Improving accuracy of air pollution exposure measurements: Statistical correction of a municipal low-cost airborne particulate matter sensor network

Calibration Methods for Low-Cost Particulate Matter Sensors Considering Seasonal Variability

Many countries use low-cost sensors for high-resolution monitoring of particulate matter (PM2.5 and PM10) to manage public health. To enhance the accuracy of low-cost sensors, studies have been conducted to calibrate them considering environmental variables. Previous studies have considered various variables to calibrate seasonal variations in the PM concentration but have limitations in properly accounting for seasonal variability. This study considered the meridian altitude to account for seasonal variations in the PM concentration. In the PM10 calibration, we considered the calibrated PM2.5 as a subset of PM10. To validate the proposed methodology, we used the feedforward neural network, support vector machine, generalized additive model, and stepwise linear regression algorithms to analyze the results for different combinations of input variables. The inclusion of the meridian altitude enhanced the accuracy and explanatory power of the calibration model. For PM2.5, the combination of relative humidity, temperature, and meridian altitude yielded the best performance, with an average R2 of 0.93 and root mean square error of 5.6 µg/m3. For PM10, the average mean absolute percentage error decreased from 27.41% to 18.55% when considering the meridian altitude and further decreased to 15.35% when calibrated PM2.5 was added.

A comprehensive review of the development of land use regression approaches for modeling spatiotemporal variations of ambient air pollution: A perspective from 2011 to 2023

Land use regression (LUR) models are widely used in epidemiological and environmental studies to estimate humans' exposure to air pollution within urban areas. However, the early models, developed using linear regressions and data from fixed monitoring stations and passive sampling, were primarily designed to model traditional and criteria air pollutants and had limitations in capturing high-resolution spatiotemporal variations of air pollution. Over the past decade, there has been a notable development of multi-source observations from low-cost monitors, mobile monitoring, and satellites, in conjunction with the integration of advanced statistical methods and spatially and temporally dynamic predictors, which have facilitated significant expansion and advancement of LUR approaches. This paper reviews and synthesizes the recent advances in LUR approaches from the perspectives of the changes in air quality data acquisition, novel predictor variables, advances in model-developing approaches, improvements in validation methods, model transferability, and modeling software as reported in 155 LUR studies published between 2011 and 2023. We demonstrate that these developments have enabled LUR models to be developed for larger study areas and encompass a wider range of criteria and unregulated air pollutants. LUR models in the conventional spatial structure have been complemented by more complex spatiotemporal structures. Compared with linear models, advanced statistical methods yield better predictions when handling data with complex relationships and interactions. Finally, this study explores new developments, identifies potential pathways for further breakthroughs in LUR methodologies, and proposes future research directions. In this context, LUR approaches have the potential to make a significant contribution to future efforts to model the patterns of long- and short-term exposure of urban populations to air pollution.

A DYNAMIC SPATIAL FILTERING APPROACH TO MITIGATE UNDERESTIMATION BIAS IN FIELD CALIBRATED LOW-COST SENSOR AIR POLLUTION DATA

Low-cost air pollution sensors, offering hyper-local characterization of pollutant concentrations, are becoming increasingly prevalent in environmental and public health research. However, low-cost air pollution data can be noisy, biased by environmental conditions, and usually need to be field-calibrated by collocating low-cost sensors with reference-grade instruments. We show, theoretically and empirically, that the common procedure of regression-based calibration using collocated data systematically underestimates high air pollution concentrations, which are critical to diagnose from a health perspective. Current calibration practices also often fail to utilize the spatial correlation in pollutant concentrations. We propose a novel spatial filtering approach to collocation-based calibration of low-cost networks that mitigates the underestimation issue by using an inverse regression. The inverse-regression also allows for incorporating spatial correlations by a second-stage model for the true pollutant concentrations using a conditional Gaussian Process. Our approach works with one or more collocated sites in the network and is dynamic, leveraging spatial correlation with the latest available reference data. Through extensive simulations, we demonstrate how the spatial filtering substantially improves estimation of pollutant concentrations, and measures peak concentrations with greater accuracy. We apply the methodology for calibration of a low-cost PM2.5 network in Baltimore, Maryland, and diagnose air pollution peaks that are missed by the regression-calibration.

Evaluating the Performance of Low-Cost PM2.5 Sensors in Mobile Settings

Low-cost sensors (LCSs) for measuring air pollution are increasingly being deployed in mobile applications, but questions concerning the quality of the measurements remain unanswered. For example, what is the best way to correct LCS data in a mobile setting? Which factors most significantly contribute to differences between mobile LCS data and those of higher-quality instruments? Can data from LCSs be used to identify hotspots and generate generalizable pollutant concentration maps? To help address these questions, we deployed low-cost PM2.5 sensors (Alphasense OPC-N3) and a research-grade instrument (TSI DustTrak) in a mobile laboratory in Boston, MA, USA. We first collocated these instruments with stationary PM2.5 reference monitors (Teledyne T640) at nearby regulatory sites. Next, using the reference measurements, we developed different models to correct the OPC-N3 and DustTrak measurements and then transferred the corrections to the mobile setting. We observed that more complex correction models appeared to perform better than simpler models in the stationary setting; however, when transferred to the mobile setting, corrected OPC-N3 measurements agreed less well with the corrected DustTrak data. In general, corrections developed by using minute-level collocation measurements transferred better to the mobile setting than corrections developed using hourly-averaged data. Mobile laboratory speed, OPC-N3 orientation relative to the direction of travel, date, hour-of-the-day, and road class together explain a small but significant amount of variation between corrected OPC-N3 and DustTrak measurements during the mobile deployment. Persistent hotspots identified by the OPC-N3s agreed with those identified by the DustTrak. Similarly, maps of PM2.5 distribution produced from the mobile corrected OPC-N3 and DustTrak measurements agreed well. These results suggest that identifying hotspots and developing generalizable maps of PM2.5 are appropriate use-cases for mobile LCS data.

Investigating the Sensitivity of Low-Cost Sensors in Measuring Particle Number Concentrations across Diverse Atmospheric Conditions in Greece and Spain

Low-cost sensors (LCSs) for particulate matter (PM) concentrations have attracted the interest of researchers, supplementing their efforts to quantify PM in higher spatiotemporal resolution. The precision of PM mass concentration measurements from PMS 5003 sensors has been widely documented, though limited information is available regarding their size selectivity and number concentration measurement accuracy. In this work, PMS 5003 sensors, along with a Federal Referral Methods (FRM) sampler (Grimm spectrometer), were deployed across three sites with different atmospheric profiles, an urban (Germanou) and a background (UPat) site in Patras (Greece), and a semi-arid site in Almería (Spain, PSA). The LCSs particle number concentration measurements were investigated for different size bins. Findings for particles with diameter between 0.3 and 10 μm suggest that particle size significantly affected the LCSs' response. The LCSs could accurately detect number concentrations for particles smaller than 1 μm in the urban (R² = 0.9) and background sites (R² = 0.92), while a modest correlation was found with the reference instrument in the semi-arid area (R² = 0.69). However, their performance was rather poor (R² < 0.31) for coarser aerosol fractions at all sites. Moreover, during periods when coarse particles were dominant, i.e., dust events, PMS 5003 sensors were unable to report accurate number distributions (R² values < 0.47) and systematically underestimated particle number concentrations. The results indicate that several questions arise concerning the sensors' capabilities to estimate PM_2.5 and PM₁₀ concentrations, since their size distribution did not agree with the reference instruments.

Merging citizen science with epidemiology: design of a prospective feasibility study of health events and air pollution in Cologne, Germany

BACKGROUND

Citizen science as an approach to merge society and science is not a new paradigm. Yet it is not common in public health, epidemiology, or medical sciences. SMARAGD (Sensors for Measuring Aerosols and ReActive Gases to Deduce health effects) assesses air pollution at participants' homes or workplaces in Cologne, Germany, as feasibility study with a citizen science approach. Personal exposure to air pollutants is difficult to study, because the distribution of pollutants is heterogeneous, especially in urban areas. Targeted data collection allows to establish connections between air pollutant concentration and the health of the study population. Air pollution is among the most urgent health risks worldwide. Yet links of individualized pollution levels and respiratory infections remain to be validated, which also applies for the feasibility of the citizen science approach for epidemiological studies.

METHODS

We co-designed a prospective feasibility study with two groups of volunteers from Cologne, Germany. These citizen scientists and researchers determined that low-cost air-quality sensors (hereafter low-cost sensors) were to be mounted at participants' homes/workplaces to acquire stationary data. The advantage of deploying low-cost sensors is the achievable physical proximity to the participants providing health data. Recruitment started in March 2021 and is currently ongoing (as of 09/22). Sensor units specifically developed for this study using commercially available electronic sensor components will measure particulate matter and trace gases such as ozone, nitrogen oxides, and carbon monoxide. Health data are collected using the eResearch system "Prospective Management and Monitoring-App" (PIA). Due to the ongoing SARS-CoV-2 pandemic, we also focus on COVID-19 as respiratory infection.

DISCUSSION

Citizen science offers many benefits for science in general but also for epidemiological studies. It provides scientific information to society, enables scientific thinking in critical discourses, can counter anti-scientific ideologies, and takes into account the interests of society. However, it poses many challenges, as it requires extensive resources from researchers and society and can raise concerns regarding data protection and methodological challenges such as selection bias.

Investigating Use of Low-Cost Sensors to Increase Accuracy and Equity of Real-Time Air Quality Information

U.S. Environmental Protection Agency (EPA) air quality (AQ) monitors, the "gold standard" for measuring air pollutants, are sparsely positioned across the U.S. Low-cost sensors (LCS) are increasingly being used by the public to fill in the gaps in AQ monitoring; however, LCS are not as accurate as EPA monitors. In this work, we investigate factors impacting the differences between an individual's true (unobserved) exposure to air pollution and the exposure reported by their nearest AQ instrument (which could be either an LCS or an EPA monitor). We use simulations based on California data to explore different combinations of hypothetical LCS placement strategies (e.g., at schools or near major roads), for different numbers of LCS, with varying plausible amounts of LCS device measurement errors. We illustrate how real-time AQ reporting could be improved (or, in some cases, worsened) by using LCS, both for the population overall and for marginalized communities specifically. This work has implications for the integration of LCS into real-time AQ reporting platforms.

Prediction of influencing atmospheric conditions for explosion Avoidance in fireworks manufacturing Industry-A network approach

This research study uses Artificial Neural Networks (ANNs) to predict occupational accidents in Sivakasi firework industries. Atmospheric temperature, pressure and humidity are the causes of explosion during chemical mixing, drying, and pellet making. The Proposed ANN model predicts the accidents and the session of accidents (FN/AN) based on atmospheric conditions. This prediction takes values from historical accident data due to the atmospheric conditions of Sivakasi (2009-2021). In the development of ANN model, the Feed-Forward Back Propagation (FFBP) with the Levenberg-Marquardt function has been employed with hidden layers of 5 and 10 to train the network. The performance accuracy of both the hidden layers is evaluated and compared with other models like Support Vector Machine (SVM), Random Forest (RF), and K-Nearest Neighbor (K-NN). The accuracy of the proposed model for accident classification is 82.7% and 67.8% for hidden layers 5 and 10, respectively. Also, the model predicts the session of accident with the accuracy of 72% and 54%, specificity of 77.7% and 60.1%, sensitivity of 69% and 52.92% for hidden layers 5 and 10, respectively. It is found that hidden layer 5 gives higher accuracy than hidden layer 10. The proposed ANN model gives the highest accuracy when compared to other models. This study is helpful in the firework industry management, and workers improve safety precautions and avoid explosions due to atmospheric conditions.

Evaluation of two collocated federal equivalent method PM2.5 instruments over a wide range of concentrations in Sarajevo, Bosnia and Herzegovina

Two widely used PM_2.5 monitors in the United States (U.S.) designated as federal equivalent methods (FEMs) by the U.S. Environmental Protection Agency were collocated for 15 months in Sarajevo, Bosnia and Herzegovina (BiH) to evaluate their comparability. With differing measurement principles, the FEMs (Met One BAM-1020 and Teledyne API T640) exhibited unique responses to the significant range in PM_2.5 over the study period. During the winter months when concentrations greatly increased (e.g., daily PM_2.5 > 100 μg m^-3), the BAM-1020 had intermittent malfunctioning nozzle contact to the collection tape, resulting in periods of data invalidation. Increased operator observation and doubling the cleaning frequency were required to maintain proper operation. The hourly data from the BAM-1020, which detects PM_2.5 via beta-attenuation of particles loaded to the collection tape, indicated higher noise at concentrations below 40 μg m^-3 relative to the T640, which detects PM_2.5 via an optical method. Above this concentration threshold, the two instruments appear to have comparable hourly fluctuations in the data. Relative to the BAM-1020, the T640 reported higher concentrations when PM_2.5 is above 80 μg m^-3. A linear regression equation was developed and applied to adjust T640 PM_2.5 high concentration values, resulting in 24-hr average T640adj PM_2.5 values closely matching that from the BAM-1020 for the full concentration range. Based on the T640adj values, the annual average for Sarajevo was calculated at the site to be 42 μg m^-3, with significant seasonality resulting in over 7-fold higher concentrations in the months of December-January compared to June-July.

A Novel Bike-Mounted Sensing Device with Cloud Connectivity for Dynamic Air-Quality Monitoring by Urban Cyclists

We present a device based on low-cost electrochemical and optical sensors, designed to be attached to bicycle handlebars, with the aim of monitoring the air quality in urban environments. The system has three electrochemical sensors for measuring NO₂ and O₃ and an optical particle-matter (PM) sensor for PM₂.₅ and PM₁₀ concentrations. The electronic instrumentation was home-developed for this application. To ensure a constant air flow, the input fan of the particle sensor is used as an air supply pump to the rest of the sensors. Eight identical devices were built; two were collocated in parallel with a reference urban-air-quality-monitoring station and calibrated using a neural network (R² > 0.83). Several bicycle routes were carried out throughout the city of Badajoz (Spain) to allow the device to be tested in real field conditions. An air-quality index was calculated to facilitate the user's understanding. The results show that this index provides data on the spatiotemporal variability of pollutants between the central and peripheral areas, including changes between weekdays and weekends and between different times of the day, thus providing valuable information for citizens through a dedicated cloud-based data platform.

Integration of complete ensemble empirical mode decomposition with deep long short-term memory model for particulate matter concentration prediction

The atmospheric particulate matter (PM) with a diameter of 2.5 μm or less (PM2.5) is one of the key indicators of air pollutants. Accurate prediction of PM2.5 concentration is very important for air pollution monitoring and public health management. However, the presence of noise in PM2.5 data series is a major challenge of its accurate prediction. A novel hybrid PM2.5 concentration prediction model is proposed in this study by combining complete ensemble empirical mode decomposition (CEEMD) method, Pearson's correlation analysis, and a deep long short-term memory (LSTM) method. CEEMD was employed to decompose historical PM2.5 concentration data to different frequencies in order to enhance the timing characteristics of data. Pearson's correlation was used to screen the different frequency intrinsic-mode functions of decomposed data. Finally, the filtered enhancement data were inputted to a deep LSTM network with multiple hidden layers for training and prediction. The results evidenced the potential of the CEEMD-LSTM hybrid model with a prediction accuracy of approximately 80% and model convergence after 700 training epochs. The secondary screening of Pearson's correlation test improved the model (CEEMD-Pearson) accuracy up to 87% but model convergence after 800 epochs. The hybrid model combining CEEMD-Pearson with the deep LSTM neural network showed a prediction accuracy of nearly 90% and model convergence after 650 interactions. The results provide a clear indication of higher prediction accuracy of PM2.5 with less computation time through hybridization of CEEMD-Pearson with deep LSTM models and its potential to be employed for air pollution monitoring.

Flexible, non-contact and multifunctional humidity sensors based on two-dimensional phytic acid doped co-metal organic frameworks nanosheets

The development of high-performance humidity sensors is of great significance to explore their practical applications in the fields of environment, energy saving and safety monitoring. Herein, a flexible, non-contact and multifunctional humidity sensor based on two-dimensional Co-metal organic frameworks (Co-MOF) nanosheets is proposed, which is fabricated by simple bottom-up synthesis method. Furthermore, environmentally friendly, renewable and abundant biomass phytic acid (PA) is modified on the surface of Co-MOF nanosheets, which releases free protons being capable of etching the framework of MOF to improve the hydrophilicity and conductivity of MOF. Compared with Co-MOF-based sensor, the Co-MOF@PA-based sensor exhibits significantly enhanced sensitivity and broadened response range within 23-95% relative humidity (RH). The humidity sensor has an excellent humidity sensing response over 2 × 10³. The Co-MOF@PA-based sensor shows good flexibility and humidity sensing properties, endowing it with multifunctional applications in real-time facial respiration monitoring, skin humidity perception, cosmetic moisturizing evaluation and fruit freshness testing. Four respiration patterns, including slow breath, deep breath, normal breath and fast breath are wirelessly monitored in real time by Co-MOF@PA-based sensor and recorded by mobile phone software. The research work presents potential applications in human-machine interactions (HMI) devices in future.

Assessment of the applicability of a low-cost sensor-based methane monitoring system for continuous multi-channel sampling

Systems that are made of several low-cost gas sensors with automatic gas sampling may have the potential to serve as reliable fast methane analyzers. However, there is a lack of reports about such types of systems evaluated under field conditions. Here, we developed a continuous methane monitoring system with automated gas sampling unit using low-cost gas sensors, TGS 2611 and MQ-4, that use a simple cloud-based data acquisition platform. We verified the consistency, repeatability, and reproducibility of the data obtained by TGS 2611 and MQ-4 low-cost gas sensors by measuring high- and low-concentration methane samples. The normalized root-mean-square errors (NRMSEs) of the samples with high methane concentrations, [CH₄] of 3, 4, 6, and 7%, were 0.0788, 0.0696, 0.1198, and 0.0719 for the TGS 2611 sensor, respectively, and were confirmed using a gas chromatograph as a reference analyzer. The NRMSEs of the samples with low [CH₄] of 0.096, 0.145, 0.193, and 0.241% measured by the TGS 2611 sensor were 0.0641, 0.1749, 0.0157, and 0.1613, whereas those NRMSEs of the same concentrations measured by the MQ-4 sensor were 0.3143, 0.5766, 0.6301, and 0.6859, respectively. Laboratory-scale anaerobic digesters were tested using the developed system. The anaerobic digesters were continuously operated for 2 months, demonstrating the potential use of sensors for detecting and monitoring methane in the field level application. This study utilized a unique way to combine the advantages of low-cost sensors and develop a reliable monitoring system by minimizing drawbacks of low-cost sensors.

From a Low-Cost Air Quality Sensor Network to Decision Support Services: Steps towards Data Calibration and Service Development

Air pollution is a widespread problem due to its impact on both humans and the environment. Providing decision makers with artificial intelligence based solutions requires to monitor the ambient air quality accurately and in a timely manner, as AI models highly depend on the underlying data used to justify the predictions. Unfortunately, in urban contexts, the hyper-locality of air quality, varying from street to street, makes it difficult to monitor using high-end sensors, as the cost of the amount of sensors needed for such local measurements is too high. In addition, development of pollution dispersion models is challenging. The deployment of a low-cost sensor network allows a more dense cover of a region but at the cost of noisier sensing. This paper describes the development and deployment of a low-cost sensor network, discussing its challenges and applications, and is highly motivated by talks with the local municipality and the exploration of new technologies to improve air quality related services. However, before using data from these sources, calibration procedures are needed to ensure that the quality of the data is at a good level. We describe our steps towards developing calibration models and how they benefit the applications identified as important in the talks with the municipality.