Thank you!

Capture and GPS tracking of Dall’s sheep, grizzly bears and wolves

Our first capture session occurred in April 2006 for Dall’s sheep and wolves, and in May 2006 for grizzly bears. We fitted six Dall’s sheep ewes, seven grizzly bears (5 females and 2 males) and eight wolves from two different packs with GPS radio-collars (Telonics TWS 3680 for grizzly bears; TWS 3580 for ewes and wolves; and also Lotek 3300SW for wolves). Captures were conducted by highly trained professionals, in accordance with guidelines approved by the Canadian Association of Zoo and Wildlife Veterinarians and the Government of the Northwest Territories. The collars were programmed to record 6 locations a day during most of the year, and 12 locations a day from May 15 to June 14, which assure higher monitoring intensity during lambing season. Grizzly bear collars enter a standby mode from November 30 to April 1 –during hibernation. All collars are equipped with an automatic release mechanism to facilitate the collar recovery after the end of the battery life and to minimize disturbances to the animals. For all of the marked animals except the six wolves wearing store-on-board collars (Lotek 3300SW), partial locations are transmitted every week, and we have so far collected over 2,500 locations, which are used for this analysis. The remaining locations will be obtained when we will retrieve the collars from the field, after the end of the battery life.

P1010012_1slide0006_image008DSCN0330


Habitat monitoring

Landscape features of the Richardson Mountains were characterized using data from the Gwich’in GIS Project (Source: GRRB and Gwich'in Tribal Council). Layers included in the resource selection analysis include topographic coverage, vegetation, rivers and streams. Digital elevation maps (Can30dem) were also used. In the future, a snow map of the study area will also be included in the analysis, as we will start monitoring snow depth and density in the study area in early 2007. Moreover, climate data will also be incorporated to this analysis in the future, as a weather station was erected last August and was programmed to record temperatures, wind speed, humidity and atmospheric pressure.

weatherstationPB240036


Analytical methods
Although some of these methods may change in the future, one of the goals of this exercise was to explore the various possible techniques and decide which ones might be useful in the future. Here I briefly outline the methods used and justify my choice.

As conducted by Doncaster (1990) and Gehrt and Fritzell (1998), I investigated the interactions between Dall’s sheep, grizzly bears and wolves at both the static and dynamic levels. Static interactions can be estimated mostly by home ranges and overlap between individual. Dynamic home ranges were evaluated by examining the locations of collared animals at simultaneous times. More specifically, I calculated the average as well as the minimum distance between pairs of individuals for all time intervals when there were two or more individuals monitored simultaneously. To ensure precision but allow for some transmission error, locations were grouped at the nearest seven minutes time. Proximity matrices were calculated based on the average and minimum Euclidian distance values, and the proximity between individuals was estimated with a metric
Multidimensional Scaling analysis (MDS). This multivariate technique is particularly useful to visualize the data and interactions between individuals.

To estimate impact of various habitat features on the interactions and habitat use of Dall’s sheep, grizzly bears and wolves, I conducted various GIS analysis, joined layers and extracted data, in order to organize my data such as each animal location, along with a set of random locations, would be characterized with an elevation, a slope, an aspect, and the presence of vegetation or not. My initial intention was to perform a resource selection function (Boyce and MacDonald, 1999), as most multivariate techniques are not usually used for habitat selection studies (Manly et al. 2002). Nevertheless, as I had many variables and many observations, I decided to try a
Principal Component Analysis, to reduce the number of variables that need to be considered to a small number of indices (the components) and see which variables should be retained in a resource selection function analysis (Manly 1986). Finally, I also performed a Canonical Correspondence Analysis (CCA), which has been developed to allow ecologists to relate the abundance of species to environmental variables (Ter Braak 1986). It can provide an assessment of the relative importance of each habitat variable to the presence of each species. The relationships were completed using a Spearmans's Rank correlation analysis. This univariate non-parametric statistical technique enables the relationship between species occurence and abiotic factors to be analyzed individually.