Validation and fine-tuning of a predictive model for air quality in livestock buildings

Abstract

Between 1997 and 1999 a comprehensive survey of airborne pollutant concentrations within, and emissions from, 160 Australian piggery buildings was undertaken. The primary aim of the study was to model the concentrations and emissions of different airborne pollutants and thus identify potential reduction techniques. The main factors identified as significantly influencing airborne pollutant concentrations were building type, pen hygiene, pig flow management, seasons, building volume, ventilation airflow rate, air temperature, relative humidity, and farm size (as expressed by number of sows on site). These effects were included in comprehensive statistical models to explain the variation in measured concentrations and emission rates. The models developed were validated and fine-tuned using the “leave-one-out” cross-validation technique. This article details the validation technique used that was aimed to maximise the value of available experimental data and further improves the practicality of the models developed. The main result of the study was the development of an improved model. The study results, and the resulting prediction models, should help improve air quality in piggery buildings by providing decision makers with an awareness of the environmental conditions inside and outside of piggery buildings. In turn, that should lead to improvements in the health and welfare of pigs and piggery staff and the sustainability of the piggery operations.