Erickson, R.A., 2016, Indiana Bat Project Data: U.S. Geological Survey data release, https://doi.org/10.5066/F75M63TN.
Summary
Our model is a full-annual-cycle population model {hostetler2015full} that tracks groups of bat surviving through four seasons: breeding season/summer, fall migration, non-breeding/winter, and spring migration. Our state variables are groups of bats that use a specific maternity colony/breeding site and hibernaculum/non-breeding site. Bats are also accounted for by life stages (juveniles/first-year breeders versus adults) and seasonal habitats (breeding versus non-breeding) during each year, This leads to four states variable (here depicted in vector notation): the population of juveniles during the non-breeding season, the population of adults during the non-breeding season, the population of juveniles during the breeding season, [...]
Summary
Our model is a full-annual-cycle population model {hostetler2015full} that tracks groups of bat surviving through four seasons: breeding season/summer, fall migration, non-breeding/winter, and spring migration. Our state variables are groups of bats that use a specific maternity colony/breeding site and hibernaculum/non-breeding site. Bats are also accounted for by life stages (juveniles/first-year breeders versus adults) and seasonal habitats (breeding versus non-breeding) during each year, This leads to four states variable (here depicted in vector notation): the population of juveniles during the non-breeding season, the population of adults during the non-breeding season, the population of juveniles during the breeding season, and the population of adults during the breeding season, Each vector's elements depict a specific migratory pathway, e.g., is comprised of elements, {non-breeding sites}, {breeding sites}The variables may be summed by either breeding site or non-breeding site to calculate the total population using a specific geographic location. Within our code, we account for this using an index column for breeding sites and an index column for non-breeding sides within the data table. Our choice of state variables caused the time step (i.e. \(t\)) to be 1 year. However, we recorded the population of each group during the breeding and non-breeding season as an artifact of our state-variable choice. We choose these state variables partially for their biological information and partially to simplify programming. We ran our simulation for 30 years because the USFWS currently issues Indiana Bat take permits for 30 years. Our model covers the range of the Indiana Bat, which is approximately the eastern half of the contiguous United States (Figure \ref{fig:BatInput}). The boundaries of our range was based upon the United States boundary, the NatureServe Range map, and observations of the species. The maximum migration distance was 500-km, which was based upon field observations reported in the literature \citep{gardner2002seasonal, winhold2006aspects}. The landscape was covered with approximately 33,000, 6475-ha grid cells and the grid size was based upon management considerations. The U.S.~Fish and Wildlife Service considers a 2.5 mile radius around a known maternity colony to be its summer habitat range and all of the hibernaculum within a 2.5 miles radius to be a single management unit. Hence the choice of 5-by-5 square grids (25 miles\(^2\) or 6475 ha). Each group of bats within the model has a summer and winter grid cell as well as a pathway connecting the cells. It is possible for a group to be in the cell for both seasons, but improbable for females (which we modeled). The straight line between summer and winter cells were buffered with different distances (1-km, 2-km, 10-km, 20-km, 100-km, and 200-km) as part of the turbine sensitivity and uncertainty analysis. We dropped the largest two buffer sizes during the model development processes because they were biologically unrealistic and including them caused all populations to go extinct all of the time. Note a 1-km buffer would be a 2-km wide path. An example of two pathways are included in Figure \ref{fig:BatPath}. The buffers accounts for bats not migrating in a straight line. If we had precise locations for all summer maternity colonies, other approaches such as Circuitscape \citep{hanks2013circuit} could have been used to model migration routes and this would have reduced migration uncertainty.
Click on title to download individual files attached to this item.
full_path_data.xml Original FGDC Metadata
View
33.47 KB
application/fgdc+xml
full_path_data.csv
78.37 MB
text/csv
Purpose
We applied a model developed by {Taylor:2010} and \citet{Erickson:2014} to understand how wind energy development would affect the Indiana Bat. The model is a network model consisting of summer nodes (maternity breeding sites) and winter nodes (hibernacula) connected by paths (migration routes). The model is also based upon the stage-structured Indiana Bat model developed by {Thogmartin:2012a}. The model differs from {Taylor:2010} and {Erickson:2014} by two important attributes. First, this model does not include density dependence within the hibernacula because Indiana Bats historically occurred at much greater densities than present levels (pre-European colonization abundances were at least one and likely more orders of magnitude greater than current population sizes). Second, this model does not include migration survival as a function of distance because our model is focused on the Indiana bat, which forages as it migrates as opposed to an avian species that does not forage as it migrates except at stopover sites. The model also differs from {Taylor:2010} because it does not include arrival order within the model. {Taylor:2010} included arrival order because it describes avian populations where the first migrants returning get the best nest sites, but cave bats live in colonies without fixed individual territories. Our model also differs from {Erickson:2014} because we apply the model to an actual (versus theoretical) landscape. Briefly, our modeling efforts consists of the following steps: 1) Parameterizing an occurrence model using ArcPy and other Python modules to manipulate data and RStan to parameterize the model, 2) Generating of a landscape by connecting hibernacula with summer sites, 3) Simulating the population over 30 years with different turbine mortality and WNS scenarios, and 4) Repeating steps 3 and 4 for 1,000 landscapes. The next section of our manuscript outlines our model using the ``Overview, Design concepts, and Details'' (ODD) protocol \citep{grimm2006standard, grimm2010odd}. Although our model is not an Agent Based Model (ABM), the ODD protocol provides a useful framework for documenting our population-level model.
Communities
Upper Midwest Environmental Sciences Center (UMESC)