growth of herbs and grasses (Ferreira and Alves 2009). In summary, rabbit recovery, in the context of Iberian lynx conservation in Iberia, should consider the following recommendations:
  1. 1.

    The creation of a standardized monitoring program for predicting rabbit occupancy and patch colonization and extinction patterns in priority reintroduction areas

  2. 2.

    The application of annual habitat adaptive management actions, such as the creation of pasturelands and artificial warrens, followed by rigorous population assessment, in the boundaries of rabbit nuclei (not exceeding 500 m from the occupied nuclei). In this case, it is important to emphasize that management procedures are only effective in areas located in the edge or near rabbit nuclei since the rabbit’s low dispersal capacity makes habitat management actions ineffective outside these areas

  3. 3.

    The performance of translocations in areas where natural colonization could be extremely difficult (areas distancing 500–1,000 m from occupied patches) using rabbits from the local population can be a viable and alternative solution to captive breeding in semi-free enclosures (Rouco et al. 2008)

  4. 4.

    A constant revision of the management process based on monitoring results


Regarding disease control, and considering the currently widespread RHD (Calvete et al. 2006) and the persistence of high myxomatosis mortalities (García-Bocanegra et al. 2010), Cotilla et al. (2010) recommended a protocol to be included in the Spanish Wildlife Disease Surveillance Strategy for monitoring viral diseases based on the prevalence of antibodies and the abundance of rabbits. Their conclusion is that efforts should be directed to rabbit populations with low antibody prevalence against the disease, and recovery strategies should be based on experimented management measures that guarantee an increase in rabbit numbers so that the populations can overcome the suppressive impact of this disease (Cotilla et al. 2010). In fact, this protocol should be applied not only in Malcata, where a possible disease outbreak occurred in 2003 (Fig. 4), but also in other important areas for lynx reintroduction in order to fully understand the potential impacts of the disease in conservation efforts.

Besides this, it is fundamental to develop a standard methodology to monitor populations to achieving rabbit population recovery not only at a regional level but also at wider scales (Delibes-Mateos et al. 2009). Particularly in the Iberian Peninsula, it is crucial to establish a long-term monitoring program for rabbit abundance (Delibes-Mateos et al. 2009). Knowledge on the effect of factors associated with rabbit population trends (particularly predation, disease and habitat features) is critical to successfully achieve the recovery of wild rabbit populations. Ultimately, one of the most vital measures is to implement a workgroup gathering the scientific community, stakeholders, hunters, game managers, conservationists, governmental agencies and other sectors engaged in rabbit management. This group would be responsible for creating a global strategy for long-term conservation of rabbit populations (Ferreira and Delibes-Mateos 2011).


Thanks to Piran White, Pelayo Acevedo and Pedro Tarroso for helpful comments and to our colleagues who helped in fieldwork. We also thank I. Barrio and two anonymous referees for providing helpful comments on a previous draft of this manuscript. Part of this research was supported by LIFE project B4-32000/99/006423 “Recovery of habitat and prey of Iberian lynx in Serra da Malcata”.

Copyright information

© Springer-Verlag 2011