Ming-Shing Ho1, Ming-Yeng Lin1, Chih-Da Wu2,3, Jung-Der Wang4, Li-Hao Young5, Hui-Tsung Hsu5, Bing-Fang Hwang5, Perng-Jy Tsai This email address is being protected from spambots. You need JavaScript enabled to view it.1 

1 Department of Environmental and Occupational Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan
2 Department of Geomatics, National Cheng Kung University, Tainan, Taiwan
3 Innovation and Development Center of Sustainable Agriculture, National Chung-Hsing University, Taichung, Taiwan
4 Department of Public Health, College of Medicine, National Cheng Kung University, Tainan, Taiwan
5 Department of Occupational Safety and Health, College of Public Health, China Medical University, Taichung, Taiwan

Received: December 11, 2023
Revised: January 15, 2024
Accepted: January 17, 2024

 Copyright The Author(s). This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are cited.

Download Citation: ||https://doi.org/10.4209/aaqr.230313  

  • Download: PDF

Cite this article:

Ho, M.S., Lin, M.Y., Wu, C.D., Wang, J.D., Young, L.H., Hsu, H.T., Hwang, B.F., Tsai, P.J. (2024). An Integrated Approach for Conducting long-term PM2.5 Exposure and Health Impact Assessments for Residents of a City-scale. Aerosol Air Qual. Res. 24, 230313. https://doi.org/10.4209/aaqr.230313


  • AQMS data is good for characterizing temporal PM2.5 variations of the rooftop level.
  • h_LUR data can characterize spatiotemporal PM2.5 variations of the rooftop level.
  • MMS data can characterize spatiotemporal PM2.5 variations of the ground level.
  • Long-term ground level PM2.5 data can assess residents’ exposure and health impact.


This study aims to show the benefits of integrating the mobile monitoring system (MMS), hybrid land use regression (h_LUR), and air quality monitoring station (AQMS) data in both conducting long-term PM2.5 exposure assessment (EA) and health impact assessment (HIA) for residents in a city-scale. A city installed with a government-operated AQMS was selected. A 1-year PM2.5 dataset was collected from AQMS (AQMS1yr, reflecting the temporal variation of the rooftop level), and was served as a basis for characterizing the spatiotemporal heterogeneity of the rooftop level (h_LUR1yr) using the h_LUR model. A 1-year dataset was simultaneously collected from an established MMS for characterizing the spatiotemporal heterogeneity of the ground level (MMS1yr). A ground-level PM2.5 concentration predictive model was established by relating hourly MMS1yr to h_LUR1yr data and significant environmental covariables using the multivariate linear regression analysis. To establish long-term exposure datasets, 9-year AQMS data (AQMS9yr) were collected, and h_LUR9yr and MMS9yr were established through the application of the h_LUR and the obtained predictive model, respectively. Results show MMS1yr (24–26 µg m–3) > h_LUR1yr (17–19 µg m–3) > AQMS1yr (13–15 µg m–3). An R2 = 0.61 was obtained for the established ground predictive model. PM2.5 concentrations consistently decrease by year for MMS9yr (29–17 µg m–3), h_LUR9yr (25–12 µg m–3), and AQMS9yr (22–11 µg m–3), respectively. The result MMS9yr > h_LUR9yr > AQMS9yr indicates both the use of h_LUR9yr and AQMS9yr would result in underestimating residents’ exposures. By reference to the results obtained from MMS9yr, using AQMS9yr and h_LUR9yr would respectively lead to the underestimation of the attributed fraction (AFs) ~21%–36% and 18%–26% for the 5 disease burdens, including ischemic heart disease, stroke, chronic obstructive pulmonary disease, lung cancer, and lower respiratory tract infection. The above results clearly indicate the importance of using the integrating approach on conducting EA and HIA.

Keywords: PM2.5, Air quality monitoring station, Hybrid land use regression, Mobile Monitor-ing system, Predictive model, Health impact assessment


Substantial scientific evidence has linked the exposures of ambient PM2.5 (i.e., particulate matter with particle sizes less than 2.5 µm) to public health (Mustafić et al., 2012; Forouzanfar et al., 2016; Monrad et al., 2017; U.S. EPA, 2019). Data from the Global Burden of Disease (GBD) shows that ambient PM2.5 was attributed to ~4.14 million pollution attributed deaths (male/female: 2.44/1.70 million), and highlights variations among different regions/countries (Fuller et al., 2022). The Global Sustainable Development Goals (SDGs) Target 11.6 shows that the reduction of PM2.5 concentration is a crucial issue for cities to achieve by 2030 (WHO, 2012). However, it should be noted that the reduction of ambient PM2.5 concentration should effectively reflect to the decrease in exposures, and even health impacts posed on residents. Therefore, conducting both the exposure assessment (EA) and health impact assessment (HIA) have become an important issue in the field of environmental health science (Chu et al., 2016; Dias and Tchepel, 2018; Morawska et al., 2018; Bai et al., 2019; Miri et al., 2019; Li et al., 2020a).

Considering the inherent spatial and temporal variations in residents' exposures to PM2.5, it is essential to develop suitable methodologies for conducting EA, particularly focusing on ground level long-term PM2.5 concentrations, which are considered more representative of residents' real exposures. In principle, methodologies for conducting EA can be broadly classified into two categories: non-ground (macro-environment) and ground (micro-environment) approaches (IARC, 2013; Dias and Tchepel, 2018; Li et al., 2018; Caplin et al., 2019; U.S. EPA, 2019). For non-ground approaches, exposure data can be obtained from the air quality monitoring station (AQMS), modeling results [such as the land use regression model (LUR), chemical transport models (CTM), dispersion models], or remote sensing techniques [such as the satellite-derived aerosol optical depth (AOD)]. Among these data collection methods, the AQMS is the most widely used in the field because of its availability and accessibility (Ye et al., 2018). The long-term data collected by AQMS offers advantages in characterizing temporal variations to the study area. With limited number of AQMS installed, the lack of spatial resolution is clearly a drawback (Xu et al., 2021; Ho et al., 2022). Alternatively, techniques that combine satellite-derived AOD data or LUR model predicting data with AQMS data (denoted as AQMS + satellite-derived AOD or AQMS + LUR) provide potential solutions to address the spatial scarcity issue associated with AQMS (WHO, 2023). Here, it should be noted that the utilization of satellite-derived AOD data is inadequate for characterizing aerosol composition and vertical concentration. Particularly, the impact of regional meteorological conditions on measured concentrations and considerable costs also limit its application (van Donkelaar et al., 2010; Chu et al., 2016). The above limitations are also applied to the AQMS + satellite-derived AOD approach. For the AQMS + LUR approach, it has demonstrated its ability to characterize the spatiotemporal heterogeneity of PM2.5 concentrations (Vlaanderen et al., 2019; Xu et al., 2021). More recently, the LUR approach has been enhanced through the integration with machine learning techniques (denoted as the Hybrid Kriging-LUR model; h_LUR) (Wu et al., 2018; Wong et al., 2021; Thongthammachart et al., 2022). Therefore, it is believed that the use of the AQMS + h_LUR approach would provide more promising results than that of the AQMS + LUR approach. Here, it is worth noting that AQMS is typically installed at elevated sites (i.e., rooftops) due to the safety consideration. Consequently, the estimated PM2.5 concentrations obtained through the AQMS + LUR (or AQMS + h_LUR) approach, might simply represent to the rooftop concentrations, and hence might not fully describe the exposures experienced by residents at the ground level (Van den Bossche et al., 2015; Borge et al., 2016; Li et al., 2018).

For ground monitoring approaches, such as the fixed airbox and the mobile monitoring system (MMS), have been widely used to characterize the spatial variation of ground-level PM2.5 concentrations. Though Airbox provides real-time ground level concentrations, data accuracy and reliability are known as its limitations (Cabada et al., 2004; Snyder et al., 2013; Castell et al., 2017; Morawska et al., 2018; Williams et al., 2019; Li et al., 2020b; Zamora et al., 2020; U.S. EPA, 2021). The MMS, though more expensive and time-consuming, offers more precise and reliable real-time measurements for assessing the PM2.5 exposure status of individuals or groups (Hankey and Marshall, 2015b; Van den Bossche et al., 2015; Okokon et al., 2017; Lin et al., 2018; Cheng et al., 2019; Shen and Gao, 2019). Considering the inherent spatial and temporal variations in residents' exposure to PM2.5, it is essential to develop suitable methodologies for conducting EA, particularly focusing on ground-level long-term PM2.5 concentrations. Many studies have indicated that the approach combining the data obtained from the MMS, AQMS, and LUR (i.e., MMS + AQMS + LUR) would be a promising EA approach for addressing spatial and temporal variations at the ground level (Hankey and Marshall, 2015a; Dröge et al., 2018; Piotrowicz and Polednik, 2019; Xu et al., 2021). Considering the LUR is enhanced by h_LUR, the integrating approach of MMS + h_LUR + AQMS might provide an even better solution.

To assess the impact of ambient PM2.5 on city residents and further propose appropriate PM2.5 abatement strategies, conducting HIA is also considered essential along with the employment of EA (Ostro, 2004; Burnett et al., 2014; WHO, 2016; van den Brenk, 2018; Burnett et al., 2018). In the present study, EA and HIA associated with residents’ PM2.5 exposures in a city scale were conducted using the data obtained from the AQMS, AQMS + h_LUR, and AQMS + h_LUR + MMS. The benefits of using the AQMS + h_LUR + MMS approach were examined by reference to the corresponding results obtained from both AQMS and AQMS + h_LUR approaches.


2.1 The Study Area

The study area (i.e., the Shalu area) includes the Shalu, Wugi, and Longjing Districts, and is situated in the west of Taichung city (the second largest city in Taiwan). The dimension of the area is ~95 km2, and is home to a population of ~97,000 residents in 2021 (MOI, 2022). The area included five industrial/science parks encompassing ~19,368 manufacturing industries; and ~2920,000 vehicles with locomotives accounting for ~61% of the total (MODA, 2022; TEPA, 2022b). The entire area also featured three significant stationary pollution sources, including a coal-fired power plant, a steel mill, and a business harbor (i.e., the Taichung Port). (Fig. 1)

 Fig. 1. The study area, h_LUR, MMS sampling, and major landmarks.Fig. 1. The study area, h_LUR, MMS sampling, and major landmarks.

2.2 Data Collection from the Air Quality Monitoring Station (AQMS)

One governmental operated AQMS is installed on the northeast side of the study area (Fig. 1). This AQMS is located on the rooftop of a school building, approximately 15 meters above ground level, with a sampling inlet at a height of 19.5 meters. The primary monitoring items include PM2.5, PM10, CO, SO2, O3, and NO2, as well as meteorological data such as UVB, rainfall, wind direction, wind speed, temperature, and humidity. The installed PM2.5 monitoring instrument (model BAM-1020, Met One Instruments Inc., Grands Pass, Oregon) uses beta-ray technology to determine its concentration. The detection range and resolution of instrument are respectively as 0–10 mg m–3 and 2 µg m–3, and employs C-14 as the beta-ray source. The filter paper used is made of glass fiber with a capture efficiency exceeding 99.999% for particles > 0.3 µm. The sampling flow rate is maintained at 16.7 LPM, and the instrument automatically records PM2.5 measurements once per hour.

To ensure data accuracy and consistency, the collected PM2.5 data is calibrated using manual monitoring data obtained from nearby reference sites, following the Federal Reference Methods/Federal Equivalent Methods (FRM/FEM) of U.S. EPA (U.S. EPA, 2022; TEPA, 2022a). Both 9-year and 1-year hourly PM2.5 datasets were acquired from the AQMS, respectively denoted as AQMS9yr (from 2013 to 2021) and AQMS1yr (from September 2013 to August 2014). The data collection period for AQMS1yr was exactly the same as the period for conducting MMS samplings (as described in the next section) (TEPA, 2022b).

2.3 The Establishment of the Mobile Monitoring System (MMS) and Data Collection

An electric vehicle platform, equipped with a PM2.5 aerosol monitor (DUSTTRAK II, model 8530, TSI Inc.), was utilized as MMS. The monitor has a detection limit ranging from 0.001 to 150 mg cm–3, operates at a flow rate of 3 L min1, and provides a time resolution of 1 second. Additionally, a GPS receiver was integrated into the platform to capture longitude and latitude coordinates with a detection limit < 3 meters and a time resolution 1 second. To ensure accurate sampling, the height of the sampling pipeline was set at 2.2 meters above the ground, designed to facilitate isokinetic sampling. PM2.5 and GPS data (in seconds) from the MMS were integrated into hourly averages and further calibrated using the Federal Reference Method/Federal Equivalent Method (FRM/FEM) from the literature (Yanosky et al., 2002; TSI, 2013) as a reference standard for comparison with corresponding AQMS measurements. The information about MMS and its quality assurance and control (QA/QC) have been described in our previous research (Ho et al., 2022).

The sampling campaign covered four seasons (i.e., September 2013–August 2014, same as the period covered by AQMS1yr) to control for possible seasonal confounding. The sampling time was set from 06:00 to 23:00 to cover most of the residents' daily activity period, and the MMS sampled twice a day on the same route forth and back, mainly covering two rush-hour periods (07:00–10:00 and 17:00–20:00). The entire sampling route was ~80 kilometers, and it took approximately three hours to complete a round-trip sampling process. In total, 73 days of data were collected, including 15, 18, 21, and 19 days from autumn (September–November), winter (December–February), spring (March–May), and summer (June–August), respectively. The sampling area and route are shown in Fig. 1.

2.4 Characterizing Spatiotemporal Variations in PM2.5 of the Rooftop Level for the Study Area

In the present study, the AQMS + h_LUR approach was adopted for characterizing spatiotemporal variations in PM2.5 of the rooftop level for the study area. The approach is similar to those used in previous studies (Vlaanderen et al., 2019; Xu et al., 2021). The Hybrid Kriging-LUR model developed by our research group was adopted as the h_LUR based on AQMS data, and 26 potential variables obtained from multiple geo-spatial databases (including the air pollutant & meteorological database, GIS database, and satellite database) (Wu et al., 2018; Wong et al., 2021). The above potential variables were further classified into three categories, including 7 air pollutants and meteorological parameters (including PM2.5, SO2, NOx, NO2, O3, temperature, and relative humidity), 18 GIS point/polygon/grid parameters (including various buffers around specific features such as Chinese restaurants, temples, paddy fields, fruit orchards, residential areas, industrial areas, commercial areas, rivers, airports, industrial parks, local roads, major roads, all roads, and elevation), and 1 satellite-derived parameters (i.e., the normalized difference vegetation index (NDVI)) (Wu et al., 2018). The bivariate association between AQMS PM2.5 concentrations and each potential variable was determined through the Spearman correlation analysis. Only potential variables with an absolute correlation coefficient greater than 0.4 were retained for further analysis. Detailed information is described in the supplemental material Table S1 of our published article (Wong et al., 2021). In summary, a total of 7 statistically significant variables were included in the h_LUR model for predicting PM2.5 concentrations (i.e., PM2.5_h_LUR), including the kriging estimated PM2.5 levels of different sites in the target city (i.e., PM2.5_kriging), 3 co-pollutants (i.e., SO2_kriging, O3_kriging, and NO2_kriging), distance to the nearest airport (i.e., distanceairport), forest within a 5000-m circular buffer (i.e., forest5000m), and farmland within a 4000-m circular buffer (i.e., farmland4000m). For the data training process, the machine learning technique XGBoost was employed. The resultant h_LUR model (R2 = 0.94, RMSE = 4.41 µg m–3, Variance Inflation Factor (VIF) < 3) is described as the following:


In the present study, both PM2.5_h_LUR of 1-year and 9-year (i.e., h_LUR1yr and h_LUR9yr) were obtained through the application of the above h_LUR model based on PM2.5 concentrations obtained from AQMS1yr and AQMS9yr, respectively.

2.5 Characterizing Spatiotemporal Variations in PM2.5 of the Ground Level for the Study Area

In the present study, the AQMS + h_LUR + MMS approach modified from previous studies was adopted for characterizing spatiotemporal variations in PM2.5 of the ground level for the study area (Hankey and Marshall, 2015a; Dröge et al., 2018; Piotrowicz and Polednik, 2019; Xu et al., 2021). Here, PM2.5 concentrations of the ground level was estimated using MMS1yr as the dependent variable, and h_LUR1yr (i.e., rooftop concentrations) and potential co-contributors as the independent variables via multivariate linear regression (MLR) analysis. In the present study, all potential co-contributors, including variables associated with anthropogenic emissions (such as CO, NOx, SO2, O3, VOCs) and meteorological conditions [such as temperature (TEMP), relative humidity (RH), rainfall (RF), wind speed (WS)], were collected (Saraswat et al., 2013; Ye et al., 2018; Zhang et al., 2018; Guo et al., 2019; Lee et al., 2020; Rittner et al., 2020; Yousefian et al., 2020; Yang et al., 2022). Through linear correlation analyses variables exhibiting significant linear correlation (i.e., r > 0.4) with the MMS1yr were selected for establishing the preliminary predictive model. Finally, the stepwise MLR analysis was adopted for further selecting significant variable in the predictive model as the following:


In the present study, the above predictive model was also used to predict the long-term ground annual concentrations of MMS9yr based on the derived h_LUR9yr.

2.6 Conducting Health Impact Analyses

In the present study, the burden of disease (BOD) was estimated while conducting health impact analyses (HIA) for the three long-term PM2.5 concentration datasets (i.e., AQMS9yr, h_LUR9yr, and MMS9y). Five specific diseases, including ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD), lung cancer (LC), and lower respiratory tract infection (LRI), were selected for determining their BOD based on the dose-response relationship obtained from the global exposure mortality model (GEMM) (Ostro, 2004; Forouzanfar et al., 2016; Burnett et al., 2018). For each selected disease, the attributed fraction (AF) was adopted to describe the contribution of specific diseases to the health burden. Here, AF can be described as the following (Steenland and Armstrong, 2006):


where RR = the relative risk.

The RR can be substituted by the hazard ration (HR) if the given disease mortality probability is small enough (e.g., < 0.1) (Burnett et al., 2018).

The annual geometric mean (GM) PM2.5 concentration of the three established datasets (i.e., AQMS9yr, h_LUR9yr, and MMS9y) were used for deriving the mortality hazard ratio (HR) through the application of the GEMM function as the following (Burnett et al., 2018):


where z = max (0, C - Crf), C is the exposed PM2.5 concentration, Crf is the counterfactual PM2.5 concentration (= 2.4 µg m–3), and θ, SE θ, α, µ, ν are the parameters can be determined based on the causes of death (Table 1).

Table 1. The values for parameter used in the GEMM.

2.7 Statistical Methods

Since distributions of all PM2.5 datasets are consistently in a log-normality form, the geometric mean (GM) is considered as a suitable proxy for characterizing the central tendency of the data distribution. All data aggregation and statistical analyses were conducted using Microsoft Excel 365 software, R software (R x64 3.6.2), and relevant R packages (Patil, 2021; R, 2021).


3.1 The Comparison of AQMS1yr, h_LUR1yr, and MMS1yr

The descriptive statistics of the three datasets of AQMS1yr, h_LUR1yr, and MMS1yr are shown in Table 2. Datasets were further classified into four seasons based on the sampling period: autumn (September–November), winter (December–February), spring (March–May), and summer (June–August).

Table 2. PM2.5 concentrations (GM (95% CI): µg m–3) of AQMS1yr, h_LUR1yr, and MMS1yr.

A total of 779 hourly PM2.5 concentration data (AQMS1yr) were acquired from the Taiwan Ministry of Environment (MOENV) website, and corresponding data were simultaneously collected from MMS during the period from September 2013 and August 2014. It can be seen that PM2.5 concentrations in AQMS1yr [14.1 (13.1, 15.2) µg m–3] met the PM2.5 year-average standard (15 µg m–3) in Taiwan. Significant difference among seasons was found (p < 0.001, ANOVA) with the lowest in summer 6.1 (5.2–7.1) µg m–3 and the highest in spring 22.5 (20.9–24.2) µg m–3. Table 2 also shows the 779 hourly h_LUR1yr data which were converted from AQMS1yr data based on the predictive model described in our previous study (Wong et al., 2021). Here, PM2.5 concentrations in h_LUR1yr [18.2 (17.3, 19.1) µg m–3] obviously exceeded the PM2.5 year-average standard (15 µg m–3). There are significant differences among the PM2.5 concentrations for the four seasons (p < 0.001, ANOVA), and the lowest was in summer 9.6 (8.8, 10.5) µg m–3 and the highest was in spring 25.5 (23.9, 27.3) µg m–3. For MMS1yr, 779 hourly data were collected from the established MMS during a 73-day sampling period (including 15, 18, 21, and 19 days from autumn, winter, spring, and summer, respectively). MMS1yr was found with PM2.5 concentrations 25.1 (24.1, 26.1) µg m–3, which also exceeded the PM2.5 year-average standard (15 µg m–3). Significant differences can also be found among seasons (p < 0.001, ANOVA), and the lowest was in autumn 18.6 (17.4, 20.0) µg m–3 and the highest in winter 31.7 (29.1, 34.5) µg m–3.

Among the three datasets, PM2.5 levels in MMS1yr were significantly higher than that in h_LUR1yr and then AQMS1yr (p < 0.001; ANOVA test). PM2.5 levels in MMS1yr and h_LUR1yr were ~11 µg m3 (78%) and 4.1 µg m–3 (29%) in magnitude higher than that of AQMS1yr, respectively (Fig. 2). This finding aligns with expectations since MMS1yr measurements were taken at the ground level, where PM2.5 levels are considered more susceptible to various nearby emission sources (e.g., traffic, factories, and construction activities, etc.). Similar results have also been reported in the study conducted by Li et al. (2018). The significant difference between AQMS1yr and h_LUR1yr indicates that there is spatial variations in PM2.5 concentrations at the roof level. Therefore, it is concluded that simply using AQMS1yr or h_LUR1yr data alone may result in the underestimation of the resident's exposures, and eventually leads to biased estimation in population health burdens.

Fig. 2. Boxplot of the annual PM2.5 concentrations of the three datasets of MMS1yr, h_LUR1yr, and AQMS1yr.Fig. 2Boxplot of the annual PM2.5 concentrations of the three datasets of MMS1yr, h_LUR1yr, and AQMS1yr.

Season environmental factors are known affecting PM2.5 levels, therefore, seasonal differences among the three datasets were also examined. Among the four seasons, PM2.5 concentrations of the ground level are consistently significantly higher than that of the rooftop (i.e., MMS1yr > h_LUR1yr > AQMS1yr, p < 0.05; ANOVA), with an exception in autumn (no significant difference among the three datasets (p > 0.05; ANOVA) (Fig. 3). Nevertheless, the above results indicate that PM2.5 concentrations of the four seasons should be collected for better characterizing residents’ exposures.

Fig. 3. Boxplot of the seasonal PM2.5 concentrations of the three datasets of MMS1yr, h_LUR1yr, and AQMS1yr.Fig. 3. Boxplot of the seasonal PM2.5 concentrations of the three datasets of MMS1yr, h_LUR1yr, and AQMS1yr.

In summary, the data collected by AQMS are not only insufficient to fully reflect the spatial variation of rooftop PM2.5 concentration, but also to reflect PM2.5 exposures of residents.

3.2 Establishing Predictive Models for Converting PM2.5 Concentrations of the Roof Level to the Ground Level

For establishing the ground PM2.5 concentration predictive model, linear correlation analyses were first conducted to identify influencing factors for those significantly correlated with the PM2.5 concentrations. Results show that lnCO, lnNOx, lnSO2, and lnWS exhibit moderate significant correlations (r > 0.4, Pearson Correlation) (Fig. 4). Consequently, lnh_LUR1yr, lnCO, lnNOx, lnSO2, and lnWS were considered as independent variables in the predictive model for predicting MMS1yr.

Fig. 4. Correlation and scatter diagram of PM2.5 and atmospheric parameters.Fig. 4. Correlation and scatter diagram of PM2.5 and atmospheric parameters.

The stepwise MLR analysis was adopted for further selecting significant variable in the predictive model. Only lnSO2 was significant and hence was included in the model. The resultant predictive model is very promising (R2 = 0.61; RMSE = 1.4 µg m–3). Since the obtained variance inflation factors (VIFs) for the model was 1.12 indicating no significant collinearity. The established predictive model is shown as the following:


In principle, MMS1yr can also be directly predicted based on AQMS1y. However, the above method was not adopted in the present study. Considering that lnh_LUR1yr data were much closed to MMS1yr spatially and it is expected to have better prediction than AQMS1y. The above inference is supported by several studies as comparing the resultant R2 and the value obtained from the present study (R2 = 0.61). For example, a study conducted in Belgium utilizing AQMS data for predicting MMS data on a 5-minute time scale, R2 ranging from 0.14 to 0.40 were yielded (Van den Bossche et al., 2015). Another study conducted in CA in 2017, which employed AQMS data to predict black carbon levels collected from MMS, an R2 value of 0.51 was obtained (Chambliss et al., 2020). A study conducted in China found a correlation 0.7 between AQMS data and personal PM2.5 (Jahn et al., 2013). Furthermore, it should be noted that h_LUR was adopted in the present study for replacing LUR models. Here, it was assumed that h_LUR would provide more promising results in prediction than those using traditional LUR models. The above inference is supported by a study where an LUR model was adopted to forecast PM2.5 concentrations measured by MMS. A higher R2 was found than those directly using AQMS data for predictions (Hankey and Marshall, 2015a).

3.2 Health Impact Assessments (HIA)

Five diseases [including ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD), lung cancer (LC), and lower respiratory tract infection (LRI)], which are known relevant to PM2.5 exposures, were selected for conducting HIA. Considering the chronic nature of these diseases, datasets adequately to characterize long-term PM2.5 exposure concentrations to residents becomes an important issue. To meet the above purpose, AQMS9yr data were first obtained from the installed AQMS. Then, AQMS9yr data were converted to h_LUR9yr using the predictive model obtained from our previous study (Wong et al., 2021). Finally, MMS9yr were estimated by applying the obtained h_LUR9yr to the ground predictive model established in the present study. Table 3 shows the yearly PM2.5 GM concentrations of AQMS9yr, h_LUR9yr, and MMS9yr. It can be seen that the PM2.5 concentrations in AQMS9yr, h_LUR9yr, and MMS9yr consistently decreased by year indicating significant efforts have been done by the government to improve air quality in Taiwan. The ground-level PM2.5 concentrations (MMS9yr) were found higher in magnitude than both AQMS9yr [7.1 (5.3, 8.7) µg m–3] and h_LUR9yr, [5.5 (4.0, 7.1) µg m–3]. Again the above finding is theoretically plausible as MMS data reflects proximity to ground traffic and other emissions. The difference between MMS9yr and h_LUR9yr indicates the existence of the vertical variation in PM2.5 concentrations. Here, it should be noted that though the PM2.5 concentration in AQMS9yr and h_LUR9yr in 2021 (11.0 µg m–3 and 12.3 µg m–3) met the WHO interim target 3 level (15 µg m–3) and slightly exceeded target 4 level (10 µg m–3) (WHO, 2021), the corresponding concentration in MMS9yr (16.8 µg m–3) still failed to comply with the above two target levels. The above result suggests that relying solely on AQMS monitoring data or h_LUR modelling results are inadequate to characterize residents' PM2.5 exposures.

Table 3. PM2.5 concentrations (GM) and attributed fraction (AF) (GM (95% CI)) of the five target diseases for AQMS9yr, h_LUR9yr, and MMS9yr.

Table 3 also shows the AF of the five target diseases (i.e., LRI, IHD, LC, COPD, and stroke) using the GEMM model. Since PM2.5 levels decrease by year, the corresponding decrease in AF is theoretically plausible. MMS9yr exhibited higher AF value for the five target diseases as in comparison with those obtained from both AQMS9yr and h_LUR9yr (Table 3). By reference to AFs obtained from MMS9yr in 2021, the use of AQMS9yr would lead to underestimation in AFs [i.e., (AFMMS-AFAQMS)/AFMMS] by 35% for LRI, 21% for IHD, 29% for LC, 27% for COPD, and 36% for stroke, respectively. The similar trend can also be seen for those using h_LUR9yr in the obtained AFs (i.e., 26%, 18%, 24%, 20%, and 21% for LRI, IHD, LC, COPD, and stroke, respectively). The above underestimations in AFs indicating the importance of obtaining adequate long-term ground level concentrations (i.e., MMS9yr) in conducting HIA.

As we examine the decrease in AF per unit PM2.5 reduction [= ΔAF/ΔPM2.5; in % (µg m–3)1] over the 9-year period for the five target diseases, the values obtained from MMS9yr (i.e., LRI: 1.44, IHD: 0.80, LC: 0.80, COPD: 0.72, and stroke: 0.80) were lower than that of AQMS9yr (LRI: 1.91, IHD: 0.91, LC: 0.91, COPD: 0.82, and stroke: 0.91) and h_LUR9yr (LRI: 1.71, IHD: 0.98, LC: 0.89, COPD: 0.73, and stroke: 0.80). These findings highlight the potential for overestimating the benefit on reducing AF per unit PM2.5 reduction if AQMS9yr or h_LUR9yr were adopted for replacing MMS9yr. Therefore, the establishment of MMS9yr has become the crucial issue for governmental agencies to enact suitable PM2.5 abatement strategies from the abatement of resident’s health impact aspect.

The present study has two limitations that should be acknowledged, since they might affect the performance of the established predictive model. Firstly, the lack of volatile organic compounds (VOCs) measurements by the AQMS monitoring resulted in the exclusion of these compounds from the predictive model. Secondly, the sample sizes for the MMS sampling campaign was limited due to financial and equipment constraints.


Though the whole study was conducted in a small-scale city, ground-level PM2.5 concentrations (i.e., MMS data) were higher than that of the rooftop level (i.e., AQMS or h_LUR data). The established ground predictive model was found with R2 = 61% indicating that the approach developed by the present studies suitable for establishing long-term ground-level PM2.5 concentrations. Over the study period, PM2.5 concentrations in AQMS9yr, h_LUR9yr, and MMS9yr datasets consistently decreased by year, reflecting significant efforts have been done by governments in Taiwan. This downward trend can also be seen in the AF of the five target diseases, highlighting the positive impact of lowering ambient PM2.5 levels on disease prevention. However, though the PM2.5 concentrations measured by AQMS in 2021 met the WHO interim target 3 level, the corresponding concentrations in MMS9yr were higher than the above target level suggesting that relying solely on AQMS monitoring data is inadequate for protecting residents from PM2.5 exposures. Additionally, the HIA results indicate that using AQMS and h_LUR may underestimate the AF for the five target diseases. However, the employment of the MMS9yr data would be able to provide more accurate estimations of disease burdens even in a small city.


The authors would like to thank the National Health Research Institute in Taiwan for funding this research project. The authors, Ming-Yeng Lin, and Chih-Da Wu have the same contribution to this research project as the corresponding author.


Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data availability Data will be made available on request.

CRediT Authorship Contribution Statement

Study design: Perng-Jy Tsai, Chih-Da Wu, and Ming-Yeng Lin; Field data collection: Ming-Shing Ho, Ming-Yeng Lin, Li-Hao Young, Yu-Cheng Chen, and Hui-Tsung Hsu; Statistical analyses: Ming-Shing Ho, Bing-Fang Hwang, and Jung-Der Wang; Manuscript preparation: Ming-Shing Ho, Ming-Yeng Lin, Perng-Jy Tsai.


  1. Bai, K., Ma, M., Chang, N.B., Gao, W. (2019). Spatiotemporal trend analysis for fine particulate matter concentrations in China using high-resolution satellite-derived and ground-measured PM2.5 data. J. Environ. Manage. 233, 530–542. https://doi.org/10.1016/j.jenvman.2018.12.071

  2. Borge, R., Narros, A., Artinano, B., Yague, C., Gomez-Moreno, F.J., de la Paz, D., Roman-Cascon, C., Diaz, E., Maqueda, G., Sastre, M., Quaassdorff, C., Dimitroulopoulou, C., Vardoulakis, S. (2016). Assessment of microscale spatio-temporal variation of air pollution at an urban hotspot in Madrid (Spain) through an extensive field campaign. Atmos. Environ. 140, 432–445. https://doi.org/10.1016/j.atmosenv.2016.06.020

  3. Burnett, R., Chen, H., Szyszkowicz, M., Fann, N., Hubbell, B., Pope, C.A., Apte, J.S., Brauer, M., Cohen, A., Weichenthal, S., Coggins, J., Di, Q., Brunekreef, B., Frostad, J., Lim, S.S., Kan, H., Walker, K.D., Thurston, G.D., Hayes, R.B., Lim, C.C., et al. (2018). Global estimates of mortality associated with long-term exposure to outdoor fine particulate matter. Proc. Natl. Acad. Sci. U.S.A. 115, 9592–9597. https://doi.org/10.1073/pnas.1803222115

  4. Burnett, R.T., Pope, C.A., Ezzati, M., Olives, C., Lim, S.S., Mehta, S., Shin, H.H., Singh, G., Hubbell, B., Brauer, M., Anderson, H.R., Smith, K.R., Balmes, J.R., Bruce, N.G., Kan, H., Laden, F., Prüss-Ustün, A., Turner, M.C., Gapstur, S.M., Diver, W.R., et al. (2014). An integrated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environ. Health Perspect. 122, 397–403. https://doi.org/10.1289/ehp.1307049

  5. Cabada, J.C., Khlystov, A., Wittig, A.E., Pilinis, C., Pandis, S.N. (2004). Light scattering by fine particles during the Pittsburgh Air Quality Study: Measurements and modeling. J. Geophys. Res. 109, 2003JD004155. https://doi.org/10.1029/2003JD004155

  6. Caplin, A., Ghandehari, M., Lim, C., Glimcher, P., Thurston, G. (2019). Advancing environmental exposure assessment science to benefit society. Nat. Commun. 10, 1236. https://doi.org/​10.1038/s41467-019-09155-4

  7. Castell, N., Dauge, F.R., Schneider, P., Vogt, M., Lerner, U., Fishbain, B., Broday, D., Bartonova, A. (2017). Can commercial low-cost sensor platforms contribute to air quality monitoring and exposure estimates? Environ. Int. 99, 293–302. https://doi.org/10.1016/j.envint.2016.12.007

  8. Chambliss, S.E., Preble, C.V., Caubel, J.J., Cados, T., Messier, K.P., Alvarez, R.A., LaFranchi, B., Lunden, M., Marshall, J.D., Szpiro, A.A., Kirchstetter, T.W., Apte, J.S. (2020). Comparison of mobile and fixed-site black carbon measurements for high-resolution urban pollution mapping. Environ. Sci. Technol. 54, 7848–7857. https://doi.org/10.1021/acs.est.0c01409

  9. Cheng, Y.H., Ninh, X.H., Yeh, S.L. (2019). Dominant factors influencing the concentrations of particulate matters inside train carriages traveling in different environments in the Taipei mass rapid transit system. Aerosol Air Qual. Res. 19, 1579–1592. https://doi.org/10.4209/aaqr.​2018.09.0335

  10. Chu, Y.Y., Liu, Y.S., Li, X.Y., Liu, Z.Y., Lu, H.S., Lu, Y.A., Mao, Z.F., Chen, X., Li, N., Ren, M., Liu, F.F., Tian, L.Q., Zhu, Z.M., Xiang, H. (2016). A review on predicting ground PM2.5 concentration using satellite aerosol optical depth. Atmosphere 7, 129. https://doi.org/10.3390/atmos7100129

  11. Dias, D., Tchepel, O. (2018). Spatial and temporal dynamics in air pollution exposure assessment. Int. J. Environ. Res. Public Health 15, 558. https://doi.org/10.3390/ijerph15030558

  12. Dröge, J., Müller, R., Scutaru, C., Braun, M., Groneberg, D.A. (2018). Mobile measurements of particulate matter in a car cabin: Local variations, contrasting data from mobile versus stationary measurements and the effect of an opened versus a closed window. Int. J. Environ. Res. Public Health 15, 2642. https://doi.org/10.3390/ijerph15122642

  13. Environmental Protection Administration, Executive Yuan (TEPA) (2022a). Manual monitoring of fine particulate matter. (accessed 2 December 2022).

  14. Environmental Protection Administration, Executive Yuan (TEPA) (2022b). Taiwan Air Quality Historical Monitoring Data. (accessed 2 December 2022).

  15. Forouzanfar, M.H., Afshin, A., Alexander, L.T., Anderson, H.R., Bhutta, Z.A., Biryukov, S., Brauer, M., Burnett, R., Cercy, K., Charlson, F.J. (2016). Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet 388, 1659–1724. https://doi.org/10.1016/S0140-6736(16)31679-8

  16. Fuller, R., Landrigan, P.J., Balakrishnan, K., Bathan, G., Bose-O’Reilly, S., Brauer, M., Caravanos, J., Chiles, T., Cohen, A., Corra, L., Cropper, M., Ferraro, G., Hanna, J., Hanrahan, D., Hu, H., Hunter, D., Janata, G., Kupka, R., Lanphear, B., Lichtveld, M., et al. (2022). Pollution and health: a progress update. Lancet Planet. Heath 6, e535–e547. https://doi.org/10.1016/S2542-5196(22)​00090-0

  17. Guo, H., Gu, X., Ma, G., Shi, S., Wang, W., Zuo, X., Zhang, X. (2019). Spatial and temporal variations of air quality and six air pollutants in China during 2015-2017. Sci. Rep. 9, 15201. https://doi.org/​10.1038/s41598-019-50655-6

  18. Hankey, S., Marshall, J.D. (2015a). Land use regression models of on-road particulate air pollution (particle number, black carbon, PM2.5, particle size) using mobile monitoring. Environ. Sci. Technol. 49, 9194–9202.  ttps://doi.org/10.1021/acs.est.5b01209

  19. Hankey, S., Marshall, J.D. (2015b). On-bicycle exposure to particulate air pollution: Particle number, black carbon, PM2.5, and particle size. Atmos. Environ. 122, 65–73. https://doi.org/10.1016/j.​atmosenv.2015.09.025

  20. Ho, M.S., Lin, M.Y., Wang, J.D., Young, L.H., Hsu, H.T., Hwang, B.F., Tsai, P.J. (2022). An integrated approach to characterize temporal–spatial variations in PM2.5 concentrations at the ground level and its implication on health impact assessments. Front. Environ. Sci. 10, 901754. https://doi.org/10.3389/fenvs.2022.901754

  21. International Agency for Research on Cancer (IARC) (2013). Air Pollution and cancer (IARC Scientific Publication No. 161). (accessed 12 Jun 2020).

  22. Jahn, H.J., Kraemer, A., Chen, X.C., Chan, C.Y., Engling, G., Ward, T.J. (2013). Ambient and personal PM2.5 exposure assessment in the Chinese megacity of Guangzhou. Atmos. Environ. 74, 402–411. https://doi.org/10.1016/j.atmosenv.2013.04.011

  23. Lee, M., Lin, L., Chen, C.Y., Tsao, Y., Yao, T.H., Fei, M.H., Fang, S.H. (2020). Forecasting air quality in Taiwan by using machine learning. Sci. Rep. 10, 4153. https://doi.org/10.1038/s41598-020-61151-7

  24. Li, J., Zhang, H., Chao, C.Y., Chien, C.H., Wu, C.Y., Luo, C.H., Chen, L.J., Biswas, P. (2020a). Integrating low-cost air quality sensor networks with fixed and satellite monitoring systems to study ground-level PM2.5. Atmos. Environ. 223, 117293. https://doi.org/10.1016/j.atmosenv.​2020.117293

  25. Li, J.Y., Mattewal, S.K., Patel, S., Biswas, P. (2020b). Evaluation of nine low-cost-sensor-based particulate matter monitors. Aerosol Air Qual. Res. 20, 254–270. https://doi.org/10.4209/​aaqr.2018.12.0485

  26. Li, Z.Y., Fung, J.C.H., Lau, A.K.H. (2018). High spatiotemporal characterization of on-road PM2.5 concentrations in high-density urban areas using mobile monitoring. Build. Environ. 143, 196–205. https://doi.org/10.1016/j.buildenv.2018.07.014

  27. Lin, M.Y., Guo, Y.X., Chen, Y.C., Chen, W.T., Young, L.H., Lee, K.J., Wu, Z.Y., Tsai, P.J. (2018). An instantaneous spatiotemporal model for predicting traffic-related ultrafine particle concentration through mobile noise measurements. Sci. Total Environ. 636, 1139–1148. https://doi.org/​10.1016/j.scitotenv.2018.04.248

  28. Ministry of Digital Affairs (MODA) (2022). Government open data, Taiwan. (accessed 1 May 2022).

  29. Ministry of the Interior (MOI) (2022). Monthly Bulletin of Interior Statistics, Taiwan. (accessed 3 February 2023).

  30. Miri, M., Ghassoun, Y., Dovlatabadi, A., Ebrahimnejad, A., Lowner, M.O. (2019). Estimate annual and seasonal PM1, PM2.5 and PM10 concentrations using land use regression model. Ecotoxicol. Environ. Saf. 174, 137–145. https://doi.org/10.1016/j.ecoenv.2019.02.070

  31. Monrad, M., Sajadieh, A., Christensen, J.S., Ketzel, M., Raaschou-Nielsen, O., Tjonneland, A., Overvad, K., Loft, S., Sorensen, M. (2017). Long-term exposure to traffic-related air pollution and risk of incident atrial fibrillation: A cohort study. Environ. Health Perspect. 125, 422–427. https://doi.org/10.1289/EHP392

  32. Morawska, L., Thai, P.K., Liu, X., Asumadu-Sakyi, A., Ayoko, G., Bartonova, A., Bedini, A., Chai, F., Christensen, B., Dunbabin, M., Gao, J., Hagler, G.S.W., Jayaratne, R., Kumar, P., Lau, A.K.H., Louie, P.K.K., Mazaheri, M., Ning, Z., Motta, N., Mullins, B., et al. (2018). Applications of low-cost sensing technologies for air quality monitoring and exposure assessment: How far have they gone? Environ. Int. 116, 286–299. https://doi.org/10.1016/j.envint.2018.04.018

  33. Mustafić, H., Jabre, P., Caussin, C., Murad, M.H., Escolano, S., Tafflet, M., Périer, M.C., Marijon, E., Vernerey, D., Empana, J.P., Jouven, X. (2012). Main air pollutants and myocardial infarction: a systematic review and meta-analysis. JAMA 307, 713–721. https://doi.org/10.1001/jama.​2012.126

  34. Okokon, E.O., Yli-Tuomi, T., Turunen, A.W., Taimisto, P., Pennanen, A., Vouitsis, I., Samaras, Z., Voogt, M., Keuken, M., Lanki, T. (2017). Particulates and noise exposure during bicycle, bus and car commuting: A study in three European cities. Environ. Res. 154, 181–189. https://doi.org/10.1016/j.envres.2016.12.012

  35. Ostro, B. (2004). Outdoor air pollution: assessing the environmental burden of disease at national and local levels. World Health Organization, Geneva. (accessed 15 May 2022).

  36. Patil, I. (2021). Visualizations with statistical details: The'ggstatsplot'approach. J. Open Res. Softw. 6, 3167. https://doi.org/10.21105/joss.03167

  37. Piotrowicz, A., Polednik, B. (2019). Exposure to aerosols particles on an urban road. J. Ecol. Eng. 20, 27–34. https://doi.org/10.12911/22998993/105329

  38. R (2021). R: A Language and Environment for Statistical Computing, Vienna, Austria. (accessed 10 Jun 2021).

  39. Rittner, R., Gustafsson, S., Spanne, M., Malmqvist, E. (2020). Particle concentrations, dispersion modelling and evaluation in southern Sweden. SN Appl. Sci. 2, 15. https://doi.org/10.1007/​s42452-020-2769-1

  40. Saraswat, A., Apte, J.S., Kandlikar, M., Brauer, M., Henderson, S.B., Marshall, J.D. (2013). Spatiotemporal land use regression models of fine, ultrafine, and black carbon particulate matter in New Delhi, India. Environ. Sci. Technol. 47, 12903–12911. https://doi.org/10.1021/​es401489h

  41. Shen, J.L., Gao, Z. (2019). Commuter exposure to particulate matters in four common transportation modes in Nanjing. Build. Environ. 156, 156–170. https://doi.org/10.1016/j.buildenv.2019.04.​018

  42. Snyder, E.G., Watkins, T.H., Solomon, P.A., Thoma, E.D., Williams, R.W., Hagler, G.S., Shelow, D., Hindin, D.A., Kilaru, V.J., Preuss, P.W. (2013). The changing paradigm of air pollution monitoring. Environ. Sci. Technol. 47, 11369–11377. https://doi.org/10.1021/es4022602

  43. Steenland, K., Armstrong, B. (2006). An overview of methods for calculating the burden of disease due to specific risk factors. Epidemiology 17, 512–519. https://doi.org/10.1097/01.ede.0000​229155.05644.43

  44. Thongthammachart, T., Araki, S., Shimadera, H., Matsuo, T., Kondo, A. (2022). Incorporating Light Gradient Boosting Machine to land use regression model for estimating NO2 and PM2.5 levels in Kansai region, Japan. Environ. Model. Software 155, 105447. https://doi.org/10.1016/j.​envsoft.2022.105447

  45. TSI (2013). Rationale for Programming: A PCF of 0.38 for Ambient Monitoring (EXPMN-007) A4. (accessed 18 May 2020).

  46. U.S. Environmental Protection Agency (U.S. EPA) (2019). Integrated Science Assessment (ISA) for Particulate Matter (Final Report, Dec 2019). (accessed 15 May 2021).

  47. U.S. Environmental Protection Agency (U.S. EPA) (2021). Air Sensor Toolbox. (accessed 2 March 2021).

  48. U.S. Environmental Protection Agency (U.S. EPA) (2022). EPA scientists develop and evaluate Federal Reference & Equivalent Methods for measuring key air pollutants. (accessed 1 September 2019).

  49. Van den Bossche, J., Peters, J., Verwaeren, J., Botteldooren, D., Theunis, J., De Baets, B. (2015). Mobile monitoring for mapping spatial variation in urban air quality: Development and validation of a methodology based on an extensive dataset. Atmos. Environ. 105, 148–161. https://doi.org/10.1016/j.atmosenv.2015.01.017

  50. van den Brenk, I. (2018). The use of Health Impact Assessment tools in European Cities. (accessed 5 March 2021).

  51. van Donkelaar, A., Martin, R.V., Brauer, M., Kahn, R., Levy, R., Verduzco, C., Villeneuve, P.J. (2010). Global estimates of ambient fine particulate matter concentrations from satellite-based aerosol optical depth: Development and application. Environ. Health Perspect. 118, 847–855. https://doi.org/10.1289/ehp.0901623

  52. Vlaanderen, J., Portengen, L., Chadeau-Hyam, M., Szpiro, A., Gehring, U., Brunekreef, B., Hoek, G., Vermeulen, R. (2019). Error in air pollution exposure model determinants and bias in health estimates. J. Exposure Sci. Environ. Epidemiol. 29, 258–266. https://doi.org/10.1038/s41370-018-0045-x

  53. Williams, R., Duvall, R., Kilaru, V., Hagler, G., Hassinger, L., Benedict, K., Rice, J., Kaufman, A., Judge, R., Pierce, G., Allen, G., Bergin, M., Cohen, R.C., Fransioli, P., Gerboles, M., Habre, R., Hannigan, M., Jack, D., Louie, P., Martin, N.A., et al. (2019). Deliberating performance targets workshop: Potential paths for emerging PM2.5 and O3 air sensor progress. Atmos. Environ.: X 2, 100031. https://doi.org/10.1016/j.aeaoa.2019.100031

  54. Wong, P.Y., Lee, H.Y., Chen, Y.C., Zeng, Y.T., Chern, Y.R., Chen, N.T., Lung, S.C.C., Su, H.J., Wu, C.D. (2021). Using a land use regression model with machine learning to estimate ground level PM2.5. Environ. Pollut. 277, 116846. https://doi.org/10.1016/j.envpol.2021.116846

  55. World Health Organization (WHO) (2012). Sustainable Development Goals target 11.6. (accessed 2 May 2021).

  56. World Health Organization (WHO) (2016). Health Risk Assessment of Air Pollution—General Principles. (accessed 2 May 2019).

  57. World Health Organization (WHO) (2021). Global air quality guidelines: particulate matter (PM2.5 and PM10), ozone, nitrogen dioxide, sulfur dioxide and carbon monoxide. (accessed 2 May 2022).

  58. World Health Organization (WHO) (2023). WHO Ambient Air quality database. (accessed 2 July 2022).

  59. Wu, C.D., Zeng, Y.T., Lung, S.C.C. (2018). A hybrid kriging/land-use regression model to assess PM2.5 spatial-temporal variability. Sci. Total Environ. 645, 1456–1464. https://doi.org/​10.1016/j.scitotenv.2018.07.073

  60. Xu, X., Qin, N., Yang, Z., Liu, Y., Cao, S., Zou, B., Jin, L., Zhang, Y., Duan, X. (2021). Potential for developing independent daytime/nighttime LUR models based on short-term mobile monitoring to improve model performance. Environ. Pollut. 268, 115951  https://doi.org/10.1016/j.envpol.​2020.115951

  61. Yang, X., Xiao, D., Bai, H., Tang, J., Wang, W. (2022). Spatiotemporal distributions of PM2.5 concentrations in the Beijing–Tianjin–Hebei region From 2013 to 2020. Front. Environ. Sci. 10, 842237. https://doi.org/10.3389/fenvs.2022.842237

  62. Yanosky, J.D., Williams, P.L., MacIntosh, D.L. (2002). A comparison of two direct-reading aerosol monitors with the federal reference method for PM2.5 in indoor air. Atmos. Environ. 36, 107–113. https://doi.org/10.1016/S1352-2310(01)00422-8

  63. Ye, W.F., Ma, Z.Y., Ha, X.Z. (2018). Spatial-temporal patterns of PM2.5 concentrations for 338 Chinese cities. Sci. Total Environ. 631–632, 524–533. https://doi.org/10.1016/j.scitotenv.​2018.03.057

  64. Yousefian, F., Faridi, S., Azimi, F., Aghaei, M., Shamsipour, M., Yaghmaeian, K., Hassanvand, M.S. (2020). Temporal variations of ambient air pollutants and meteorological influences on their concentrations in Tehran during 2012-2017. Sci. Rep. 10, 292. https://doi.org/10.1038/s41598-019-56578-6

  65. Zamora, M.L., Rice, J., Koehler, K. (2020). One year evaluation of three low-cost PM2.5 monitors. Atmos. Environ. 235, 117615. https://doi.org/10.1016/j.atmosenv.2020.117615

  66. Zhang, N., Huang, H., Duan, X., Zhao, J., Su, B. (2018). Quantitative association analysis between PM2.5 concentration and factors on industry, energy, agriculture, and transportation. Sci. Rep. 8, 9461. https://doi.org/10.1038/s41598-018-27771-w 

Share this article with your colleagues 


Subscribe to our Newsletter 

Aerosol and Air Quality Research has published over 2,000 peer-reviewed articles. Enter your email address to receive latest updates and research articles to your inbox every second week.

77st percentile
Powered by
   SCImago Journal & Country Rank

2022 Impact Factor: 4.0
5-Year Impact Factor: 3.4

Aerosol and Air Quality Research partners with Publons

CLOCKSS system has permission to ingest, preserve, and serve this Archival Unit
CLOCKSS system has permission to ingest, preserve, and serve this Archival Unit

Aerosol and Air Quality Research (AAQR) is an independently-run non-profit journal that promotes submissions of high-quality research and strives to be one of the leading aerosol and air quality open-access journals in the world. We use cookies on this website to personalize content to improve your user experience and analyze our traffic. By using this site you agree to its use of cookies.