Animal-based indicators for the welfare assessment of dairy goats under extensive farming systems

The present thesis focused on the readaptation in pasture conditions of the original AWIN protocol, developed for goats raised in intensive and semi-intensive systems. The feasibility and reliability of both the individual and of the group-level animal-based welfare indicators were evaluated when the readapted protocol was applied by three raters. Specifically, some of the group-level indicators assessed during the readaptation of the protocol were specifically rearranged to properly assess the welfare of goats kept at pasture, but many of them, such as the Familiar Human Approach Test (FHAT), used to assess the Human Animal Relationship (HAR) of goats kept at pasture, needed to be validated in extensive conditions to better improve the feasibility and the reliability for such indicators when assessing the welfare of goats kept outdoors. The validation of FHAT was carried out by checking its convergent validity with the Latency to the First Contact Test (LFCT), which was already developed, validated, and included in the original AWIN protocol to evaluate HAR for dairy goats kept indoors. The LFCT was performed both by unfamiliar and familiar (farmer) assessors. The LFCT time was compared among four different classes of reactions, obtained basing on the reactions of goats towards the farmer at pasture during the application of the test [1. AVOIDANCE (withdrawal of the whole flock from the farmer); 2. APPROACH (at least one goat allows the farmer to approach and/or follows the farmer; no physical interaction); 3. CONTACT (at least one goat voluntarily approaches the farmer; contact lasts ≤3 seconds); 4. ACCEPTANCE (at least one goat accepts to be touched by the farmer; contact lasts >3 seconds)]. When the readapted protocol was applied, the reliability of the individual welfare indicators was analysed through Gwet’s γ(AC1), which demonstrated to be the best agreement index for assessing the concordance in the presence of both two and many raters, and in the presence of variables characterised by any number of categories. Concerning the evaluation of IOR for the group-level welfare indicators in the presence of both two and three raters, the Intra-class correlation coefficient (ICC) is the most documented index. The obtained results showed an excellent reliability for almost all the individual welfare indicators (i.e. Udder asymmetry: P0 = 94%; γ(AC1) = 0.93), but also showed poor reliability for the group-level indicators (i.e. Oblivion; ICC = -0.45) (with FHAT as the only exception), probably due to the adverse climatic conditions encountered in alpine summer ranges, which could have limited the visibility of goats and the overall feasibility of the evaluation, the latter negatively affecting the reliability results. For this reason, some methodological issues still need to be solved for group level indicators. For example, recent Precision Livestock Farming (PLF) tools may hopefully contribute in the near future to provide valid, feasible, and reliable welfare indicators to be included in assessment protocols for these peculiar and welfare-challenging farming systems. In relation to the validation of FHAT, LTFC time resulted higher in the classes without contact (AVOIDANCE and APPROACH) when compared to the classes with contact (CONTACT and ACCEPTANCE). However, the differences among the classes were significant in response to the presence of the familiar assessor, only (P = 0.016). The obtained results support the use of FHAT for the evaluation of HAR in goats raised in pasture-based systems