Reliability Assessment of Low-Cost PM Sensors under High Humidity and High PM Level Outdoor Conditions

Abstract

Low-cost particulate matter (PM) sensors have become increasingly popular due to their compact size, low power consumption, and cost-effective installation and maintenance. While several studies have explored the effects of meteorological conditions and pollution exposure on low-cost sensor (LCS) performance, few have addressed the combined impact of high PM concentration and high humidity levels. In contrast to most evaluation studies, which generally report $\text{PM}_{2.5}$ levels below $150~μ\text{g/m}^3$, our study observed hourly average $\text{PM}_{2.5}$ concentrations ranging from $6-611~μ\text{g/m}^3$ (mean value of $137~μ\text{g/m}^3$), with relative humidity between $25-95\%$ (mean value of $72\%$), and temperature varying from $6-29^\circ$C (mean value of $16^\circ$C). We evaluate three LCS models (SPS30, PMS7003, HPMA115C0-004) in outdoor conditions during the winter season in New Delhi, India, deployed alongside a reference-grade beta attenuation monitor (BAM). The results indicate a strong correlation between LCS and BAM measurements (${R^2} > 90\%$). The RMSE increases with increasing PM concentration and humidity levels but the narrow $95\%$ confidence interval range of LCS as a function of the reference BAM suggests the importance of LCS in air pollution monitoring. Among the evaluated LCS models, SPS30 showed the highest overall accuracy. Overall, the study demonstrates that LCS can effectively monitor air quality in regions with high PM and high humidity levels, provided appropriate correction models are applied.

Figures (7)

• Figure S1: Time-series plots depicting various parameters: (a) Reference BAM $\text{PM}_{2.5}$ concentration, (b) Sensirion $\text{PM}_{2.5}$, (c) Plantower $\text{PM}_{2.5}$, (d) Honeywell $\text{PM}_{2.5}$, (e) Humidity (%), (f) Temperature ($^\circ \text{C}$). Missing data points are attributed to power failure and maintenance.
• Figure S2: Mean absolute error (MAE) for hourly averaged granular data. With higher humidity levels under the same exposure, as well as higher $\text{PM}_{2.5}$ levels with similar humidity levels, an increase in MAE is observed.
• Figure S3: Sensirion-3 unit after being opened. Red circle shows the region where cobweb was found inside the Sensirion-3 unit.
• Figure S4: Scatterplot between LCS and reference BAM for daily averaged data. The solid red line represents the linear regression fit to the daily average data points, with shaded regions indicating the $95\%$ confidence interval. Additionally, parameters such as linearity and error are displayed within each subplot corresponding to the LCS-reference pair.
• Figure S5: Time-series plot of reference BAM ($\text{PM}_{2.5}$), humidity and temperature based on day of the week. The subplots show Weekly, hourly, and monthly variation of these parameters.