Elsevier

Science of The Total Environment

Volume 658, 25 March 2019, Pages 1256-1264
Science of The Total Environment

Estimation of spatiotemporal PM1.0 distributions in China by combining PM2.5 observations with satellite aerosol optical depth

https://doi.org/10.1016/j.scitotenv.2018.12.297Get rights and content

Highlights

  • Fusing PM2.5 observations is significant to improve estimation accuracy of PM1.0.

  • Estimation performance shows significant daily, seasonal and spatial variations.

  • Estimation accuracy largely dependences on quality of AOD and interpolated PM2.5

Abstract

Particulates smaller than 1.0 μm (PM1.0) have strong associations with public health and environment, and considerable exposure data should be obtained to understand the actual environmental burden. This study presented a PM1.0 estimation strategy based on the generalised regression neural network model. The proposed strategy combined ground-based observations of PM2.5 and satellite-derived aerosol optical depth (AOD) to estimate PM1.0 concentrations in China from July 2015 to June 2017. Results indicated that the PM1.0 estimates agreed well with the ground-based measurements with an R2 of 0.74, root mean square error of 19.0 μg/m3 and mean absolute error of 11.4 μg/m3 as calculated with the tenfold cross-validation method. The diurnal estimation performance displayed remarkable single-peak variation with the highest R2 of 0.80 at noon, and the seasonal estimation performance showed that the proposed method could effectively capture high-pollution events of PM1.0 in winter. Spatially, the most polluted areas were clustered in the North China Plain, where the average estimates presented a bimodal distribution during daytime. In addition, the quality of satellite-derived AOD, the robustness of the interpolation algorithm and the proportion of PM1.0 in PM2.5 were confirmed to affect the estimation accuracy of the proposed model.

Introduction

As the major components of aerosols, particulate matter (PM) exert remarkable impacts on the climate system through their direct and indirect effects (Qin et al., 2017; Wang et al., 2015). The report of IPCC5 indicates that, the impact of aerosol–cloud interaction on climate remains the most uncertain factor in global radiation changes (IPCC, 2013). Many studies on aerosol climatic effects have been conducted (Charlson et al., 1992; Pan et al., 2017; Lu et al., 2018; Mao et al., 2018). Large amounts of particle emissions seriously reduce atmospheric visibility and affect daily traffic. Moreover, particles smaller than 10 μm are inhalable by humans and threaten human health (Bartell et al., 2013; Kulshrestha, 2018).

Studies on the regional spatiotemporal evolution of PM with different diameters acquired through station measurements have indicated that fine particles, especially PM1.0, are the major contributor to air pollution (Shi et al., 2014; Zhuang et al., 2014). Statistically, PM1.0 accounts for more than 80% of the average PM2.5 and even reaches 98% in several regions of China (Li et al., 2015). PM1.0 data is more useful than PM2.5 data in assessing anthropogenic emissions. Morawska et al. (2008) investigated the size distribution of ambient particles and found that PM1.0 enables improved distinction between combustion and mechanically generated aerosols. Furthermore, the health hazards of PM increase with the decrease in particulate diameter (Huang et al., 2003; Elder et al., 2015) because fine particles contain considerable harmful substances and could enter the respiratory system, especially the alveoli, deeply, thereby causing serious threats to humans (Zhang et al., 2016).

Given the strong associations of PM1.0 with public health and the environment, large amounts of exposure data should be obtained to strengthen particle pollution control. However, PM1.0 monitoring stations are sparse and unevenly distributed, especially in sparsely populated western parts of China; this feature is not conducive for conducting continuous spatiotemporal monitoring of ultrafine particulates. Advanced satellite remote sensing with wide observation horizons can effectively compensate for limitations in ground measurements (Guo et al., 2009). Studies have demonstrated that aerosol extinction is closely related to particles with diameters between 0.4 and 1.0 μm (Madronich and Flocke, 1999). Therefore, satellite-derived aerosol optical depth (AOD), which is the column integral of aerosol's extinction coefficient, could be a useful tool for mapping PM1.0 distributions in theory. Current studies on PM1.0 estimation with satellite remote sensing are only in the exploratory stage. Chen et al. (2018) investigated PM1.0 estimation on the basis of moderate resolution imaging spectroradiometer (MODIS) AOD in China and obtained 59% variability for daily estimations. Qin et al. (2018) estimated PM1.0 concentrations from MODIS over Yangtze River Delta, China, by using a geographically and temporally weighted regression model with R2 of 0.74 for daily estimations. However, hourly PM1.0 concentration with high spatial coverage is needed when evaluating a diurnal transport and control approach of air pollution. As a geostationary satellite, Himawari-8 provides an unprecedented opportunity. Our previous work (Zang et al., 2018) showed that the limited number of PM1.0 sites and insufficient ground-based observations cannot effectively constrain the estimated results of PM1.0 (R2 of 0.65), and model accuracy needs further improvement.

In this study, considering the high correlation between PM1.0 and PM2.5 and large amount of available observation data on PM2.5, we estimated PM1.0 by combining ground-based PM2.5 measurements with hourly satellite aerosol retrievals in China into an improved generalised regression neural network (GRNN) model (hybrid GRNN model with the principal component analysis [PCA] method, namely, PCA–integrated GRNN model). We designed three groups of comparative experiments, which consisted of multiple linear regression (MLR), GRNN and PCA-integrated GRNN models with or without PM2.5 observations. We compared the tenfold cross-validation (CV) results of these models by fusing particle observations to evaluate the improvement in each model. In addition, we estimated hourly PM1.0 concentrations by using a well-trained model and depicted the spatial distribution of annual and seasonal averages of PM1.0. The error sources of estimates were discussed and analysed for future improvements of the proposed method.

Section snippets

Datasets

The datasets employed in the PM1.0 estimation are summarised in Table 1.

Evaluation of model performance

Table 2 summarises the PM1.0 estimation performance of each model from three groups of comparative experiments. Comparisons of the first and second experimental groups indicated that the nonlinear ANN model could better learn the potential relationship between PM1.0 and its predictors, and its estimation accuracy was improved by nearly twice. Specifically, the values of R2, RMSE and MAE of the MLR model without PM2.5 observations were 0.21, 33.3 μg/m3 and 23.0 μg/m3, respectively, whereas the

Conclusions

Fine PM poses a greater health risk than coarse particulates (PM10) (Oberdörster et al., 2005), and the extinction coefficient increases with the increase in PM1.0 fraction (Sabbagh-Kupelwieser, 2010). Despite the abundance of studies on PM2.5 and PM10 estimations from satellite remote sensing, only a few studies have been conducted on PM1.0 by using satellite observations. In this study, we estimated hourly PM1.0 concentrations in China by using a PCA–integrated GRNN model that combines

Conflicts of interest

The authors declare no conflict of interest.

Acknowledgments

This study was supported by the National Key Research and Development Program of China (grant numbers 2017YFC0212600, 2016YFC0200900 and 2017YFC1501401) and the National Natural Science Foundation of China (grant numbers 41701381 and 41627804). The authors are also grateful to CMA, Japan Aerospace Exploration Agency, ECMWF, Data Center of the US NASA, and USGS for providing the datasets used in this work.

References (48)

  • K. Qin et al.

    Estimating PM1 concentrations from MODIS over Yangtze River Delta of China during 2014–2017

    Atmos. Environ.

    (2018)
  • Y. Shi et al.

    Airborne submicron particulate (PM1) pollution in Shanghai, China: chemical variability, formation/dissociation of associated semi-volatile components and the impacts on visibility

    Sci. Total Environ.

    (2014)
  • W. Sun et al.

    Daily PM2.5 concentration prediction based on principal component analysis and LSSVM optimized by cuckoo search algorithm

    J. Environ. Manag.

    (2016)
  • L. Wen et al.

    Enhanced formation of fine particulate nitrate at a rural site on the North China Plain in summer: the important roles of ammonia and ozone

    Atmos. Environ.

    (2015)
  • Y. Wu et al.

    Synergy of satellite and ground based observations in estimation of particulate matter in eastern China

    Sci. Total Environ.

    (2012)
  • L. Zang et al.

    Estimating hourly PM1 concentrations from Himawari-8 aerosol optical depth in China

    Environ. Pollut.

    (2018)
  • J. Zhang et al.

    Chemical characteristics and influence of continental outflow on PM1.0, PM2.5 and PM10 measured at Tuoji island in the Bohai Sea

    Sci. Total Environ.

    (2016)
  • T. Zhang et al.

    Estimation of ultrahigh resolution PM2.5 concentrations in urban areas using 160 m Gaofen-1 AOD retrievals

    Remote Sens. Environ.

    (2018)
  • Q. Zhou et al.

    A hybrid model for PM₂.₅ forecasting based on ensemble empirical mode decomposition and a general regression neural network

    Sci. Total Environ.

    (2014)
  • X. Zhuang et al.

    Haze insights and mitigation in China: An overview

    J. Environ. Sci.

    (2014)
  • J.C. Anderson et al.

    Accuracy assessment of Aqua-MODIS aerosol optical depth over coastal regions: importance of quality flag and sea surface wind speed

    Atmos. Meas. Tech. Discuss.

    (2012)
  • S.M. Bartell et al.

    Particulate air pollution, ambulatory heart rate variability, and cardiac arrhythmia in retirement community residents with coronary artery disease

    Environ. Health Perspect.

    (2013)
  • R.J. Charlson et al.

    ClimateForcing by antlropogenic aerosols

    Science

    (1992)
  • A.V. Donkelaar et al.

    Estimating ground-level PM2.5 using aerosol optical depth determined from satellite remote sensing

    J. Geophys. Res. Atmos.

    (2006)
  • Cited by (60)

    • Revisiting the atmospheric particles: Connecting lines and changing paradigms

      2022, Science of the Total Environment
      Citation Excerpt :

      Perrone et al. (2013) highlighted the importance of submicron PM measurements as it provides a better estimation of anthropogenic particles than PM2.5. Zang et al. (2019) estimated PM1 through a strategy that combined PM2.5 ground-based observations and aerosol optical depth (AOD) satellite-derived data in China. However, the uncertainties brought by the chemical differences between PM1 and PM2.5 were remarked upon by Sun et al. (2021b) in a highly polluted environment of China.

    View all citing articles on Scopus
    View full text