Validation guidelines of sensors used for the automatic assessment of animal welfare

The use of sensor technology to monitor animal welfare continuously is of interest to farmers and other stakeholders. However, for sensor data to be used effectively in welfare monitoring, it is imperative that sensors provide valid and reliable measurements, and sensor data needs to be relevant to animal welfare. Therefore, the validation of sensor outputs is paramount in both commercial and research contexts. Additionally, the establishment of a standardized validation procedure is necessary to ensure the comparability of validation trials and the attainment of reliable outcomes. Within the framework of the aWISH (HEurope, 101060818) and ClearFarm (H2020, 862919) projects, a guideline has been devised to formulate a standardized protocol for sensor validation in welfare assessment, structured into steps organized under two principal blocks. Block 1 pertains to output validation, assessing the accuracy of sensor measurements with respect to their intended outputs, while Block 2 concerns welfare relevance validation, evaluating whether sensor outputs can serve as informative and trustful indicators of animal welfare. This abstract predominantly focuses on Block 2, which encompasses four sequential steps: (1) selection of a welfare consequence under a specific context, (2) identification of validated indicator(s) relevant to the chosen welfare consequence, (3) refinement of validated welfare indicator(s) to align with the outputs of relevant sensors identified in Block 1, and (4) determination of sample size and statistical analysis for the validation trial. Additionally, the limitations of this guideline are addressed, delineating scenarios where its applicability may be constrained. Illustrative case studies involving cattle, pigs, and poultry will demonstrate all steps of this guideline, encompassing a spectrum of prevalent sensor technologies including cameras, microphones, load cells, and a combination of sensors.