Predictive modelling for dairy cow welfare assessment

Animal welfare is a rather complex concept. It includes various different facets that make it difficult for scientists to produce an ultimate definition. Leaving out ethical discussions, science aims at finding key concepts that could work as objective indicators for the welfare status of farmed animals. Among the commonly accepted indicators, disease presence in dairy farms is of main importance. The negative impact of diseases on animal welfare and productive performances is straightforward. Recently, however, great attention has been given to the environmental conditions where the animals are reared. Specifically, the heat stress issue has been addressed in different papers available in scientific literature. In this PhD thesis, rumination fills a key role in animal welfare definition. This physiological process, which differentiates and characterises ruminants from the rest of the mammals, has been thoroughly studied. Mechanisms involved in rumination are well known, as well as the diseases that strike the rumen. However, rumination is still poorly considered in the general context of welfare in dairy cows. In the papers produced during this PhD project and presented in Part II (some of them already published on international journals), rumination relationships with the main factors influencing animal welfare were studied. In the first paper, heat stress effects on rumination time were described. The paper shows that rumination decreases when discomfort caused by heat stress increases. In the second paper, rumination was hypothesised as a predictor for various diseases groups, commonly affecting dairy farms. Other results are presented as drafts of papers that will be submitted to international journals as soon as ready. The first draft explores the heritability of rumination and its genetic correlation with milk production traits. Only few studies can be found in literature about rumination heritability, and none of them evaluated the same genetic relationships we investigated. Lastly, in the second draft a first rough attempt to build a predictive model for diseases prediction was made. Fine tuning is still needed to improve the obtained results. Minor results are presented as abstracts (oral presentations at two international meetings) in the last section of Part II.