Research Priorities of Applying Low-Cost PM2.5 Sensors in Southeast Asian Countries

The low-cost and easy-to-use nature of rapidly developed PM_2.5 sensors provide an opportunity to bring breakthroughs in PM_2.5 research to resource-limited countries in Southeast Asia (SEA). This review provides an evaluation of the currently available literature and identifies research priorities in applying low-cost sensors (LCS) in PM_2.5 environmental and health research in SEA. The research priority is an outcome of a series of participatory workshops under the umbrella of the International Global Atmospheric Chemistry Project–Monsoon Asia and Oceania Networking Group (IGAC–MANGO). A literature review and research prioritization are conducted with a transdisciplinary perspective of providing useful scientific evidence in assisting authorities in formulating targeted strategies to reduce severe PM_2.5 pollution and health risks in this region. The PM_2.5 research gaps that could be filled by LCS application are identified in five categories: source evaluation, especially for the distinctive sources in the SEA countries; hot spot investigation; peak exposure assessment; exposure–health evaluation on acute health impacts; and short-term standards. The affordability of LCS, methodology transferability, international collaboration, and stakeholder engagement are keys to success in such transdisciplinary PM_2.5 research. Unique contributions to the international science community and challenges with LCS application in PM_2.5 research in SEA are also discussed.

Correction: Panel study using novel sensing devices to assess associations of PM2.5 with heart rate variability and exposure sources

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Panel study using novel sensing devices to assess associations of PM2.5 with heart rate variability and exposure sources

BACKGROUND/OBJECTIVE

This work applied a newly developed low-cost sensing (LCS) device (AS-LUNG-P) and a certified medical LCS device (Rooti RX) to assessing PM_2.5 impacts on heart rate variability (HRV) and determining important exposure sources, with less inconvenience to subjects.

METHODS

Observations using AS-LUNG-P were corrected by side-by-side comparison with GRIMM instruments. Thirty-six nonsmoking healthy subjects aged 20-65 years were wearing AS-LUNG-P and Rooti RX for 2-4 days in both Summer and Winter in Taiwan.

RESULTS

PM_2.5 exposures were 12.6 ± 8.9 µg/m³. After adjusting for confounding factors using the general additive mixed model, the standard deviations of all normal to normal intervals reduced by 3.68% (95% confidence level (CI) = 3.06-4.29%) and the ratios of low-frequency power to high-frequency power increased by 3.86% (CI = 2.74-4.99%) for an IQR of 10.7 µg/m³ PM_2.5, with impacts lasting for 4.5-5 h. The top three exposure sources were environmental tobacco smoke, incense burning, and cooking, contributing PM_2.5 increase of 8.53, 5.85, and 3.52 µg/m³, respectively, during 30-min intervals.

SIGNIFICANCE

This is a pioneer in demonstrating application of novel LCS devices to assessing close-to-reality PM_2.5 exposure and exposure-health relationships. Significant HRV changes were observed in healthy adults even at low PM_2.5 levels.

Application of Machine Learning for the in-Field Correction of a PM2.5 Low-Cost Sensor Network

Many low-cost sensors (LCSs) are distributed for air monitoring without any rigorous calibrations. This work applies machine learning with PM_2.5 from Taiwan monitoring stations to conduct in-field corrections on a network of 39 PM_2.5 LCSs from July 2017 to December 2018. Three candidate models were evaluated: Multiple linear regression (MLR), support vector regression (SVR), and random forest regression (RFR). The model-corrected PM_2.5 levels were compared with those of GRIMM-calibrated PM_2.5. RFR was superior to MLR and SVR in its correction accuracy and computing efficiency. Compared to SVR, the root mean square errors (RMSEs) of RFR were 35% and 85% lower for the training and validation sets, respectively, and the computational speed was 35 times faster. An RFR with 300 decision trees was chosen as the optimal setting considering both the correction performance and the modeling time. An RFR with a nighttime pattern was established as the optimal correction model, and the RMSEs were 5.9 ± 2.0 μg/m³, reduced from 18.4 ± 6.5 μg/m³ before correction. This is the first work to correct LCSs at locations without monitoring stations, validated using laboratory-calibrated data. Similar models could be established in other countries to greatly enhance the usefulness of their PM_2.5 sensor networks.

Fusion of Environmental Sensing on PM2.5 and Deep Learning on Vehicle Detecting for Acquiring Roadside PM2.5 Concentration Increments

Traffic emission is one of the major contributors to urban PM2.5, an important environmental health hazard. Estimating roadside PM2.5 concentration increments (above background levels) due to vehicles would assist in understanding pedestrians' actual exposures. This work combines PM2.5 sensing and vehicle detecting to acquire roadside PM2.5 concentration increments due to vehicles. An automatic traffic analysis system (YOLOv3-tiny-3l) was applied to simultaneously detect and track vehicles with deep learning and traditional optical flow techniques, respectively, from governmental cameras that have low resolutions of only 352 × 240 pixels. Evaluation with 20% of the 2439 manually labeled images from 23 cameras showed that this system has 87% and 84% of the precision and recall rates, respectively, for five types of vehicles, namely, sedan, motorcycle, bus, truck, and trailer. By fusing the research-grade observations from PM2.5 sensors installed at two roadside locations with vehicle counts from the nearby governmental cameras analyzed by YOLOv3-tiny-3l, roadside PM2.5 concentration increments due to on-road sedans were estimated to be 0.0027-0.0050 µg/m3. This practical and low-cost method can be further applied in other countries to assess the impacts of vehicles on roadside PM2.5 concentrations.

Laboratory Evaluations of Correction Equations with Multiple Choices for Seed Low-Cost Particle Sensing Devices in Sensor Networks

To tackle the challenge of the data accuracy issues of low-cost sensors (LCSs), the objective of this work was to obtain robust correction equations to convert LCS signals into data comparable to that of research-grade instruments using side-by-side comparisons. Limited sets of seed LCS devices, after laboratory evaluations, can be installed strategically in areas of interest without official monitoring stations to enable reading adjustments of other uncalibrated LCS devices to enhance the data quality of sensor networks. The robustness of these equations for LCS devices (AS-LUNG with PMS3003 sensor) under a hood and a chamber with two different burnt materials and before and after 1.5 years of field campaigns were evaluated. Correction equations with incense or mosquito coils burning inside a chamber with segmented regressions had a high R2 of 0.999, less than 6.0% variability in the slopes, and a mean RMSE of 1.18 µg/m3 for 0.1-200 µg/m3 of PM2.5, with a slightly higher RMSE for 0.1-400 µg/m3 compared to EDM-180. Similar results were obtained for PM1, with an upper limit of 200 µg/m3. Sensor signals drifted 19-24% after 1.5 years in the field. Practical recommendations are given to obtain equations for Federal-Equivalent-Method-comparable measurements considering variability and cost.

A versatile low-cost sensing device for assessing PM2.5 spatiotemporal variation and quantifying source contribution

This study evaluated a newly developed sensing device, AS-LUNG-O, against a research-grade GRIMM in laboratory and ambient conditions and used AS-LUNG-O to assess PM_2.5 spatiotemporal variations at street levels of an Asian mountain community, which represented residents' exposure (at the interface of atmosphere and human bodies leading to potential health impacts). In laboratory, R² of 1-min AS-LUNG-O and GRIMM was 0.95 ± 0.04 (n = 64,179 for 40 sets). After conversion with individual correction equations, their correlation in ambient tests was 0.93 ± 0.05, with absolute % difference of only 10 ± 9%. Ten AS-LUNG-O sets were installed at street sites with another one at 10 m above ground on July 1-28 and December 2-31, 2017 in Nantou, Taiwan. Important source contributions to PM_2.5 were quantified with regression analysis. Temporal variation expressed as the daily max/mean of 5-min PM_2.5 reached 13.7 in July and 12.2 in December. Spatial variation expressed as the percent coefficients of variance (%CV) across ten community locations was 22% ± 20% (max: 199%) in July and 19 ± 18% (max: 206%) in December. Incremental contribution from the stop-and-go traffic, market, temple, and fried-chicken vendor to PM_2.5 at 3-5 m away were 4.38, 3.90, 2.72, and 1.80 μg/m³, respectively. Significant spatiotemporal variations and community source contributions revealed the importance of assessing neighborhood air quality for public health protection. For long-term air quality monitoring, the percentage of available power and signals of G-sensor provided indicative information of maintenance required. Advantages of low cost (USD 650), small size, light weight, solar power supply, backup data storage, waterproof housing, multiple-sensor flexibility, and high precision and accuracy (after correction) enable AS-LUNG-O to be widely applied in environmental studies.