On dairy farms, management of calving is important for the health of dairy cows and the survival rate of calves born. Although an expected calving date is known, farmers need to check their cows regularly to estimate the moment when a cow will start calving. A sensor system which predicts the moment of calving could help farmers to check cows effectively for the occurrence of dystocia. In this study, a total of 450 cows on two farms were equipped with Agis SensOor sensors (Agis Automatisering B.V., Harmelen, the Netherlands), which measure rumination activity, activity and temperature hourly. Data were collected over a one-year period. During that period, the exact moment of 417 calvings was recorded using camera images of the calving pen taken every 5 minutes. In total 110 calvings could be linked with sensor data. The moment when calving started was defined as the hour in which the camera images showed the cow having contractions or labour initially started. Two logit models were developed: a reduced model with the expected calving date as the independent variable and a full model which additionally included independent variables based on sensor data. The areas under the Receiver Operating Characteristic curves were 0.682 and 0.878 for the reduced and full model with, at a false positive rate of 10%, sensitivities of 22 and 69%, respectively. Results indicated that the inclusion of sensor data improved prediction of the start of calving and thus that the sensor data used have some potential for predicting the moment of calving.
Keywords: sensor technology, calving management, dairy farming
