Short ‐ term anuran community dynamics in the Missouri River floodplain following an historic flood

Extreme ecological disturbances such as floods, wildfires, and droughts are difficult to study because they are rare and unpredictable. We had the opportunity to study the effect of an historic flood on an anuran community in the Missouri River floodplain. We used occupancy estimation to estimate the proportion of wetlands occupied by calling adult male anurans. Three species—the plains leopard frog (Lithobates blairi), Woodhouse's toad (Anaxyrus woodhousii), and Blanchard's cricket frog (Acris blanchardi)—had only minor changes in occupancy rate two years after the flood. Colonization rates for these species were positively associated with wetlands that were shallower near the shore and they did not appear to be affected by reduced vegetation. Three other species or species complexes—the northern leopard frog (Lithobates pipiens), the gray treefrog complex (Hyla versicolor and Hyla chrysocelis), and the boreal chorus frog (Pseudacris maculata)—had greatly reduced occupancy rates two years after the flood. Colonization rates for these species were relatively low, while they had high extinction rates. Colonization rates for these species were not associated with any habitat characteristic we measured. Future flood events will likely continue to make northern leopard frogs, gray treefrogs, and boreal chorus frogs a less important part of the ecological community. While some species may fare well under extreme climate events, such as this flood, that are forecast under climate change scenarios, many species will struggle.


INTRODUCTION
Disturbance regimes are a natural part of ecosystem dynamics and species are adapted to varying magnitudes of disturbance at varying intervals.However, biological communities occasionally experience disturbances so extreme in magnitude that the consequences may be severe.Such disturbances may include wildfires, floods, hurricanes, and droughts, which cause drastic alterations in community composition and diversity, and extirpation of sensitive species (Poff 2002, McKenzie et al. 2004, Ryan et al. 2015).Such perturbations are, by definition, rare, and opportunities to directly study ecosystem impacts are rarer (Bender et al. 1984).
Historically large-river floodplains were very dynamic systems (Junk et al. 1989) and local extinction and recolonization of floodplain species were probably regular processes of varying magnitudes.However, damming and channelization of the Missouri River in the 1940s and 1950s (Nilsson et al. 2005) created a much more static floodplain ecosystem that has a different response to flooding.We would expect that extreme water flows through a floodplain today are likely to compromise conservation and management goals which have been largely developed and implemented based on the assumption of a modified static system.Effects of floods on plant communities include decreased species richness and diversity (Bornette and Amoros 1996).Intermediate levels of flooding disturbance usually have relatively minor effects on most invertebrates and vertebrates because of their mobility, life history, or ability to immigrate from surrounding areas (Crandall et al. 2003).Normal flood regimes have been found to have a determinative effect on the species richness of amphibians in floodplains.Drainage basins in southern Spain that varied in flood regimes were found to vary in amphibian species richness from three to 12 (Real et al. 1993).But we are unaware of any studies on herpetofauna quantifying effects of extreme flooding.
In 2010 we initiated a study of the anuran community in Missouri River floodplains along the Iowa border.Our objectives were to assess functionality of restored and natural wetlands using anuran species population status as our metric, and to investigate the association between population presence and landscape and habitat covariates to inform future wetland management efforts.Modifications to the floodplain have likely changed anuran distribution and abundance in the floodplain (Eskew et al. 2012).In 2011, an historic flood occurred on our study area (Fig. 1).Water flow volume exceeded records dating back to 1898 (Grigg et al. 2011).Large areas of the historical flood plain in the reach from Sioux City, Iowa, to Omaha, Nebraska, USA, where the study area is located, were flooded that had not been flooded since 1952 (Grigg et al. 2011).Flood levels were sustained for several months; it was not the rapid increase and decrease in water level that is common in Midwest floods (Grigg et al. 2011).River processes of scouring and deposition that occurred during the flood drastically changed the wetlands and surrounding habitat matrix.Many areas of the floodplain were submerged by 6 m of water and more than 2 m of sand were deposited in some places ( personal observation).Flood waters extended up to 1 km or more from the river channel.The wetlands were completely inundated for months, displacing or eliminating This natural disaster provided us with an opportunity to investigate several questions about resultant short-term anuran population dynamics with pre-and post-flood data.In the most extreme scenario for a species, flooding may extirpate a species from an area.The ability of species to recolonize areas they have been extirpated from is a critical concern in conservation.We availed ourselves of this opportunity to investigate recolonization of anuran species after extreme flooding, determine what factors may contribute to recolonization failure, and determine what management actions could be undertaken to improve recolonization rates.Because most floodplain environments are already highly modified, it seems reasonable to manage for recolonization following a historic flood, rather than assume that natural recolonization processes will occur.
Recolonization may fail because habitat is not suitable when colonization is attempted, barriers to movement into the area of interest, such as inhospitable habitat, have developed (Rothermel andSemlitsch 2002, Mazerolle andDesrochers 2005), or other factors.Identification of the barriers to recolonization is paramount in determining what action may need to be taken to facilitate recolonization.Species may recolonize unaided from adjacent habitat or be reintroduced through human efforts.In all cases, it is of interest whether species can recolonize unaided, if they can become re-established, how long it will take to recolonize, and if any efforts can aid the restoration of the ecosystem.
We employed occupancy estimation to quantify the number of sites occupied by seven anuran species in the floodplain before and after the flood.We employed habitat and landscape measurements to assess covariates of occupancy, colonization, and extinction rates.A priori, we expected high extinction rates and dramatic reductions in occupancy rates following the flood.Our habitat and landscape covariates aided in explaining the recolonization process that we observed.Our results provide insight into the dynamics of an anuran community following a rare event that is likely to become increasingly common with climate change (CCSP 2008, Walls et al. 2013).

METHODS
The study area included two adjacent mitigation sites along the Missouri River (Fig. 2).We surveyed wetlands in Tieville Bend, Upper Decatur, and Middle Decatur Bends of the Blackbird-Tieville-Upper Decatur Bend mitigation site (Blackbird Bend was not included) and all areas of the Louisville Bend mitigation site (U.S. Army Corps of Engineers 2013).We initially inventoried wetlands in the study area in 2009, and defined survey sites using this inventory.If a wetland was relatively small, the entire wetland was defined as the survey site.For large backwaters and channels, we designated 100 m radius survey plots.In subsequent years, we updated the wetland inventory, especially after the historic flooding of 2011.Additionally, we did not survey wetlands in dry years if there was not sufficient water for anuran reproduction.We surveyed 105 different sites during the study but not all sites were present each year.Depending on the time of year, 41-53 sites were present in 2010, 82 sites in 2012, and 48-52 sites in 2013.Eighteen sites were present across all three years.
We surveyed for presence of calling anurans in 2010, 2012, and 2013.Flooding in 2011 precluded surveys.We completed only a single survey for American bullfrogs in 2010 because of midseason flooding.We surveyed sites repeatedly for analysis in an occupancy estimation framework (MacKenzie et al. 2003).We conducted five surveys of each site in 2010, seven surveys in 2012, and 13 surveys in 2013.We truncated the data at the calling season for each species, as determined by temporal pattern in the dataset.We surveyed beginning 30 min after sunset.We surveyed each site for 5 min and recorded the presence or absence of calling male anurans.At 100 m radius plots, if we were unsure whether a calling adult male was in the plot, we measured the distance to its location using a Global Positioning System (GPS) unit after the survey.We measured water temperature at the edge of the wetland, air temperature, and time of day after the survey.We canceled surveys if precipitation precluded detection, wind speed was greater than 24 km/hour, water temperature was less than 108C, or air temperature less than 08C.We constructed a covariate for the amount of moonshine during the surveys.A full moon on a clear night was a 1 and no moonshine, either from cloud cover, a new moon, or the moon was below the horizon, was a 0. We calculated these values using fraction of the moon illuminated values (U.S. Naval Observatory 2012).
We investigated relationships between occupancy parameters and one or more habitat covariates for each species.The species-specific sets of covariates were chosen based on commonly accepted life history characteristics.Depth at 1 m from the shoreline of the wetland was measured at 20 evenly spaced points around the wetland and the average depth used to calculate the average slope.We characterized emergent vegetation and shoreline cover (bare ground and grass/herbs) using releve ´measures (Mueller-Dombois and Ellenberg 1974).We characterized proportion of the wetland with emergent vegetation as the proportion of the water 1 m from the shoreline with any emergent vegetation.We assessed herbaceous vegetation cover within the strip of terrestrial habitat 1 m from the waterline.One person performed nearly all the assessments to reduced variability among different observers.We conducted habitat surveys once in 2010 and twice in 2012 and 2013, once early in the growing season (April) and later when the vegetation had matured (late May and June).
We calculated distance to nearest wetland and wetland size in a Geographic Information System (GIS).We recorded the boundaries of wetlands , ; 1.5 ha with a GPS unit April-May of 2012 and uploaded it into the GIS.We delineated wetlands .; 1.5 ha using January-April 2012 leaf-off aerial photography.Wetland size fluctuated during the season, but relative to variation in wetland sizes (up to ;15 ha), we determined that this variability was negligible.During the 2011 flood year, we conducted shoreline surveys for metamorphs of anuran species where access to the study area was possible.We divided the waterline of the flooded area into 271 segments of 50 m and recorded the presence or absence of species for each segment (Fig. 3).

Analysis
We analyzed the data using the multi-season occupancy model (MacKenzie et al. 2003).This model estimates parameters that describe the dynamics of change in site occupancy over time.Occupancy rate (w) is the probability a site is occupied, or in the realized sense, the proportion of sites occupied during a particular season, during which the site occupancy state is considered to remain constant.In between seasons, sites could be colonized or go extinct.The extinction rate (e) is the probability that a site that was occupied becomes unoccupied in the next season and the colonization rate (c) is the probability that a site that was unoccupied becomes occupied.Detection probability ( p), i.e., the probability that individuals are detected at an occupied site during a survey, is used to control for false absences.The dynamics described by this model are often referred to as metapopulation dynamics (Hanski 1994).The existence/non-existence of sites in some years (either from being created/ destroyed by the flood or from being dry for an entire season) during the study presented additional modeling challenges.MacKenzie et al. ( 2011) developed a model to specifically address presence/absence of habitat availability.We used a dummy covariate structure in program MARK v www.esajournals.org(White and Burnham 1999) using the multiseason occupancy model to model parameters in manner similar to MacKenzie et al. (2011).We used RMark (Laake 2013) to facilitate computation of the large number of alternative models.
All of our models assumed that colonization and extinction rates were year-specific.We modeled covariates on colonization and extinction rates as additive rather than multiplicative.Colonization and extinction rates are defined in terms of the interval between years but habitat covariates were measured at the time of the last detection occasion, i.e., at the end of the interval.Thus, we assumed that colonization and extinction processes were ultimately decided at the end of the interval, when the anurans decided if a site was appropriate for breeding.Estimates of occupancy, extinction, and colonization rates were estimated with w 1 (.), e(year), c(year) with the best detection probability model appropriated from the covariate modeling.Occupancy rate estimates for 2012 and 2013 were derived from the above parameter estimates.
For covariate modeling, we generally followed the philosophy of Doherty et al. ( 2012) that all possible combinations of models should be run.However, the number of models using all possible additive combinations was prohibitive, so we used a two-step process.We first modeled each parameter as a function of a single covariate.There were four parameters (w 1 , e, c, and p) and we refer to the model for each parameter as a submodel.There were eight, six, seven, and seven submodels (Table 1) for the four parameters, for a total of 2,352 models.
We calculated the cumulative AIC c (Akaike's information criterion corrected for small sample size; Burnham and Anderson 2002) weight for each submodel and eliminated submodels with less than 0.05 weight.The final model set then contained all additive combinations of covariates that had not been eliminated in the first step.We then calculated cumulative AIC c model weights for the final model set.We examined the b estimates from the linear model for each covariate for indications of non-convergence.Models that had convergence problems with more covariates were simplified until covariate betas converged.The sign of each beta estimate indicates the direction of the relationship between the covariate and parameter.We considered covariates to be important if their cumulative AIC c weight was .0.80.
To represent the relationship between a parameter and a covariate graphically, we controlled for model selection uncertainty and other potential covariates in the model by modelaveraging the parameter of interest along the range of the covariate of interest.If other covariates were also present in the sub-model, they were held constant at the mean unless otherwise stated.If parameter estimates from a model were invalid because of numerical convergence problems, that model was removed for model-averaging that parameter, and model weights recalculated for calculating the modelaveraged estimates.

RESULTS
We detected five species and one species complex in sufficient numbers for statistical analysis: northern leopard frog (Lithobates pipiens), plains leopard frog (L.blairi ), Wood- v www.esajournals.orghouse's toad (Anaxyrus woodhousii ), gray treefrog complex (Hyla chrysocelis and H. versicolor), Blanchard's cricket frog (Acris blanchardi ), and American bullfrog (L.catesbeianus).The gray treefrog species are a diploid-tetraploid species complex and no differences in life history are known (Ptacek 1996); therefore we combined them for analysis.Occupancy rate for the boreal chorus frog (Pseudacris maculata) was estimated as 0.77 (SE ¼ 0.08) in 2010 but it was rare and irregular in occurrence after the flood, precluding statistical analysis.They were detected at four sites in 2012 and at each site they were detected only once.In 2013, they were detected at six sites; at five of those they were only detected once.

Species dynamics
Because some sites were only wet for part of the year and different species call at different times of the year, the number of sites surveyed for each species was sometimes different within a year.Consequently, the occupancy rates need to be carefully interpreted because, to give an example, an occupancy of 0.80 for 10 sites is not the same as an occupancy rate of 0.80 for 100 sites from a biological perspective (MacKenzie et al. 2011).However, the number of sites was similar in 2010 (n ¼ 41-53) and 2013 (n ¼ 48-52), so the occupancy rates are approximately comparable for those years.
Three species (plains leopard frogs, Woodhouse's toads, and Blanchard's cricket frogs) had occupancy rates around 0.60 in 2013, which is similar to their pre-flood levels (Fig. 4).Three other species (northern leopard frogs, gray treefrogs, and boreal chorus frogs), on the other hand, were at occupancy rates in 2013 that were much lower than in 2010.No estimates were possible for the boreal chorus frog but clearly it was at very low levels.No occupancy estimates were possible for American bullfrogs in 2010, because we had only surveyed 31 of 52 sites once v www.esajournals.orgduring bullfrog calling season before flooding prevented more surveys.We detected bullfrogs at 13 sites (42% of the sites surveyed), so they were present in at least 25% of the sites before the flood.Bullfrogs likely had a roughly similar occupancy rate in 2010 and 2013, while increasing in 2012.No species was entirely extirpated from the study area.
Colonization and extinction rates (Fig. 5) provided insight into the dynamics that produced the annual occupancy rates, because occupancy rates could remain the same even though different sites were occupied.Plains leopard frog had a relatively high colonization rate in 2012 and did not go extinct at any sites, but in 2013 they went extinct at more sites than they colonized, though they still maintained a high occupancy rate.Woodhouse's toad had approximately equal colonization and extinction rates in 2012, but did not go extinct at any sites in 2013 while colonizing other sites, increasing their overall occupancy rate.Blanchard's cricket frog had approximately equal colonization and extinction rates in 2012, but had a higher colonization rate in 2013.Northern leopard frogs and gray treefrogs had a very high extinction rate in 2012 and an estimated colonization rate of 0.
Northern leopard frogs increased in 2013, but gray treefrogs continued to do poorly, going extinct at all sites they colonized after the flood.Bullfrogs also decreased from 2012 to 2013.
In 2011, the 50-m metamorph survey segments were somewhat irregular when the waterline fluctuated and where the shorelines were convoluted.Consequently, we report observed number of metamorphs and the naı ¨ve occupancy rate (the proportion of segments at which individuals were actually observed) of the survey segments and interpret them broadly (Table 2).We likely missed the emergence of most plains leopard frog metamorphs and Blanchard's cricket frog metamorphs because they were only beginning

Habitat and landscape covariates
Covariates of parameter estimates provided insight into habitat characteristics likely affecting species dynamics (Tables 3 and 4).Slope was an important wetland attribute for all species except for American bullfrogs.Increasing slope was negatively correlated with colonization rate in Woodhouse's toads, plains leopard frogs, and Blanchard's cricket frogs, i.e., these species avoided colonizing sites with steep slopes.Greater slope was positively correlated with extinction rate in gray treefrogs and northern leopard frogs, i.e., these species tended to go extinct at sites with steeper slopes.Besides slope, habitat and landscape covariates associated with colonization and extinction rates tended to be species specific.

Plains leopard frog
Before the flood, the greatest predictor of plains leopard frog occurrence was wetland size, with greater probability of occurrence at smaller wetlands.The extinction rate relationship was not biologically reasonable and is not reported.Colonization rate increased with decreasing slope (Fig. 6).Detection probability was simply dependent on year.

Woodhouse's toad
The strongest correlate of occupancy for Woodhouse's toad before the flood was bare ground.The extinction rate parameter estimate failed to converge when paired with slope in the models with the most weight, so it was dropped from the model set.We constrained w 1 and e to two covariates maximum each so that betas would converge.No covariates for extinction rate reached our threshold of 80% of the cumulative weight.Colonization rate decreased as shoreline slope became steeper (Fig. 6).Air temperature and moonshine were positively associated with detection probability.

Blanchard's cricket frog
No covariate was highly correlated with Blanchard's cricket frog occupancy in 2010.After the flood, increasing percentage of bare ground was the strongest correlate of increasing extinction rate.However, extinction rate was only high at sites with a very high proportion of bare ground (Fig. 7).In 2012, when little vegetation was present, extinction rate was higher, but in 2013, when there was more vegetation and more organic litter, extinction rate was very low and only likely in sites with very high proportions of bare ground.Shallow slope was the strongest correlate of colonization rate (Fig. 6).Detection probability was best predicted by water temperature, with some evidence for a quadratic relationship.

Gray treefrog complex
In 2010, gray treefrog occupancy was positively associated with distance from other wetlands, i.e., more isolated sites had a greater probability of being occupied.In 2012, they tended to v www.esajournals.orgdisappear from large wetlands and wetlands with steep slopes (Fig. 7).Confidence intervals were wide, likely because the small number of sites at which they did not go extinct.Extinction rate was estimated at nearly 1.00 except for very small sites.Extinction rate estimates for 2013 were unreliable.No effects on colonization rate met our standard of 80% cumulative weight.Detection probability varied by year.

Northern leopard frog
Northern leopard frog occupancy before the flood was more likely with increasing emergent vegetation.After the flood, extinction rate increased with shoreline slope (Fig. 6).Colonization rate was not strongly correlated with any habitat covariates.Detection probability decreased with increasing air temperature.

American bullfrog
Occupancy probability in 2012 was positively correlated with wetland size.Conversely, extinction rate in 2013 increased with wetland size (Fig. 7).Estimates for covariates of colonization rate did not converge.Detection probability was primarily correlated with water temperature.

Boreal chorus frog
Boreal chorus frogs had an occupancy rate of 0.77 (SE ¼ 0.08) the first year, but irregularity and rarity of detections after the flood prevented reliable determination of a time period of population closure and therefore we could not conduct reliable statistical analyses.

DISCUSSION
Plains leopard frogs, Woodhouse's toads, and Blanchard's cricket frog were remarkably resilient to the flood, occupying ;60% of sites in 2013, two years after the likely loss of the entire adult cohort.In contrast, northern leopard frogs, gray treefrogs, and chorus frogs were greatly reduced in distribution, occupying ;20% of sites or less in 2013.Consequently the flood did have an effect on community composition by reducing distribution of nearly half of the species, though based on previous research (Real et al. 1993) we expected the effect to be greater.Remarkably, however, no species was extirpated.
From a conservation standpoint, it is encour-aging that no species was extirpated and at least some portion of the community was able to recolonize quickly.The mechanisms of recolonization were probably different for different species.One reason for rapid recolonization may have been the fact that they were able to successfully reproduce in 2011, despite the flooding.Real et al. (1993) considered the amphibian larval stage as sensitive to flooding, but in our case the current was not fast at all at the edges of the flood water, allowing larvae to remain in the flooded area.We conducted call surveys in 2011 up until the early June steep increase in river stage (unpublished data).Species whose breeding season occurred before June were calling.We assume that mating occurred before the flooding for all species except for American bullfrogs and perhaps Blanchards' cricket frog.Tadpoles must have then rose with the flood water.One mechanism of recolonization for leopard frogs and bullfrogs in particular was that adults followed the waterline down as the flood receded until it reached the level of the current wetlands, just as they followed the waterline up as the water was rising.These are species that can commonly be found at the water's edge and so would have followed the water as it receded, especially if there was no vegetation for cover.These species also hibernate in the wetlands, not in the uplands, and so they would benefit from following the flood waters down.Thus, in our case, the amphibious nature of anurans allowed some species to reproduce the year of the flood and quickly replace the adult cohort that was entirely lost.Species that spend their time outside of the breeding season in the uplands and hibernate in the uplands, such as chorus frogs and gray treefrogs, would not be likely to follow the water's edge as it receded.For these species the study area was small enough that it was well within dispersal ranges of many species (Semlitsch andBodie 2003, Semlitsch 2008).However, large areas devoid of vegetation may have served as barriers to colonization.It is also likely that they often found the wetlands themselves unsuitable.
After the flood the landscape was dramatically changed.Most vegetation was scoured away or buried under sediment.The landscape between wetlands in 2012 was initially devoid of vegeta-tion.As vegetation grew back, it was patchy in distribution and not as thick as it was prior to the flood.Emergent vegetation was nearly nonexistent in 2012 and still rare in 2013.In general, species that were more desiccation tolerant or more tolerant of open, unvegetated habitats had higher occupancy rates after the flood.Plains leopard frogs are known to be more desiccation tolerant than northern leopard frogs (Gillis 1979).Species that are associated with heavy emergent and upland vegetation had lower occupancy rates after the flood.
All species except for American bullfrogs were strongly affected by the slope of the wetland, whether it was an effect on extinction rate or colonization rate.For the three species that did well, colonization rate was negatively associated with slope.However, in the species that did poorly, extinction rate was positively associated with slope.Steep slopes may expose adults, eggs, and tadpoles to larger potential fish predators.Flooding introduced predatory fish into many, if not all, wetlands.Shallow water along the shoreline is a safe refuge from large predatory fish for adults, eggs, and tadpoles.Thus, breeding frogs may have preferred wetlands with shallow slopes to avoid predation.It is also possible that they selected wetlands with shallow slopes because it was beneficial to egg and tadpole survival.Tadpoles of some species, toads in particular, congregate in shallow water where presumably it is warmer and they can develop faster (Cupp 1980).Availability of attachment sites for eggs is also greater in shallow water where there is more emergent vegetation.
Woodhouse's toads occurred primarily at wetlands with more bare ground in 2010.Bare ground was rarer before the flood.After the flood bare ground was very common and may have contributed the relatively high occupancy rates of Woodhouse's toads.Woodhouse's toads grow quickly and reach sexual maturity approximately one year after metamorphosis (Clarke 1974) though they may breed later in the breeding season (Sullivan 1987).Many of the toads observed in 2012 appeared to be the 2011 cohort because of their relatively small adult size.It seems likely that many of the Woodhouse's toads that colonized the wetlands in 2012 were toads that metamorphosized at the flood's edge in 2011.Under normal circumstances these first-year males would have little success reproducing (Sullivan 1983).Woodhouse's toad detection probability increased with moonlight, which was unexpected.For most species, it is thought that on moonlit nights they curtail activity to avoid predators (Granda et al. 2008).However, toads have chemical defenses and thus may not be as constrained by predator activity.Moonlit nights may help them find mates more easily.
Plains leopard frogs preferred small wetlands, perhaps because of their adaptation to temporary prairie wetlands (Crawford et al. 2005).Temporary wetlands, which are likely to be small wetlands, are more likely to be devoid of fish predators.Like other species, slope was the primary correlate of their colonization rate after the flood.
Blanchard's cricket frog is successful at inhabiting sandy or muddy shorelines due to its coloration (Smith et al. 2003), but it can also frequent more vegetated habitats (Gray et al. 2005).Bare ground was relatively uncommon before the flood and the species was successful in that environment.After the flood its ability to inhabit sandy or muddy shorelines allowed it to successfully recolonize study area wetlands.The positive correlation between bare ground and extinction rate might seem counter to the literature at first.However, after the flood many sites were virtually devoid of any vegetation or cover and Smith et al. (2003) note that Blanchard's cricket frogs use small objects for cover.It would appear that without at least a few refuges, the habitat is not suitable for Blanchard's cricket frogs.Blanchard's cricket frogs appear to be declining for unknown reasons at the northern edge of their range (Youngquist and Boone 2014) but our results suggest their current status in the study area should not be of concern.
Pre-flood gray treefrog occupancy was positively associated with distance to wetland.This result seems counterintuitive at first, but highlights the fact that suitable wetlands are not necessarily clustered on the landscape next to each other.At the scale of the study area, gray treefrogs were likely not choosing breeding sites based on nearness to other sites.Gray treefrogs had very high extinction rates, with extinction rate being low only at very small sites with very shallow shorelines.Precision was poor, however, because so few sites did not go extinct.
Northern leopard frogs had a positive association with emergent vegetation before the flood.After the flood little or no emergent vegetation was present, which likely is one reason they declined precipitously.They had very high extinction rates and low colonization rates.Though northern leopard frogs reproduced successfully the year of the flood and were observed at many sites in 2012, it appears that, in contrast to Woodhouse's toads, they found the habitat unsuitable for breeding in 2012.Northern leopard frogs need suitable non-breeding habitat in addition to suitable breeding habitat (Pope et al. 2000), which was likely not present in 2012 and present in limited amounts in 2013.
American bullfrogs occurred in larger wetlands, likely because they need permanent wetlands for their two year life cycle in Iowa (Casper and Hendricks 2005).However, they tended to go extinct at larger wetlands after the flood.This may be because some of the new large wetlands were scour holes with no vegetation and bullfrogs prefer wetlands with heavy vegetation (Casper and Hendricks 2005).Bullfrogs at the study site are on the edge of their historical range, but still often considered to have a negative impact in Iowa in anthropomorphic landscapes (Lannoo et al. 1994).In 2012, surveys suggested that bullfrog occupancy in 2012 was comparable to pre-flood rate.However, in 2013, their occupancy rate declined substantially.While vegetation increased in 2013 relative to 2012, there was still not thick stands of shoreline and emergent vegetation that bullfrogs prefer.
Chorus frogs often call from emergent vegetation (Moriarty and Lannoo 2005).Observations of chorus frog choruses in and near the study site were nearly always associated with emergent or flooded vegetation.It is likely that lack of vegetative cover was a severe limiting factor for chorus frogs.With increasing vegetation chorus frogs should return to some extent.
Absent a reoccurrence of severe flood events, emergent vegetation and shoreline vegetation can be expected to increase in our study area, thereby creating improved habitat conditions for gray treefrogs, northern leopard frogs, and chorus frogs.How long it will take them to reach occupancy rates similar to their pre-flood rates is difficult to predict, however.On the other hand, large dunes of deposited sand seem likely to remain poorly vegetated and perhaps vegetation in other areas will not be as dense, creating better habitat for Woodhouse's toads.Plains leopard frogs, Blanchard's cricket frogs, and American bullfrogs seem likely to do well under most scenarios.However, if more extreme flood events occur, gray treefrogs, northern leopard frogs, and chorus frogs may be largely extirpated from the floodplain.
Our results suggest several management actions that could be taken to mitigate the effects of another flood and encourage recovery from the 2011 flood.Firstly, sites with shallow slopes need to be available.Shallow slopes may need to be created in currently existing wetlands or new wetlands could be created with shallow slopes.Secondly, refugia sites above the maximum expected flood boundary should be created.These could be created in areas with clay soils that will hold water or near a tributary that will provide water for an adequate hydroperiod in most years.Sites created would preferably be small and temporary, with a hydroperiod from a few weeks to a few years, representing the ephemeral half of the ephemeral to permanent wetland gradient (Wellborn et al. 1996).Permanent wetlands will always likely be well represented in the floodplain, but temporary, fishless wetlands will not.Sites should be managed to encourage emergent vegetation.With adequate refugia sites, anuran species will be available to recolonize the flood zone when possible.
Call surveys provide important insight into one stage in the life of anurans, but reproduction and emergence of metamorphs may not have occurred at all sites at which males were heard calling (e.g., Greenberg and Tanner 2005).Data from additional life stages such as tadpoles and metamorphs could provide more insight into the factors driving population dynamics.
Our study reinforces the premise that extreme environmental events will differentially affect species and community structure depending on life history characteristics.Extreme environmental events are likely to increase in frequency (CCSP 2008) and amphibians may be particularly susceptible to these events (Walls et al. 2013).Identification of species that may be most at risk to such events and the actions that can be proactively taken to mitigate risk should be an important component of informed conservation

Fig. 1 .
Fig. 1.Mean river stage with 1 standard deviation for the period from October 30, 1998 to June 12, 2014 compared to the 2011 river stage for USGS streamgage 06601200 at Decatur, Nebraska, USA, near the center of the study site.Data courtesy of the USGS Iowa Water Science Center.

Fig. 2 .
Fig. 2. Mitigation sites and study area in mid-western Iowa.

Fig. 3 .
Fig. 3. Beginning points of 50-m segments for metamorph surveys during the 2011 flood year.Satellite photography shows flooding in darker shades.

Fig. 4 .
Fig. 4. Occupancy rate estimates with standard errors for six species in this study.The vertical line designates the flood year.Points have been offset slightly so standard errors are visible.The boreal chorus frog is excluded because estimates from after the flood were unreliable.

Fig. 5 .
Fig. 5. Extinction and colonization rate estimates with one SE for the intervals from 2010 to 2012 and 2012 to 2013.Confidence intervals are not given for estimates near 0 or 1.

Fig. 6 .
Fig. 6.Estimated colonization rates by slope with 95% confidence intervals for plains leopard frog, Woodhouse's toad, Blanchard's cricket frog, and estimated extinction rates by slope with 95% confidence intervals for northern leopard frog.The first column is rates from 2010 to 2012 and the second column is rates from 2012 to 2013.

Fig. 7 .
Fig. 7.Estimated extinction rates with 95% confidence intervals for Blanchard's cricket frog by the proportion of bare ground at the shoreline from 2010 to 2012 and 2012 to 2013, gray treefrog complex by slope with wetland size fixed to 0.5 ha, and American bullfrogs by wetland size.

Table 1 .
Submodels used for the four parameters of the multi-season occupancy model.Extinction rate and colonization rate had identical submodels.One or the other of shoreline bare ground and grass/herbs was used.

Table 2 .
Number of metamorphs observed and the naı ¨ve occupancy rate at 271 segments along the edge of the waterline during the flood.

Table 3 .
Cumulative weights and sign (positive or negative) of covariates with .80%cumulative weight from final model sets for three species of anurans.The parameters are occupancy rate in 2010 (w 1 ), the extinction rate (e), the colonization rate (c), and detection probability ( p).WetSize ¼ wetland size (ha), DistWet ¼ distance to nearest wetland (m), Slope ¼ average depth of the wetland 1 m from the waterline, BareGr ¼ proportion of the shoreline within 1 m of the waterline that was bare ground, AirT ¼ air temperature (8C), WatT ¼ water temperature (8C), EmerVeg ¼ proportion of shoreline with emergent vegetation within 1 m of the waterline, MS ¼ proportion of the moon illuminated, Year ¼ detection probability varied by year of surveys.Parameters with multiple covariates with .80%cumulative weight have multiple rows.

Table 4 .
Cumulative weights and sign (positive or negative) of covariates with .80%cumulative weight from final model sets for three species of anurans.The parameters are occupancy rate in 2010 (w 1 ), the extinction rate (e), the colonization rate (c), and detection probability ( p).WetSize ¼ wetland size (ha), DistWet ¼ distance to nearest wetland (m), Slope ¼ average depth of the wetland 1 m from the waterline, BareGr ¼ proportion of the shoreline within 1 m of the waterline that was bare ground, AirT ¼ air temperature (8C), WatT ¼ water temperature (8C), EmerVeg ¼ proportion of shoreline with emergent vegetation within 1 m of the waterline, MS ¼ proportion of the moon illuminated, year ¼ detection probability varied by year of surveys.Parameters with multiple covariates with .80%cumulative weight have multiple rows.