(The full thesis can be accessed here.)

Research Purpose

For my undergraduate senior thesis, "Comparing the Effects of Landscape Habitat Data Collection Methods in a Species Occupancy Model," I investigated how land cover data resolution changed the results of multi-season occupancy models for river otter (Lontra canadensis) and mink (Neovison vison).

Overview of Methods

I used data from "River otter and mink occupancy dynamics in riparian systems" (Holland et al. 2019) paired with land cover data from online databases (National Land Cover Database, National Wetlands Inventory) instead of the original study’s 10m, hand-digitized raster file that was created from field measurements. To simplify the land cover data and align it with the original study's categories, I reclassified deciduous forest, evergreen forest, and mixed forest into one category: “forest”. Additionally, the NLCD land cover data categorized many rivers and streams within the study sites as “woody wetlands”. Therefore, I used a separate file for wetlands, rivers, and streams from the U.S. Fish & Wildlife Service National Wetlands Inventory that more accurately categorized the water sources. All other data (e.g. environmental, weather, temporal) were not changed from the original study.

Using ArcGIS Pro, I created buffers for the 103 sites at distances of 1km, 500m, 250m, and 100m. I also created a 50m buffer to calculate riparian area forest. I then used these buffer areas to calculate percentages for each land cover type, including wetlands, and density of roads for each of the 103 sites at each buffer distance.

After additional calculations and reformatting in spreadsheets, I analyzed the data in R Studio. I followed the original study's criteria to combine variables and find each species' top occupancy, colonization, and extinction models. The top detection variables were kept the same, as the replaced data did not affect those variables. All models were created in R Studio using the package “unmarked”. I compared overall models using the Akaike information criterion. I then compared the top models from my approach to the top models from the original study, to assess whether using the coarser data would result in top models different enough to affect management implications.