The presented simulation study provides understanding of how vaccination may impact the spread and maintenance of an infection in wild boar populations under different viral constraints . The course of a potential outbreak was represented by an established model of Classical Swine Fever (CSF) in wild boar populations . Therein, alternative spatial application schemes of oral mass vaccination were implemented. The comparative evaluation of alternatives was focused on two output quantities: (1) the extent of the outbreak was measured by the maximum distance from the release point, and (2) the risk of endemicity was measured by the probability of virus circulation after 10 years. The first measure is essential for planning the extent of restriction zones for the pig production sector according to existing disease legislation (EU Council Directive 80/217/EEC ). The second measure prescribes whether spatially restricted host populations can maintain the infection and, hence, translates into the necessity of disease control within an infected population [29, 41, 44, 72–74]. Indeed, following the epidemic phase of an outbreak, either recurrent outbreaks have to occur inside parts of the area that already had been affected ("endemic phase" according to ), or the infection will fade out after spreading trough the population. The first defines endemicity as used in our analysis. The latter, although leading only to a transient infection of the host population, often also is referred to as "long" persistence because depending on the size of the connected and populated wild boar area, the epidemic phase of an outbreak might take a long time till fade out. For example, with a simulation area of 200 km × 50 km and spatial spread calibrated to 32 km per annum  an infection without potential to become endemic might still be present for 3-6 years depending on the place of introduction. Therefore, instead of referring to infections "persisting for a long time" in a population, we used "probability of endemicity" to describe possible maintenance of the infection after it had spread through the population.
Effectiveness of population vaccination is paralleled to the reduction in the number of secondary cases produced by e.g. an infected wild boar group [75, 76]. The number depends on the probability of transmission and the average infectious period of infected individuals . Under different viral constraints the latter might be highly variable between individual cases or different virus strains. For example, different experimental and field studies of CSFV exhibited a variety of outcome with regards to the lethality of individual infections (represented in the model by the parameter M) and the variability in the survival time of lethally infected hosts (from acute to rather chronic infections; represented by the parameter μ in the model) . To cope with this uncertainty, the simulation results were produced by systematically sampling over the possible values of the two parameters that define individual disease courses. Simulations in the same wild boar population but assuming different viral constraints (Figure 6a2), subjected the full range from fade out (e.g. highly virulent strain CSFV/1.1/dp/CSF0382/XXXX/Koslov ) to endemicity (e.g. moderately virulent strain CSFV/2.3/wb/CSF1045/2009/Roesrath ) to the same vaccination schemes enabling most general comparison of effectiveness. Moreover, the results of the comparative evaluation were tested for their sensitivity to qualitative alteration of transmission probability (see Additional file 2) but were regained in full.
Complete area vaccination and the maximum outcome
Compared to the simulations without vaccination (Figure 6a) the performance of vaccination was overviewed by the reduction of Pend and Dmax for the different viral constraints (Figure 6c1 + 2). The results showed that maximum reduction of spatial spread of the infection was achieved for viral constraints that resulted in an effective average infectious period Tinf between 2 and 6 weeks (Figure 2a). Thereof maximum reduction in the probability of endemicity was achieved only for moderate case mortalities (M < 60%) combined with sufficiently long infectious periods of lethally infected hosts (μ > 3 weeks). Interestingly recent outbreaks are characterised by moderate lethality of less than 50% and few acute courses i.e. few of the lethally infected animals dying within less than four weeks [41, 44, 58, 62]. Thus, with the recent CSFV strains in wild boar (e.g. moderately virulent strain CSFV/2.3/wb/CSF1045/2009/Roesrath ) the treatment of the total wild boar population - although impractical - is expected to prevent spatial spread and the endemicity of the infection.
If high case mortality and long mean infectious periods of lethally infected animals were assumed, vaccination had no effect on the spatial spread but induced a prolongation of virus circulation. This is reasonable as without vaccination the decline of the population density behind the epidemic front due to high lethality already favours fade-out although spatial spread covered the whole area (Figure 6a1 + 2). Indeed, the initial average density of the simulated population was 5 hosts per km2, lethality of 80% left an approximate population density of 1 host per km2 behind the epidemic front. Guberti et al.  estimated the threshold for CSFV spread as 1 host per km2 making self-eradication a consistent outcome. Vaccination now prevented the collapse of the host population. As a consequence, the infection became endemic when the mean life expectancy of lethally infected hosts was set sufficiently long (i.e. μ above 4 weeks).
With CSFV such highly lethal infections are expected to coincide with rather short mean life expectancies of infected hosts (e.g. highly virulent strain CSFV/1.1/dp/CSF0382/XXXX/Koslov ). Therefore, such a negative impact of vaccination is less likely with CSFV control. More general, however, our results suggest that vaccination planning should be performed with caution if an outbreak is reported to kill the majority of infected animals: If the mean infectious period of the disease is very short, the outbreak is expected to be self-limiting in the population; if not, vaccination might even create endemicity.
Alternative baiting strategies - feasible expectations
Astonishingly, the buffered vaccination approach backed up the efficacy of complete area vaccination with regard to both criteria "prevention of disease spread" (Figure 7a1 and Figure 6c1) and "eradication" (Figure 7a2 and Figure 6c2). Baiting one year ahead of disease spread sufficiently mimicked the large proactive component of complete area vaccination. The backward component of the buffered strategy shows parallels with recent control proposals  that recommend repeated vaccination for at least one year after the last case detection in a local area.
The key assets in the strategy are the temporal raise of population immunity and the spreading distance of CSFV during one year:
Considering the devolution of population immunity by number of campaigns in a susceptible population (Figure 4), the assumed bait uptake led to saturation of population immunity after about 3 campaigns (i.e. 1 year of regular vaccination schedule). This dynamics is comparable to the dynamics observed in areas where oral vaccination was practiced before the reporting of CSF infections .
In the simulation model, the known spreading velocity value of 8 km per annual quarter  was scaled to transmission between a sufficient number of wild boar groups. If, however, the spreading distance of 32 km per annum is less general, e.g. an infection spreads differently fast in other eco-regions, then the results of our simulations are robust and will be repeated if the width of the buffer is aligned to the alternatively reported distance value. The success of this strategy might be favoured by the regular boar habitat structure in the model. Whether alteration of spreading velocity in structured landscapes of boar habitat requires adjustment of buffer width needs further detailed examination [79–81].
Baiting of areas that were ever infected during the outbreak is comparable to recent baiting strategies of successive extension of vaccination zones in accordance with disease spread. The strategy has a strong backward component but no proactive component. Hence, it is reasonable that this strategy cannot decrease spatial disease spread. With continued, uncontrolled spread of the infection through the wild boar area the strategy successively converged to the complete area design along with an increasing proportion of the landscape affected. Virus eradication can thus be expected to take place in late-stage disease control, while annual baiting costs increase with ongoing disease presence. More relevant, however, is the lacking potential to prevent disease spread i.e. the growth of the infected area.
Baiting of actually infected habitat areas (Figure 7c) was revealed to be completely ineffective. This appears reasonable as both the proactive and the backward component of the complete area baiting are no longer realised. Although the approach is often taken to be most cost effective , this judgement usually ignores total effort, which accumulates quickly if no eradication can be achieved.
At first glance, these findings contradict vaccination successes reported from the field with ever infected area vaccination . However, in this particular area the vaccination protocol changed during the control program  because the unavailable marker property of the applied vaccine limited the follow up investigation. Vaccination started with a wide buffer around the infected area (compare Figures 26a and 27a in von Rüden ), but later protocols foresee only the newly infected area to be added to the baiting area (ever infected area vaccination scenario) which did not actually happen. Following our findings, the positive effect of the "buffer" included in the initial campaigns already might have caused the observed success.
The concept of vaccinating the infected area or an additional buffer around it obviously depends on the monitoring of the infected area . If monitoring relies on serological investigations (e.g. CSF), oral mass vaccination will hamper the applicability of the strategies if no marker vaccines are available for oral application which allow differentiation of vaccinated from infected animals. For vaccination against CSFV infections in wild boar efforts are on the way to substitute the recent vaccine by an orally applicable marker version with the same protective characteristics. The vaccine already was tested experimentally in the field and thus motivated the systematic evaluation of possible new spatial vaccination schemes that make use of the dynamic actual infected area (see, e.g. [51, 84, 85]).