A Habitat Model for the Detection of Two-lined Salamanders at C. F. Phelps Wildlife Management Area, Fauquier and Culpeper Counties, Virginia

Aquatic salamanders represent an important component of Virginia river watersheds, but despite potential declines, few specifics are known about their habitat preferences. We surveyed the habitats of the northern two-lined salamander and collected data on an array of habitat variables associated with the species. We used a logistic regression analysis to develop a model predicting its presence or absence for a given 50m-transect. Our final model incorporated the variation in stream depth and direction of stream flow and accounted for 25% of the variation in our data. We conclude that stream depth variation is an important feature of salamander habitat ecology, and surmise that direction of flow is of site-specific importance possibly related to stream order. Both features may be behavioral adaptations to avoid fish predation. INTRODUCTION Stream-dwelling salamanders are an important component of aquatic ecosystems. They account for a significant proportion of the biomass of a stream ecosystem, and act as a key trophic link, important as both predators and prey (Spight 1967, Burton and Likens 1975, Rocco and Brooks 2000). Consequently, these salamanders have potential to act as an indicator of stream health (Rocco and Brooks 2000, Barr and Babbitt 2002). This is particularly true for headwater streams were salamanders may act as the dominant vertebrate predator (Davic and Welsh 2004). Accordingly, it would be beneficial to better understand how these species make use of their available habitat. This is especially important in the face of on-going amphibian declines (Alford and Richards 1999). Knowledge of this type may provide better insights into the conservation of these species and their associated ecosystems (Cushman 2005). Previous surveys of stream and terrestrial amphibian diversity have been carried out in the Rappahannock River watershed of northern Virginia; however, more needs to be done to quantify the habitat preferences of important stream species (Mitchell 1998, McGhee and Killian 2010). To begin addressing this need, we conducted a preliminary study of salamander habitat at C.F. Phelps Wildlife Management Area (WMA) located in the Rappahannock River watershed and developed a simple habitat model for the Virginia Journal of Science Volume 61 Number 4 Winter 2010 1 Corresponding author. E-mail: jaymcghee@rmc.edu Virginia Journal of Science, Vol. 61, No. 4, 2010 http://digitalcommons.odu.edu/vjs/vol61/iss4 152 VIRGINIA JOURNAL OF SCIENCE northern two-lined salamander (Eurycea bislineata), a common stream species for the area (McGhee and Killian 2010). Northern two-lined salamanders are common to northern Virginia forest streams within the Rappahannock River watershed (Mitchell and Reay 1999). While they are considered potentially important components of the local ecosystems in which they occur, few studies have developed predictive models of habitat use (Davic and Welsh 2004). They occupy stream margins and seeps, using submerged rocks and woody debris for cover; but may periodically be found in upland terrestrial sites (Petranka 1998). Females attach eggs beneath submerged rocks of varying surface area in headwater streams (Jakubanis et al. 2008). Larvae of this species are benthic predators associated with stream pools with low silt (Smith and Grossman 2003, Petranka 1998). Two-lined salamanders are able to access low-order streams typically inaccessible to predatory fishes, and have become adapted to these headwater stream environments (Vannote et al. 1980, Davic and Welsh 2004). We hypothesized that two-lined salamanders would be detected in or near cool narrow, shallow streams. From this hypothesis, we predicted that important habitat variables in a logistic regression model would be stream temperature, stream depth, and stream width. METHODS We chose sampling sites by randomly selecting a GPS starting location constrained to occur within C. F. Phelps WMA, and moving from that point to the nearest stream. We then moved upstream or downstream a randomly selected distance of up to 50m, and laid a 50m transect running downstream. We sampled stream transects by searching five 1-m quadrats placed within each of the five 10-m sections of the 2 transect. The particular location of the quadrat within these 10-m sections was randomly selected (Jaeger 1994, Jaeger and Inger 1994). We searched quadrats by looking under larger cover objects such as rocks or decaying logs, leaf pack, leaf litter, and using a standard-mesh aquarium dip net (1/16 inch mesh size) to sample stream bottoms (Mitchell 2000). We identified captured salamanders to species (Petranka 1998). Data were collected at both transect and quadrat levels (Table 1). We used logistic regression to select models with those predictive variables most associated with salamander captures at the transect level. Variables measured at the quadrat level were averaged and averages and standard deviations were used as separate predictor variables. As synergistic effects may occur between the variables we measured, we created a priori multiplicative variables for testing as well (Table 1). We used forward stepwise selection (P = 0.05 to enter and 0.10 to remove) in SPSS (SPSS Inc., Chicago IL). We assessed variable coefficients using the change in -2 loglikelihood and evaluated the explanatory value of models using Nagelkerke’s r (Ryan 1997, Hosmer and Lemeshow 1989, Nagelkerke 1991). For all statistical