PM_2.5 Sink-Source Dichotomy in Urban Clusters: Land Cover Efficiency Gradients and Socio-Meteorological Interactions

Abstract

Urban clusters face escalating atmospheric challenges, with elevated PM_2.5 concentrations representing a critical environmental constraint. This study investigates spatiotemporal evolution mechanisms of PM_2.5 within China’s Taiyuan-Yuci-Xinzhou (TYX) urban cluster. Daily PM_2.5 concentrations (July–September 2000–2020) characterized spatiotemporal distributions. The Urban Forest Effects (UFORE) model integrated multisource data—remote sensing, leaf area index (LAI), wind speed, and precipitation—within a Geographically and Temporally Weighted Regression (GTWR) framework, quantifying PM_2.5 dry deposition across land cover types and identifying drivers. Key findings reveal the following: (1) PM_2.5 concentrations followed an initial-rise-then-decline trajectory, with pollution hotspots concentrated in Taiyuan City and neighboring industrial corridors; (2) Construction lands sprawl markedly elevated PM_2.5 levels, whereas green areas and agricultural lands expansion promoted deposition; water areas exhibited no significant effect; (3) Wind speeds and precipitation positively modulated PM_2.5 in green areas and agricultural lands, contrasting with NDVI’s negative influence; elevation showed null correlation, while agricultural lands deposition correlated positively with PM_2.5; (4) GDP displayed an inverted U-curve association with PM_2.5; positive correlations emerged with AVSI, TIOV, NOP, and EC indices, but negative linkage with TOC. This clarifies land cover impacts on urban atmospheric particulates, providing empirical foundations for pollution control.

1. Introduction

Integrated urban development is a fundamental approach to sustainable urban planning and a critical pathway for enhancing regional economic competitiveness [1]. In 1933, German geographer W. Christaller first systematically introduced the concept of urban agglomeration [2]. Later in 1957, Gottmann proposed that future urbanization trajectories would depend on the amalgamation of multiple cities around central urban cores, leading to mega-city formation [3]. This phenomenon is driven by the pursuit of economic advancement at regional levels [4,5].

In China, urban integration development was formally proposed and initiated in 2014 to synergize economic growth across urban clusters. This strategy expands both the scope and depth of cooperation within regional urban networks [6]. As integration progresses, however, expanded production, construction, and industrial development inevitably increase emissions of pollutants such as PM_2.5. These emissions constitute significant pollution sources with demonstrable adverse effects on regional air quality [7].

The study area (Figure 1) encompasses central Shanxi Province. Since the 1990s, Taiyuan City has pursued integration with surrounding urban areas, functioning as the regional core. Initial development centered on the Tai-Yu co-citizenship initiative, focusing on establishing transportation infrastructure, high-tech industries, and university parks [8]. In recent years, Tai-Yu’s urbanization has advanced to foster service sector clusters, promote ecological tourism, and develop high-tech industrial parks.

For decades, Shanxi’s economic growth has been driven primarily by heavy industry and energy-intensive sectors, which account for approximately 60% of the region’s industrial output value [9]. This industrial structure is characterized by high energy consumption, high emissions, and high pollution. Research indicates that PM_2.5 contributes ~30% of air pollutants in Taiyuan and adjacent areas, followed by other pollutants including PM₁₀ [10].

Rapid urbanization and industrialization have been identified by the World Health Organization (WHO) as major contributors to air pollution [11]. In northern China’s persistent haze regions, peak PM_2.5 concentrations exceed 800 μg/m³—over 30 times the WHO guideline value [12]. As the primary component of haze, PM_2.5 poses substantial threats to human health, particularly respiratory systems. This pollutant increases incidence rates of respiratory diseases and contributes to cardiovascular and cerebrovascular conditions, especially among elderly and vulnerable populations [13,14,15,16]. The detrimental effects of PM_2.5 on human health and the environment remain significant concerns [17].

Regional PM_2.5 concentrations are substantially influenced by multiple sources, including coal combustion, transportation emissions, biomass burning (producing inorganic aerosols), and soil dust. These sources collectively account for 7% to 27% of total regional PM_2.5 [18]. Despite governmental policies promoting energy conservation, emissions reduction, and cleaner production, critical scientific questions persist regarding how urban integration affects PM_2.5. This knowledge gap requires urgent resolution, particularly given ongoing high energy consumption and substantial land use alterations [19].

Most studies have focused on PM_2.5 in relation to natural and socioeconomic factors. Natural factors include wind speed, temperature, relative humidity, precipitation, and seasonal variations. Socioeconomic factors encompass regional economic development, urbanization, industrial structure, industrial agglomeration, market segmentation, and environmental regulation [20,21,22]. Research consistently demonstrates that urbanization substantially impacts regional ecological environments. This is primarily manifested through infrastructure development and industrial activities that cause environmental contamination and energy resource pollution. Fossil fuel combustion for seasonal heating in urban areas constitutes an important source of PM_2.5 emissions, elevating atmospheric concentrations of this pollutant [23,24,25]. In recent years, research on PM_2.5 within urban agglomerations or integrated urban development has gained significant attention. Scholars have comprehensively analyzed the spatiotemporal evolution of PM_2.5 pollution, its relationship with meteorological conditions, and its impacts on human health. Findings indicate that PM_2.5 in integrated urban areas exhibits strong spatial aggregation. Studies examining PM_2.5 drivers reveal significant spatiotemporal heterogeneity in influencing factors. Meteorology, socioeconomic conditions, pollution sources, and ecological factors demonstrate interacting effects on PM_2.5 concentrations [26,27]. While existing research has extensively analyzed factors affecting the spatiotemporal distribution of PM_2.5, the retention effect of landscape pattern evolution on PM_2.5 during integrated urban development remains underexplored.

The examination of PM_2.5 concentration trends and their underlying causes during urban integration presents an urgent environmental challenge requiring resolution. This study employs the Tai-Yu-Xin (TYX) urban integration region as a case study to investigate spatiotemporal distribution patterns of PM_2.5 and its dry deposition, alongside the response of PM_2.5 to land use/cover changes. We further examine relationships between PM_2.5 and natural/social indicators through dry deposition estimation across land classes, land use transitions, landscape pattern analysis, and geographically and temporally weighted regression (GTWR). Specifically, this research addresses three questions: (1) the spatial and temporal distribution characteristics of PM_2.5 concentration in the context of urban integration of Tai-Yu-Xin (TYX); (2) response patterns of PM_2.5 and its dry deposition to land use and landscape evolution; and (3) influence mechanisms of natural and social indicators on PM_2.5. This work provides theoretical insights for understanding the spatiotemporal distribution of PM_2.5 in integrated urban development, supporting regional PM_2.5 reduction efforts.

2. Materials and Methods

2.1. Study Area

The urban integration of Taiyuan-Yuci-Xinzhou (TYX) encompasses the city of Taiyuan, various districts and counties within Xinzhou City, and Yuci District (YC) within Jinzhong City. This integration results in a total of 14 administrative districts and counties, as illustrated in Figure 1b. The seven administrative counties that constitute the Taiyuan City administrative division are Yingze (YZ), Xinghualing (XHL), Jiancaoping (JCP), Wanbailin (WBL), Xiaodian (XD), Gujiao (GJ), and Yangqu (YQ). The remaining six administrative counties are under the jurisdiction of the city of Xinzhou. They are Xinfu (XF), Dingxiang (DX), Yuanping (YP), Wutai (WT), Daixian (DC), and Fanshi (FS). TYX is situated in the north-central region of Shanxi, to the north of the Jinzhong Basin. It is bordered to the west by Lvliang Mountain and to the east by Taihang Mountain. The geographical characteristics of the study area are intricate and varied, comprising mountainous terrain as the predominant feature, interspersed with basins, valleys, and plains. The elevation of the area ranges from 519 to 3091 m above mean sea level (a.m.s.l.). The study area is classified as a warm-temperate climate zone. The climate is classified as a temperature continental monsoon climate, with an annual average precipitation of 456 mm, calculated as the mean over the period 2000–2020. The precipitation is concentrated in July and August, accounting for approximately 60% of the yearly total. Average annual sunshine duration in the study area is 2 808 h for the period 2011–2013, and the mean wind speed is 3.027 m/s for 2016–2017 [28,29]. Although our analysis covers 2000–2020, these values are used as reference averages due to data availability, and we acknowledge that actual conditions during 2000–2020 may differ slightly.

The study area is dominated by coniferous forests and mesic deciduous scrubs, with deciduous broad leaf forest representing a secondary vegetation type. Field investigations revealed that the southern part of the study area corresponds to the provincial capital, characterized as a densely populated urban region. In this area, crops such as maize, vegetables, and fruits are predominantly cultivated, whereas drought-tolerant crops such as wheat and potatoes are mainly distributed across mountainous and hilly zones.

2.2. Data Source

2.2.1. Land Use and Land Cover Data

In this study, Landsat imagery for five reference years was used: Landsat 5 TM for 2000 and 2005, Landsat 7 ETM+ (SLC-off) for 2010, and Landsat 8 OLI/TIRS for 2015 and 2020. The multispectral bands of these sensors have a native spatial resolution of 30 m. The pre-processing procedures, including radiometric calibration, atmospheric correction, image splicing, and vector cropping, were conducted on the image data in ENVI 5.3. Subsequently, the selection of training samples and supervised classification were conducted to distinguish the study area into four land use types: construction lands, water areas agricultural lands, and green areas. After completing the series of supervised classification procedures in ENVI 5.3, the class confusion matrix tool was applied by importing the training sample data for each period of land use classification. The classification accuracy was then evaluated by comparing the manually selected reference samples with the results obtained from automated supervised classification. The results indicated that: the overall validation accuracy of the classification results in each period reached over 85%, with Kappa coefficients exceeding 0.83, indicating superior classification outcomes (Figure 2). The DEM data were obtained from the ASTER GDEM 30 m resolution dataset, as illustrated in Figure 1c. The aforementioned images were sourced from the Chinese Academy of Sciences Geospatial Data Cloud Platform “http://www.gscloud.cn (accessed on 16 January 2024)”. It should be noted that the acquired data were all obtained during the summer months (July–September) of the corresponding year, with cloud coverage below 5%.

2.2.2. PM_2.5 Concentration Data

The data regarding the concentration of PM_2.5 are derived from the National Science & Technology Infrastructure of China “http://www.geodata.cn (accessed on 21 January 2024)”. The daily seamless near-surface PM_2.5 concentration dataset at 1 km resolution for China from 2000 to 2020 exhibits a correlation coefficient of over 0.95 with a root-mean-square error (RMSE) of 12.03 and 19.56 μg/m³ when compared to the measured data at state-controlled stations [30].

2.2.3. Wind Speed and Precipitation Data

Wind speed data were sourced from the monthly near-surface mean wind speed dataset for China (2000–2020), released by NSTIC at 1 km resolution [31]. Precipitation data were obtained from historical reanalysis datasets of the European Centre for Medium-Range Weather Forecasts (ECMWF) and National Aeronautics and Space Administration (NASA), accessed via www.xihe-energy.com. This dataset exhibits an average percentage discrepancy of 1.41%, demonstrating sufficient precision for this study [32,33].

2.2.4. NDVI-LAI

LAI was calculated by remote sensing inversion using NDVI values [34,35], and the results of LAI data in the study area are presented in Figure 3. The formula for NDVI and LAI is below:

$$NDVI={\displaystyle \frac{NIR-R}{NIR+R}}$$

(1)

$$LAI=0.57\times e^{2.33\times NDVI}$$

(2)

2.2.5. Socioeconomic Indicators

This study analyzes the following variables: gross domestic product (GDP), added value of the secondary industry (AVSI), total industrial output value (TIOV), population (POP), electricity consumption (EC), and total output value of the construction industry (TOC) to examine how socioeconomic development changes across administrative districts influence PM_2.5 patterns. Socioeconomic data were obtained from the annual Shanxi Statistical Yearbook and China Urban (County) Statistical Yearbook.

2.3. Methods

Figure 4 presents the technical roadmap for analyzing spatiotemporal evolution patterns of dry deposition and PM_2.5 concentrations in TYX over the past two decades. This is achieved by estimating the dry deposition of different land use types in the study area and using an 8 km × 8 km fishing net as the basic research unit for green areas and agricultural land types, respectively. This allows for the exploration of the relationship between PM_2.5. Concentration and Landscape Pattern, Dry Deposition (DD), Wind Speed (WS), NDVI (ND), Precipitation (PR), Elevation (EL), and their responses to regional socioeconomic indicators.

2.3.1. Methods of Estimating Dry Deposition

(1) Methods for estimating dry deposition in green areas and agricultural lands.

The dry deposition of PM_2.5 pollutants was estimated using the Urban Forest Effects (UFORE) model for green areas and agricultural lands categories [36]. Additionally, the percentage of deposition over a specified time range and the percentage of air quality improvement effect can be calculated [37,38,39]. The calculation is performed in accordance with the specifications outlined in Equation (3):

$$Q_{ij}={\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^{m}Flu{x}_{ij}\cdot LA{I}_{ij}}}$$

(3)

where Q_ij represents the PM_2.5 deposition in the ith row and jth column of the image element. LAI_ij denotes the current image element LAI value. Flux_ij signifies the dry deposition flux of the image element, which represents the cumulative amount of particulate matter deposition per unit area and per unit time of the dry deposition state. This is calculated using the formula presented in Equation (4):

$$Flu{x}_{ij}=\sum _{i=1}^n\sum _{j=1}^{m}VO{D}_{ij}·PM\_Valu{e}_{ij}·(1-Re)·T$$

(4)

where VOD_ij represents the dry deposition velocity on the leaf surface of vegetation in this image element (m·s⁻¹), PM_Value_ij represents Current PM_2.5 concentration value in the image element (μg·m⁻³), Re represents resuspension rate, T represents the dry deposition time(s) which is calculated by Equation (5):

$$T=R\_D_{d\mathrm{ay}}\cdot 24\cdot 60\cdot 60$$

(5)

where R_D_day represents the number of days over which the actual dry deposition occurred. The elution rate of different grain sizes on different tree species has been demonstrated to reach 90% when the cumulative rainfall in a single day reaches 15 mm [40,41]. In order to obtain the effective dry deposition days in July–September of each year, the days with single-day precipitation accumulation exceeding 15 mm were excluded, in accordance with the requirements of the dry deposition calculation.

In the calculation of dry deposition flux, the settling velocity and resuspension rate of particulate matter are both closely related to natural factors [42]. In this paper, the dry deposition velocity (VOD) and resuspension rate (Re) are derived based on wind speed, and the correlation index R² of this model is above 0.98 [43]. The calculation method is as in Equations (6) and (7).

$$Re=0.0158·WindSpeed$$

(6)

$$VOD=0.0002\cdot exp(0.4314\cdot WindSpeed)$$

(7)

Equations (6) and (7) are employed to calculate the corresponding dry deposition velocity and resuspension rate values for each month based on the wind speed (m·s⁻¹) data for that period. Figure 5 illustrates the deposition velocity and resuspension rate of PM_2.5 for September 2000.

The PM_2.5 particulate matter deposition velocity for agricultural crops was calculated using the PM_2.5 gravitational deposition velocity, aerodynamic drag above the canopy, empirical LULC parameters, and particulate matter friction velocity, as demonstrated in Equation (8) [44].

$$V_{\mathrm{d}}(P{M}_{2.5})=V_{\mathrm{g}}(P{M}_{2.5})+{\displaystyle \frac{1}{R_{\mathrm{a}}+1/(a_{1}u)}}$$

(8)

where V_d(PM_2.5) represents the dry deposition velocity of agricultural lands (m·s⁻¹), Vg(PM_2.5) represents the gravitational settling velocity of PM_2.5 (m·s⁻¹), R_a represents the aerodynamic resistance above the canopy, α₁ is an empirical parameter based on LULC, and u is the friction velocity, which can be used to estimate the value of the dry deposition velocity of agricultural lands. Given the study area’s location in northern China, which is predominantly dry land, and in light of the findings of related scholars in analogous regions [45], the dry deposition velocity of agricultural lands in this experiment was determined to be 5.6·10⁻² (m·s⁻¹).

(2) Estimation of dry deposition in water areas

The accurate estimation of the dry deposition velocity of fine particulate matter in the surface layer of a water body represents a challenging area of study, with the results obtained by using various estimation models exhibiting significant geographical differences. In this experiment, the dry deposition velocity of fine particles in the surface layer of the water body in the study area was estimated to be 2·10⁻² (m·s⁻¹) by a process of comparison and synthesis involving the research results of numerous scholars and the geographic location of the study area in conjunction with the natural environmental conditions. In light of the aforementioned considerations, the following Equation (9) represents the formula for calculating the total dry deposition in the water areas [46]:

$$Q_{water}=VO{D}_{water}\cdot C\cdot T$$

(9)

where Q_water represents the value of total dry deposition in the water body (μg·m⁻²). VOD_water denotes the dry deposition velocity of PM_2.5 in the surface layer of the water body (m·s⁻¹). C signifies the value of particulate matter concentration (μg·m⁻³); T denotes the time when the dry deposition actually occurs (equivalent to Equation (4)).

(3) Calculation of dust retention rate

The dust retention rate is defined as the ratio of PM_2.5 dry deposition to the total PM_2.5 concentration, and it is calculated using the following Equation (10) [46]

$$P={\displaystyle \frac{Q}{Q+E}}$$

(10)

where P is the dust retention rate, Q is the dry deposition (μg), and E represents the regional atmospheric PM_2.5 background value (μg). Where E can be calculated by referring to Equation (11) [47]:

$$E=C\times T\times H_{MLH}\times S$$

(11)

where C represents the concentration of particulate matter (μg·m⁻³), T denotes the dry deposition time (s), H_MLH signifies the height of the atmospheric mixed layer within the designated study area (m), and S refers to the area encompassed by the specified land class (m²). In consideration of the elevation characteristics of the study area and related literature, a height of 650 m was selected as the atmospheric mixed layer height of the study area in this paper [48].

2.3.2. Evolution of the Landscape Patterns

To more accurately characterize the evolutionary patterns of regional landscape structure during urban integration development, and to more effectively investigate correlations between regional PM_2.5 dry deposition and landscape pattern transformation, this study selected ten landscape indicators (

Table 1. Selected indices and descriptions.

Landscape Index	Abbreviat-Ions	Ecological Implications	Formulas	References
Number of Patches	NP	Number of patches	$NP=N$	[49]
Patch Cohesion Index	COHESION	Aggregation state analysis of patches	$COHESION=\left[1-{\displaystyle \frac{{\displaystyle \sum _{i=1}^m{\displaystyle \sum _{j=1}^n{p}_{ij}^{*}}}}{{\displaystyle \sum _{i=1}^m{\displaystyle \sum _{j=1}^n{p}_{ij}^{*}\times \sqrt{a_{ij}^{*}}}}}}\right]\times {\left[1-{\displaystyle \frac{1}{\sqrt{z}}}\right]}^{-1}\times 100$
Landscape Shape Index	LSI	Complexity of patch morphology	$LSI={\displaystyle \frac{0.25E}{\sqrt{A}}}$	[50]
Aggregation Index	AI	Analysis of Connectivity Between Landscape Types	$AI=\left[{\displaystyle \frac{g_{ii}}{\max \to g_{ii}}}\right]\times 100\%$
Patch Density	PD	Overall landscape heterogeneity and fragmentation	$PD={\displaystyle \frac{n_i}{A}}(10000)(100)$	[51]
Edge Density	ED	Extent of landscape fragmentation	$ED={\displaystyle \frac{{\displaystyle \sum _{k=1}^m{e}_{ik}}}{A}}(10000)$
Contagion	CONTAG	The degree of aggregation and the trend of extension among patches	$CONTAG=\left\{1+{\displaystyle \frac{{\displaystyle \sum _{i=1}^m{\displaystyle \sum _{k=1}^m\left[p_i\left(\frac{g_{ik}}{{\displaystyle \sum _{k=1}^m{g}_{ik}}}\right)\right]}}g\left[\ln p_i\left(\frac{g_{ik}}{{\displaystyle \sum _{k=1}^m{g}_{ik}}}\right)\right]}{2\ln m}}\right\}(100)$
Largest Path Index	LPI	Identify dominant landscape types	$LPI={\displaystyle \frac{\max a_i}{A}}(100)$
Shannon’s Diversity Index	SHDI	Trend distribution of patch types across the landscape	$SHDI=-{\displaystyle \sum _{i=1}^m(p_i\times \ln p_i)}$
Shannon’s Evenness Index	SHEI	Analysis of landscape dominance types and their distribution	$SHEI={\displaystyle \frac{-{\displaystyle \sum _{i=1}^m(p_i\times \ln p_i)}}{\ln m}}$

). Landscape pattern indices for 2000, 2005, 2010, 2015, and 2020 were calculated using Fragstats 4.2.681 Stable Version “https://fragstats.org/index.php/downloads (accessed on 1 March 2024)” to quantify fundamental spatiotemporal characteristics of landscape configuration.

2.3.3. Spatiotemporal Analysis

(1) Geographically and Temporally Weighted Regression

The heterogeneity of the spatial and temporal distribution of PM_2.5 and its influencing factors in the TYX area were analyzed and studied using the geographically and temporally weighted regression (GTWR) model constructed by Huang et al. (2010) [52]. This is expressed in the following equation:

$$Y_i={\beta }_0(u_i,v_i,t_i)+{\displaystyle \sum _k{\beta }_k}(u_i,v_i,t_i)X_{ik}+e_i$$

(12)

where Y_i represents the value of the explanatory variable at point i. The variables u_i and v_i denote the latitude and longitude coordinates of point i, respectively, while t_i represents the time coordinate of point i. β₀(u_i,v_i,t_i) represents the value of the regression intercept at point i, β_k(u_i,v_i,t_i) denotes the regression coefficient of the kth explanatory variable at the point, and if β_k > 0, it signifies that the explanatory variable is positively correlated with the dependent variable, and vice versa, it is negatively correlated. X_ik denotes the value of the kth explanatory variable at point i, and e_i is the random error value.

(2) Spatial Auto-Correlation

The spatial correlation of PM_2.5 concentration distribution in TYX was examined through the calculation of the Moran’s I [53]. The specific calculation formula is as follows:

$$\mathrm{Moran}’\mathrm{s} I=\frac{{\displaystyle \sum _{i=1}^m}{\displaystyle \sum _{j=1}^m}{W}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{S^2{\displaystyle \sum _{i=1}^m}{\displaystyle \sum _{j=1}^m}{W}_{ij}}$$

(13)

where the x_i and x_j represent the attribute values of different research units, while n represents the number of research objects. W_ij represents the spatial weight matrix, and $\overline{x}$ represents the average of attribute values. Moran’s I is calculated between −1 and 1. A Moran’s I value of 1 indicates a significant spatial agglomeration effect, while a value of −1 indicates a significant spatial convergence effect. A Moran’s I value of 0 indicates a random distribution of attribute values.

The local Moran’s I [54] was employed to examine the distinctive attributes of the localized areas within the study region. This was accomplished through the application of the following formula:

$$Local Moran’s I={\displaystyle \frac{n(x_i-\overline{x}){\displaystyle \sum _{j=1}^m{W}_{ij}(x_j-\overline{x})}}{{\displaystyle \sum _{j=1}^n{(x_j-\overline{x})}^2}}}$$

(14)

where the value of m represents the number of neighboring units of a research unit in space. The results of the local Moran’s I were calculated according to the following four classifications: “high-high,” “high-low,” “low-high,” and “low-low,” which correspond to the four types of agglomeration.

3. Results

3.1. Spatial and Temporal Characteristics of PM_2.5

As illustrated in Figure 6, the spatial distribution of cumulative PM_2.5 concentrations was examined for the months of July to September in 2000, 2005, 2010, 2015, and 2020, utilizing a uniform color scale ranging from 0 to 2500 µg/m³. As evident in the figure, the southern part of the study area exhibited persistently higher values, indicative of consistently elevated concentrations of particulate matter. The central and northern regions also exhibited varying degrees of PM_2.5 pollution. Regarding the temporal dimension, the year 2000 (a–c) displayed the lightest hue and the lowest baseline pollution levels. Conversely, 2005 (d–f) and 2010 (g–i) showed a gradual deepening of the hue, while 2015 (j–l) exhibited the darkest and most extensive coverage, reaching a peak in July 2015 (j). In 2020 (m–o), the overall hue remained higher than in 2000 but lower than in 2015, suggesting that while pollution levels had decreased, they remained elevated. Furthermore, except for 2015, pollution levels in September of each year were marginally higher than those observed in July and August of the same year. Overall, PM_2.5 concentrations exhibited a protracted trend of initial increase followed by a decline. The southern region of the study area demonstrated the most pronounced air pollution.

The Moran’s I spatial correlation of the spatial and temporal distribution of PM_2.5 indicates a significant spatial aggregation and positive spatial correlation in the distribution of PM_2.5 within the study area, as illustrated in

Table 2. Global auto-correlation of Moran’s I.

Time Phase	Moran’s I	Z	p
2000	0.54	12.469	<0.001
2005	0.52	12.162	<0.001
2010	0.57	13.810	<0.001
2015	0.60	13.250	<0.001
2020	0.68	15.597	<0.001

. The local Moran’s I (Figure 7) demonstrate that the high-high aggregation area is predominantly concentrated in the primary urban zone of Taiyuan and the northern region of Yuci (YC). Furthermore, the high-concentration aggregation areas observed in Yangqu (YQ) and Xinfu (XF) during the 2000–2015 period underwent a gradual dissolution in 2020. The northern part of the study area exhibited a low-low aggregation scenario, with an expanding trend in the northern end. The local spatial auto correlation analysis indicated that the localized spatial variation pattern of PM_2.5 concentration was consistent with the distribution of high and low concentration values. Furthermore, the study area did not exhibit high-low aggregation or low-high aggregation types during the study period.

3.2. Landscape Mosaic Evolution

In order to facilitate a more detailed analysis of the evolution of landscape pattern in the study area, a series of landscape indices were calculated, with a specific focus on the different land types present in the area.

(1) Type-Level Index

In this study, a number of representative indicators (Table 1) were selected for calculation, and the results are presented in Figure 8.

The NP and PD of the construction land have been increasing over the past 20 years, indicating a gradual increase in the degree of fragmentation. The LPI and LSI continue to increase, indicating that the construction land is subjected to a heightened degree of interference from human activities. The NP and PD indicate that the urbanization construction of the study area began to experience a relatively large amount of rapid development during the period of 2000–2005. In the subsequent period, as urbanization construction becomes more rational, the situation of urbanization development becomes more stable. Urbanization development is characterized by a greater degree of stability.

The NP, PD, and LSI of water areas all show a decreasing trend during 2000–2010, a marked increase between 2010 and 2015, and a subsequent decline from 2015 to 2020. Meanwhile, the LPI increases, indicating enhanced landscape dominance and reduced irregularity of shape.

The NP and PD, LPI and LSI of agricultural lands exhibited an initial increase, followed by a subsequent decrease. This suggests that the degree of fragmentation of agricultural land underwent an initial enhancement, followed by a subsequent weakening, and that the degree of landscape dominance also demonstrated an initial enhancement, followed by a subsequent weakening.

The NP and PD of green areas landscapes exhibited a rapid increase from 2000 to 2005, followed by a yearly decline from 2005 to 2020. The LPI demonstrated a gradual decrease from 2000, subsequently stabilizing. The LSI initially increased and then stabilized. The cohesion values of the various landscapes within the study area were found to be high, indicating a greater degree of connectivity between the different patch types.

(2) Landscape level indicators

As evidenced by the landscape-level index data presented in

Table 3. Landscape index of land types from 2000 to 2020 in study area.

Year/Type	ED	CONTAG	SHDI	SHEI	AI
2000	11.0535	67.4035	0.9738	0.6051	98.3522
2005	19.1381	63.9234	1.0423	0.6476	97.1399
2010	11.1038	66.4040	1.0038	0.6237	98.3450
2015	14.4106	64.7912	1.0361	0.6438	97.8493
2020	12.9671	65.2507	1.0295	0.6397	98.0658

, the ED exhibited an initial increase followed by a decline and subsequent stabilization during the 2000–2020 period. In contrast, CONTAG and AI demonstrated greater stability, with minimal fluctuations in their values. This suggests that the landscape is weakly discrete, more aggregated, and more stable. The value of CONTAG is high, indicating that the dominant plaques have good connectivity. The SHDI and SHEI appeared to increase from 2000 to 2005 and then tend to stabilize, indicating that the proportion of various types of patches in the landscape increased first and then tended to stabilize, presenting a uniform spatial distribution. This indicates that land use is more abundant. The data (Table 3) demonstrate that during the period between 2000 and 2020, the landscape pattern of the study area has been influenced by human activities to a certain extent, with the degree of disturbance gradually increasing over time.

3.3. Spatial and Temporal Characteristics of Dry Deposition in Different Land Use Contexts

(1) Spatial and temporal distribution of dry deposition in green areas

Figure 9 illustrates the spatiotemporal distribution of dry deposition over green areas in July–September for 2000, 2005, 2010, 2015, and 2020. In the high-elevation northern and central zones, blue shading indicates consistently low deposition under strong ventilation. By contrast, the southern lowlands and urban fringe exhibit a progressive transition from light yellow to deep red, reflecting concentrated and increasing particulate settling over time. Each September panel shows the most extensive warm-color coverage, suggesting that late-summer atmospheric stability and agricultural residue burning drive peak deposition. Temporally, high-value hotspots are limited in 2000 but expand in both extent and intensity by 2015; by 2020, these hotspots contract and shift southeastward, and overall deposition levels decline. The persistence of elevated deposition in the southern region underscores the need for continued targeted monitoring and mitigation.

(2) Spatial and temporal distribution of dry deposition in agricultural lands

Figure 10 shows that dry deposition over agricultural lands during July–September in 2000, 2005, 2010, 2015, and 2020, using a unified color scale of 0.01–0.55 µg/m². Most agricultural lands remain in the low-deposition range, indicating that crop cover, residue retention, and irrigation practices effectively limit PM_2.5 settling. However, discrete hotspots appear in mid-elevation zones and valley bottoms, where exposed soils and irrigation channels favor particle retention. Temporally, both deposition intensity and hotspot extent increase steadily from 2000 to a peak in 2015, then contract by 2020 in magnitude and spatial continuity. Seasonally, except for 2010 and 2015, September panels exhibit the most concentrated hotspots, reflecting post-harvest tillage and residue disturbance. Overall, the rise-and-fall pattern of dry deposition underscores the combined influence of evolving agricultural management and climatic factors. Continued soil conservation and targeted dust-control measures are recommended for vulnerable mid-elevation areas.

(3) Dry deposition of water areas

As evidenced by the calculation results presented in

Table 4. Dry deposition value in water areas.

Year	Month	Monthly Average Settling Velocity (μg/m2)	Days of Dry Deposition	Dry Deposition Value (t)
2000	7	0.1920	15	8.3383
	8	0.1926	19	10.5948
	9	0.2020	16	9.3574
2005	7	0.2593	14	9.7487
	8	0.2469	14	9.2825
	9	0.2652	17	12.1070
2010	7	0.3182	15	9.1670
	8	0.2334	16	7.1723
	9	0.2485	15	7.1590
2015	7	0.3515	22	23.4081
	8	0.2012	19	11.5718
	9	0.2217	17	11.4086
2020	7	0.2142	19	7.3013
	8	0.1501	14	3.7699
	9	0.1734	25	7.7770

, the monthly average value of the dry deposition of the water areas is 9.8776 t. Notably, the highest dry deposition of the water body was observed in July 2015, reaching 23 t, while the lowest was recorded in August 2020, at 3.7699 t. The data on dry deposition indicate that the magnitude of the value is closely related to the number of days of dry deposition and the rate of dry deposition.

3.4. PM_2.5 Responses to Land Use and Landscape Evolution

(1) Spatial and temporal evolution of PM_2.5 dust retention rate

Figure 11 presents the spatiotemporal distribution of PM_2.5 retention rates on green areas during July–September in 2000, 2005, 2010, 2015, and 2020. The northwestern and central high-elevation regions exhibit consistently higher retention rates, averaging 3–4%. In contrast, the northeastern region shows persistently lower retention rates across all years. Over time, the spatial extent of high-retention zones fluctuates, with southern areas showing more pronounced and persistent hotspots.

Figure 12 illustrates a gradual increase in the PM_2.5 dry-deposition retention rate on agricultural lands from 2000 to 2020, along with distinct changes in spatial distribution. In 2000, retention rates were generally low, and high-value areas were scattered. From 2005 to 2015, retention rates rose steadily, with high-retention zones increasingly concentrated in the central and southern parts of the study area. By 2020, these areas further expanded and exhibited more spatial continuity.

(2) PM_2.5 Correlations with land use and landscape configuration

Based on the correlation coefficients between PM_2.5 concentration and different land use types and landscape indices over the 20-year period (

Table 5. Correlation coefficient between the land use area and PM_2.5 Deposition.

Correlation Coefficient Between the Land Use Area and PM2.5	Construction Lands	Water Areas	Agricultural Lands	Green Areas
	0.476 *	0.282	−0.425 *	−0.621 *

* Significant correlation at 0.01 level, value without an asterisk are not significant (p > 0.01).

and

Table 6. Correlation coefficient between land use type pattern index and PM_2.5.

	Variable	Construction Lands	Water Areas	Agricultural Lands	Green Areas
Type level	NP	0.316 *	0.215	−0.367 *	−0.431 *
	PD	0.438 *	0.415 *	−0.451 *	−0.571 *
	LPI	0.312 *	0.271	−0.317 *	−0.537 *
	LSI	0.412 *	0.239 *	−0.439 *	−0.498 *
	COHESION	0.218	0.197	−0.213	−0.376 *

* Significant correlation at 0.01 level; values without an asterisk are not significant (p > 0.01).

), the NP, PD, LSI, and LPI of construction lands are positively correlated with PM_2.5 concentration. In contrast, the area and landscape indices of agricultural lands and green areas exhibit significant negative correlations with PM_2.5 levels, indicating that both land types contribute to reducing atmospheric particulate concentrations. The correlation between water areas and PM_2.5 is positive but not statistically significant.

This study also analyzes the relationship between PM_2.5 concentration and six landscape-level indices—ED, CONTAG, SHDI, SHEI, and AI (

Table 7. Correlation coefficient between land use landscape metrics and PM_2.5.

Year/Type	ED	CONTAG	SHDI	SHEI	AI
2000	0.521 *	0.235	0.141	0.402 *	−0.409 *
2005	0.437 *	0.307 *	0.172 *	0.513 *	−0.398 *
2010	0.476 *	0.283 *	0.210	0.569 *	−0.511 *
2015	0.512 *	0.290 *	0.153	0.607 *	−0.581 *
2020	0.497 *	0.313 *	0.107	0.524 *	−0.497 *

* Significant correlation at 0.01 level; values without an asterisk are not significant (p > 0.01).

). Among them, AI shows a significant negative correlation with PM_2.5 concentration, whereas SHEI, ED, and CONTAG exhibit significant positive correlations. The correlation between SHDI and PM_2.5 is weak and statistically insignificant. Overall, landscape fragmentation, diversity, evenness, and connectivity have a measurable impact on PM_2.5 levels in the study area, though the influence is relatively moderate.

3.5. GTWR Model Outputs for PM_2.5 Driving Factors

(1) Natural indicators

The spatial and temporal evolution of green areas PM_2.5 concentration was found to be significantly correlated with each of the natural factors, with the overall pattern of ND > WS > PR > DD > EL (Figure 13). The correlation between PM_2.5 concentration in agricultural lands and the natural factors, with the exception of EL, was more significant, with the overall pattern indicating a higher concentration of PM_2.5 in areas designated DD, ND, WS, and PR (Figure 14).

(2) Social indicators

Based on the GTWR model, this study calculated the spatiotemporal regression coefficients between PM_2.5 concentrations and six socioeconomic indicators across 14 counties within the study area (

Table 8. (a,b) Regression coefficients of social indicators and GTWR of PM_2.5 concentration.

(a)
	GDP					AVSI					TIOV
	2000	2005	2010	2015	2020	2000	2005	2010	2015	2020	2000	2005	2010	2015	2020
DC	−18.66	−43.97	−3.81	−1.13	−0.51	12.26	−25.89	0.27	−5.09	−0.63	−2.72	25.30	0.99	1.87	−0.05
DX	−5.49	−12.96	−3.75	−0.32	−0.38	−48.23	−18.41	0.21	−2.71	0.48	−2.52	6.92	0.91	2.66	−0.30
FS	−17.03	−45.91	−3.86	4.97	−0.49	−4.85	−55.55	0.34	−16.55	−0.03	−2.12	27.05	1.07	2.62	−0.26
GJ	5.16	−3.30	−3.58	−0.22	−0.27	−62.76	−27.22	0.11	0.73	0.57	−2.09	0.75	0.69	8.05	−0.42
JCP	5.31	−3.77	−3.63	−0.19	−0.28	−74.58	−24.58	0.13	0.49	0.59	−2.19	0.69	0.76	7.11	−0.40
WBL	6.56	−3.74	−3.61	−0.15	−0.28	−72.87	−25.17	0.12	0.65	0.58	−2.17	1.14	0.72	8.68	−0.42
WT	−9.77	−6.37	−3.81	−0.64	−0.52	−43.87	−33.96	0.27	−3.44	0.57	−2.31	−1.88	0.99	0.35	−0.38
XD	9.12	−3.97	−3.60	−0.09	−0.28	−78.97	−25.38	0.11	0.65	0.58	−2.20	2.09	0.72	9.80	−0.44
XF	−4.26	−5.07	−3.69	−0.18	−0.35	−50.14	−26.59	0.17	0.03	0.63	−2.60	1.99	0.83	5.81	−0.41
XHL	7.40	−4.32	−3.64	−0.07	−0.29	−80.92	−26.30	0.13	0.55	0.59	−2.27	2.34	0.76	8.62	−0.44
YQ	3.04	−4.62	−3.67	−0.09	−0.33	−72.67	−27.23	0.14	0.35	0.61	−2.42	2.10	0.89	6.66	−0.42
YZ	8.72	−3.95	−3.63	−0.06	−0.28	−82.16	−26.67	0.12	0.54	0.60	−2.31	2.49	0.75	8.06	−0.44
YC	11.98	−3.38	−3.62	−0.05	−0.27	−87.94	−26.09	0.12	0.59	0.61	−2.31	3.42	0.74	9.00	−0.45
YP	−14.78	−3.00	−3.74	−0.02	−0.52	−8.81	−30.48	0.20	−0.77	0.59	−3.41	0.67	0.89	4.90	−0.28
(b)
	NOP					EC					TOC
	2000	2005	2010	2015	2020	2000	2005	2010	2015	2020	2000	2005	2010	2015	2020
DC	0.25	47.41	11.79	18.63	17.59	−550.86	−78.75	10.47	−314.28	−126.31	126.46	1.60	0.13	−1.33	−0.70
DX	−5.71	21.47	11.57	5.73	14.67	−324.27	−31.74	9.74	−165.50	−162.4	247.44	0.65	0.11	−2.57	−0.88
FS	10.14	48.20	11.92	19.39	19.67	−440.20	−88.27	11.19	−359.43	−101.72	−124.29	9.83	0.16	−1.57	−0.90
GJ	−10.95	0.15	10.88	−11.18	0.44	293.24	−8.30	7.36	−17.56	88.34	214.42	−0.32	0.02	−8.57	−0.29
JCP	−11.13	1.65	11.10	−9.84	2.11	67.01	−7.16	8.13	−3.99	47.25	260.79	0.31	0.05	−7.58	−0.36
WBL	−11.63	0.36	10.99	−12.28	0.20	211.42	−12.03	7.78	−25.99	92.45	243.73	0.34	0.04	−9.19	−0.28
WT	1.35	16.55	11.74	8.42	11.23	−325.49	23.19	10.49	−310.63	−129.98	123.68	1.64	0.14	−0.55	−0.62
XD	−12.44	−2.55	10.95	−13.88	−0.50	315.75	−21.97	7.74	−36.11	109.71	242.97	0.71	0.04	−10.31	−0.27
XF	−8.69	0.91	11.35	−9.25	2.44	−239.99	−19.73	8.92	30.19	25.04	296.52	0.46	0.08	−6.27	−0.36
XHL	−12.03	−3.59	11.10	−13.55	−0.60	137.35	−24.97	8.18	2.27	105.07	267.96	0.81	0.06	−9.13	−0.27
YQ	−10.54	−2.33	11.23	−12.64	0.31	−44.32	−23.12	8.58	70.72	78.27	285.62	0.80	0.07	−7.17	−0.29
YZ	−12.70	−5.55	11.06	−14.00	−0.92	203.83	−28.94	8.09	43.18	114.64	267.72	0.94	0.06	−8.57	−0.27
YC	−13.77	−8.89	10.98	−14.58	−1.04	371.34	−40.84	7.94	20.37	122.22	256.17	1.04	0.05	−9.51	−0.28
YP	−7.52	−15.42	11.52	−11.29	3.63	−453.94	−157.54	9.47	160.28	−19.22	396.71	0.48	0.10	−5.31	−0.29

a,b). The aim is to quantitatively reveal the spatial heterogeneity in the relationship between PM_2.5 concentrations and socioeconomic indicators.

This study also calculated the correlations between various socioeconomic indicators and both PM_2.5 concentration and its dry deposition, aiming to examine the influence of socioeconomic factors on air pollution levels and the ecological capacity for particulate removal (

Table 9. Statistics on the correlation between various economic indicators and PM_2.5.

	PM2.5 Concentration	PM2.5 Dry Deposition
GDP	0.851 *	−0.691 *
AVSI	0.691 *	−0.655 *
TIOV	0.527 *	−0.637 *
NOP	0.533 *	−0.505 *
EC	0.446 *	−0.218
TOC	0.721 *	−0.741 *
R2	0.849	0.859
Adjusted R2	0.719	0.647

* Significant correlation at 0.05 level; values without an asterisk are not significant (p > 0.05).

). The results show that all socioeconomic indicators are significantly positively correlated with PM_2.5 concentration. Except for EC, all indicators exhibit significant negative correlations with PM_2.5 dry deposition. The model’s R² and adjusted R² values indicate that the selected socioeconomic indicators effectively explain the variations in both PM_2.5 concentration and dry deposition.

4. Discussion

4.1. Mechanism of PM_2.5 Responses to Land Use and Landscape Patterns

Figure 11 presents the spatiotemporal distribution of PM_2.5 dust retention rate on green areas during July–September in 2000, 2005, 2010, 2015, and 2020. The relatively higher retention rates observed in the northwestern and central high-elevation regions may be explained by dense vegetation cover and favorable meteorological conditions that promote particle deposition. This pattern is consistent with the findings of Li et al., who also reported stronger dust retention capacity in high-elevation environments [47]. This likely reflects dense vegetation cover and stable plant growth, which promote pollutant–leaf contact and effective particle adsorption [55]. In contrast, the northeastern region maintains lower retention year-round; its high altitude and complex terrain support scattered coniferous forests and alpine meadows with low canopy density, limiting dry-deposition efficiency [56]. These spatial patterns and their temporal fluctuations underscore the combined effects of vegetation management practices and interannual climate variability on PM_2.5 capture by green spaces. Nevertheless, the urban areas in the southern sector generally exhibit relatively low dust retention rates, indicating the need for targeted measures—such as optimizing green infrastructure and implementing regular foliar cleansing—to further enhance particulate retention in these regions.

Figure 12 illustrates a steady increase in PM_2.5 dust retention rate on agricultural lands from 2000 to 2020, accompanied by notable shifts in spatial distribution. In 2000, dust retention rates were generally low, with high-rate patches dispersed across the landscape. Between 2005 and 2015, dust retention rates climbed annually, and prominent hotspots gradually concentrated in the central and southern sectors. By 2020, these high-retention areas expanded further, forming more continuous clusters. This pattern reflects the dominance of broadleaf crops—such as maize, orchards, and vegetables—that offer large foliar surfaces for particle capture [57]. In contrast, regions sown with wheat, soybean, or potato exhibit lower retention due to smaller leaf areas. The overall rise in retention by 2020 may also stem from integrated clean-production initiatives that have reduced ambient PM_2.5 levels, thereby enhancing deposition efficiency [58]. Nevertheless, the expansion of high-retention zones signals increased environmental pressure in vulnerable areas, underscoring the need for targeted soil-management and conservation practices.

Land use factors play a critical role in shaping the “source–sink” landscape system’s impact on PM_2.5 concentration. As shown in Table 5 and Table 6, construction lands functions predominantly as a “source” landscape [59]. The expansion of construction land, along with increased patch fragmentation and edge density, tends to exacerbate PM_2.5 pollution. Therefore, future planning of construction land should aim to regulate edge growth and reduce fragmentation to help mitigate PM_2.5 emissions. In contrast, vegetated landscapes serve as “sink” landscapes. Through direct adsorption of PM_2.5 by plant surfaces and indirect effects such as transpiration and airflow reduction by tree canopies, vegetation can retain particles and modify local microclimates to suppress dust dispersion [60]. Agricultural and green spaces thus contribute significantly to particulate matter reduction. Given the limited area of water bodies and their exposure to pollutant diffusion from nearby construction zones, the observed rise in PM_2.5 concentration in water landscapes may be attributed to these external influences [61]. According to the results in Table 7, enhancing patch aggregation in green areas can effectively reduce PM_2.5 levels, whereas increased landscape fragmentation may weaken the capacity of vegetated land types to mitigate air pollution.

4.2. Driving Mechanisms of Natural and Socioeconomic Factors on PM_2.5

(1) Driving mechanisms of natural factors

As illustrated in Figure 13, a notable positive correlation exists between PM_2.5 concentration and WS in green areas and agricultural lands. WS exerts a pronounced influence on PM_2.5 concentrations in green areas within the south-central and southwestern regions of the study area. Conversely, its impact is relatively limited in the northwestern and northeastern zones. This is due to the differing rates of ground-transfer winds at varying elevations, which in turn result in disparate levels of transport and dispersion of PM_2.5 particulate matter. The flat terrain facilitates the diffusion of PM_2.5 particulate matter. However, as elevation increases in the northwest and northeast, it impedes the diffusion of PM_2.5 particulate matter, thereby exerting a pronounced interactive effect of EL and WS on the diffusion of PM_2.5 particulate matter in this region. This, in turn, gives rise to disparities in the accumulation of PM_2.5 concentrations [62]. As illustrated in Figure 14, the impact of WS on PM_2.5 concentration in agricultural land exhibited a progressive intensification from Daixian (DC) and Fanshi (FS) to the central urban zone of Taiyuan City, Yuci (YC), and other regions. This is due to the fact that the areas designated as Daixian (DC), Fanshi (FS), and other areas predominantly comprise villages, whereas the southeastern region constitutes the core and primary urban center of the study area. The presence of urban buildings with varying heights exerts a discernible impact on the transmission and diffusion of PM_2.5, thereby imparting a degree of clustering effect on the PM_2.5 concentration [63]. As illustrated in Figure 9 and Figure 11, the interaction between EL and WS on PM_2.5 in green areas has been observed to enhance the dust retention effect of green areas on PM_2.5, while also increasing the dry deposition of PM_2.5 to a certain extent. The agricultural landscapes were primarily concentrated in the narrow plains in the central part of the study area, which resulted in a negligible and non-significant impact of EL on the spatial and temporal distribution of PM_2.5.

As illustrated in Figure 9 and Figure 10, the maximum values of dry deposition on green areas and agricultural lands in August of each year were observed to be lower than those recorded in July and September of the same year. Given that August falls within the study area’s rainy season, it can be inferred that the impact of green areas and agricultural land on dry deposition of PM_2.5 will diminish with an increase in PR. This is due to the fact that precipitation processes will elute the PM_2.5 particulate matter deposited on the surface of vegetation, thereby reducing the retention time of PM_2.5 on the surface of vegetation and consequently decreasing the dry deposition of PM_2.5 on green areas and agricultural lands [64]. It has been demonstrated in related studies that when precipitation is weak or accompanied by the compounding effect of WS, the eluted particulate matter will be resuspended into the atmospheric environment along with the evapotranspiration process of rainwater, resulting in a rebound of PM_2.5 concentration [65].

The magnitude of the ND value is indicative of the growth of vegetation and crops, as well as the degree of regional vegetation cover. Since 2005, the regression coefficient of ND in the southwestern region has shown an increasing trend, gradually expanding toward central areas from 2010 onward. This indicates a progressively enhanced PM_2.5 retention effect of ND in these regions. This suggests that the influence of ND on PM_2.5 in this region is undergoing gradual intensification. This can be attributed to the initiative to increase the regional forest greening rate, as outlined in the “11th Five-Year Plan” (2006–2010) for Taiyuan. During this period, the forest greening rate in Taiyuan increased by 4.45%, and the greening coverage rate of the built-up area increased from 36.1% to 41% [66]. During the “12th Five-Year Plan” (2010–2015), the existing ecological woodland in Taiyuan City increased from 76,446.74 ha to 97,482.14 ha, while in Xinzhou City, it increased from 335,115.36 ha to 427,598.03 ha [67]. The “13th Five-Year Plan” (2016–2020) for gardening and greening in Taiyuan highlighted the importance of increasing the forest vegetation coverage rate in the mountains surrounding the city to 42.87% by the end of 2020. Additionally, the plan set a target of 1240.31 ha for the creation of new urban green areas [68]. The spatial distribution of ND regression coefficients indicates that the ND of green areas in the vicinity of Taiyuan city is more responsive to PM_2.5 in the presence of relevant policies. Additionally, the vegetation cover of green areas in the eastern and northern parts of the study area exhibits a slower rate of increase, resulting in a diminished impact on the PM_2.5 concentration value in these regions. Figure 14 illustrates a robust negative correlation between the ND of agricultural land and PM_2.5 concentration, which contrasts with the positive correlation observed in previous studies [69]. This is due to the fact that during the period of July to September in the study area, crops commenced the transition from the growing to the maturity stage. During this period, the natural growth process of vegetation was not affected by human activities such as mechanical harvesting and straw burning. To a certain extent, arable crops can be considered to have a similar adsorption and degradation function with respect to PM_2.5 pollutants as green areas vegetation.

The correlation between DD and PM_2.5 concentration in green areas is negative. Figure 13 illustrates a significant negative correlation between DD and PM_2.5 in the Gujiao (GJ) area between the years 2000 and 2020. This indicates that the dry deposition process of the green areas has a considerable impact on reducing the concentration of PM_2.5 in the area, thereby optimizing the regional atmospheric environment. As illustrated in Figure 6, the PM_2.5 concentration in the Gujiao (GJ) area has consistently remained below that of the main city of Taiyuan over an extended period. This can be attributed to the presence of a mountainous barrier between the two regions, which has effectively impeded the diffusion of PM_2.5 particulate matter into the Gujiao (GJ) area, maintaining a relatively low PM_2.5 concentration in the region over time. Additionally, Figure 11 illustrates that the green areas dust retention rate in Gujiao (GJ) has remained consistently high for an extended period. The elevated dust retention rate, coupled with the low pollutant concentration in the background, has resulted in a pronounced negative correlation between DD and PM_2.5 in the region. As illustrated in Figure 6, It can be observed that the green area landscapes of Wutai (WT) and Fanshi (FS) in the northeast of the study area are situated at a considerable distance from the pollution sources of industrial production and the emission sources of human activities. Additionally, the background value of PM_2.5 concentration is relatively low in this area. The green areas in this region is characterized by a single type of vegetation, which is influenced by the high-altitude mountainous landform. This vegetation has a weaker effect on the dry deposition of PM_2.5, which in turn has resulted in a weak correlation between the DD of green areas in this area and PM_2.5 over time. The value of the DD regression coefficient for agricultural land (Figure 14) indicates that during the period of 2000–2010, the concentration area with the highest value in the entire study area is the surrounding districts and counties centered on Taiyuan City. This is due to the proximity of arable crops to populated areas, which are subject to human activities that interfere with the effective process of dry deposition of the agricultural land vegetation. Concurrently, the mechanical utilization of agricultural activities, among other factors, gives rise to the formation of novel PM_2.5 particles, thereby establishing a positive correlation between DD and PM_2.5. The regression coefficients of DD and PM_2.5 in the northern part of Yuanping (YP) and Daixian (DC) exhibited a decline, accompanied by a shift from a positive to a negative correlation in the areas of Daixian (DC), Wutai (WT), and others. This phenomenon, when considered alongside Figure 6, can be attributed to the relatively low concentration of PM_2.5 in this region. The abatement of PM_2.5 by the dry deposition process of crops has a significant impact on the optimization of regional air quality. From 2015 to 2020, the regression coefficient for DD exhibited an increase for Xinfu (XF), Dingxiang (DX), and Fanshi (FS) (Figure 14), indicating that the abatement of PM_2.5 concentration by agricultural land landscapes in the process of urban cluster construction is influenced by a multitude of factors and does not significantly optimize regional air quality.

(2) Driving mechanisms of socioeconomic factors

Some scholars have investigated the significance of urban integration in socioeconomic development from the vantage point of socio-spatial dynamics. This perspective posits that spatial transformations associated with urban integration exert an influence on social structure and community relations [70]. Urban integration represents a significant process whereby urban and rural socio-spatial forces converge and coexist. It can be conceptualized as a form of “spatial succession,” encompassing the transformation of residential, production, and public spaces [71]. The transformation of social networks subsequent to the migration of rural residents to urban areas, a consequence of urban integration, has influenced regional social relations and regional economic development [72]. By employing a socio-spatial dynamics lens, it is feasible to conduct a more exhaustive examination of the manner in which regional PM_2.5 concentration fluctuates in response to an array of socioeconomic indicators throughout the process of urban integration.

The regression coefficient value of GDP in the study area in 2000 (Table 8a) indicates a combination of positive and negative coefficients. The positive correlation is primarily evident in Taiyuan City and YC, which are the core areas of economic and social development and human activities in the study area. Since 2005, as the regional economic development model began to prioritize cleaner production, the Shanxi Provincial Government implemented a series of energy-saving and emission-reduction initiatives and goals [73]. Under the guidance of relevant policies, GDP and PM_2.5 concentration began to exhibit a negative correlation, indicating that economic growth and environmental quality are increasingly aligned. This suggests that economic growth and environmental quality are developing in a more harmonious manner. However, due to the varying stages of socioeconomic development across regions, the predominant types of industries and production efficiency vary considerably, leading to spatial and temporal heterogeneity in the regression coefficient values.

AVSI represents the economic added value of regional industrial processing and production industries, including those in the energy, extractive, manufacturing, and construction sectors. The actual production process of each of these industries is accompanied by the emission of various forms of pollutants. From 2000 to 2020, the regression coefficient value of AVSI exhibited a gradual shift from negative to positive values (Table 8a), indicating a transition from a negative to a positive correlation between AVSI and PM_2.5 concentration across all counties and districts. Additionally, the incremental increase in pollutant emissions associated with the ongoing urban integration construction in the study area contributed to the rising regional PM_2.5 pollution level [74].

TIOV regression coefficients from the time series (Table 8b) demonstrate that the overall positive value is predominantly evident from 2000 to 2015, indicating a positive correlation between TIOV and PM_2.5. This indicates that the integrated development of the city has resulted in regional economic growth, which has been accompanied by a surge in fossil energy consumption. This, in turn, will lead to an increase in PM_2.5 concentration [75]. The 2020 TIOV regression coefficient exhibited a negative value, indicating that the implementation of a series of related policies, such as cleaner production and energy saving, has a more pronounced positive effect on reducing the concentration of PM_2.5 in the region. The results of this study also suggest that the management of pollutant sources can contribute to sustained improvements in the regional atmospheric environment.

From the perspective of the NOP (Table 8b), the impact of population size on PM_2.5 has been predominantly positive in all years of the last 20 years, with the exception of 2000, when the correlation was negative. The construction of urban integration has led to an increase in the frequency of rural-urban population movement, which has in turn led to an increase in regional energy consumption and automobile exhaust emissions. This has resulted in an indirect increase in the concentration of pollutants such as PM_2.5, as well as an increase in regional alkaline pollutant gases such as ammonia [76]. Furthermore, the process of population migration is accompanied by a notable increase in emissions from utilities such as power generation and heating, as well as from the heavy industry sector [77].

EC is an essential metric for assessing the energy consumption of regional equipment and products (Table 8b). In 2000, a notable positive correlation was observed between EC and PM_2.5 concentration in Taiyuan and YC, indicating that electric energy consumption is a primary contributor to regional PM_2.5 [78]. From 2005 to 2020, the EC regression coefficients of the major districts and counties in the study area all exhibited a positive shift, particularly in the core areas of economic development. This indicates that with the process of urban integration, the energy consumption of all industries increased significantly. The integrated urban development process, which relies primarily on fossil fuels such as coal and petroleum as its source of energy, will inevitably result in an intensification of localized air pollution, resulting in high PM_2.5 concentrations [79]. The rising concentration of PM_2.5 in the outdoor environment and the relocation of human activities from outdoor to indoor settings to circumvent the health effects of air pollution also elevate the EC value, which in turn reinforces the positive correlation between EC and PM_2.5 even further [80].

The construction industry plays a pivotal role in the process of urban integration. The regression coefficient value of TOC demonstrates a precipitous decline from 2000 to 2005 (Table 8b), indicating that the contribution of TOC to PM_2.5 commenced a sharp decrease during this period. This is attributable to the enactment and implementation of the inaugural version of the cleaner production regulations by the local government in 1999, which established the mandates for energy conservation and emission reduction in the construction industry and strictly regulated the contribution of TOC to the regional PM_2.5 [81]. From 2005 to 2015, the value of the TOC regression coefficient continued to decrease and began to change from positive to negative, indicating that the production process in the construction industry began to no longer have an impact on regional air quality. This can be attributed to the fact that since 2011, China has formulated the “12th Five-Year Plan” for the development of the construction industry, which sets out clear targets for energy saving and emission reduction in construction [82].

Table 9 illustrates a significant correlation between PM_2.5 concentration and dry deposition with each economic indicator. PM_2.5 concentration demonstrates a notable positive correlation with GDP, indicating that regional GDP exerts a considerable influence on PM_2.5 concentration, with TOC and AVSI also exerting a notable effect. Conversely, the dry deposition of PM_2.5 exhibits a negative correlation with each economic indicator, with the exception of EC, which exerts a notable effect on PM_2.5 dry deposition. The remaining indicators exert a significant influence on PM_2.5 dry deposition, with TOC and GDP exerting a more pronounced effect on PM_2.5 dust retention. In the context of integrated urban development, the growth of GDP and other economic indicators is likely to have a negative impact on regional air quality. Consequently, it is essential to implement further controls on pollutant emissions associated with industrial and construction activities. This will entail optimizing production processes, reducing reliance on the energy industry, exploring alternative avenues of economic growth and formulating low-carbon and environmentally friendly industrial development plans.

5. Conclusions

This study quantitatively assessed the multi-scale effects of natural and socioeconomic factors on PM_2.5 concentrations and dry deposition in the TYX region under the context of urban integration. The results demonstrate that PM_2.5 distribution is jointly influenced by topographic, meteorological, land use, and socioeconomic variables. The spatial configuration of source–sink landscape systems plays a critical role, with construction land acting as a PM_2.5 source and green areas serving as sinks. The spatial spillover of PM_2.5 further underscores the importance of regional coordination rather than isolated local governance. In particular, promoting urban integration and improving the spatial layout of green spaces—by increasing patch size, connectivity, and dominance—can enhance PM_2.5 removal capacity. Simultaneously, controlling the fragmentation and edge complexity of construction land is essential for mitigating pollution.

Methodologically, the use of an 8 km × 8 km grid unit in GTWR analysis helped to reduce the limitations imposed by administrative boundaries and revealed spatial heterogeneity in PM_2.5 responses. The study also highlights the importance of selecting appropriate spatial scales when analyzing landscape-ecological processes, as different scales may affect the interpretation and reliability of results. These findings provide theoretical support for atmospheric environmental governance and offer practical guidance for land planning and regional integration strategies.