Occupancy Rates and Detection Probabilities of Red-Backed Salamanders on the Virginia Fall Line

Abstract

To meet the conservation needs of declining amphibian populations, there is a need to assess monitoring techniques in various habitat types and seasons. I assessed detection rates and proportion of area occupied via transect monitoring for red-backed salamanders at a site along the Virginia Fall Line in Doswell, Virginia. I established 24 transects in a 3.2-ha area in both riparian and upland habitats. Objects providing natural cover along these transects were sampled 3 times a week in fall and spring over a twoyear period. Models of occupancy and detection were developed and compared using Akaike’s Information Criterion. Based on 113 captures, model selection indicated a low fixed initial occupancy of transects with seasonal changes in colonization and extinction. Detection probability was uniformly low, possibly contributing to model uncertainty in determining the best explanatory variables. I infer that the increased colonization of transect cover objects during fall and increased emigration from cover objects during spring is a result of changing moisture conditions and feeding opportunities. It is likely that occupancy and detection will vary substantially for survey sites based on habitat, season, or even by transect placement, and thus managers will need to estimate these parameters for any population monitoring program of red-backed salamanders. INTRODUCTION In view of the worldwide decline of amphibian populations and the increased interest to conserve these populations, biologists have been highly interested in developing robust monitoring methods for specific populations, habitats, or sites of conservation interest (Dodd and Barichivich, 2007; Adams et al., 2013; Petitot et al., 2014). One problem in developing these methods is that, regardless of the sampling technique, there is a possibility that an observer will fail to detect an individual of the population of interest when they are present on the sampling site. Thus, the apparent absence of a taxon could mean that members of the population of interest are truly absent from the sampling site or that they are present but not detected. Because this problem has significant implications for whether management strategies are implemented, any robust monitoring method will require the estimation of detection probabilities, the probability (ranging between 0 – 1) that an individual will be detected by a survey procedure, given that is available to be found (Schmidt, 2003). Unfortunately, the detection probability is likely to vary extensively based on a suite of environmental Virginia Journal of Science, Vol. 67, No. 1, 2016 http://digitalcommons.odu.edu/vjs/vol67/iss1 doi: 10.25778/BHQ9-3R58 Virginia Journal of Science, Vol. 67, No. 1, 2016 https://digitalcommons.odu.edu/vjs/vol67/iss1 VIRGINIA JOURNAL OF SCIENCE 10 2 variables, species differences, population size differences, and even individual behavioral differences (Lancia et al., 1996; Dodd and Dorazio, 2004; Tanadini and Schmidt, 2011). Monitoring is especially needed for a species integral to specific ecosystem functions (Davic and Welsh, 2004). Red-backed salamanders (Plethodon cinereus) are one such species, serving as a potentially useful candidate for long-term monitoring because they are fairly common within their range, often represent a high biomass in specific habitats, and may be tightly linked to the health of their environment (Welsh and Droege, 2001). Red-backed salamanders are likely to have low detection rates, however, largely as a function of their ecology. Bailey et al. (2004a), for example, have attempted to determine detection rates for the congener southern red-backed salamander (Plethodon serratus) using a Pollock’s Robust Design for mark-recapture estimation, but found that because the majority of the population remained below ground, the likelihood of a redbacked salamander being available (above ground) to be found and then actually being detected could be as low as 4%. In addition, there was a tendency for individuals to be “trap-shy”, and thus unlikely to be recaptured, biasing abundance estimates. These results led Bailey et al. (2004a) to suggest using count data to determine the proportion of an area occupied instead of estimating abundance. This reduces the sampling effort required while still allowing for the estimation of detection probabilities and occupancy rates (MacKenzie et al., 2002). This means that larger areas can be sampled with less intensity for the same amount of effort, which may be more functional for conservation management purposes than more intensive studies on smaller sites. Indeed, Bailey et al. (2004b) accomplished this for multiple species in an area in the Great Smoky Mountains. For researchers wishing to establish long-term monitoring at a particular field site, using the proportion of area occupied (PAO) survey seems a robust possibility. Given their current conservation needs, it is desirable that long-term monitoring projects for amphibians, including the red-backed salamander, are initiated in multiple habitat types to gain a better picture of patterns in occupancy and detectability. Redbacked salamanders are particularly associated with well-drained mature montane forest with extensive leaf-litter and deep, pH-neutral soils, with plenty of cover objects (Burger, 1935; Petranka, 1998; Milanovich et al., 2010; McGhee and Killian, 2013). Many studies establishing monitoring methods are understandably in these types of habitats (Dodd and Dorazio, 2004; Hyde and Simons, 2005; Williams and Berkson, 2004). To establish a complete picture of occupancy for a species, however, a wide variety of habitats should be assessed, along with estimates of the amount of effort required to establish occupancy for these sites (Mackenzie and Royle, 2005). For example, the Randolph-Macon College Environmental Field Station occurs along the Virginia Fall Line, demarcating the eastern boundary between the piedmont and coastal region of the state. In hopes of testing the potential of a long-term monitoring protocol on this site, my objective was to assess occupancy and detection rates for a series of transects on a small subsection of the field station. The estimation of a detection rate is important to determine the likelihood of detecting an individual of a monitored species if present, and so it is helpful to translate the number of individual transect surveys that would be required to detect an individual that is actually present. To accomplish this I calculated the amount of sampling effort Virginia Journal of Science, Vol. 67, No. 1, 2016 http://digitalcommons.odu.edu/vjs/vol67/iss1 3 that would be required to establish occupancy for a given transect (O’Connell et al., 2006). STUDY SITE Fourteen of Virginia’s salamander species (approximately 28% of Virginia species) occur within the York River drainage (Mitchell and Reay, 1999). The Randolph-Macon College Environmental Field Station (EFS), owned by Martin Marietta Quarry, encompasses a small ridge in this drainage, next to a local rock quarry and bordering the Little River in Doswell, Virginia. The EFS is a 26.7 hectare (66 acres) property in northern Hanover County, Virginia, located about 8 miles from Ashland, Virginia (Randolph-Macon College, 2010). This site is located on the Fall Line, a sharp rise in elevation that runs through the state that acts as a geological border between the piedmont and coastal ecoregions of Virginia. It contains mature hardwood forest. Elevation ranges between 150 – 220 m. METHODS In August 2011, I established 24 permanent transects, each 25m in length. These transects were arranged in groups of four, radiating from the corners of a central silt fence enclosure being used for a related study (McGhee, 2013). Each array of 4 transects was established from a randomly located line moving perpendicular to the Little River through both a riparian and upland zone (elevation difference = ~50m). From randomly selected points along this line, I placed the center points of these arrays (a 25-m2 silt fence enclosure) between 0 – 50 m away, only constraining the center of the transect arrays to be a minimum of 25 m away from each other. This resulted in 3 arrays of 4 transects in the upland zone, and 3 arrays of 4 transects in the riparian zone, over an area covering approximately 3.2 ha. I sampled transects following Pollock’s Robust Design (1982), wherein primary sampling periods are comprised of a series of secondary sampling occasions occurring over a short enough time period that a closed population can be assumed. An open population can be assumed across primary sampling periods (MacKenzie et al., 2003). Within an occupancy modeling framework, this allows for the estimation of 4 probabilities: initial occupancy of transect (), colonization of transect (), extinction (or emigration) from transect (), and detection at transect (p: MacKenzie et al., 2003). The colonization and extinction parameters allow for testing changes in occupancy over the time of the study. Primary samples were taken between 17 August – 4 December of 2011, 16 February – 23 May of 2012, 3 October – 6 December of 2012, and 26 February – 8 May 2013, with each separated by a 10-day period on average, and comprised of 3 secondary samples occurring over a 3 – 4 day period. A sample consisted of walking each 25m transect and searching under each natural cover object intersecting the transect. Salamanders found under natural cover objects were measured for total length (TL), snout-vent length (SVL), and identified to species. I assigned an age to red-backed salamanders (juvenile or adult) based on their SVL (adult SVL ≥ 34 mm; Petranka, 1998), and documented the color morphology of the individual (red-stripe on dorsum, or unstriped morph; Petranka,1998). Only detections of red-backed salamanders were used Virginia Journal of Science, Vol. 67, No. 1, 2016 http://digitalcommons.odu.edu/vjs/vol67/iss1 Virginia Journ