Page Banner

United States Department of Agriculture

Agricultural Research Service

Title: Classification of Cattle Activity Based Upon High Frequency Spatial Positioning

Authors
item Johnson, Michael - UCLA
item Sheehy, Cody - OREGON STATE UNIV
item Harris, Norman - UNIV ALASKA FAIRBANKS
item Clark, Patrick
item Ganskopp, David
item Louhaichi, Mounir - OREGON STATE UNIV

Submitted to: Society for Range Management Meeting Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: January 1, 2007
Publication Date: February 9, 2007
Citation: Johnson, M., Sheehy, C., Harris, N., Clark, P., Ganskopp, D.C., Louhaichi, M. 2007. Classification of cattle activity based upon high frequency spatial positioning [abstract]. Society for Range Management Meeting Abstracts. Paper No. 221.

Technical Abstract: Animal behaviorists are interested in the spatial pattern of animals on landscapes because of their implications for species survival, resource use, and niche definition, as well as inter- nd intra -specific dominance, facilitation and competition. In the last several years it has become possible to continuously record animal positions via GPS logging at very short intervals (1 second to 60 seconds). We hypothesized that these high frequency GPS locations could be used to identify resting (stationary) locations, travel routes, and time budgets as animals go about their daily routines. Unfortunately, continuous logging produces enormous amounts of data which are difficult to analyze without algorithms to facilitate the classification and interpretation of data. Our firs task was to extract sites on a landscape where animals were stationary for long periods. Two methods were evaluated that automatically separated moving positions from stationary positions: Algorithm 1 identifies and classifies as non-moving (inactive, low velocity points with large turning angles: Algorithm 2 searched the data for a user-defined number of positions that had the closest proximity to one another, then used as a seed group: mean and standard deviation were calculated for the seed group then adjacent positions (in time) were added as long as they were within a specified distance (or are not significantly different) from the core group. Groups were constructed until the desired number had been made or until all points were classified, then minimum convex polygons were created around the point clusters. We also tested a third algorithm that was designed to identify travel routes on a landscape. It searched the sequential coordinate list and identified a seed group of locations with the highest velocity. Consecutive points are added until the velocity dropped significantly and the points are converted into a GIS line file. All Algorithms were tested against field observations and incorporated into the KRESS Modeler version 3.0.5

Last Modified: 10/21/2014
Footer Content Back to Top of Page