Apparent survival of tree voles
We estimated annual apparent-survival for tree voles using a Cormack-Jolly-Seber model implemented in R package RMark (Laake & Rexstad, 2008; White & Burnham, 1999). During annual nest platform checks we attempted to capture any tree voles that were present and marked tree voles by clipping the tips of their dorsal fur, making a high contrast mark readily identifiable using remote camera photographs. We then resighted those individuals and analyzed 1-day bin-width encounter histories of marked tree voles that began on the date of capture. Capture and handling methods were approved by the U.S. Forest Service Institutional Animal Care and Use permit #2016-009 and Oregon Department of Fish and Wildlife Scientific Taking Permit 041-18.
Because tree voles molt, growing their fur at unknown intervals and we had only one capture occasion per year, we right-censored encounter histories to a survey period such that we minimized the uncertainty that animals were likely to leave the sample due to their mark fading while also retaining most of the data. To estimate the longevity of marks we reviewed sequential photographs of marked female tree voles (n = 40) for which we observed their unique marking fading to became indistinguishable from an unmarked vole. We estimated that 20% of marks became indistinguishable at 63-days or fewer although some lasted longer as we estimated marks faded at a median of 88 days and a mean of 83.4 ± 28.7 days (n = 20). We used 63-days to estimate apparent survival to minimize underestimates of survival due to mark-loss. To assess sensitivity of our survival models to encounter history length, we present survival estimates using encounter histories of 35-days to 84-days.
As tree voles were only marked in the summer months (June – August), our inferences were limited to June – October. To provide an estimate of annual survival comparable to previous studies (Swingle et al., 2010), however, we assumed that if survival and predation risk was constant year-round, extrapolating our 1-day survival rate to a 1-year period (365 days) would provide a valid comparison. To assess the assumption that predation was consistent year-round in our study, we summarized our qualitative assessment of weasel predation for each month (Fig. S1a). We pooled data across years but provided separate estimates for males and females. To estimate the variance of average annual survival, we temporally rescaled our data using the delta method (Powell, 2007).
Apparent survival can underestimate actual survival compared to known-fate estimates as it is impossible to distinguish emigration from mortality within the model. We presented data on observations of known fates of marked tree voles, including observations of mortality due to predators and emigration when a marked tree vole was observed at a different nest platform than the capture location. In 2017 we applied passive integrated transmitters to tree voles and recaptured tree voles in 2018 to provide further information on apparent survival based on the rate of recapture of these transmitters in live tree voles.