Uri Lerner, Barak Fishbain
Portable air-quality (AQ) Micro-Sensing Units (MSU) allow the utilization of AQ monitoring in exposure analysis, mobile measurements and citizen science. Typically, the accuracy of these devices is assessed using the mean error or correlation coefficients with respect to laboratory equipment. However, these criteria do not represent how such sensors would perform outside the lab for different large-scale field applications. The aim of this project is to develop algorithmic tools to evaluate data and performance of sensor nodes within a distributed air quality sensing network, focusing on mobile measurements. In Year 5, we developed a comprehensive Sensor Evaluation Toolbox (SET) for comparing pairs of AQ sensors by an integrated score based on a range of criteria. The criteria include root mean square difference (RMSE); correlation coefficients (Pearson, Kindell and Spearman) calculated for time series and for pollutant bi-variate polar plots; and lower frequencies energy of the time series. The SET scheme correspond to the sensors’ reliability; capability to locate pollution sources; ability to represent the pollution level on a coarse scale; and capability to capture the high temporal variability of the observed pollutant. The SET toolbox was applied to measurements acquired by small AQ sensors deployed in eight cities across Europe showing that the suggested scheme indeed facilitates a comprehensive cross platform analysis that can be used to determine sensors’ performance envelope and its favorable working conditions. The SET was implemented in R and uploaded to Barak Fishbain’s website, and a manuscript has been submitted recently. Another manuscript discussing the possible impact of velocity on the function of vehicle-mounted air-quality sensors, based on Year 4 field and wind tunnel campaigns, has been published in a peer reviewed journal (Lerner et al. 2015, see Year 4 report for more details).