Some abstracts do not have video files because ASAS was denied recording rights.
875
Simultaneous monitoring of water consumption in eight double pens as a tool for improving welfare and predicting diseases and unwanted behavioral changes in finisher pigs
Increasing animal welfare and heightening the level of management by sensor based monitoring of water consumption in finisher pigs are the overall objectives in this study. It has previously been shown that water monitoring can be used to predict outbreaks of diarrhea in weaner pigs at section level, and diarrhea or unwanted behavioral changes in finisher pigs at double pen level. A double pen is defined as two neighboring pens getting their water supply from the same pipe on which the sensor is placed. In this study the hourly water consumption is measured in a commercial farm by water flow sensors in eight double pens each containing 36 finisher pigs and distributed with two in each of four sections. The eight individual time series are modeled simultaneously in one dynamic linear model (DLM) and variance components are estimated by an EM-algorithm. Insertion dates are not synchronized but follow the production cycle demanding the DLM to handle a varying number of time series at any given time. The diurnal drinking pattern is described by a combination of three harmonic waves (24 h, 12 h, and 8 h wave length) as well as underlying levels and trends for herd, sections, and double pens. Preliminary results indicate a strong correlation between double pens in the same section as well as some pen specific effect.
In this study simultaneous monitoring is used to detect diseases (diarrhea, influenza and respiratory diseases), as well as the unwanted behavioral changes preceding outbreaks of tail bite and fouling. Early warnings can be generated either independently at double-pen level or merged at section level or herd level. This quality can be used as a prioritizing tool minimizing the occurrence of false positive alarms if the warning pattern is highly different from what can be expected based on knowledge of the specific disease or behavioral change. This study is a part of an ongoing project aiming to improve welfare and productivity in growing pigs using advanced ICT methods.
Keywords: Sensor based, early warning, dynamic linear model