Water, Air, & Soil Pollution

, 224:1508

Reductions of PM2.5 Air Concentrations and Possible Effects on Premature Mortality in Japan


    • Environmental Research Center (ERC)Sohar University

DOI: 10.1007/s11270-013-1508-2

Cite this article as:
Nawahda, A. Water Air Soil Pollut (2013) 224: 1508. doi:10.1007/s11270-013-1508-2


The current study estimates premature mortality caused by long-term exposure to elevated concentrations of PM2.5 (particulate matter with aerodynamic diameter equal to or less than 2.5 μm) in Japan from 2006 to 2009. The premature mortality is calculated based on a relative risk of 1.04 (95 % CI, 1.01–1.08) per 10 μg m−3 increase above the annual mean limit of 10 μg m−3 taken from the World Health Organization Air Quality Guidelines. The spatiotemporal variations of PM2.5 are estimated based on the measurements of suspended particulate matter (SPM) (with aerodynamic diameter approximately less than 7.0 μm) at 1,843 monitors. The improvements of air quality in Japan by reducing the emissions of SPM from 2006 to 2009 could save 3,602 lives based on a reduction target of 10 μg m−3 annual mean concentration. This finding could be a tangible benefit gained by reducing the emissions of particulate matter in Japan.


Premature mortalitySPMExposurePM2.5Japan

1 Introduction

In recent years, the enforcement of exhaust emission regulations in Japan has succeeded in reducing the emissions of suspended particulate matter (SPM). According to the annual reports of the Ministry of the Environment in Japan (MOEJ), during the period from 1974 to 2009, the mean annual concentration of SPM decreased from 0.058 to 0.021 mg m−3 in stationary monitors and from 0.162 to 0.024 mg m−3 in mobile monitors (roadside monitors). In 2009, the air quality standards of SPM, 0.10 mg m−3 daily average and 0.20 mg m−3 hourly values, were met at more than 98.8 % of the 1,792 monitors located throughout Japan (MOEJ 2009). Therefore, quantifying the health impacts caused by these improvements using the available air quality monitoring data is crucial and produces more relevant benefits to policy decision makers compared to the effects estimated by a Chemical Transport Model (CTM). This is because of high uncertainty and limited validation especially for secondary air pollutants (Nawahda et al. 2012); Wang et al. 2006; Liu et al. 2009; Saikawa et al. 2009).

In Japan, there are few epidemiological studies that clearly demonstrate effects of SPM on human health among different age groups, e.g., Yamazaki et al. (2011), Kagawa (1994), Shima et al. (2003), Iwai et al. (2005), Yorifuji et al. (2008), Katanoda et al. (2011), Ueda et al. (2012), and Omori et al. (2003). Also, there are limited studies on valuating the effect of elevated concentrations of SPM on human health such as Nakatani et al. (2007). Most of these studies were limited in space and time and could not identify statistically representative health effects and relative risk values of premature mortality among different age groups. Western studies have identified strong associations between elevated concentrations of PM2.5 and mortality, lung cancer, and reduced lung function (Pope III et al. 2002; Pope III and Dockery 2006; Lippmann 2009; Jerrett et al. 2009). However, the spatial variability and the fact that there are no thresholds for mortality and morbidity caused by exposure to PM2.5 complicate the process of establishing clear standards and exposure guidelines that can be realized from the World Health Organization (WHO) guidelines, the U.S. Environmental Protection Agency (USEPA), and Japanese air quality standards. These levels of PM2.5 are 25/10 (WHO 2005) and 35/15 (USEPA 2010c; MOEJ 2009) [daily mean (in micrograms per cubic meter) / annual mean limit (in micrograms per cubic meter)]. The USEPA 35/15 standards were discussed during the 2010 public meeting of the USEPA Clean Air Scientific Advisory Committee. Also, there were requests to assess the environmental risks of PM2.5 below 15 μg m−3 (USEPA 2010a). In Japan, due to lack of comprehensive epidemiological studies on the effects of PM2.5 on premature mortality for all age groups, we use the adjusted mortality relative risk (RR) (1.04) (95 % confidence interval (CI), 1.01–1.08) associated with a 10 μg m−3 change in PM2.5 mean annual concentration from Pope III et al. (2002) to estimate the premature mortality among the age groups +30 years. Also, we assume achievement of the WHO Air Quality Guideline for PM2.5 (10 μg m−3 annual mean), which is the lowest level at which total, cardiopulmonary, and lung cancer mortality have been shown to increase with more than 95 % confidence in response to long-term exposure to PM2.5 (WHO 2005). The findings of this study are a step toward a better understanding of the health benefits of reducing the emissions of PM2.5 in Japan. Also, this study estimates premature mortality among different age groups.

2 Methodology

2.1 Monitoring Data

Our analyses are dependent on the reliability and high quality of the monitoring concentrations of SPM. The mean annual concentrations of SPM at each monitor from 2006 to 2009 (Figs. 1 and 2) are obtained from the National Institute of Environmental studies. These monitors are managed and operated by the MOEJ and local governments. However, only few air quality monitors have both measurements of SPM and PM2.5; at these monitors, the ratio of PM2.5 to SPM varied from 0.6 to 0.8 and could be affected by sampling method (filter collection, Anderson's sampler, light-scattered estimation, and β-ray absorption) (Iwai et al. 2005). Accordingly, a conversion factor of 0.7 is assumed.
Fig. 1

Mean annual concentration of SPM in Japan from 2006 to 2009

Fig. 2

Distributed mean annual concentration of SPM in 2006

2.2 Exposure and Premature Mortality Analysis

The distributed annual mortality is calculated as follows using Eq. (1) for elevated PM2.5 mean annual concentrations:
$$ \mathrm{Mortalit}{{\mathrm{y}}_{{\mathrm{P}{{\mathrm{M}}_{2.5 }}}}}\left( {i,j,t} \right)=\mathrm{pop}\left( {i,j,t} \right){M_{\mathrm{b}}}\left( {i,j,t} \right)\ {\beta_{{\mathrm{P}{{\mathrm{M}}_{2.5 }}}}}\vartriangle \mathrm{P}{{\mathrm{M}}_{2.5 }}\left( {i,j,t} \right) $$
$$ \beta ={{{\ln \left( {\mathrm{RR}} \right)}} \left/ {{\vartriangle {C_{{\mathrm{P}{{\mathrm{M}}_{2.5 }}}}}}} \right.} $$
where i and j specify the location, t is the year of simulation, pop is the exposed population, △PM2.5(i,j,t) is the change in the annual mean concentrations, Mb is the annual mean baseline mortality rate, and β is the PM2.5 concentration–response (CR) coefficient, which can be calculated using Eq. (2). Equation (1) is an approximation of the nonlinear CR function of the BenMAP model (USEPA 2010b) as follows:
$$ 1- \exp \left( {-\mathrm{RR}\vartriangle C} \right)\approx \mathrm{RR}\vartriangle C. $$

Additionally, many recent publications show similar formulations, e.g., Liu et al. (2009), Wang et al. (2006), and Saikawa et al. (2009).

According to Pope III et al. (2002), an increase of 10 μg m−3 annual average of PM2.5, within a range from around 7.0 to 30 μg m−3, caused a 4 % (95 % CI, 1.01–1.08) increase in mortality rate for the age group of 30 years and above. This gives β a value around 0.004. We use the same β value also for mean annual concentrations above 30 μg m−3 similar to Cohen et al. (2005); they linearly extrapolated the PM2.5 CR function to cover a wider range from 0 to 90 μg m−3. Additionally, we extend the analysis below 10 μg m−3, mainly above 7.0 μg m−3, and not less than this value because the published CR relationships from epidemiological studies such as Pope III et al. (2006) on the health effects of PM2.5 have low limits around 7.0 μg m−3.

2.3 Population Distribution

Population distribution in Japan in 2005 is derived from the Gridded Population of the World version 3 (GPWv3) prepared by the Center for International Earth Science Information Network (CIESIN), Columbia University. Also, we assume constant percentages of the four main age groups, g1 (0–14 years), g2 (15–64 years), g3 (65–74 years), and g4 (+75 years), and constant baseline mortality from 2005 to 2009. According to the official statistics of Japan (e-stat 2011), the fractions of population in 2005 and the corresponding baseline mortality for each group based on mortality statistics are shown in Table 1. We assume that the population will not change during the study period. Also, g1 and g2 groups are neglected since Eq. (1) is applicable for age group +30 years, and their baseline mortality is small.
Table 1

Summary of baseline mortality parameters in Japan in 2005 (Japan Statistics Bureau, 1996–2008)

Age group


Population (×1,000)

Age group (%)

Baseline mortality (per 1,000)


























We use ArcGIS system to evaluate Eq. (1) based on population distribution map and to generate interpolation surfaces of SPM using kriging method. This interpolation surface estimates SPM in places that have no air quality monitors. Validating the interpolated SPM, which has to be done in case of limited number of monitors, could be of less importance in this study, and even if it is done for few places and it shows strong or weak associations, this does not necessarily prove the weakness or robustness of the analysis and findings because of the spatiotemporal variability of hydro-climatic parameters and primary components of the monitored SPM.

3 Results and Discussion

The estimated premature mortality cases caused by exposure to elevated concentrations of PM2.5, higher than 7.0 and 10 μg m−3, in Japan from 2006 to 2009 for each age group are shown in Table 2. The distributed estimated total premature mortality cases for the age groups +65 years are shown in Fig. 3. The estimated premature mortality based on 7.0 μg m−3 (40,000 cases) is about 1.4 times the estimations based on 10 μg m−3 (28,400 cases) for the period from 2006 to 2009. About 77 % of these cases are among the age groups of +75 years. The improvements of air quality in Japan by reducing the emissions of SPM from 2006 to 2009 could save 3,568 and 3,602 lives based on 7.0 and 10 μg m−3analysis, respectively. The benefits, 3,568 and 3,602 saved lives, are estimated by adding the deaths in 2006 minus the deaths in 2007, the deaths in 2007 minus deaths in 2008, and the deaths in 2008 minus deaths in 2009. These estimations are based on RR value of 1.04. If we assumed the maximum RR value reported by Jerrett et al. (2009), which is 1.08, the estimated number of premature mortality cases would double.
Table 2

Total mortality cases caused by exposure to PM2.5 in Japan from 2006 to 2009


PM2.5 > 7.0 (μg m−3)

PM2.5 > 10 (μg m−3)

65–74 years

+75 years


95 % CI

65–74 years

+75 years


95 % CI



















































Fig. 3

Distributed premature mortality cases for the age groups +65 years in Japan from 2006 to 2009

During the past 10 years, there have been numerous studies on estimating the effects of PM on human health in Asia; however, the estimated effects were based on simulated concentrations by CTMs coupled with emission inventories. Validation of these concentrations is not easy and most of these studies avoided discussing the validation of the simulated concentrations, which is one of the main sources of uncertainty, e.g., Wang et al. (2006), Saikawa et al. (2009), Anenberg et al. (2010), and Liu et al. (2009). Limited validation of the simulations of CTMs could be understood in regions where there are no air quality monitors. In Japan, we believe that the extensive air quality monitoring network provides a better characterization of the spatiotemporal concentrations of PM2.5 compared to CTMs. However, our estimates of the premature mortality caused by exposure to PM2.5 are still conservative. This is mainly caused by the common high uncertainties in health risk assessments, which were discussed in details in many studies such as Saikawa et al. (2009) who discussed the uncertainty in the CR relationships caused by different background pollution levels in the USA and other countries. Additionally, we raise the following uncertainties:

3.1 Interpolated SPM

There is uncertainty in the interpolated SPM in places where no air quality monitoring exists. We believe that it will have a relatively minor impact on our results because most of the heavy populated areas have numerous monitors as shown in Figs. 1 and 3. Also, kriging interpolation depends on distances among monitors. Therefore, when monitors are far away from each other, the accuracy of the interpolated SPM decreases and vice versa. Additionally, assessing premature mortality based on actual readings of the monitors and not based on an interpolation surface is not practical because population distribution has a resolution of 5 km, and in heavily populated places, adjacent monitors are within a few kilometers.

3.2 Population Data

The GPWv3 data sets are an approximation of the real population and involve uncertainty. According to the CIESIN (2005), the errors in population estimations are caused by the interpolation method, the timeliness of the census, the number of estimations and their accuracy, and the boundaries. Additionally, we assume constant population and baseline mortality for each age group in Japan from 2006 to 2009. This involves uncertainty because of the difference between rural and urban areas; however, the percentages of each age group are almost the same in heavily populated prefectures as shown in Fig. 4. Additionally, there is an unavoidable uncertainty related to the fact that actual exposed populations are greater than the registered populations in prefectures such as Tokyo and Osaka, and according to the official statistics in 2005, population densities per square kilometer are 2,186.96 and 1,894.3, respectively. These exposed populations are much smaller than the actual population due to commuting and business activities there.
Fig. 4

Ratios of age groups in Japan in 2005 from Japan Statistics Bureau, 1996–2008 (e-stat 2011)

3.3 RR Value of PM2.5

The recognized relationship between race and health may affect our estimations of premature mortality based on CR functions from studies on American and European communities. This issue has been discussed extensively in Japan since 2007. In the scientific discussions of MOEJ, it was agreed that Western CR functions for PM2.5 are applicable in Japan and most of the available Japanese epidemiological studies, with exception of inconsistency between Japanese and Western cases by Ueda et al. (2012), produced similar results, e.g., Omori et al. (2003) and Katanoda et al. (2011). Accordingly, Japan has set the air quality standard to be the same as the US standards for PM2.5. The findings in this study are based on the RR values taken form Pope III et al. (2002). However, Jerrett et al. (2009) reported higher values; the RR values of PM2.5 from both models (single and two pollutants) vary from 1.048 (95 % CI, 1.024–1.071) to 1.08 (95 % CI, 1.048–1.113). Therefore, reevaluating Eq. (1) based on these values could give double of the estimated premature mortality in this study.

4 Conclusion

Assessing premature mortality caused by exposure to elevated concentrations of PM2.5 in Japan based on monitored data from 1,843 air quality monitors provides an efficient tool for assessing the health effects. This study shows clearly that the improvements in air quality in Japan have reduced premature mortality among sensitive age groups. The estimated premature mortality based on an assumption of 7.0 μg m−3 (40,000 cases) is about 1.4 times the estimations based on air concentrations of 10 μg m−3 (28,400 cases) for the period from 2006 to 2009. About 77 % of these cases are among age groups of +75 years. The improvements of air quality in Japan by reducing the emissions of SPM from 2006 to 2009 could save 3,568 and 3,602 lives based on 7.0 and 10 μg m−3analysis, respectively. These findings are tangible benefits gained by reducing the emissions of SPM in Japan. We think these estimations include levels of uncertainty. However, quantifying the uncertainty in our estimations such as uncertainty in RR values of PM2.5 cannot be achieved easily unless there are comprehensive Japanese epidemiological studies.


Thanks for Allah, my parents, and my wife and for Asia Center for Air Pollution Research, Japan for their support. Also, thanks for Dr. Scott Voorhees, USEPA for his comments and thanks for the Research Council in Oman for the support.

Copyright information

© Springer Science+Business Media Dordrecht 2013