Performance characterization of low-cost air quality sensors for off-grid deployment in rural Malawi

Abstract

Abstract. 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, Ox, 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.

Long-term Evaluation of Commercial Air Quality Sensors: An Overview from the QUANT Study

“…The accuracy of the reported data for a particular device can be impacted by a variety of factors, from the calibrations applied to its location or seasonality. This has important implications for the way sensorbased technologies are deployed and supports the case made by others (Bittner et al, 2022;Farquhar et al, 2021;Crilley et al, 2018;Williams, 2020;Bi et al, 2020) that practical methods to monitor sensor bias will be crucial in uses where data accuracy is paramount.…”

“…Numerous studies have investigated the effectiveness of calibration methods for sensors e.g. (Bigi et al, 2018;Bittner et al, 2022;Malings et al, 2020;Spinelle et al, 2017;Zimmerman et al, 2018), including selecting appropriate reference instruments (Kelly et al, 2017), the need for regular calibration to maintain accuracy (Gamboa et al, 2023), the necessity of rigorous calibration protocols to ensure consistency (Kang et al, 2022), and transferability (Nowack et al, 2021) of results. Ultimately, the reliability and associated uncertainty of any applied calibration will influence the final sensor data quality.…”

Study of the Suitability of a Personal Exposure Monitor to Assess Air Quality

“…It is important to note that the ARISense is a low-cost sensor; measurement errors in this instrument are possible [30].…”

Nanostructural evolution of hydrothermally grown SrTiO3 perovskite and its implementation in gaseous phase detection of ethanol

“…However, it was noted that the response of the sensor was a near-linear relationship with ethanol concentration varying from 0.5 to 200 ppm (figure 7(e)). The linear regression equations for S1, S2, S3 and S4 are shown in figure 7(e) with R 2 values >0.92, 0.91, 0.89 and 0.933, respectively [57]. The theoretical detection limit was calculated by using the slope of the logarithmic plot (figure 7(e)) and standard deviation from the baseline of figures 7(a)-(d) at 150 °C [58,59].…”