Exploring the relationships between 3D urban landscape patterns and PM 2.5 pollution using the multiscale geographic weighted regression model

Abstract

Fine particulate matter (PM_2.5) is a major source of air pollution and exerts serious impacts on human health. The 3D urban landscape patterns can significantly affect the diffusion and emissions of PM_2.5. However, studies on the relationships between 3D urban landscape patterns and PM_2.5 pollution across different seasons remain understudied. With the ground-level air pollutants estimated by the remote sensing and fine-scale building information, this study applied the multiscale geographically weighted regression model to explore such relationships. Wuhan, the largest metropolis in Central China, was selected as the study area for the application of our methodology. The results showed that the direction, degree, and scale of the effect of 3D urban landscape patterns on PM_2.5 pollution varied across seasons. For building height, the standard deviation of building height had a significant positive correlation with PM_2.5 all year round. For building density, the building count density showed a significant positive correlation with PM_2.5 in general, with the bandwidth in winter and autumn smaller than in spring and summer. The building plan area fraction exerted both positive and negative influences on PM_2.5, dependent on season and location. The bandwidth of it gradually increased from spring to winter, with the effect changing from local to regional scale. For building volume, the floor area ratio showed a significant negative correlation with PM_2.5 in winter and autumn, and a localized effect was found, especially in winter. The findings of this study provide practical implications for urban planning and policy making to mitigate PM_2.5 pollution in the rapidly urbanizing regions.

Keywords

3D urban landscape pattern MGWR seasonality diffusion and emission

Introduction

Air pollution has become one of the most serious environmental problems in China (Lu et al., 2020). PM_2.5 (particulate matter with aerodynamic diameter less than 2.5 μm) is one of the main air pollutants and poses a great threat to public health (Ulpiani, 2021). PM_2.5 can induce asthma and respiratory inflammation (Liu et al., 2022), jeopardize lung functions (Hou et al., 2020), and even promote the incidence of cancer (Park et al., 2022). It is estimated that the deaths related to PM_2.5 pollution in China increased from 714,000 in 2000 to 971,000 in 2017 (Yue et al., 2020). To prevent and control PM_2.5 pollution, the Chinese Ministry of Ecology and Environment revised the Ambient Air Quality Standards in 2012 (GB3095-2012) and PM_2.5 pollution was included in air quality assessment for the first time. Moreover, the State Council of China launched the Air Pollution Prevention and Control Action Plan (APPCAP) in 2013. As the largest air pollution control action ever implemented, the plan aims to reduce PM_2.5 concentrations by 10%–25% by 2017 in the cities at prefecture level and above. Up to 2018, however, only 121 out of the 338 cities (accounting for 35.8%) met the standard value of PM_2.5 concentrations regulated by the GB3095-2012. Thus, PM_2.5 pollution is still a major environmental problem in China and remains a perennial concern to the public. To identify the influencing factors of PM_2.5 pollution and take effective measures to abate PM_2.5 pollution in Chinese cities have become urgent issues.

During periods of heavy PM_2.5 pollution, measures such as promotion of clean energy technology, limitation of the number of vehicles, and relocation of industry are effective at controlling PM_2.5 emissions and improving air quality. However, when air quality enters the phase of moderate and low pollution, the above-mentioned measures may become less effective as the sources of PM_2.5 emissions become concealed and changeable (Li et al., 2023). At this point, shaping urban landscape patterns is found to be one of the most effective ways to alleviate PM_2.5 pollution (Mao et al., 2022). Based on remote sensing data, the impacts of urban size, fragmentation, and compactness on PM_2.5 pollution at regional and national scales have been investigated (Sun et al., 2022; Tan et al., 2023; Zhao et al., 2022). In particular, urban compactness has a complex relationship with PM_2.5 pollution. On the one hand, a large and compact city tends to have heavier PM_2.5 pollution than a decentralized and multi-center city due to its poor diffusion conditions (Mao et al., 2022). On the other hand, the compact urban form can shorten commuting distances and reduce the PM_2.5 emissions from public transportation (Wang et al., 2022a). These studies take the city as a whole to explore the relationships between 2D urban forms and PM_2.5 pollution, whereas the intra-urban variability of PM_2.5 has been ignored. Recent study has shown that the spatial variability of PM_2.5 is non-negligible and is closely related to urban landscape heterogeneity (Yuan et al., 2019), which should stimulate our attention.

Recently, advances in high-resolution remote sensing, SAR, and LiDAR facilitate the acquisition of 3D urban information, and the influence of 3D urban landscape patterns on PM_2.5 concentrations has been increasingly investigated (Ke et al., 2022; Liu et al., 2021; Xu and Chen 2021; Yuan et al., 2019). A majority of studies have explored such influence based on fixed and mobile air quality monitoring stations (Chen et al., 2021; Ke et al., 2022; Liu et al., 2021; Wang et al., 2022b; Zeng et al., 2022). It has been shown that the 3D urban landscape patterns are closely related to the diffusion and emissions of PM_2.5. First, the 3D urban landscape patterns can change ventilation, air temperature, and humidity through the canyon effect, shading effect, and wind-blocking effect, thus affecting the diffusion of PM_2.5 (Ke et al., 2022). Second, the 3D urban landscape patterns have a direct or indirect effect on the emissions of PM_2.5. For example, Chen et al. (2021) reported that the area with higher floor area ratio (FAR) tends to have larger population and vehicles, resulting in more PM_2.5 emissions. In addition to station-based observations, recent development of remote sensing and data processing techniques also enable acquisition of high-resolution and spatially continuous PM_2.5 data. Based on remote-sensing-estimated PM_2.5, Yuan et al. (2019) explored the associations between built environments (i.e., land use, land cover, and urban form) and PM_2.5 concentrations in the central urban area of Wuhan using the spatial lag and error models. Yang et al. (2022) assessed the impact of urban landscape patterns on PM_2.5 pollution in Nanchang through adoption of the local climate zones. Undoubtedly, these studies provide a new perspective for urban planning to mitigate PM_2.5 pollution.

Although recent studies have made great progress, they have some limitations. First, few studies have taken into account the seasonal effect. The meteorological conditions and PM_2.5 pollution sources can vary largely in different seasons. These may cause the relationships between 3D urban landscape patterns and PM_2.5 pollution to vary through the year. For example, Cao et al. (2021) found that the influence of 3D urban landscape patterns on PM_2.5 pollution is more pronounced in spring and autumn than in winter and summer in Beijing. Second, PM_2.5 pollution and urban landscape patterns have high spatial variability. Global regression models assume that the relationships between explanatory and dependent variables are spatially invariable (Hu et al., 2022). By contrast, the geographically weighted regression (GWR) model allows regression coefficients to vary spatially (i.e., local fitting), thus producing higher explanatory power than global regression models (Su et al., 2012). However, GWR assumes that all spatial processes (i.e., explanatory variables) operate at the same spatial scale (Wang et al., 2022c). Previous studies have shown that the influence of 3D urban landscape patterns on PM_2.5 pollution had a significant multiscale effect. For example, Shi et al. (2019b) found that compact and open high-rise built environments had a wider range of influence on PM_2.5 pollution than the other types. Zeng et al. (2022) estimated PM_2.5 concentrations using the land-use regression model and found that the optimal buffer sizes of building height and building coverage ratio were larger than building volume density. Alternatively, the multiscale GWR model (MGWR) allows the relationships between explanatory and dependent variables to vary on different scales with changeable bandwidths (Mansour et al., 2021). To our knowledge, few studies have used MGWR to explore the relationships between urban landscape patterns and PM_2.5 pollution, although this approach has been increasingly used in geographical and ecological research (Bi et al., 2022; Mollalo et al., 2020; Rong et al., 2022).

With the above issues in mind, this study is designed to investigate the impact of 3D urban landscape patterns on PM_2.5 pollution in the largest metropolis in Central China – Wuhan. Wuhan has experienced rapid urbanization since 2000, resulting in serious PM_2.5 pollution (Fan et al., 2022; Xiong et al., 2017). Meanwhile, urban landscape patterns in Wuhan are complex with high spatial heterogeneity. Thus, it is an ideal place to carry out relevant research. Prior to the analysis, the spatiotemporal changes of PM_2.5 concentrations in Wuhan during 2000–2018 are first analyzed using recently developed high-resolution remote sensing data to provide essential background information. The relationships between 3D urban landscape patterns and PM_2.5 concentrations across different seasons corresponding to 2018 in the central urban area of Wuhan are then investigated using the MGWR model. The specific research objectives are to: (1) characterize the spatial patterns of PM_2.5 concentrations across different seasons and the 3D urban landscape patterns in central Wuhan; (2) explore the relationships between 3D urban landscape patterns and PM_2.5 concentrations in central Wuhan across different seasons; (3) analyze the seasonal and multiscale effects of the 3D urban landscape patterns on PM_2.5 pollution. This study will hopefully improve understanding of the relationships between 3D urban landscape patterns and PM_2.5 pollution, and provide valuable information on urban planning to mitigate PM_2.5 pollution in rapidly urbanized regions like Wuhan.

Study area

Wuhan, the capital of Hubei Province, is the largest city in Central China (113°41′–115°05′ E, 29°58′–31°22′ N; Figure 1(a)). It lies at the confluence of the Yangtze River and Han River, covering an area of 8569.15 km² and supporting a permanent population of 13.73 million in 2022 (NBSC, 2023). Wuhan has a subtropical monsoon climate and a flat terrain with low altitudes (generally less than 50 m). Summer (May, June, July, August, and September) is extremely hot and rainy, with daily maximum temperature frequently exceeding 35°C and accumulated rainfall between 600 and 1200 mm. Winter (December, January, and February) is cold and humid, with daily temperature range between 3°C and 12°C on average. Spring (March and April) is windy due to the strong air convection. Autumn (October and November) is sunny and cool with less rainfall (https://hb.cma.gov.cn/). With the rapid urbanization and industrialization, Wuhan has suffered PM_2.5 pollution with varying degrees in recent decades (Supplemental Figures S1–S3), which threatens the health of local residents (Bi et al., 2022).

Figure 1.

On the left: the location of Wuhan in China (a) and the administrative division of Wuhan (b). On the right: the spatial distribution of buildings categorized by the number of floors in central Wuhan (c). The red dot pinpoints the geographically weighted city center.

In response to the APPCAP, the Wuhan Municipal Ecology and Environment Bureau formulated a series of measures to abate PM_2.5 pollution in 2014 (https://hbj.wuhan.gov.cn/). These measures included motor vehicle emissions reduction, dust control and removal, and clean energy utilization. By 2018, the PM_2.5 pollution in Wuhan had been largely controlled, with the annual mean PM_2.5 concentrations (45.35 μg/m³) being 44% lower than those in 2013 (80.71 μg/m³; Supplemental Figures S1–S3). Therefore, we chose 2018 to investigate the relationships between 3D urban landscape patterns and PM_2.5 pollution. That is, we manage to reduce the influence of vehicle emissions, coal burning, and dust emissions on PM_2.5 pollution and focus particularly on the impact of 3D urban landscape patterns. Our study area is focused on the central urban area of Wuhan (Figure 1(b) and 1(c)). The central Wuhan is composed of three towns (Wuchang, Hankou, and Hanyang) and seven districts (Jiang’an, Jianghan, and Qiaokou in the town of Hankou; Wuchang, Hongshan, and Qingshan in the town of Wuchang; Hanyang as a whole), where a large number of buildings formed a complex 3D urban landscape pattern.

Materials and methods

Data acquisition and processing

The PM_2.5 data used in this study were collected from the China High Air Pollutants (CHAP) dataset (Wei et al., 2021; https://weijing-rs.github.io/product.html/). The dataset was produced based on the Space-Time Extra-Trees model to establish the relationship between PM_2.5 concentrations and aerosol optical depths. Auxiliary data were used to improve the prediction accuracy of the model, including meteorological data (e.g., air temperature, relative humidity, and surface pressure), multi-resolution emission inventory for China, vegetation index, land cover data, and population data. The dataset covered a period of 2000–2022, with a spatial resolution of 1 km × 1 km and a temporal resolution of days. The dataset captured well the spatial variability of PM_2.5 concentrations in China, with high prediction accuracy (CV-R² between 0.86 and 0.90) and strong predictive power (R² between 0.80 and 0.82). This dataset has been increasingly used in research focusing on public health, ecological quality, air pollution, and so forth (Feng et al., 2023; Jiang et al., 2023; Wang et al., 2023).

The 1:2000 building vector data covering central Wuhan in 2018 were produced by field surveys and provided by Wuhan Geomatics Institute (https://www.whkc.com/), with a position accuracy of 0.2 m. The data contained detailed information on building footprint, floor number, and building function (Figure 1(c)). According to the Chinese Design Code for residential buildings (GB50096-2011), we set the height of one building floor to 3 m so that the height of each building could be estimated. The building vector data were divided into 1 km × 1 km grids corresponding to the spatial resolution of the PM_2.5 concentrations data using the fishnet tool in ArcGIS 10.7. The 3D urban landscape metrics were then calculated in each grid cell. As our study was focused on central Wuhan, a total number of 588 grid cells (i.e., samples) were used for subsequent statistical analysis.

Calculation of 3D urban landscape metrics

We calculated eight 3D urban landscape metrics on a 1-km scale based on the 1:2000 building vector data to characterize the 3D urban landscape patterns in central Wuhan. The eight metrics contained three height metrics, two density metrics, two volume metrics, and one compound metric. Specifically, the height metrics included building height density (BHD), building height range (BHR), and standard deviation of building height (H_STD). The density metrics consisted of building count density (BCD) and building plan area fraction (BAF). The volume metrics comprised FAR and building evenness index (BEI). BEI represents the uniformity of the spatial distribution of buildings (Yu et al., 2021). The larger the BEI is, the lower the similarity of morphology between buildings is. The one compound metric referred to sky view factor (SVF) in particular. SVF measures the degree to which the sky is obscured by the surrounding obstacles at a given point, and estimation of SVF is based on the sum of the angle elements reaching the visible sky in the hemisphere (Grimmond et al., 2001). Here, we adopted the method proposed by Li et al. (2021) to calculate SVF by subtracting the local shadows in all directions of a given point. The detailed formulas to calculate the eight urban landscape metrics can be found in Supplemental Table S1.

Previous studies have shown that these metrics have a direct or indirect impact on the emissions and diffusion of atmospheric pollutants. BHD, H_STD, and BHR indicate the roughness length of urban underlying surfaces, which are closely related to turbulent exchange in the urban canopy layer and the diffusion of PM_2.5 (Cao et al., 2021; Xu and Chen, 2021). BCD and FAR are widely used to describe the intensity of human activities and reflect the level of PM_2.5 emissions (Lin et al., 2017; Yuan et al., 2019). In addition, buildings with dense distribution and inconsistent directions represented by high BAF and BEI could block air convection and thus the diffusion of PM_2.5 (Chen et al., 2021; Yu et al., 2021). SVF could also affect the diffusion of PM_2.5 through regulating incoming solar energy and ventilation in the urban canopy layer (Chen et al., 2021). On the one hand, emissions of air pollutants from buildings (e.g., residential, commercial, and industrial emissions) contribute to PM_2.5 pollution. On the other hand, the 3D urban landscape patterns can enhance or weaken the vertical and horizontal diffusion of PM_2.5 pollutants through modification of microclimate.

Spatial autocorrelation analysis

The global Moran’s I was calculated to determine the degree of spatial aggregation and dispersion of PM_2.5 concentrations in central Wuhan. The global Moran’s I ranges between −1 and 1. The closer the value is to 1, the higher the degree of aggregation is; conversely, the closer the value is to −1, the higher the degree of dispersion is. If the global Moran’s I is equal to 0, the spatial distribution of an object is deemed to be random (Li et al., 2018). Calculation of the global Moran’s I is as follows

I = \frac{n \sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j} \sum_{i = 1}^{n} (x_{i} - \bar{x}) (x_{j} - \bar{x})}{\sum_{i = 1}^{n} \sum_{j = 1}^{n} w_{i j} \sum_{i = 1}^{n} {(x_{j} - \bar{x})}^{2}}

(1)

where

I

is the global Moran’s I,

x_{i}

and

x_{j}

are the PM_2.5 concentrations in grids

i

and

j

\bar{x}

is the mean value of PM_2.5 concentrations,

W_{i j}

is the spatial weight matrix, and

n

is the total number of girds in the region of interest.

In addition to the global Moran’s I, the local Moran’s I was further calculated to identify spatial clusters of PM_2.5 with high or low values in central Wuhan. There are four types of spatial clustering patterns indicated by the local Moran’s I: high-high (HH), high-low (HL), low-high (LH), and low-low (LL). A significant positive value for the local Moran’s I indicates that a feature has neighboring features with similarly high or low feature values (HH or LL), while a significant negative value means that a feature has neighboring features with dissimilar feature values (HL or LH; Li et al., 2018). Calculation of the local Moran’s I is as follows

I_{i} = \frac{(x_{i} - \bar{x}) \sum_{j = 1 . j \neq i}^{n} W_{i j} (x_{j} - \bar{x})}{S_{i}^{2}}

(2)

S_{i}^{2} = \frac{\sum_{j = 1 . j \neq i}^{n} {(x_{j} - \bar{x})}^{2}}{n - 1} - {(\bar{x})}^{2}

(3)

In the equations (2) and (3),

I_{i}

is the local Moran’s I,

S_{i}^{2}

is the variance of the PM_2.5 concentrations, and the meanings of the other variables are the same as the equation (1).

Spatial regression analysis

We used the MGWR model to explore the spatial relationships between 3D urban landscape patterns and seasonal mean PM_2.5 concentrations in central Wuhan relative to 2018. To validate the effectiveness of the MGWR model, the GWR model was also adopted and compared with the MGWR model. The GWR model is developed for the analysis of spatial data in particular and allows the regression coefficients to vary locally by associating the explanatory variables with geographical locations (Hu et al., 2022). The GWR model is formulated as follows

y_{i} = a_{i o} (u_{i}, v_{i}) + \sum_{k = 1}^{8} a_{i k} (u_{i}, v_{i}) x_{i k} + ε_{i}

(4)

where

y_{i}

is the PM_2.5 concentration in grid

i

x_{i k}

is the kth urban landscape metric in grid

i

(u_{i}, v_{i})

is the spatial location of grid

i

a_{i o} (u_{i}, v_{i})

and

a_{i k} (u_{i}, v_{i})

are the intercept and local parameter estimate for

x_{i k}

, respectively, and

ε_{i}

is the error term.

The GWR model assumes that the spatial relationships between explanatory and response variables vary on the same spatial scale (i.e., all explanatory variables have the same bandwidth). However, previous studies have shown the multiscale effect of 3D urban landscape patterns on air pollution (Shi et al., 2019a; Zeng et al., 2022). To relax the assumption of GWR, we used the MGWR model proposed by Fotheringham et al. (2017) that allows the relationships between explanatory and response variables to vary on different scales (i.e., changeable bandwidths). Compared with GWR, the MGWR model considers the spatial heterogeneity of influencing factors and thereby reduces the collinearity and bias in parameter estimation (Hu et al., 2022). The MGWR model can well capture the local, regional, and global effects of explanatory variables, which is formulated as follows

y_{i} = \sum_{k = 1}^{8} β_{b w k} (u_{i}, v_{i}) x_{i k} + ε_{i} i

(5)

where

y_{i}

is the PM_2.5 concentration in grid

i

x_{i k}

is the kth urban landscape metric in grid

i

β_{b w k} (u_{i}, v_{i})

is the regression coefficient between the kth urban landscape metric and the PM_2.5 concentration in grid

i

(

b w k

represents the bandwidth, which is used to calibrate the kth conditional relationship), and

ε_{i}

is the error term.

The two regression models were run in the MGWR software version 2.2 developed by Oshan et al. (2019; https://sgsup.asu.edu/sparc/multiscale-gwr). The R² (determination coefficients), adjusted R², AIC (Akaike Information Criterion), and AICc (bias-corrected version of AIC) were used to evaluate the performance of the two models. The lower the AIC (and AICc) and the higher the R² (and adjusted R²) are, the better the model performance is.

Results

Spatial patterns of seasonal PM_2.5 concentrations

Prior to conducting the spatial regression analysis, we first show the spatial patterns of seasonal mean PM_2.5 concentrations in central Wuhan relative to 2018. As illustrated in Figure 2, the urban core area had the lowest PM_2.5 concentrations across all seasons, while the Qingshan District bore the highest PM_2.5 concentrations. Generally, the northern part of central Wuhan had higher PM_2.5 concentrations than the southern part. The exception was found in winter, when the concentrations of PM_2.5 exhibited a decreasing trend from west to east. Throughout the year, the PM_2.5 concentrations were the lowest in summer (≤38 μg/m³) and the highest in winter (73 μg/m³ on average), while the PM_2.5 concentrations in autumn and spring were generally between 42 and 60 μg/m³. According to the China Ambient Air Quality Standard, the air quality of central Wuhan in 2018 was fine but not good. Obviously, the PM_2.5 concentrations showed a spatial aggregated distribution (with the global Moran’s I between 0.89 and 0.94 at the 0.001 confidence level; Supplemental Table S2). The spatial aggregation patterns in spring and summer were similar (Supplemental Figure S4). That is, the urban core area along the Yangtze River was identified as a low-low clustering, while the Qingshan District and its surrounding areas showed a high-high clustering. In autumn and winter, the distribution of the spatial clusters was dispersed. The high-high clustering appeared on the western periphery of central Wuhan and, once again, the Qingshan District, while the low-low clustering was primarily found in the Wuchang and Hongshan Districts. A notable phenomenon was that the low-low clustering in winter was all around the inland lakes and rivers.

Figure 2.

Spatial patterns of seasonal PM_2.5 concentrations in central Wuhan corresponding to 2018: (a) spring, (b) summer, (c) autumn, and (d) winter.

Spatial patterns of 3D urban landscape metrics

We next show the 3D urban landscape patterns in central Wuhan characterized by building height, density, and volume features (Figure 3). BHD, BAF, FAR, as well as BEI along the two sides of the Yangtze River were generally larger than other locations. Across the three towns, the four feature values were the highest in the town of Hankou. The variations of building height (i.e., H_STD and BHR) in the towns of Hankou and Hanyang were larger than in the town of Wuchang on the whole. By contrast, the spatial distribution of BCD was relatively homogenous. For the seven districts in central Wuhan, the Qingshan District had the lowest height and volume features but the highest density features. This is because the Qingshan District is an industrial zone possessing a large number of low-rise buildings with large building footprints. The height, density, and volume of buildings in the Hongshan District were the minimum across central Wuhan. To summarize, the town of Hankou had a high level of development with the spatial differences in building features being small, whereas the development in the town of Wuchang was imbalanced. Finally, the landscape openness represented by SVF on the east bank of the Yangtze River was lower than on the west bank. The Qingshan District, the locations around the East Lake, and the southern part of the Hongshan District had the largest SVF.

Figure 3.

Spatial patterns of (a) building height density, (b) building height range, (c) standard deviation of building height, (d) building count density, (e) building plan area fraction, (f) floor area ratio, (g) building evenness index, and (h) sky view factor in central Wuhan corresponding to 2018.

Relationships between PM_2.5 concentrations and 3D urban landscape patterns

As the PM_2.5 concentrations and 3D urban landscape patterns show significant spatial heterogeneity, we chose local regression models (i.e., GWR and MGWR) instead of global regression models to investigate the relationships between them. We first compared the performance of GWR and MGWR in different seasons. The results showed that MGWR performed better than GWR, with higher R² and lower AIC across all seasons (Supplemental Table S3). In comparison, the performance of the MGWR model in spring was the best (R² = 0.82 and AIC = 866.46), while that in winter was the worst (R² = 0.67 and AIC = 1288.11). Herein, the MGWR model was utilized to explore the spatial relationships between 3D urban landscape patterns and seasonal mean PM_2.5 concentrations relative to 2018 in central Wuhan. As shown in Figure 4, the local R² of the MGWR model had higher values and spatial variability in spring and summer (0.63 and 0.62 on average) than in autumn and winter (0.54 and 0.52 on average). Moreover, the spatial variability of the local R² in spring and summer showed similar patterns. Except the winter season, the locales around the East Lake had the maximum local R² (0.8–0.9 in spring and summer and 0.7–0.8 in autumn), indicating that the 3D urban landscape patterns exerted a strong influence on the PM_2.5 concentrations here. By contrast, the locales in the southern part of the Hongshan District had the minimum local R² (less than 0.4 in general) in all seasons, especially spring and summer. Overall, the magnitude of the local R² demonstrated the importance of 3D urban landscape patterns on PM_2.5 concentrations in central Wuhan.

Figure 4.

The local R² of MGWR between seasonal PM_2.5 concentrations and urban landscape metrics in central Wuhan corresponding to 2018: (a) spring, (b) summer, (c) autumn, and (d) winter.

Compared with the fixed bandwidth in GWR (125 for spring, 127 for summer, 110 for autumn, and 85 for winter), the bandwidths of the regression coefficients in MGWR varied between different metrics (Table 1). For example, the bandwidths of BHD (≥578) and BEI (≥586) were the largest, accounting for more than 98.3% and 99.7% of the samples, respectively. This meant that the influences of BHD and BEI on PM_2.5 concentrations were global with low spatial variability. By contrast, the bandwidth of SVF was the minimum (<70), indicating that the influence of SVF on PM_2.5 concentrations was local with large spatial variability. The bandwidths also varied across different seasons. For example, the bandwidths of BCD and BAF showed opposite seasonal characteristics. For BCD, the bandwidths in spring and summer (256 and 110) were larger than those in autumn and winter (71), while for BAF, the bandwidths in spring and summer (46 and 105) were much smaller than those in autumn and winter (318 and 363). In addition, BHR and FAR had a local or regional effect on PM_2.5 concentrations in a season but a global effect in the other seasons, while H_STD had global effect in a season but a local or regional effect in the other seasons.

Table 1.

Bandwidths of the MGWR model for each urban landscape metric.

	Spring	Summer	Autumn	Winter
BHD	587	578	587	587
BHR	587	532	122	587
H_STD	234	289	562	78
BCD	256	110	71	71
BAF	46	105	318	363
FAR	585	459	587	98
BEI	587	586	587	587
SVF	54	47	70	44

Finally, we show the spatial correlations between PM_2.5 concentrations and 3D urban landscape metrics across different seasons in central Wuhan. In spring, BHR had a significant negative correlation with PM_2.5 concentrations, while H_STD had a significant positive correlation with PM_2.5 concentrations (Figure 5). The correlation between H_STD and PM_2.5 concentrations was the strongest in the southern part of central Wuhan (the town of Hanyang and the southern Hongshan District). For the density metrics, a positive correlation between BCD and PM_2.5 concentrations was found across the entire region, and the correlation was significant primarily in the northern part of the region, especially the Qingshan District. The correlation between BAF and PM_2.5 concentrations was location-dependent with a small effect scale. In the urban core area around the city center, it was a significant negative correlation; in the Qingshan District to the north and the Hongshan District to the south, it was a significant positive correlation. The volume metrics (FAR and BEI) were negatively correlated with PM_2.5 concentrations, but the correlations were extremely weak and insignificant. SVF exhibited a significant positive correlation with PM_2.5 concentrations in the urban core area and the area in and around the Qingshan District, and the effect scale of SVF was close to BAF. Among these metrics, the absolute regression coefficients of BAF and SVF were the highest (up to 0.52), while those of the other metrics were less than 0.35 in spring. The spatial correlations between 3D urban landscape metrics and PM_2.5 concentrations in summer were similar with those in spring (BEI was excluded from analysis because of the minimal regression coefficients across all seasons; Figure 6). However, the strength of the correlations in summer was generally stronger than that in spring (e.g., local regression coefficients up to 0.69 for SVF and 0.52 for BCD).

Figure 5.

The local regression coefficients of MGWR between PM_2.5 concentrations and urban landscape metrics in central Wuhan in the spring of 2018: (a) building height density, (b) building height range, (c) standard deviation of building height, (d) building count density, (e) building plan area fraction, (f) floor area ratio, (g) building evenness index, and (h) sky view factor.

Figure 6.

The local regression coefficients of MGWR between PM_2.5 concentrations and urban landscape metrics in central Wuhan in the summer of 2018: (a) building height density, (b) building height range, (c) standard deviation of building height, (d) building count density, (e) building plan area fraction, (f) floor area ratio, (g) building evenness index, and (h) sky view factor.

In autumn, FAR, for the first time of the year, showed a significant negative correlation with PM_2.5 concentrations in all locales, and the regression coefficients equal to 0.35 on average (Figure 7). Except for FAR, H_STD and BAF also showed much stronger correlations with PM_2.5 concentrations. The regression coefficients of H_STD became higher in all locales (0.31–0.35), and BAF had a significant positive correlation with PM_2.5 concentrations across almost the entire region. Notably, the maximum regression coefficient for SVF reached 0.79 in autumn, especially in the Qingshan District. BHR had a smaller effect scale on PM_2.5 concentrations in autumn, and the correlation was the highest among all seasons (up to −0.44). Once again, the correlation between BCD and PM_2.5 concentrations was the strongest in the Qingshan District. In winter, the influence of H_STD, BAF, and FAR continued to enhance, with the regression coefficients up to 1.12, 0.44, and −0.69, respectively (Figure 8). In addition, the effect scales of H_STD and FAR became much smaller compared with the other seasons, with the regression coefficients showing greater spatial variability. Meanwhile, the correlation between BHR and PM_2.5 concentrations became weak (−0.14 on average) and was significant merely in the southern part. In terms of BCD and SVF, the correlations between them and PM_2.5 concentrations in winter were complex. For BCD, a significant negative correlation was found (especially in the town of Hanyang and the southern Hongshan District), with the maximum regression coefficient equal to −0.68. For SVF, the positive coefficients in the Qingshan District reached 0.97, while the negative coefficients in other locales were up to −0.89.

Figure 7.

The local regression coefficients of MGWR between PM_2.5 concentrations and urban landscape metrics in central Wuhan in the autumn of 2018: (a) building height density, (b) building height range, (c) standard deviation of building height, (d) building count density, (e) building plan area fraction, (f) floor area ratio, (g) building evenness index, and (h) sky view factor.

Figure 8.

The local regression coefficients of MGWR between PM_2.5 concentrations and urban landscape metrics in central Wuhan in the winter of 2018: (a) building height density, (b) building height range, (c) standard deviation of building height, (d) building count density, (e) building plan area fraction, (f) floor area ratio, (g) building evenness index, and (h) sky view factor.

Discussion

Both global and local autocorrelation analyses show that the seasonal mean PM_2.5 concentrations in central Wuhan had significant spatial variability. Generally, the PM_2.5 concentrations in the northern part of central Wuhan were higher than the southern part. The industrial emissions from the Qingshan District and external sources of air pollutants invading from the northern provinces (Hebei, Shandong, and Henan) contributed to the PM_2.5 pollution in the northern part (Zhao et al., 2023). Meanwhile, the city center acted as a buffer zone to prevent the spread of PM_2.5 pollutants from the north to the south to some degree (Yu, 2023). However, the spatial pattern of PM_2.5 pollution in winter was different from the other seasons. The meteorological conditions (e.g., atmospheric inversion) in winter resulted in a weak impact of 3D urban landscape patterns on the diffusion of PM_2.5 pollution through modification of microclimate (Cao et al., 2021). In addition, the combustion of biomass fuel and insufficient combustion of vehicle emissions aggravated PM_2.5 pollution in winter. The exception was found around lakes where the PM_2.5 concentrations were lower than in other places. The wet deposition of PM_2.5 pollutants by the two large water bodies, that is, the East Lake and Tangxun Lake, significantly reduced the PM_2.5 pollutants in the surrounding areas (Zhu and Zhou, 2019). Therefore, a decrease in PM_2.5 concentrations from the west to the east was found in winter.

As global regression models may have significant estimated deviation (Bi et al., 2022), we used two local regression models (GWR and MGWR) to explore the relationships between 3D urban landscape patterns and PM_2.5 pollution in central Wuhan. Compared with global regression models (e.g., ordinary least squares, spatial lag model, and spatial error model), GWR and MGWR considered geospatial heterogeneity and were more accurate in exploring the spatial relationships between explanatory and response variables (Supplemental Table S4). By contrast, the MGWR model produced higher explanatory power and accuracy than the GWR model due to the fact that spatial processes may operate at different scales in the real world (Fotheringham et al., 2017). The local R² indicated that the 3D urban landscape patterns had lesser impacts on PM_2.5 pollution in autumn and winter than in spring and summer. The results were consistent with previous research focused on the impact of urban form and size on PM_2.5 pollution across China (Liu et al., 2018). Wu et al. (2015) also reported that urban landscape patterns had larger explanatory power on PM_2.5 pollution in spring and summer than in autumn and winter in Beijing. The strong air convection and high temperature in spring and summer may strengthen the impact of urban landscape patterns on PM_2.5 pollution through modification of microclimate, whereas the temperature inversion and increment of PM_2.5 emissions may weaken such impact in autumn and spring. Further, we found that the 3D urban landscape metrics had the strongest association with PM_2.5 pollution around the East Lake (except in winter). The lake breeze circulation triggered by the lake-land temperature difference may strengthen the ability of urban landscape patterns to modify microclimate and thus the diffusion of PM_2.5 pollutants.

The varying bandwidths of the MGWR model captured the scale effects of 3D urban landscape patterns on PM_2.5 pollution across different seasons. For example, BCD was closely related to the PM_2.5 emissions (Lin et al., 2017). The PM_2.5 emissions in autumn and winter were high with a relatively larger spatial heterogeneity in central Wuhan, which led to a localized relationship between BCD and PM_2.5 pollution (i.e., smaller bandwidths). FAR was also closely related to the emissions of PM_2.5 (Yuan et al., 2019). During the extreme cold and high PM_2.5 pollution period, FAR also strongly modified airflow and thus the diffusion of PM_2.5 (Jung and Yoon, 2021). Hence, a localized effect of FAR on PM_2.5 pollution was found especially in winter. Conversely, BAF had larger bandwidths in autumn and winter than in spring and summer. BAF was closely related to ventilation and a densely built environment could block air convection and hinder the diffusion of PM_2.5 (Chen et al., 2021). Spring had the highest wind speed in Wuhan, followed by summer, while autumn and winter had relatively low wind speed (Yang et al., 2019). The magnitude of wind speed corresponded well to the bandwidth of BAF across the four seasons. That is, high wind speed with large spatial variability led to a localized effect of BAF on PM_2.5, whereas low wind speed with small spatial variability led to a globalized effect of BAF on PM_2.5. We also found that the effect of BHD on PM_2.5 was global all year round, although the correlation between them was not significant. Meanwhile, Liu et al. (2019) showed that BHD had a global effect on land surface temperature in central Wuhan when using MGWR. Essentially, this was due to the small spatial variability of BHD compared with other explicit urban landscape metrics (e.g., BCD, BAF, and FAR) in the study area.

It was also shown that the 3D urban landscape patterns had different relationships with seasonal PM_2.5 concentrations in central Wuhan. For the height metrics, H_STD had a positive correlation with PM_2.5 concentrations. Liu et al. (2021) found that H_STD more than 10 m led to a higher concentration and stagnation of air pollutants. BHR showed a negative correlation with PM_2.5 concentrations. However, Cao et al. (2021) argued that the increase of BHR would reduce wind speed due to increased surface roughness and consequently accumulate PM_2.5. In fact, BHR may not well characterize the variation of building height on a large spatial scale. For the density metrics, BAF had a complex relationship with PM_2.5 concentrations. On the one hand, the densely built environment can raise air temperature in the street valley through radiation trapping, thus increasing the aerodynamic energy of atmospheric turbulence and promoting the diffusion of PM_2.5 (Xu and Chen, 2021). On the other hand, it can also increase ground friction and reduce wind speed, thus hindering the diffusion of PM_2.5 (Chen et al., 2021). The effect of BAF on PM_2.5 concentrations could be positive and negative, dependent on which process is dominant. We show that BAF had a significant positive correlation with PM_2.5 concentrations in general, indicating that the effect of BAF on ground friction might dominate. However, a negative correlation was found in the urban core area around the city center in spring and summer, indicating that the thermodynamic process might dominate. BCD generally had a positive correlation with PM_2.5 concentrations because a higher BCD usually indicated higher PM_2.5 emissions (e.g., more traffic loads and vehicle emissions; Lin et al., 2017). Notably, BCD and BAF showed a strong positive correlation with PM_2.5 concentrations in the Qingshan District owing to the industrial emissions generated by a large number of factories.

In addition to the height and density metrics, we also examined some more complex metrics such as FAR and SVF. FAR is determined by both building height and density, and it is an important indicator of urban capacity. Usually, a higher FAR may result in more anthropogenic emissions and thus a higher level of PM_2.5 pollution (Yuan et al., 2019). However, we found that FAR had a significant negative correlation with PM_2.5 concentrations in autumn and winter across central Wuhan. Due to the effective measures taken by the government to abate PM_2.5 pollution, the urban core area around the city center had lower PM_2.5 concentrations than its surrounding areas in 2018. However, FAR along the two sides of the Yangtze River was generally larger than other locales. In addition, the Qingshan District owned the highest PM_2.5 and the lowest FAR due to the large number of low-rise industrial buildings. From another perspective, a higher FAR can lead to a deeper urban canyon and accelerate the airflow velocity (Luo et al., 2023). Similarly, Jung and Yoon (2021), based on microscale simulation, also showed that FAR had a negative correlation with PM_2.5 concentrations during the period of extreme cold and high PM_2.5 pollution. SVF is a more complex metric measuring the openness of urban space and exerts a strong influence on microclimate through modification of incoming solar radiation and ventilation (Li et al., 2021). A high SVF will increase energy absorption while enhancing ventilation, and vice versa. Previous research frequently showed that SVF was negatively correlated with PM_2.5 concentrations (Chen et al., 2021; Shi et al., 2018; Silva and Monteiro, 2016). In this study, we found that SVF had a significant positive correlation with PM_2.5 concentrations in locales with low SVF (e.g., the Qingshan District). This was mainly due to industrial emissions from many low-rise industrial plants rather than the modification of microclimate.

Research on the effect of urban landscape patterns on PM_2.5 pollution helps urban planning to improve air quality in urban environments. Our study makes a step forward, but there are limitations. First, the relationships between 3D urban landscape patterns and PM_2.5 pollution may vary across different scales. Ke et al. (2022) found that BHD had a significant negative correlation with PM_2.5 concentrations on a 500-m scale, while an insignificant negative correlation was found in this study on a 1-km scale. With advances in remote sensing and mobile monitoring techniques, future research should understand the effect of 3D urban landscape patterns on PM_2.5 concentrations on different scales to facilitate cross-scale urban planning. Second, vegetation and water also play an important role in regulating PM_2.5 pollution. Vegetation can hinder the dispersion of PM_2.5 through modification of air flow and promote the deposition of PM_2.5 through adsorption of leaves (Łowicki, 2019). Water bodies can lower ambient temperature and increase humidity owing to the high heat storage and evaporation, which promote the wet deposition of particulate matters and inhibit the secondary formation of precursors to particulate matters (Lou et al., 2017; Zhou et al., 2021). Third, the mechanisms for the impact of 3D urban landscape patterns on PM_2.5 pollution need further exploration. Recently, mesoscale models have been used to explore such impact (Xu and Chen, 2021). However, advanced mesoscale models (e.g., WRF) only allow incorporation of limited urban canopy parameters (Deng et al., 2023). Some parameters exerting a strong influence on atmospheric environments have not been but should be considered in future model development. Finally, although the MGWR model reduces over fitting and under fitting of the GWR model by allowing the regression coefficients to vary on local, regional, and global scales, the problems may exist and should not be ignored (Su et al., 2022).

Conclusion

This study explored the relationships between 3D urban landscape patterns and PM_2.5 concentrations across different seasons in central Wuhan using the MGWR model. The effect of 3D urban landscape patterns on PM_2.5 pollution varied in different seasons in terms of the direction and degree of the effect: (1) the increase of H_STD significantly aggravated PM_2.5 pollution through the year; (2) BAF had a positive correlation with PM_2.5 concentrations in general except that a negative correlation was found in the urban core area in spring and summer due to the thermodynamic process; (3) BCD showed a positive correlation with PM_2.5 concentrations especially in spring and summer; (4) the increase of FAR significantly reduced PM_2.5 concentrations in autumn and winter; (5) SVF showed a significant positive effect on PM_2.5 concentrations in the urban core area and the industrial zone. In addition, the effect scales of 3D urban landscape patterns also varied in different seasons: (1) SVF had the smallest bandwidths, indicating a localized effect on PM_2.5 pollution; (2) the bandwidths of BCD in spring and summer were larger than those in autumn and winter, while the reverse was true for BAF; (3) FAR had a localized effect in winter while a global effect in the other seasons. To alleviate PM_2.5 pollution, we suggest limiting the building count and reducing the building height variation to render a compact urban landscape pattern formed by high-rise buildings. Meanwhile, urban planners and policymakers should pay special attention to the PM_2.5 pollution in the norther part of central Wuhan especially the Qingshan District, and take effective measures such as construction of public transport facilities, utility of clean energy, upgradation of emissions standards, and so forth.

Supplemental Material

Supplemental Material - Exploring the relationships between 3D urban landscape patterns and PM_2.5 pollution using the multiscale geographic weighted regression model

Supplemental Material for Exploring the relationships between 3D urban landscape patterns and PM_2.5 pollution using the multiscale geographic weighted regression model by Haoyan Duan, Qian Cao, Lunche Wang, Xihui Gu and Khosro Ashrafi in Progress in Physical Geography: Earth and Environment

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Natural Science Foundation of China (Grant numbers 42371115, 42371354) and Natural Science Foundation of Hubei Province (2022CFB039).

ORCID iDs

Qian Cao

Lunche Wang

Supplemental Material

Supplemental material for this article is available online.

References

Dai

Chen

, et al. (2022) A new framework for analysis of the morphological spatial patterns of urban green space to reduce PM_2.5 pollution: a case study in Wuhan, China. Sustainable Cities and Society 82: 103900. DOI: 10.1016/j.scs.2022.103900

Cao

Luan

Liu

, et al. (2021) The effects of 2D and 3D building morphology on urban environments: a multi-scale analysis in the Beijing metropolitan region. Building and Environment 192: 107635. DOI: 10.1016/j.buildenv.2021.107635

Chen

Bai

Zhu

, et al. (2021) The influence of neighborhood-level urban morphology on PM_2.5 variation based on random forest regression. Atmospheric Pollution Research 12(8): 101147. DOI: 10.1016/j.apr.2021.101147

Deng

Cao

Wang

, et al. (2023) Characterizing urban densification and quantifying its effects on urban thermal environments and human thermal comfort. Landscape and Urban Planning 237: 104803. DOI: 10.1016/j.landurbplan.2023.104803

Fan

Zhan

Liu

, et al. (2022) Investigating the interactive and heterogeneous effects of green and blue space on urban PM_2.5 concentration, a case study of Wuhan. Journal of Cleaner Production 378: 134389. DOI: 10.1016/j.jclepro.2022.134389

Feng

Wang

Yuan

, et al. (2023) Changes in air pollution, land surface temperature, and urban heat islands during the COVID-19 lockdown in three Chinese urban agglomerations. The Science of the Total Environment 892: 164496. DOI: 10.1016/j.scitotenv.2023.164496

Fotheringham

Yang

Kang

(2017) Multiscale geographically weighted regression (MGWR). Annals of the Association of American Geographers 107(6): 1247–1265. DOI: 10.1080/24694452.2017.1352480

Grimmond

Potter

Zutter

, et al. (2001) Rapid methods to estimate sky-view factors applied to urban areas. International Journal of Climatology 21(7): 903–913. DOI: 10.1002/joc.659

Hou

Chen

, et al. (2020) Associations of long-term exposure to ambient fine particulate matter and nitrogen dioxide with lung function: a cross-sectional study in China. Environment International 144: 105977. DOI: 10.1016/j.envint.2020.105977

10.

Zhang

(2022) Exploring the spatial and temporal driving mechanisms of landscape patterns on habitat quality in a city undergoing rapid urbanization based on GTWR and MGWR: the case of Nanjing, China. Ecological Indicators 143: 109333. DOI: 10.1016/j.ecolind.2022.109333

11.

Jiang

Chen

Dai

, et al. (2023) Change of the spatial and temporal pattern of ecological vulnerability: a case study on Cheng-Yu urban agglomeration, Southwest China. Ecological Indicators 149: 110161. DOI: 10.1016/j.ecolind.2023.110161

12.

Jung

Yoon

(2021) Analysis of the effects of floor area ratio change in urban street canyons on microclimate and particulate matter. Energies 14(3): 714. DOI: 10.3390/en14030714

13.

Huang

, et al. (2022) Three-dimensional building morphology impacts on PM_2.5 distribution in urban landscape settings in Zhejiang, China. The Science of the Total Environment 826: 154094. DOI: 10.1016/j.scitotenv.2022.154094

14.

Zhang

(2018) Exploring the spatial spillover effects of industrialization and urbanization factors on pollutants emissions in China’s Huang-Huai-Hai region. Journal of Cleaner Production 195: 154–162. DOI: 10.1016/j.jclepro.2018.05.186

15.

Wang

, et al. (2021) Quantifying 3D building form effects on urban land surface temperature and modeling seasonal correlation patterns. Building and Environment 204: 108132. DOI: 10.1016/j.buildenv.2021.108132

16.

Han

Zhou

, et al. (2023) Uncertainties in research between urban landscape and air quality: summary, demonstration, and expectation. Landscape Ecology 38(10): 2475–2485. DOI: 10.1007/s10980-023-01744-5

17.

Lin

Chiang

Y-Y

Pan

, et al. (2017) Mining public datasets for modeling intra-city PM_2.5 concentrations at a fine spatial resolution. Proc ACM SIGSPATIAL Int Conf Adv Inf 2017: 25. DOI: 10.1145/3139958.3140013

18.

Liu

(2018) Disentangling the complex effects of socioeconomic, climatic, and urban form factors on air pollution: a case study of China. Sustainability 10(3): 776. DOI: 10.3390/su10030776

19.

Liu

Zhan

Gao

, et al. (2019) Seasonal variation of the spatially non-stationary association between land surface temperature and urban landscape. Remote Sensing 11(9): 1016. DOI: 10.3390/rs11091016

20.

Liu

Chen

Wei

, et al. (2021) Nonlinear relationship between urban form and street-level PM_2.5 and CO based on mobile measurements and gradient boosting decision tree models. Building and Environment 205: 108265. DOI: 10.1016/j.buildenv.2021.108265

21.

Liu

Hua

Song

(2022) PM_2.5 exposure and asthma development: the key role of oxidative stress. Oxidative Medicine and Cellular Longevity 2022: 3618806. DOI: 10.1155/2022/3618806

22.

Lou

Liu

, et al. (2017) Relationships of relative humidity with PM _2.5 and PM₁₀ in the Yangtze River delta, China. Environmental Monitoring and Assessment 189: 582–616. DOI: 10.1007/s10661-017-6281-z

23.

Łowicki

(2019) Landscape pattern as an indicator of urban air pollution of particulate matter in Poland. Ecological Indicators 97: 17–24. DOI: 10.1016/j.ecolind.2018.09.050

24.

Zhang

Xing

, et al. (2020) Progress of air pollution control in China and its challenges and opportunities in the ecological civilization era. Engineering 6(12): 1423–1431. DOI: 10.1016/j.eng.2020.03.014

25.

Luo

, et al. (2023) How 2D and 3D built environments impact urban surface temperature under extreme heat: a study in Chengdu, China. Building and Environment 231: 110035. DOI: 10.1016/j.buildenv.2023.110035

26.

Mansour

Al Kindi

Al-Said

, et al. (2021) Sociodemographic determinants of COVID-19 incidence rates in Oman: geospatial modelling using multiscale geographically weighted regression (MGWR). Sustainable Cities and Society 65: 102627. DOI: 10.1016/j.scs.2020.102627

27.

Mao

Wang

Pan

, et al. (2022) A study on the dynamic spatial spillover effect of urban form on PM_2.5 concentration at county scale in China. Atmospheric Research 269: 106046. DOI: 10.1016/j.atmosres.2022.106046

28.

Mollalo

Vahedi

Rivera

(2020) GIS-based spatial modeling of COVID-19 incidence rate in the continental United States. The Science of the Total Environment 728: 138884. DOI: 10.1016/j.scitotenv.2020.138884

29.

NBSC (2023) National Bureau of Statistics of China: China Statistical Yearbook 2023. Beijing: China Statistics Press.

30.

Oshan

Kang

, et al. (2019) mgwr: a Python implementation of multiscale geographically weighted regression for investigating process spatial heterogeneity and scale. ISPRS International Journal of Geo-Information 8(6): 269. DOI: 10.3390/ijgi8060269

31.

Park

S-H

Yoon

S-J

Choi

, et al. (2022) Particulate matter promotes cancer metastasis through increased HBEGF expression in macrophages. Experimental & Molecular Medicine 54(11): 1901–1912. DOI: 10.1038/s12276-022-00886-x

32.

Rong

Guo

, et al. (2022) Multi-scale spatio-temporal analysis of soil conservation service based on MGWR model: a case of Beijing-Tianjin-Hebei, China. Ecological Indicators 139: 108946. DOI: 10.1016/j.ecolind.2022.108946

33.

Shi

Xie

Fung

JC-H

, et al. (2018) Identifying critical building morphological design factors of street-level air pollution dispersion in high-density built environment using mobile monitoring. Building and Environment 128: 248–259. DOI: 10.1016/j.buildenv.2017.11.043

34.

Shi

Wang

Yang

, et al. (2019a) Exploring the relationships between urban forms and fine particulate (PM_2.5) concentration in China: a multi-perspective study. Journal of Cleaner Production 231: 990–1004. DOI: 10.1016/j.jclepro.2019.05.317

35.

Shi

, et al. (2019b) Investigating the influence of urban land use and landscape pattern on PM_2.5 spatial variation using mobile monitoring and WUDAPT. BMC Gastroenterology 19: 15–26. DOI: 10.1016/j.landurbplan.2019.04.004

36.

Silva

Monteiro

(2016) The influence of urban form on environmental quality within a medium-sized city. Procedia Engineering 161: 2046–2052. DOI: 10.1016/j.proeng.2016.08.801

37.

Xiao

Zhang

(2012) Multi-scale analysis of spatially varying relationships between agricultural landscape patterns and urbanization using geographically weighted regression. Applied Geography 32(2): 360–375. DOI: 10.1016/j.apgeog.2011.06.005

38.

Chen

Tan

, et al. (2022) Estimating regional PM_2.5 concentrations in China using a global-local regression model considering global spatial autocorrelation and local spatial heterogeneity. Remote Sensing 14(18): 4545. DOI: 10.3390/rs14184545

39.

Sun

Zhou

Wang

(2022) Relationships between urban form and air quality: a reconsideration based on evidence from China’s five urban agglomerations during the COVID-19 pandemic. Land Use Policy 118: 106155. DOI: 10.1016/j.landusepol.2022.106155

40.

Tan

Zhou

Wang

(2023) Relationships between urban form and PM_2.5 concentrations from the spatial pattern and process perspective. Building and Environment 234: 110147. DOI: 10.1016/j.buildenv.2023.110147

41.

Ulpiani

(2021) On the linkage between urban heat island and urban pollution island: three-decade literature review towards a conceptual framework. The Science of the Total Environment 751: 141727. DOI: 10.1016/j.scitotenv.2020.141727

42.

Wang

Niu

Fan

, et al. (2022a) Urban configuration and PM_2.5 concentrations: evidence from 330 Chinese cities. Environment International 161: 107129. DOI: 10.1016/j.envint.2022.107129

43.

Wang

Cheng

Dai

(2022b) Strategies for sustainable urban development and morphological optimization of street canyons: measurement and simulation of PM_2.5 at different points and heights. Sustainable Cities and Society 87: 104191. DOI: 10.1016/j.scs.2022.104191

44.

Wang

Zhang

, et al. (2022c) Investigating the spatiotemporal pattern of urban vibrancy and its determinants: spatial big data analyses in Beijing, China. Land Use Policy 119: 106162. DOI: 10.1016/j.landusepol.2022.106162

45.

Wang

Shen

Yan

, et al. (2023) Individual exposure of ambient particulate matters and eosinophilic chronic rhinosinusitis with nasal polyps: dose-response, mediation effects and recurrence prediction. Environment International 177: 108031. DOI: 10.1016/j.envint.2023.108031

46.

Wei

Lyapustin

, et al. (2021) Reconstructing 1-km-resolution high-quality PM_2.5 data records from 2000 to 2018 in China: spatiotemporal variations and policy implications. Remote Sensing of Environment 252: 112136. DOI: 10.1016/j.rse.2020.112136

47.

Xie

, et al. (2015) Effects of urban landscape pattern on PM_2.5 pollution—a Beijing case study. PLoS One 10(11): e0142449. DOI: 10.1371/journal.pone.0142449

48.

Xiong

Zhou

Schauer

, et al. (2017) Seasonal and spatial differences in source contributions to PM_2.5 in Wuhan, China. Science of The Total Environment 577: 155–165. DOI: 10.1016/j.scitotenv.2016.10.150

49.

Chen

(2021) Impact of urban morphology on the spatial and temporal distribution of PM_2.5 concentration: a numerical simulation with WRF/CMAQ model in Wuhan, China. Journal of Environmental Management 290: 112427. DOI: 10.1016/j.jenvman.2021.112427

50.

Yang

Cheng

Wang

, et al. (2019) Chemical characteristics of atmospheric carbonyl compounds and source identification of formaldehyde in Wuhan, Central China. Atmospheric Research 228: 95–106. DOI: 10.1016/j.atmosres.2019.05.020

51.

Yang

Leng

Xiao

, et al. (2022) Investigating the impact of urban landscape composition and configuration on PM_2.5 concentration under the LCZ scheme: a case study in Nanchang, China. Sustainable Cities and Society 84: 104006. DOI: 10.1016/j.scs.2022.104006

52.

(2023) Correlation analysis of urban building form and PM_2.5 pollution based on satellite and ground observations. Frontiers in Environmental Science 10: 1111223. DOI: 10.3389/fenvs.2022.1111223

53.

Liu

Zhang

, et al. (2021) Influences of buildings on urban heat island based on 3D landscape metrics: an investigation of China’s 30 megacities at micro grid-cell scale and macro city scale. Landscape Ecology 36(9): 2743–2762. DOI: 10.1007/s10980-021-01275-x

54.

Yuan

Song

Huang

, et al. (2019) Exploring the association between the built environment and remotely sensed PM_2.5 concentrations in urban areas. Journal of Cleaner Production 220: 1014–1023. DOI: 10.1016/j.jclepro.2019.02.236

55.

Yue

Huang

, et al. (2020) Stronger policy required to substantially reduce deaths from PM_2.5 pollution in China. Nature Communications 11(1): 1462. DOI: 10.1038/s41467-020-15319-4

56.

Zeng

Hang

Wang

, et al. (2022) Influence of urban spatial and socioeconomic parameters on PM_2.5 at subdistrict level: a land use regression study in Shenzhen, China. Journal of Environmental Sciences 114: 485–502. DOI: 10.1016/j.jes.2021.12.002

57.

Zhao

Zhou

, et al. (2022) The impacts of urban structure on PM_2.5 pollution depend on city size and location. Environmental Pollution 292: 118302. DOI: 10.1016/j.envpol.2021.118302

58.

Zhao

Zhang

Cheng

, et al. (2023) Investigate the effects of urban land use on PM2.5 concentration: an application of deep learning simulation. Building and Environment 242: 110521. DOI: 10.1016/j.buildenv.2023.110521

59.

Zhou

Zhang

Zhu

(2021) Impact of urban water networks on microclimate and PM_2.5 distribution in downtown areas: a case study of Wuhan. Building and Environment 203: 108073. DOI: 10.1016/j.buildenv.2021.108073

60.

Zhu

Zhou

(2019) Effect of urban water bodies on distribution characteristics of particulate matters and NO₂. Sustainable Cities and Society 50: 101679. DOI: 10.1016/j.scs.2019.101679

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.38 MB

Exploring the relationships between 3D urban landscape patterns and PM 2.5 pollution using the multiscale geographic weighted regression model

Abstract

Keywords

Introduction

Study area

Materials and methods

Data acquisition and processing

Calculation of 3D urban landscape metrics

Spatial autocorrelation analysis

Spatial regression analysis

Results

Spatial patterns of seasonal PM2.5 concentrations

Spatial patterns of 3D urban landscape metrics

Relationships between PM2.5 concentrations and 3D urban landscape patterns

Discussion

Conclusion

Supplemental Material

Supplemental Material - Exploring the relationships between 3D urban landscape patterns and PM2.5 pollution using the multiscale geographic weighted regression model

Footnotes

Declaration of conflicting interests

Funding

ORCID iDs

Supplemental Material

References

Supplementary Material

Spatial patterns of seasonal PM_2.5 concentrations

Relationships between PM_2.5 concentrations and 3D urban landscape patterns

Supplemental Material - Exploring the relationships between 3D urban landscape patterns and PM_2.5 pollution using the multiscale geographic weighted regression model