Use of Simulation Modeling to Evaluate Management Strategies for Reintroduced Riparian Brush Rabbits in California

Population viability analysis (PVA) was developed to evaluate the status of populations and to assess factors affecting their persistence in a quantitative way (Morris et al. 1999). In recent years, the use of PVA and other demographic decision models has become increasingly common to predict extinction risk of populations and identify likely outcomes of alternative management options aimed at mitigating these risks (Wittmer et al. 2010; Armstrong and Reynolds 2012; Converse et al. 2013; Pe'er et al. 2013). In the context of species reintroductions, PVA has been used to model likely consequences of initial release strategies as well as management strategies following population establishment (Converse et al. 2013). For example, PVA has assisted in determining the number of captive-bred individuals that can be safely allocated for reintroductions (Zeoli et al. 2008), identifying appropriate levels of supplementation for small populations (McCleery et al. 2005), and evaluating potential effects of management actions on a population's growth rate (Armstrong et al. 2007).

Although population models have been used to guide species reintroductions for many years, model structure, complexity, and the choice of modeling platform remain the focus of current research (Armstrong and Wittmer 2011; Armstrong and Reynolds 2012; Bell et al. 2013; Gedir et al. 2013; Lacy et al. 2015). There is, however, longstanding agreement about the overwhelming importance of incorporating uncertainty, both in terms of parameter estimates and model selection, into any modeling approach aimed at quantifying the dynamics of populations (e.g., Beissinger and Westphal 1998; Burnham and Anderson 2002; Goodman 2002; McGowan et al. 2011; Armstrong and Reynolds 2012). Once uncertainty is incorporated, purpose-built spreadsheet models (White 2000; Armstrong and Reynolds 2012) as well as widely available modeling packages (e.g., Lacy et al. 2015) can be used effectively to predict population viability and to inform management.

Brush rabbits Sylvilagus bachmani occur in diverse habitats of western North America from Baja California to northern Oregon. The riparian brush rabbit S. b. riparius (Figure 1) is a distinct subspecies restricted to areas of thick, brushy cover along rivers and tributaries in the Central Valley of California, USA (Williams et al. 2008; Kelt et al. 2014) and is considered endangered under the U.S. Endangered Species Act (ESA 1973, as amended). By the mid-1980s, suitable riparian forest within the historic range of this subspecies had been reduced to a few small and widely scattered fragments, totaling approximately 2,100 ha (Williams et al. 2002, 2008), and only two extant populations of riparian brush rabbits remained (i.e., Caswell Memorial State Park and South Delta; Williams et al. 2002). Both remaining populations were small and considered at risk of imminent extinction from demographic and environmental stochasticity (including or resulting from flooding, wildfire, habitat conversion, disease, or predation), and possibly from competition with desert cottontails S. audubonii (USFWS 1998; Williams and Basey 1986). The perceived risk of extinction led to their listing as an endangered subspecies by the state of California in 1994 (CDFW 2015) and subsequently by the federal government (USFWS 2000). In response to the urgent threats faced by the wild populations, the Endangered Species Recovery Program at California State University, Stanislaus (Turlock, California) initiated a controlled propagation and reintroduction program in November 2001 (Williams et al. 2008; Hamilton et al. 2010).

Using a release strategy that allowed individuals to acclimate to their new surroundings (Kleiman 1989), the Endangered Species Recovery Program reintroduced 325 captive-bred individuals to unoccupied habitat at the San Joaquin River National Wildlife Refuge (“Refuge” hereafter) between July 2002 and July 2005 (Williams et al. 2008; Hamilton et al. 2010). Habitat in the Refuge was restored concurrently with the reintroduction, and restoration included revegetation of levees and provision of flood refugia. Postrelease survival of rabbits was monitored with radiotelemetry, and Hamilton et al. (2010) documented that monthly survival was affected by acclimation time as well as environmental conditions at the release site. In particular, postrelease mortality was highest during the first 4 wk following reintroduction, and both body mass and length of time in the release enclosure generally improved survival. An analysis accounting for release mortality during the acclimation period suggested that subsequent monthly survival probabilities were most strongly influenced by release year (year 1 vs. years 2 and 3) and a catastrophic flooding event. However, length of acclimation time in soft-release enclosures remained an important variable in longer-term survival.

Hamilton et al. (2010) provided data with which to evaluate the probability of persistence of the reintroduced population as well as the likely effect of release and postrelease management strategies. Using the available survival estimates and estimated reproductive parameters, we developed a PVA to determine the risk of extinction of the reintroduced population under a series of alternate management scenarios. Our specific objectives were to 1) develop a PVA model for the reintroduced riparian brush rabbit population at the Refuge, and determine its probability of persistence over a time-period relevant to both conservation and management (e.g., 20 y); 2) determine the impact of potential management strategies (e.g., release strategies, including supplementations) on population persistence; and 3) identify model parameters that may need additional refinement. Based on the latter objective we provide recommendations for key ecological features in need of further study. Although we apply this approach to riparian brush rabbits, it should be of general interest to managers charged with developing optimal release and postrelease management strategies for reintroduction projects.

The San Joaquin River National Wildlife Refuge was established in 1987 on the San Joaquin River approximately 18 km west of Modesto in Stanislaus County, California (37.615°N, 121.213°W). The Refuge comprises 2,688 ha and lies within the historic range of riparian brush rabbits. Habitat restoration began in 2002 and included planting of riparian habitat, revegetation of levees, and establishment of raised mounds to serve as flood refugia at strategic locations within the Refuge (Figure 2; Williams et al. 2008; Kelt et al. 2014). For a more detailed description of the study area, see Hamilton et al. (2010).

We propagated captive-bred riparian brush rabbits in three 0.5-ha outdoor enclosures (Figure 3) established in habitat similar to that at release sites and approximately 64 km north of the Refuge (Williams et al. 2008). Between November 2001 and July 2005, 476 offspring were born within the breeding enclosures. Of these, we reintroduced 325 individuals weighing ≥400 g (a single exception was 1 rabbit weighing 395 g) to the Refuge over a period of 3 y from July 2002 to July 2005 (49, 231, and 45 individuals, respectively). We transported individuals designated for reintroduction by vehicle from the breeding enclosures to the Refuge where we released them temporarily into additional fenced enclosures. Release enclosures varied in size between 0.30 and 0.40 ha. Enclosures were surrounded by custom-made 1.5-m-tall fences consisting of poultry netting with a 2.54-cm mesh size. Rabbits remained confined in release enclosures for 2–20 d, allowing them to become familiar with general habitats in the area. The duration of confinement varied by individual and averaged 9.3 d in 2002–2003, 4.6 d in year 2003–2004, and 5.6 d in year 2004–2005 (Hamilton et al. 2010). For actual release into the wild, we opened multiple 6–7-m-wide sections of the netting so rabbits could exit enclosures at will.

We fit all except 1 of the 325 released individuals with radiocollars prior to release, and we attempted to recapture them for transmitter replacement before batteries failed. We monitored the status of collared individuals at least twice per week throughout the study and used encounter histories to determine postrelease survival. Further details of postrelease monitoring, identification of likely causes of mortality and survival analyses are reported in Hamilton et al. (2010). All capture and handling procedures were approved via permits from the U.S. Fish and Wildlife Service and the California Department of Fish and Wildlife, as well as the University of California, Davis, Animal Use and Care Administrative Advisory Committee, and procedures met guidelines recommended by the American Society of Mammalogists (Sikes et al. 2011). The reintroduced population continued to be supplemented annually with captive-bred rabbits until 2013 (total = 576 individuals, range = 4–134 individuals/y).

Because of the apparent variation in survival probabilities associated with the duration of acclimation periods in release enclosures as well as environmental conditions such as flooding (Hamilton et al. 2010), we developed a range of PVA models to simulate the dynamical consequences of alternate reintroduction and management strategies and to estimate population persistence (Table 1). We performed simulations using Vortex 10.1 (Lacy et al. 2015), which applies Monte Carlo simulations to model populations as individuals progress through events in their life cycle (births, deaths, catastrophes, etc.). Additionally, Vortex allows users to explore the effects of deterministic forces and stochastic events on the dynamics of these populations. We chose Vortex over more flexible, purpose-built spreadsheet models because it enabled researchers and managers with little previous modeling experience to contribute effectively to model development while simultaneously incorporating stochasticity (Lacy et al. 2015).

Our modeled populations consisted of two demographic stages—reproductively inactive juveniles aged 1–3 mo and all adults ≥4 mo old (Figure 4). We used a “time cycle” of 30 d (equivalent to 1 Vortex “year”) for all simulations and adjusted estimates of vital rates accordingly. We selected a 30-d interval because it was most compatible with the gestation period of riparian brush rabbits.

For each scenario (see below), we ran 1,000 simulations over a 20-y period. Such relatively short time-frames have been recommended as more appropriate for population modeling of threatened and endangered species (Beissinger and Westphal 1998; Pe'er et al. 2013). We set a demographically based extinction threshold at one sex remaining. Criteria used to evaluate riparian brush rabbit viability included risk of terminal extinction as well as population trajectory.

We estimated population-specific reproductive parameters for riparian brush rabbits including age at first and final reproduction, proportion of females producing litters, litter size, and offspring sex ratio both from data collected during controlled propagation and from monitoring of translocated individuals (Williams et al. 2008). We trapped in captive-propagation enclosures at 2-wk intervals to assess age-specific reproductive status, litter size, and sex ratio of offspring. In addition, we trapped an average of 5 d/mo at the reintroduction site to assess reproductive parameters of the translocated population, mark young rabbits, and refit lost or failing transmitters (Williams et al. 2008). Note that we based the proportion of females breeding (46%) exclusively on data collected from radiocollared animals at the Refuge. When necessary, we supplemented estimates of reproductive parameters with observations from previously published studies of lagomorph species with similar life-history strategies (Orr 1940; Mossman 1955; Sowls 1957; Chapman and Harman 1972). We summarize resulting parameter estimates used in our PVA in Table 2.

We took information pertinent to adult survival from Hamilton et al. (2010), who presented survival estimates for two time-periods. An initial assessment of postrelease survival indicated that survival in the first 4 wk after release was significantly improved with additional time in release enclosures (acclimation time) as well as by body mass. Additionally, after excluding release-related mortalities, monthly survival probabilities following this 4-wk postrelease period were greater during the first year of reintroduction than in the subsequent 2 y. The decrease in survival after the initial releases in year 1 may have been affected by competition for resources with established rabbits, increase in proficiency of predators, and reduced time in release enclosures; acclimation time was almost twice as long in year 1 as in years 2 and 3 (mean = 9.3, 4.6, and 5.6 d, respectively). However, variation in survival also was affected by environmental conditions. For example, survival was reduced by approximately 30% during March 2005 when large areas of the Refuge were inundated by severe flooding (Hamilton et al. 2010). We therefore derived two different estimates of mortality from results presented in Hamilton et al. (2010). First, we estimated monthly survival probabilities based on the first year of translocations when the population likely was at a lower density and rabbits had longer acclimation periods. Second, we estimated monthly survival probabilities based on the second and third years of the translocations, when the population likely was at a higher density and acclimation periods were shorter. Several models in the original survival analysis had similar levels of support (ΔAIC_c < 2), so we used results averaged across top models for all estimates of adult survival (Hamilton et al. 2010). We converted survival probabilities into monthly mortality estimates for use in our PVA (Table 2).

To account for the large effect of flooding on rabbit survival (Hamilton et al. 2010), we incorporated seasonal flooding in alternative model scenarios. We used river stage data to determine flood frequency at the Refuge (California Department of Water Resources Data Exchange Center; http://cdec.water.ca.gov/). River levels exceeding 12,000 and 17,000 ft³/s cause moderate and severe flooding, respectively (J. Rentner, personal communication; Figure 5). Historical records indicate that three high-severity floods and five moderate floods occurred between 1989 and 2009. In response, we included two flood scenarios—a 15% annual probability of severe flood and a 25% annual probability of moderate flood. To simulate the effect of catastrophes on survival and reproduction, Vortex multiplies the “normal” survival rates by a severity factor. These values can range from 0 (no survival or reproduction) to 1 (no effect). We derived severity factors from observations during moderate flooding in March 2005 (severity factor = 0.70; Hamilton et al. 2010; known-fate model-adjusted values) and severe flooding in April 2006 (severity factor = 0.09; Endangered Species Recovery Program, unpublished data).

We set juvenile survival equal to that of adults because data on this parameter are unavailable. Because juvenile survival commonly is lower than that of adults in many mammals (Caughley 1966), this simplification may have positively biased our estimates of population persistence. We derived survival data from time of translocation to death or disappearance; most individuals had attained or nearly attained reproductive ability during this interval. It is possible that our estimates of persistence are unduly optimistic, although our results underscore that this does not complicate model interpretation.

We knew initial population size (23 females, 26 males) and age structure from the founder reintroduction. Williams (1993) estimated the population size for riparian brush rabbits at Caswell Memorial State Park at 304 rabbits over 104 ha, and proposed that the rabbit population was probably at or near carrying capacity (K). Using this estimate and the approximate amount of available habitat at the Refuge (850 ha), we calculated a potential K = 2,550 rabbits at the Refuge. This likely is a conservative estimate because the Refuge contains more early successional riparian forest areas than Caswell Memorial State Park and brush rabbits prefer this habitat type to climax communities (Kelt et al. 2014). Thus, the Refuge could likely sustain higher densities of rabbits than Caswell Memorial State Park. For our PVA scenarios, we set K conservatively at 1,300 to obtain growth rates and population trajectories unaffected by density dependence until K is reached. Results from preliminary analyses indicated that increasing K had no qualitative effect on persistence.

Using estimates of vital rates, we first built a baseline scenario to model growth and persistence of riparian brush rabbits at the Refuge. The baseline model had a founder population of 49 individuals released over the first year and was supplemented monthly by 12 individuals (6 females, 6 males) for an additional 23 mo (the end of year 3). The total number of rabbits released thus equaled 325 individuals, matching the exact number released at the Refuge during the first 36 mo of the study. We then compared qualitative and quantitative results from the baseline scenario with a range of alternate model scenarios (see Table 3 for a complete list of models considered).

• a) 

  Founder effects: To determine if the initial number of rabbits released affected population growth and persistence, we built two models where the founder population consisted of 325 individuals using slightly male-biased sex ratios as well as adjusted sex ratios (i.e., equal number of females and males), respectively.

• b) 

  Population supplementation: Because we released additional rabbits into the Refuge for several years past the initial reintroduction period, we built additional models aimed at understanding the possible effects of population supplementation. Specific models evaluated the effect of 1) supplementing individuals (n = 276) in a single release event to the initial founder population of 49 rather than monthly over 2 y as in the baseline model (total n = 325 individuals); 2) supplementing 48 individuals (24 females, 24 males) annually between year 2 and year 10 (total n = 432 individuals); and 3) supplementing 48 individuals (24 females, 24 males) annually between year 2 and year 20 (total n = 912 individuals).

• c) 

  Release strategies: For baseline as well as 10- and 20-y supplementation scenarios, we evaluated how increased survival probabilities associated with longer acclimation times affected population growth and persistence. We parameterized these models using estimates of increased survival probabilities described by Hamilton et al. (2010).

• d) 

  Flooding: For baseline as well as 10- and 20-y supplementation scenarios, we also evaluated if lower survival probabilities associated with moderate and severe flooding affected population growth and persistence. We summarize probability of flood events to occur and total reduction in survival and reproduction in Table 2.

• e) 

  Sensitivity analyses: Finally, we evaluated sensitivity of model predictions to uncertainty in parameter estimates by systematically varying each vital rate by 10% and 25% for all models (see Table S1, Supplemental Material). We evaluated uncertainty associated with age of first reproduction by increasing minimum age of reproduction from 4 to 5 and 6 mo instead. We then calculated the difference (Δ) between the intrinsic growth rate (deterministic; hence Δdet-r) based on the original scenarios and those accounting for uncertainty in vital rates.

All model scenarios resulted in high probabilities of extinction of introduced populations of riparian brush rabbits in the absence of either ongoing supplementations or significant improvements to vital rates. For example, all baseline scenarios resulted in terminal extinction risks of 1.0 and median times to extinction ranging from 4.6 to 6.8 y (Table 3). Extinction risks under the baseline scenarios were independent of initial release strategies. The risk of extinction remained high (>99%) even when the initial founder population (n = 49 individuals) was supplemented with captive-bred brush rabbits (48 individuals annually) for up to 10 y. Supplementing the population with 48 individuals annually for the entire 20-y period would be required for the population to persist given current estimates of vital rates. Population declines, however, would resume once supplementation ceases because the deterministic intrinsic rate of population growth remains negative. Population persistence is also likely under the baseline scenario (9% extinction probability) and 10-y supplementation scenario (0% extinction probability) once survival rates increased to those observed for individuals kept in release enclosures for prolonged time-periods. Expected future population sizes under these scenarios, however, would be low (Figure 6).

The probability of population persistence is further reduced once the possibility of moderate and severe floods occurring is incorporated into models. Assuming an effect of flooding on reproduction and survival, even populations that were considered viable otherwise (e.g., 10-y supplementation scenario with increased survival associated with longer periods in acclimation pens) were converted to populations at risk of extinction. As with earlier scenarios, population persistence was likely only under the 20-y supplementation scenario.

Estimates of population growth were almost twice as sensitive to changes in survival (Δdet-r = 0.011) as to changes in female reproductive activity and number of offspring per female (Δdet-r = 0.006). Estimates of population growth were least sensitive to possible delayed onset of age of first reproduction (Δdet-r = 0.003; see Table S1, Supplemental Material). Results also showed that even a 25% increase in survival probabilities of females would be insufficient to change estimates of population persistence qualitatively.

Results suggest that, in every scenario, either continued supplementations of riparian brush rabbits or improved vital rates will be required for population persistence. In the absence of ongoing releases, local extirpation appears inevitable given available data. Although the probability of persistence likely would have been higher had rabbits been retained in release enclosures for longer periods of time, our models suggest that the gain from this strategy would be insufficient to yield a viable population of brush rabbits at or around the presumed current carrying capacity for the Refuge. However, even with continued supplementation, optimized release strategies may not result in a viable population (e.g., Letty et al. 2000; Van Zant and Wooten 2003). Although continued supplementation might delay local extirpation, thereby providing researchers additional time to diagnose limiting factors and correct habitat deficiencies (McCleery et al. 2005), they also may mask unrecognized problems, which would prolong the time required to detect declines in population trajectories and inadvertently increase the time it takes for managers to adjust procedures. Thus, managers of this subspecies face a challenging situation: is it better to continue local supplementation while attempting to clarify further the factors that appear to lead inevitably to local extirpation, or to cease releases temporarily while a better understanding of these factors is pursued?

Population persistence is a primary goal of reintroduction programs (Armstrong and Seddon 2008), but all modeling scenarios yielded deterministic estimates of the intrinsic rate of population growth (r) < 0, suggesting that the reintroduced population would not be self-sustaining and that releases may not provide the anticipated benefits at the Refuge. Suboptimal population growth reflects either unexpectedly low reproductive output or unexpectedly high mortality. Future research should thus focus on identifying the most limiting demographic parameter. Given the sensitivity of model predictions to changes in estimates of survival, particular effort should be directed at understanding mechanisms underlying current mortality rates. Although the majority of sources of riparian brush rabbit mortality were unknown, predation from avian and mammalian species was by far the most frequent known cause (figure 3 in Hamilton et al. 2010). This suggests that both active and passive predator management could be attempted to improve brush rabbit survival at the Refuge.

An additional explanation for the predicted population decline is that habitat conditions necessary for population persistence at the Refuge were not as suitable as initially assumed. Although the Refuge falls within the historic range of brush rabbits, it is not yet clear how well the available habitat reflects the needs of riparian brush rabbits. At the Refuge, reintroduced riparian brush rabbits preferentially use dense, shrubby understory vegetation over other available habitat types (Kelt et al. 2014), and the Refuge appears to provide more and better habitat for riparian brush rabbits than that at Caswell Memorial Park or the South Delta, where populations have remained relatively stable over many years. However, no formal studies have evaluated other integral components of habitat for brush rabbits such as food preferences, resource availability, or predator and parasite responses in the reintroduced population. Results presented here call for investigation of all other potential limiting factors for riparian brush rabbits to identify potential causes of decline (Caughley 1994). When faced with similar challenges, Cabezas and Moreno (2007) showed that the effectiveness of translocations of European rabbits Oryctolagus cuniculus increased once they also corrected for additional factors (e.g., shelter, food resources) that limited population growth.

Further investigations into causes of current low estimates of vital rates are essential to reconcile our predictions of rapid extirpation at the Refuge with the apparent persistence of established brush rabbit populations at Caswell Memorial Park, the South Delta, and within one region of the Refuge (i.e., Chrisman Island). Specifically, RBR numbers on Chrisman Island were greatly affected by the spring 2011 floods, yet appear to have rebounded in the absence of further supplementation under recent drought conditions (P. Kelly, unpublished data). Furthermore, populations at Caswell Memorial Park and the South Delta have persisted without management intervention despite sharing some of the same environmental conditions, most notably a severe risk of flooding. The estimated population at Caswell Memorial Park varied from a low of approximately ≤10 following flooding in winter of 1985–1986, to an estimated maximum of 303 (range = 212–758) individuals (Williams 1993). Despite these large temporal fluctuations in population size, this population has persisted while being restricted to an area of <10% of that available at the Refuge (Endangered Species Recovery Program, unpublished data). The South Delta population—first confirmed in 1998—currently exists in approximately 109 ha of habitat and has endured despite extensive habitat fragmentation. Habitat there may support, on average, only 25–100 rabbits (Williams et al. 2002). These observations suggest that survivorship and reproduction in established populations may be higher than those observed in the reintroduced population through 2005.

Although sensitivity analyses did not indicate that extinction risk strongly was influenced by the proportion of females breeding, we believe that estimating this parameter in field conditions under differing population densities is one of the most critical parameters needed by managers and conservation planners. Whereas the proportion of females breeding in our propagation facility was approximately 89%, the proportion at the Refuge was only 46%. This parameter may be artificially high in propagation pens where rabbits are protected from predators and have ample natural food and some veterinary care (Zeoli et al. 2008). However, our estimate from the Refuge is similar to values reported for wild populations of lagomorphs: approximately 58–60% for brush rabbits (Chapman and Harman 1972; Basey 1990), approximately 47% for desert cottontails (Sowls 1957), and approximately 54% for European rabbits (Bell 1977). None of these studies, however, extended longer than 1–2 y, and none, including those used in our analyses, extended through a complete population cycle.

Another key mortality factor is flooding (Hamilton et al. 2010). Historical data indicate that the Refuge floods regularly and severely. Although flooding was, and likely remains, a threat to this subspecies, the greatly reduced availability of Refuge habitat resulting from stream channelization and agricultural development greatly elevates the threat of flooding to remaining populations. Until recently, the reintroduced Refuge population had very limited available refugia, and individuals likely were required to traverse long distances before encountering suitable high ground. Probably only a few individuals near levees were successful at this, although proximity to higher ground is not the only factor influencing survival because flood waters may rise rapidly and typically include abundant debris that forces animals into shrubs and small trees where they may be trapped and exposed to predation, hypothermia, or starvation. Newly constructed mounds and berms (n = 33) may serve as refugia in future flood events, and planned breaching of levees will allow floodwaters to disperse through the property at lower levels in the future. The intent is to allow a return of a normal floodplain and flood dynamics to this area, lowering the flood level so that existing levees can be vegetated and serve as additional refugia. Higher ground also will be available on the perimeter of the restored areas (Williams et al. 2008). Nonetheless, in our simulations, extinction risk was high for all model scenarios, even in the absence of flood risk, indicating that factors other than flooding may be limiting the growth rate of this subspecies.

Because data for most Sylvilagus fail to support substantial density-dependence in fertility and fecundity, Edwards et al. (1981) suggested that population persistence for eastern cottontail S. floridanus in Illinois was dependent primarily on survival, and this seems likely to be true for other Sylvilagus. Brush rabbits, like other Sylvilagus, are subject to various predators (Chapman 1974; Chapman and Litvaitis 2003), and while this is probably the major factor in the death of most Sylvilagus species (Chapman and Litvaitis 2003:118), little work has assessed the role of predation relative to other sources of mortality. For riparian brush rabbits at the Refuge, monthly survival was 71% during the first 4 wk postrelease, and 89% thereafter (Hamilton et al. 2010). The latter corresponds to roughly 24% annual survival, but we lack comprehensive data on causes of mortality. The Refuge hosts a diverse suite of potential predators, including coyote Canis latrans, red fox Vulpes vulpes, bobcat Lynx rufus, striped skunk Mephitis, raccoon Procyon lotor, domestic dog Canis familiaris and cat Felis catus, and a variety of diurnal raptors (esp. red-tailed hawk Buteo jamaicensis) and owls (esp. great horned owls Bubo virginianus). Given that brush rabbits rely on shrub cover to avoid predation, further efforts to understand the interactions among habitat and food availability, rabbit movement patterns, and context-dependent predation on riparian brush rabbits at the Refuge may be instrumental in reducing mortality (Bond et al. 2001). Information about predation should be incorporated into future PVA assessments (e.g., Wittmer et al. 2014).

Reintroduction programs can be divided into phases representing population establishment and population persistence (Armstrong and Seddon 2008). During the former phase, release-related management decisions can significantly affect the success of a reintroduction project. The ultimate persistence of reintroduced populations, however, depends on the availability of habitat, and populations without access to habitat are unlikely to persist without continued intervention. In such cases, managers may consider suspending reintroductions until habitat becomes available, or until limiting factors are better understood. Available data (vital rates estimated through July 2005) suggest that the population at the Refuge may not be able to persist in the longer term, underscoring concerns that our understanding of habitat needs of riparian brush rabbits remains incomplete. In 2013, we therefore made a decision to suspend release efforts of riparian brush rabbits temporarily to allow time for additional monitoring and calibration of model parameters. Preliminary observations (e.g., via driving or walking on the Refuge) suggest that abundances may have remained similar to those before releases were terminated, suggesting that the population may be more stable than indicated by our analyses.

Finally, we emphasize that reintroductions in general, and further work on riparian brush rabbit in particular, should be structured within an adaptive management framework and include review by experts (Armstrong et al. 2007; Runge 2013). For example, to understand the role of refugia in reducing mortality risks from flooding, survival estimates of rabbits with access to artificially raised mounds could have been compared with those of individuals without. Using adaptive management would have thus enabled us to assess the effects of management strategies earlier on and use such an understanding to refine our management strategies. Although a posteriori analyses can resolve some problems in reintroductions (Armstrong and Seddon 2008), earlier simulations may allow managers to evaluate sensitivity of extinction risk to different population parameters before extensive resources are allocated, thereby targeting key research needs. Expert-review of our PVA, on the other hand, helped resolve issues of model structure and interpretation of model outcomes. With targeted research and real-time attention to key demographic parameters there is no reason to believe that this reintroduction cannot be successful and provide a model for further reintroductions of this endangered subspecies to restored habitat.

Table S1. Model scenarios to evaluate sensitivity of population viability analysis (PVA) model predictions over a 20-y time-period for riparian brush rabbits Sylvilagus bachmani riparius at the San Joaquin River National Wildlife Refuge, California, to parametric uncertainty in vital rates collected between 2001 and 2005. For each scenario (see Table 1), we varied the following vital rates by 10% or 25% (monthly intervals for age of reproduction): proportion of females producing litters (repro), litter size (offspring), age of first reproduction (age repro), and mortality rates (mort). Results shown as intrinsic rate of increase (deterministic and stochastic ±SD), probability of extinction, extant population size ± SD, and median and mean times to extinction (in years).

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S1 (51 KB XLS).

Reference S1. [USFWS] U.S. Fish and Wildlife Service. 1998. Recovery plan for upland species of the San Joaquin Valley, California. Region 1, Portland, Oregon.

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S2; also available at http://ecos.fws.gov/docs/recovery_plans/1998/980930a.pdf (37.7 MB PDF).

Reference S2. Williams DF. 1993. Population censuses of riparian brush rabbits and riparian woodrats at Caswell Memorial State Park during January 1993. Lodi: California Department of Parks and Recreation, Final Report.

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S3 (232 KB PDF).

Reference S3. Williams DF, Basey GE. 1986. Population status of the riparian brush rabbit, Sylvilagus bachmani riparius. Sacramento: California Department of Fish and Game, Wildlife Management Division, Nongame Bird and Mammal Section, Final Report.

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S4 (1.4 MB PDF).

Reference S4. Williams DF, Kelly PA, Hamilton LP. 2002. Controlled propagation and reintroduction plan for the riparian brush rabbit. Turlock: Endangered Species Recovery Program, California State University; and Sacramento, California: U.S. Fish and Wildlife Service.

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S5: (1.8 MB PDF).

Reference S5. Williams DF, Kelly PA, Hamilton LP, Lloyd MR, Williams EA, Youngblom JJ. 2008. Recovering the endangered riparian brush rabbit (Sylvilagus bachmani riparius): reproduction and growth in confinement and survival after translocation. Pages 349–361 in Alves PC, Ferrand N, Hacklander K, editors. Lagomorph biology: evolution, ecology, and conservation. Berlin Heidelberg: Springer-Verlag.

Found at DOI: http://dx.doi.org/10.3996/052015-JFWM-045.S6 (279 KB PDF).

Funding for this cooperative project has been provided by many agencies and programs, including U.S. Bureau of Reclamation (Conservation Program; Habitat Restoration Program; South Central California Area Office in Fresno); CALFED; U.S. Fish and Wildlife Service (San Luis National Wildlife Refuge Complex; Burned Area Emergency Response; Sacramento Fish and Wildlife Office; Pacific Southwest Region Office); and the California Department of Fish and Wildlife. We thank members of the Riparian Mammals Technical Group for ongoing funding, discussion, advice, logistical and other support. We are grateful to the veterinarians of the University of California Davis Wildlife Health Center and Veterinary Medical Teaching Hospital, particularly K. Gilardi and S. Larsen, for their care and expertise. We also thank River Islands, LLC, especially S. Dell'Osso, for allowing us to capture riparian brush rabbits on their property to use as brood stock in the controlled propagation program. For their many hours of hard work and their dedication to this project, we thank the past and present members of the riparian brush rabbit team and the California State University, Stanislaus-Endangered Species Recovery Program, particularly L. Hamilton, M. Lloyd, S. Phillips, and E. Williams. Finally, we are grateful to D. Armstrong, C. McGowan, the Associate Editor, and one anonymous reviewer whose comments greatly improved this manuscript.

Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.