Guideline to validate sensor output of animal-based measurements

Sensor technology is used in precision livestock farming applications to monitor amongst others animal health
and welfare. However, in order to effectively use sensor data for welfare monitoring, the sensors need to provide valid and reliable measurements and the measured data needs to be relevant for animal welfare. Thus, sensor validation is crucial for applications in commercial and research settings. For this validation process, there is a need for a standardized procedure, so that the performance of validation trials are comparable and can lead to a reliable outcome. In the context of the aWISH (HEurope, 101060818) and Clearfarm (H2020, 862919) projects, a guideline is set up to develop a standardized protocol for sensor validation, divided into steps organized into two main blocks. Block 1 is dedicated to output validation, meaning whether the sensor is measuring the output it is designed to measure accurately. Block 2 is dedicated to welfare relevance validation, meaning whether sensor output can provide information regarding animal welfare and can be used as an animal welfare indicator. This presentation will focus mainly on Block 1, which includes different steps: (1) description of the sensor output, (2) determination of the gold standard and reference output, (3) determination of sample size and statistical analysis, (4) limitations of the technology. The guidelines will be illustrated using a number of case studies in cattle, pigs and poultry covering cameras, microphones, wearables, load cells and a combination of sensors.