Revising VT State Deer Wintering Area Data
The white-tailed deer (Odocoileus viginianus) is the species of deer found in Vermont and an important cultural symbol to the state. During the harsh winters, the deer must find functional deer winter habitats, commonly referred to as deer wintering areas (DWAs), which are characterized by features such as close canopy forest cover to shield the deer from the snow.
It is important to identify DWAs so they can be properly managed and protected under Act 250 and provide information for hunters so they can effectively manage deer populations. While the VT Agency of Natural Resources created a map of DWAs, limitations to their methodology included a lack of field research and focused aroudn the presence of softwood tree cover. To improve on their efforts, I developed a habitat suitability index (HSI) for white-tailed deer informed by a wider range of variables and compared the resulting DWA map in Ripton, VT with the state's from 2011.
Comparing VT ANR's DWA data and mine, it is clear that more research must be done in order to more accurately DWAs. It was confusing how to classify what a DWA was and my HSI determined that a much larger area in Ripton has the highest score on my HSI for white-tailed deer. This was confusing as even though the official state DWAs focused on softwood forest cover, as indicated in dark red and in yellow green in a map further down explaining forest cover types, there is no clear correlation between the presence of softwood forests and the state data. It says on description for the data that VFWD District Biologists were consulted to verify DWAs and it would be helpful to understand how these biologists evaluated the areas so towards improving the HSI and therefore DWA mapping.
While my geography class did collect field data on the presence of white-tailed deer, it was not sufficient enough to be able to run the multivariate statistical models needed to evaluate my HSI. Therefore, more deer pellet surveys would be needed to remedy this and spending more time on investigating a wide range of variables such as the presence of browse and specific species of trees that are known to attract white-tailed deer would allow the development of a more accurate HSI to base my statistical analyses off.
Methodology and Limitations
I focused on three main variables to develop my habitat suitability model (HSI) for white-tailed deer:
- Proximity to water sources
- Depth of snow cover
- Type of forest cover
Access to water is vital for the deer's survival. A previous study showed that deer are usually within 1 to 1.5 miles of permanent water, and with the maximum distance of 2.4 miles (Rodgers et al. 1978). With this information, a score from 1 to 3 was given to areas around the water sources (1 -- <1.5 miles, 2-- 1.5 - 2.4 miles, 3-- >2.4 miles). To achieve this, identified water sources from the land cover data generated by the Multi-Resolution Land Characteristics Consortium (MRLC) were joined to stream data generated by the USGS and the EPA. Euclidean distance was calculated from the streams and water sources, and then reclassified.
Low (<20 cm), Moderate (20 cm - 50 cm) and Deep (> 50 cm) were used to score areas 1, 2 and 3 respectively, with the lowest score going to low snow depth as deer are more likely to use areas with shallow snow during the winter (Morrison et al. 2001). The snow depth data was collected from the National Snow and Ice Data Center and reclassified for scoring from 1 to 3.
The same land cover data used to identify water sources helped identify the different types of forest cover available as well. Three types of forest cover (deciduous, evergreen and mixed) were identified in the land cover data with deciduous forest scored the lowest and evergreen forest scored the highest. Studies have shown a positive correlation between the density of coniferous forest cover and deer presence (Kirchhoff and Schoen 1987; Shi et al. 2006).
The final map layer used to compare my findings with the VT ANR layer was a weighted sum of the different scores to create an aggregate score for the different areas. I ran an original least squares regression on the estimated deer density (estimated from the deer pellet surveys my geography class conducted) with my weighted sum layer to try and correlate my model that was based on aerial satellite data with field research. However, GIS warned me there was a spatial bias in my data, meaning it would have been better to run a geographically weighted regression. Unfortunately I was not able to do so because of the quantity of data we had collected.
Further limitations inherent to the use of aerial satellite imagery and associated data include constraints related to the time when the data for the various layer was taken and the spatial resolution of the data which affected how the areas were scored according to my HSI. Given more time and knowing the results of the three variables I tested, I would have chosen another variable other than proximity to water as it seemed that all areas in Ripton were within an ideal range for the white-tailed deer as well.
Cited References (Not Online)
Rodgers, K.J.; Ffolliott, P.F.; Patton, D.R. Home range and movement of five mule deer in a semidesert grass-shrub community. In Rocky Mountain Forest and Range Experiment Station; Forest Service US, Department of Agriculture: Fort Collins, CO, USA, 1978; Volume 355, pp. 1–6.