Seasonal Field Calibration of Low-Cost PM2.5 Sensors in Different Locations with Different Sources in Thailand

Abstract

Low-cost sensors (LCS) have been increasingly deployed to monitor PM2.5 concentrations. More than 1500 LCS have been installed in Thailand to increase public awareness of air quality. However, performance of these sensors has not been systematically investigated. In this study, PM2.5 LCS were co-located next to a PM2.5 federal equivalent method (FEM) reference instrument at three Thai locations—in the north, center and northeast. We evaluated the performance of a PM2.5 LCS (PMS7003, Plantower) to understand the key factors affecting performance, including emission sources, relative humidity, temperature and PM2.5 concentration. Low PM concentration and high humidity levels had a significant impact on performance. Sensors in a high traffic emission area showed low correlation. The unadjusted PM2.5 LCS performance varied with locations. Errors were mainly observed at low concentrations. They significantly underestimated concentrations in congested urban environments. After calibration, accuracy was improved with multiple regression models. The performance of sensors only at Chiang Mai (CM) during the dry season and Ubon Ratchathani (URT) during the dry and wet seasons were acceptable with coefficient of variation: 5.8 ± 4.7–6.8 ± 5.0%, slope: 0.829–0.945, intercept: 1.12–5.49 µg/m3, R2: 0.880–0.934 and RMSE: 4.3–5.1 µg/m3. In the congested area in Bangkok (BKK), they underestimated concentrations of small particles.

AirMLP: A Multilayer Perceptron Neural Network for Temporal Correction of PM2.5 Values in Turin

“…By accounting for both seasonality and temporal dynamics, the models can become more adaptive and provide accurate corrections across various environmental scenarios. This approach aligns with the idea of developing context-aware models that tailor their corrections based on the prevailing conditions [64], ultimately advancing the effectiveness of low-cost sensor data in air quality monitoring.…”

“…The hyperparameter values under consideration were as follows: the number of neurons per layer spanned from [300, 500, 700], the number of consecutive records fluctuated between [6,12,20], and batch sizes covered [64,256,512]. We assessed these hyperparameter combinations across five distinct architectures, resulting in a total of 135 potential configurations, each requiring training and evaluation.…”

Using a Low-Cost Sensor to Estimate Fine Particulate Matter: A Case Study in Samutprakarn, Thailand

“…The average concentration of PM 2.5 in Thailand was attributed to a high level in the dry season. Moreover, Chunitiphisan et al [44] revealed that the ambient PM 2.5 in Northern Thailand had considerably increased, and the mass level in the wet season was found to be lower than that in the dry season at the sampling place [33,64].…”

“…Air temperature had a strong effect on PM 2.5 in the winter and had a weak effect on PM 2.5 in summer [33]. Dejchanchaiwong et al [64] found that low PM concentrations and high RH levels had a substantial impact on LCS performance. Specifically, during the wet season, LCS showed higher relative inaccuracy than during the dry season.…”