Modulation of synoptic circulation to dry season PM2.5 pollution over the Pearl River Delta region: An investigation based on self-organizing maps

https://doi.org/10.1016/j.atmosenv.2020.117482Get rights and content

Highlights

  • A synoptic catalogue is developed for the PRD region by self-organizing maps.

  • The synoptic modulation to PM2.5 pollution over the PRD region is investigated.

  • The synoptic circulations play a role via redistributing the air pollutants within the PRD region.

Abstract

Fine particle (PM2.5) pollution is of concern for the Pearl River Delta (PRD) region, particularly during the dry season (October to March), when the PM2.5 concentrations can exceed Chinese national standards. In part due to sparse observations, the dispersion/transport mechanisms that lead to PM2.5 pollution over the PRD region are not yet fully understood, and the potential synoptic controls have not been investigated. In this study, systematic analyses were conducted using the multisite surface meteorological observations, two-site radiosonde measurements, and regional PM2.5 observations, based on the application of self-organizing map neural network to large-scale mean sea-level pressure (MSLP) data. The results suggested that the relative position of the PRD to high-pressure systems exerted significant effects on the dispersion conditions and the PM2.5 distribution over the PRD region. When cold high pressure invaded, the PRD region was dominated by synoptic-scale northerly flow, which caused southward pollution transport and hence positive PM2.5 anomalies in the coastal area. When cold high pressure weakened and moved eastward, the PRD region was dominated by meso/local-scale flows, resulting in strong atmospheric recirculation and elevated PM2.5 concentrations in the inland area. However, the regional average of PM2.5 concentrations was not sensitive to the changes in synoptic circulation patterns, which implied that the synoptic circulation played roles most in redistributing the air pollutants within the PRD region and hence a regional PM2.5 pollution emission reduction is needed to improve regional air quality.

Introduction

With the rapid urban growth and industrial development in recent decades, high concentrations of ambient pollutants have become major environmental problems because of their adverse effects on public health (Chen et al., 2017; Lelieveld et al., 2015; Shiraiwa et al., 2017). Many studies have investigated the formation mechanisms of air pollution, revealing that the major contributors are anthropogenic emissions and meteorological conditions. Due to the quasi-stable nature in anthropogenic emissions, the short-term fluctuations in air quality are thought to be attributed to variable transport/dispersion conditions, which are largely subject to synoptic-scale circulations (Demuzere et al., 2009; Zhang et al., 2012). Therefore, understanding the synoptic modulation to air pollution is very important for the early prediction of air pollution episodes.

Synoptic-scale circulations represent certain atmospheric conditions in a given area through their close association with various meteorological variables (Zheng et al., 2015). By classifying large-scale circulation states into different patterns, synoptic climatology serves as a powerful tool to develop direct linkages between synoptic circulation and local environment (Dayan et al., 2012; Yarnal, 1993; Yarnal et al., 2001). The last ten years have witnessed a growing application of automated circulation classification in the air pollution research (Demuzere et al., 2009; Hegarty et al., 2007; Liao et al., 2017; Mao et al., 2020; Miao et al., 2017, 2019; Pope et al., 2015; Russo et al., 2014; Santurtun et al., 2015; Shu et al., 2017; Ye et al., 2016; Zhang et al., 2012, 2013). However, the traditional synoptic classification methodologies (e.g., principal component analysis) suffer from a general shortcoming, i.e., the clustered circulation patterns cannot be organized into a continuum (Hewitson and Crane, 2002; Sheridan and Lee, 2011). Such shortcoming could propagate to subsequent analysis, inhibiting a detailed identification of synoptic modulation to air pollution.

The abovementioned shortcoming can be solved by the self-organized map (SOM) (Hewitson and Crane, 2002), an unsupervised neural network algorithm based on competitive and cooperative learning that projects data onto a topologically ordered array (Kohonen, 1995). Unlike traditional classification methods, the SOM array represents the atmospheric continuity and could visualize relationships between synoptic circulation and local environment (Liu et al., 2006). The advantages have made the SOM become an increasingly popular tool in atmospheric research (Gallagher et al., 2018; Gibson et al., 2016; Horton et al., 2015; Jiang et al., 2015; Juliano and Lebo, 2019; Lennard and Hegerl, 2015; Liu et al., 2016; Mattingly et al., 2016; Yu and Zhong, 2018). For the first time, Pearce et al. (2011) extended the SOM method into air pollution research and proved the SOM technique to be a robust medium for examining the relationships between synoptic circulations and air pollution. After that, Jiang et al. (2017) conducted an investigation on how subtle changes in the synoptic systems affected the local meteorology and air quality in Sydney based on the SOM classification; Han et al. (2019) quantified the local and synoptic meteorological influences on the daily variability of summertime surface ozone in eastern China with the aid of the SOM method.

The Pearl River Delta (PRD) region (i.e., the Guangdong-Hong Kong-Macao Greater Bay area), which is located in the coastal part of southern China, has recently been identified as the largest morphologically continuous urban agglomeration in the world (Taubenböck et al., 2019). As a side effect of megacity urbanization, the regional air pollution caused by photochemical smog and haze-fog have become urgent environmental problems (Deng et al., 2019; Wang et al., 2015; Wu et al., 2005). In particular, frequent fine particle (PM2.5) pollution during the dry season (October to March) poses a health threat to the 70 million inhabitants living in this region (Lin et al., 2016). Topographically, the PRD region is situated within a basin bounded by elevated terrain to the east, north and west, and the South China Sea to the south (Fig. 1). The local meteorological conditions in the PRD region are hence influenced by a range of synoptic circulation features and some complex terrain-related local and mesoscale phenomena, such as sea/land breezes, slope/valley winds and urban heat island circulation (Fan et al., 2006, 2008, 2011; Lu et al., 2010; Wang et al., 2017). Previous studies have demonstrated that each type of local and mesoscale circulation plays an important role in the status of the air quality over the PRD region (Ding et al., 2004; Li et al., 2018; Liu et al., 2017; Lo et al., 2006; Wang et al., 2018; Wu et al., 2013; Xia et al., 2016). However, the demonstrated importance of the local to mesoscale structures for regional air pollution is not detached from the larger-scale synoptic circulation. There is still a gap in the understanding of the response of local meteorological conditions to large-scale circulation patterns and the modulation of synoptic circulation to PM2.5 pollution over the PRD region.

This study combined the SOM method with quantitative measures of regional wind regimes to investigate how the local, mesoscale and large-scale synoptic meteorological conditions jointly affect the PM2.5 air quality across the PRD region based on multisite observations during the dry seasons from 2013 to 2019, aiming to establish a visualized linkage between synoptic circulation patterns, local meteorological conditions and daily air quality in this complex coastal-basin environment. The results would improve our understanding of PM2.5 pollution over the PRD region from multiscale perspective.

Section snippets

Meteorological data

In this analysis, daily mean sea-level pressure (MSLP) data from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-Interim reanalysis dataset were used to determine the synoptic circulation patterns affecting the PRD during the dry seasons from 2013 to 2019. The analyzed domain covers the latitudes 12.5°N–32.5°N and longitudes 97.5°E−127.5°E (horizontal resolution 0.5° × 0.5°), in which the PRD is located near the center. To obtain more homogeneous data for the SOM-based

Self-organizing maps

Self-organizing maps were employed to categorize the large-scale circulation conditions to develop a dry season synoptic climatology for the PRD region. The SOM technique, which conceptually dated back to the late 1980s and was first presented in detail in Kohonen (1995), is a neural network algorithm that uses unsupervised classification to perform nonlinear mapping of high-dimensional datasets onto regular two-dimensional arrays (i.e., nodes). It is preferred over other circulation

Self-organized synoptic circulation patterns

The SOM of the MSLP provided a clear visualization of the atmospheric continuum by presenting nine archetypes of large-scale circulation patterns (Fig. 3). Due to the nature of the SOM method, the nodes on the SOM plane (Fig. 3) were topologically ordered (i.e., self-organized), with the corner nodes representing the circulation patterns that differed most from each other. There is a gradual transition between the neighboring patterns on the SOM plane, which corresponds to subtle change in the

Conclusion

Based on SOM neural network algorithm, a dry season synoptic catalogue was developed for the PRD region, comprising 9 circulation patterns that were self-organized on the SOM plane. The self-organized circulation patterns facilitated a systematic investigation on how the synoptic circulation variability affected regional meteorological conditions and subsequently the PM2.5 air quality over the PRD region.

The results suggested that the synoptic circulation patterns played an important role in

CRediT authorship contribution statement

Zhiheng Liao: Conceptualization, Methodology, Software, Data curation, Visualization, Writing - original draft. Jielan Xie: Investigation, Data curation, Formal analysis, Validation. Xingqin Fang: Investigation, Data curation, Writing - review & editing. Yu Wang: Investigation, Data curation. Yu Zhang: Investigation, Data curation. Xinqi Xu: Investigation, Data curation. Shaojia Fan: Conceptualization, Resources, Supervision, Project administration.

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.

Acknowledgments

This study was supported by the National Key Research and Development Plan of China (Nos. 2017YFC0209606 and 2016YFC0203305) and the National Natural Science Foundation of China (No. 41630422).

References (57)

  • A. Santurtun et al.

    Surface ozone concentration trends and its relationship with weather types in Spain (2001-2010)

    Atmos. Environ.

    (2015)
  • H. Taubenböck et al.

    A new ranking of the world's largest cities—do administrative units obscure morphological realities?

    Rem. Sens. Environ.

    (2019)
  • X. Wang et al.

    Progress in understanding the formation of fine particulate matter and ground-level ozone in Pearl River Delta region, China

    Atmos. Environ.

    (2015)
  • D. Wu et al.

    An extremely low visibility event over the Guangzhou region: a case study

    Atmos. Environ.

    (2005)
  • X. Ye et al.

    Study on the synoptic flow patterns and boundary layer process of the severe haze events over the North China Plain in January 2013

    Atmos. Environ.

    (2016)
  • Y. Zhang et al.

    Impact of synoptic weather patterns on spatio-temporal variation in surface O-3 levels in Hong Kong during 1999-2011

    Atmos. Environ.

    (2013)
  • R.J. Chen et al.

    Fine particulate air pollution and daily mortality A nationwide analysis in 272 Chinese cities

    Am. J. Resp. Crit. Care

    (2017)
  • U. Dayan et al.

    On the importance of synoptic classification methods with respect to environmental phenomena

    Int. J. Climatol.

    (2012)
  • M. Demuzere et al.

    The impact of weather and atmospheric circulation on O-3 and PM10 levels at a rural mid-latitude site

    Atmos. Chem. Phys.

    (2009)
  • S.J. Fan et al.

    Atmospheric boundary layer characteristics over the Pearl River Delta, China, during the summer of 2006: measurement and model results

    Atmos. Chem. Phys.

    (2011)
  • S.J. Fan et al.

    Atmospheric boundary layer features of Pearl River Delta and its conception model (in Chinese)

    China Environ. Sci.

    (2006)
  • M.R. Gallagher et al.

    Impact of atmospheric circulation on temperature, clouds, and radiation at summit station, Greenland, with self-organizing maps

    J. Clim.

    (2018)
  • P.B. Gibson et al.

    Projected changes in synoptic weather patterns over New Zealand examined through self-organizing maps

    Int. J. Climatol.

    (2016)
  • H. Han et al.

    Local and synoptic meteorological influences on daily variability of summertime surface ozone in eastern China

    Atmos. Chem. Phys. Discuss.

    (2019)
  • J. Hegarty et al.

    Synoptic controls on summertime surface ozone in the northeastern United States

    J. Geophys. Res. Atmos.

    (2007)
  • B. Hewitson et al.

    Self-organizing maps: applications to synoptic climatology

    Clim. Res.

    (2002)
  • D.E. Horton et al.

    Contribution of changes in atmospheric circulation patterns to extreme temperature trends

    Nature

    (2015)
  • N.B. Jiang et al.

    Insights into the implementation of synoptic weather-type classification using self-organizing maps: an Australian case study

    Int. J. Climatol.

    (2015)
  • Cited by (21)

    • Quantitative evaluation of impacts of the steadiness and duration of urban surface wind patterns on air quality

      2022, Science of the Total Environment
      Citation Excerpt :

      Air pollution is a complex phenomenon that is influenced by multiple variables, including anthropogenic emissions, meteorological conditions, chemical processes, and topography (Carvalho et al., 2006; Lazaridis, 2011; Monks et al., 2009; Thompson, 2001). Of these four dominant factors, meteorological conditions significantly modulate air quality by affecting multiple physical and chemical processes, including pollutant transport, diffusion, dry/wet deposition, and photochemistry (Cassano et al., 2016; Fan et al., 2008; Guo et al., 2021; Leung et al., 2018; Liao et al., 2020; Miao et al., 2017; Wang et al., 2017; Xiao et al., 2006). Meteorological parameters are important for air-quality forecasts, including quantitative models and empirical and statistical forecasts (Agarwal et al., 2020; Shi et al., 2021; Stirnberg et al., 2021).

    • Spatiotemporal variations of PM<inf>2.5</inf> pollution and its dynamic relationships with meteorological conditions in Beijing-Tianjin-Hebei region

      2022, Chemosphere
      Citation Excerpt :

      The first is to simulate the nexus between haze and weather conditions through physical-chemical models such as the Chemical Transport Models (Bove et al., 2014; Sun et al., 2021). The second is to identify the synoptic-scale circulation modes in a specific area through the self-organizing map, and further examine how they affect PM2.5 pollution (Chang et al., 2020; Liao et al., 2020; Liu et al., 2021a). The third is to quantify the impact of meteorology on air quality through statistical models such as the Generalized Additive Model (Xiao et al., 2021; Yanez et al., 2017).

    • Insights into the source-specific health risk of ambient particle-bound metals in the Pearl River Delta region, China

      2021, Ecotoxicology and Environmental Safety
      Citation Excerpt :

      Urban heat islands (UHI) may induce thermal low pressure areas that produce convergence zones, which facilitate the meso/local-scale transport of primary pollutants (Lai and Cheng, 2009). Liao et al. (2020) found that when the PRD region was dominated by this kind of meso/local-scale circulations, severe PM2.5 pollution occurred in the inland area and extreme pollution was found in the northwestern inland region. All of these facilitated the accumulation of particle-bound metals in the northwest city cluster, resulting in a distinct spatial difference among the three city clusters.

    View all citing articles on Scopus
    View full text