Urban network of cost-efficient particulate matter measurement devices: Performance against reference observations and scientific benefit

22 cost-efficient (aka low-cost) commercially available particulate matter (PM) measurement devices were installed in a diverse urban area in Leipzig, Germany. The instruments measure mostly PM2.5, some additionally PM10, and are equipped with methods for quality assurance such as conditioning to a defined temperature and regular internal calibration. In order to investigate the spread between the instruments and to enable a pre-campaign calibration, all instruments were setup in the laboratory and the outside air and compared against the same reference measurements. After calibration the measurement network was installed and run for 14 months. It covers roughly 2x2 km2 and holds different urban features like residential and commercial buildings, important main roads, city parks, and small open building gaps. Within the network there is an official air quality monitoring station located directly at a main road. In addition, at two further monitoring stations instruments were installed to study the long-term performance, dependence on meteorological conditions and comparison to reference measurements. The cost-efficient instruments perform generally quite well after the calibration. In particularly for higher PM loads > 10 g m-3 the agreement against references is mostly satisfying, where the relative spread between instruments (while mounted in the same location) is often far below 50 %. Under very high relative humidity (RH > 95 %), which were only observed for cold temperatures during the campaign, reference observations were overestimated. Below this RH threshold no additional deviation between reference and sensors was found, hence, suggesting a stable signal. Overall, the chosen instruments have the potential for application in monitoring of air quality limit values, i.e. the answer the question how frequently are certain limit values exceeded. Furthermore, differences between different local features in the observation area could be observed in e.g., the diurnal cycle but also peak and mean concentrations. Due to the high time resolution (10s raw data), short peak events such as New Years fireworks or summer barbeque can be detected and compared to background conditions at other stations in the network.