Analysis of spatial and seasonal distributions of air pollutants by incorporating urban morphological characteristics

Highlights

• It is the one of the first studies that incorporates urban morphological characteristics as an essential predictor for air pollution distribution.

• The research is designed in consideration of sensitivity of parameter choices, and the approach is generally applicable to study other cities;

• Parallel analyses are conducted simultaneously for both warm and cold seasons;

Abstract

Due to the worldwide trend of industrialization and urbanization, air pollutants were emitted heavily on a global scale particularly in developing countries, which produces adverse effects on human health by causing health problems such as respiratory and lung diseases. Many regression models based on land use types and urban fabrics have been built to analyze the spatiotemporal distribution of air pollutants, however, few of them examined the relationship between urban morphological characteristics and the distribution of air pollutants in a megacity. This study investigates such relationships for six types of air pollutants (PM2.5, PM10, SO₂, NO₂, O₃, and CO) and a composite AQI (Air Quality Index) based on hourly data at 35 monitoring stations in Beijing in 2016, with morphological characteristics (Morphological building index), meteorological factors (Land Surface Temperature, LST), land use (vegetation, road length, gas station and industry point data), and population distribution data. We also analyzed the results with spatiotemporal regression and SSH (Spatial Stratified Heterogeneity) models respectively. According to the spatiotemporal regression model, the morphological building index (MBI) shows a strong correlation with the dispersion of PM2.5 (R² = 0.81) and AQI (R² = 0.80) in the warm season and this finding was reinforced through the Leave-one-out-cross-validation (LOOCV) analysis. From the SSH analysis, the road length in a large proximal region impacts air pollutants the most, especially for O₃; and population density significantly affects PM 2.5, AQI, SO₂, and NO₂ in the cold season. From an integrated interpretation, distance to nearest industry impacts the spatial distribution of NO₂ in cold season, while it impacts that of PM2.5 and AQI in both warm and cold seasons. The research finds that these two models supplement each other well and together help to give us a better understanding of how air quality is affected in the urban landscape.

Introduction

Since the rapid urbanization and industrialization, China has experienced deteriorating air condition, especially in large cities during the past two decades. For Chinese cities, the emissions of NO₂ (nitrogen dioxide) and SO₂ (sulfur dioxide) in 2010 were as 2.5 times and 1.5 times the values in 1990, respectively (Mendenhall, Sincich, & Boudreau, 1996). Among them all, the most common air pollutant is PM2.5 in recent years and a lot of attention has been paid to it (Wang, Hu, Chen, Chen, & Xu, 2013, Yang et al., 2017, Yuan, Liu, Castro, & Pan, 2012). In Beijing, the annual average concentration of PM2.5 was around 70–100 μg/m³, which were two to three times higher than the level 1 Interim Target (35μg/m³) assigned by WHO (Cheng et al., 2013). Besides, these emissions have adversely impacted air quality at regional, national, and even global scales and played an essential role in global climate change (Grimm et al., 2008). In addition, air pollutants have been damaging the environment due to their chemical properties. For example, the acidification of soil, lakes, and rivers are resulted from NO, SO₂, and Ammonia, which caused the loss of plant life, the habitat of animals, and the reduction of crop yield (Rao, Rajasekhar, & Rao, 2014). Moreover, annually, almost 3.7 million people die prematurely because of outdoor air pollution around the world (WHO, 2014). In China, air pollution has become the fourth greatest risk factor in all deaths (Fang, Liu, Li, Sun, & Miao, 2015). Many epidemiological researchers have demonstrated that air pollution leads to a variety of health problems by long-term exposure to air pollutants (Brunekreef & Holgate, 2002), which produces the considerable medical cost for individuals and economic loss for the government by the reduction of productivity.

Air pollutants are released from a variety of sources, such as industry and transportation, and are driven by various social economic factors. Many researchers believe air pollution is closely related to the urban land use patterns and optimization of spatial planning could improve the air quality in the long run. Thus, the relationship between the spatial land use and land cover and the distribution of air pollution attracts our attention. The conception of LUR (Land Use Regression) was first introduced and termed regression mapping (Briggs et al., 1997) and was applied in the SAVIAH (Small Area Variation in Air Quality and Health) project. The development and accessibility of GIS data and techniques contributed a lot to the popularity of LUR models. Ross et al. (2007) predicted PM2.5 in New York City and surrounding counties by LUR models, integrated with traffic density and census data. The model interpreted 60% or more of the variation of PM2.5 concentration with prediction errors below 10% (Ross, Jerrett, Ito, Tempalski, & Thurston, 2007). Meng et al. combined population, length of major roads, agricultural land area, and the number of the industry sites to the LUR model and explained a large part of the variability of NO₂ concentration in Shanghai, which outperformed the interpolation methods of Kriging and IDW (Meng et al., 2015). Liu, Henderson, Wang, Yang, and Peng (2016) used LUR model to interpret the variances of NO₂ and PM2.5 concentrations in Shanghai, and found anti-correlation with coastal regions and correlation with industry and highway intensity (Liu et al., 2016). Zheng et al. analyzed the land use patterns with the spatiotemporal distribution of multiple air pollutants (i.e. O₃, SO₂, NO₂, and CO) to find their relationships (Zheng et al., 2017). To improve the temporal granularity of air pollution monitoring data, Anand et al. built a mixed-effects LUR method to model daily NO₂ concentrations in Hong Kong from 2005 to 2015 with satellite datasets, which realized the daily mapping of ambient surface NO₂ (Anand & Monks, 2017). When it comes to the global scale, Larkin et al. created the first global NO₂ LUR model to find the spatial variability of NO₂ concentration. The model explained 54% of the annual change of NO₂ and continental R² ranges from 0.42 to 0.67 (Larkin et al., 2017).

Some studies constructed multiple LUR models for different scenarios or purposes. For example, Wu et al. constructed eight LUR models to explain diurnal, seasonal and annual spatial changes of PM2.5 concentration in Beijing (Wu et al., 2015). Huang et al. explored the relationship of fifty-nine potential variables (e.g. land use, traffic and industry emission, and population density) with four air pollutants (i.e. PM2.5, SO₂, NO₂, and O₃) through LUR models based on national monitoring network, and the variance of the four pollutants could be explained to a certain degree (Huang et al., 2017a). Yang et al. set four LUR models to estimate the air pollution concentration through ground-based measurements, remote sensing data, air quality model, and other spatial inputs, and found the best model to explain the NO₂ and PM2.5, respectively (Yang et al., 2017). Yang et al. developed seasonal LUR models to simulate the change of PM2.5 in the urban area, and these models had a good fit and explained the variation of spatial distribution of PM2.5 concentration well (Yang et al., 2017). However, such multiple LUR models may not always show good performance. For instance, Muttoo et al. stated that seasonal models did not show clear differences for measuring NOx with similar R² values, and this was due to the high correlation between seasonal measurements for each of the monitoring sites (Muttoo et al., 2018).

Even though the spatial land use patterns are related to the distribution of air pollutants, the accuracy of the final results depends heavily on the accuracy of classification of land use types and most of them interpreted the results based on LUR model only. Besides, the urban morphology, as the three-dimension form of a set of buildings and urban shapes (Chen, 2013), has significant influences on the concentration of air pollution (Cárdenas Rodríguez, Dupont-Courtade, & Oueslati, 2016). Bereitschaft et al. quantified urban form by preexisting sprawl indices and spatial metrics and analyzed their relationship with air pollution among 86 U.S. cities (Bereitschaft & Debbage, 2013). Yuan et al. quantified the impact of urban morphological parameters, urban permeability, and building geometries, on the dispersion of air pollution by CFD (Computational fluid dynamics) approach in the high-density urban regions (Yuan, Ng, & Norford, 2014). Rodríguez et al. proved that urban morphology produced significant influences on pollution centration and found that higher concentration of NO₂ and PM10 was related to fragmented and highly constructed cities and the higher concentration of SO₂ was related to densely populated cities (Cárdenas Rodríguez et al., 2016). She et al. (2017) analyzed the correlation between urban form which was described by six spatial metrics and group-based measurements of six air pollutants in the largest metropolitan zone, Yangtze River Delta. The study demonstrated that urban form did affect the urban air quality (She et al., 2017). Lu et al. developed the geographically weighted regression model to analyze the relationship between urban form and the density of NO₂ and SO₂ derived by satellite data, and the results showed that urban form produced significant effects on the air quality in urban areas of China (Lu & Liu, 2016).

As for street-level analysis, Maignant evaluated the dispersion of air pollution through the MISKAM (Mikroskaliges Klima und Ausbreitungsmodell) model integrating with urban morphology, buildings volume and roughness, and climatology factors in a street canyon (Maignant, 2006). Edussuriya et al. found the linkage between urban morphology and air quality, and the final results showed that six morphological variables significantly explained the variance of air pollution at the street level (Edussuriya, Chan, & Ye, 2011). Shen et al. investigated the connection of street morphology or canyons to dispersion of air pollution for six cities around the world. It was concluded that an open central street would greatly improve the air quality due to a larger vertical exchange of air flow via the street roof (Shen, Gao, Ding, & Yu, 2017). Shi et al. used vehicle-based mobile measurements and regression models to estimate the spatial distribution of PM2.5 and PM10, and found the most decisive factor of urban morphology was the frontal area index on street-level air quality in the central area of Hong Kong (Shi et al., 2016a).

However, these previous studies emphasized those building morphology parameters (e.g. sky view factor (Shi et al., 2016b, Silva & Monteiro, 2016), and building orientation (Lu & Liu, 2016)) or urban form indices (e.g. frontal area index (Ghassoun & Löwner, 2017, Lu & Liu, 2016, Shi et al., 2016a)) that are largely affected by meteorological factors such as wind direction and wind speed, or by the detailed building fabrics, which is more suitable for small research areas with clear 3D building maps and accurate group-based field measurements. For a megacity like Beijing, this process would be time-consuming and need too much human interference. This current study proposes to investigate morphological characteristics at various spatial levels based on data extracted from high-resolution (HR) images. Such urban form characteristics has not been investigated before in statistical approaches to analyzing the spatiotemporal distribution of air pollution in megacities. Another shortcoming in many of the previous LUR models is the assumption of linear relationship. Because any type of models (e.g. LUR model) may have its own limitations due to its assumption (e.g. linearity) and internal mechanism, the current study will apply multiple models to supplement each other in order to provide findings for a better understanding.

This study aims to answer three questions for the study area: (1). What is the potential utility of morphological information for estimating the spatiotemporal distribution of air pollution? (2). Are there seasonal differences in the distributions of air pollutants and if so how are these differences distributed spatially? (3). To what extent does each factor of interest affects the spatial and seasonal distributions of air pollutants? The innovation of the research lies in four aspects: i) it is the one of the first studies that incorporates urban morphological characteristic, which is not sensitive to the accuracy of land use classification and is easy for general adoption, as an essential predictor for air pollution distribution. We combine morphological, meteorological, land use and socioeconomic variables together to analyze the distribution of air pollutants' concentrations; ii) the research is designed carefully in consideration of possible sensitivity of parameter choices, thus the approach is generally applicable to study other cities; iii) parallel analyses of spatial distribution of seven air pollution indexes are conducted simultaneously for both warm and cold seasons, which helps to capture the changing dynamics of air pollutants throughout the year; iv) the use of both spatiotemporal regression and SSH models provides a more comprehensive understanding of how air pollutants concentrations are affected by different factors.