Model Validation and Improvement Using New Data on Habitat Characteristics Important to Forest Salamanders, and Short-Term Effects of Forestry Practices on Salamander Movement and Population Estimates

Abstract

Amphibians, because of their semi-permeable skin, sensitivity to changing microclimates, and important role in ecosystems, are often viewed as indicators of ecosystem health. They make excellent organisms for studies on the effects of silvicultural practices. My goal was to provide recommendations for forest management in the southern Appalachians so that harvesting operations are compatible with maintaining healthy populations of forest amphibians.

I tested previously created habitat models that determined the most important habitat characteristics for salamanders. I counted salamanders in 240 10 x 10 m plots located in the MeadWestvaco Wildlife and Ecosystem Research Forest in north-central West Virginia. We also collected a variety of habitat data in these plots to predict salamander abundance with previously created models. These simple linear regression analyses of predicted versus observed values suggest for most models (7 out of 9) a weak relationship between predicted and observed values (R2 from 0.0033 to 0.2869, p from < 0.0001 to 0.7490). However, one of the models showed characteristics suggesting that it predicted new data as well or better than the original data, and therefore was the most accurate at predicting salamander abundance, and could be used for management purposes, although there was still much unexplained variation. This model included the variables woody stems (< 7.5 cm DBH), available rock, riparian status (i.e., within 15 m of a stream), percent overstory canopy cover, and available highly decomposed woody debris (decomposition classes 3 to 5). All of these relationships were positive except for woody stems, suggesting that in order to maintain healthy populations of salamanders, we should protect areas next to streams, with high amounts of rock, decomposed woody debris, overstory canopy cover, and few woody stems.

I also examined the immediate effects of clearcuts on salamander movement and population estimates. I batch marked salamanders in plots at the edges of a clearcut, and in a control plot. Using the Schnabel estimator, I estimated population sizes in each plot. I then compared population estimates pre- and post-harvest on the interior (harvested) and exterior (unharvested) sides of the plots, taking into account the control plot. I also examined adult-juvenile ratios and movements from one side of the plot to the other. I found no significant changes (p > 0.05) following harvest in any of these measures, suggesting that salamanders do not move out of the harvested area post-harvest, at least over the short term (10 months of this study). This suggests that a longer period of time (> 1 year) is required to observe the population declines detected in most studies.