Behavioural complexity in fattening pigs I. Video-based validation of a novel welfare indicator according to varying intensity of housing conditions

The interest in individual animal welfare has increased in recent years. One of the reasons is a growing awareness of the importance of animals’ individual differences in their individual capability to adapt to their environment. However, animal- and individual-based welfare indicators are often time-consuming to record and it is challenging to combine several indicators into a meaningful measure for welfare assessment. Behavioural complexity, including previously proposed features (e.g. number of behaviours, Shannon diversity index) as well as new features (e.g. variances of bout durations within and between behavioural transitions), has the theoretical potential to reflect animal welfare in a valid and relatively simple way. For a first practical validation of the complexity measure, 3912 hours of continuous video data from 163 fattening pigs - recorded on 85 days in three age classes, two seasons, and four housing levels of different intensities - were analysed. Twelve complexity-features were reduced using principal component analysis and the resulting principal components (PCs) were analysed using linear mixed effects models. General complexity (PC 1) was on average higher with higher housing levels (p [housing level] = 0.007; p [global] = 0.003). This pattern was modulated by age, with older pigs having a lower general complexity and a more pronounced relationship with housing levels (p [housing level x age class] = 0.009) but not by season (main effect and interactions: p > 0.422). The transition variability (PC 2) was on average also higher with higher housing levels (p [housing level] = 0.008; p [global] = 0.149) and neither age nor season had a statistically supported effect (p > 0.10). The Shannon diversity index for the frequency of behaviours as possible simplified measure showed the same pattern as the general complexity Moreover, this Shannon-diversity index had a more pronounced relationship with housing level in winter than in summer (p [housing level x season] = 0.012). These results suggest, that behavioural complexity could potentially be used as an individual-based indicator to detect changes in animal welfare if measured longitudinal in order to take age and season into account. Further advances in the automated analysis of video data are needed for practical use on farms. Additionally, further studies will need to correlate individual behavioural complexity against other individual measures of welfare. • Increased interest in individual animal welfare due to adaptability differences. • Behavioural complexity shows potential for valid, simple welfare measurement. • Higher behavioural complexity observed in more animal-friendly housing systems. • Age and season had modulating effect on behavioural complexity. • Automation of video data analysis is crucial for practical farm applications.