Inference of Timber Harvest Effects on Survival of Stream Amphibians Is Complicated by Movement
The effects of contemporary logging practices on headwater stream amphibians have received considerable study but with conflicting or ambiguous results. We posit that focusing inference on demographic rates of aquatic life stages may help refine understanding, as aquatic and terrestrial impacts may differ considerably. We investigated in-stream survival and movement of two stream-breeding amphibian species within a before-after timber harvest experiment in the Oregon Coast Range. We used recaptures of marked individuals and a joint probability model of survival, movement, and capture probability, to measure variation in these rates attributed to stream reach, stream gradient, pre- and post-harvest periods, and the timber harvest intensity. Downstream biased movement occurred in both species but was greater for Coastal Tailed Frog (
Ascaphus truei
) larvae than aquatic Coastal Giant Salamanders (
Dicamptodon tenebrosus
). For
D. tenebrosus
, downstream biased movement occurred early in life, soon after an individual's first summer. Increasing timber harvest intensity reduced downstream movement bias and reduced survival of
D. tenebrosus
, but neither of these effects were detected for larvae of
A. truei
. Our findings provide insight into the demographic mechanisms underlying previous nuanced studies of amphibian responses to timber harvest based on biomass or counts of larvae.
Proportions of marked individuals captured in subsequent primary sampling occasions, by plot of origin (left) and histograms of observed movements (right). Regression lines are fitted values from logistic regression with symbol size proportional to numbers of individuals marked.
Directional bias in movement (left; 30 days elapsed) for the 14 Trask study reaches during pre- (circles), and post- (squares) harvest periods (upper 150 m). Harvested reaches are solid symbols. On the right are estimated percentiles of the predictive distribution of movement distances over 30 days for a new un-harvested stream of a given gradient, depicting the additive effect of dispersion onto the systematic directional bias (bold).
An example time-sequence for the probability of Zt, the location of an individual along the stream thalweg, over 30-day time steps from 30 days to 390 days given an initial location of Z0 = 15 at time 0. The figure depicts downstream drift (systematic movement bias) and increasing dispersion over time for larvae of Ascaphus truei at Pothole Creek 2 prior to the experimental harvest treatment occurring. Vertical dotted lines represent upstream and downstream boundaries of the study reach at Z = 0 and Z = 60, respectively, and the boundary of the harvested region at Z = 30. For any value of Z, the plotted densities are the medians of the posterior distributions of the probability density function for the t-distribution.
Probability of being located downstream of the harvested region boundary [Pr(Zt > 30)] by elapsed time, and Z0 the starting location at time 0. Estimates, which are conditional on survival, are for larvae of Ascaphus truei from Pothole Creek 2 prior to the experimental harvest occurring, and only starting locations upstream of Z0 = 30 are depicted. The line plotted in bold for Z0 = 15 equates to, at each 30-day time-step on the x-axis, the area under the curve in Figure 3 that occurs to the right of Z = 30.
Probability of surviving Pr(Xt = 1) and being located downstream of the harvested region boundary Pr(Zt > 30) by elapsed time t, and Z0, the starting location at time 0. Estimates are for larvae of Ascaphus truei from Pothole Creek 2 prior to the experimental harvest occurring, and only starting locations upstream of Z0 = 30 are depicted. The line plotted in bold is for Z0 = 15.
Posterior distributions of model parameters for Dicamptodon tenebrosus marked when > 2 grams (diagonal hashed boxes), marked when ≤ 2 g (solid gray boxes), and for Ascaphus truei (open boxes). Horizontal ticks represent 2.5th (lower credibility interval), 25th, 50th, 75th, and 97.5th (upper credibility interval) percentiles of posterior distributions. Points are posterior means.
Contributor Notes
Associate Editor: J. W. Snodgrass.