A Comparison of Passive and Active Acoustic Sampling for a Bat Community Impacted by White-Nose Syndrome

White-nose syndrome (WNS) is a recently described fungal disease (Pseudogymnoascus destructans) of cave-dwelling bats (Blehert et al. 2009; Minnis and Lindner 2013). Since 2006, WNS has caused the deaths of >5.5 million bats (USFWS 2012) and has rapidly spread from its first appearance in New York to ≥25 states and 5 Canadian provinces. In the context of severe population declines after the appearance of WNS, the use of acoustic monitoring to track changes in population sizes and species assemblages during the summer maternity season has been shown to be effective (Ford et al. 2011). Moreover, as WNS continues to spread and bat populations decline further, biologists will likely need to rely more on acoustic detection as the primary method of monitoring bat presence on the landscape as a matter of practicality because of greater cost and effort required for mist-netting studies of declining species.

Traditional monitoring protocols typically rely on capture methods (netting) at roosts, water sources, or along flyways. Capture techniques are only able to sample an extremely small portion of the area used by free-flying bats, typically missing high-flying or alert bats (Murray et al. 1999; O'Farrell and Gannon 1999; MacCarthy et al. 2006), and thereby resulting in biased samples and incomplete documentation of species assemblages. Individuals and species of bats are not equally vulnerable to capture on a given night (O'Farrell et al. 1999), and the number of bats that are now susceptible to physical capture on a given night is even less because of severe declines. Acoustic monitoring is a noninvasive sampling technique to quantify activity of echolocating bats in the environment. These methods have become routine for investigating bat ecology, species assemblages, and relative abundance on landscapes and relative to land management (Britzke et al. 2002; Johnson et al. 2002, 2011; Milne et al. 2004; Ford et al. 2011). For example, O'Farrell and Gannon (1999) detected 86.9% of the total possible bat species assemblage for the Southwest with acoustic detectors versus 63.5% by netting only. In Missouri, Murray et al. (1999) concluded that acoustic detectors accounted for greater bat species richness with less effort over time than did mist-netting. Although research has suggested that using capture methodologies in combination with acoustic detectors is ideal (Patriquin et al. 2003; Flaquer et al. 2007; Robbins et al. 2008), in WNS-impacted areas, the high levels of sampling effort to detect declining bat species using netting currently is logistically and fiscally prohibitive (Coleman 2013).

Acoustic monitoring methodologies are broadly categorized as either active or passive sampling. Additionally, mobile driving transects are increasingly incorporated, but these methods are restricted to areas that can safely support 30-mile (48.3-km) transects at 20 miles/h (32.2 km/h; Britzke and Herzog 2009). Active sampling refers to surveying where the surveyor is present and able to change the orientation of the microphone relative to bats for optimization of echolocation pass recording (Britzke 2002). Active sampling typically records high-quality echolocation passes because contact can be maintained between microphone directionality and bats. However, active sampling is limited by the availability of the observer(s) because a user must remain present for the duration of the survey. Consequently, standard active sampling protocols typically terminate sampling by 0200 hours, after which bat activity decreases (Johnson et al. 2002), whereas standard passive sampling protocols suggest that sampling continue throughout the entire night (USFWS 2014). Passive sampling refers to automatic recording of passes without a surveyor present, with the microphone fixed in a predetermined direction. Passive sampling designs often include multiple sampling stations allowing sampling over wide spatial scales. This method typically records lower quality passes than does active sampling because of the fixed directionality of the microphone, but it allows for longer data collection periods at a lower time expense to observers than does active sampling. Both techniques have been widely tested in the field (Johnson et al. 2002; Britzke 2003; Milne et al. 2004; Brooks and Ford 2005; Britzke et al. 2011), but no direct comparisons of standard active and passive sampling protocols have been made in the context of WNS-impacted landscapes.

Nine species of bat occur in northern New York in the vicinity of Fort Drum Military Installation (Fort Drum). These include the following species impacted by WNS: the big brown bat Eptesicus fuscus, the federally endangered Indiana bat Myotis sodalis, the eastern small-footed bat Myotis leibii, the little brown bat Myotis lucifugus, the proposed-for-listing northern bat Myotis septentrionalis, and the tri-colored bat Perimyotis subflavus. The silver-haired bat Lasionycteris noctivagans, the hoary bat Lasiurus cinereus, and the eastern red bat Lasiurus borealis occur in the study area, but are not thought to be impacted by WNS. Previous analyses of passive acoustic activity patterns at Fort Drum showed significant declines in overall summer foraging activity in little brown bats, northern bats, and Indiana bats, as well as a decline in foraging activity from early to late summer season in post-WNS years—indicating a probable decrease in reproductive success of surviving individuals and subsequent lack of juvenile recruitment—in little brown bats (Ford et al. 2011).

As WNS and the associated population declines continue to spread, monitoring of impacted species will become increasingly difficult in terms of required effort, yet exceedingly important for management and regulatory purposes. An important step in developing a monitoring protocol for bat species in a WNS-impacted area is determining whether active or passive sampling is the most efficient acoustical sampling technique, in terms of logistical costs and detectability of declining species. Although implementation of capture techniques remains effective for detecting declining species in regions that have not yet been impacted by WNS, these techniques are now time- and cost-prohibitive in WNS-impacted areas. Furthermore, mobile transects cannot be properly implemented in areas where 30-mile (48.3-km) transects at 20 miles/h (32.2 km/h) cannot be performed, such as military training areas with strict speed-limit enforcement like Fort Drum, mountainous regions with sinuous and steep roadways, or urban areas with high speed limits and heavy traffic. The objective of our study was, therefore, to directly compare universally implementable, temporally distinct passive and active acoustic sampling designs to maximize detectability of bat species at Fort Drum.

We conducted our study at Fort Drum Military Installation (Fort Drum). Fort Drum covers approximately 43,000 ha in Jefferson and Lewis counties in northern New York (44°00′N, 75°49′W). The installation lies at the intersection of the St. Lawrence–Great Lakes Lowlands, the foothills of the Adirondack Mountains, and the Tug Hill Plateau ecoregions within the Black River and Indian River drainages. Approximately 57% of the landscape is made up of forested habitat dominated by northern hardwood types. Wetland systems such as wet meadows and beaver- Castor canadensis impacted streams and ponds make up 20% of the landscape. The nearby Niagara Escarpment (10–15 km west of Fort Drum) contains karst formations and caves (Fenton 1966). Development is concentrated in the Cantonment area, with the remainder of the installation consisting of 18 training areas, an airfield, and a large, centralized main impact zone that are all largely undeveloped. White-nose syndrome was first detected at Fort Drum in 2008.

During the summers of 2011 (7 Jul–10 Aug) and 2012 (2 Jul–8 Aug), we collected acoustic data at 31 wetland and forested corridor sites across the installation to compare passive to active sampling. We used Anabat II frequency-division detectors¹ connected to a compact flash-storage Zero-Crossings Analysis Interface Module, as well as the SD1 and SD2 units (Titley Electronics, Ballina, New South Wales, Australia) to record bat echolocation passes. We calibrated all units using an ultrasonic insect-deterring device following the methods of Larson and Hayes (2000) prior to use in the field. Because of increased military training in 2012, training sites sampled in 2011 could not be replicated during 2012; therefore, we chose additional sites in the Cantonment area to compare active and passive sampling in the latter year.

Passive and active sampling methods were not always conducted simultaneously (i.e., on the same night at a given site) because of logistical constraints. For passive sampling, we placed Anabat units in weatherproof boxes with polyvinyl chloride tubes containing a small weep hole for water drainage (O'Farrell 1998). Boxes were placed on 1.5-m tripods aligned in a manner that allowed sound to enter the tubes at a 45° reflective angle to be received by Anabat transducers perpendicularly (refer to diagrams in Britzke et al. 2010; Coleman 2013). We programmed Anabats to record data continuously from approximately 1900 to 0700 hours for two to three consecutive sampling nights per site. This practice is in accordance with standard passive sampling recommendations by the United States Fish and Wildlife Service Range-wide Indiana Bat Summer Survey Guidelines (USFWS 2014). Typically, we sampled passively at three sites simultaneously. For active sampling, we hand-held Anabat detectors for 30 min, pointing and sweeping the microphone toward areas of expected activity (Ford et al. 2005) in order to mimic commonly implemented active sampling methods described by O'Farrell et al. (1999) and Johnson et al. (2002). We sampled multiple sites in a random order each night beginning at sunset and ending no later than 0200 hours to correspond with peak hours of bat activity. We sampled three to four sites on a given night, and sampled each site using active methods two to three times throughout a season to account for between-night variation. We conducted simultaneous active and passive sampling (both methods on the same night at a given site) on 10 and 11 occasions in 2011 and 2012, respectively. Surveys in both years were conducted in July and August, so effects of pre- versus postvolancy of young were not anticipated between years. We avoided surveys using either sampling method during periods of continuous rain, rapid wind, or low temperatures (<10°C).

We downloaded data nightly onto a laptop computer using the CFCread program (Titley Electronics, Ballina, New South Wales, Australia). We used EchoClass Version 1.1 (U.S. Army Engineer Research and Development Center, Vicksburg, MS, USA) to clean and classify the echolocation call data. Given that accuracy rates using automated software to identify specific species are <100% (Britzke et al. 2011), we considered a species present at a site when the maximum likelihood value for identification was ≥90%. We created nightly presence–absence detection histories from the acoustic data (Gorresen et al. 2008). We considered each nightly survey independent because of the separation of sites and break in sampling during daylight hours. For each species, we attempted to fit single-season, single-species occupancy models to determine constant detection probability and unbiased occupancy estimates (MacKenzie et al. 2002) over each sampling type and year using Program PRESENCE (Version 2.4; Hines and Mackenzie 2008). Typically, it is preferred that a candidate set of models are attempted and that models are ranked using Akaike's Information Criterion and compared using Akaike weights (Burnham and Anderson 2002). However, because we were only interested in constant, mean detection and occupancy estimates and did not measure for potential site or sampling covariates, we chose to examine a single model for each species at each combination of sampling type and year. Rather than conducting a single, multiple-season model for each species using sampling covariates, we separated models because some sites were not replicated between years. We examined four models for each bat species: 2011 passive, 2011 active, 2012 passive, 2012 active.

In addition to determining detection probability and occupancy estimates, we determined the relative activity for each species for both sampling types and years, standardized as echolocation passes per hour. We defined a “pass” as an acoustic file that contained three or more bat sound pulses. We performed a paired t-test to discern differences in recorded relative foraging activity between active and passive sampling when sampling occurred in the same night using SAS statistical software (Version 9.2; SAS Institute 2012, Cary, NC, USA) and defined statistical significance at α  =  0.05. Lastly, for comparisons, we calculated the relative equipment, labor, and logistic costs for both acoustical techniques as well as for a comparable contracted mist-netting effort.

In 2011, we recorded 903 and 2,771 bat echolocation passes using active and passive sampling, respectively, at 18 sites located in training areas of the installation. Approximately 52% of calls were identified by EchoClass to the species level using active sampling, and approximately 53% of calls were identified using passive sampling. All other calls were considered low quality and identified as unknown. In 2012, we recorded 364 and 3,467 bat echolocation passes with active and passive sampling, respectively, at 13 sites located in the Cantonment Area of the installation. Approximately 42% of calls were identified to the species level using active sampling and approximately 37% of calls were identified using passive sampling. We detected seven of the nine possible species that could occur at Fort Drum during the 2011 and 2012 monitoring seasons, including Indiana bats, northern bats, little brown bats, big brown bats, eastern red bats, hoary bats, and silver-haired bats (Table S1, Supplemental Material). We did not detect eastern small-footed bats or tri-colored bats. We determined estimates of occupancy and detection probabilities for the seven species that were detected during this study. Overall, passive sampling yielded higher detection probabilities than did active sampling for all detected species in both years (Figure 1). Additionally, there was a decline in detection probabilities in WNS-impacted species across years for passive sampling. Detection probabilities for little brown bats decreased from 0.51 (SE  =  0.13) to 0.26 (SE  =  0.21), for Indiana bats from 0.62 (SE  =  0.17) to 0.08 (SE  =  0.05), and for northern bats from 0.39 (SE  =  0.30) to 0.03 (SE  =  0.03). Although occupancy estimates were higher using active sampling than passive sampling (Table 1) in 2012 for the Indiana bat (1.0 versus 0.38) and northern bat (0.04 versus 0.03) and in both years for the little brown bat (1.0 versus 0.88 in 2011 and 1.0 versus 0.41 in 2012), the probability of detecting these species was substantially higher using passive sampling in both years (Figure 1). Although occupancy estimates were similar between active and passive sampling for big brown bats (0.83 versus 0.74 in 2011 and 0.94 versus 0.87 in 2012) and species not impacted by WNS (eastern red bats, hoary bats, and silver-haired bats; Table 1), higher detection probabilities were usually achieved using passive sampling (Figure 1).

From comparisons on mean hourly foraging activity, we found that sampling method had a significant effect on number of echolocation passes recorded for big brown bats (t  =  2.36, P  =  0.029) and little brown bats (t  =  2.64, P  =  0.015) when active and passive sampling occurred in the same night (Table 2). Higher relative foraging activity was detected for big brown bats using active sampling and for little brown bats using passive sampling (Table 2). Indiana bats were not detected on nights when passive and active sampling both occurred, and were thus not included in the comparison of relative foraging activity between sampling methods. There was not a significant difference in relative foraging activity recorded by passive or active sampling for other detected species. Observed relative foraging activity was generally lower in WNS-impacted species than non-impacted species using either method of sampling in both years.

Our cost assessment revealed similar equipment, gas, and labor costs of $2,420 (U.S. dollars) and $2,330 for passive and active sampling, respectively. However, because of the much longer duration of 24 h by passive sampling versus 1 h by active sampling over two nights, we determined a cost of $101/h of data collected with passive sampling versus $2,330/h of data collected with active sampling. Finally, our cost assessment revealed that contracted mist netting costs approximately $270/h of data collected at a rate of 10 h of data collected over two nights (Table 3).

Previous research has suggested that active sampling produces higher quality recordings (i.e., less sound distortion because of ability to adjust directionality of microphones), greater total number of echolocation passes, and higher species richness than does passive sampling (Britzke 2002; Johnson et al. 2002; Milne et al. 2004). However, the results of our study suggest active sampling provides lower detection probabilities and performs no better in recording relative foraging-activity patterns than passive sampling for most WNS-impacted species at Fort Drum. Although passive and active acoustic sampling using Anabat detectors have been widely used, previous work has compared measures of relative foraging activity but not the more critical measure of detection probability. Furthermore, previous studies have compared passive and active sampling during comparable and/or simultaneous sampling durations. Alternatively, our study compared two standard, temporally distinct protocols, thereby allowing passive sampling to display its inherent “duration” advantage over active sampling. Additionally, Johnson et al. (2002) recorded active acoustics directly to a laptop and passive acoustics to a tape recorder system for comparison. We deployed passive detectors for much longer periods than active detectors in a given night and for several nights in a row with data recorded to compact flash storage media for both methods in our study. Therefore, it is difficult to directly compare previous work to our study because of differences in methodologies and advancement in technologies available to us that preceding researchers did not have.

Johnson et al. (2002) in northwestern Georgia and Milne et al. (2004) in Australia report that active sampling was superior to passive sampling in detecting higher species richness, detecting high-flying bats, and recording better quality calls, particularly for myotine bats. We did not inspect the quality of recordings directly between methods in our study; however, the percentages of calls identified to the species level using passive versus active sampling, respectively, were similar in 2011 (53% versus 52%) and 2012 (37% versus 42%). Although there was no difference in species richness detected by the two sampling methods over the duration of our study, passive sampling consistently yielded higher observed detection probabilities than active sampling for all species, despite potentially higher occupancy estimates for some species using active sampling. MacKenzie et al. (2002) suggest that occupancy estimates very close to 1.0 should be cautiously interpreted if obtained when the detection probability is <0.15 or when sampling occurs over fewer than seven events. Such estimates are based on low amounts of presence data, making it difficult for the model to distinguish between true absences and false nondetections. This is the case herein when active sampling was used for all WNS-impacted myotines, as well as for hoary bats in both years and in 2012 for silver-haired bats. Additionally, MacKenzie and Royle (2005) and Gorresen et al. (2008) indicate that when occurrence of a species is low, survey designs should employ as many sites as possible at the expense of repeated visits to the same sites. In our study, 31 unique sites were sampled over 2 y, yet occupancy estimates for active sampling were not reliable, suggesting that higher levels of effort are needed to successfully monitor for rare bat species, such as those impacted by WNS, when using active sampling. Indeed, Weller (2008) suggested that data-deficient species that are rare or difficult to detect may require >50% more sampling effort than common species when using acoustic monitoring. Passive sampling produced the highest detection probabilities and, therefore, the most robust estimates of occupancy for the majority of species our study.

Although detection of rare or declining species requires high levels of effort, and is hence costly (Weller 2008), acoustic monitoring as a tool for estimating occupancy and detection probability clearly is still more efficient and cost-effective than traditional capture methodologies for detecting rare bat species at Fort Drum. For example, Indiana bats were encountered within two nights of passive acoustic sampling, whereas no Indiana bats were captured using netting around known historical maternity areas until the 35th net night in 2012 (Coleman 2013). Furthermore, passive sampling produced overall higher detection probabilities in this study for all species that were successfully detected. Accordingly, we believe that passive sampling should be the preferred method for maximizing acoustic detectability of species not only in areas that have observed severe post-WNS declines similar to those at Fort Drum, but also in areas where there is interest in monitoring for non-impacted species.

Higher relative foraging activity was recorded for big brown bats using active sampling than with passive sampling. We documented that big brown bats are relatively easy to detect using either acoustical method relative to the other bat species. This is unsurprising given what is known about its population stability relative to WNS and its ubiquitous use of foraging habitat (Shump and Shump 1982; Kurta and Baker 1990). Big brown bats have not exhibited drastic declines locally as have the myotine species (Ford et al. 2011; Coleman 2013). For species such as the big brown bat that are still common on the landscape, active acoustic sampling may still be a suitable monitoring technique, because these species are more likely to be detected in any given 30-min sampling period than are rare or declining species. Moreover, it is somewhat surprising that northern bats did not demonstrate measured effects based on sampling method as was exhibited by little brown bats, which were recorded with higher relative foraging activity when using passive sampling than when using active sampling. However, WNS-impacted species are now detected so infrequently at Fort Drum—as demonstrated by the lack of Indiana bat recordings when active and passive sampling were conducted simultaneously—that much greater sampling effort using either acoustic method may be required to determine changes in relative foraging activity or adequately model these species with reliable parameter estimates and low standard errors. Unfortunately, because of severe WNS-induced population declines (Ford et al. 2011; USFWS 2012), some species (e.g., Indiana bat, little brown bat, and northern bat) may require much greater effort than common species to be detected when they are present, regardless of which method is used. Furthermore, species such as northern bats are difficult to record acoustically because of their low-intensity call amplitude (Broders et al. 2004). Finally, some species were not detected acoustically at all, such as the eastern small-footed bat and the tri-colored bat. Although both of these species have been captured and recorded previously at Fort Drum (C.A. Dobony, Fort Drum, unpublished data), encounter rates have always been low and it is possible that these species have become locally extirpated since the arrival of WNS.

From a management perspective, initial equipment costs for active and passive sampling are similar for the level of effort we performed. However, the cost per hour of data is substantially higher for active than for passive sampling because passive sampling has an inherent advantage of longer sampling durations and, therefore, a greater amount of data collected with lower effort required to implement each survey (Table 3). For rare species that require rigorous sampling effort, the ability to implement passive monitoring over much longer periods and simultaneously at multiple sites is clearly more cost-efficient than intensive active sampling that compounds labor-cost requirements.

The U.S. Fish and Wildlife Service is currently reviewing the status of some WNS-impacted species (the little brown bat and the northern bat) to determine whether these species warrant a threatened or endangered listing under the Federal Endangered Species Act (Kunz and Reichard 2010; CBD 2010; USFWS 2013). Successful implementation of conservation actions for rare species is contingent upon detecting those species where they are present in suitable habitat. Thus, a monitoring technique that can detect the most species under the most efficient time and effort conditions is required. In many WNS-impacted areas where mist-netting now is unrealistic for determining presence or probable absence of rare species because of the cost and time-prohibitive effort that is required, managers will need to rely on acoustic monitoring. We recommend passive acoustic sampling as the preferred method for maximizing detection probabilities of declining species in areas where severe impacts associated with WNS have been observed. Unfortunately, as WNS impacts compound and bat populations continue to decline, any technique will probably require far more effort (and potentially cost) than was previously required pre-WNS or immediately post-WNS.

Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.

Table S1. Acoustic bat activity data collected using active and passive acoustic sampling at 31 sites of acoustic echolocation detectors at Fort Drum, New York, summers of 2011 and 2012. Variables are Site (1–31), Design (active or passive), Year (2011 or 2012), Month, Day, and number of pulses for each of nine potential species. Active sampling is defined as acoustic sampling that occurred in 30-min intervals between the hours of sunset and 0200 with a user present to manipulate the directionality of the microphone. Passive sampling is defined as acoustic sampling that occurred over a 12-h period (1900–0700 hours) without a user present and with the microphone set in a predetermined direction. The number of pulses for each species represents the number of “passes” (files that contain >3 sound pulses) identified as a given species by EchoClass Version 1.1 (U.S. Army Engineer Research and Development Center, Vicksburg, MS, USA). Species include: big brown bat Eptesicus fuscus, eastern red bat Lasiurus borealis, hoary bat L. cinereus, silver-haired bat Lasionycteris noctivagans, eastern small-footed bat Myotis leibii, little brown bat Myotis lucifugus, northern bat M. septentrionalis, Indiana bat M. sodalis, and tri-colored bat Perimyotis subflavus.

Found at DOI: 10.3996/082103-JFWM-057.S1 (19 KB XLSX)

Reference S1. Britzke ER. 2002. Designing monitoring programs using frequency-division bat detectors: active versus passive sampling. Pages 79–83 in Bringham RM, Kalko EKV, Jones G, Parsons S, Limpens HJGA, editors. Bat Echolocation Research: tools, techniques, and analysis. Austin, Texas: Bat Conservation International.

Found at DOI: 10.3996/082103-JFWM-057.S2 (38 KB PDF).

Reference S2. Britzke ER. 2003. Use of ultrasonic detectors for acoustic identification and study of bat ecology in the eastern United States. Doctoral dissertation. Cookeville, Tennessee: Tennessee Technological University.

Found at DOI: 10.3996/082103-JFWM-057.S3 (229 KB PDF).

Reference S3. Britzke ER, Herzog C. 2009. Using acoustic surveys to monitor population trends in bats. U.S. Army Engineer Research and Development Center.

Found at DOI: 10.3996/082103-JFWM-057.S4 (88 KB PDF).

Reference S4. Britzke ER, Murray KL, Heywood JS, Robbins LW. 2002. Acoustic identification. Pages 220–224 in Kurta A, Kennedy J, editors. The Indiana bat: biology and management of an endangered species. Austin, Texas: Bat Conservation International.

Found at DOI: 10.3996/082103-JFWM-057.S5 (532 KB PDF).

Reference S5. [CBD] Center for Biological Diversity. 2010. Petition to list the eastern small-footed bat Myotis leibii and northern long-eared bat Myotis septentrionalis as threatened or endangered under the Endangered Species Act.

Found at DOI: 10.3996/082103-JFWM-057.S6 (726 KB PDF).

Reference S6. Coleman LS. 2013. Assessing the impacts of white-nose syndrome induced mortality on the monitoring of a bat community at Fort Drum Military Installation. Master's thesis. Blacksburg, Virginia: Virginia Polytechnic Institute and State University.

Found at DOI: 10.3996/082103-JFWM-057.S7 (1.64 MB PDF).

Reference S7. Kunz TH, Reichard JD. 2010. Status review of the little brown myotis (Myotis lucifugus) and determination that immediate listing under the Endangered Species Act is scientifically and legally warranted. Boston University.

Found at DOI: 10.3996/082103-JFWM-057.S8 (452 KB PDF).

Reference S8. [USFWS] United States Fish and Wildlife Service. 2012. North American bat death toll exceeds 5 million from white-nose syndrome. Arlington, Virginia News Release: 17 January 2012.

Found at DOI: 10.3996/082103-JFWM-057.S9 (109 KB PDF).

Reference S9. [USFWS] United States Fish and Wildlife Service. 2014. 2014 Range-wide Indiana bat summer survey guidelines.

Found at DOI: 10.3996/082103-JFWM-057.S10 (635 KB PDF).

Funding for this study was provided by the Fort Drum Natural Resources Branch through National Park Service, Southern Appalachian Cooperative Ecosystem Study Unit contract W9126G-11-2-SOI-0029 and the U.S. Geological Survey Cooperative Research Unit Research Work Order VA-RWO-142. We thank R. Rainbolt, A. Dale, S. Dedrick, G. Luongo, and N. Grosse for field assistance. Earlier drafts of this manuscript were reviewed by D. Stauffer and J.B. Johnson. Additionally, this manuscript received helpful peer review from the Subject Editor and two anonymous reviewers.

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