Field Evaluation of Low Cost Sensors Array for Air Pollution Monitoring

In recent years, many laboratory studies show that low cost systems based on electrochemical and metal oxide gas sensors can be a promising alternative to densify the actual air pollution monitoring networks. However, real environmental conditions have an impact on the effectiveness of these sensors. Therefore, measurements carried in laboratory and those in field are different. In this paper we investigate whether low cost sensors keep providing good performances in field under real environmental conditions. We use an electronic nose containing one electrochemical sensor and one metal oxide gas sensor dedicated to monitor nitrogen dioxide. The results show that electrochemical sensors keep providing a good performance while the metal oxide sensors are affected by other interfering gases. We present also how electrochemical sensors could be combined to calculate the ozone concentration. We highlight also the advantage of using machine learning algorithms such as support vector machine regression (SVR) to improve significantly the concentration estimation accuracy by combining data from all sensors.