In search of an optimal bio-logger timeframe and device combination for quantifying activity budgets in free-ranging cattle

Authors: CUNNINGHAM, STEPHANIE - Mississippi State University; Augustine, David; Derner, Justin; Smith, David; BOUDREAU, MELANIE - Mississippi State University

Submitted to: Smart Agricultural Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/7/2024
Publication Date: N/A
Citation: N/A

Interpretive Summary: Sensors attached to a collar placed on free-ranging cattle could potentially be used to monitor not only the animal's location, but also its daily behavior patterns (e.g. walking, grazing, resting, ruminating). We studied the use of a GPS tracking device combined with an accelerometer (a device that measures movements in 3 dimensions) placed on yearling steers grazing rangeland in eastern Colorado. We examined the optimal time frame to summarize measurements by the devices in order to best predict the animals activity patterns. We found that the greatest accuracy (90%) could be achieved by measuring a GPS fix every 90 seconds combined with summarizing the accelerometer movements over each associated 90-second interval. Alternatively, only slightly lower accuracy (0.86) was achieved by summarizing the accelerometer at a 30 second interval, and collecting a GPS every 300 seconds, which required less battery power for the GPS, and would allow the sensors to run for a longer time period. Thus, these sensors can be used to assess daily activity patterns of free-ranging cattle with relative high accuracy. The steers we studied grazed for an average of 8.3 hours per day and ruminated for an average of 6.5 hours per day. Data from the sensors can also be used to assess the animals velocity during grazing bouts, mean grazing bout duration, and the tortuosity of grazing pathways, which in turn can serve as an indicator of forage conditions.

Technical Abstract: Precision livestock management incorporates new technologies, including bio-loggers, to remotely monitor livestock health and behavior. Despite the potential benefits in extensive beef cattle systems, limited adoption of these sensors has occurred potentially due to cost, technical, or processing challenges. For free-ranging cattle, we aimed to provide recommendations for streamlined data acquisition and processing by determining how frequency of sensor measurements affects the accuracy of classification of various behaviors. We resampled GPS and accelerometer data (collected at 1 Hz and 12 Hz, respectively) across multiple timeframes (spanning 10 seconds to 15 minutes) to evaluate which combinations of devices, data features, and timeframes might be considered optimal for assessing resting, grazing, travel, and rumination activities. We used random forest models to predict cattle behavior using two classification schemes: the first separated stationary, grazing, and walking activity, while the second split the stationary activity category into rumination vs. non-rumination. We used our trained model to classify cattle behavior for data collected during the growing season (May – Sept) to assess how variations in model accuracy were reflected in inference of daily activity budgets. In the simplified behavioral scheme, classification accuracy was greatest (>0.90) when GPS information was combined with at least one accelerometer metric. Timeframes of 30–90 s provided the greatest classification accuracy, although timeframes up to 300 s had similar classification accuracies, but with increased variability in accuracy. Classification accuracies decreased when we included rumination, but similarly had the greatest performance when both GPS and a full suite of accelerometer metrics was used (accuracy of ~0.90). Average daily grazing time (8.3 h day-1) was within 2 hours across devices and timeframes, with or without rumination included. Estimates of rumination time were again similar across devices and timeframes, averaging 6.5 h day-1. Estimates of daily travel distance decreased by nearly 4 km as the GPS fix interval varied from 10 s (8.7 km day-1) to 15 min (3.5–8.7 km day-1). We suggest using a combination of GPS and accelerometer data at timeframes between 90 and 300 s for classification of grazing and rumination behavior, depending on study objectives and battery constraints. This study provides guidance for balancing accurate fine-scale data collection with data processing and battery limitations for assessing cattle behaviors in extensive rangelands.