Heterogeneous Occupancy and Density Estimates of the Pathogenic Fungus Batrachochytrium dendrobatidis in Waters of North America

Biodiversity losses are occurring worldwide due to a combination of stressors. For example, by one estimate, 40% of amphibian species are vulnerable to extinction, and disease is one threat to amphibian populations. The emerging infectious disease chytridiomycosis, caused by the aquatic fungus Batrachochytrium dendrobatidis (Bd), is a contributor to amphibian declines worldwide. Bd research has focused on the dynamics of the pathogen in its amphibian hosts, with little emphasis on investigating the dynamics of free-living Bd. Therefore, we investigated patterns of Bd occupancy and density in amphibian habitats using occupancy models, powerful tools for estimating site occupancy and detection probability. Occupancy models have been used to investigate diseases where the focus was on pathogen occurrence in the host. We applied occupancy models to investigate free-living Bd in North American surface waters to determine Bd seasonality, relationships between Bd site occupancy and habitat attributes, and probability of detection from water samples as a function of the number of samples, sample volume, and water quality. We also report on the temporal patterns of Bd density from a 4-year case study of a Bd-positive wetland. We provide evidence that Bd occurs in the environment year-round. Bd exhibited temporal and spatial heterogeneity in density, but did not exhibit seasonality in occupancy. Bd was detected in all months, typically at less than 100 zoospores L−1. The highest density observed was ∼3 million zoospores L−1. We detected Bd in 47% of sites sampled, but estimated that Bd occupied 61% of sites, highlighting the importance of accounting for imperfect detection. When Bd was present, there was a 95% chance of detecting it with four samples of 600 ml of water or five samples of 60 mL. Our findings provide important baseline information to advance the study of Bd disease ecology, and advance our understanding of amphibian exposure to free-living Bd in aquatic habitats over time.

Of the over 3,000 fungal species described from aquatic habitats [37], chytrid fungi are the earliest of extant fungi to diverge phylogenetically, and now have a global distribution [38,39]. Bd is one of over 1,200 chytrid species described from freshwater, marine, and terrestrial systems occurring across temperate, tropical, and tundra environments [40][41][42]. Although chytrids function primarily as plant saprobes and parasites [37], some also parasitize animals [43,44]. Bd is one of only two chytrids known to infect vertebrate hosts [44,45].
To understand the pathology and conservation implications of Bd, a better understanding of its distribution, ecology, and the life history of Bd in the wild is needed. The life cycle of Bd includes forms that are free-living in the aquatic environment [46,47]. Bd has been detected by filtering water samples to capture free-living zoospores and zoosporangia, then performing a genetic analysis on the filtered particulates [48][49][50]. Bd has not been reliably detected from sediments [48]. Laboratory experiments have demonstrated that Bd has survived on sterilized moist sand for up to three months and remained infective in lake water for up to seven weeks [51]. Bd cultures can be maintained under laboratory conditions for several years (TC, ARB personal observation), which suggests that Bd can survive in the environment without a host as long as conditions are favorable. In laboratory settings, Bd growth and reproduction depended on temperature (4-25uC, ideal 17-23uC) and pH (4-10, 6-7 ideal) [52], and differences in both generation time and fecundity of Bd in response to different thermal regimes are observed in at least one Bd strain [53].
A greater understanding of Bd distribution and spread is of paramount importance to more fully understand its threat to amphibians [36]. Although it has an impact on the persistence of selected amphibian populations around the world [21,35,54], the full scope of the effects of chytridiomycosis on global amphibian population declines is not well understood. Most Bd research efforts have focused on Bd in amphibian hosts per se, with little focus on the ecology of free-living Bd outside of the amphibian host [55][56][57]. In amphibian hosts, Bd exhibits sensitivity to a number of environmental variables including temperature [36,[58][59][60][61]and elevation [59,62], suggesting a tendency to associate highelevation areas or regions with cool temperatures with increased risk for Bd-related declines and extinctions [63]. From a large global data set of amphibians sampled for Bd infections, it was found that Bd occurrences were inversely correlated with the temperature range at sites [36]. Seasonal patterns of amphibians infected with Bd have emerged in several studies [58,[64][65][66][67], further implicating microclimatic associations and potential differences between strains [68]. In wild amphibian populations from temperate areas, Bd infections appear to follow predictable patterns, with the highest prevalence and intensity in the cooler spring months, and decreasing prevalence in the warmer summer and autumn months, sometimes to non-detectable limits [66,67]. We describe spatial and temporal patterns in occupancy and density of free-living Bd in aquatic habitats of the United States, outside of the amphibian host, to better understand the geographic distribution and ecology of this pathogen, and to further inform pathogen dynamics research.
Occupancy modeling is a powerful tool to estimate site occupancy and detection probability when detection is imperfect [69]. Occupancy models have been used to investigate disease ecology in amphibians, fishes, and birds, where the focus was on pathogen occurrence in the host [70][71][72]. We used an occupancy modeling approach to investigate free-living Bd in North American surface waters, from water samples collected across several sites in North America (Figure 1) to assess: 1) the seasonality, if any, of Bd detection; 2) the relationships between Bd site occupancy and habitat attributes including water quality (temperature, pH, specific conductance), geography (latitude, longitude, elevation), climate metrics (minimum/maximum/range in precipitation and temperature, dew point), and time (Julian day); and 3) the probability of Bd detection from water samples as a function of the number of samples taken, sample volume, and water quality (temperature, pH, specific conductance). We predicted that free-living Bd occupancy and density would follow the same trends as observed in Bd-infected amphibians, with the highest detections in the North American spring season and the lowest in the late summer/autumn [64,66,71]. Based on reported Bd-amphibian host relationships and laboratory studies of Bd, we expected to see corresponding temperature and pH associations in our analyses. We also predicted that detection probability would increase as the number of samples and volume of water per sample increased, and would increase as temperature and pH approached the ideal range for Bd growth [52]. Finally, we report on the temporal patterns of Bd density from a four-year case study of a Bd-positive wetland, and the relationship between water quality and Bd density.
Our findings address important baseline information to advance the study of Bd disease ecology in temperate-zone systems, and specifically to advance our understanding of the likelihood of amphibian exposure to free-living Bd in aquatic habitats over time. Our results are applicable to Bd risk assessment, sampling protocols, policy development, and regulatory decision-making processes.

Ethics Statement
Field studies involved collecting water samples. No animals were sampled as part of this study. Sampling did not involve endangered or protected species, although they may have occurred at or near the sites sampled. Permission to access National Park Service lands was obtained by Bill Commins, NPS Research Permit Coordinator, Washington, DC (bill_commins@ nps.gov). Private lands were accessed with landowner permission; TC may be contacted for future permissions. Public right-of-ways did not require permission/permits to sample. Given the sensitive nature of some sites, GPS coordinates are available upon request.

Field Methods
All Sites. We collected Bd water samples from 41 amphibian breeding sites along a latitudinal gradient spanning the North American continent between 45 and 48 degrees North, and a longitudinal gradient focused on, but not exclusive to, the western United States west of 105 degrees West ( Figure 1). Site selection occurred opportunistically at locations where on-going amphibian research was being conducted; hence our scope of inference is restricted to the sampled sites. Sites were located in temperate ecoregions in a variety of land-use settings ranging from urban/ suburban to designated wilderness, and at elevations ranging from near sea level to ,3500 m-39% were low-elevation, 54% were mid-elevation, and 7% were high-elevation. Three replicate water samples were collected, separated by .50 m, during a single survey occasion that occurred between May and September, 2007 to 2010. Most sites (63%) were sampled in June and July. We measured water quality (i.e., temperature, pH, specific conductance) at 22 of 41 sites, herein WQ sites ( Figure 1).
Case Study Site (Beaverton, Oregon). We also conducted a four-year case study in Beaverton, Oregon (45.3108, 2122.4752, Datum: WGS84), which we monitored monthly between 2007 and 2011 to assess temporal variation in Bd occurrence and density. The site was a palustrine wetland in a suburban setting, dominated by reed canary grass (Phalaris arundinacea). It was approximately 1.2 hectares in area, bisected by a road, and connected to larger wetland complexes 6.4 hectares in area associated with Johnson Creek, both upstream and downstream of the site. Surrounding land use was historically agriculture (row crops and horse pasture), and is now primarily residential. In the 1990s the site was converted to a wetland mitigation site and has undergone recent efforts to reestablish a forested wetland.
Water samples were collected using the protocol described in [48], with modifications to improve Bd DNA recovery and sampling efficiency, which are described below. At each site, three spatial replicate water samples were collected for Bd testing, also at least 50 m apart in the pond or wetland. Each sample was collected from shallow water, between 5 and 20 cm below the water surface. A general overview of sample collection is as follows: using a sterile 60-mL syringe rinsed 3 times with native water, we filtered water drawn from below the surface through a Sterivex 0.22-mm capsule filter (luer-loc with male end) until the filter was nearly clogged. The volume was recorded (range: between 20 mL and 2.4 L, mean 350 mL) and the filter was flushed using a new sterile syringe with 50 mL of DNAase/RNAse-free 0.01 Mol Phosphate Buffered Saline (PBS) to remove excess dissolved organic carbon which can inhibit the PCR reaction. The filter was purged of water by removing it from the syringe, drawing air into the syringe, reattaching the filter and pushing air through the filter and repeating if necessary. The outflow end of the capsule filter was sealed with Hemato-Seal clay (we found that other types dried and cracked in the field, compromising the seal), then we injected the filter with 0.9 mL cell lysis buffer solution, and sealed it with a luer-loc cap. Each sample was placed in an individual Whirl-pak bag and kept cool. Upon returning from the field, samples were refrigerated until DNA extraction occurred, within three months of sample collection. Water quality was measured with a single YSI multiparameter data sonde, calibrated before each sampling event using standardized U.S. Geological Survey (USGS) protocols [73]. The YSI was cleaned between each use.
The method we used to collect Bd samples involves water filtration and molecular techniques but differs from what is usually termed environmental DNA (eDNA) sampling [74,75]. Typical eDNA methods sample the environment for DNA from parts of an organism (e.g., shed skin, tissue fragments), whereas the method we used is a standard method for characterizing microbial communities, where whole microbes are captured on the filter and dissolved DNA is passed through the filter. This ensured we were sampling only Bd that occupied the site at the time the sample was collected. However, it is possible that if we filtered recently shed amphibian skin infected with Bd, it may have been captured by our water sampling process.

Laboratory Methods
DNA was extracted from the cell lysis solution using the Gentra Puregene Tissue Kit [48]. We added proteinase K (0.1 mg mL 21 ) directly to the filter by opening the luer-loc cap, injecting the solution and resealing the capsule. We placed each filter in a sterile 50-mL Falcon tube, and incubated it at 55uC for 60 min in a continuously rotating rack inside the incubator to ensure the filter was completely bathed in the lysis/proteinase K solution. Following this step, we drew the solution out of the filter with a sterile syringe, and continued the extraction following the manufacturer's protocol. Quantitative PCR assays were conducted using the Qiagen QuantiTect SYBR Green PCR kit following standard protocols [48]. To quantify the Bd zoospore equivalents, we calculated a conversion factor from a dilution series of known zoospore numbers (i.e., 169), which is within the range of 60 to 220 for rRNA fungal gene locus copy numbers [76]. Bd genomic equivalents are therefore 169 times greater than the numbers we report. A site was considered Bd-positive if at least one of three replicate filters from a site returned a positive qPCR result.

Statistical Methods
Occupancy models require multiple observations per site. We treated each water sample as an independent observation, for a total of 3 observations per site. We used a single-season occupancy model to simultaneously estimate the site-level probability of Bd occurrence in amphibian habitats (Y) while accounting for the observation-level probability of detecting Bd in amphibian habitats when it is present (p), and relating both parameters to our hypothesized covariates. We investigated eight covariates that we hypothesized were related to the probability of occurrence (Y) and five covariates we hypothesized were related to detection probability (p) in amphibian habitats, and built models for all possible combinations of the a priori covariates (Table 1). Temporal (day of year) and geographic (latitude, longitude, and elevation) predictors of occupancy were recorded at the time of sampling. Other environmental predictors of occupancy (precipitation, minimum temperature, maximum temperature, temperature range) were derived from PRISM (Parameter-elevation Regressions on Independent Slopes Model) climate mapping system data at a spatial resolution of 1 km 2 [77]. Two covariates of p were recorded from all sites (p as a function of volume, and p as constant across surveys), and three covariates of water quality (temperature, pH, specific conductance) were recorded from the subset of 22 WQ sites. Continuous covariates were standardized to mean = 0 and standard deviation = 1. We evaluated the degree of support for each competing model based on Akaike's Information Criteria (AIC) and the resulting Akaike's weights. We considered that models within 2 delta AIC received similar degree of support.
Using the maximum likelihood estimates from the top-performing model for all 41 sites, we also estimated the detection probability given the number of samples collected and the volume of the water filtered per sample. In the subset of WQ sites, we further investigated three water quality covariates (temperature, pH, and specific conductance) that potentially related to the probability of Bd detection, using the top-performing detection probability models ( Table 2). In the Oregon case study, we examined associations between Bd detection and four water quality metrics (temperature, pH, specific conductance, and turbidity) using Pearson's correlation coefficient. To include this site in our occupancy analysis, we randomly selected one sampling event that occurred within the sampling period of the other sites, (between May and September, and between the years 2007 and 2010). Data were summarized using the base packages available in R 2.14.0 [78].

Results
Bd exhibited temporal and spatial heterogeneity in detection and density in amphibian habitats. In both our landscape-scale study of North America and the Beaverton, Oregon case study we detected Bd in all months surveyed, but not in all samples. Bd density was highest in spring (Figures 2 & 3), and at two Nevada sites sampled in May, Bd densities were more than 50 times the densities that we observed at other sites (2,109 and 4,924 zoospores L 21 ). At all sites, mean zoospore densities were highest in April and August (Figure 2, overall mean: 40 zoospores L 21 ; median: 15 zoospores L 21 ; range: 0-4,924 zoospores L 21 .). In some cases, we observed marked heterogeneity between the individual samples, e.g., at the two Nevada sites, one sample recovered 2,109 zoospores L 21 from the wetland and the other two filters recovered an order of magnitude less, 192 and 227 zoospores L 21 . At the site where 4,924 zoospores L 21 were recovered, Bd was not detected from the other two filter samples. Our Oregon case study had a higher mean density than observed in our landscape-scale study (i.e., 100 zoospores L 21 ) and the highest Bd density observed from a field collected sample, 3,159,289 zoospores L 21 (Figure 3). There was a 95% chance of detecting Bd at a site when it was present with 4 samples of 600 mL of water or 5 samples of 60 mL of water (Figure 4), estimated by calculating 1-(1-p) ' #samples.

All Sites
We detected Bd in 19 of 41 (47%) of sites sampled, however our occupancy estimate was much higher, suggesting 61% of amphibian habitats were occupied by Bd (SE = 11%, CI = 39%-80%), underestimating Bd in our study by 14%. Models with elevation were the best fit to predict Bd occupancy (Table 2); as elevation decreased, Bd occupancy increased ( Figure 5). At 100 m  Amphibian Chytrid Fungus Occupancy of Wetlands and Density Estimates PLOS ONE | www.plosone.org elevation, the probability of detecting Bd when it was present was between 92-99%. At 500 m elevation, the probability of detection decreased slightly to 81-96%. However, at higher elevations, detection probability decreased sharply, 36-59% at 1500 m, and only 10-20% at 2500 m. The best-supported model structure for parameter 'p' was p(pH); the probability of detecting Bd, when present, increased as pH increased, however differences in detection probability were negligible when the pH range was considered ( Figure 6). Observed pH ranged from 6.5 to 10.4 (mean = 7.8, median = 7.6). Bd was detected in samples with a pH between 6.5 and 8.5 (mean = 7.1, median = 7.1, n = 14 samples from 8 sites). Because there was a negligible difference between the p(ph) model and the p(.) model (DAIC = 1.1), we proceeded with the most parsimonious model for estimates of Y (Table 2).

Case Study Site (Beaverton, Oregon)
Across all years, we detected Bd at the Oregon site at least once per month, although not necessarily in every month in a given year (e.g., Bd was detected in December during 2 of 4 years sampled). As predicted, Bd zoospore density was highest in the spring, March to June, and contrary to our landscape-scale analysis, lowest in summer (Figure 3). Mean Bd density was similar in March and October, and an unexpected peak was observed in December in 2 of 4 years (2007 and 2008). From one sampling event in March 2011, we recovered 3,159,289 zoospores L 21 from one filter, approximately four orders of magnitude larger than any previous observation in our study (e.g., Nevada sites mentioned above). The Bd densities from the other two filters from this March 2011 sampling event were 0 and 4 zoospores L 21 . Bd detection and density at the Oregon site were not correlated with any of the water quality variables that we measured (temperature R 2 = 0.01, pH R 2 = 0.02, specific conductance R 2 = 0.02 and turbidity R 2 = 0.02).

Discussion
Bd is emerging as an extremely challenging pathogen to fully comprehend, and our baseline information of the temporal and spatial patterns of Bd occurrence in USA waters, free-living, apart from amphibian hosts, adds a new dimension to understanding the dynamics of the pathogen. Host-pathogen systems are more complex when a pathogen can infect multiple host species rather than a single species [79]. Unlike most chytrid parasites, Bd is a generalist, infecting at least 42% of the species of amphibians examined [36]. Globally, Bd occurrence in amphibians increased as species richness increased, although in North America when species richness was ,10, as was the case in our study area, the odds of Bd detection were constant [36]. Bd detection probability in water also increased as amphibian density at a site increased [80]. This may be due to an increase in the number of competent hosts, direct skin contact with infected individuals, or exposure to free-living Bd. Also, Bd-positive amphibians arriving at breeding sites from upland overwintering sites may release Bd into the aquatic environment, causing an increase in Bd density in the water. Alternatively (or concurrently), spring-season conditions (e.g., increasing temperature and light) and amphibian arrival at a breeding site may stimulate free-living Bd to grow in the aquatic environment. Adding to the complexity is context-dependency, where different species, life stages, and populations of amphibians may respond to a pathogen such as Bd differently, depending on   biotic and abiotic factors acting in concert with each other, including strain differences and variation in organismal responses between sites [81][82][83].
Bd did not exhibit seasonality in occupancy (e.g., presence) in our study. We detected Bd in all months, which is evidence that Bd persists in the environment year-round in sites in the U.S. In amphibians within the geographic scope of our study, Bd has been estimated to occur in 53% of amphibian populations sampled, and in 66% of non-native bullfrog (Lithobates catesbeianus) populations sampled [71], estimates that are similar to our occupancy estimates for Bd in amphibian habitats (61%). Areas for future research include analyses of: (i) the substrates upon which Bd lives in the environment in the cold season when amphibian densities at the ponds are low or zero; (ii) spatial heterogeneity within a wetland or pond; (iii) alternative hosts, vectors, and reservoirs in different seasons; (iv) Bd strain diversity, especially in areas with introduced species; and (v) pathogenicity, or infectious potential, of Bd when amphibians are absent for months or years.
The occurrence of a species and its abundance are influenced by habitat factors at different temporal and spatial scales, from microhabitat to landscape [84][85][86][87][88][89]. In temperate areas, there is seasonal variation in Bd infection prevalence (i.e., the proportion of a population infected) in amphibian populations, with the highest prevalence in the early spring and lowest in the autumn [66,90]. Our study calls attention to the heterogeneity that may exist within a site and across the landscape over space and time. We observed a clear signal of increased Bd density in the spring. Our results were consistent with our prediction that Bd density in water would be high in the spring, and contrary to our prediction that Bd density would be low in the summer. The effect of season may be driven by various factors; the temperatures during the spring months are within the range for optimal Bd growth [52], and Bd-infected animals are returning to breeding sites, shedding zoospores into the water and transmitting the infection to each other as they congregate. Alternatively or additionally, springtime increases in host abundance may elicit a chemotactic response from free-living Bd [91], which could contribute to an observed increase in the zoospore density in water.
Adding complexity to the effort of understanding Bd's temporal dynamics in prevalence, alternative Bd hosts and reservoirs need to be ascertained, and their potential interactive role in amphibian-Bd systems is only beginning to be explored. Several documented and suggested alternative Bd hosts and reservoirs (e.g., crayfish [92]; nematodes [93]; birds [94,95]) occur within the same water bodies as amphibians, and some potential host/ reservoir species (e.g., zooplankton) consume chytrid spores, including Bd [43,96], although their post-consumption fate is unknown. When consumed, up to 50% of parasitic fungal spores may survive passage through the gut, which may further fuel disease epidemics [97]. However, there is experimental evidence that consumption by zooplankton can negatively affect free-living Bd such that infections in tadpoles are reduced [98], which suggests the aquatic-community context is an important consideration with regard to disease dynamics. Furthermore, Bd can survive in pond water for months without a host [51], and occurrence in benthic sediments appears to be likely, although detection is inhibited, perhaps due to high levels of organic carbon inherent in sediment samples from these habitats [48]. These combined factors make interpreting the Bd-amphibian disease system in a field-based setting particularly challenging, and highlight the need for a more focused effort to study Bd life history in the aquatic environment.
In our study, Bd exhibited seasonality in density, but the pattern depended on the spatial scale considered (North American landscape-scale or single-site case study). At the landscape level, the mean densities observed did not follow all of the patterns we predicted. While we did observe high densities in spring months, we also observed high densities in at least one of the summer months, August. This may have been a limitation of our study design; if sites had been visited more than once, we could have made comparisons of relative densities between visits. At our Oregon case-study site, we observed the predicted temporal patterns, with exceptions. We repeatedly observed similar mean densities in late spring and late autumn (March and October), as well as a secondary peak in winter (December), when we expected Bd density to be low. There were no major storm or flood events in the days prior to our sampling that would have flushed or resuspended Bd or potential physical reservoirs such as sediments into the wetland. Our observations lead to more questions about the role of potential secondary host species and reservoirs in regulating Bd densities in amphibian habitats. Crayfish are identified as a potential secondary host [92], but we did not observe them at the Oregon case-study site in our four-year monitoring period. Although it was a single observation, our sample of over 3 million zoospore equivalents from a liter of water is notable because it was several orders of magnitude higher than any other sample collected (the sample was run multiple times to ensure no lab errors). The sample was collected in March 2011, during the predicted spring peak, and it is possible that we captured a large free-floating colony of zoosporangia. Such colonies might be expected to be patchy in distribution, and our sample may have haphazardly collected such a localized highdensity cluster. This observation strongly suggests that Bd zoospores can have a patchy/aggregated distribution and highlights the importance of collecting several spatial replicates from a pond or wetland during water sampling. The observed temporal variation in our study is intriguing, and suggests an interplay of abiotic and biotic factors, highlighting the relevance of finer-scale assessments. Our results demonstrate that we can correct for this clustering and reliably detect Bd by adjusting the number of samples and volume of water collected. Investigations of the patterns and processes influencing Bd occupancy in amphibian habitats may be most appropriate at an ecoregional or local scale [21,36]. Factors influencing detection probability may be sitespecific and context-dependent [66,71]. Importantly, such studies can be informed by both landscape approaches (like ours), and insights from laboratory experiments (e.g., investigating primary and secondary hosts, reservoir species, and amphibian species and life stage) that can shed light on local, regional, and global amphibian declines [83,92,98].

Elevation as a predictor of occupancy
The fact that Bd occupancy within our study area was highest at low-elevation sites suggests that distribution patterns and geographic range may be influenced by human-induced landscape change, or differences in host-species composition. The probability of detecting Bd when it is present increases as the Human Footprint, an index human activities, increases (i.e., human population size, density of secondary roads, proportion of agricultural lands) [71]. This generalized index doesn't identify specific human activities, but provides a direction for future studies of ecophysiology, or the links between environmental stress, endocrine-immune interactions, and disease [99]. Bullfrogs are a common non-native species implicated as disease vectors throughout our study area at lower elevations that are absent from higher elevations, although they demonstrate differential responses to North American versus Globally Distributed Bd strains [68]. The proportion of sites occupied by Bd is higher when non-native bullfrogs are present (66%, versus 53% when bullfrogs were absent), which may explain why we observed higher occupancy as elevation decreased, though bullfrog presence alone does not explain widespread Bd occupancy of amphibian habitats [71,100].
Although temperature was not a predictor of Bd occupancy, the temperature at high elevations often has a greater range than at lower elevations. Temperature influences both generation time and fecundity of Bd, and differences in long-term responses to different thermal regimes are observed in at least one Bd strain [53,57]. Cold-adapted Bd develops faster than warm-adapted Bd, however, overall developmental time is longer; it may take 4 times longer to encyst, and take 6 times longer to mature and release zoospores [53]. These factors may explain differences in occupancy between our high-and low-elevation sites, and they could influence detection probability if Bd density is lower as a result of decreased generation time or fecundity.
At a global scale, there are family-level differences in the odds of Bd infection related to elevation. Elevation was found to be (i) positively correlated with Bd detection/infection in toads (Bufonidae); (ii) negatively correlated with Bd detection in true frogs (Ranidae); and (iii) not a predictor for Bd detection in treefrogs (Hylidae) [36]. However, at the North American scale, elevation was not a predictor for Bd detection, rather species richness, temperature, and biome were [36]. This lends further support that regional and local spatial scales may be more appropriate levels of investigation to assess risk to amphibian populations and species. Species representing these three amphibian families occur throughout our study area, and the effect of the amphibian community assemblage on Bd occupancy of their breeding habitats warrants further investigation.

Water quality as a predictor of detection probability
Our study revealed a potential relationship between pH and Bd detection probability. Other water quality covariates that we had predicted could be related to Bd occupancy and detection showed no relationships, perhaps because of the scale of our measurements. Specific conductance (conductivity in microSiemens, mS cm 21 , standardized to 25uC) and turbidity are coarse measurements that indicate the relative amount of dissolved ions and optical properties of the water but do not identify the specific constituents in the water that may be important to Bd occupancy and detection (e.g., Ca +2 , Mg +2 , Na +1 , K +1 , Cl 21 , nitrogen and phosphorus species, organic carbon, turbidity source, sediment loading, phytoplankton community). The observed temperature and pH ranges at our sites were always within the range that Bd grows in the laboratory, and both metrics were often within the range for optimal growth. Our results did not support our prediction that detection probability would be highest when pH was within the range for ideal Bd growth [52]. Although the best models included pH and suggested detection probability increased as pH increased, the differences in detection probability were negligible when the pH range was considered ( Figure 6). While there may be a narrow pH range that is optimal for Bd growth (pH 6-7) in a laboratory setting [52], this narrow range may not be ecologically relevant given the fluctuations in pH at different time scales, such as diel, episodic, and seasonal [101,102]. pH is influenced by abiotic (e.g., acid-neutralizing capacity) and biotic (e.g., organic carbon, aquatic plant community) characteristics of a system [103][104][105], and it is not clear if pH itself is a mechanism influencing Bd detection probability within our study area or if pH is instead reflective of other processes of the aquatic system that are important to Bd ecology. Lower pH can inhibit microbial metabolism [106], which could decrease detection probability. Changes in pH are related to the acid-neutralizing capacity in a system, which is strongly tied to the amount of organic carbon present [104,105], which is in turn an important nutrient for aquatic fungi [107]. Higher pH may indicate higher organic carbon [105], which could increase detection probability. Generally, chytrids decompose organic particulate matter [107], and it is not known if Bd uses non-animal organic carbon sources. Diel variation of up to pH 1.8 is not uncommon in mesotrophic and eutrophic aquatic systems, which may be driven by the plant communities present [101,102]. pH increases with the addition of algae and vascular plants [103]. If Bd is associating with plants, perhaps as a secondary host, pH as the predictor of detection probability would be indirect. This result provides direction for future research into how pH influences Bd site occupancy and detection probability.

Management Implications
Our results show that water filtration sampling is a reliable way to assess the occurrence of Bd in amphibian habitats when 4-5 samples of a small volume, i.e., 60-600 mL of water, are filtered directly from ponds and wetlands, especially at sites less than 1500 m elevation. Bd densities recovered were consistently above the detection limit of the qPCR for a quantitative result (10 genomic equivalents), which affirms water filtration as a viable method for both Bd detection and quantification in amphibian habitats. Our results are context-dependent, as found in other studies from the southwestern United States where more samples were necessary to achieve 95% confidence that Bd was detected when it was present [80]. Our results suggest that a single visit collecting multiple samples in the spring or summer months may be sufficient (and perhaps more effective) for Bd detection from water samples, rather than multiple visits collecting one sample throughout the season, however spatial replicates within the same site visit are important [50]. The number of samples required to detect Bd is representative of our area of inference and is not intended to serve as a recommendation for a standard protocol, as regional and site-level variation in Bd occupancy may require fewer or additional samples [80].
There are few, if any systems, where species detection is perfect [69]. Imperfect detection (false negatives) may occur because a species is rare or cryptic [108,109], or alternatively, a false positive report or misidentification may occur [110][111][112]. We underestimated Bd occupancy of amphibian habitats within our study area by 14%, which is greater than other observations from a different geographic region that used the same filtration method [80]. These results stress the importance of using methods that account for imperfect detection when evaluating the potential occurrence and distribution of organisms when detection probability is less than 1. The probability of detecting Bd in amphibians varies based on life stage, and is highest in adults at low elevations [71]. In temperate areas, lentic-breeding amphibians are often explosive breeders [113,114], and the likelihood of encountering many adult amphibians is limited to the narrow window of time (hours to days) they are present at a site. While larval amphibians are present at a site for longer time periods, the probability of detecting Bd when it is present in them is less than 20% [71]. If the objective of a study is to determine the Bd status of a site, sampling the habitat rather than the animals may be the most cost-effective approach in terms of time and resources. Sampling the environment also resolves problems with detection due to differences in capture probability between infected and uninfected individuals related to behavioral changes, however the results are limited to Bd occupancy of the aquatic habitat sampled, and inferences to the disease status of an amphibian population, or potential secondary hosts or vectors, cannot be made except by sampling them directly.
By focusing research efforts on understanding the ecology of this fungal pathogen outside of the host, conservation efforts can be more informed and focused to meet the management goals and objectives for species at-risk and common species alike. Patterns in amphibian response to Bd are different based on biome [36]. The strategy for assessing Bd status at a site may be different according to the ecoregion, and may include sampling the habitat, other species that are present, as well as the amphibians that occur at a site. When water sampling is coupled with amphibian sampling, scientists can begin to understand the relationships between Bd occupancy and density in the environment, and the occurrence of disease in amphibian populations.