Regression methods in the calibration of low-cost sensors for ambient particulate matter measurements

Calibration and validation-based assessment of low-cost air quality sensors

BACKGROUND

Air pollution poses a significant threat to public health. Low-cost air quality sensors (LCSs) can provide a data foundation for air quality monitoring, particularly supplementing the regulatory monitoring network and identifying local air quality issues. However, the performance varies considerably, and questions remain regarding reliability and accuracy of LCS data.

METHODS

We evaluated the accuracy, stability and precision of six LCSs over a three-month period of collocation with reference instruments at two locations. A mathematical workflow including calibration and validation was developed for accuracy and stability, incorporating a combination of environmental factors (e.g., temperature, relative humidity), linear and nonlinear regression, followed by precision evaluation by Bland-Altman plots.

RESULTS

For particulate matter, data from LCSs was found to be reliable after simple linear regression (R2 > 0.9 for both calibration and validation). For gas sensors including nitrogen dioxide, carbon monoxide, and Ozone, nonlinear methods that met the validation requirements also performed well using simple linear regression models (R2 > 0.7 for both calibration and validation), whereas machine learning models, such as random forest, could not pass the validation, and require cautious application. In non-laboratory environments, incorporating environmental factors into the calibration function may lead to subsequent performance instability. Regarding precision between LCSs, unstable measurement biases among devices have been observed.

CONCLUSIONS

Linear regression method is recommended as the preferred method for onsite calibration of LCSs, with caution advised when incorporating environmental factors due to increased uncertainty. Furthermore, when deploying LCSs, it is important to consider their varying responses to high or low pollutant concentrations.

A wearable real-time particulate monitor demonstrates that soaking hay reduces dust exposure

BACKGROUND

Affordable particulate matter (PM) monitors suitable for use on horses will facilitate the evaluation of PM mitigation methods and improve the management of equine asthma.

OBJECTIVE

Calibrate a real-time wearable PM monitor (Black Beauty [BB]) and compare the PM exposures of horses fed dry or soaked hay.

STUDY DESIGN

Laboratory calibration; complete cross-over feed trial.

METHODS

Side-by-side sampling with BB monitors and tapered element oscillating microbalances (TEOMs) was performed under varying concentrations of PM from alfalfa hay. Linear regression was used to derive a calibration formula for each unit based on TEOM PM measurements. Precision was evaluated by calculating the coefficient of variation and pairwise correlation coefficients between three BB monitors. PM exposure was measured at the breathing zone of 10 horses for 8 h after they were fed dry or soaked hay. Repeated measures generalised linear models were constructed to determine the effect of hay treatment and measurement duration (initial 20-min vs. 8-h) upon exposure to PM with diameters smaller than or equal to 10 μm (PM10) and 2.5 μm (PM2.5).

RESULTS

BB monitor PM2.5 and PM10 measurements were linearly correlated with TEOM data (coefficient of determination r2 > 0.85 and r2 > 0.90 respectively), but underestimated PM2.5 mass concentrations by a factor of 4 and PM10 concentrations by a factor of 44. Measures from the three BB monitors had a coefficient of variation <15% and pairwise r > 0.98. Feeding soaked hay significantly reduced average PM2.5 exposures (20-min: dry: 160 μg/m3, soaked: 53 μg/m3, p < 0.0001; 8-h: dry: 76 μg/m3, soaked: 31 μg/m3, p = 0.0008) and PM10 exposures (20-min: dry: 2829 μg/m3, soaked: 970 μg/m3, p < 0.0001; 8-h: dry: 1581 μg/m3, soaked: 488 μg/m3, p = 0.008).

MAIN LIMITATIONS

No health outcome measures were collected.

CONCLUSIONS

With appropriate corrections, the BB monitor can be used to estimate horse PM exposure. While 20-min measurements yielded higher estimates of exposure than 8-h measurements, both intervals demonstrate that soaking hay reduces PM exposures by more than 50%.

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.

Indication Variability of the Particulate Matter Sensors Dependent on Their Location

Particulate matter (PM) suspended in the air significantly impacts human health. Those of anthropogenic origin are particularly hazardous. Poland is one of the countries where the air quality during the heating season is the worst in Europe. Air quality in small towns and villages far from state monitoring stations is often much worse than in larger cities where they are located. Their residents inhale the air containing smoke produced mainly by coal-fired stoves. In the frame of this project, an air quality monitoring network was built. It comprises low-cost PMS7003 PM sensors and ESP8266 microcontrollers with integrated Wi-Fi communication modules. This article presents research results on the influence of the PM sensor location on their indications. It has been shown that the indications from sensors several dozen meters away from each other can differ by up to tenfold, depending on weather conditions and the source of smoke. Therefore, measurements performed by a network of sensors, even of worse quality, are much more representative than those conducted in one spot. The results also indicated the method of detecting a sudden increase in air pollutants. In the case of smokiness, the difference between the mean and median indications of the PM sensor increases even up to 400 µg/m3 over a 5 min time window. Information from this comparison suggests a sudden deterioration in air quality and can allow for quick intervention to protect people's health. This method can be used in protection systems where fast detection of anomalies is necessary.

Designing State-of-the-Art Gas Sensors: From Fundamentals to Applications

Gas sensors are crucial in environmental monitoring, industrial safety, and medical diagnostics. Due to the rising demand for precise and reliable gas detection, there is a rising demand for cutting-edge gas sensors that possess exceptional sensitivity, selectivity, and stability. Due to their tunable electrical properties, high-density surface-active sites, and significant surface-to-volume ratio, nanomaterials have been extensively investigated in this regard. The traditional gas sensors utilize homogeneous material for sensing where the adsorbed surface oxygen species play a vital role in their sensing activity. However, their performance for selective gas sensing is still unsatisfactory because the employed high temperature leads to the poor stability. The heterostructures nanomaterials can easily tune sensing performance and their different energy band structures, work functions, charge carrier concentration and polarity, and interfacial band alignments can be precisely designed for high-performance selective gas sensing at low temperature. In this review article, we discuss in detail the fundamentals of semiconductor gas sensing along with their mechanisms. Further, we highlight the existed challenges in semiconductor gas sensing. In addition, we review the recent advancements in semiconductor gas sensor design for applications from different perspective. Finally, the conclusion and future perspectives for improvement of the gas sensing performance are discussed.

Application of ADMS-Urban for an area with a high contribution of residential heating emissions - model verification and sensitivity study for PM2.5

Air pollution poses a significant risk to both human health and the environment in the contemporary world. Among the various pollutants, particulate matter with a diameter <2.5 μm (PM2.5) is regarded as the most hazardous. It has been implicated in over four million global fatalities in 2019 alone. This research paper divulges the outcomes of modelling the spatial-temporal fluctuations of PM2.5 concentrations within the confines of Wroclaw, a city situated in Poland, Central Europe. The model's output was evaluated through comparison with collected data from two government-operated monitoring stations within the city. For this study, we used the ADMS-Urban model and tested two different sources of background data (low-cost sensors and the EMEP MSC-W atmospheric chemistry transport model). The statistical analysis conducted in the paper indicates that the model reproduces the temporal variability of PM2.5. The conclusions from this research indicate that the average annual PM2.5 concentration within Wroclaw is 13.8 μg/m3, with the concentration peaking in the month of March. The spatial distribution reveals the highest PM2.5 concentrations primarily in the southern and western zones of the city, with additional elevated concentrations observed sporadically throughout the city. The study unveils that 1.3 % of Wroclaw's area experiences PM2.5 concentrations exceeding the EU's annual limit of 20 μg/m3. When considered in relation to the WHO's suggested annual average level of 5 μg/m3, Wroclaw city experiences exceedances throughout. When background concentrations are excluded from the model, the annual average PM2.5 concentration across the city is noted to be reduced by >50 %. A thorough investigation of the city's emission structure, taking into account only emissions from the city without background, indicates that the residential sector contributes about 77.3 % of the total annual average PM2.5 concentration in Wroclaw. The transportation and industrial sectors account for nearly 19.5 % and 3.2 %, respectively.

Aerial monitoring of atmospheric particulate matter produced by open-pit mining using low-cost airborne sensors

Mining is an economic activity that entails the production and displacement of significant amounts of atmospheric particulate matter (PM) during operations involving intense earthcrushing or earthmoving. As high concentrations of PM may have adverse effects on human health, it is necessary to monitor and control the fugitive emissions of this pollutant. This paper presents an innovative methodology for the online monitoring of PM10 concentrations in air using a low-cost sensor (LCS, <300 USD) onboard an unmanned aerial vehicle. After comprehensive calibration, the LCS was horizontally flown over seven different areas of the large Riotinto copper mine (Huelva, Spain) at different heights to study the PM10 distribution at different longitudes and altitudes. The flights covered areas of zero activity, intense mining, drilling, ore loading, waste discharge, open stockpiling, and mineral processing. In the zero-activity area, the resuspension of PM10 was very low, with a weak wind speed (3.6 m/s). In the intense-mining area, unhealthy concentrations of PM10 (>51 μgPM10/m3) could be released, and the PM10 can reach surrounding populations through long-distance transport driven by several processes being performed simultaneously. Strong dilution was also observed at high altitudes (> 50 m). Mean concentrations were found to be 22-89 μgPM10/m3, with peaks ranging from 86 to 284 μgPM10/m3. This study demonstrates the potential applicability of airborne LCSs in the high-resolution online monitoring of PM in mining, thus supporting environmental managers during decision-making against fugitive emissions in a cost-effective manner.

Analysis of aerosol optical depth and relation to covariates during pre-monsoon season (2002-2019) over Pakistan using ARIMAX model and cross-wavelet analysis

The pre-monsoon season heavily influences the precipitation amount in Pakistan. When hydrometeorological parameters interact with aerosols from multiple sources, a radiative climatic response is observed. In this study, aerosol optical depth (AOD) space-time dynamics were analyzed in relation to meteorological factors and surface parameters during the pre-monsoon season in the years 2002-2019 over Pakistan. Level-3 (L3) monthly datasets from Moderate Resolution Imaging Spectroradiometer (MODIS) and Multi-Angle Imaging Spectroradiometer (MISR) were used. Tropical Rainfall Measuring Mission (TRMM) derived monthly precipitation, Atmospheric Infrared Sounder (AIRS) derived air temperature, after moist relative humidity (RH) from Modern-Era Retrospective analysis for Research and Applications, Version-2 (MERRA-2), near-surface wind speed, and soil moisture data derived from Global Land Data Assimilation System (GLDAS) were also used on a monthly time scale. For AOD trend analysis, Mann-Kendall (MK) trend test was applied. Moreover, Autoregressive Integrated Moving Average with Explanatory variable (ARIMAX) technique was applied to observe the actual and predicted AOD trend, as well as test the multicollinearity of AOD with covariates. The periodicities of AOD were analyzed using continuous wavelet transformation (CWT) and the cross relationships of AOD with prevailing covariates on a time-frequency scale were analyzed by wavelet coherence analysis. A high variation of aerosols was observed in the spatiotemporal domain. The MK test showed a decreasing trend in AOD which was most significant in Baluchistan and Punjab, and the overall trend differs between MODIS and MISR datasets. ARIMAX model shows the correlation of AOD with varying meteorological and soil parameters. Wavelet analysis provides the abundance of periodicities in the 2-8 months periodic cycles. The coherency nature of the AOD time series along with other covariates manifests leading and lagging effects in the periodicities. Through this, a notable difference was concluded in space-time patterns between MODIS and MISR datasets. These findings may prove useful for short-term and long-term studies including oscillating features of AOD and covariates.

Performance-based protocol for selection of economical portable sensor for air quality measurement

An effective micro-level air quality management plan requires high-resolution monitoring of pollutants. India has already developed a vast network of air quality monitoring stations, both manual and real time, located primarily in urban areas, including megacities. The air quality monitoring network consists of conventional manual stations and real time Continuous Ambient Air Quality Monitoring Stations (CAAQMS) which comprise state-of-the-art analysers and instruments. India is currently in the early stages of developing and adopting economical portable sensor (EPS) in air quality monitoring systems. Protocols need to be established for field calibration and testing. The present research work is an attempt to develop a performance-based assessment framework for the selection of EPS for air quality monitoring. The two-stage selection protocol includes a review of the factory calibration data and a comparison of EPS data with a reference monitor, i.e. a portable calibrated monitor and a CAAQMS. Methods deployed include calculation of central tendency, dispersion around a central value, calculation of statistical parameters for data comparison, and plotting pollution rose and diurnal profile (peak and non-peak pollution measurement). Four commercially available EPS were tested blind, out of which, data from EPS 2 (S2) and EPS 3 (S3) were closer to reference stations at both locations. The selection was made by evaluating monitoring results, physical features, measurement range, and frequency along with examining capital cost. This proposed approach can be used to increase the usability of EPS in the development of micro-level air quality management strategies, other than regulatory compliance. For regulatory compliance, additional research is needed, including field calibration and evaluating EPS performance through additional variables. This proposed framework may be used as starting point, for such experiments, in order to develop confidence in the use of EPS.

Smart and Portable Air-Quality Monitoring IoT Low-Cost Devices in Ibarra City, Ecuador

Nowadays, increasing air-pollution levels are a public health concern that affects all living beings, with the most polluting gases being present in urban environments. For this reason, this research presents portable Internet of Things (IoT) environmental monitoring devices that can be installed in vehicles and that send message queuing telemetry transport (MQTT) messages to a server, with a time series database allocated in edge computing. The visualization stage is performed in cloud computing to determine the city air-pollution concentration using three different labels: low, normal, and high. To determine the environmental conditions in Ibarra, Ecuador, a data analysis scheme is used with outlier detection and supervised classification stages. In terms of relevant results, the performance percentage of the IoT nodes used to infer air quality was greater than 90%. In addition, the memory consumption was 14 Kbytes in a flash and 3 Kbytes in a RAM, reducing the power consumption and bandwidth needed in traditional air-pollution measuring stations.

Correction Model for Metal Oxide Sensor Drift Caused by Ambient Temperature and Humidity

For decades, Metal oxide (MOX) gas sensors have been commercially available and used in various applications such as the Smart City, gas monitoring, and safety due to advantages such as high sensitivity, a high detection range, fast reaction time, and cost-effectiveness. However, several factors affect the sensing ability of MOX gas sensors. This article presents the results of a study on the cross-sensitivity of MOX gas sensors toward ambient temperature and humidity. A gas sensor array consisting of temperature and humidity sensors and four different MOX gas sensors (MiCS-5524, GM-402B, GM-502B, and MiCS-6814) was developed. The sensors were subjected to various relative gas concentrations, temperatures (from 16 °C to 30 °C), and humidity levels (from 75% to 45%), representing a typical indoor environment. The results proved that the gas sensor responses were significantly affected by the temperature and humidity. The increased temperature and humidity levels led to a decreased response for all sensors, except for MiCS-6814, which showed the opposite response. Hence, this work proposed regression models for each sensor, which can correct the gas sensor response drift caused by the ambient temperature and humidity variations. The models were validated, and the standard deviations of the corrected sensor response were found to be 1.66 kΩ, 13.17 kΩ, 29.67 kΩ, and 0.12 kΩ, respectively. These values are much smaller compared to the raw sensor response (i.e., 18.22, 24.33 kΩ, 95.18 kΩ, and 2.99 kΩ), indicating that the model provided a more stable output and minimised the drift. Overall, the results also proved that the models can be used for MOX gas sensors employed in the training process, as well as for other sets of gas sensors.

Low-Cost Sensors for Air Quality Monitoring - the Current State of the Technology and a Use Overview

In recent years the monitoring of air quality using cheap sensors has become an interesting alternative to conventional analytical techniques. Apart from vast price differences conventional techniques need to be performed by the trained personnel of commercial or research laboratories. Sensors capable of measuring dust, ozone, nitrogen and sulphur oxides, or other air pollutants are relatively simple electronic devices, which are comparable in size to a mobile phone. They provide the general public with the possibility to monitor air quality which can contribute to various projects that differ in regional scale, commercial funding or community-base. In connection with the low price of sensors arises the question of the quality of measured data. This issue is addressed by a number of studies focused on comparing the sensor data with the data of reference measurements. Sensory measurement is influenced by the monitored analyte, type and design of the particular sensor, as well as by the measurement conditions. Currently sensor networks serve as an additional source of information to the network of air quality monitoring stations, where the density of the network provides concentration trends in the area that may exceed specific measured values of pollutant concentrations and low uncertainty of reference measurements. The constant development of all types of sensors is leading to improvements and the difference in data quality between sensors and conventional monitoring techniques may be reduced.

Real-Time Low-Cost Personal Monitoring for Exposure to PM2.5 among Asthmatic Children: Opportunities and Challenges

This study aims to evaluate the accuracy and effectiveness of real-time personal monitoring of exposure to PM concentrations using low-cost sensors, in comparison to conventional data collection method based on fixed stations. PM2.5 data were measured every 5 min using a low-cost sensor attached to a bag carried by 47 asthmatic children living in the Seoul Metropolitan area between November 2019 and March 2020, along with the real-time GPS location, temperature, and humidity. The mobile sensor data were then matched with station-based hourly PM2.5 data using the time and location. Despite some uncertainty and inaccuracy of the sensor data, similar temporal patterns were found between the two sources of PM2.5 data on an aggregate level. However, average PM2.5 concentrations via personal monitoring tended to be lower than those from the fixed stations, particularly when the subjects were indoors, during nighttime, and located farther from the fixed station. On an individual level, a substantial discrepancy is observed between the two PM2.5 data sources while staying indoors. This study provides guidance to policymakers and researchers on improving the feasibility of personal monitoring via low-cost mobile sensors as an alternative or supplement to the conventional station-based monitoring.

Calibrating low-cost sensors for ambient air monitoring: Techniques, trends, and challenges

Low-cost sensors (LCSs) are widely acknowledged for bringing a paradigm shift in supplemental traditional air monitoring by air regulatory agencies. However, there is concern regarding its data quality and performance stability, which has greatly restricted its large-scale applications. Knowing the recent techniques, progress, and challenges of LCS calibration is of immense significance to promote the field of environmental monitoring. By summarizing the published evidence, this review shows that the global sensor market is rapidly expanding due to the surging needs, but the calibration efforts have been focused on a limited selection of sensors. Relative humidity correction, regression, and machine learning are the three mainstream calibration techniques. Although there is no one-size-fits-all solution, a feature of the latest research tendency is machine learning. The duration of calibration is largely neglected in the experiment design, but it is found to affect the performance of different calibration methods, especially those that are data-driven. Geographically, China and the United States gained the most research attention in the sensor calibration field, but the spatial mismatch between particulate matter hotspots and calibration sites is quite evident for the rest of the world. Incomplete and unevenly distributed research footprints could limit the large-scale test of method generalizability, as well as diminish the monitoring capacity in underserved areas that suffer greater environmental justice crises. In general, model performance is enhanced by including the key influencing factors, but the degree of improvement is not evidently related to the number of explanatory variables. Overall, studies prove the critical importance of field calibration before sensor deployment, but more studies are needed to establish experiment protocols that can be customized to specific needs.

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

Low-cost air quality sensors can help increase spatial and temporal resolution of air pollution exposure measurements. These sensors, however, most often produce data of lower accuracy than higher-end instruments. In this study, we investigated linear and random forest models to correct PM_2.5 measurements from the Denver Department of Public Health and Environment (DDPHE)'s network of low-cost sensors against measurements from co-located U.S. Environmental Protection Agency Federal Equivalence Method (FEM) monitors. Our training set included data from five DDPHE sensors from August 2018 through May 2019. Our testing set included data from two newly deployed DDPHE sensors from September 2019 through mid-December 2019. In addition to PM_2.5, temperature, and relative humidity from the low-cost sensors, we explored using additional temporal and spatial variables to capture unexplained variability in sensor measurements. We evaluated results using spatial and temporal cross-validation techniques. For the long-term dataset, a random forest model with all time-varying covariates and length of arterial roads within 500 m was the most accurate (testing RMSE = 2.9 μg/m³ and R² = 0.75; leave-one-location-out (LOLO)-validation metrics on the training set: RMSE = 2.2 μg/m³ and R² = 0.93). For on-the-fly correction, we found that a multiple linear regression model using the past eight weeks of low-cost sensor PM_2.5, temperature, and humidity data plus a near-highway indicator predicted each new week of data best (testing RMSE = 3.1 μg/m³ and R² = 0.78; LOLO-validation metrics on the training set: RMSE = 2.3 μg/m³ and R² = 0.90). The statistical methods detailed here will be used to correct low-cost sensor measurements to better understand PM_2.5 pollution within the city of Denver. This work can also guide similar implementations in other municipalities by highlighting the improved accuracy from inclusion of variables other than temperature and relative humidity to improve accuracy of low-cost sensor PM_2.5 data.

Spatial calibration and PM2.5 mapping of low-cost air quality sensors

The data quality of low-cost sensors has received considerable attention and has also led to PM_2.5 warnings. However, the calibration of low-cost sensor measurements in an environment with high relative humidity is critical. This study proposes an efficient calibration and mapping approach based on real-time spatial model. The study carried out spatial calibration, which automatically collected measurements of low-cost sensors and the regulatory stations, and investigated the spatial varying pattern of the calibrated low-cost sensor data. The low-cost PM_2.5 sensors are spatially calibrated based on reference-grade measurements at regulatory stations. Results showed that the proposed spatial regression approach can explain the variability of the biases from the low-cost sensors with an R-square value of 0.94. The spatial calibration and mapping algorithm can improve the bias and decrease to 39% of the RMSE when compared to the nonspatial calibration model. This spatial calibration and real-time mapping approach provide a useful way for local communities and governmental agencies to adjust the consistency of the sensor network for improved air quality monitoring and assessment.