Dying like rabbits: general determinants of spatio-temporal variability in survival


  • Zulima Tablado,

    Corresponding author
    1. Departamento Biología de la Conservación, Estación Biológica de Doñana (CSIC), Avda Américo Vespucio s/n, 41092 Sevilla, Spain
    2. Université de Savoie, CNRS-UMR 5553 Laboratoire d’Ecologie Alpine, 73376 Le Bourget du Lac, France
    Search for more papers by this author
  • Eloy Revilla,

    1. Departamento Biología de la Conservación, Estación Biológica de Doñana (CSIC), Avda Américo Vespucio s/n, 41092 Sevilla, Spain
    Search for more papers by this author
  • Francisco Palomares

    1. Departamento Biología de la Conservación, Estación Biológica de Doñana (CSIC), Avda Américo Vespucio s/n, 41092 Sevilla, Spain
    Search for more papers by this author

Correspondence author. E-mail: zutal@ebd.csic.es


1. Identifying general patterns of how and why survival rates vary across space and time is necessary to truly understand population dynamics of a species. However, this is not an easy task given the complexity and interactions of processes involved, and the interpopulation differences in main survival determinants.

2. Here, using European rabbits (Oryctolagus cuniculus) as a model and information from local studies, we investigated whether we could make inferences about trends and drivers of survival of a species that are generalizable to large spatio-temporal scales. To do this, we first focused on overall survival and then examined cause-specific mortalities, mainly predation and diseases, which may lead to those patterns.

3. Our results show that within the large-scale variability in rabbit survival, there exist general patterns that are explained by the integration of factors previously known to be important at the local level (i.e. age, climate, diseases, predation or density dependence). We found that both inter- and intrastudy survival rates increased in magnitude and decreased in variability as rabbits grow old, although this tendency was less pronounced in populations with epidemic diseases. Some causes leading to these higher mortalities in young rabbits could be the stronger effect of rainfall at those ages, as well as, other death sources like malnutrition or infanticide.

4. Predation is also greater for newborns and juveniles, especially in population without diseases. Apart from the effect of diseases, predation patterns also depended on factors, such as, density, season, and type and density of predators. Finally, we observed that infectious diseases also showed general relationships with climate, breeding (i.e. new susceptible rabbits) and age, although the association type varied between myxomatosis and rabbit haemorrhagic disease.

5. In conclusion, large-scale patterns of spatio-temporal variability in rabbit survival emerge from the combination of different factors that interrelate both directly and through density dependence. This highlights the importance of performing more comprehensive studies to reveal combined effects and complex relationships that help us to better understand the mechanisms underlying population dynamics.


The population dynamics of a species is governed by demographic parameters, such as survival rates, which may widely change in time and space (Mazaris & Matsinos 2006; Ozgul et al. 2006). Processes, environmental and intrinsic properties, and interactions determining survival at a specific time and population may differ importantly from those of other areas and years. Thus, it is far from trivial to know to which extent local inferences about drivers of survival can be integrated and generalized over large spatio-temporal scales.

One way of attaining a comprehensive understanding of the key factors affecting species survival, while avoiding casual results, is considering as much variability as possible, and investigating simultaneously all processes and factors that can be involved. The importance of exploring this large-scale demographic variation is increasingly being acknowledged (Lester, Gaines & Kinlan 2007; Moles et al. 2007). Nevertheless, the complexity of mortality processes together with economic and time constrains lead authors to usually focus on single death causes and particular areas, considering greater levels of variability only, in the best cases, at the temporal scale (Ozgul et al. 2006; Baker & Thompson 2007).

In this study, we took European rabbit survival as model system and information published on specific populations as input data to investigate the existence of general patterns in the survival variability of a species and identify the combination of factors explaining that spatio-temporal variation. We first focused on exploring total survival patterns, and then, we examined the mortality causes that may lead to those survival rates.

At local level, rabbit survival has been studied in detail given the need for both controlling and preserving some of the populations (Fenner & Fantini 1999; Virgós, Cabezas-Díaz & Lozano 2007). Predation, two introduced infectious diseases [i.e. myxomatosis and rabbit haemorrhagic disease (RHD)] and to a lesser extent other causes of mortality (e.g. floods) have been found to importantly affect survival in this species (Villafuerte 1994; Calvete et al. 2002; Rodel et al. 2009). However, despite rabbits being one of the most widely distributed mammals in the world (Flux & Fullagar 1992; Thompson & King 1994), no studies had previously investigated variation in rabbit survival and mechanisms controlling it at large spatio-temporal scales.

We expected that global variability in rabbit survival would be explained by factors and processes already identified at the population scale (i.e. age, climate, diseases and predation among others). However, by integrating all information on local studies, we also hoped to disentangle some not always straightforward relationships between the drivers of rabbit morality. Given previous studies showing that at younger age classes predation rates are higher (Villafuerte 1994; Henning et al. 2008), RHD lethality is reduced and there exist maternal antibodies against myxomatosis and RHD (Ross 1972; Robinson et al. 2002), we predicted that the influence of infectious diseases on survival would be higher as age increased as opposed to predation. In addition, since diseases may weaken animals and make them more vulnerable to predation (Dunsmore, Williams & Price 1971; Villafuerte, Lazo & Moreno 1997), we also expected an increase in predation rates in populations where the diseases were present. Finally, we hoped to find general patterns relating climate and reproductive trends to the occurrence of disease outbreaks, as they influence vector abundance, contact rates and number of susceptible rabbits. We believe this type of research is essential to truly understand the determinants of population dynamics of a species and be able to anticipate and extrapolate population trends beyond study areas.

Materials and methods

We collected information on survival of wild rabbits from 91 publications, between 1953 and 2008, mainly from Spain, France, Germany, UK, Australia and New Zealand. We gathered data on survival, predation rates, rabbit consumption by predators, disease occurrence, disease mortality and other sources of mortality. To make survival, predation and other mortalities comparable across studies and ages, rates given or graphed by different authors for a period (L = length in days) were transformed into monthly (30·4 days) rates according to inline image, inline image and inline image. Only rates referring to a specific rabbit age period were used.

Regarding rabbit consumption by predators, we only considered data describing the percentage of biomass and frequency of occurrence of rabbits in each predator diet. As for disease occurrence, we looked for information on the presence/absence of outbreaks of myxomatosis or RHD in a given month, and we also gathered data on disease case-fatality (i.e. percentage of infected rabbits dying from the disease) from studies describing epidemics. Finally, related attributes like rabbit age, season, or year were also collected when possible. We also obtained information on climatic conditions in each study area and period from nearby meteorological stations. To accomplish our objectives, we took a two-step approach in which we first examined variability in total survival rates and afterwards performed specific analyses for each mortality cause.

Overall survival patterns

To investigate temporal and spatial variation in overall survival, we classified data in three age categories: newborn (<3 weeks, that is, before emergence from the warren/breeding stop), juvenile (from 3 weeks to 4 months, i.e. until sexual maturity) and adults (older than 4 months) and compared monthly survival variability [i.e. coefficient of variation (CV) and SD] of different age classes both across and within studies. Survival probabilities from the different studies were also used to perform pairwise correlations between age groups. Note that for the sake of comparison across age classes, survival of newborns was also expressed as a monthly rate, even though it applies only to the first 3 weeks of life.

Additionally, generalized linear mixed models (GLMM) were performed using monthly (30·4 days) survival rates of different studies (Appendix S1) as dependent variable to understand which factors are determining those survival variability patterns. We tested for the effect of rabbit age (in months; age_months), disease presence in the population, climate and the interactions between diseases and climate with age. Unfortunately, appropriate data on predation pressure were not available to be included in the model.

In cases where rabbit weight was provided instead of age, we used data from publications of the same or nearby populations to find equivalences between rabbit body mass and age. If rabbit age was expressed as an interval, we used the mean between the extreme values, and when the higher limit of the interval was not given (e.g. over 4 month of age), we set a maximum value of 33 months, which is a reasonable life expectancy as shown by Gibb & Fitzgerald (1998) and von Holst et al. (1999).

Regarding diseases, only the two diseases causing the highest mortalities in rabbit populations were considered, that is, myxomatosis and RHD. There were three categories of this variable (diseases) depending on the number of diseases having reached the population by the study period: No diseases, only myxomatosis, and both myxomatosis and RHD. The case of only RHD is found only in New Zealand, and not enough suitable data were obtained to consider this forth category. As high rainfall and low temperatures might be a source of mortality in some areas, we also included mean annual precipitation (total_p) and temperature (mean_t) as a measure of the climatic conditions in the area, and as rainfall may affect rabbits differently depending on their age (e.g. warren flooding), we also incorporated the interaction between total_p and age_months.

The low availability of rabbit density information in the reviewed studies prevented us from including also the effect of density dependence in the model. However, we were able to apply a second GLMM to a few data provided by Lombardi et al. (2003) and Parer (1977) to test for the association between adult survival rates and density (rabbit_dens). To make densities comparable across areas, we scaled them from 0 to 1 within each study ((density − densitymin)/(densitymax − densitymin)). We always used the distribution best fitting to the data, which in the two previous models was the beta distribution, and Location_in_study was incorporated, here and in all other GLMMs of this paper, as random factor to account for the autocorrelation among measures taken within sites (populations) in each given study.

Specific sources of mortality

Data concerning specific causes of mortality were also compiled and analysed to understand which variables where controlling its spatio-temporal variability. To investigate factors affecting predation upon rabbits, we applied a GLMM to predation rates available within 12 studies (Appendix S1). As a result of the lower sample size, we only tested for the effect of rabbit age (months) and its interaction with disease presence.

Two of these previous studies (Robson 1993; Moriarty, Saunders & Richardson 2000) provided data on relative densities of rabbits and predators along the year. For these studies, we carried out an additional model in which predation rate was regressed against rabbit_dens and relative predator density (predator_dens). For comparison’s sake, predator_dens was also standardized from 0 to 1 ((density − densitymin)/(densitymax − densitymin)). In both cases, the dependent variable, which best followed a beta-distribution, corresponded to predation rates, previously standardized to periods of 30·4 days.

To further explore patterns of predation on rabbits, we performed four GLMMs with information on rabbit consumption by predators (Appendix S1). In all four cases, the response variables were best fitted by a negative binomial distribution, and apart from Location_in_study, predator species was also included as random factor to account for autocorrelation.

First, we examined the impact of season and diseases on rabbit frequency in diet samples. Then, using percentages of rabbit biomass in predator’s diet, we tested for the effect of predator_type (generalist or specialist) and the interaction between the latter and season. We considered as specialist species only the Iberian lynx (Lynx pardinus) and Spanish Imperial Eagle (Aquila adalberti), which have been previously referred as ‘super specialists’ of rabbits (Ferrer & Negro 2004; Delibes-Mateos et al. 2008) and that are well known to depend on rabbits all year-round. Any other predators that may switch between different preys when available were considered as generalists, regardless of their local functional responses.

In the two last diet models, the response variable was the relative percentage of each rabbit age class in predator’s diet. This allowed to investigate differences in the rabbit ages preferred by generalists vs. specialists through the inclusion of the interaction between age_class (newborn, juvenile and adult) and predator_type as the independent term of one of the GLMMs. In the other model, we considered instead the interaction between season and age_class to evaluate whether the percentages of each age class consumed by predators changed throughout the year.

Another important source of mortality in rabbits is infectious diseases, that is, myxomatosis and RHD, which became endemic in the populations causing annual or biennial outbreaks (Fenner & Fantini 1999; Bruce, Twigg & Gray 2004). Thus, data from studies describing these epidemics in different areas and years (Appendix S1) were analysed to determine which variables control the spatio-temporal variability in disease occurrence and case-mortality in rabbit populations. First, presence/absence of disease activity in a given month, which was binomially distributed, was regressed against climatic variables and stage of the reproductive season. Data corresponding to first epidemic outbreaks were omitted as the absence of antibodies in the population could act as a confounding factor in disease phenology.

As climate may affect disease transmission (e.g. modifying vector abundance or virus persistence), we evaluated the linear and quadratic effect of mean annual precipitation (total_p), and monthly rainfall (precip) and temperature (temp), on the probability of observing a disease. Time since breeding season onset (breed_month) and its quadratic effect were also included, as reproduction drives the input of susceptible young entering a population, which is necessary for a disease to erupt. This variable was calculated using the model obtained by Tablado, Revilla & Palomares (2009). Months outside the breeding season were assigned the value 0.

After investigating what drives disease outbreaks, we examined the factors determining disease-induced mortality of rabbits infected during epidemics. For both diseases, the GLMM response variable was case-fatality, which best followed a beta distribution. As explanatory variables, we used rabbit age (age_months) and years since diseases reached the population (time; to account for disease resistance development). Sometimes, dates of myxomatosis or RHD arrival were obtained from publications of nearby areas. When laboratory rabbits were used, time was assigned the value 1 simulating a population first contact with the diseases. Because there are several myxoma virus strains with different a priori grades of virulence, we added this variable (Virulence) to the independent term of the model. Virulence of a given strain is determined using domestic rabbits without resistance and goes from 1 to 5 (see Fenner & Marshall 1957).

All models were fitted using sas proc glimmix (SAS Institute Inc., Cary, NC, USA), and their goodness of fit was evaluated through the ratio of the obtained generalized chi-square to the degrees of freedom. Values near one imply that data variability is properly described by the model (Fernandez et al. 2010). Variables tested in all cases were those that have been previously proved to be related, at least individually, to rabbit survival and mortality at local scales. Finally, variability in other mortality causes was briefly described. These deaths caused by, for example, floods or warren collapses have not received much attention in literature; however, they might also contribute to explain the differences in survival rates among rabbit populations.


Patterns and variability in total survival

Examining directly age-specific survival, rates were observed to increase in magnitude and decreased in variability with age, both spatially and temporally (Table 1). We also found, through pairwise correlations between age-specific rates, that newborn and juvenile survivals tended to be negatively associated contrary to adult and juvenile rates, whose correlation was positive (Table 1).

Table 1.   General determinants of rabbit survival. Sample size (N) is the number of age-specific survival rates available. Lower panel presents mean and variability of rabbit survival according to three age groups and pairwise correlations between survival rates at different ages
EffectsaParameter estimate(±SE) (N = 57) FP-value
  1. **P-value greater than standard level 0·05

  2. aRandom factor = Location_in_study, distribution = beta, link = logit.

  3. bIn Wood (1980) and Villafuerte (1994); newborns corresponded to the period 1–28 days, and in the site Cape Naturaliste of King and Wheeler 1981, data on juvenile survival start at 1 month of age.

  4. cBetween studies.

  5. dMean of the within study SD and CVs.

  6. ns, nonsignificant variable.

Intercept2·041 ± 0·834  
 Age_months0·142 ± 0·062 6·650·017
 Diseases ns
 Age_months*diseases  3·370·053**
  No diseases0  
  Myxo−0·037 ± 0·055  
  Myxo + rhd−0·108 ± 0·051  
 Total_p−0·002 ± 0·001 9·410·006
 Age_months*total_p0·0001 ± 0·00005 3·410·070**
 Mean_t ns
Age classMean monthly survivalSDCVCorrelation coefficient
Newborn (<3 weeks)b0·32 (0·05–0·61)0·19c0·59c 
Juvenile (3 weeks–4 months)b0·59 (0·33–0·78)0·16c0·26c−0·76 (P = 0·13; N = 5)
Adults (>4 months)0·89 (0·63–0·96)0·08c0·08c0·601 (P = 0·07; N = 10)

The positive effect of age on survival was confirmed by the GLMM, although it varied depending on the presence of diseases in the population (Table 1; Fig. 1a). This interaction between diseases and age_months showed, as predicted, that the presence of diseases in a population decreases survival much more in adults than in young rabbits, whose survival probabilities appear to remain similar to disease-free populations. In fact, disease presence per se did not reduce survival significantly. Rabbit age also interacted positively with the amount of rainfall in the area, although standard significance levels were not reached. That is, the overall negative effect of annual precipitation on survival decreases in importance in older rabbits (Table 1; Fig. 1b,c). Mean temperature, however, did not seem to be significantly affecting survival. With the small sample size (n = 6) obtained from using data from Parer (1977) and Lombardi et al. (2003), we were not able to find a significant effect of population density on adult survival.

Figure 1.

 Factors determining European rabbit survival. Interaction between age and diseases (a). Effect of precipitation on survival as age increases, in a population without diseases (b) and in a population with myxomatosis and rabbit haemorrhagic disease (c).

Causes of mortality

Rabbit age and its interaction with diseases also appeared to influence predation rates. Overall predation risk decreases as rabbits grow older; however, the trend is sharper in populations with only myxomatosis than in populations with both diseases (Table 2; Fig. 2a). That is, the presence of both infectious diseases increases predation probability more in juveniles and adults (especially under 12 months) than in newborn rabbits. With other models we also found that, within years, predation varies positively with rabbit and predator density (Table 2).

Table 2.   Variables controlling rabbit predation (between and within years) and predator diet (rabbit occurrence and biomass). ‘N’ (sample size) corresponds to either age-specific predation rates or diet data for each predator species and season
EffectsOverall predation ratea (N = 18)Predation rate (within year)b (N = 8)
Parameter estimate(±SE)FP-valueParameter estimate(±SE)FP-value
  1. a,bRandom factor = Location_in_study, distribution = beta, link = logit.

  2. c,dRandom factor = Location_in_study and predator species, distribution = negative binomial, link = log.

  3. ns, nonsignificant variable; RHD, rabbit haemorrhagic disease.

Intercept0·203 ± 0·377     
 Age_months−1·371 ± 0·33918·840·0019   
 Age_months*diseases 13·580·0050   
  No diseases     
  Myxo + rhd1·211 ± 0·329     
Intercept   −3·277 ± 1·455  
 Rabbit_dens   1·526 ± 0·5318·270·045
 Predator_dens   1·943 ± 0·49015·740·017
EffectsRabbit occurrence in predator dietc (N = 32)Rabbit biomass in predator dietd (N = 61)
 Parameter estimate(±SE)FP-valueParameter estimate(±SE)FP-value
Intercept3·188 ± 0·245     
 Season 3·380·0365   
  Spring0·517 ± 0·182     
  Summer0·239 ± 0·183     
  Autumn0·021 ± 0·198     
 Diseases 4·370·0252   
  No diseases0     
  Myxo0·409 ± 0·214     
  Myxo + RHD0·737 ± 0·251     
Intercept   4·480 ± 0·511  
 Predator_type   ns
 Predator_type*season   (Fig. 2b)3·340·0086
Figure 2.

 Determinants of rabbit predation risk and predator diet. Effect of rabbit age on predation probability in populations with different number of diseases present (a). Seasonal patterns in rabbit consumption by specialist and generalist predators (b). Relative consumption of each rabbit age class depending on predator type (c) and season (d).

Regarding rabbit occurrence in predator diet, we observed significant intra-annual variations depending on the season and higher rabbit frequency in diet samples when diseases were present in the population (Table 2). When using percentage of rabbit biomass in diet, although we could not prove the effect of predator type individually, we found a significant interaction between predator_type and season. Specialists consume similar amounts of rabbit biomass in every season, whereas generalist predators include much higher percentages of rabbit biomass in winter and spring than in autumn (Fig. 2b), being the latter the moment when rabbit populations usually reach the annual minimum.

The last two diet GLMMs showed how predator_ type and season affect the relative portion of each rabbit age class present in predator diet. Generalist species make a greater use of juveniles and newborn rabbits, while specialists prefer the older age classes (F = 3·45; d.f. = 44; P = 0·01; Fig. 2c). Throughout the year, juveniles were the most consumed rabbit age group; however, the relative percentage of newborns in predator diet increased during winter and spring, contrarily to adult rabbits whose relative importance was higher in summer and autumn (F = 8·94; d.f. = 81; P = <0·0001; Fig. 2d).

As for infectious diseases, myxomatosis activity in rabbit populations appeared to be associated with breeding month and annual precipitation, both linearly and quadratically, but not with monthly climate variables (Table 3). The probability of occurrence of this disease decreases during the first months of the breeding season but increases again as the season progresses and may still be relatively high after the end of the reproductive period (breed_month = 0) or in the first months of the next one (Fig. 3a). Myxomatosis outbreaks are also more frequent in locations with moderately high mean annual precipitations; however, their probability declines again in extremely wet areas (Fig. 3b).

Table 3.   Factors affecting disease phenology and lethality. Sample sizes (N) refers to the number of months available (with or without disease activity) for the phenological models and of age-specific disease lethalities in the case-fatality analyses
EffectsMyxomatosis phenologya (N = 556)Rabbit haemorrhagic disease phenologyb (N = 165)
Parameter estimate(±SE)FP-valueParameter estimate(±SE)FP-value
  1. **P-value greater than standard level 0·05.

  2. a,bRandom factor = Location_in_study, distribution = binomial, link = logit.

  3. d,eRandom factor = Location_in_study, distribution = beta, link = logit.

  4. ns, nonsignificant variable.

Intercept−3·724 ± 1·318  −6·282 ± 3·612  
Total_p0·015 ± 0·0059·250·0025ns
Total_p2−1E−5 ± 4·4E−67·150·0077ns
Tempns0·842 ± 0·3206·920·009
Temp2ns−0·025 ± 0·0098·160·005
Breed_month−0·385 ± 0·11211·870·00060·437 ± 0·242**3·270·073
Breed_month20·045 ± 0·01311·900·0006−0·049 ± 0·028**3·150·078
 Rabbit occurrence in predator dietc (N = 61)Rabbit biomass in predator dietd (N = 26)
EffectsParameter estimate(±SE)FP-valueParameter estimate(±SE)FP-value
Intercept3·535 ± 0·643  −2·286 ± 1·217  
 Age_monthsns1·988 ± 0·37627·96<0·0001
 Time−0·055 ± 0·01513·210·0009ns
 Virulence−0·875 ± 0·14636·07<0·0001   
Figure 3.

 Factors affecting disease phenology. Effect of breeding season on the probability of finding myxomatosis (a) and rabbit haemorrhagic disease (c) active in the population. Values of Breed_month imply the number of months since the beginning of the reproductive period, with 0 corresponding to months without reproduction. Influence of annual precipitation (b) and monthly temperature (d) on the probability of myxomatosis and rabbit haemorrhagic disease occurrence, respectively.

In contrast, the phenology of the RHD does not appear to be significantly related to annual or monthly precipitation but to monthly temperature, in both a linear and quadratic way. Therefore, it will be more probable to find RHD in months with milder temperatures and in those with extreme temperatures (Table 3; Fig. 3d). Although standard significance levels were not reached (Table 3), we found that RHD could be also affected by the breeding period but in an opposite way to myxomatosis. For RHD, the probability of encountering infected rabbits increases towards the middle of the reproductive season and not in the end or beginning of it (Fig. 3c).

Regarding case-fatality, mortality of sick rabbits during a myxomatosis outbreak will decrease as time since disease arrival to the population and strain virulence grade increase (Table 3), but we did not find a significant effect of rabbit age. In the case of RHD, however, lethality increases importantly with age of sick rabbits but does not seem to be influenced by time since the disease reached the population (Table 3).

Finally, there are other death sources that have been suggested to be, at least locally, important, such as flooding or warren collapse (Copson, Brothers & Skira 1981; Palomares 2003), infanticide (Rodel et al. 2008, 2009) and coccidiosis (Tyndale-Biscoe & Williams 1955; Mykytowycz 1962). These types of deaths may greatly vary among areas and affect mainly newborn, causing mortalities ranging from 7% to 86% (Appendix S2).


Our findings, integrating data from local studies, confirmed that there exist variability patterns in rabbit survival that are generalizable over large spatio-temporal scales and that can be explained by the combination of factors already identified as important individually. We observed, both within and among studies, that newborn and juvenile rabbits had not only lower but also more variable survival probabilities than adults. This is most likely due to young rabbits being more affected by environmental variability than adult ones, as shown by the interaction between age and rainfall. A similar pattern was observed, for large herbivores, by Gaillard, Festa-Bianchet & Yoccoz (1998) who suggested that higher survival variability in younger classes was due to their greater sensitivity to mortality factors, independently of whether they are density-dependent or stochastic, and of the taxa involved.

Pairwise correlations between age-specific survival rates, although not significant, probably a cause of the small sample sizes available, seem to show a positive association between juvenile and adult survival implying that both groups are usually affected by similar death sources (Reed & Slade 2006; Baker & Thompson 2007). Conversely, the negative apparent correlation between newborn and juvenile rates (i.e. low newborn survival leads to relatively high juvenile survival and vice versa) suggests that, regardless of the proximate mortality causes, younger ages are more sensitive to density dependence, as stated by Gaillard, Festa-Bianchet & Yoccoz (1998).

The age effect was corroborated by our GLMM, which showed that the positive influence of age on survival appears to have been attenuated with the arrival of myxomatosis and RHD to populations through a decrease in adult survival, while overall mortality of younger rabbit remains similar. This agrees with Henning et al. (2008) who found that predation was more important than diseases for young rabbits, as opposed to adults. Therefore, the well-known negative influence of these diseases on rabbit population trends (Angulo & Cooke 2002; Henning et al. 2008) may result from their impact on the oldest ages, both directly and possibly also through increased vulnerability to predation (Fig. 4).

Figure 4.

 Conceptual scheme of age-specific determinants of rabbit survival. Size of objects and arrows represents the effect strength. Direct impacts of climate on diseases are represented with dashed arrows, while continuous lines symbolize effects generalized to both diseases and predator types. Δd corresponds to intra-annual variations in rabbit density (i.e. breeding season stage), and Δp refers to intra-annual variations in predator pressure.

Our results also show that annual rainfall has a general negative influence on rabbit survival. Wet areas are detrimental for rabbits owing to warren flooding/collapse, hypothermia, and increased transmission of endoparasites such as coccidia (Bull 1958; Robson 1993; Palomares 2003; Rodel et al. 2004). These would, at least partially, explain why precipitation has a greater impact on younger age classes than on older ones. Improved pasture growth in regions with abundant rainfall may also lead to denser/less digestible vegetation, and thus, rabbit body condition and survival may worsen (Williams et al. 1995; Dekker 2007).

The positive relationship between myxomatosis and annual precipitation could also be partially responsible for this decrease in rabbit survival with increased rainfall (Fig. 4). This is also consistent with the lower effect of rainfall on older rabbits, which may have antibodies from previous myxomatosis epidemics. Although some authors have shown an association between low survival and cold temperatures (Marshall 1959; Rodel et al. 2004), we did not find a significant relationship in our study, probably owing to not having enough temperature variation.

We could not either demonstrate the influence of density on survival. This may be partly due to the small sample sizes available for some analyses, which reduce the power to detect effects, and also to the overriding effect of environmental variability over density (i.e. the latter can be more easily detected under similar environmental conditions). However, the apparent negative trends observed in Lombardi et al. (2003) and Parer (1977) agree with results of Rodel et al. (2004) who found a decrease in subadult survival as population density increased. Also, as seen later, density appeared to have direct effects on specific mortality causes, like predation or disease activity (Fig. 4).

When closely examining the different causes of rabbit mortality, we observed that the increase in survival with age could be partially driven by predation patterns, which are also significantly affected by age, with adults being less predated than younger rabbits. As predicted, disease presence was found to influence predation rates through its interaction with rabbit age. Unfortunately, we did not have enough predation data on populations without diseases, but we found that predation decreased more abruptly with age in populations with only myxomatosis than where both myxomatosis and RHD were present, probably because older animals that would usually escape predation become easy targets when weakened by the additional disease (Fig. 4). This effect is especially noticeable in adult rabbits under 12 months, since beyond that age animals are more likely to have been previously exposed and immunized against the disease (Fenner & Fantini 1999; Calvete et al. 2002).

This pattern of interaction could be, in part, responsible for the similar trends observed in overall survival. However, unfortunately, from the bibliography we cannot distinguish how much of this extra predation corresponds to rabbits eaten as carrion after dying from the disease. In a separate analysis, predation rates also appeared to be positively associated with seasonal increases in predator numbers and with rabbit density (Fig. 4). The latter agrees with previous studies indicating that predation is density-dependent (Erlinge et al. 1984; Sinclair & Pech 1996).

Food habit analyses showed that rabbits are consumed more often in spring and summer, which coincides with density peaks at the end of the breeding season in most regions (Myers et al. 1994; Tablado, Revilla & Palomares 2009). Higher disease activity during these seasons could also contribute to this pattern. Indeed, the greater probability of sick rabbits to be predated or eaten as carrion (Rogers, Arthur & Soriguer 1994; Cabezas 2005) is reflected by the increased rabbit frequency in diet samples of populations with myxomatosis and RHD (Fig. 4). With data on rabbit biomass in diet, we found that those seasonal variations are mainly caused by generalist predators as opposed to specialists that depend on rabbits and consume it in high amounts without important temporal variations. These findings agree with Calzada (2000) and Revilla & Palomares (2002), which examined diets of single predator species separately.

This coincides also with the type of functional response expected for this two predator categories. Specialist predators are usually associated with type II functional responses, where the increase in prey consumption with prey density is limited by a constant search rate, generating a simple convex curve (Holling 1959; Davey et al. 2006). Generalists, however, show a sigmoid response (type III) owing to increased search rate as prey density increases (Holling 1959; Turchin & Hanski 1997). The latter allows faster and greater functional responses, and could explain the strong variations of rabbit consumption across seasons, as could represent prey switching in generalist predators according to rabbit availability (Andersson & Erlinge 1977; Pech et al. 1992).

With further analyses, we also observed that rabbit age classes are differently preyed upon depending on the season and thus, on their relative availability. The proportion of juveniles in diet is high year-round, but newborns are consumed mainly in winter and spring, which generally corresponds to main breeding seasons (Myers et al. 1994; Tablado, Revilla & Palomares 2009). For the rest of the year, when newborn availability is lower, the relative importance of adults in diet increases. This coincides also with the decrease in rabbit consumption by generalists, who feed mainly on younger classes, and thus, specialist species may account for the relative increase in adult predation (Figs 2 and 4). Nonetheless, we should be careful when interpreting predator food habits. They help us to understand the system, but the actual impact of predation on rabbit populations cannot be derived from diet data without additional information on predator abundance, rabbit density and attack rates.

Apart from predation, other main sources of rabbit mortality are infectious diseases (Appendix S3), that is, myxomatosis and RHD (Fig. 4; Ross et al. 1989; Calvete 2006). These became enzootic lingering in, for example, remnant susceptible rabbits not immunized during epidemics or infected vectors (Chapple & Lewis 1965; Ross 1972) and breaking out annually or biennially owing to the input of new susceptible hosts during reproductive pulses (Altizer et al. 2006; Begon et al. 2009). Mortality due to myxomatosis or RHD depends on, first, the occurrence of an outbreak that infects rabbits, and secondly, the dying probability of those sick rabbits. Thus, we examined these two epidemic components (i.e. phenology and lethality), even though we acknowledge that this is not enough to capture all the epidemiological complexity concerning these two infectious diseases.

Our phenological analyses showed that, as expected, there exist general trends in disease occurrence determined by reproduction and climate, although the type of relationships differs between myxomatosis and RHD. The period in which myxomatosis is more active goes from the end of a breeding season to the beginning of the next one (i.e. from spring to autumn in most locations). These are months when high numbers of susceptible rabbits (Fig. 4) coincide with abundant insects, which are the main vector of myxomatosis transmission (Appendix S3), even though direct contact might also apply (Parer & Korn 1989; Merchant et al. 2003).

We could not demonstrate the influence of monthly temperature and rainfall on myxomatosis phenology, but we found a negative quadratic effect of annual precipitation (Fig. 4). Wet areas have higher abundance and activity of insect vectors, and thus, greater myxomatosis incidence than drier areas (Soriguer 1981; Parer & Korn 1989; Ross et al. 1989; Simón et al. 1998). However, in locations with extremely high precipitations, the occurrence of myxomatosis decreases again. Greater mortalities caused by warren flooding in these populations (Copson, Brothers & Skira 1981) may reduce the availability of susceptible rabbits necessary to sustain myxomatosis outbreaks.

As for RHD, neither monthly nor annual precipitation showed a significant effect on phenology. However, temperature seemed to affect RHD occurrence, which is more likely in milder months (Fig. 4). High temperatures decrease virus survival (e.g. in infected carcasses), and thus, RHD transmission and activity during hot months (Xu & Chen 1989; McColl et al. 2002). Additionally, extreme temperatures, both hot and cold, may prevent RHD outbreaks by reducing reproduction, and thus contact rates and also vector abundance. Despite the latter not being as important as direct transmission (Appendix S3), vectors might also help spreading the disease, mainly between locations (Asgari et al. 1998; Cooke & Berman 2000).

The probability of finding rabbits infected with RHD is also affected by the population reproductive stage, even though standard significance levels were not reached. However, contrarily to myxomatosis, it increases during the main breeding season (Fig. 4). This may be because of greater numbers of sick-to-susceptible contacts during reproduction (Cooke 1999; Calvete et al. 2002; Mutze et al. 2002).

Here, we also showed differences in case-fatality between diseases (Appendix S3). The high case-fatality of RHD seems to be modified by rabbit age. This effect was already showed by Robinson et al. (2002), who found that young rabbits under 4 weeks might get infected but usually do not die from RHD, this survival decreases importantly with age especially when maternal antibodies are not present, and by 3 months of age the disease kills around 90% of sick rabbits. Some authors have suggested that genetic resistance to RHD could develop in European rabbits (Cooke 1999; White et al. 2001); however, more time might be necessary, as we did not encounter any significant effect of time since disease arrival on resulting case-fatalities.

In contrast, myxomatosis lethality has considerably decreased since the disease introduction. Shortly after virus release, new strains with lower virulence but enhanced transmission started to establish in rabbit populations (Fenner 1983; Kerr & Best 1998), allowing for the selection of genetically resistant animals (Marshall & Douglas 1961; Fenner 1983), which in turn might have promoted the selection again of strains slightly more virulent (Fenner & Fantini 1999; Aparicio, Solari & Bonino 2006). As a result, strain attenuation together with genetic resistance has led to the relatively low myxomatosis case-fatalities in current populations (i.e. average 33%; Appendix S3).

Although several authors have shown an increase in survival to myxomatosis infection with age (Fenner & Ratcliffe 1965; Parer et al. 1994), we were not able to demonstrate it when pooling data across studies. This may be attributed to confounding factors increasing young survival that could not be controlled in this study (e.g. genetic resistance or sire effects). The latter is a mechanism, not well understood yet, which protects kittens born within 9 months of paternal infection with myxomatosis (Parer et al. 1995; Appendix S3). Maternal antibodies, however, could be ruled out because rabbits used in all survival experiments were previously tested for antibody presence and excluded if positive.

The low variability in age data could also explain the lack of significant correlation between age and myxomatosis case-fatality, as most studies used rabbits older than 3–4 months. It has also been suggested that Grade I strains could behave in an opposite way to the others (Sobey et al. 1970; Parer et al. 1994), thus contributing to obscure the overall influence of age. Unfortunately, available data did not allow to further examine this interaction.

Finally, rabbits may also die from other causes (e.g. malnutrition, flooding, collapse, coccidiosis or infanticide), which despite being anecdotally highlighted as important at local level (Copson, Brothers & Skira 1981; Palomares 2003; Rodel et al. 2008), are not generally considered relevant at larger scales. These death sources seem especially important for newborns (i.e. within warrens) and juveniles (Gibb & Fitzgerald 1998; Rodel et al. 2009), agreeing with the interaction between rainfall and rabbit age in our survival analyses (Fig. 4). Sometimes these mortality types may be underestimated through their assignment to predation or infectious diseases (e.g. dead rabbits eaten as carrion; Tyndale-Biscoe & Williams 1955; Webb 1993) or not detected globally because of the lack of adequate data. However, some studies like Rodel et al. (2009) emphasize their importance, as they found a loss of around 40% of newborns after eliminating myxomatosis, RHD and most predation from their enclosed population. This similarly low survival probabilities found in younger rabbits even without epidemic diseases or predation suggests that, to some extent, mortality causes might be compensatory at those ages.

There are other factors whose global effect we could not consider or demonstrate, but that may be still relevant at local scale. For example, hunting, type of insects transmitting diseases and ‘ecological interference’ between diseases (Rohani et al. 2003). However, this does not invalidate the generalized patterns found here, which conclude that large-scale variability in European rabbit survival results from the complex combination of different factors and processes (i.e. age, predation, myxomatosis, RHD and climate), which at the same time interact directly, and through density-dependent and compensatory effects. Further research should focus on the potential interactions and compensation between different mortality causes (e.g. climate and diseases), and on the derived population’s reproductive response. Only then, will we be able to understand current and future population trends of a given species.


Climate and day length information were obtained from KNMI (Royal Netherlands Meteorological institute) and US Naval Observatory websites, respectively. This research was supported by projects CGL2004-00346 and CGL2009-07301 from the Spanish Ministerio de Ciencia e Innovación. ZT was also partially supported by a FPU grant from the Ministerio de Educación.