Utilizing Plethodontid Occupancy to Assess Initial Ecosystem Responses to Invasive Plant Control Treatment in Central Hardwoods

Description

This data reflects occupancy sampling of Plethodontid salamanders in West Virginia, USA across lands treated for invasive plant control. We hypothesized negative impacts on odds of salamander occupancy on treated sites due to loss of vegetative structure and increased predation risk. Our data identified decreased occupancy odds on treated sites, decreased occupancy odds on sites with lower canopy cover, and high interspecific variation in occupancy. Our results reflect potential indirect damage to Plethodontids and correlated ecosystem services from invasive plant control efforts. We collected vegetation and salamander data across 60 sites in West Virginia, USA, across 12 sampling occasions. Vegetation sampling was completed once and included vegetation data in 0.02 ha plot. Salamander sampling took place within the same plot, and included the use of visual and coverboard sampling, completed 12 times at each site. We also measured observation level variates, such as soil temperature at the time of sampling. We used a single-season, single-species occupancy model to test hypotheses, using the "unmarked" package in RStudio. To include multiple species, we coded each species as a factor with its own occupancy intercept. Additionally, to address a potential closure assumption violation, we coded Season as a covariate to observe occupancy odd differences between 'Spring' and 'Fall' sampling. The 'Treatment' covariate in our modeling is a binary covariate, meaning sites were simply treated for invasive species (1), or did not receive treatment (0). Therefore, while we see a decrease in occupancy odds between treated and untreated sites, we do not know what is driving this difference. Herbicide impacts, loss of vegetative structure, increased predation risk, or a combination of these could all be driving a larger 'Treatment' effect.

Steps to reproduce

We established 15 0.02 ha sampling plots throughout each WMAs, totaling 60 sites. Sites included a gradient of various environmental conditions. Soil temperature and soil moisture were measured during each salamander sampling occasion and during our vegetation sampling using a digital soil tester inserted at a 2 cm depth. We measured lumen amounts during each diurnal sample. pH was only recorded once per plot. Leaf litter depth was collected by measuring five points within our 0.02 ha, by using a ruler to estimate leaf litter depth. Understory vegetation sampling was completed in the late summer. Per site, 4 1m2 quadrats were used to estimate the understory vegetation and coarse woody debris cover. We estimated the percent cover of each species and coarse woody debris within the quadrat. Overstory sampling was conducted by recording species and diameter at breast height of all woody vegetation with a DBH ≥ 10.16 cm within our 0.02 ha plot and for all stems between 2.54 cm and 10.16 cm within a 0.009 ha plot, located around the same plot center. Canopy cover was determined by taking a photo at approximately breast height (approx. 1.5m) and using pixel % to estimate cover. Salamander sampling was conducted during the mid-spring to mid-fall of 2025. This was done to increase likelihood of detection. Sampling across all sites was conducted in a stretch of 2 to 4 consecutive days, to increase the probability of similar weather conditions between sites. There were six sampling occasions conducted during diurnal conditions, and six sampling occasions conducted during nocturnal conditions. These samples alternated, beginning with diurnal sampling in our first week, doing nocturnal sampling during our second, and continuing respectively. At each site, six yellow-poplar coverboards were placed within our plot. Coverboards were placed on both treated and untreated plots. Coverboards were checked in tandem with the visual sampling at each plot. Along with the coverboards surveys, visual sampling was conducted during each visit. Visual sampling consisted of actively searching for salamanders within the site, overturning refugia and returning it to original placement if moved. To account for imperfect detection, we used single-season, single-species occupancy modeling in RStudio (ver. 2026.01.0+392, R ver. 4.5.1) using package “unmarked” (ver. 1.5.1). We assumed similar detectability among Plethodontids due to their comparable surface activity patterns. We estimated the occupancy for our 8 detected species by coding each species as a factor. Due to multiple species being modeled within the same sites, species-specific occupancy estimates shared site-level covariates. All continuous covariates were scaled by using the methods described in Gelman (2008), in which the mean is subtracted by each value and divided by two standard deviations.

Institutions

• West Virginia University

  West Virginia, Morgantown

Categories

Funders

• EQT Corporation
• West Virginia University

  Morgantown

Licence