Skip to main content
NCBI home page
VA Author Manuscripts logoLink to VA Author Manuscripts
. Author manuscript; available in PMC: 2022 Jun 1.
Published in final edited form as: Environ Int. 2021 Feb 19;151:106445. doi: 10.1016/j.envint.2021.106445

Estimation of ambient PM2.5 in Iraq and Kuwait from 2001 to 2018 using machine learning and remote sensing

Jing Li a, Eric Garshick b,c, Jaime E Hart a,c, Longxiang Li a, Liuhua Shi a,d, Ali Al-Hemoud e, Shaodan Huang a,*, Petros Koutrakis a
PMCID: PMC8023768  NIHMSID: NIHMS1678488  PMID: 33618328

Abstract

Iraq and Kuwait are in a region of the world known to be impacted by high levels of fine particulate matter (PM2.5) attributable to sources that include desert dust and ambient pollution, but historically have had limited pollution monitoring networks. The inability to assess PM2.5 concentrations have limited the assessment of the health impact of these exposures, both in the native populations and previously deployed military personnel. As part of a Department of Veterans Affairs Cooperative Studies Program health study of land-based U.S. military personnel who were previously deployed to these countries, we developed a novel approach to estimate spatially and temporarily resolved daily PM2.5 exposures 2001–2018. Since visibility is proportional to ground-level particulate matter concentrations, we were able to take advantage of extensive airport visibility data that became available as a result of regional military operations over this time period. First, we combined a random forest machine learning and a generalized additive mixed model to estimate daily high resolution (1 km × 1 km) visibility over the region using satellite-based aerosol optical depth (AOD) and airport visibility data. The spatially and temporarily resolved visibility data were then used to estimate PM2.5 concentrations from 2001 to 2018 by converting visibility to PM2.5 using empirical relationships derived from available regional PM2.5 monitoring stations. We adjusted for spatially resolved meteorological parameters, land use variables, including the Normalized Difference Vegetation Index, and satellite-derived estimates of surface dust as a measure of sandstorm activity. Cross validation indicated good model predictive ability (R2 = 0.71), and there were considerable spatial and temporal differences in PM2.5 across the region. Annual average PM2.5 predictions for Iraq and Kuwait were 37 and 41 μg/m3, respectively, which are greater than current U.S. and WHO standards. PM2.5 concentrations in many U.S. bases and large cities (e.g. Bagdad, Balad, Kuwait city, Karbala, Najaf, and Diwaniya) had annual average PM2.5 concentrations above 45 μg/m3 with weekly averages as high as 150 μg/m3 depending on calendar year. The highest annual PM2.5 concentration for both Kuwait and Iraq were observed in 2008, followed by 2009, which was associated with extreme drought in these years. The lowest PM2.5 values were observed in 2014. On average, July had the highest concentrations, and November had the lowest values, consistent with seasonal patterns of air pollution in this region. This is the first study that estimates long-term PM2.5 exposures in Iraq and Kuwait at a high resolution based on measurements data that will allow the study of health effects and contribute to the development of regional environmental policies. The novel approach demonstrated may be used in other parts of the world with limited monitoring networks.

Keywords: PM2.5, Aerosol optical depth (AOD), Visibility, High resolution, Exposure

1. Introduction

Since 2001, many U.S. military personnel and coalition military personnel have been deployed in support of operations in the Southwest Asia and experienced with frequent exposure to high concentrations of fine particulate matter (PM2.5) attributable to large arid and semiarid areas, local pollution from vehicles and industry, oil exploration and refining, and refuse-burning, including burn pits (Alolayan et al., 2013; Engelbrecht et al., 2009; Garshick et al., 2019). Epidemiologic studies based on military health encounter data have demonstrated more frequent post-deployment health encounters than non-deployed personnel for respiratory symptoms and for airway disease, predominantly asthma (Abraham et al., 2014; Pugh et al., 2016; Abraham et al., 2012; Sharkey et al., 2016). There is concern that deployers may have experienced reductions in pulmonary function or other health effects attributable to exposure (U.S. Institute of Medicine, 2011; Garshick et al., 2019). Moreover, exposure to air pollution has been directly linked to health effects worldwide (Pappin and Hakami, 2013; Stafoggia et al., 2017; Brauer et al., 2016; Wang et al., 2020; Zhang et al., 2021). Unlike health studies assessing PM-related health effects in the US, Europe, and parts of Asia, the lack of a comprehensive network of regional monitoring stations in countries such as Iraq and Kuwait has hindered the design of health effect studies, both in the native populations and previously deployed military personnel.

Visibility data are routinely collected at airports or meteorology stations throughout the world. The relationship between visibility and PM2.5 is relatively stable due to the light extinction (scattering and absorption) effects of particles with sizes similar to the wavelengths of visible light, particularly in arid areas (Abbey et al., 1995; Zhao et al., 2017; Liu et al., 2017; Vajanapoom et al., 2001). Visibility has been used previously to estimate historical ground-level PM2.5 exposures in China prior to the establishment of a PM2.5 monitoring network (Liu et al., 2017). Masri et al. (2017) conducted a pilot study that demonstrated the feasibility of using airport visibility data in Southwest Asia and Afghanistan. Visibility-based approach takes advantage of the large database of historical visibility collected by airport or meteorology stations, but the spatial resolution of estimated PM2.5 is still coarse for epidemiological studies.

Satellite-based aerosol optical depth (AOD), a measure of light attenuation in the column of air from Earth’s surface to the top of atmosphere which is affected by meteorology and land conditions (Paciorek and Liu, 2009), has been widely used to estimate spatiotemporally resolved PM2.5 exposure, given its extensive spatial coverage, high spatial resolution, and reliable repeated daily observations (Liu et al., 2005; Paciorek et al., 2008; Chudnovsky et al., 2012; Kloog et al., 2014; Beloconi et al., 2016; Brokamp et al., 2017; Huang et al., 2018; Di et al., 2019). The satellite-based approach has been applied to different study areas, such as the United States (Kloog et al., 2014; Di et al., 2016), China (Xue et al., 2019), and Europe (Beloconi et al., 2018).

Quantifying a relationship between AOD and PM2.5 usually involves incorporating ground-level PM2.5 measurements, metrological variables, and geographic covariates at high spatial and temporal resolutions. However, this method has not been applied to areas with a paucity of PM2.5 monitoring sites due to the lack of spatial coverage. There are no monitoring networks in the Middle East designed to provide comprehensive measurements. Instead of using ground-level PM2.5 measurements, some studies use a Chemical Transport Model (CTM) or observational meteorological values to simulate the relationship between AOD and PM2.5 globally (van Donkelaar et al., 2015; Lin et al., 2018). However, CTMs rely on accurate emission inventories, which are poorly developed in the Middle East. Therefore, current models that estimate PM2.5 globally using CTM inventories are unreliable in this region. Until our efforts, there have been no previous studies assessing historical PM2.5 exposures at high spatial resolution in the Middle East based on measurement data.

As part of a Department of Veterans Affairs Cooperative Studies Program health study of land-based U.S. military personnel, we predicted spatially and temporally resolved PM2.5 concentrations by a multi-staged approach. First, we used a hybrid machine learning model and incorporated multiple variables (including satellite-based AOD data, land-use terms, and meteorological variables) to predict high spatiotemporal resolution visibility, as visibility data throughout the region were available as result of military airport activity. We then derived relationships between daily PM2.5 and visibility using existing monitoring stations in the region to convert visibility to PM2.5. In the absence of extensive air quality monitoring sites, this approach leverages historical visibility data, satellite AOD, and existing regional PM2.5 monitoring stations to understand the intensity and spatiotemporal variability in PM2.5 exposures over Iraq and Kuwait.

2. Material and methods

2.1. Data collection

2.1.1. AOD data

We collected 1 km × 1 km AOD data for Iraq and Kuwait between January 1st, 2001 through December 31st, 2018 from the MODIS instrument, which is onboard Terra and Aqua, two Earth Observing System (EOS) satellites (Salomonson et al., 1989). Terra was launched in December 1999, and Aqua was launched in May 2002. The local crossing time over Kuwait and Iraq for Terra is approximately 10:30, while for Aqua it is approximately 13:30. We used Collection-6 Multi-Angle Implementation of Atmospheric Correction (MAIAC) AOD data (MCD19A2), which includes a correction that adjusts for reflection of bright light from desert regions, from the Terra for 2001–2018 and the Aqua for 2003–2018. (Lyapustin et al., 2018) Four sections (tiles) of the NASA global AOD database (tiles h22v05, h22v06, h21v05, h21v06) that included approximately 360,000 grid cells for each tile were used. We extracted AOD at 550 nm that met MODIS quality assurance criteria (flag of “good”), and data with extreme values (above 4) were excluded (Di et al., 2016). Due to cloud cover and technical factors such as imaging over some bright surfaces and cloud, AOD has missing approximately 40% (see Fig. S2 in supplement) (Xiao et al., 2017). Therefore, we calculated linear associations between the Aqua and Terra AOD for each day and imputed the missing MAIAC AOD values when values from only one instrument was available (Huang et al., 2018). Grid cells over large water bodies were excluded (https://www.diva-gis.org/datadown) (Fig. S1). The average AOD from Aqua and Terra was used for modeling.

2.1.2. Visibility data

Hourly visibility data for Kuwait and Iraq were provided by the U.S. Air Force 14th Weather Squadron for all airports and meteorological stations in Kuwait and Iraq. We also obtained additional hourly visibility data from National Oceanic and Atmospheric Administration (NOAA) Integrated Surface Database (ISD, https://www.ncdc.noaa.gov/isd). Seventy-five percent of the visibility data in this study was from U.S. Air Force and 25% was from NOAA. We calculated average daily visibility from the available hourly data. The distribution of the 263 visibility monitoring stations available in Iraq and Kuwait during the study period is shown in Fig. S3.

2.1.3. PM2.5 monitoring data

We obtained PM2.5 monitoring data in the study regions that were available over 3 different time periods and sites. A total of 1942 daily PM2.5 and concurrent visibility data were used in our model (Table S1). These included daily PM2.5 concentrations from U.S. Embassy Kuwait air quality monitor for 2017–2018 (https://kw.usembassy.gov/embassy/kuwait-city/air-quality-monitor/), PM2.5 concentrations we previously collected in three monitoring sites in Kuwait during 2004–2005 (Brown et al., 2008), and daily PM2.5 observations from two monitoring sites from 2017 to 2018 our group collected in Kuwait City.

2.1.4. Meteorological data

Meteorological data were obtained from the fifth generation European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) (https://cds.climate.copernicus.eu/cdsapp#!/search?type=dataset). ERA5 is a widely used global reanalysis, which has been verified by comparisons with observations (Hersbach et al., 2020). We considered 18 metrological variables, including temperature and dew point temperature at 2 m, u-wind (east–west component of the wind) speed at 10 m, v-wind (north–south component of the wind) speed at 10 m, instantaneous 10 m wind gust, total precipitation, surface pressure, downward UV radiation at the surface, cloud cover (high, medium, low, and total) evaporation, planetary boundary layer height, relative humidity at 1,000 hPa, vegetation cover (high and low), and forecast albedo (a measure of solar radiation)

2.1.5. Dust and particle data

Since Iraq and Kuwait were largely covered by the desert and are impacted by intense and frequent dust storms, we assessed a number of surface dust related parameters as predictors of PM2.5 levels. Dust-related (e.g. “Dust surface Mass Concentration”, “Dust Extinction AOD”, “Dust Angstrom Parameter”) and additional particle parameters (e.g. “Organic Carbon Column Mass Density”, “Black Carbon Column Mass Density”) (Supplemental Table S2) were obtained from the Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA-2) dataset (https://disc.gsfc.nasa.gov/datasets/M2TUNXAER_5.12.4/summary). MERRA-2 is an atmospheric reanalysis platform provided by NASA, which is agreed well with the results of other satellite observations and those of ground-based measurements (Ginoux et al., 2001; Rienecker et al., 2011; Colarco et al., 2010). MERRA-2 has a spatial resolution of 0.625°×0.5° (approximately 62.5 km × 50 km) and temporal resolution of one day.

2.1.6. Land use data

Land use parameters including Normalized Difference Vegetation Index (NDVI), elevation, distance to the nearest industrial area, and traffic density were included as model predictors (Table S2). NDVI data were obtained from the NOAA Climate Data Record (CDR) of Normalized Difference Vegetation Index (NDVI), Version 4 (https://data.nodc.noaa.gov/cgi-bin/iso?id=gov.noaa.ncdc:C00813). This dataset was acquired by the Advanced Very High Resolution Radiometer (AVHRR) sensor at a spatial resolution of 5 km × 5 km and a temporal resolution of 1 day. Elevation of each grid cell was extracted from ETOPO1 Global Relief Model (https://www.ngdc.noaa.gov/mgg/global/global.html). ETOPO1 is a 1 arc-minute global relief model of Earth’s surface that integrates land topography and ocean bathymetry. Information regarding industrial areas was provided by the U.S. National Geospatial-Intelligence Agency. Since industrial emissions may influence the relationship between AOD and visibility, we calculated the distance from the center of each grid cell to the nearest industrial area for use as a potential adjustment variable. Road data were obtained through OpenStreetMap (https://www.openstreetmap.org) and traffic density in each 1 km2 grid cell was calculated by the total Highway road length in the grid cell.

2.2. Model

In order to estimate PM2.5 concentrations in each 1 km × 1 km grid cell, we developed a 4-stage model (Fig. 1).

Fig. 1.

Fig. 1.

Open in a new tab

Flowchart of the four-stage modeling approach.

2.2.1. Stage 1

In stage 1, we used a random forest model to predict daily visibility based on MAIAC AOD, meteorological variables, MERRA-2 derived surface dust variables, and land use parameters. Random forest is a machine learning method for classification and regression which has been used to estimate PM2.5 (Huang et al., 2018; Di et al., 2019; Brokamp et al., 2017; Hu et al., 2017). The detailed predictors are shown in Supplemental Table S2. A total of 134,265 paired AOD-visibility observations were used for model training. We optimized model input parameters using a grid search approach (Di et al., 2019), with the input parameters listed in Supplemental Table S3. The final trained model was then used to predict daily visibility for each 1 km × 1 km grid cell with AOD.

We performed a 10-fold cross-validation by randomly dividing the visibility stations into 10%−90% splits (10 random groups). For each split, the model was trained with data from 90% of the visibility sites and was used to predict visibility in the 10% of sites not included in the training. This process was repeated 10 times, and the cross-validation R2 (CV R2) were computed.

2.2.2. Stage 2

For grid cells with AOD but without visibility measurements, we used the Stage 1 model to estimate daily visibility for each grid.

2.2.3. Stage 3

For grid cells missing both AOD and visibility measurements, a generalized additive mixed model (GAMM) was used to predict daily visibility based on the output from Stage 2 model (Kloog et al., 2014). This method was previously used in (Kloog et al., 2012). The model contains a smooth function of longitude and latitude, and a random intercept for each grid cell. A separate spatial surface was fit for each two-month period of each year to capture the temporal variations (Kloog et al., 2014). In addition, we included the mean daily visibility and a random cell-specific slope. The semi parametric regression model is shown as Eq. (1):

PredVisij=(α+μi)+(β+vi)MVisj+Smooth(X,Y)k(j)+ϵi,j(ujvj)~((00),Ωβ) (1)

where PredVisij is the predicted visibility at grid-cell i on a day j from the stage 2 model; MVisj is the mean value of predicted visibility from stage1 on a day j across the study region; α and β are the fixed intercept and fixed slope, respectively;μi and vi are the random intercept and grid-cell specific random slope, respectively. X is longitude, and Y is latitude; “Smooth” is a thin plate spline function; and k(j) is the two-month period in which day j falls. To estimate the goodness of stage 3 model, we use the same cross-validation validation method as in stage 1.

2.2.4. Stage 4

First, we fit a linear mixed effects model to describe the relationship between PM2.5 and visibility Because of the paucity of PM2.5 data in this region, we used data during 2017–2018 to train the model and used data during 2004–2005 for validation. Inverse of visibility and squared form of relative humidity were used as variables in the model based on the comparison of models using different combinations of variables (Text S1 and Table S5). The model can be expressed as:

PM2.5i,j=(μ+μm′)+(β1+β1m′)(1VISi,j)+β2RHi,j+β3RHi,j2+εi,j(μm′βm′)~N((0,0),Ψ) (2)

where PM2.5i,j is the measurement daily PM2.5 concentrations for a PM2.5 site i on a day j; VIS is visibility; RH is relative humidity; μ is the fixed intercept; μ′m is the month -specific random intercept, that is, the regression intercept in month m; β1 is the fixed slope for the inverse of visibility; β1′m is the random slope for the inverse of visibility range in month j; β2 and β3 are the slopes for RH and squared form of relative humidity, respectively; εi,j is the error term at site i on day j. Ψ is the unstructured variance covariance matrix for the random effects.

Finally, we predicted PM2.5 concentration using Eq. (2) based on daily visibility for each 1 km × 1 km grid cell during 2001–2018 in Kuwait and Iraq. We used the results from the daily prediction model to calculate monthly and yearly PM2.5 averages.

All programming was implemented in R software version 3.5.0.

3. Results

The spatial distribution of yearly average MAIAC AOD downloaded from NASA for the study region is shown in Supplemental Fig. S4. The CV R2 between fitted and predicted daily visibility was 0.71 (Table S4). Fig. 2 illustrates the variable importance for the top 30 predictors in the stage 1 random forest model. MAIAC AOD, year, dust column mass density, dust extinction AOD, and dust surface mass concentration are the five most important predictors.

Fig. 2.

Fig. 2.

Open in a new tab

Top 30 most importance variables for the random forest model predicting daily visibility (x-axis: variable importance relative to the importance of AOD, calculated by function “h2o.varimp_plot” of R Package ‘h2o’).

The stage 3 model performed well with a cross validation R2 of 0.68. A statistically significant association was found between visibility and daily PM2.5 in stage 4 model. This model resulted in a high CV R2 value of 0.70 in the years (2017–2018) and R2 value of 0.74 in 2004–2005, indicating good model fit (Table S5). The predicted and measured PM2.5 concentrations for stage 4 model are shown in Fig. S5.

We found considerable spatial and temporal variations over the study region. Fig. 3 shows the spatial distribution of the predicted PM2.5 concentrations in the study area, averaged over the full study period. The predictions for each year are shown in Fig. 4 and mean values for Iraq and Kuwait were 37 and 41 μg/m3, respectively. Annual and monthly median PM2.5 predictions for Iraq and Kuwait are shown in Fig. 5. Monthly average PM2.5 concentrations for each year are shown in Fig. S6.

Fig. 3.

Fig. 3.

Open in a new tab

Mean PM2.5 concentrations (μg/m3) in each 1 km × 1 km grid during the entire modeling period predicted by our models (black line: boundary; blue line: water body).

Fig. 4a.

Fig. 4a.

Open in a new tab

Spatial distribution of predicted mean PM2.5 concentrations (μg/m3) for each year (2001–2009; Black line: boundary; Blue line: water body).

Fig. 5.

Fig. 5.

Open in a new tab

Annual and monthly average PM2.5 concentrations in Kuwait and Iraq.

In general, the PM2.5 concentrations in Kuwait were slightly higher than Iraq during 2001–2018. The annual PM2.5 concentration for Kuwait ranged from 36 to 49 μg/m3, and the values for Iraq ranged from 33 to 44 μg/m3. As expected, the predicted PM2.5 concentrations were higher in urban areas compared to rural areas. Some big cities such as Bagdad, Karbala, Najaf, and Diwaniya have annual average PM2.5 concentrations above 45 μg/m3. In contrast, Al-Anbar, the largest governorate west of Iraq (138,501 km2) which is mostly desert had low average PM2.5 concentrations, except for large urban areas in the east (Al-Fallujah and Al-Ramadi). PM2.5 levels in the study region also showed distinct temporal variability. The highest annual PM2.5 concentration for both Kuwait and Iraq were observed in 2008, followed by 2009. The lowest values for both Iraq and Kuwait were observed in 2014. It is possible that the extreme drought in 2008–2009 contributed to the high PM2.5 level in 2008 and heavy rain in the end of 2013 contributed to lower levels in 2014. By regions, the higher PM2.5 levels in 2008 and 2009 were predominantly in the urban areas. The monthly average predictions for Kuwait ranged from 32 to 49 μg/m3, and for Iraq ranged from 29 to 44 μg/m3. July had the highest predicted concentrations, and November had the lowest values. This is consistent with seasonal patterns of PM2.5 observations and indicates that PM2.5 pollution in the region has been a long-term problem.

As demonstrated by the considerable spatial and temporal variability over the region, deployers in Iraq and Kuwait experienced variability in PM2.5 exposures. Some bases were located in regions with lower PM2.5 concentrations such as in the Syrian Desert, which have average PM2.5 concentrations below 30 μg/m. However, many U.S. bases located in urban area and had very high PM2.5 concentrations. The weekly average PM2.5 at Baghdad International Airport, the US base in Balad, and Kuwait City International Airport are shown in Fig. S7. The weekly PM2.5 concentrations ranged to more than 150 μg/m3 for the bases in Baghdad and Balad in summer of 2008 and 2009.

4. Discussion

We developed a model to predict daily PM2.5 exposures at 1 km × 1 km resolution for Iraq and Kuwait, where the monitoring system is very limited. Our results indicated that this region had very high PM2.5 levels, which varied by location and year.

The annual mean predicted PM2.5 concentrations for Kuwait (41 μg/m3) and Iraq (37 μg/m3) from 2001 to 2018 were considerably higher than the U.S. National Air Quality Standard of 12 μg/m3 and the WHO guideline of 10 μg/m3. Very few studies conducted PM2.5 measurements in Kuwait or Iraq. Engelbrecht et al. (2009) collected PM2.5 samples approximately four times per month at five sampling sites in Iraq and four sampling sites in Kuwait in 2006. Al-Hemoud et al. (2018a) assessed daily air concentration of PM10 for the year 2012 in a suburban area of Kuwait. Al-Hemoud et al. (2018b) and Al-Hemoud et al. (2019) collected PM2.5 measurements in three monitoring stations in Kuwait from 2014 to 2017. Our estimates have the same spatial distribution with the PM2.5 samples of Engelbrecht et al. (2009). Our predictions and the PM10 in Al-Hemoud et al. (2018a) have the same seasonal pattern. The PM2.5 concentrations predicted in this study are consistent with a previous WHO air pollution database (WHO, 2018). However, the PM2.5 concentrations in (Al-Hemoud et al., 2019) (34–80 μg/m3) are generally higher than our results, possibly because they selected hourly PM2.5 values above a minimum background concentration of 8.83 μg/m3 to report.

Our study has some major findings that are of public health significance. In previous studies, the AOD-PM2.5 relationship was directly used to examine the spatial distribution of PM2.5. However, for many countries, there are no available historical PM2.5 measurements. Our novel exposure assessment approach indicates it possible to assess air pollution exposures in countries without extensive PM2.5 monitoring locations, and without extensive PM2.5 historical data. And the other predictors used in our model are predominantly from global public datasets, thus our method could be applied in other countries. Secondly, the random forest model approach provided the relative importance of each factor associated with historical PM2.5 concentrations to assess exposures in this region or similar areas. For example, in addition to AOD, we found that surface dust related variables are also associated with PM2.5 in arid environments (Li et al., 2020). The seasonal and monthly trends of PM2.5 are consistent with the seasonal occurrence of dust storms in the region, demonstrating the impact of dust storms on the PM2.5 model predictions. (Al-Hemoud et al., 2017; Al-Hemoud et al., 2018a; Li et al., 2020).

A limitation of our approach is that the stage 4 model regression results was based on PM2.5 measurement data collected at three sampling sites for two years because of the paucity of PM2.5 data in this region. We used the PM2.5 data collected 2017–2018 to train the model and used data collected 2004–2005 for validation, which could have potentially influenced the accuracy of historical predictions. However, the relationship between PM2.5 and visibility is relatively stable under specific conditions of relative humidity as in Iraq and Kuwait (Zhao et al., 2017; Liu et al., 2017). It has also been shown that PM2.5 particles in Iraq and Kuwait exhibit similar hygroscopic properties (Abdeen et al., 2014). In support of our approach is that the cross validation of the 2004–2005 PM2.5 data indicated good model fit. Another limitation is that the AOD data used in this study has nonrandom missing data (such as days with heavy cloud clover) that was accounted for by modeling, and the local crossing time of the AOD data were relatively fixed between 10:30 am and 1:30 pm. Although these factors could result in systematic bias of the PM2.5 predictions, the model would still be internally consistent.

To the best of our knowledge, this is the first model to estimate historical PM2.5 concentrations at a high resolution in Iraq and Kuwait that takes advantage of all available measurement data. This high-resolution model makes it possible to assess the health effects of PM2.5 exposures both in the native populations and the military personnel. This study is also helpful to the construction of regional environmental policies in areas with different pollution levels.

Supplementary Material

supplement

Fig. 4b.

Fig. 4b.

Open in a new tab

Spatial distribution of predicted mean PM2.5 concentrations (μg/m3) for each year (2010–2018; Black line: boundary; Blue line: water body).

Acknowledgments

This work was supported by the VA Cooperative Studies Program #595: Pulmonary Health and Deployment to Southwest Asia and Afghanistan, from the United States (U.S.) Department of Veterans Affairs, Office of Research and Development, Clinical Science Research and Development, Cooperative Studies Program. This publication was also made possible by U.S. Environmental Protection Agency (EPA) grant RD-835872, NASA grant 80NSSC19K0225, and NIH/NIEHS grant P30 ES000002. The contents do not represent the views of the U.S. Department of Veterans Affairs, U.S. EPA or the U.S. Government. The computations in this paper were run on the Odyssey cluster of Harvard University.

Footnotes

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.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.envint.2021.106445.

References

  1. Abbey DE, Ostro BE, Fraser G, Vancuren T, Burchette RJ, 1995. Estimating fine particulates less-than 2.5 microns in aerodynamic diameter (PM2.5) from airport visibility data in California. J. Expo. Anal. Env. Epid 5, 161–180. [PubMed] [Google Scholar]
  2. Abdeen Z, Qasrawi R, Heo J, Wu B, Shpund J, Vanger A, Sharf G, Moise T, Brenner S, Nassar K, Saleh R, Al-Mahasneh QM, Sarnat JA, Schauer JJ, 2014. Spatial and temporal variation in fine particulate matter mass and chemical composition: the Middle East Consortium for Aerosol Research Study. Scientific World J. 2014, 878704. 10.1155/2014/878704. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Abraham JH, DeBakey SF, Reid L, Zhou J, Baird CP, 2012. Does Deployment to Iraq and Afghanistan Affect Respiratory Health of US Military Personnel. J. Occup. Environ. Med 54, 740–745. 10.1097/JOM.0b013e318252969a. [DOI] [PubMed] [Google Scholar]
  4. Abraham JH, Eick-Cost A, Clark LL, Hu Z, Baird CP, DeFraites R, Tobler SK, Richards EE, Sharkey JM, Lipnick RJ, Ludwig SL, 2014. A retrospective cohort study of military deployment and postdeployment medical encounters for respiratory conditions. Mil. Med. 179, 540–546. 10.7205/MILMED-D-13-00443. [DOI] [PubMed] [Google Scholar]
  5. Al-Hemoud A, Al-Sudairawi M, Neelamanai S, Naseeb A, Behbehani W, 2017. Socioeconomic effect of dust storms in Kuwait. Arabian J. Geosci. 10 10.1007/s12517-016-2816-9. [DOI] [Google Scholar]
  6. Al-Hemoud A, Al-Dousari A, Al-Shatti A, Al-Khayat A, Behbehani W, Malak M, 2018a. Health Impact Assessment Associated with Exposure to PM10 and Dust Storms in Kuwait. Atmosphere 9. 10.3390/atmos9010006. [DOI] [Google Scholar]
  7. Al-Hemoud A, Gasana J, Al-Dabbous AN, Al-Shatti A, Al-Khayat A, 2018b. Disability Adjusted Life Years (DALYs) in Terms of Years of Life Lost (YLL) Due to Premature Adult Mortalities and Postneonatal Infant Mortalities Attributed to PM2.5 and PM10 Exposures in Kuwait. Int. J. Environ. Res. Public Health 15, 15, 2018b. 10.3390/ijerph15112609. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Al-Hemoud A, Gasana J, Al-Dabbous A, Alajeel A, Al-Shatti A, Behbehani W, Malak M, 2019. Exposure levels of air pollution (PM2.5) and associated health risk in Kuwait. Environ. Res 179, 108730. 10.1016/j.envres.2019.108730. [DOI] [PubMed] [Google Scholar]
  9. Alolayan MA, Brown KW, Evans JS, Bouhamra WS, Koutrakis P, 2013. Source apportionment of fine particles in Kuwait City. Sci. Total. Environ 448, 14–25. 10.1016/j.scitotenv.2012.11.090. [DOI] [PubMed] [Google Scholar]
  10. Beloconi A, Kamarianakis Y, Chrysoulakis N, 2016. Estimating urban PM10 and PM2.5 concentrations, based on synergistic MERIS/AATSR aerosol observations, land cover and morphology data. Remote Sens. Environ. 172, 148–164. 10.1016/j.rse.2015.10.017. [DOI] [Google Scholar]
  11. Beloconi A, Chrysoulakis N, Lyapustin A, Utzinger J, Vounatsou P, 2018. Bayesian geostatistical modelling of PM10 and PM2.5 surface level concentrations in Europe using high-resolution satellite-derived products. Environ. Int 121, 57–70. 10.1016/j.envint.2018.08.041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Brauer M, Freedman G, Frostad J, van Donkelaar A, Martin RV, Dentener F, van Dingenen R, Estep K, Amini H, Apte JS, Balakrishnan K, Barregard L, Broday D, Feigin V, Ghosh S, Hopke PK, Knibbs LD, Kokubo Y, Liu Y, Ma S, Morawska L, Sangrador JL, Shaddick G, Anderson HR, Vos T, Forouzanfar MH, Burnett RT, Cohen A, 2016. Ambient Air Pollution Exposure Estimation for the Global Burden of Disease 2013. Environ. Sci. Technol 50, 79–88. 10.1021/acs.est.5b03709. [DOI] [PubMed] [Google Scholar]
  13. Brokamp C, Jandarov R, Rao MB, LeMasters G, Ryan P, 2017. Exposure assessment models for elemental components of particulate matter in an urban environment: A comparison of regression and random forest approaches. Atmos Environ 1994 (151), 1–11. 10.1016/j.atmosenv.2016.11.066. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Brown KW, Bouhamra W, Lamoureux DP, Evans JS, Koutrakis P, 2008. Characterization of particulate matter for three sites in Kuwait. J Air Waste Manage 58 (8), 994–1003. [DOI] [PubMed] [Google Scholar]
  15. Chudnovsky AA, Lee HJ, Kostinski A, Kotlov T, Koutrakis P, 2012. Prediction of daily fine particulate matter concentrations using aerosol optical depth retrievals from the Geostationary Operational Environmental Satellite (GOES). J. Air. Waste. Manage 62, 1022–1031. 10.1080/10962247.2012.695321. [DOI] [PubMed] [Google Scholar]
  16. Colarco P, da Silva A, Chin M, Diehl T, 2010. Online simulations of global aerosol distributions in the NASA GEOS-4 model and comparisons to satellite and ground-based aerosol optical depth. J. Geophys. Res 115 10.1029/2009jd012820. [DOI] [Google Scholar]
  17. Di Q, Kloog I, Koutrakis P, Lyapustin A, Wang Y, Schwartz J, 2016. Assessing PM2.5 Exposures with High Spatiotemporal Resolution across the Continental United States. Environ. Sci. Technol 50, 4712–4721. 10.1021/acs.est.5b06121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Di Q, Amini H, Shi L, Kloog I, Silvern R, Kelly J, Sabath MB, Choirat C, Koutrakis P, Lyapustin A, Wang Y, Mickley LJ, Schwartz J, 2019. An ensemble-based model of PM2.5 concentration across the contiguous United States with high spatiotemporal resolution. Environ. Int 130, 104909. 10.1016/j.envint.2019.104909. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Engelbrecht JP, McDonald EV, Gillies JA, Jayanty RKM, Casuccio G, Gertler AW, 2009. Characterizing Mineral Dusts and Other Aerosols from the Middle East—Part 1: Ambient Sampling. Inhal. Toxicol 21, 297–326. 10.1080/08958370802464273. [DOI] [PubMed] [Google Scholar]
  20. Garshick E, Abraham JH, Baird CP, Ciminera P, Downey GP, Falvo MJ, Hart JE, Jackson DA, Jerrett M, Kuschner W, Helmer DA, Jones KD, Krefft SD, Mallon T, Miller RF, Morris MJ, Proctor SP, Redlich CA, Rose CS, Rull RP, Saers J, Schneiderman AI, Smith NL, Yiallouros P, Blanc PD, 2019. Respiratory Health after Military Service in Southwest Asia and Afghanistan. An Official American Thoracic Society Workshop Report. Ann. Am. Thorac. Soc 16, e1–e16. 10.1513/AnnalsATS.201904-344WS. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Ginoux P, Chin M, Tegen I, Prospero JM, Holben B, Dubovik O, Lin S-J, 2001. Sources and distributions of dust aerosols simulated with the GOCART model. J. Geophys. Res.-Atmos 106, 20255–20273. 10.1029/2000jd000053. [DOI] [Google Scholar]
  22. Hersbach H, Bell B, Berrisford P, Hirahara S, Horányi A, Muñoz-Sabater J, Nicolas J, Peubey C, Radu R, Schepers D, Simmons A, Soci C, Abdalla S, Abellan X, Balsamo G, Bechtold P, Biavati G, Bidlot J, Bonavita M, Chiara G, Dahlgren P, Dee D, Diamantakis M, Dragani R, Flemming J, Forbes R, Fuentes M, Geer A, Haimberger L, Healy S, Hogan RJ, Hólm E, Janisková M, Keeley S, Laloyaux P, Lopez P, Lupu C, Radnoti G, Rosnay P, Rozum I, Vamborg F, Villaume S, Thépaut J, 2020. The ERA5 global reanalysis. Q. J. Roy. Meteor. Soc. 146, 1999–2049. 10.1002/qj.3803. [DOI] [Google Scholar]
  23. Hu X, Belle JH, Meng X, Wildani A, Waller LA, Strickland MJ, Liu Y, 2017. Estimating PM2.5 Concentrations in the Conterminous United States Using the Random Forest Approach. Environ. Sci. Technol 51, 6936–6944. 10.1021/acs.est.7b01210. [DOI] [PubMed] [Google Scholar]
  24. Huang K, Xiao Q, Meng X, Geng G, Wang Y, Lyapustin A, Gu D, Liu Y, 2018. Predicting monthly high-resolution PM2.5 concentrations with random forest model in the North China Plain. Environ. Pollut 242, 675–683. 10.1016/j.envpol.2018.07.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kloog I, Nordio F, Coull BA, Schwartz J, 2012. Incorporating local land use regression and satellite aerosol optical depth in a hybrid model of spatiotemporal PM2.5 exposures in the Mid-Atlantic states. Environ. Sci. Technol 46, 11913–11921. 10.1021/es302673e. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kloog I, Chudnovsky AA, Just AC, Nordio F, Koutrakis P, Coull BA, Lyapustin A, Wang Y, Schwartz J, 2014. A New Hybrid spatio-temporal model for estimating daily multi-year PM2.5 concentrations across Northeastern USA using high resolution aerosol optical depth data. Atmos. Environ 1994 (95), 581–590. 10.1016/j.atmosenv.2014.07.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Li J, Garshick E, Al-Hemoud A, Huang S, Koutrakis P, 2020. Impacts of meteorology and vegetation on surface dust concentrations in Middle Eastern countries. Sci. Total. Environ 712 10.1016/j.scitotenv.2020.136597. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Lin CQ, Liu G, Lau AKH, Li Y, Li CC, Fung JCH, Lao XQ, 2018. High-resolution satellite remote sensing of provincial PM2.5 trends in China from 2001 to 2015. Atmos. Environ 180, 110–116. 10.1016/j.atmosenv.2018.02.045. [DOI] [Google Scholar]
  29. Liu M, Bi J, Ma Z, 2017. Visibility-Based PM2.5 Concentrations in China: 1957–1964 and 1973–2014. Environ. Sci. Technol 51, 13161–13169. 10.1021/acs.est.7b03468. [DOI] [PubMed] [Google Scholar]
  30. Liu Y, Sarnat JA, Kilaru A, Jacob DJ, Koutrakis P, 2005. Estimating ground-level PM2.5 in the eastern united states using satellite remote sensing. Environ. Sci. Technol 39, 3269–3278. 10.1021/es049352rn. [DOI] [PubMed] [Google Scholar]
  31. Lyapustin A, Wang YJ, Korkin S, Huang D, 2018. MODIS Collection 6 MAIAC algorithm. Atmos. Meas. Tech 11, 5741–5765. 10.5194/amt-11-5741-2018. [DOI] [Google Scholar]
  32. Masri S, Garshick E, Hart J, Bouhamra W, Koutrakis P, 2017. Use of visual range measurements to predict fine particulate matter exposures in Southwest Asia and Afghanistan. J. Air Waste Manag. Assoc. 67, 75–85. 10.1080/10962247.2016.1243169. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Paciorek CJ, Liu Y, Moreno-Macias H, Kondragunta S, 2008. Spatiotemporal associations between GOES aerosol optical depth retrievals and ground-level PM2.5. Environ. Sci. Technol. 42, 5800–5806. 10.1021/es703181j. [DOI] [PubMed] [Google Scholar]
  34. Paciorek CJ, Liu Y, 2009. Limitations of remotely sensed aerosol as a spatial proxy for fine particulate matter. Environ. Health Perspect. 117, 904–909. 10.1289/ehp.0800360. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Pappin AJ, Hakami A, 2013. Source attribution of health benefits from air pollution abatement in Canada and the United States: an adjoint sensitivity analysis. Environ Health Perspect. 121, 572–579. 10.1289/ehp.1205561. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Pugh MJ, Jaramillo CA, Leung KW, Faverio P, Fleming N, Mortensen E, Amuan ME, Wang CP, Eapen B, Restrepo M, Morris MJ, 2016. Increasing Prevalence of Chronic Lung Disease in Veterans of the Wars in Iraq and Afghanistan. Mil. Med 181, 476–481. 10.7205/MILMED-D-15-00035. [DOI] [PubMed] [Google Scholar]
  37. Rienecker MM, Suarez MJ, Gelaro R, Todling R, Bacmeister J, Liu E, Bosilovich MG, Schubert SD, Takacs L, Kim G-K, Bloom S, Chen J, Collins D, Conaty A, da Silva A, Gu W, Joiner J, Koster RD, Lucchesi R, Molod A, Owens T, Pawson S, Pegion P, Redder CR, Reichle R, Robertson FR, Ruddick AG, Sienkiewicz M, Woollen J, 2011. MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate 24, 3624–3648. 10.1175/jcli-d-11-00015.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Salomonson VV, Barnes WL, Maymon PW, Montgomery HE, Ostrow H, 1989. MODIS - advanced facility instrument for studies of the earth as a system. Ieee. T. Geosci. Remote 27, 145–153. 10.1109/36.20292. [DOI] [Google Scholar]
  39. Sharkey JM, Abraham JH, Clark LL, Rohrbeck P, Ludwig SL, Hu Z, Baird CP, 2016. Postdeployment Respiratory Health Care Encounters Following Deployment to Kabul, Afghanistan: A Retrospective Cohort Study. Mil. Med 181, 265–271. 10.7205/MILMED-D-14-00690. [DOI] [PubMed] [Google Scholar]
  40. Stafoggia M, Schwartz J, Badaloni C, Bellander T, Alessandrini E, Cattani G, De’ Donato F, Gaeta A, Leone G, Lyapustin A, Sorek-Hamer M, de Hoogh K, Di Q, Forastiere F, Kloog I, 2017. Estimation of daily PM10 concentrations in Italy (2006–2012) using finely resolved satellite data, land use variables and meteorology. Environ. Int 99, 234–244. 10.1016/j.envint.2016.11.024. [DOI] [PubMed] [Google Scholar]
  41. U.S. Institute of Medicine, 2011. Long-term Health Consequences of Exposure to Burn Pits in Iraq and Afghanistan. National Academies Press. [DOI] [PubMed] [Google Scholar]
  42. Vajanapoom N, Shy CM, Neas LM, Loomis D, 2001. Estimation of particulate matter from visibility in Bangkok, Thailand. J. Expo. Anal. Env. Epid 11, 97–102. 10.1038/sj.jea.7500148. [DOI] [PubMed] [Google Scholar]
  43. van Donkelaar A, Martin RV, Brauer M, Boys BL, 2015. Use of satellite observations for long-term exposure assessment of global concentrations of fine particulate matter. Environ. Health Perspect. 123, 135–143. 10.1289/ehp.1408646. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Wang HM, Zheng JH, Yang T, He ZC, Zhang P, Liu XF, Zhang MX, Sun LH, Yu XF, Zhao J, Liu XY, Xu BP, Tong LP, Xiong JY, 2020. Predicting the emission characteristics of VOCs in a simulated vehicle cabin environment based on small-scale chamber tests: Parameter determination and validation. Environ. Int 142, 10. 10.1016/j.envint.2020.105817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. WHO: Ambient (outdoor) air pollution in cities database, 2018.
  46. Xiao Q, Wang Y, Chang HH, Meng X, Geng G, Lyapustin A, Liu Y, 2017. Full-coverage high-resolution daily PM 2.5 estimation using MAIAC AOD in the Yangtze River Delta of China. Remote Sens. Environ. 199, 437–446. 10.1016/j.rse.2017.07.023. [DOI] [Google Scholar]
  47. Xue T, Zheng Y, Tong D, Zheng B, Li X, Zhu T, Zhang Q, 2019. Spatiotemporal continuous estimates of PM2.5 concentrations in China, 2000–2016: A machine learning method with inputs from satellites, chemical transport model, and ground observations. Environ. Int 123, 345–357. 10.1016/j.envint.2018.11.075. [DOI] [PubMed] [Google Scholar]
  48. Zhang M, Xiong J, Liu Y, Misztal PK, Goldstein AH, 2021. Physical-chemical coupling model for characterizing the reaction of ozone with squalene in realistic indoor environments. Environ. Sci. Technol [DOI] [PubMed] [Google Scholar]
  49. Zhao T, Yang L, Yan W, Zhang J, Lu W, Yang Y, Chen J, Wang W, 2017. Chemical characteristics of PM1/PM2.5 and influence on visual range at the summit of Mount Tai, North China. Sci. Total. Environ 575, 458–466. 10.1016/j.scitotenv.2016.09.173. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supplement
NIHMS1678488-supplement-supplement.docx (3.5MB, docx)

RESOURCES