Low-cost sensors for the measurement of fine particulate matter mass (PM_2.5) enable dense networks to increase the spatial resolution of air quality monitoring. However, these sensors are affected by environmental factors such as temperature and humidity and their effects on ambient aerosol, which must be accounted for to improve the in-field accuracy of these sensors. We conducted long-term tests of two low-cost PM_2.5 sensors: Met-One NPM and PurpleAir PA-II units. We found a high level of self-consistency within each sensor type after testing 25 NPM and 9 PurpleAir units. We developed two types of corrections for the low-cost sensor measurements to better match regulatory-grade data. The first correction accounts for aerosol hygroscopic growth using particle composition and corrects for particle mass below the optical sensor size cut-point by collocation with reference Beta Attenuation Monitors (BAM). A second, fully-empirical correction uses linear or quadratic functions of environmental variables based on the same collocation dataset. The two models yielded comparable improvements over raw measurements. Sensor performance was assessed for two use cases: improving community awareness of air quality with short-term semi-quantitative comparisons of sites and providing long-term reasonably quantitative information for health impact studies. For the short-term case, both sensors provided reasonably accurate concentration information (mean absolute error of ∼4 µg/m³) in near-real time. For the long-term case, tested using year-long collocations at one urban background and one near-source site, error in the annual average was reduced below 1 µg/m³. Hence, these sensors can supplement sparse networks of regulatory-grade instruments, perform high-density neighborhood-scale monitoring, and be used to better understand spatial patterns and temporal air quality trends across urban areas.

Assessing the intracity spatial distribution and temporal variability in air quality can be facilitated by a dense network of monitoring stations. However, the cost of implementing such a network can be prohibitive if traditional high-quality, expensive monitoring systems are used. To this end, the Real-time Affordable Multi-Pollutant (RAMP) monitor has been developed, which can measure up to five gases including the criteria pollutant gases carbon monoxide (CO), nitrogen dioxide (NO₂), and ozone (O₃), along with temperature and relative humidity. This study compares various algorithms to calibrate the RAMP measurements including linear and quadratic regression, clustering, neural networks, Gaussian processes, and hybrid random forest–linear regression models. Using data collected by almost 70 RAMP monitors over periods ranging up to 18 months, we recommend the use of limited quadratic regression calibration models for CO, neural network models for NO, and hybrid models for NO₂ and O₃ for any low-cost monitor using electrochemical sensors similar to those of the RAMP. Furthermore, generalized calibration models may be used instead of individual models with only a small reduction in overall performance. Generalized models also transfer better when the RAMP is deployed to other locations. For long-term deployments, it is recommended that model performance be re-evaluated and new models developed periodically, due to the noticeable change in performance over periods of a year or more. This makes generalized calibration models even more useful since only a subset of deployed monitors are needed to build these new models. These results will help guide future efforts in the calibration and use of low-cost sensor systems worldwide.

Air quality monitoring has traditionally been conducted using sparsely distributed, expensive reference monitors. To understand variations in PM_2.5 on a finely resolved spatiotemporal scale a dense network of over 40 low-cost monitors was deployed throughout and around Pittsburgh, Pennsylvania, USA. Monitor locations covered a wide range of site types with varying traffic and restaurant density, varying influences from local sources, and varying socioeconomic (environmental justice, EJ) characteristics. Variability between and within site groupings was observed. Concentrations were higher near the source-influenced sites than the Urban or Suburban Residential sites. Gaseous pollutants (NO₂ and SO₂) were used to differentiate between traffic (higher NO₂ concentrations) and industrial (higher SO₂ concentrations) sources of PM_2.5. Statistical analysis proved these differences to be significant (coefficient of divergence > 0.2). The highest mean PM_2.5 concentrations were measured downwind (east) of the two industrial facilities while background level PM_2.5 concentrations were measured at similar distances upwind (west) of the point sources. Socioeconomic factors, including the fraction of non-white population and fraction of population living under the poverty line, were not correlated with increases in PM_2.5 or NO₂ concentration. The analysis conducted here highlights differences in PM_2.5 concentration within site groupings that have similar land use thus demonstrating the utility of a dense sensor network. Our network captures temporospatial pollutant patterns that sparse regulatory networks cannot.

Low-cost gas and particulate matter sensor packages offer a compact, lightweight, and easily transportable solution to address global gaps in air quality (AQ) observations. However, regions that would benefit most from widespread deployment of low-cost AQ monitors often lack the reference-grade equipment required to reliably calibrate and validate them. In this study, we explore approaches to calibrating and validating three integrated sensor packages before a 1-year deployment to rural Malawi using colocation data collected at a regulatory site in North Carolina, USA. We compare the performance of five computational modeling approaches to calibrate the electrochemical gas sensors: k-nearest neighbors (kNN) hybrid, random forest (RF) hybrid, high-dimensional model representation (HDMR), multilinear regression (MLR), and quadratic regression (QR). For the CO, O_x, NO, and NO₂ sensors, we found that kNN hybrid models returned the highest coefficients of determination and lowest error metrics when validated. Hybrid models were also the most transferable approach when applied to deployment data collected in Malawi. We compared kNN hybrid calibrated CO observations from two regions in Malawi to remote sensing data and found qualitative agreement in spatial and annual trends. However, ARISense monthly mean surface observations were 2 to 4 times higher than the remote sensing data, partly due to proximity to residential biomass combustion activity not resolved by satellite imaging. We also compared the performance of the integrated Alphasense OPC-N2 optical particle counter to a filter-corrected nephelometer using colocation data collected at one of our deployment sites in Malawi. We found the performance of the OPC-N2 varied widely with environmental conditions, with the worst performance associated with high relative humidity (RH >70 %) conditions and influence from emissions from nearby residential biomass combustion. We did not find obvious evidence of systematic sensor performance decay after the 1-year deployment to Malawi. Data recovery (30 %–80 %) varied by sensor and season and was limited by insufficient power and access to resources at the remote deployment sites. Future low-cost sensor deployments to rural, low-income settings would benefit from adaptable power systems, standardized sensor calibration methodologies, and increased regional regulatory-grade monitoring infrastructure.