In May 2017, a two-day workshop was held in Los Angeles (California, U.S.A.) to gather practitioners who work with low-cost sensors used to make air quality measurements. The community of practice included individuals from academia, industry, non-profit groups, community-based organizations, and regulatory agencies. The group gathered to share knowledge developed from a variety of pilot projects in hopes of advancing the collective knowledge about how best to use low-cost air quality sensors. Panel discussion topics included: (1) best practices for deployment and calibration of low-cost sensor systems, (2) data standardization efforts and database design, (3) advances in sensor calibration, data management, and data analysis and visualization, and (4) lessons learned from research/community partnerships to encourage purposeful use of sensors and create change/action. Panel discussions summarized knowledge advances and project successes while also highlighting the questions, unresolved issues, and technological limitations that still remain within the low-cost air quality sensor arena.

Abstract. Volatile organic compounds (VOCs) present a unique challenge in air quality research given their importance to human and environmental health, and their complexity to monitor resulting from the number of possible sources and mixtures. New technologies, such as low-cost air quality sensors, have the potential to support existing air quality measurement methods by providing data in high time and spatial resolution. These higher-resolution data could provide greater insight into specific events, sources, and local variability. Furthermore, given the potential for differences in selectivities for sensors, leveraging multiple sensors in an array format may even be able to provide insight into which VOCs or types of VOCs are present. During the FRAPPE and DISCOVER-AQ monitoring campaigns, our team was able to co-locate two sensor systems, using metal oxide (MOx) VOC sensors, with a proton-transfer-reaction quadrupole mass spectrometer (PTR-QMS) providing speciated VOC data. This dataset provided the opportunity to explore the ability of sensors to estimate specific VOCs and groups of VOCs in real-world conditions, e.g., dynamic temperature and humidity. Moreover, we were able to explore the impact of changing VOC compositions on sensor performance as well as the difference in selectivities of sensors in order to consider how this could be utilized. From this analysis, it seems that systems using multiple VOC sensors are able to provide VOC estimates at ambient levels for specific VOCs or groups of VOCs. It also seems that this performance is fairly robust in changing VOC mixtures, and it was confirmed that there are consistent and useful differences in selectivities between the two MOx sensors studied. While this study was fairly limited in scope, the results suggest that there is the potential for low-cost VOC sensors to support highly resolved ambient hydrocarbon measurements. The availability of this technology could enhance research and monitoring for public health and communities impacted by air toxics, which in turn could support a better understanding of exposure and actions to reduce harmful exposure.

The increased use of low-cost air quality sensor systems, particularly by communities, calls for the further development of best-practices to ensure these systems collect usable data. One area identified as requiring more attention is that of deployment logistics, that is, how to select deployment sites and how to strategically place sensors at these sites. Given that sensors are often placed at homes and businesses, ideal placement is not always possible. Considerations such as convenience, access, aesthetics, and safety are also important. To explore this issue, we placed multiple sensor systems at an existing field site allowing us to examine both neighborhood-level and building-level variability during a concurrent period for CO2 (a primary pollutant) and O3 (a secondary pollutant). In line with previous studies, we found that local and transported emissions as well as thermal differences in sensor systems drive variability, particularly for high-time resolution data. While this level of variability is unlikely to affect data on larger averaging scales, this variability could impact analysis if the user is interested in high-time resolution or examining local sources. However, with thoughtful placement and thorough documentation, high-time resolution data at the neighborhood level has the potential to provide us with entirely new information on local air quality trends and emissions.

An array of low-cost sensors was assembled and tested in a chamber environment wherein several pollutant mixtures were generated. The four classes of sources that were simulated were mobile emissions, biomass burning, natural gas emissions, and gasoline vapors. A two-step regression and classification method was developed and applied to the sensor data from this array. We first applied regression models to estimate the concentrations of several compounds and then classification models trained to use those estimates to identify the presence of each of those sources. The regression models that were used included forms of multiple linear regression, random forests, Gaussian process regression, and neural networks. The regression models with human-interpretable outputs were investigated to understand the utility of each sensor signal. The classification models that were trained included logistic regression, random forests, support vector machines, and neural networks. The best combination of models was determined by maximizing the F1 score on ten-fold cross-validation data. The highest F1 score, as calculated on testing data, was 0.72 and was produced by the combination of a multiple linear regression model utilizing the full array of sensors and a random forest classification model.

No abstract