Making sense of low cost sensors. Air quality monitoring in Gothenburg Sweden
Examensarbete för masterexamen
Industrial ecology (MPTSE), MSc
Over the recent decade, Emergence and commercialization of the new, low-cost, sensor technologies have created the possibility of major paradigm shifts in air quality monitoring. Their price of three orders of magnitude lower than standard/reference instruments provides the opportunity for new applications such as higher geographical and temporal resolutions of the measurements. There have been studies on the performance of a network of these sensors, however, their individual reliability is still questionable. This study aimed to evaluate the performance of one of the most common low-end sensors available on the market, SDS011, as well as a middle-end sensor, SDS019 under different circumstances such as temperature and humidity. The main research questions were: how reliable are these sensors and what are the causes of errors for these sensors and is it possible to find correction factors based on meteorological data? To address the research questions, a range of experiments in different environments, including field and laboratory, have been conducted under several humidity and temperatures. The results of the experiments illustrated a high linear correlation between the SDS011 and SDS019 sensors with the reference sensor(Optical Particle Sizer) at laboratory experiments. The data were fitted to the reference sensor using a linear regression model. additionally, a multiple linear regression was applied to include the temperature and relative humidity as additional input parameters to the regression model. The results of the multiple and normal regression were compared and discussed under different circumstances for both SDS011 and SDS019 sensors. The field experiments showed significant differences between the SDS011 and reference instruments and these could not be explained by humidity alone. They were not significantly reduced when applying laboratory correction factors either. A three week comparison of the SDS011 against the golden standard for PM in air quality monitoring (TEOM) showed periods with both decent and poor agreement, illustrating that the SDS011 sensors respond to PM but that they are rather unreliable when used as single devices. Further research work is needed to understand this. Nevertheless the sensors are suitable for operation in a network to obtain spatial air quality information , both as stationary and mobile.
air quality monitoring, low-cost sensors , multiple linear regression, signal processing, aerosols , particulate matter , PM2.5, particles , dust, human health.