Data paper: Behavioural and production data of sows fed tailored diets over three consecutive gestations

This database was created to better understand the long-term impact of precision feeding ( PF , i.e. delivery of an individualised mix of different diets to meet estimated individual requirements) on the performances of gestating sows, compared to sows fed a conventional feeding strategy (i.e., delivery of a single diet throughout gestation). Two experiments were carried out with feed supplies to PF sows adjusted individually and daily when they were grouped, based on energy and amino acids requirements (trial 1) and additionally on minerals during the last month of gestation (trial 2). Sows were group-housed from day 3 (trial 1) or day 25 (trial 2) to day 108 of gestation. Data were collected from sows over one to three successive gestations cumulating 58 (trial 1) or 77 (trial 2) gestations. The gestation pens were equipped with electronic feeder and water dispensers to characterise the feeding and drinking behaviours, with an automated scale to characterise the BW and with cameras to characterise the behaviour and location of sows, on a daily and individual basis. An automatic video analysis software analysed sows’ physical activity at the group scale, on three different stages of gestation. Physical activity was also recorded individually with accelerometers for a few numbers of sows. Social interactions, activities, and pen location were characterised at the individual level on specific periods from video manual analysis. Backfat thickness ( BT ) was measured manually with an ultrasound portable device. Cleanliness, scratches and lameness were recorded weekly during gestation. Each gestation room was equipped with devices which enable to follow ambient temperature, relative humidity, and other parameters such as noise or methane concentration. At farrowing, litter size and birthweight were measured. Data were used to analyse the effect of feeding strategies on feeding behaviour, variation of BW and BT during gestation, health status, social interactions when sows were group-housed, and litter characteristics at farrowing. The database contains a large amount of data including feed composition and intakes, behaviours, ambient parameters, and sow performances of group-housed gestating sows enabling to use unsupervised data mining approaches. Longitudinal database may be used to correlate feeding behaviour with physical activity or ambient parameters. Common phenotypes (BW, BT) collected at different key stages of gestation may be used to extend the existing dataset to have a greater statistical power.