Modeling the response of an endangered rabbit population to RHDV2 and vaccination

Rabbit hemorrhagic disease virus 2 (RHDV2), recently detected in the western United States, has the potential to cause mass mortality events in wild rabbit and hare populations. Currently, few management strategies exist other than vaccination. We developed a spatially explicit model of RHDV2 for a population of riparian brush rabbits (Sylvilagus bachmani riparius), a subspecies of brush rabbit classified as endangered in the United States, on a subsection of the San Joaquin River National Wildlife Refuge. The goal of our model was to provide guidance regarding vaccination strategies for an endangered rabbit species. Our model predicts that increased interactions between rabbits (a proxy for landscape connectivity) and disease transmission rates among susceptible hosts (individual brush rabbits and conspecifics) have the greatest influence on the outcome of a potential vaccination campaign. Our model projects that across a range of parameter estimates (given an RHDV2 incursion), the median estimated population size with a 0%–10% vaccination rate after 1 year is 538 rabbits (95% Confidence Interval [C.I.] 69–1235), approximately 36% of the expected size of the study population of 1470 rabbits without an RHDV2 introduction. With a 10%–20%, 20%–30%, or 30%–40% vaccination rate, the median estimated population size increased to 628 rabbits (95% C.I. 130–1298), 723 rabbits (95% C.I. 198–1317), and 774 rabbits (95% C.I. 228–1410), respectively. These estimates represent 43%, 49%, and 53% of the expected population size without an RHDV2 introduction. Overall, a 1% increase in vaccination rate was associated with a six rabbit (95% C.I. 5–7) increase in total remaining population size. This result is dependent on assumptions regarding environmental transmission, home range size (and contact rates of rabbits). Given the relatively short lifespan of rabbits and the potential need for boosters, vaccination programs are most likely to be successful for small target populations where relatively high vaccination rates can be maintained.


| INTRODUCTION
Modeling host-pathogen dynamics in wildlife species is particularly challenging (Huyvaert et al., 2018).Many pathogens have multiple transmission pathways, abundance, and distributions of wildlife species are often poorly documented, and the true frequency of disease events may be unreported due to low detection probabilities.Rarely is treatment of all individuals feasible, and pro-active management, although often the most effective, can be difficult to achieve (Grider et al., 2022).However, developing a model that captures the essential dynamics of a disease can be useful for providing guidance to decision makers.Host-pathogen models that include both population dynamics of the host species as well as disease transmission pathways can be useful for disease management.These types of models can illuminate key uncertainties in transmission dynamics (Russell et al., 2015(Russell et al., , 2021)), including uncertainty in the effects of potential mitigation strategies (Bernard & Grant, 2019), and provide a transparent mechanism for evaluating trade-offs of different management decisions (Grant et al., 2017), particularly for emerging diseases.
Rabbit hemorrhagic disease is a frequently fatal and highly contagious viral hepatitis disease of domestic and wild lagomorphs caused by several closely related pathogenic lagoviruses (Family Caliciviridae) (Abrantes et al., 2012).Rabbit hemorrhagic disease virus (RHDV) was first described in China during the 1980s in imported domestic European rabbits (Oryctolagus cuniculus) from Europe.RHDV causes high mortality rates in rabbits but has not been documented to affect hares or other mammalian species (Abrantes et al., 2012).RHDV rapidly spread and became endemic in many wild European rabbit populations and was associated with negative population impacts for not only the rabbit but also rabbit predators (Abrantes et al., 2012;Delibes-Mateos et al., 2014;Monterroso et al., 2016).In 2010, a new variant rabbit hemorrhagic disease virus 2 (RHDV2) was first detected in France (Le Gall-Reculé et al., 2013), where it spread rapidly in farmed and wild rabbits throughout Europe.
RHDV2 is a Lagovirus in the family Caliciviridae that includes European brown hare syndrome virus (EBHV) and RHDV both of which can result in severe disease in rabbits.Like RHDV, RHDV2 causes high mortality rates in European rabbits.In contrast, other Lagoviruses such as EBHV do not affect European rabbits but do cause disease in hares (Lepus spp.; Lavazza et al., 1996).Currently, RHDV2 is an emerging disease in the United States.In March of 2020, RHDV2 was detected in the United States in wild populations of desert cottontail (Sylvilagus audubonii) and black-tailed jackrabbit (Lepus californicus) (Asin et al., 2021).Within 10 months of initial detection, RHDV2 mortalities were confirmed in five wild North American lagomorph species, including black-tailed jackrabbit, antelope jackrabbits (Lepus alleni), desert cottontail, eastern cottontail (Sylvilagus floridanus), and mountain cottontail (Sylvilagus nuttallii) throughout the southwestern United States, multiple other U.S. States, and Mexico (Clifford & Moriarty, 2022).RHDV2 causes similar clinical disease as RHDV; however, RHDV2 infects a wider range of lagomorph species and shows increased mortality among juveniles (Rouco et al., 2018).Similar to RHDV, a cyclical seasonal pattern in mortality rates has been noted in European rabbits for RHDV2, but in Australia, where RHDV was introduced in an effort to control the non-native European rabbit, cyclical patterns appeared to vary between the two viruses (Taggart et al., 2021).Information is currently lacking regarding the ability of animals to survive and recover from RHDV2 infections (and whether those animals continue to shed virus), and the potential contribution of survivors to the disease transmission cycle is unknown.Previous studies have detected seropositive European rabbits, demonstrating that infected domestic rabbits may survive (Rouco et al., 2018;Taggart et al., 2021).In an experimental setting, Eastern cottontails were susceptible to RHDV2 and can shed viral DNA, indicating that this widespread species may play a role in transmission dynamics of the disease in North America (Mohamed et al., 2022).This is concerning for conservation efforts for lagomorph species as well as domestic rabbits, as eastern cottontails and other cottontail species are found across a large geographic range and may be sympatric with more vulnerable lagomorph populations.
Few mitigation measures exist for RHDV other than vaccination.Rabbits vaccinated against classic RHDV have been documented with RHDV2 infections that resulted in death, thus vaccination against RHDV does not offer cross protection against RHDV2 (Dalton et al., 2012;Le Gall-Reculé et al., 2013;Le Minor et al., 2019).Recently, vaccines specifically developed for RHDV2 have been tested and were shown to be effective in domestic rabbits (Oryctolagus cuniculus; Le Minor et al., 2019;Müller et al., 2019).However, Hänske et al. (2021) documented four cases of RHDV2 deaths of domestic rabbits vaccinated with Filavac VHD K C + V (ECUPHAR).The effectiveness of the vaccine in wild rabbit species and wild environments is unknown.Overall, RHDV2 has the potential to have substantial effects on wild rabbit and hare populations due to its ability to overcome immunity imparted by other RHDV variants and the classic RHDV vaccine (Peacock et al., 2017), increased virulence (Capucci et al., 2017), ability to cause severe mortality in young rabbits (Hall et al., 2015), ability to spread rapidly (Mahar et al., 2018), and wide host range including both rabbits and hares (Camarda et al., 2014).There is some debate regarding long-term consequences of RHDV2 to rabbit populations.Larger populations may recover more quickly than smaller populations (Guerrero-Casado et al., 2016); however, higher rabbit abundance after disease emergence may lead to an increased population of susceptible hosts and potentially longer-term effects if the disease persists in the system (Calvete, 2006).Generally, RHDV2 causes non-specific symptoms such as lethargy and inappetence (Hänske et al., 2021), and results in death within 70 h (Capucci et al., 2017).In clinical trials rabbits developed a fever, and multiple lesions in the liver, spleen, and bone marrow, were evident upon necropsy (Neimanis et al., 2018).Once RHDV2 has entered the population, infected individuals shed virus infecting the environment, and carcasses of rabbits that succumbed to RHDV2 further contribute to the viral load on the landscape (Taggart et al., 2021).Wild European rabbit population declines of $60% to 80% were attributed to the spread of RHDV2 in Australia (Mutze et al., 2018;Ramsey et al., 2020) and Spain (Guerrero-Casado et al., 2016).However, some of these population declines may have been short-term (Ramsey et al., 2020), and these populations may have different susceptibility than North American wild lagomorph species.Habitat quality may also influence the response of populations to disease, with poor quality habitat leading to larger effects, at least in the short-term or immediately after an outbreak (Calvete, 2006;Guerrero-Casado et al., 2016).Regardless, lagomorphs are an integral component of trophic webs, whether mediating top-down (Kuijper & Bakker, 2005;Van der Wal et al., 2000;Zeevalking & Fresco, 1977) or bottom-up (Guerrero-Casado et al., 2016;Monterroso et al., 2016) dynamics, and their potential decline from disease is likely to cause cascading trophic effects throughout ecosystems.
Emerging diseases have the potential to severely reduce wildlife populations.Although it is too early to know the long-term effects from RHDV2 emergence in North America, its ability to cause substantial population reductions is particularly concerning for small populations occupying a limited geographic range (such as the riparian brush rabbit, Sylvilagus bachmani riparius).The riparian brush rabbit is listed as endangered under both the United States and California Endangered Species Acts (U.S. Fish and Wildlife Service, 2000).The subspecies occupies riparian-oak forest and successional shrub habitat along riparian corridors in the northern San Joaquin Valley of California, a habitat type that has been reduced to less than 1% of its historical extent (Kelly, 2018).Down to just a few hundred individuals, a captive propagation program was implemented for this species (Williams et al., 2008), and between 2002 and 2013 approximately 1500 rabbits were introduced into the San Joaquin River National Wildlife Refuge (hereafter the refuge; California Department of Fish and Wildlife, 2020;Kelly, 2018).This introduced population, located primarily on the refuge and adjacent conserved lands, is now considered the only remaining population that currently exhibits the ability to withstand demographic and environmental stochasticity and remain viable (U.S. Fish and Wildlife Service, 2020).Conservatively, the 852 ha of restored riparian habitat on the refuge supports a population of approximately 1500 rabbits (Landers et al., 2021).However, this area represents the portion of the total refuge for which population estimates exist.Additional small populations of riparian brush rabbits and native riparian habitat occur outside of these areas; however, the abundance in these areas is unknown but estimated as a few hundred individuals (Figure 1; California Department of Fish and Wildlife, 2020).In addition to the threat of disease, this population is threatened by flooding, wildfire, drought, predation, and anthropogenic changes to hydrology (U.S. Fish and Wildlife Service, 2020).
Due to the small population size of the riparian brush rabbit and extremely limited geographic range, RHDV2 represents a substantial threat to the continued persistence of this species.In May 2020, RHDV2 was confirmed in black-tailed jack rabbits in southeastern California.Cases in jackrabbits and cottontail species rapidly spread throughout the southern portion of the State, such that within 2 months of emergence, known mortalities were occurring within 500 km of the riparian brush rabbit range.In response, an emergency plan was designed to vaccinate riparian brush rabbits against RHDV2.The goal of the vaccination program was to reduce diseaseinduced extinction risk in advance of the anticipated arrival of RHDV2 to the refuge population.Subsequently the disease has spread to several counties in California, and the first riparian brush rabbit diagnosed with RHDV2 was reported in May 2022.At the time RHDV2 first started spreading among wild rabbit populations, no domestically produced, licensed vaccine was available in the United States for RHDV2.Filavac VHD K C + V ® vaccine (Filavie, France), developed for the active immunization of domestic rabbits against RHDV and RHDV2 was imported for the riparian brush rabbit vaccination effort.Importation of the RHDV2 vaccine was costly, and therefore, efficient planning was needed to utilize labor and financial resources appropriately.
We developed a spatially explicit model of the RHDV2 riparian brush rabbit system to evaluate the effectiveness of vaccination for preventing precipitous declines in a population of riparian brush rabbits threated by RHDV2.We used this model-based framework to help guide wildlife disease management decisions such as these.Additionally, we used this model to evaluate the effects of uncertainty in transmission dynamics on the model output, and determined which variables were most important for contributing to the number of rabbits remaining at the end of the simulation.Our model assisted management agencies in determining whether a vaccination effort (based on plausible levels of effort) would be effective in managing the risk of RHDV2 to the riparian brush rabbit population while balancing the effort and costs of vaccinating a wild population.

| METHODS
We developed a spatially explicit agent-based model of host-pathogen dynamics in R (R Core Development Team, 2022; see Supporting Information for model code; code was developed by the U.S. Fish and Wildlife Service).The area modeled (852 ha) included portions of high-quality restored habitat south of the San Joaquin River on the refuge and excluded some portions of the refuge (Figure 1).We incorporated information on lifehistory, including home range size, age at reproduction, survival, and number of kits expected to survive to adulthood as parameters in the model to describe the dynamics of riparian brush rabbits (Table 1).To begin the simulations, 1470 adult riparian brush rabbits, a population estimate for the modeled area from October 2020 (Landers et al., 2021), were distributed randomly across the 852 ha.Each rabbit was assigned a home range selected from the spatial locations of all available rabbit habitat on the portion of the refuge included in the model.The age distribution of the rabbits was designed to reflect natural mortality rates (Kelly & Holt, 2011).We generated 100 parameter sets using a Latin hypercube design and drew parameter estimates from our specified distributions for each parameter (Tables 1 and 2).We first calibrated the model for no vaccination and no RHDV2 incursion to verify that we could achieve a stable population size.We then ran 10 simulations with each of the 100 parameter draws for 1000 simulations total across a range of vaccination rates (0%-40%).Thirty percent vaccination was the maximum feasible vaccination rate estimate by refuge staff based on costs of both the vaccine and the personnel time required to capture and vaccinate rabbits.
Each model iteration simulated the population dynamics of the riparian brush rabbit population (births, natural mortality, and juvenile dispersal) over the course of 1 year.We introduced RHDV2 into this system on day one of the simulation (i.e., two rabbits were randomly selected for infection).Direct transmission of RHDV2 occurred between rabbits (direct contact) and environmental transmission occurred between rabbits and the home ranges of infected rabbits (simulating infection from contact with contaminated environments).Infected rabbits can die or recover, and recovered individuals can continue to shed virus for 30 to 100 days, and viable virus can remain in a carcass for 90 days (Dixon & Krauer, 2022).However, it is currently unknown how long the virus may remain viable in the environment (and this is likely dependent on environmental conditions); therefore, in our models we explore a range of values from 60 to 120 days.Vaccinated rabbits were present as the simulations began (i.e., prior to introduction of the disease), and a second vaccination occurred 180 days later.This second vaccination is applied randomly to the remaining population (i.e., there is no assumption that previously vaccinated rabbits are targeted for a second round of vaccinations).

| Population dynamics (reproduction, natural mortality, and dispersal)
Simulated population dynamics of the rabbits included reproduction for mature females, age-specific natural mortality rates, and juvenile dispersal.Simulations started at the beginning of the breeding season (200 days) and ran for 365 days.We began the simulations with the expected age distribution based on the mortality rates.With a mortality rate of 0.001/day for rabbits 80 days old (independent from females) to 2 years old and a mortality rate of 0.005/day for rabbits older than two, the expected age distribution is approximately 55% one-year-olds, 39% two-year-olds, and 6% three-year-olds.On each day of the simulation, births, juvenile dispersal, and natural mortality occurred while each rabbit aged by 1 day.Births occurred throughout the breeding season (rather than in a single pulse), and uninfected adult females not caring for young produced litters at a rate of r, the daily probability of producing a litter, during the breeding season.In other words, we conducted a binomial flip for each reproductive female that does not have a litter with a success rate of r to simulate births (a litter) on a daily rate, based on the annual reproductive potential of riparian brush rabbits (Table 2).If a reproductive female (age 1 year or older) gave birth, the number of juveniles per litter that survived to adulthood was sampled from a uniform distribution of 1, 2, or 3 (Table 1).We did not explicitly model infection in juvenile rabbits less than 80 days old.We made the assumption that the adult female was the primary transmitter of disease to dependent young.Juveniles remained with the adult female until they were 80 days old, and if the female became infected and died from RHDV2, all juveniles <80 days old also died.Juveniles of recovered adult rabbits that were immune to RHDV2 survived unless the female died from natural mortality.Juvenile dispersal took place at 80 days old.Juveniles dispersed within a distance that was twice the home range radius of the adult rabbits (Table 2).In other words, a new home range was established randomly for the juvenile within the defined dispersal range from their mother.New home ranges were restricted to the 852 ha of the spatially explicit model area.Once the juvenile dispersed, it could be infected through direct contact with other individuals or the environment in the same fashion as an adult rabbit.Rabbits died of natural mortality at age-specific rates (Table 1).

| Host-to-host transmission of RHDV2
All unvaccinated rabbits began in the susceptible state, and two infected rabbits were introduced to the population on Day 1.Thereafter, two infected rabbits were introduced every 30 days (we discuss this assumption below) and placed randomly in the environment.Host-to-host transmission rates were based on estimated rates of contact between hosts, which were influenced by the size of the home range and the number of rabbits in the population.
Seasonality of RHDV2 is currently unknown.Observations of RHDV2 outbreaks in wild rabbits in the Southwest United States were bimodal with peaks in spring and fall and declines in summer and winter (Figure 2; U.S. Geological Survey, 2022).It is not known if patterns in observed mortalities are due to rabbit behavioral changes over the course of a season, or changes in the virulence of the virus due to environmental factors.Therefore, we modeled the probability of infection from a contact with an infected host (P i ) as a seasonally changing variable with a unique P i for each day (Figure 2).
Probabilities of host contact are based on the amount of overlap of neighboring home ranges where P c is probability of host contact and P i is probability of infection from an infected host given a contact (Russell et al., 2021).Specifically, the probability of an infected rabbit A transmitting the disease to another rabbit B is P c Â P i , where P c is the probability of contact between neighboring rabbits (between A and B) and P i is probability of infection from an infected host given a contact.In this model, P c was estimated as the proportion of the combined host ranges that overlap.Essentially the larger the home range of the rabbit, the greater the area covered, and the greater the probability of contact with an adjacent individual.Overlap was calculated by assuming each rabbit has a roughly circular home range with a radius HR and a center point x c ,y c .We calculated the distance between each individual rabbit on the landscape and estimated the amount of overlap based on the size of the home range and the distance between the center points (Weisstein, 2022).The probability of rabbit A interacting with rabbit B was the area of overlap within the home range of rabbit A divided by the total area of the home range (note the probability of rabbit A interacting with rabbit B was equal to the probability of rabbit B interacting with rabbit A because the home ranges in any given scenario were the same size).

| Nontarget species transmission of RHDV2
Cottontail rabbit species may also be involved in the spread of RHDV2 (Mohamed et al., 2022).Therefore, we estimated the contribution of this sympatric species to overall probability of infection.Camera trap studies reported an estimated 0.5 desert cottontail per riparian brush rabbit home range (Tarcha, 2020).Therefore, we included a probability of infection from these nontarget rabbits by increasing the number of rabbits in each home range from 1.0 to 1.5.Once a host rabbit was infected, we assumed that any nontarget rabbit species within the home range was also infected.Both individuals (riparian brush rabbit and desert cottontail) could infect rabbits in overlapping home ranges, leading to an overall probability of infection from an infected rabbit in each overlapping home range of Equation (1) 1 À (1 À P i ) 1.5 Â P c .Post-infection, infected rabbits died at a rate of m i (Table 2) or recovered at a rate of r i (Table 2).Recovered rabbits remained infectious for up to 120 days (Table 2).

| Environmental transmission
RHDV2 may also be transmitted through contaminated soil.Therefore, we assumed that once a rabbit died from RHDV2, other rabbits could become infected with the virus upon contact with the home range (Table 2).When an infected rabbit died, its home range was infectious for n E days (Table 2).

| Evaluation of the relative influence of variables
Boosted regression trees used algorithm-based methods (regression trees and boosting) to develop predictive models that could automatically account for nonlinear relationships and correlated variables.These methods are useful for providing information regarding the relative importance of each variable on the predicted outcome.We evaluated the relative influence of home range size, probability of environmental transmission, probability of recovery, vaccination rate, infection length, and probability of mortality from RHDV2 infection using generalized boosted regression trees (Elith et al., 2008;Greenwell et al., 2020) in R (R Core Development Team, 2022).We evaluated the number of rabbits remaining at the end of the simulations to explore which parameters were most influential on the outcome of our model simulations.We increased the number of trees until we achieved stable results (i.e., the relative influence of the variables changed less than 1% every time the generalized boosted regression tree was conducted).We used cross validation with 10 subsets to identify the best fitting model and compared the predicted fits from the boosted regression to the observed data using a generalized linear model to calculate an R 2 .

| RESULTS
Overall, simulated vaccination resulted in larger population sizes versus no or low vaccination scenarios (Figure 3).Our model projects that across a range of parameter estimates, a 20%-30% vaccination rate results in a median estimated population size of 723 rabbits (95% Confidence Interval [C.I.] 198-1317) which is 1.4 times higher than the median estimate population size with a 0%-10% vaccination rate (538 rabbits [95% C.I. 69-1235]) (Figure 3).A vaccination strategy of 30%-40% resulted in a median population size of 775 rabbits (95% C.I. 228-1410), approximately 1.1 times higher than a 20%-30% vaccination strategy.A 0%-10% vaccination scenario resulted in a median 64% decline over all scenarios, whereas 20%-30% vaccination resulted in a median 51% decline in the population size, and a 30%-40% vaccination resulted in a median 48% decline.However, a wide range of population sizes was achieved depending on values of other parameters in the model.
We evaluated the relative influence of variables included in the model on our response variable (population size of adult rabbits).Our generalized boosted regression tree analyses indicated that the probability of infection from the environment, the home range size, and the daily probability of mortality from RHDV2 were the most influential variables affecting the population size of rabbits at the end of the simulation (Figure 4).A regression of simulation outcomes on predictions from F I G U R E 4 Relative influence of variables from generalized boosted regression tree results for all simulations.Graph projects the relative influence of variables on the remaining population size at the end of the scenario.Env, per day probability of environmental transmission given a contact with an infected rabbit's home range or the carcass of an infected rabbit's home range; hr, simulated home range size; inf length, the length of time the environment in the home range of a rabbit remains infectious after the death of an infected host and length of time recovered rabbits remain infectious; mort, daily probability of mortality from RHDV2 for infected rabbits; rec, percentage of the population that recovers from RHDV2 infections; vacc, the proportion of the population vaccinated.
F I G U R E 5 Expected change in number of rabbits remaining at the end of the simulations for a change in the parameter estimate.Numbers in parenthesis indicate the unit of change in the parameter estimate that results in the expected change.Error bars represent 95% confidence intervals.Env trans, per day probability of environmental transmission given a contact with an infected rabbit's home range or the carcass of an infected rabbit's home range; home range size, simulated home range size; infection length, the length of time the environment in the home range of a rabbit remains infectious after the death of an infected host and length of time recovered rabbits remain infectious; natural mortality, daily probability of mortality from disease for infected rabbits; recovery, percentage of the population that recovers from RHDV2 infections; vaccination, the proportion of the population vaccinated.
the boosted regression tree resulted in an R 2 value of 0.93 indicating a good model fit.Environmental transmission accounted for 41% of the relative influence, home range size for 31%, mortality for 11%, probability of recovery for 9%, vaccination rate for 6%, and infection length for <1% (Figure 4).
Environmental transmission was the most influential variable (Figure 4).As expected, increasing environmental transmission rates resulted in smaller overall population sizes at the end of the simulation (Figure 4).Specifically for every 1% increase in the per day probability of infection from the environment (given a contact), the total number of rabbits remaining at the end of the simulation declined by 67 rabbits (95% C.I. 62-72) (Figure 5).Similarly, with every 1-m increase in the radius of the home range size, the expected number of rabbits remaining at the end of the simulation declined by 10 rabbits (95% C.I. 9-11).In essence, larger home ranges (resulting in more areas of overlap) led to further spread of disease and a smaller resulting population size.With a 20%-30% vaccination rate and home range radii of 50 m, the median population size at the end of the simulation was 995 rabbits (95% C.I. 820-1300) while a home range radii of 80 m resulted in a median ending population size of 695 rabbits (95% C.I. 273-1224).Finally, an increase of 1% in the vaccination rate resulted in an increase of approximately 6 rabbits (95% C.I. 5-7) at the end of the simulation (Figure 5).

| DISCUSSION
Spatially explicit modeling of RHDV2 in the endangered riparian brush rabbit population predicted that vaccination of roughly 30% or more of the population would have positive conservation benefits.Overall, our model demonstrated that vaccination of wild riparian brush rabbits would prevent the population from declining precipitously in the face of a disease outbreak.Although we demonstrated that vaccination can potentially be effective for riparian brush rabbits, vaccination as a long-term strategy is costly.Rabbits are short-lived and population turnover is high; thus, consistent and frequent vaccination campaigns may be beneficial.Long breeding seasons with a constant influx of susceptible juvenile rabbits reduces the likelihood of achieving herd immunity (Ashby & Best, 2021).We assumed in our model that the vaccine is 100% effective in protecting wild rabbits from RHDV2.Although research is ongoing, the efficacy of this vaccine in riparian brush rabbits remains unknown.Therefore, if management objectives aim to achieve 30% vaccination rates, the actual vaccination rate would need to be higher to account for the possibility of imperfect protection against RHDV2.The duration of immunity following vaccination with Filavac is 12 months in healthy domestic rabbits under laboratory conditions.The manufacturer recommends administering a single dose of the vaccine, followed by annual re-vaccination (booster) in low-risk settings or every 6 months in high-risk settings, such as a rabbitry or rescue center.Annual re-vaccination of wild rabbits may be difficult; therefore, vaccination is most likely to be effective for small, targeted populations rather than broad-scale landscape-wide use.Finally, the spatial configuration of vaccination may contribute to the success or failure of the vaccination campaign.Unexposed unvaccinated rabbits may be clustered in pockets of the habitat indicating that both vaccination and habitat configuration may have slowed the spread of the disease.Spatially, informed vaccination strategies that result in barriers to disease spread may reduce the required vaccination rates by preventing transmission of the disease into new areas.
Currently, there are many information gaps regarding the transmission and spread of RHDV2 in wild rabbit populations in the United States.A recent study determined that eastern cottontails are susceptible to and can potentially transmit the disease (Mohamed et al., 2022).This species is widespread throughout the eastern, central, and southern United States and into Canada (Chapman et al., 1982).Additionally, its range overlaps with New England cottontails (Sylvilagus transitionalis), a species of concern in several states (Connecticut, Massachusetts, Maine, New Hampshire, New York, and Rhode Island; populations are presumed extirpated from Vermont) (Kays & Wilson, 2009).Our models indicate that dense rabbit populations with more contacts (i.e., more connectivity) between individuals are more likely to transmit RHDV2, resulting in more severe outbreaks.Although starting density was the same across all simulations in our study (only home range size and dispersal distance varied), higher density rabbit populations may spread disease more quickly.However, high density populations may also reflect high quality habitat, which could enable population resilience following disease outbreaks or prevent severe outbreaks if well-distributed quality resources prevent high densities of individuals from congregating around or competing for scarce patchy resources (Calvete, 2006).Although we did not explore the effect of population density in our model, our results demonstrate that factors that increase contact rates and likelihood of transmission have a strong influence on the effect of transmissible disease on a population.
Disease transmission is likely more complex in multispecies rabbit communities.In our model, we included the estimated number of sympatric unvaccinated rabbits (i.e., desert cottontails) that occupied the same home ranges as our target species (riparian brush rabbits) (Tarcha, 2020), which contributed to the overall probability of infection.This addition accounts for transmission by other sympatric rabbit species that are susceptible to disease and could inhibit the development of herd immunity in a vaccinated population by lowering the ratio of vaccinated-to-unvaccinated hosts.Additionally, RHDV2 has been detected in domestic rabbits in the United States, which poses a threat to free-ranging wild populations (Clifford & Moriarty, 2022).Inadequate containment facilities for domestic rabbits and improper disposal of contaminated food or bedding can provide an additional source transmission.Finally, Mahar et al. (2018) concluded that the rapid spread of RHDV2 in Australia was likely attributable to multiple reintroductions and potential humanmediated movement of the virus.Thus, there are multiple transmission pathways for RHDV2 to enter and re-enter a population.We did not explore human-mediated transmission or explicitly include domestic rabbit introduction in the current study, but future modeling efforts could include additional risks associated with humans and domestic rabbits.
The role of environmental transmission in RHDV2 dynamics in the United States is also poorly understood.Infected rabbit carcasses contain viable RHDV viral particles for up to 90 days in the environment (Dixon & Krauer, 2022).Possibly similar to RHDV, RHDV2 may degrade more rapidly in dry environments and the ability of the virus to remain viable in feces, carcasses, or contaminated soil is unknown but likely contributes to the spread of the disease (Henning et al., 2005;McColl et al., 2002).Dispersing rabbits, or rabbits recolonizing areas where RHDV2 has extirpated populations, may be at risk of contracting the disease from the environment months after infected rabbits have died.It is also likely that environmental persistence of this pathogen is dependent on temperature, precipitation, and other factors.Seasonal and geographic variability likely affect RHDV outbreak timing and duration, with more pronounced effects of RHDV often observed in arid regions compared to cooler, humid areas (Cooke, 2002).If environmental transmission is a substantial contributor to the maintenance and spread of RHDV2 in wild populations, the risk of RHDV2 becoming established in the United States is high.RHDV can be spread via fomites, such as contaminated food and vegetation, soil, insects, scavengers, or people (Schwensow et al., 2014).We did not explicitly include this low probability event in our model.However, we simulated an introduction of two infected rabbits into the population every 30 days to reflect a worst-case scenario of continuous disease reintroduction; therefore, it is possible our results are more pessimistic than a real-world situation.Currently, how often or whether RHDV2 will reappear in populations once the disease is introduced is unknown.However, given that RHDV2 can be transmitted through both environmental and host-to-host transmission routes, the likelihood of RHDV2 persisting once introduced is high.
In conclusion, models can be useful for evaluating the trade-offs of disease mitigation strategies, creating transparency in decision making, and identifying areas where further research would be beneficial to reduce the uncertainty associated with making management decisions.Our model helped inform the decision to vaccinate riparian brush rabbits for RHDV2, by demonstrating that although labor intensive, RHDV2 vaccination of riparian brush rabbits in the near-term could help to mitigate disease-associated mortality while additional recovery actions, such as habitat restoration (Calvete, 2006) and translocations, are implemented to increase population size and geographic distribution and decrease the potential effects of outbreaks.Trapping and handling of animals always presents the possibility of unintended mortality.However, given this population of riparian brush rabbits represents one of the last remaining viable populations, the risk of catastrophic losses to the population from an RHDV2 outbreak or from RHDV2 outbreak combined with wildfire or flooding, was determined to be unacceptably high in this context.Additionally, implementation of vaccination prior to arrival of the virus was intended to avoid a worst-case scenario of greater than anticipated mortality in an endangered species that could trigger even more costly and risky conservation interventions (i.e., captive rescue).In May 2022, approximately 1.5 years after initiation of the vaccination program, RHDV2 caused mortalities were confirmed in unvaccinated riparian brush rabbits and desert cottontails in the study area.A large mortality event was not observed over summer and fall of 2022.Anecdotally, there may have been a small decrease in rabbit numbers captured in fall of 2022, but extreme drought conditions along with the emergence of RHDV2 could both have contributed to a possible reduction in brush rabbit numbers in our trapping areas.Despite the uncertainties and information gaps regarding transmission dynamics of RHDV2 in the United States, model predictions from this study were critical for informing the implementation of a subsequent vaccination for this endangered species.
Takahashi.Designed the methods, led the analysis, led the interpretation of the data, and created the final figures: Robin E. Russell.
Map of location of San Joaquin River National Wildlife Refuge within the State of California.(b) Map of San Joaquin River National Wildlife Refuge and associated conservation lands surrounding the refuge, units in gray are units with documented estimates of abundance that were included in our study.

F
I G U R E 2 Probability of infection as a function of time of year.Infection rates decrease over the summer, increase in the fall, and are low throughout the winter.

F
I G U R E 3 Proportion of rabbits remaining under a range of vaccination scenarios.Dots are median values, and error bars represent 95% confidence intervals.