A hierarchical modeling approach to estimate terrestrial salamander abundance prior to prescribed fire and timber harvest

Background/Question/Methods

Prescribed fire and timber harvest are anthropogenic disturbances that can have substantial effects on forest ecosystems. Timber harvest often causes drying of the soil and leaf litter, making forests less capable of supporting amphibian populations. Effects of prescribed fire on wildlife in general, and amphibians specifically, are inadequately understood. Terrestrial salamanders are thought to play an integral role in nutrient cycling and forest productivity, but may be negatively affected by disturbances that alter their microhabitat. With prescribed fire becoming common in forest management, it is essential to understand its effects on wildlife. Therefore, we investigated the effects of prescribed fire and timber harvest on terrestrial salamanders. We were particularly interested in robustly determining population size prior to treatment, and identifying which factors best predict the abundance and detection of terrestrial salamanders. In spring and fall 2010-2011, we conducted 5 repeated samples of 20 5-hectare experimental plots, which were harvested in early 2012 or will be burned in late 2012. We performed area-constrained searches of natural cover and leaf litter, measured (SVL), and recorded the capture location of salamanders (predominantly Plethodon serratus). Terrestrial salamanders have notoriously low detectability; thus, we used program unmarked, which fits hierarchical models of abundance for species subject to imperfect detection. 

Results/Conclusions

We recorded 1883 captures of P. serratus and 21 captures of other Plethodontid salamanders (Plethodon albagula, Eurycea longicauda). The average number of captures per 3 x 3 meter sample plot ranged from 1.6 (fall 2011) to 2.9 salamanders (fall 2010). Most salamanders (75%) were found within or under leaf litter; the rest were either under rocks (11.4%) or woody cover (13.6%). Salamander captures were strongly correlated with recent rainfall; approximately 63% of the variation in captures-per-plot was explained by the number of days elapsed since rain. Rainfall had the strongest effect on detection; slope, aspect, and soil water-holding capacity were the best predictors of true abundance. The overall detection probability was between 10-15%. We found no significant differences between treatment groups prior to burning or harvesting. Our results illustrate the importance of accounting for imperfect detection when sampling wildlife. They also suggest that including leaf litter in area-constrained searches for terrestrial salamanders allows for a more complete understanding of their distribution among microhabitats. This sampling technique will allow us to detect changes in cover object use following prescribed fire and timber harvest and reveal the role of leaf litter loss in population persistence.