Biodiversity in Vermont

OVERVIEW
My post-doctoral research focused on Vermont ’s bird, amphibian, and reptile diversity.  Field work in 2003 and 2004 at over 600 locations throughout the state documented species occurrences in a variety of habitats.  The goals of the project included: 1) documenting how specific species are distributed, 2) building predictive models to illustrate each species’ likely range in Vermont, 3) modeling the expected effects of land use change and increased urbanization on species, and 4) using optimization modeling to explore the possibility of protecting species while simultaneously meeting human needs.  This research was part of a collaboration with Therese Donovan, Austin Troy, Bill Keeton, and Alexey Voinov at the University of Vermont.

 

Links

Home

Coyote
Communication

Publications

Software

Teaching

NPS Northeast Temperate Network

USGS Vermont
Co-op Unit

FIELD METHODS
Research sites were selected according to a stratified random sample: 20% of sites were in agricultural areas, 20% were in developed areas, and 60% were in forested areas with varying road densities (and varied levels of fragmentation).  Because it was not always possible to get permission to access randomly placed sites, some sampling locations were moved to the nearest place where we were able to secure permission.  Each site contained 4 sampling stations separated by at least 500 m; at each of these stations we searched for birds, reptiles, and amphibians.  Birds were detected using two 10-minute point counts, followed by a third count during playback of a chickadee mobbing tape.  Reptiles and amphibians were located during four five-minute transect searches along 12-meter swaths, followed by a 10-minute area search that ranged up to 25 meters from the central sampling point.  We collected vegetation data and other data at each sampling point; this data included temperature, cloud cover, wind, vegetation density, percent canopy cover, and diameter and species of large trees.

DATA ANALYSES
Field data was entered into a computer database and double-checked for accuracy.  This field data will be supplemented with a variety of fragmentation metrics acquired via GIS.  Our landscape information comes from a 1992 land cover layer, that has been updated to reflect the extent of urbanization in the study area as of 2002. 

One essential feature of survey data is that the probability of finding a species is contingent on two probabilities: the probability that the species is actually present at a given station, times the probability that the species is detected if it is present.  While detecting a species clearly indicates that it is present, NOT detecting a species could be due to the species truly being absent, or to the species being present but not found during our survey.  Luckily there are statistical techniques available (through a program called MARK) that allows us to determine the probability of detecting a species, and how that probability changes depending on different covariates (such as temperature and habitat).  Once we determine the detection probability for a species, we can estimate the true probability that the species will be present in each major habitat type.

Once we have accurate detection probabilities, we can build species-specific habitat suitability models.  These models predict the likelihood of a species being present in an area based on important landscape-level variables like habitat, amount of fragmentation, and distance from water.  These models will be used as inputs for analyses in Austin Troy’s lab that will investigate how different development scenarios for northeastern Vermont will affect the state’s biodiversity.  Alexey Voinov plans to use these models in an optimization context.  One potential optimization question can be framed as follows: given a certain area where a transportation or development project will occur, what is the best configuration that will cost the least while protecting as many species as possible?

STATUS
As of December 2004, all bird, reptile, and amphibian data has been checked.  I am currently working with graduate students in Austin Troy’s lab as they work to update the 1992 land cover data to reflect the extent of urbanization in Vermont in 2002.  In the near future I will be analyzing the detection probabilities for bird species.  For this analysis, Therese Donovan and I are investigating the effects of the chickadee mobbing tape, time of day, and season on the probability of detecting several common species.  Part of this analysis involves determining the best functional form to describe the effect of time and season; this form likely varies by species and depending on whether the tape playback was used.