Discussion

The results of our analysis showed no strong environmental predictors of fix rate in the Yukon Territory. Given past research, some results obtained were quite contradictory to what was expected. Aspect was predicated to have a major impact on fix rate considering the northern latitude of the study area combined with the relatively southern latitude of GPS satellite orbits. However, six of the nine categories were found to have perfect fix rate. Furthermore, the categories which did not have a perfect fix rate (south, southeast, and west) were the areas where high fix rates were expected. Available sky was also predicted to have a major effect on fix rate as it embodies the obstructions caused by varying terrain. However, results showed that while available sky had a small relationship with PDOP, it had little (to no) effect on fix rate. Overall, fix rates were extremely high, and thus the necessity of correcting for fix rate bias is minimal.

Limitations

There were a number of limitations in the analysis:

  • Given the number of variables that can influence GPS performance, the sample size was relatively low, making the isolation of variables difficult.
  • Sample variance was low. The median and mode for fix rate between the 30 collars was 100%, with a standard deviation of 1%. This limited the analysis, not only due to the difficulty of determining relationships using the five collars below 100% fix rate, but also because a single failed fix in many of the collars would have met, or exceeded the standard deviation.
  • Data was collected using three different collar types. Unfortunately, due to results from Chi squared testing, two of the three collar types (Gen 4 SOB and Gen 4 ARGOS) had to be removed from the data set.
  • As data was collected through stationary testing (collars placed on spikes facing upward at a constant elevation), our analysis cannot account for changes in device orientation and elevation. Therefore, results should be used cautiously in future studies where collars are mobile.
  • Data used in our analysis was only collected during the summer months of 2009 and 2010 when vegetation is in full bloom. In result, our analysis could not take seasonality into account.