Epsilon Archive for Student Projects

Abstract

The aim of the study was to find repetitive behavioral patterns pre calving that could be used as indications of calving in beef cattle. In the study sixteen Standing- and lying down sensors were attached to the leg of sixteen beef cows. Ten of the sixteen beef cows gave birth to calves during the trial period. From these ten cows, data from six cows were used to assess the sensors fitness for use as calving indicators. The analytical part of the experiment was divided into two parts, calculation of divergence of mean lying time within different time intervals, and the number of lying bouts.

The first part consisted of optimization of an equation used to highlight divergence in mean lying time within eight different time intervals (3, 6, 9, 12, 15, 18, 21 and 24 hours). The equation used was: z(n) = y(n) – y(n-1) + k*z(n-1). Five different constants (0.1, 0.3, 0.5, 0.6 and 0.7) were tested to optimize the equation, which should give rise to as low false positive calving indications as possible among the six cows in the study.

Mean lying time calculated under the longest time interval and with the highest constant gave rise to less false positive calving indications, than if mean lying time was calculated under shorter time periods and with a smaller constant. The goal of this study was to find one general threshold value for all cows that indicated calving. Since individual variation in behavioral pattern was shown to be large, the threshold value was set low, not to miss a calving. This gave rise to unacceptable high numbers of false positive calving indications for some individuals. To reduce the problem with false positive calving indications, an individually based model for detection of deviations from the cow´s normal lying down behavior would more accurate indications of calving.

The second part of the study included analysis of number of lying bouts per day to find differences in the behavioral pattern that could be used to predict calving. The results indicated that this is a better way of predicting calving than measuring mean lying time under a given time interval. Although one threshold value was used for all six cows, the results were more reliable with less false positive calving indications. An individual threshold value would most probably give a more certain prediction of calving.

For both methods described above a computer program for individual interspersion of the cow´s behavior could be developed, which could give the sensors a great value in detecting calving in the future.