Behavior is an extremely relevant trait for efficient, sustainable and economical pig production. However, behavior is a complex, multifactorial and difficult-to-measure trait. Therefore, influencing a desired behavior is challenging. Defining behavior, then collecting data to accurately reflect differences in a specific behavior, is also difficult. For example, positive behavior of a sow towards her pen mates (group behavior) is defined differently from positive behavior towards her piglets (maternal behavior). Some examples of factors that influence behavior include housing, access to/type of enrichment, diet, feed, water management, climate, health and genetics. Although selection for improved maternal ability and group behavior is already part of the breeding program at Topigs Norsvin, we continue to investigate ways to complement our current breeding strategy.
Collecting data for this purpose is not trivial. In contrast to traits like body weight and back fat, behavior cannot be defined using data collected at a single point in time. Rather, accurate phenotyping of behavior traits requires data collection over extended periods of time, which is either not possible, or not feasible, to collect via visual observation. Cameras, however, can be used to collect a high volume of data on a continuous basis. This dramatically reduces the time and labor needed to collect data on behavior traits, but collecting data is just the beginning of the pipeline. Next steps in this pipeline are related to translating camera data into information that can be used for genetic evaluation. A critical part of this process is training models to automatically extract information from video footage. This information may be as simple as detection of a specific body position or movement, to complex behaviors like tail biting, belly nosing or play behavior.
Topigs Norsvin has data collection pipelines in place at several locations, including farms in Canada and in Europe. In Norway and Germany, cameras are being used to study group behavior. The first major milestone of this project was the ability to identify individual animals and track those same animals over time. These data are being used to study posture traits (i.e. standing, sitting, lying), as a first step towards evaluating differences in more intricate behavior, such as feeding behavior. In Canada, cameras are mounted above farrowing pens at Topigs Norsvin’s new nucleus facility (Innova) to collect data from purebred sows and piglets. These data are being collected in a free farrowing system, enabling sows to express their full genetic potential for maternal ability, which is not obvious in a conventional, crated system. Data is collected from these pens on a continuous basis and used to study maternal behavior, such as changes in posture or time spent maintaining a particular posture. Models have also been developed to detect individual piglets, and are being extended to study differences in piglet behavior from farrowing to weaning.
In summary, behavior is an important aspect of pig production; yet, given the difficulty of defining behavior and collecting representative data, assigning an economic value to behavioral traits is difficult. However, new tools, like cameras, can be used to capture data which are otherwise impossible to obtain. These data are being used to define novel behavior traits, which will add to our current selection strategy to breed pigs for improved maternal ability and group behavior.
Click here to see more...