Environmental monitoring and disease prediction

Abstract Suboptimal environmental conditions impair health and performance in swine farms. Nowadays, it is possible to use wireless sensor technology to monitor key environmental indicators, including barn temperature, CO2 concentration, relative humidity, and water consumption. This chapter provides information about the determination of the role of variability in the physical environment on the temporal expression of production diseases. To understand variation over time within a farm environment, a system to monitor key environmental indicators (temperature, water consumption, relative humidity, and CO2 concentration) real-time has been developed. Moreover, information on the incidence and prevalence of the respiratory disease and mortality, medication, and vaccination use were collected from daily health surveys using a digital pen and were processed automatically. The chapter describes how health and sensor data can be combined with productive data into a single dataset that was able to be used for training two statistical models applied for respiratory disease. The first methodology used was logistic regression, a classical statistical method by which it was possible to determine significant relationships between temperature and CO2, and the presence of respiratory disease. The second methodology was a novel deep learning approach which can be utilised to handle incomplete and noisy raw data and detect the onset of anomalous environmental conditions in farms that may relate to onset of respiratory disease. It is concluded that changes in environmental conditions have a management value as they are predictors of disease. Development of methods to capture farm data and model to utilise them are likely to increase our ability to predict and hence manage the onset of disease on pig and poultry farms.