Spatiotemporal Heterogeneity of Land-Use Landscape Pattern Effects on CO₂ Emissions at the City-Level Scale in China

Abstract

Climate change has emerged as a critical global issue. Land-use/cover change (LUCC) plays a pivotal role in influencing terrestrial ecosystem carbon cycles and further regulates carbon emission intensity by reshaping the spatial characteristics of landscape patterns. Taking 300 Chinese cities as the study area, an analytical framework encompassing carbon emission accounting, regional land-use landscape pattern analysis, spatiotemporal correlation between landscape patterns and carbon emissions, and economic “core-periphery” disparities was presented. The land-use carbon emissions and landscape pattern indices of each city from 2005 to 2020 were calculated, and the geographically weighted regression (GWR) model was employed to examine the impact of land-use landscape pattern changes on carbon emissions from an urban perspective. Furthermore, the cities were categorized into developed and underdeveloped groups based on the median per capita GDP to compare how economic development levels moderate this impact mechanism. The results indicate that the relationship between landscape patterns and carbon emissions exhibits significant spatial heterogeneity, highlighting the complexity of the influence of land-use morphology on carbon emissions. Sustainable land-use strategies must account for regional disparities in economic levels, planning capacity, and administrative characteristics rather than pursuing a uniform urban form. Economic development significantly moderates the carbon mitigation effects of landscape patterns through its influence on spatial governance capacity, leading to pronounced differences between cities at varying development levels. Moving forward, regionally tailored approaches that integrate landscape optimization with industrial transformation and ecological conservation should be prioritized to provide spatial solutions for achieving the carbon peaking and carbon neutrality goals.

1. Introduction

Against the backdrop of intensifying global warming, environmental challenges stemming from technological progress and economic development have become increasingly severe [1]. Scientific research indicates that human activities since the Industrial Revolution have increased atmospheric CO₂ concentrations by over a quarter of a fold [2]. Since 2007, China has accounted for 27% of global carbon emissions, emerging as the world’s largest emitter [3]. In response to rapid climate change, many countries have adopted concrete carbon neutrality targets and implemented dual-strategy plans emphasizing both emission reduction and carbon sequestration [4]. To achieve sustainable development, China has prioritized climate action, committing to a low-carbon economic pathway. At the 2015 Paris Climate Conference, China submitted Enhanced Actions on Climate Change: China’s Nationally Determined Contributions, outlining targets for emission control, energy restructuring, and forest stock expansion by 2030. At the 75th UN General Assembly in 2020, China officially announced its goals to peak CO₂ emissions before 2030 and achieve carbon neutrality before 2060. The “dual carbon” targets represent an unprecedented elevation of China’s low-carbon strategy [5].

Land-use/cover change (LUCC) drives regional and global climate change by altering surface energy and material exchange [6,7]. Studies have shown that LUCC contributes to approximately 30% of anthropogenic carbon emissions, making it the second-largest driver of global climate change after fossil fuel combustion [8,9,10]. Thus, analyzing carbon emissions from a land-use perspective is critical for regional emission management and carbon neutrality goals [11]. Research on land-use emissions has focused primarily on (1) LUCC impacts on emissions [12,13], (2) emission accounting methods [14,15], and (3) land-use optimization for low-carbon targets [16,17,18]. Among the factors influencing emissions (e.g., population, economic development, energy consumption, and transportation), researchers increasingly recognize the potential role of land-use structure and spatial patterns (e.g., landscape pattern changes) in emission reduction [19,20,21,22].

Landscape patterns and their dynamics represent the spatial manifestation of LUCC under combined natural and anthropogenic forces. Their compositional types, geometric morphology, scale characteristics, and spatial configuration profoundly influence regional ecological processes and edge effects [23,24]. LUCC directly shapes landscape patterns, thereby affecting regional carbon emissions by modifying surface cover, human activity intensity, and ecological processes [25,26,27]. Analyzing how landscape pattern changes impact carbon cycling is essential for understanding the mechanisms behind emission reduction [28,29]. For example, urban expansion often increases impervious surfaces, reducing natural vegetation and carbon sequestration, whereas energy-intensive land uses (e.g., transportation, construction, and industry) exacerbate fossil fuel combustion, increasing emissions [30]. Conversely, protecting or restoring ecological lands (e.g., forests and wetlands) enhances carbon sinks, offsetting anthropogenic emissions [31]. Moreover, landscape fragmentation or aggregation influences emission efficiency; thus, compact urban layouts may reduce transportation energy use, whereas low-density, sprawling development increases energy demand [32].

Further studies have revealed that LUCC not only alters surface cover but also modulates emission intensity by reshaping the spatial features of landscape patterns (e.g., fragmentation, patch connectivity, and morphological complexity) [33,34,35]. Urban spatial attributes—such as patch connectivity and shape complexity—affect emissions by altering energy transmission efficiency, resident lifestyles, and urban heat island effects [36,37]. A study on the Yangtze River Delta urban agglomeration demonstrated that fragmented and disordered urban patches significantly increase emissions [38]. More complex and irregular spatial patterns (indicated by higher SHAPE_MN index values) are correlated with higher emissions, whereas optimized connectivity (reflected by larger CONNECT index values) improves emission efficiency [39]. In specific research contexts, scholars typically employ landscape pattern indices (e.g., LPI, PD, PAFRAC, MSIEI, CONTAG) to quantitatively characterize landscape patterns. These indices can effectively reflect various landscape structural attributes, including patch adjacency characteristics, fragmentation degree, and spatial aggregation.

Land-use patterns exert differential impacts on carbon emissions across varied regional contexts through multiple pathways, including industrial production, transportation, and residential activities. In areas yet to achieve economies of scale, high agglomeration may yield environmental benefits. By contrast, in mature urban areas, excessive population density and over-agglomeration could lead to increased energy demand and environmental pressures [40]. For instance, numerous highly urbanized metropolitan regions worldwide have encountered rising carbon emissions attributable to high-density development [41,42]. Economically less-developed regions often suffer from irregular and fragmented landscape patterns [43,44]. Dispersed urban patches are typically associated with fragmented urban functions and sprawl, which pose challenges to commuting systems and transportation networks [45], consequently elevating carbon emissions. Research demonstrates that irregular land-use configurations tend to increase transportation-related carbon emissions [46]. However, in cities prioritizing ecological functions over economic output, moderate spatial complexity can facilitate synergistic integration between urban spaces and green infrastructure. Such cities effectively mitigate carbon emissions by enhancing natural cooling capacities and air purification through these integrated systems [47,48,49]. Existing studies indicate that higher spatial contiguity can significantly reduce average travel distances, thereby generating measurable carbon mitigation effects [50]. This mechanism proves particularly prominent in highly populated cities with advanced regional integration. Nevertheless, in less-developed cities, enhanced contiguity may induce unidirectional human capital flow to core areas, triggering negative externalities, including surging cross-regional commuting demands [51,52].

China’s urbanization has transitioned from rapid growth to a more stabilized phase, which has also led to a turning point in the evolution of carbon emissions. This phased development has resulted in non-linear characteristics in the carbon emission effects of landscape patterns. However, few studies have concurrently accounted for the influences of temporal dynamics and spatial heterogeneity. Existing research often focuses on single land-use types or specific ecosystems [53,54,55,56], neglecting interactive carbon cycling across ecosystems and regional systemic integrity. Such fragmented perspectives hinder the identification of holistic emission mechanisms, potentially leading to misguided policies [57,58,59]. Thus, investigating emission drivers within a regional ecosystem framework is urgently needed. Examining how landscape patterns influence CO₂ emissions over extended time periods—particularly urban land-use structure and configuration—holds practical significance. To examine how landscape patterns affect carbon emissions differently across varying levels of economic development, this study introduces the Core–Periphery Theory. This is a framework for understanding urban spatial interactions and diffusion processes, providing a crucial analytical lens for examining regional disparities [60]. Friedman conceptualized this theory as a universally applicable model primarily employed to explain uneven development processes between regions or urban–rural areas. He posited that any spatial economic system can be delineated into core and peripheral zones with distinct characteristics. The core area accumulates technological and capital advantages through agglomeration effects, while the peripheral area experiences developmental lag due to factor outflows.

To examine the spatiotemporal heterogeneity of how land-use landscape patterns affect CO₂ emissions, with a particular focus on how economic development levels moderate these effects, this study conducts the following research. (1) The spatiotemporal heterogeneity of carbon emissions and landscape pattern indices was calculated. Using land-use data and relevant statistics from 2005 to 2020, CO₂ emissions were calculated via the emission factor method, while landscape pattern indices were quantified using FRAGSTATS 4.2 software. The spatiotemporal evolution characteristics of the results were analyzed. (2) The spatiotemporal heterogeneity was examined regarding the impacts of landscape characteristics on carbon emissions. Geographically weighted regression (GWR) was employed to quantify regional variations in regression coefficients of landscape pattern indices. The evolutionary trends across four time periods (2005–2020) were identified. (3) Heterogeneity was analyzed under the influence of economic development levels. Guided by the core–periphery theory, kernel density analysis was applied to assess the moderating role of economic development levels. As the urbanization process has transitioned from rapid expansion to a stabilized phase, we hypothesize that carbon emissions followed an initial increase before reaching a plateau. Urban development leads to the continual encroachment of built-up areas into other land-use types, which would increase patch fragmentation while reducing inter-patch adjacency and aggregation. However, rational urban planning could decrease patch complexity and enhance the evenness of different patch types. We hypothesize that the influence of landscape patterns on CO₂ emissions exhibits significant spatiotemporal heterogeneity. Developed cities, benefiting from economies of scale and scientific planning, demonstrate superior carbon regulation effects. The findings will provide scientific support for formulating region-specific low-carbon land-use policies, contributing to regional green development and carbon neutrality goals.

2. Materials and Methods

2.1. Study Area and Data Sources

This study focuses on 300 cities in central and eastern China (Figure 1), aiming to investigate the impact of land-use landscape patterns on city-level CO₂ emissions and the relative importance of various driving factors. The study area covers 5327.93 × 10³ km², accounting for 55.5% of China’s total land area. As the most urbanized, economically dynamic, and densely populated region in China, this study area encompasses several major economic hubs and population centers, including the Beijing–Tianjin–Hebei region, the Yangtze River Delta, the Pearl River Delta, and the Mid-Yangtze River urban agglomerations, which collectively represent primary sources of carbon emissions in the country.

Recent decades have witnessed significant land-use/cover change (LUCC) in the study area. The region has experienced rapid socioeconomic development coupled with accelerated urbanization, including the expansion of small- and medium-sized towns. These changes have led to substantial increases in carbon emissions. Concurrently, ecological conservation measures have strengthened natural carbon sinks, resulting in dynamic shifts in regional carbon balance patterns. The selected cities reflect diverse developmental stages. This stratified representation enables a comprehensive assessment of how various urbanization modes influence carbon emissions. Although urban–rural disparities have gradually narrowed with accelerated development, significant intercity variations in CO₂ emission levels persist due to regional heterogeneity in industrial structures. Continued economic growth has driven dynamic land-use transitions, particularly through intensified industrialization and urbanization, resulting in increasingly complex landscape patterns across these cities.

This study utilized land-use data and relevant statistical data from 2005 to 2020. The land-use data were obtained from the China Land-Use/Cover Change Dataset (CNLUCC), a multitemporal remote sensing monitoring dataset provided by the Data Center for Resources and Environmental Sciences (RESDC, http://www.resdc.cn (accessed on 13 January 2025)) of the Chinese Academy of Sciences. Based on the CNLUCC dataset, we further reclassified the original land-use types into the following six categories: (1) cropland, (2) forest, (3) grassland, (4) water, (5) built-up land, and (6) unused land. The statistical data (including energy consumption, agricultural production activity, crop yield, GDP per capita, etc.) were collected primarily from the China City Statistical Yearbook, China Regional Economic Statistical Yearbook, and statistical yearbooks and bulletins of various provinces and cities in China from 2005 to 2020.

2.2. Research Framework

Based on land-use remote sensing images covering 300 cities in China from 2005 to 2020, spatiotemporal variations in land-use quantity and conversion processes were systematically investigated over the past 15 years. The carbon emission factor method was employed to calculate carbon emissions/sequestration of different land-use types, thereby deriving the net carbon emissions for each city. Landscape pattern indices were selected from the following five dimensions: contiguity, fragmentation, complexity, diversity, and aggregation. After Pearson’s correlation coefficient testing, landscape pattern indices for each city from 2005–2020 were computed. The Ordinary Least Squares (OLS) model was integrated into the Geographically Weighted Regression (GWR) model to examine the impact of landscape patterns on carbon emissions. Incorporating the economic “core-periphery” theory, the study area was categorized into developed and underdeveloped cities. Kernel density analysis was applied to investigate how economic development levels moderate the effects of landscape patterns on carbon reduction. It should be noted that the CNLUCC dataset adopted in this study has the following two inherent limitations: (1) the 1-km resolution may lead to inadequate identification of small-scale landscape characteristics within urban areas; (2) the modified classification system employed for the 2020 data may introduce continuity deviations when compared with previous years. The research framework is illustrated in Figure 2.

2.3. Land-Use Carbon Emissions Quantification

This study quantified carbon emissions from built-up land, both carbon emissions and sequestration from cropland [61], and carbon sequestration from forest, grassland, and water. The net carbon emissions were calculated as the algebraic sum of these components, with carbon emissions assigned positive values and carbon sequestration assigned negative values.

2.3.1. Carbon Emissions of Built-Up Land

This research estimated carbon emissions from built-up land based on energy consumption, which was calculated following the methodology established by the Intergovernmental Panel on Climate Change (IPCC, 2006) [62]. The calculation formula is as follows:

$${\mathrm{E}}_{\mathrm{i}}=\sum _{\mathrm{i}=1}^6{\mathrm{C}}_{\mathrm{i}}\times {\mathrm{N}\mathrm{C}\mathrm{V}}_{\mathrm{i}}\times {\mathsf{\delta }}_{\mathrm{i}}\times {\mathrm{O}\mathrm{R}}_{\mathrm{i}}$$

(1)

where ${\mathrm{E}}_{\mathrm{i}}$ represents carbon emissions from energy consumption; ${\mathrm{C}}_{\mathrm{i}}$ denotes energy consumption; ${\mathrm{N}\mathrm{C}\mathrm{V}}_{\mathrm{i}}$ is the average low calorific value; ${\mathsf{\delta }}_{\mathrm{i}}$ represents the ${\mathrm{C}\mathrm{O}}_2$ emission factor; ${\mathrm{O}\mathrm{R}}_{\mathrm{i}}$ indicates the oxidation rate of energy combustion; and $\mathrm{i}(1,2\dots 6)$ represents the type of energy. In the actual calculation, these parameters were determined based on the China Energy Statistical Yearbook and the guidelines proposed by the IPCC (2006). The specific parameters are listed in

Table 1. Coefficients of different energy sources.

Type	${\mathbf{N}\mathbf{C}\mathbf{V}}_{\mathbf{i}}$(TJ/104 t)	${\mathsf{\delta }}_{\mathbf{i}}$(tC/TJ)	${\mathbf{O}\mathbf{R}}_{\mathbf{i}}$
coals	209.08	25.8	0.98
coke	263.44	29.2	0.98
oven gas	1735	12.1	0.995
petroleum	3893.1	15.3	0.995
petrol	430.7	20.2	0.99
diesel oil	426.25	20.2	0.99

.

2.3.2. Carbon Emissions of Cropland

The carbon emission accounting for croplands is calculated primarily by summing the products of various agricultural input consumption amounts and their corresponding emission coefficients. Key carbon sources include fertilizer application, pesticide use, agricultural plastic film coverage, diesel consumption, and agricultural machinery operations [63]. This accounting framework specifically reflects the direct carbon emission effects induced by agricultural production inputs [64]. The calculation formula is as follows:

$${\mathrm{E}}_{\mathrm{m}}=\sum {\mathrm{T}}_{\mathrm{m}}·{\mathsf{\gamma }}_{\mathrm{m}}$$

(2)

where ${\mathrm{E}}_{\mathrm{m}}$ represents the carbon emissions of cropland; ${\mathrm{T}}_{\mathrm{m}}$ is the volume of agricultural production activities; ${\mathsf{\gamma }}_{\mathrm{m}}$ represents the coefficient of agricultural production activities; and $\mathrm{m}(1,2\dots )$ is the type of agricultural production activities. Based on the inventory of agricultural production inputs established by relevant scholars [65,66], the carbon emission factors for various agricultural production activities were determined as shown in

Table 2. Carbon emission factors of various agricultural production activities.

Type	Carbon Emission Factor	Comparative Analysis
agricultural fertilizers (t/kg)	8.956 × 10−4	The emission factor is low, but the application rate per unit area is large.
Pesticides (t/kg)	4.934 × 10−3	The emission intensity is 5.5 times that of fertilizers, which is related to the chemical properties of organophosphorus pesticides.
agricultural plastic film (t/kg)	5.180 × 10−3	Due to the long-term carbon release effect of microplastics, the residue rate of mulch film is as high as 20%.
agricultural irrigation (t/hm2)	2.048 × 10−2	60% of China’s farmland is irrigated. Diesel pumps contribute major carbon emissions.
agricultural machinery (t/kw)	1.800 × 10−4	The unit power emission is the lowest, but the total power of China’s agricultural machinery exceeds 1 billion kW.

. The data on the quantities of various agricultural production activities were sourced from publicly available materials provided by the National Bureau of Statistics and the Ministry of Agriculture and Rural Affairs.

2.3.3. Carbon Absorption of Cropland

The assessment of carbon sink effects in crop systems requires comprehensive consideration of species-specific biomass accumulation characteristics and carbon sequestration efficiency [67]. During climate regulation processes, crops exhibit significant interspecies variation in their carbon absorption capacity through photosynthesis [68,69]. Current research typically quantifies cropland carbon sequestration potential by developing models based on the per-unit-area biomass yield of crops and their respective carbon absorption coefficients during growth periods. The formula is as follows:

$${\mathrm{S}}_{\mathrm{n}}={\mathrm{C}}_{\mathrm{n}}\times {\mathrm{D}}_{\mathrm{n}}$$

(3)

where ${\mathrm{S}}_{\mathrm{n}}$ is the carbon sequestration of a specific crop variety; ${\mathrm{C}}_{\mathrm{n}}$ represents the carbon sequestration coefficient; ${\mathrm{D}}_{\mathrm{n}}$ is the biological yield; and $\mathrm{n}(1,2\dots )$ denotes different crop species. The carbon absorption coefficients of major crops in China based on related research results [70] are shown in

Table 3. Proportion of each crop’s yield to total yield and carbon absorption coefficients.

Type		Rice	Wheat	Sorghum	Millet	Soya	Potato	Oilseed	Vegetable	Cotton	Tobacco	Fruit
${\mathrm{C}}_{\mathrm{n}}$		0.414	0.485	0.471	0.450	0.450	0.423	0.450	0.450	0.450	0.450	0.450
Proportion/%	2005	14.56	7.86	11.24	0.15	1.74	2.80	2.48	45.51	0.46%	0.22	13.00
	2010	13.54	7.97	12.26	0.10	1.31	2.15	2.23	45.02	0.41%	0.21	14.80
	2015	12.24	7.65	13.21	0.11	0.93	1.96	2.08	45.22	0.33%	0.17	16.10
	2020	11.63	7.41	14.47	0.15	1.18	1.60	1.94	45.97	0.33%	0.12	15.20

. The data on the yields of various crops and cash crops used in this table were sourced from publicly available materials provided by the National Bureau of Statistics and the Ministry of Agriculture and Rural Affairs.

2.3.4. Carbon Absorption of Other Categories

The carbon absorption of forest, grassland, water, and unused land was calculated by multiplying their respective areas by the carbon emission coefficients. For the forest and grassland coefficients, we referenced findings from studies on China’s vegetation carbon absorption, which reported average coefficients of −0.56 for forests, −0.134 for shrublands, and −0.021 for grasslands from 1949–2003 [71,72]. As most studies commonly use −0.6 for forests and −0.02 for grasslands, and considering that our study covers over half of China’s territory, we adopted mean values of −0.5 for forests and −0.02 for grasslands. For water bodies and unused land, we determined coefficients of −0.045 and −0.005, respectively, based on research showing their stable long-term CO₂ absorption patterns. All coefficients represent net carbon sequestration capacities (negative values indicate absorption).

2.4. Land-Use Landscape Pattern Quantification

Landscape pattern indices refer to metrics used to quantify the structural, morphological, and spatial configuration characteristics of landscapes, which encapsulate landscape pattern information. To comprehensively and accurately capture landscape pattern changes in 300 Chinese cities from 2005 to 2020, this study selected eight landscape pattern indices from five dimensions—adjacency, fragmentation, complexity, evenness, and aggregation—to quantify these changes (

Table 4. Description of the landscape pattern indices.

Type	Factor	Abbreviation	Description
Contiguity	Largest patch index	LPI	Largest patch index (LPI) quantifies the percentage of total landscape area occupied by the largest patch, reflecting its dominance and connectivity within the landscape. Higher LPI values indicate greater contiguity of the same land cover type. Metric range: 0 < LPI ≤ 100.
Fragmentation	Number of patches	NP	The number of patches (NP) measures landscape heterogeneity, with NP values positively correlated with the fragmentation degree. Metric range: NP > 0.
	Patch density	PD	Patch density (PD) quantifies the number of patches per unit area, reflecting the spatial fragmentation intensity of the landscape. Metric range: PD > 0.
Complexity	Landscape shape index	LSI	Landscape shape index (LSI) quantifies patch shape complexity. The LSI value equals 1 when all patches in the landscape are perfect squares. The index increases as patch shapes become more irregular. Metric range: LSI ≥ 1.
	Perimeter-area fractal dimension	PAFRAC	Perimeter-area fractal dimension (PAFRAC) measures the complexity of urban form, with higher values indicating greater spatial complexity. Metric range: 1 ≤ PAFRAC ≤ 2.
Diversity	Modified Simpson’s evenness index	MSIEI	Modified Simpson’s evenness index (MSIEI) quantifies the uniformity of patch-type distribution within a landscape. Higher values indicate a more even distribution of patch types without dominant types, reflecting greater landscape evenness. Metric range: 0 ≤ MSIEI ≤ 1.
Aggregation	Interspersion and juxtaposition index	IJI	The interspersion and juxtaposition index (IJI) measures the degree of intermixing and spatial adjacency among different patch types within a landscape. Higher IJI values indicate greater spatial interdigitation of patch types with complex adjacency relationships, while lower values suggest clustered distributions with simpler adjacency patterns. Metric range: 0 ≤ IJI ≤ 100.
	Contagion index	CONTAG	Contagion index (CONTAG) quantifies connectivity between patches within a landscape. Higher values indicate stronger connectivity among dominant patch types, while lower values reflect more dispersed configurations with reduced connectivity. Metric range: 0 ≤ CONTAG ≤ 100.

). To eliminate the impact of multicollinearity, this study conducted a Pearson correlation analysis to examine the relationships between variables and their directional associations. Based on the Pearson correlation coefficients, indicators with strong correlations were screened and excluded. All landscape pattern indices were computed using FRAGSTATS 4.2.

2.5. Model Construction

2.5.1. Ordinary Least Squares (OLS)

Ordinary least squares (OLS) is a regression analysis method used to estimate unknown parameters in linear regression models [73]. The OLS approach determines regression coefficients by minimizing the sum of squared errors (residual sum of squares, RSS) between observed values and model-predicted values, thereby quantifying the relationship between independent and dependent variables. The sum of squared errors can be expressed as follows:

$$RSS=\sum _{i=1}^n{\left(Y_i-{\widehat{Y}}_i\right)}^2$$

(4)

where $Y_i$ represents the observed carbon emission value and ${\widehat{Y}}_i$ denotes the predicted carbon emission value from the regression model. By minimizing the RSS, we obtain optimally fitted regression coefficients that help quantify the relationship between land-use landscape pattern indices and carbon emissions. This method is widely applied in statistics and geographic information systems (GIS) for spatial data analysis and predictive modeling. In this study, we hypothesize that city-level land-use landscape pattern indices influence carbon emissions and employ the OLS regression model to analyze this relationship. The specific regression model can be expressed as follows:

$$Y_i=\alpha +{\beta }_1{X}_{1i}+{\beta }_2{X}_{2i}+\cdots +{\beta }_n{X}_{ni}+{\mu }_i$$

(5)

where $Y_i$ denotes the carbon emissions (dependent variable) of city $i$; $X_{1i}$, $X_{2i}$, …, $X_{ni}$ represent various land-use landscape pattern indices (independent variables) for city $i$; $\alpha$ serves as the model intercept, indicating baseline carbon emissions when all predictors are zero; ${\beta }_1$, ${\beta }_2$, …, ${\beta }_n$ are regression coefficients quantifying the influence magnitude of each landscape pattern index on emissions; and ${\mu }_i$ constitutes the error term accounting for unexplained stochastic variations. In ArcGIS, this OLS regression processes geospatial data to generate both city-specific coefficient estimates and overall model fit statistics (R²).

2.5.2. Geographically Weighted Regression Model (GWR)

The ordinary least squares (OLS) method employs a conventional linear regression approach that considers only variable magnitudes while ignoring spatial autocorrelation. However, the influence of land-use patterns on carbon emissions may exhibit spatial heterogeneity. Moran’s Index (Moran’s I) is a statistical measure used to assess spatial autocorrelation in spatial data. It is divided into global Moran’s I and local Moran’s I, and is widely applied in geographic information systems (GIS) and spatial statistical analysis. This study employs the global Moran’s I for spatial autocorrelation analysis to describe the average degree of association among all spatial units across the entire region [74]. To further investigate this spatial variation in how landscape pattern indices affect emissions, this study advances the OLS framework by incorporating geographically weighted regression (GWR), a localized modeling technique that accounts for potential spatial non-stationarity in the land-use-emission relationship [75,76]. Unlike global OLS, GWR generates location-specific regression coefficients, thereby better capturing spatial heterogeneity and revealing regionally distinct associations that may be influenced by localized economic and planning characteristics. While maintaining a similar functional form to traditional regression, GWR introduces spatial dependency through geographically varying parameters. The model is expressed as follows:

$$Y_i={\beta }_0(u_i+v_i)+{\beta }_1{(u_i+v_i)X}_{1i}+{\beta }_2{(u_i+v_i)X}_{2i}+\cdots +{\beta }_n(u_i+v_i)X_{ni}+{\mu }_i$$

(6)

where $Y_i$ represents the carbon emissions of city $i$ (dependent variable); $X_{1i}$, $X_{2i}$, …, $X_{ni}$ denote various land-use landscape pattern indices for city $i$ (independent variables); ${\beta }_0(u_i+v_i)$, ${\beta }_1(u_i+v_i)$, …, ${\beta }_n(u_i+v_i)$ are location-specific regression coefficients at geographic coordinates $(u_i+v_i)$, capturing spatially varying relationships; and ${\mu }_i$ is the error term accounting for unexplained random variations.

3. Results

3.1. Changes in Land-Use Area

As shown in

Table 5. Land-use changes from 2005 to 2020.

Type	2005/km2	2010/km2	2015/km2	2020/km2	Rate (2005–2020)
Cropland	1,668,702	1,653,214	1,643,256	1,627,459	−2.47%
Forest	1,890,082	1,893,215	1,887,758	1,890,883	0.04%
Grassland	1,103,343	1,095,833	1,090,841	1,087,760	−1.41%
Water	136,139	137,483	136,441	136,587	0.33%
Built-up land	190,011	211,887	237,332	254,089	33.72%
Unused land	338,494	335,856	332,138	331,152	−2.17%

, within the study area encompassing 300 Chinese cities, cropland and forestland constitute the predominant land-use types, accounting for approximately 30% and 35% of the total area, respectively, followed by grassland. Notably, over the past 15 years, the following significant land-use transformations have occurred: cropland, grassland, and unused land areas have decreased by 2.47%, 1.41%, and 2.17%, respectively. Conversely, built-up areas exhibited remarkable expansion, growing from 190,000 km² in 2005 to 254,000 km² in 2020, a substantial increase of 33.72%. This is consistent with our hypothesis that built-up areas continuously encroach upon other land-use types. Forestland and water bodies demonstrated more complex dynamic patterns, undergoing alternating phases of growth and decline before ultimately registering net increases of 0.04% and 0.33%, respectively, across the 15-year period.

As illustrated in Figure 3, the most significant land-use change observed across China’s 300 cities has been the marked expansion of built-up areas, predominantly through the conversion of cropland, which constituted 23.54% of the total transferred area within the study region. Regionally, the most rapid urbanization occurred in peripheral areas surrounding provincial capitals and regional core cities, with Shandong and Jiangsu Provinces exhibiting particularly pronounced growth, whereas Shanghai recorded the highest density of built-up area expansion. However, these urbanizing trends have been partially counterbalanced by the implementation of ecological restoration policies, such as land reforestation programs and Yangtze River conservation strategies, leading to subsequent increases in wetland and forest coverage. The most substantial reduction occurred in the cropland area, with the majority being converted to built-up areas. Additionally, the mutual conversion between cropland and forestland emerged as another significant transition type, each accounting for approximately 10% of the total transferred area.

3.2. Spatial-Temporal Heterogeneity of Carbon Emissions

As shown in Figure 4, urban carbon emissions have rapidly increased overall, with a growing number of cities exceeding the 5 × 10⁷ t emission threshold. The boxplot reveals the data distribution characteristics, displaying quartiles (upper and lower box boundaries) and outliers (points beyond the whiskers). Temporal analysis reveals the following three key trends: first, the overall increase in box height (interquartile range) indicates increasing variability in emission levels, reflecting widening disparities among cities; second, a slight reduction in box height between 2015 and 2020 suggests a potential convergence trend; and third, the persistent presence of outliers across all years represents cities with exceptionally high or low emissions, with their changing distribution possibly reflecting differential progress in emission control—with some cities making notable improvements while others face mounting challenges. Most significantly, the consistent upward shift in median values (horizontal lines within boxes) year after year confirms that the majority of cities experienced continuous growth in carbon emissions throughout the study period.

As we hypothesized, carbon emissions exhibited significant spatiotemporal heterogeneity from 2005–2020. Figure 5 illustrates the spatial distribution and temporal evolution of carbon emissions across China’s 300 cities for the years 2005, 2010, 2015, and 2020. The emission levels are categorized into five distinct classes, with a color gradient from light to dark representing increasing emission intensity; the darkest red hue signifies the highest emission category. Spatially, the patterns reveal the following pronounced regional disparities: coastal eastern China and central regions consistently present elevated emission levels, predominantly represented by dark red and red zones, indicative of their intensive economic activities and rapid industrial development, which drive substantial energy consumption and, consequently, carbon output.

The temporal analysis highlights the following three significant spatial transformations: initially, high-emission zones were concentrated in the eastern coastal and central regions; by 2015, these high-emission zones had expanded radially, with several cities in western China and Inner Mongolia beginning to exhibit increased emission levels; most notably, by 2020, select regions started showing measurable emission reductions. This spatial progression reflects the dynamic interplay between regional development policies, industrial relocation, and emission control measures implemented during China’s 15-year urbanization surge. This aligns with our hypothesis regarding the trend of carbon emissions—an initial increase followed by stabilization.

Overall, China’s carbon emissions generally increased from 2005 to 2020, although with marked regional variations in growth rates and temporal patterns. This overall growth trajectory appears to be closely tied to concurrent processes of urbanization, economic expansion, and rising energy consumption. However, the environmental consequences of this growth have raised significant sustainability concerns. Notably, the period between 2015 and 2020 witnessed a discernible deceleration in emission growth rates, which was likely attributable to China’s implementation of comprehensive carbon mitigation policies during this phase. These measures—including large-scale renewable energy deployment, increased energy efficiency standards, and the establishment of national carbon market mechanisms—appear to have effectively moderated the previously accelerating emission trends.

3.3. Spatial-Temporal Heterogeneity of Land-Use Landscape Pattern Indices

3.3.1. Indicator Selection

This study first conducted a Pearson correlation analysis to examine the intervariable relationships and their directional associations (Figure 6). The results revealed statistically significant correlations between LSI and NP, LPI and CONTAG, MSIEI and LPI, and MSIEI and CONTAG, but minimal correlations between NP/LSI and carbon emissions across most study years. Consequently, from the original eight landscape pattern indices, we systematically selected five indicators that exhibited strong correlations with carbon while maintaining low multicollinearity. These selected metrics, PD (patch density), LPI (largest patch index), PAFRAC (perimeter-area fractal dimension), IJI (interspersion and juxtaposition index), and MSIEI (modified Simpson’s evenness index), quantify key dimensions of land-use morphology, including fragmentation, adjacency, complexity, aggregation, and spatial evenness characteristics, respectively.

3.3.2. Changes in Landscape Pattern Indices

Figure 7 illustrates the temporal trends of the land-use landscape pattern indices in the study area. The boxplot (Figure 7a) shows the distribution characteristics and temporal variations in these indices. Overall, from 2005 to 2020, the median and quartile values of all indices remained relatively stable, indicating consistent overall distribution patterns. The LPI and MSIEI exhibited the highest rates of change, with average annual changes of −3.60% and 3.20%, respectively, followed by the IJI (1.76%) and PD (1.55%), whereas the PAFRAC showed the most stable trend, with an average change of only −0.18%. Notably, the LPI and MSIEI values were concentrated predominantly within the interquartile ranges without significant outliers, suggesting relatively uniform spatial distributions across cities.

As we hypothesized, these indices demonstrated marked spatial heterogeneity in their changes from 2005–2020 (Figure 7b). Using Jenks natural breaks classification, we categorized the change rates of each index into six levels. Most cities presented positive PD change rates, indicating increasing landscape fragmentation, except for Shandong Province and adjacent areas, where negative values suggested reduced fragmentation and more consolidated patches. Negative LPI changes prevailed, reflecting decreased patch adjacency. The PAFRAC changes, although predominantly negative, had minimal absolute values (−0.18%), suggesting slightly reduced but generally stable shape complexity. Positive IJI changes in most cities indicated more interspersed patch distributions and lower aggregation, in contrast to Shandong, Jiangsu, and neighboring cities, which showed increased aggregation. The changes in the MSIEI ranged from 0 to 7.88%, implying moderately enhanced evenness in patch-type distribution across most urban landscapes. This finding is consistent with our initial hypothesis.

3.4. Spatiotemporal Heterogeneity of Land-Use Patterns’ Impact on Carbon Emissions

Global spatial autocorrelation analysis using Moran’s I index revealed statistically significant positive spatial dependence across all examined years. Specifically, the Moran’s I values for 2005, 2010, 2015, and 2020 were 0.14, 0.17, 0.19, and 0.23, respectively (all significantly greater than 0 with p values < 0.01), indicating strong spatial autocorrelation in the data. This spatial dependence necessitated the application of geographically weighted regression (GWR) to analyze the spatiotemporal heterogeneity of the driving factors. The detailed regression results for 2005 (

Table A1. Regression results of GWR model in 2005.

ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI
1	0.37	0.40	−0.10	0.68	0.39	61	−0.50	−0.22	0.23	0.48	0.21	121	0.28	−0.08	0.03	0.42	−0.68	181	−0.61	0.04	0.24	0.66	−0.66	241	−0.71	−1.43	0.52	0.40	−1.14
2	−0.12	−0.38	−0.31	0.03	−0.30	62	−0.32	0.04	0.01	0.06	0.21	122	−1.60	1.78	0.15	1.27	1.01	182	0.18	0.41	0.17	0.31	−0.01	242	0.25	−0.36	−0.01	0.38	−0.27
3	−0.06	−0.34	0.61	0.89	−0.44	63	0.14	−0.65	1.03	0.16	−0.13	123	−0.19	0.19	0.21	0.31	0.15	183	0.18	0.39	0.12	0.15	0.09	243	−0.10	−1.56	0.10	1.34	−2.39
4	−1.06	−1.65	1.30	1.69	−2.78	64	−0.23	0.45	−0.27	−0.33	0.91	124	−0.02	−0.39	0.93	1.66	−1.11	184	0.26	0.48	−0.09	−0.36	0.41	244	−0.45	0.24	−0.43	0.43	0.14
5	0.74	−0.90	−0.30	−1.05	−0.89	65	−0.13	−0.02	0.13	0.45	0.17	125	−0.39	0.29	0.26	0.77	0.57	185	−0.29	0.47	0.13	−0.04	0.67	245	−1.19	−1.62	0.92	1.32	−2.41
6	0.38	0.05	0.09	0.04	−0.54	66	0.16	0.33	−0.36	−0.30	0.59	126	0.06	0.33	0.45	0.55	−0.16	186	−0.18	0.41	0.09	0.72	0.15	246	−0.22	−0.12	0.29	−0.10	0.36
7	−0.87	−0.11	0.28	0.48	0.02	67	−0.11	−0.80	0.26	0.06	−0.53	127	−0.10	−0.47	0.72	1.30	−1.02	187	−0.44	−4.69	−0.07	0.37	−1.87	247	0.53	−0.78	−0.10	0.10	−0.83
8	0.18	−0.19	−0.50	−0.61	−0.29	68	0.31	−0.14	0.04	0.45	−0.77	128	0.78	−0.95	−0.37	−1.12	−0.96	188	−0.52	−1.98	−1.14	−1.74	−0.67	248	−0.56	1.13	0.58	0.47	0.97
9	0.54	−0.43	−0.80	−0.84	−0.48	69	−1.25	0.37	1.52	0.96	−0.14	129	0.47	−0.19	−0.62	−0.70	−0.44	189	−0.01	−0.04	0.64	0.60	−0.35	249	−0.01	0.20	0.30	0.51	−0.25
10	0.34	0.07	0.01	0.37	−0.58	70	0.21	0.08	−1.00	−0.19	0.50	130	−0.35	−0.74	0.41	0.72	−0.49	190	−0.10	−0.02	0.03	−0.25	0.61	250	0.47	0.90	0.16	0.33	0.65
11	0.20	−0.43	0.13	0.36	−0.40	71	−0.45	−0.05	0.05	−0.02	0.21	131	1.06	0.00	−0.13	0.18	−0.08	191	−0.40	0.19	0.35	0.15	0.43	251	−0.35	0.65	0.07	0.43	0.26
12	2.18	1.40	−1.15	−1.12	2.16	72	0.17	−0.21	−0.42	−0.25	−0.13	132	0.01	−1.14	−0.06	0.51	−1.78	192	−0.46	−0.42	−0.95	−0.62	−0.08	252	0.15	−0.10	−0.48	−0.42	−0.20
13	0.26	−0.35	0.05	0.19	−0.85	73	0.55	0.28	−0.59	−0.27	0.42	133	−0.45	0.08	0.84	1.02	−0.42	193	−0.16	−0.54	0.71	0.28	−0.12	253	−0.14	0.31	0.24	0.35	0.23
14	−0.15	−0.15	−0.72	0.56	−0.18	74	−0.43	0.83	0.31	−0.38	1.34	134	0.43	0.22	−0.51	−0.30	0.35	194	0.57	−0.81	−0.39	−0.87	−0.91	254	−0.60	0.02	0.82	0.39	−0.18
15	−1.55	0.77	−0.04	1.81	−0.52	75	−0.74	−9.55	0.77	1.37	−6.32	135	−0.08	−0.29	−0.21	0.06	−0.42	195	0.32	0.25	−0.45	0.04	0.26	255	−1.25	−0.77	0.96	2.15	−1.71
16	1.00	0.05	−0.06	0.29	−0.08	76	−0.16	0.58	−0.65	−0.57	1.10	136	−0.18	−0.69	0.99	1.90	−1.31	196	−2.67	0.80	0.79	1.47	−0.19	256	0.15	−0.56	−0.32	−0.17	−0.39
17	−0.22	0.27	−0.05	0.01	0.65	77	−1.55	−2.33	1.45	1.84	−3.45	137	0.26	0.47	0.07	−0.08	0.25	197	0.55	−0.02	0.44	0.90	−1.29	257	0.45	0.52	−0.68	−0.09	1.28
18	−1.24	2.50	0.61	3.67	−1.25	78	−0.01	−0.60	0.00	−0.06	−0.37	138	−0.56	−0.06	0.41	0.22	0.18	198	−0.05	−0.53	0.06	−0.05	−0.30	258	0.71	0.61	−0.94	−0.24	1.53
19	0.71	−0.88	−0.42	−0.97	−0.92	79	−0.07	−0.72	−0.15	−0.16	−0.36	139	−0.13	0.24	0.16	0.29	0.34	199	−0.29	0.39	−0.26	0.33	0.90	259	0.22	−0.29	0.08	0.38	−0.13
20	−0.21	−0.22	0.52	1.20	−0.52	80	−0.41	1.32	0.36	0.56	0.95	140	−0.73	−0.13	0.19	0.82	−1.07	200	0.17	0.00	−0.39	−0.35	−0.36	260	−0.96	0.31	−0.71	0.78	−0.19
21	0.07	1.32	0.43	−0.34	2.01	81	−0.05	0.08	0.28	−0.03	0.19	141	0.98	0.13	−0.04	−0.26	1.09	201	0.02	−0.31	−0.48	−0.52	−0.40	261	0.00	−0.65	−0.05	−0.12	−0.39
22	2.64	−0.19	−1.65	−1.71	1.50	82	0.50	−0.81	−0.11	0.63	−0.90	142	−0.02	−0.59	−0.31	0.39	−1.23	202	−0.32	−0.81	−0.90	−1.15	0.31	262	−0.18	0.15	0.18	0.27	0.17
23	0.23	1.01	−0.03	−0.48	2.06	83	−0.12	0.14	0.15	0.50	0.10	143	0.66	0.15	−0.66	−0.56	0.25	203	0.66	1.02	−0.11	0.07	0.90	263	−0.08	−0.53	0.04	0.44	−0.60
24	−0.49	−0.16	0.05	−0.09	−0.09	84	−0.21	−0.51	−0.24	0.00	−0.25	144	0.24	−0.10	−0.15	0.23	−0.08	204	1.66	1.08	−0.86	−0.78	1.48	264	−0.08	−0.25	0.09	−0.19	0.20
25	0.79	0.67	−1.29	−0.53	1.42	85	−1.62	4.24	1.47	0.31	3.90	145	−0.23	0.32	0.18	0.24	0.33	205	0.08	−0.74	−0.38	−0.83	−0.29	265	−0.52	−0.09	0.22	0.07	0.01
26	−0.97	2.72	0.97	0.02	2.49	86	−0.12	−1.11	0.02	0.82	−1.34	146	0.14	0.49	0.05	−0.46	0.56	206	0.41	−0.89	−0.16	0.39	−0.69	266	0.38	−0.03	0.51	1.22	−1.52
27	0.21	0.45	0.02	−0.35	0.42	87	−0.43	−0.20	−1.02	−0.50	−0.54	147	−0.25	0.40	0.09	0.09	0.51	207	−0.39	0.11	0.15	0.10	0.32	267	0.26	−0.03	0.07	0.19	−0.16
28	0.20	0.53	−0.51	−0.36	0.96	88	−0.62	0.21	0.32	0.21	0.26	148	−0.06	−0.61	0.17	0.10	−0.43	208	−0.40	−0.32	−0.47	−0.56	0.40	268	0.50	−0.45	−0.11	−0.76	−0.43
29	−0.19	0.38	0.16	0.20	0.41	89	0.90	0.94	−0.39	0.02	1.12	149	0.10	−0.15	0.04	−0.08	0.16	209	−0.49	0.18	−1.11	0.26	−0.02	269	0.13	0.83	−0.14	−0.33	1.01
30	−1.26	3.69	0.62	−0.70	4.14	90	2.60	0.49	−1.61	−1.61	1.60	150	−0.71	−0.10	0.24	0.15	0.10	210	−1.88	2.03	0.45	1.15	2.25	270	−0.28	−0.85	0.64	1.51	−2.00
31	−0.12	−0.51	0.59	0.73	−0.73	91	0.68	0.48	−0.88	−0.57	0.75	151	0.82	0.46	−0.85	−0.46	1.18	211	0.67	0.45	−1.21	−0.49	1.61	271	−0.18	0.26	−0.16	−0.11	0.61
32	−0.26	0.80	0.24	0.18	0.86	92	−0.43	−1.31	0.16	1.97	−2.54	152	−0.22	0.35	−0.10	0.00	0.79	212	−2.50	4.67	1.27	−0.40	4.95	272	0.74	−0.53	−0.01	0.48	−0.68
33	−0.18	0.23	−0.06	0.18	0.54	93	−0.82	−1.20	1.05	1.47	−1.38	153	0.09	0.30	−0.14	0.06	0.18	213	−1.76	2.46	0.45	0.65	2.25	273	−0.15	−0.47	0.09	0.15	−0.31
34	1.14	0.63	−0.60	−0.45	1.12	94	−0.45	−1.25	0.02	1.60	−2.26	154	−0.07	−1.08	−1.16	−1.37	0.44	214	0.13	−0.01	−0.32	0.15	0.12	274	0.21	−0.21	0.03	0.13	0.04
35	0.58	−0.41	0.39	0.81	−0.75	95	−0.36	−0.54	0.64	0.31	−0.08	155	0.38	0.15	−0.53	−0.02	0.39	215	0.39	−0.45	−0.48	−0.61	−0.68	275	−1.07	−0.21	0.64	1.03	−1.19
36	−1.66	−0.04	0.78	1.03	0.04	96	−0.17	−0.50	0.15	−0.07	−0.21	156	−0.26	0.28	0.15	0.17	0.38	216	0.53	−0.45	−0.52	−0.75	−0.60	276	−1.13	1.09	0.73	1.63	0.38
37	0.81	−0.09	0.18	0.58	−0.36	97	−0.04	−0.10	−0.01	−0.08	−0.02	157	−0.51	−1.17	−1.10	−1.08	−0.56	217	−0.22	0.11	−0.06	0.01	−0.18	277	0.01	−0.05	−0.02	0.21	0.13
38	−0.01	0.72	0.26	−0.16	0.84	98	0.29	−0.82	−0.64	−0.67	−0.71	158	−0.10	0.34	−0.07	0.02	0.38	218	−0.82	−0.11	0.35	1.35	−0.61	278	−0.55	−0.73	0.22	0.07	−0.48
39	0.23	−0.56	−0.85	−1.21	−0.33	99	−0.13	0.00	0.56	0.68	−0.29	159	−0.42	−0.76	1.30	2.20	−1.63	219	0.42	−0.15	−0.01	−0.19	−0.24	279	−0.02	1.56	−0.55	0.50	1.36
40	−0.37	0.71	−0.51	0.13	1.05	100	−2.76	1.85	0.22	1.56	2.42	160	0.45	−0.65	−0.16	−0.80	−0.73	220	−0.43	−1.07	−0.58	0.43	−0.63	280	0.43	−0.13	0.01	0.11	−0.41
41	−0.22	−0.73	−0.19	−0.08	−0.30	101	−0.91	−0.33	0.57	0.92	−1.21	161	−0.30	0.41	−0.25	0.31	0.13	221	−0.52	−2.08	0.41	1.17	−2.42	281	0.19	0.42	0.13	−0.21	0.34
42	0.77	−1.01	−0.38	−1.03	−1.00	102	−0.04	−0.55	0.04	−0.04	−0.33	162	−0.19	−0.05	0.16	0.77	−0.26	222	−0.48	−0.49	0.76	0.79	−0.78	282	0.04	−0.52	−0.41	−0.42	−0.55
43	1.06	0.12	−0.15	0.19	0.04	103	−0.44	−0.77	−0.99	−1.03	−0.21	163	−0.22	−0.10	0.23	−0.04	0.14	223	−1.16	−0.03	0.80	1.53	−0.77	283	−0.25	−0.10	0.26	−0.02	0.14
44	−0.24	0.29	0.14	0.18	0.37	104	0.43	−0.13	0.02	0.10	−0.41	164	0.17	0.54	0.13	0.35	0.07	224	−0.33	−0.08	0.18	0.68	0.31	284	−0.06	−0.32	0.61	0.29	−1.07
45	−0.06	0.35	0.44	0.21	0.62	105	−0.04	−0.50	0.60	0.53	−0.06	165	−0.16	−5.44	0.74	0.66	−2.83	225	−0.05	−1.06	−0.25	0.74	−0.90	285	0.12	−0.57	0.40	0.87	−0.94
46	0.21	−0.03	−0.19	0.07	−0.27	106	−0.98	−0.64	0.65	1.10	−1.57	166	1.13	−0.04	−0.25	0.06	−0.06	226	0.59	0.02	0.36	0.59	−1.03	286	0.13	−0.31	0.60	0.10	0.21
47	−0.03	−0.61	0.86	0.25	−0.15	107	0.05	0.56	0.02	−0.33	0.62	167	−0.07	−0.56	0.11	−0.04	−0.32	227	−0.20	0.54	−0.37	0.48	1.07	287	0.90	0.82	−1.26	−0.31	1.69
48	−0.60	2.30	1.02	−0.29	3.00	108	0.43	−0.10	0.02	0.14	−0.38	168	0.14	−0.53	−0.64	−0.97	−0.24	228	0.42	0.13	−1.19	−0.80	1.01	288	−1.20	−0.06	0.41	1.16	−1.21
49	0.27	−0.21	0.19	0.32	−0.87	109	−0.20	0.40	0.36	0.28	0.89	169	0.11	0.04	−0.32	−0.14	0.27	229	0.52	0.01	−0.40	−0.53	−0.02	289	−0.04	−0.47	0.69	0.30	−0.02
50	−0.38	−1.34	0.10	1.61	−2.34	110	0.17	−0.51	−0.06	−0.48	−0.69	170	0.03	−0.13	0.16	0.35	0.01	230	0.18	−0.10	0.41	1.08	−1.24	290	0.51	0.35	0.03	0.43	−0.60
51	−1.12	−0.97	0.79	0.57	−0.58	111	−0.75	−0.52	0.80	0.91	−1.39	171	0.18	−0.76	0.37	0.99	−0.98	231	0.27	0.30	−0.53	−0.33	0.52	291	−0.80	−0.27	0.32	1.44	0.44
52	−0.22	0.28	−0.02	0.10	0.67	112	0.53	0.79	−0.70	−1.28	1.16	172	−0.23	0.33	−0.04	0.00	0.73	232	−0.27	−1.36	0.09	1.36	−2.14	292	−0.56	−1.91	−1.17	−1.52	−0.96
53	0.06	−0.41	0.76	0.22	0.08	113	0.13	−1.19	0.08	0.83	−1.36	173	−2.61	6.85	2.39	−0.97	6.80	233	0.37	−0.04	0.06	0.38	−0.62	293	1.23	−0.85	−0.20	−0.46	0.62
54	0.11	0.49	0.06	0.25	0.14	114	−0.06	−0.93	0.00	0.46	−0.87	174	−0.49	−0.88	0.72	1.02	−1.28	234	−0.57	0.22	0.07	0.96	−0.06	294	−0.07	−0.50	0.68	0.41	−0.08
55	0.62	−0.69	−0.48	−0.88	−0.81	115	−0.14	−0.24	0.42	0.63	−0.45	175	0.21	−0.39	0.36	−0.08	0.12	235	0.36	−0.14	0.04	0.41	−0.72	295	−0.16	−0.04	0.16	−0.06	0.07
56	0.26	0.13	0.03	0.11	0.07	116	−0.22	−0.89	−0.18	0.66	−1.33	176	−0.88	−0.43	0.95	1.86	−1.20	236	1.18	−0.03	−0.38	−0.07	0.03	296	1.78	−0.46	−0.16	−0.64	1.24
57	−0.02	−0.48	−0.11	−0.17	−0.74	117	0.43	−0.13	0.02	0.11	−0.42	177	−0.03	−1.11	0.52	1.76	−1.75	237	−0.51	1.47	0.06	−0.16	2.27	297	−0.26	0.27	0.12	0.12	0.43
58	1.41	−0.34	−0.77	−0.66	0.27	118	−0.22	−0.98	0.18	0.29	−0.52	178	0.10	−0.55	0.42	1.32	−0.91	238	−0.33	−1.16	−0.09	0.79	−1.47	298	−0.26	−0.58	0.57	0.67	−0.78
59	0.17	0.54	−0.21	−0.35	0.48	119	0.07	−0.97	0.53	1.43	−1.48	179	0.14	0.43	0.17	0.33	0.00	239	−0.51	−0.06	−0.09	0.98	−0.36	299	−0.25	0.27	0.16	0.19	0.33
60	0.01	−0.03	0.10	0.44	−0.32	120	−0.02	0.45	−0.36	−0.28	0.86	180	−0.39	−0.50	0.88	1.00	−1.06	240	−0.31	−0.08	0.37	0.12	0.18	300	−0.52	0.02	−0.27	0.48	0.78

), 2010 (

Table A2. Regression results of GWR model in 2010.

ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI
1	−0.27	−0.03	−0.05	0.47	−0.12	61	−0.70	−0.42	0.47	0.70	−0.03	121	0.27	−0.11	0.00	0.38	−0.56	181	−1.15	0.21	0.65	0.84	−0.57	241	−0.68	−2.09	0.69	0.61	−1.78
2	−0.06	−0.35	−0.56	−0.45	0.01	62	−0.65	0.05	0.27	0.23	0.07	122	0.35	2.41	0.78	3.96	−1.94	182	0.01	0.83	0.39	0.58	0.43	242	0.17	−0.11	0.00	0.38	−0.13
3	−0.07	−0.88	0.88	1.09	−1.19	63	0.23	0.19	0.18	0.17	0.10	123	−0.21	−0.18	0.16	0.29	−0.13	183	−0.12	0.20	0.59	0.70	−0.07	243	−0.16	−1.23	0.19	1.39	−2.22
4	−1.31	−0.88	1.72	1.51	−1.91	64	−0.14	0.03	0.12	0.21	0.18	124	−0.10	−0.52	0.52	1.06	−0.60	184	0.05	0.59	−0.26	0.14	0.27	244	−0.46	0.20	0.34	0.82	−0.41
5	0.21	0.18	−0.05	−0.57	0.16	65	−0.08	−0.62	0.46	0.60	−0.58	125	−0.21	0.20	0.05	0.81	0.47	185	−0.18	0.37	−0.41	0.25	0.16	245	−1.12	−1.82	1.14	1.40	−2.52
6	0.63	1.47	0.20	−0.52	0.01	66	−0.05	0.59	−0.32	−0.25	1.03	126	−0.06	1.44	0.26	0.33	1.23	186	−0.29	0.07	0.15	0.80	−0.27	246	0.03	−0.14	0.06	−0.27	0.30
7	−1.56	−0.17	0.85	0.85	−0.28	67	0.07	−0.48	0.42	−0.20	−0.15	127	−0.35	−0.74	0.50	0.99	−0.85	187	−0.67	−3.71	0.02	0.43	−2.03	247	0.34	−0.58	0.06	0.33	−0.93
8	0.24	−0.04	−0.42	−0.63	0.00	68	0.30	−0.19	0.00	0.40	−0.63	128	0.12	0.13	−0.04	−0.39	0.04	188	−0.62	−3.22	−0.31	−0.39	−2.31	248	−0.35	0.62	0.58	0.59	0.06
9	0.54	−0.23	−0.89	−1.29	−0.32	69	−1.58	0.69	1.81	1.01	−0.02	129	0.45	0.02	−0.86	−1.16	−0.20	189	−0.10	0.12	0.69	0.59	−0.13	249	−0.08	0.26	0.25	0.34	0.07
10	0.31	0.08	−0.02	0.28	−0.38	70	−0.03	0.04	−0.36	0.00	0.13	130	−0.70	−1.17	0.15	0.22	−0.15	190	−0.05	0.29	−0.02	0.01	0.59	250	0.00	1.69	−0.13	0.07	1.65
11	−0.05	−0.76	0.65	0.61	−1.04	71	−0.95	−0.08	0.20	0.15	0.50	131	0.40	−0.39	0.24	0.54	−0.50	191	−0.20	−0.08	0.09	−0.08	0.12	251	−0.47	0.21	0.22	0.44	−0.06
12	1.85	1.18	−1.30	−1.22	1.86	72	0.26	−0.32	−0.79	−0.82	0.24	132	−0.61	−0.58	0.59	0.94	−1.51	192	−0.55	−0.86	−0.87	−0.23	−0.22	252	−0.39	−0.86	0.16	−0.32	−0.33
13	0.41	0.32	0.17	−0.28	−0.85	73	−0.04	0.11	−0.18	0.04	0.08	133	−0.67	−0.33	0.51	1.11	−0.55	193	0.02	0.12	−0.05	0.21	0.00	253	−0.20	−0.01	0.19	0.34	−0.02
14	0.06	−0.02	−0.34	0.27	−0.20	74	−0.35	0.45	−0.18	0.26	0.43	134	−0.25	0.41	−0.25	−0.07	0.77	194	−0.55	0.48	0.21	0.61	0.15	254	−1.07	0.37	0.83	1.52	−0.11
15	−0.41	0.25	−0.35	1.35	−0.69	75	−0.70	−5.39	0.52	0.92	−3.79	135	−0.03	−0.38	−0.27	−0.01	−0.46	195	−0.12	−0.02	0.00	0.47	−0.11	255	−1.40	−0.95	0.55	1.74	−1.46
16	0.34	−0.33	0.23	0.59	−0.45	76	−0.06	0.33	−0.04	0.03	0.67	136	−0.26	−1.10	1.23	1.91	−1.68	196	−1.30	0.45	0.55	1.01	−0.37	256	0.11	−0.62	−0.10	−0.38	−0.40
17	−0.08	−0.03	0.09	0.18	0.06	77	−1.58	−2.08	1.75	1.81	−3.13	137	−0.01	0.67	0.24	0.26	0.44	197	0.56	0.34	1.04	1.92	−2.13	257	0.23	−0.25	−0.69	−0.46	0.76
18	0.13	3.09	1.29	5.19	−2.95	78	0.14	−0.54	0.00	−0.47	−0.19	138	−0.18	−0.15	0.08	−0.26	0.11	198	0.16	−0.57	0.04	−0.39	−0.25	258	0.24	−0.77	−0.69	−0.49	0.25
19	0.33	0.02	−0.19	−0.68	−0.11	79	0.04	−0.48	−0.12	−0.45	−0.06	139	−0.10	−0.22	0.07	0.24	0.02	199	−0.18	0.19	0.23	0.64	0.12	259	0.08	−0.08	0.16	0.53	−0.02
20	−0.11	−0.80	0.61	1.08	−1.10	80	0.33	0.49	0.17	0.46	0.09	140	−1.06	0.24	0.88	1.09	−0.96	200	−0.03	0.31	−0.36	−0.36	0.08	260	−0.80	0.68	0.60	0.68	0.01
21	−0.01	1.05	0.55	0.29	1.23	81	0.31	0.01	−0.09	−0.60	0.15	141	0.29	0.15	0.17	0.08	0.53	201	0.10	−0.05	−0.37	−0.55	−0.03	261	0.14	−0.51	−0.04	−0.55	−0.14
22	1.16	−0.56	−0.95	−0.69	0.25	82	0.46	−0.96	−0.13	0.50	−1.01	142	−0.94	−0.02	0.59	1.41	−1.41	202	−0.55	−2.57	−0.21	−0.39	−1.70	262	−0.13	−0.30	0.12	0.22	−0.17
23	1.01	1.30	0.00	0.56	0.87	83	0.08	0.27	0.46	0.21	−0.69	143	0.24	−0.54	−0.43	−0.57	0.04	203	0.76	0.44	−0.23	−0.35	0.57	263	0.03	−0.50	0.03	0.49	−0.69
24	−0.33	−0.22	0.03	−0.05	−0.26	84	−0.21	−0.30	−0.19	−0.17	0.09	144	0.01	−0.09	0.13	0.36	0.01	204	0.94	1.51	−0.95	−0.68	1.82	264	0.06	−0.23	−0.01	−0.30	0.24
25	−0.59	−1.29	0.03	−0.37	−0.36	85	−1.44	2.56	0.96	1.06	2.52	145	−0.22	0.04	0.14	0.24	0.08	205	0.21	−0.40	−0.22	−0.85	−0.03	265	−0.26	−0.16	0.05	−0.16	−0.07
26	−0.85	1.84	0.65	0.34	2.07	86	−0.07	−0.86	0.01	0.71	−1.15	146	−0.02	0.66	−0.14	0.13	0.41	206	0.25	−0.55	−0.02	0.52	−0.69	266	0.08	−0.24	1.39	3.00	−3.07
27	−0.03	0.57	0.02	0.13	0.37	87	−1.07	−1.08	−0.39	−0.15	−0.64	147	−0.34	0.32	0.16	0.18	0.34	207	−0.47	−0.04	0.22	0.14	0.15	267	0.10	0.14	0.22	0.22	0.04
28	0.20	0.19	−0.51	−0.41	0.61	88	−0.44	−0.14	0.22	0.13	−0.09	148	0.12	−0.61	0.12	−0.26	−0.34	208	−0.45	−0.44	−0.48	−0.11	0.40	268	0.09	0.50	0.07	−0.44	0.55
29	−0.22	0.18	0.13	0.21	0.21	89	0.92	−0.56	−0.48	−0.62	−0.29	149	0.13	−0.06	−0.17	−0.40	0.42	209	−0.09	0.92	0.64	−0.59	0.51	269	0.26	0.37	−0.45	−0.45	0.79
30	−0.60	3.59	0.65	−0.71	3.77	90	1.09	1.12	−1.13	−0.81	1.63	150	−1.02	−0.20	0.49	0.27	−0.09	210	−1.02	−0.90	0.24	0.70	−0.48	270	−1.49	−0.57	1.04	2.62	−2.38
31	0.05	−0.33	0.18	0.36	−0.39	91	1.05	0.74	−1.28	−1.03	1.31	151	0.04	−0.61	−0.41	−0.36	0.28	211	−0.03	−1.58	−0.58	−0.65	−0.38	271	−0.06	−0.04	0.12	0.24	0.07
32	−0.31	0.61	0.17	−0.12	0.93	92	−0.49	−1.14	0.11	1.58	−2.04	152	−0.35	0.11	−0.02	0.04	0.31	212	−1.43	4.49	1.60	4.17	−0.38	272	0.60	−0.68	−0.09	0.28	−0.61
33	−0.28	−0.02	−0.02	0.14	0.02	93	−0.67	−0.86	0.96	1.74	−1.16	153	0.03	0.60	0.16	0.46	0.37	213	−0.66	3.54	1.30	5.06	−2.21	273	0.01	−0.44	0.01	0.07	−0.32
34	1.00	0.05	−0.50	−0.58	0.51	94	−0.48	−1.15	−0.05	1.10	−1.76	154	−0.32	−3.43	−0.18	−0.68	−2.44	214	0.22	0.08	−0.24	0.01	0.00	274	0.06	−0.05	0.22	0.54	0.06
35	0.20	−0.59	0.38	0.71	−0.56	95	−0.18	−0.09	−0.10	0.37	−0.15	155	−0.01	0.29	−0.22	0.12	0.35	215	−0.03	0.24	−0.14	−0.18	−0.07	275	−1.78	0.16	1.22	1.32	−0.95
36	−2.41	−0.22	1.55	1.69	−0.45	96	0.21	−0.58	0.07	−0.31	−0.31	156	−0.23	0.02	0.12	0.17	0.15	216	0.39	−0.07	−0.40	−0.85	−0.25	276	−0.26	−0.29	0.27	0.52	−0.22
37	0.27	−0.39	0.30	0.69	−0.48	97	0.11	−0.12	−0.13	−0.25	0.04	157	−0.59	−1.57	−0.83	−0.30	−0.74	217	−0.20	0.29	0.06	0.09	0.22	277	−0.03	−0.08	0.10	0.25	0.01
38	−0.24	0.96	0.30	0.28	1.07	98	0.27	−0.38	−0.58	−0.83	−0.30	158	−0.20	0.88	0.17	0.51	0.77	218	−1.37	−0.34	0.91	1.76	−1.13	278	−0.48	−0.99	0.30	0.18	−0.80
39	0.26	−0.23	−0.51	−0.90	−0.06	99	−0.47	−0.66	0.41	0.60	−0.53	159	−0.58	−0.88	0.91	1.60	−1.01	219	0.38	−0.29	−0.44	−0.74	0.06	279	0.41	1.49	−0.34	0.71	0.80
40	−0.09	0.40	−0.25	0.34	0.38	100	−1.28	−0.91	0.05	0.54	0.08	160	0.02	0.33	0.07	−0.35	0.33	220	−0.54	−1.69	−0.33	0.78	−1.26	280	0.32	−0.07	0.07	0.21	−0.37
41	−0.22	−0.53	−0.12	−0.22	0.00	101	−0.98	−0.02	1.29	1.23	−1.20	161	−0.48	0.38	0.15	0.43	0.01	221	−0.76	0.14	0.65	1.16	−0.35	281	−0.11	−0.02	0.61	0.32	−0.08
42	0.39	−0.04	−0.17	−0.75	−0.10	102	0.15	−0.56	0.02	−0.42	−0.23	162	0.11	0.03	−0.08	0.24	−0.17	222	−0.76	−0.38	0.59	1.03	−0.73	282	0.09	−0.11	−0.41	−0.56	−0.08
43	0.36	−0.28	0.19	0.53	−0.40	103	−0.58	−1.87	−0.55	−0.28	−1.09	163	0.17	−0.12	−0.08	−0.44	0.09	223	−2.07	−0.31	1.55	2.18	−1.35	283	0.12	−0.12	−0.02	−0.42	0.12
44	−0.31	0.13	0.17	0.21	0.18	104	0.32	−0.06	0.08	0.21	−0.38	164	−0.14	1.07	0.47	0.82	0.57	224	−0.31	0.06	0.01	0.65	0.57	284	−0.43	1.29	1.51	0.02	0.63
45	−0.19	−0.58	0.28	0.32	−0.24	105	0.08	0.05	0.05	0.37	0.06	165	−0.41	−2.20	0.66	0.23	−0.45	225	−0.41	−1.28	0.15	1.01	−1.26	285	−0.11	−0.89	0.28	0.69	−0.91
46	0.28	0.07	−0.15	0.10	−0.25	106	−0.84	−0.42	0.97	0.99	−1.20	166	0.57	−0.41	0.12	0.38	−0.50	226	0.68	0.71	0.75	1.05	−1.36	286	0.14	0.04	0.48	0.52	0.16
47	0.13	0.16	0.04	0.18	0.03	107	−0.02	0.63	−0.36	0.15	0.34	167	0.18	−0.57	0.07	−0.36	−0.27	227	−0.16	0.55	0.33	0.73	0.34	287	−0.34	−0.94	−0.17	−0.33	0.06
48	−0.21	2.24	0.90	0.31	2.35	108	0.38	−0.06	0.06	0.08	−0.15	168	0.17	−0.24	−0.40	−0.86	0.03	228	−0.66	−1.91	−0.05	−0.34	−1.04	288	−1.01	−0.10	0.56	0.96	−0.95
49	0.17	0.66	0.32	0.12	−0.62	109	−0.27	0.47	0.12	0.37	1.04	169	−0.02	0.01	−0.29	0.06	0.14	229	0.24	−0.52	−0.24	−0.73	−0.11	289	0.11	0.17	0.01	0.19	0.10
50	−0.42	−1.13	0.09	1.34	−1.93	110	−0.21	0.25	0.14	0.01	0.19	170	−0.05	−0.73	0.45	0.67	−0.82	230	−0.19	−0.35	1.07	2.36	−2.21	290	0.40	0.40	−0.07	0.12	−0.11
51	−1.37	−1.52	1.31	0.90	−1.12	111	−0.85	0.06	1.49	1.18	−1.10	171	−0.05	−1.12	0.68	1.22	−1.46	231	0.10	0.13	0.03	0.30	0.26	291	−1.10	−0.69	0.88	2.33	−0.08
52	−0.25	0.04	0.00	0.07	0.17	112	0.19	1.57	0.86	−1.37	1.49	172	−0.15	0.02	0.07	0.14	0.14	232	−0.30	−1.08	0.12	1.26	−1.83	292	−0.62	−2.20	−0.70	−0.36	−1.18
53	0.16	0.17	0.12	0.23	0.16	113	0.13	−0.87	0.05	0.78	−1.22	173	−1.81	4.96	1.79	0.58	3.87	233	0.32	−0.04	0.07	0.29	−0.38	293	0.62	−0.52	0.02	−0.07	0.27
54	−0.11	1.18	0.41	0.76	0.71	114	0.01	−0.74	−0.01	0.38	−0.82	174	−0.98	−0.29	0.62	1.53	−0.88	234	−0.37	0.03	0.06	0.75	−0.34	294	0.10	0.17	−0.02	0.21	0.08
55	0.29	0.08	−0.24	−0.65	−0.15	115	0.06	−0.11	0.24	0.35	−0.21	175	0.32	0.19	0.33	0.20	0.27	235	0.33	−0.12	0.06	0.36	−0.50	295	0.26	−0.06	−0.15	−0.57	0.07
56	0.17	0.22	0.00	0.04	0.24	116	−0.22	−0.73	−0.33	0.12	−0.77	176	−0.38	−0.29	0.41	1.06	−0.77	236	0.77	−0.40	−0.02	0.22	−0.51	296	0.63	0.05	0.09	−0.05	0.55
57	−0.49	0.22	0.20	0.52	−0.01	117	0.31	−0.07	0.08	0.21	−0.40	177	−0.74	−1.26	0.69	1.89	−1.44	237	−0.06	1.48	0.19	−0.20	1.55	297	−0.26	0.08	0.11	0.11	0.24
58	0.18	0.14	−0.35	0.00	−0.15	118	−0.39	−1.42	0.24	0.40	−0.93	178	0.12	−0.46	0.44	1.36	−0.75	238	−0.31	−0.98	−0.09	0.65	−1.24	298	−0.56	−0.53	0.53	0.69	−0.55
59	0.06	0.82	−0.21	−0.10	0.60	119	−0.51	−1.06	0.57	1.36	−1.02	179	−0.19	0.46	0.62	0.90	0.05	239	−0.52	−0.09	0.15	0.89	−0.58	299	−0.21	−0.08	0.12	0.18	0.04
60	0.28	0.14	−0.21	−0.03	−0.06	120	−0.38	0.23	−0.03	−0.01	0.59	180	−0.75	−0.13	0.52	1.33	−0.44	240	0.16	−0.10	−0.02	−0.48	0.15	300	−0.63	−0.47	−0.19	0.70	0.42

), 2015 (

Table A3. Regression results of GWR model in 2015.

ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI
1	−0.39	−0.15	0.28	0.95	−0.18	61	−0.77	−0.25	0.43	0.96	0.27	121	0.69	0.34	0.15	0.72	−0.43	181	−1.04	−0.30	0.18	0.34	−0.49	241	−0.55	−2.26	0.51	0.44	−1.77
2	0.05	−0.14	−0.44	−0.32	−0.03	62	−0.48	0.18	0.11	0.14	0.29	122	0.10	1.91	0.25	2.79	−1.19	182	0.01	1.13	0.31	0.53	0.67	242	0.24	0.19	−0.02	0.30	0.24
3	0.06	−0.26	0.47	0.54	−0.32	63	−0.05	−0.47	0.60	0.80	−0.26	123	−0.37	0.43	0.43	0.55	0.32	183	0.11	0.22	0.73	0.78	−0.15	243	0.08	−1.10	0.25	1.27	−2.13
4	−1.11	−0.72	0.76	0.74	−1.20	64	0.14	−0.13	0.19	0.23	−0.16	124	0.02	−0.19	0.22	0.50	−0.09	184	0.11	−0.53	0.39	0.50	−0.68	244	−0.68	−0.08	−0.15	1.20	−0.52
5	0.16	0.05	−0.25	−0.51	0.20	65	−0.04	−0.05	0.01	0.16	0.20	125	−0.18	−0.03	0.09	0.83	0.26	185	−0.30	−0.47	0.29	0.72	−0.40	245	−0.95	−1.07	0.53	0.90	−1.61
6	0.63	1.15	0.29	−0.23	0.19	66	−0.19	0.49	−0.13	−0.05	0.83	126	−0.32	0.56	0.33	0.61	0.12	186	0.30	0.11	−0.64	−0.07	0.20	246	−0.17	−0.16	0.20	−0.09	0.35
7	−1.40	−0.33	1.14	1.63	−0.56	67	−0.02	−0.53	0.49	0.07	−0.23	127	−0.45	−0.37	0.51	0.98	−0.51	187	−1.37	−5.74	−0.05	0.42	−2.96	247	0.44	−0.45	0.02	0.16	−0.70
8	0.14	0.08	−0.32	−0.50	0.20	68	0.72	0.11	0.16	0.80	−0.68	128	0.04	0.02	−0.21	−0.25	0.07	188	−0.27	−3.94	−0.23	−0.90	−3.20	248	−0.60	1.11	0.01	−0.21	1.58
9	−0.06	0.22	0.04	0.07	0.24	69	−1.53	0.83	0.89	0.13	1.04	129	−0.44	1.05	0.49	0.85	0.69	189	−0.23	−0.20	0.32	0.15	−0.05	249	−0.14	0.44	0.27	0.37	0.29
10	0.79	0.73	0.10	0.52	−0.04	70	0.19	−0.07	−0.62	−0.04	0.16	130	−0.46	−0.72	−0.61	−0.71	1.34	190	0.08	−0.65	0.55	0.65	−0.31	250	0.13	−0.02	−0.11	0.18	−0.24
11	0.30	−0.47	0.34	0.20	−0.63	71	−0.71	−0.04	0.36	0.47	0.46	131	0.40	−0.06	−0.07	0.17	0.03	191	−0.05	−0.07	0.03	−0.17	0.24	251	−0.36	0.95	0.02	0.51	0.40
12	0.36	−0.72	−0.42	−0.33	−0.71	72	0.43	−0.22	−0.64	−0.67	0.14	132	−1.25	−1.31	0.53	0.77	−1.48	192	−0.53	−0.68	−1.03	−0.09	0.03	252	−0.18	−0.54	0.01	−0.21	−0.24
13	0.47	−0.28	0.05	−0.14	−1.27	73	0.28	0.02	0.09	0.15	−0.05	133	−0.97	0.03	0.43	1.18	−0.45	193	−0.24	−0.46	0.49	0.86	−0.25	253	−0.36	0.69	0.48	0.57	0.51
14	0.16	−0.08	−0.34	0.25	−0.01	74	−0.34	−0.64	0.58	0.85	−0.40	134	−0.17	0.43	−0.13	−0.02	0.79	194	−0.33	0.00	−0.09	0.37	−0.14	254	−0.95	−0.59	0.98	−0.11	−0.88
15	−0.32	0.41	−0.65	1.01	−0.17	75	−0.62	−5.63	−0.11	0.02	−2.64	135	0.10	−0.24	−0.29	0.03	−0.47	195	−0.01	0.27	−0.20	0.24	0.30	255	−1.91	−2.02	1.77	3.13	−2.05
16	0.37	0.01	−0.08	0.17	0.09	76	−0.30	0.73	−0.10	0.02	1.22	136	−0.08	−0.46	0.88	1.46	−0.99	196	−1.96	0.70	0.35	0.57	0.58	256	−0.23	−0.54	0.38	0.25	−0.37
17	0.29	−0.41	0.15	0.12	−0.48	77	−1.32	−1.25	0.84	0.93	−1.80	137	0.16	0.40	0.56	0.44	0.11	197	1.16	0.70	1.26	2.05	−1.53	257	−0.12	0.14	−0.37	0.14	0.92
18	−0.21	1.87	0.30	3.07	−1.54	78	0.02	−0.51	0.02	−0.18	−0.20	138	0.08	−0.10	−0.02	−0.41	0.40	198	0.00	−0.51	0.08	−0.12	−0.24	258	0.10	−0.33	−0.52	−0.15	0.57
19	0.04	0.06	−0.22	−0.15	0.07	79	−0.09	−0.54	−0.14	−0.49	0.06	139	−0.27	0.25	0.27	0.55	0.32	199	−0.30	0.24	−0.05	0.51	0.47	259	0.20	−0.12	0.31	0.52	−0.05
20	−0.12	−0.19	0.69	1.22	−0.75	80	−0.63	0.29	−0.29	0.32	0.48	140	−0.68	−0.23	0.14	0.77	−0.97	200	−0.78	1.33	0.73	1.28	0.89	260	−0.76	0.54	−0.49	0.28	0.73
21	−1.32	2.10	1.21	1.36	0.92	81	0.24	−0.06	0.05	−0.29	0.31	141	0.83	0.51	0.07	−0.03	1.10	201	−0.22	0.15	0.09	0.11	0.13	261	0.04	−0.50	−0.02	−0.28	−0.15
22	0.59	−3.97	−0.54	−0.24	−3.03	82	0.25	−0.54	−0.13	0.34	−0.44	142	−0.74	−1.21	0.96	2.11	−3.17	202	−0.23	−3.37	−0.14	−0.73	−2.81	262	−0.29	0.28	0.35	0.52	0.25
23	0.24	1.98	0.14	1.18	0.61	83	0.06	0.00	0.43	0.28	−0.83	143	0.45	−0.29	−0.32	−0.57	0.11	203	0.68	−0.39	−0.13	−0.34	−0.43	263	0.08	−0.34	−0.03	0.44	−0.52
24	−0.32	−0.08	−0.10	−0.17	−0.03	84	−0.33	−0.41	−0.59	−0.74	0.43	144	0.20	−0.26	0.11	0.18	−0.19	204	1.07	−1.38	−0.46	−0.43	−1.22	264	−0.20	−0.16	0.26	−0.16	0.48
25	−0.36	−1.21	−0.08	−0.29	−0.51	85	−0.99	1.08	0.10	0.63	1.16	145	−0.42	0.65	0.40	0.42	0.57	205	0.13	−0.45	−0.33	−0.96	0.10	265	0.04	−0.11	−0.12	−0.45	0.31
26	−0.72	0.67	0.07	0.44	0.82	86	0.01	−0.53	−0.03	0.50	−0.70	146	0.10	−0.67	0.60	0.56	−0.76	206	0.32	−0.21	0.06	0.39	−0.21	266	0.85	0.64	1.44	2.74	−1.94
27	0.09	−0.64	0.71	0.63	−0.76	87	−0.82	−1.56	−0.52	−0.26	−1.62	147	−0.50	0.83	0.30	0.20	0.82	207	−0.76	0.43	0.39	0.23	0.58	267	−0.04	−0.15	0.17	−0.02	0.02
28	0.19	−0.16	−0.34	−0.02	0.28	88	−0.18	−0.07	−0.01	−0.18	0.15	148	−0.06	−0.55	0.28	0.09	−0.36	208	−0.51	−0.47	−0.66	0.03	0.49	268	0.18	0.16	−0.25	−0.65	0.37
29	−0.50	0.90	0.41	0.34	0.79	89	0.66	−0.62	−0.35	−0.46	−0.34	149	−0.24	−0.09	0.33	−0.14	0.44	209	−1.07	−0.08	−0.57	1.32	−0.36	269	0.12	0.26	−0.33	−0.21	0.65
30	−0.75	2.12	0.05	−0.50	2.55	90	0.98	−1.91	−0.69	−0.41	−1.51	150	−1.11	0.03	0.55	0.49	0.16	210	−1.63	−0.44	−0.06	0.86	−0.19	270	−0.43	−0.93	1.87	3.78	−3.08
31	0.29	−0.20	−0.30	−0.23	−0.10	91	−0.37	−0.60	−0.57	−0.10	−0.72	151	0.22	−0.31	−0.22	−0.31	0.48	211	0.13	−1.50	−0.61	−0.49	−0.42	271	0.24	−0.29	0.15	0.18	−0.33
32	−0.39	−0.02	−0.07	0.52	0.14	92	−0.35	−0.71	0.31	1.78	−1.85	152	0.20	−0.27	0.22	0.21	−0.37	212	−1.32	2.95	0.95	3.14	−0.69	272	0.41	−0.32	−0.20	0.14	−0.14
33	0.18	−0.36	0.19	0.13	−0.47	93	−0.54	−0.30	0.53	1.52	−0.40	153	0.17	−0.06	0.08	0.28	−0.20	213	−0.79	2.69	0.80	3.73	−1.63	273	0.00	−0.24	−0.07	−0.19	0.07
34	0.89	−0.02	−0.29	−0.58	0.43	94	−0.40	−0.75	0.10	1.34	−1.62	154	0.08	−4.16	−0.22	−1.18	−3.31	214	0.39	0.09	−0.56	−0.27	0.09	274	0.13	−0.02	0.37	0.52	0.09
35	0.23	−0.20	0.10	0.32	−0.02	95	−0.38	−0.58	0.52	0.95	−0.30	155	0.47	0.36	−0.52	0.06	0.55	215	−0.33	0.50	−0.06	0.33	0.33	275	−2.55	−0.81	1.63	1.42	−1.28
36	−2.55	−0.08	2.25	3.69	−0.50	96	−0.03	−0.52	0.15	−0.06	−0.26	156	−0.48	0.53	0.36	0.39	0.57	216	−0.04	0.32	−0.12	−0.01	0.31	276	−0.70	0.03	0.05	0.71	−0.22
37	0.33	−0.04	−0.02	0.22	0.08	97	0.02	−0.25	−0.05	0.03	−0.13	157	−0.51	−1.35	−0.93	−0.22	−0.40	217	−0.19	0.30	−0.08	−0.09	0.53	277	0.23	−0.36	0.10	0.10	−0.33
38	−0.09	0.16	−0.12	−0.18	0.49	98	−0.24	−0.04	0.20	0.24	0.07	158	−0.03	0.07	0.26	0.63	−0.06	218	−3.26	−2.15	3.82	5.34	−2.68	278	−0.45	−0.98	0.20	0.04	−0.65
39	0.18	−0.30	−0.71	−1.21	0.15	99	−1.23	0.01	0.54	1.19	−0.71	159	−0.38	−0.42	0.54	0.95	−0.37	219	0.46	−0.32	−0.52	−0.66	−0.13	279	0.24	1.04	−0.67	0.29	0.88
40	−0.29	0.49	−0.29	0.18	0.67	100	−1.56	−0.48	−0.29	0.55	0.33	160	0.15	0.04	−0.23	−0.60	0.20	220	−0.50	−1.72	−0.90	0.53	−1.04	280	0.65	0.14	0.15	0.47	−0.54
41	−0.40	−0.68	−0.26	−0.48	0.22	101	−1.28	−0.82	0.90	0.91	−1.45	161	−0.19	0.47	−0.36	0.35	0.18	221	−0.59	−1.00	0.70	−0.34	−1.30	281	0.13	−0.61	1.02	0.86	−0.83
42	0.11	0.04	−0.24	−0.30	0.12	102	0.00	−0.52	0.05	−0.12	−0.23	162	−0.02	−0.08	0.15	0.55	−0.17	222	−1.15	−0.22	0.77	0.86	−0.84	282	−0.40	0.15	0.37	0.50	0.08
43	0.38	0.04	−0.12	0.13	0.13	103	−0.37	−2.30	−0.55	−0.38	−1.75	163	−0.05	−0.15	0.09	−0.15	0.27	223	−3.43	−2.44	3.43	4.67	−3.02	283	0.04	−0.09	0.05	−0.36	0.45
44	−0.58	0.72	0.40	0.33	0.68	104	0.64	0.13	0.16	0.44	−0.56	164	−0.04	1.64	0.34	0.67	1.03	224	−0.38	0.02	0.16	1.04	0.56	284	−1.32	0.23	1.19	0.10	−0.06
45	−0.85	0.30	0.29	0.76	0.02	105	−0.02	−0.19	0.51	0.79	−0.04	165	−0.01	−1.58	−0.22	−0.81	1.09	225	−0.16	−0.66	0.06	0.61	−0.55	285	−0.22	−0.23	0.38	0.76	−0.36
46	0.42	0.11	−0.28	0.09	−0.44	106	−0.89	−0.88	0.34	0.96	−1.54	166	0.50	−0.09	−0.16	0.09	0.00	226	1.12	0.79	0.99	1.23	−0.89	286	0.10	0.09	0.67	0.46	0.27
47	−0.18	−0.49	0.54	0.91	−0.27	107	0.00	−0.47	0.26	0.46	−0.54	167	0.01	−0.52	0.14	−0.08	−0.26	227	−0.20	0.38	0.07	0.53	0.36	287	−0.16	−0.57	−0.25	−0.15	0.28
48	−1.12	2.10	1.23	0.87	1.41	108	0.89	0.32	0.21	−0.01	0.48	168	0.11	−0.31	−0.72	−1.42	0.31	228	−0.37	−2.43	−0.07	−0.50	−1.86	288	−0.96	0.18	0.06	0.99	−0.83
49	0.33	0.26	0.56	0.62	−0.60	109	−0.37	0.10	0.14	0.76	0.67	169	0.12	0.36	−0.45	−0.28	0.63	229	0.40	−0.34	−0.41	−0.60	−0.16	289	−0.11	−0.36	0.52	0.85	−0.18
50	−0.30	−0.75	0.17	1.36	−1.61	110	−0.09	−0.10	−0.13	−0.17	−0.04	170	0.10	−0.13	0.05	0.14	0.04	230	0.39	0.47	1.37	2.60	−1.42	290	1.05	1.35	0.05	0.29	0.41
51	−1.34	−1.37	0.88	0.28	−0.61	111	−1.62	−0.82	1.20	0.89	−1.43	171	0.18	−0.47	0.42	0.72	−0.64	231	0.04	0.27	0.00	0.12	0.47	291	−0.74	−0.37	−0.10	1.81	0.93
52	0.20	−0.38	0.15	0.08	−0.45	112	−0.34	1.17	−0.24	−1.16	2.08	172	0.28	−0.38	0.18	0.17	−0.47	232	−0.16	−0.71	0.13	1.10	−1.37	292	−0.48	−2.30	−0.72	−0.40	−1.24
53	−0.03	−0.28	0.58	0.76	−0.09	113	0.25	−0.62	0.07	0.62	−0.86	173	−1.62	2.07	0.57	0.82	1.46	233	0.81	0.50	0.28	0.39	0.18	293	1.21	−1.53	−0.07	−0.36	0.22
54	−0.02	0.40	0.54	0.95	−0.05	114	0.03	−0.47	−0.06	0.24	−0.45	174	−1.20	−0.60	0.77	1.08	−1.34	234	−0.07	−0.10	−0.11	0.57	−0.08	294	−0.09	−0.31	0.49	0.84	−0.17
55	−0.05	0.14	−0.18	0.02	0.05	115	−0.15	−0.20	0.14	0.27	−0.06	175	0.07	−0.43	0.72	0.69	−0.18	235	0.77	0.16	0.22	0.70	−0.48	295	0.01	−0.12	0.03	−0.15	0.18
56	0.05	0.11	0.07	0.07	0.10	116	−0.24	−0.40	−0.17	0.26	−0.68	176	−1.47	−0.97	1.96	3.18	−0.99	236	0.64	−0.10	−0.29	−0.04	−0.02	296	1.56	−0.04	0.09	−0.33	1.22
57	−0.43	−0.21	−0.02	0.55	−0.36	117	0.62	0.10	0.15	0.43	−0.58	177	−0.38	−0.54	0.44	1.15	−0.56	237	−0.06	1.81	0.09	−0.19	1.82	297	−0.61	0.69	0.37	0.24	0.76
58	0.26	−2.42	−0.49	−0.12	−2.24	118	0.04	−1.14	0.01	0.12	−0.36	178	−0.34	−0.15	0.47	0.81	−0.11	238	−0.29	−0.61	−0.02	0.58	−0.88	298	−1.23	0.16	0.77	1.01	−0.55
59	0.19	0.24	−0.14	0.03	0.09	119	−0.29	−0.48	0.28	0.75	−0.25	179	0.04	0.99	0.51	0.69	0.50	239	−0.69	−0.42	0.73	1.49	−0.55	299	−0.41	0.46	0.35	0.43	0.47
60	1.04	0.73	−1.04	−1.13	0.85	120	0.25	−0.11	0.21	0.37	−0.21	180	−1.04	−0.08	0.54	1.37	−0.70	240	0.12	−0.07	0.06	−0.38	0.47	300	−0.79	−0.09	−0.43	0.60	1.00

), and 2020 (

Table A4. Regression results of GWR model in 2020.

ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI	ID	PD	LPI	PAFRAC	IJI	MSIEI
1	−0.76	−0.09	0.82	1.49	−0.21	61	−0.52	−0.32	0.37	0.89	0.18	121	0.64	0.43	0.10	0.60	−0.20	181	−0.71	0.31	0.49	0.56	−0.19	241	−0.51	−2.11	0.42	0.25	−1.56
2	0.17	0.01	−0.55	−0.23	0.03	62	−0.46	−0.01	0.17	0.14	0.07	122	0.50	2.29	0.34	2.88	−1.09	182	0.23	0.52	0.29	0.58	−0.07	242	0.08	0.11	0.08	0.27	0.26
3	0.15	−0.32	0.74	0.60	−0.60	63	−0.03	−0.27	0.80	0.91	−0.22	123	−0.15	−0.10	0.13	0.26	0.03	183	0.19	−0.23	0.70	0.93	−0.70	243	−0.04	−0.62	0.01	0.92	−1.02
4	−0.57	−1.04	1.06	0.94	−1.62	64	0.01	0.22	0.01	0.21	0.35	124	−0.34	−0.35	0.62	0.76	−0.48	184	0.32	−0.43	0.21	0.87	−0.95	244	−0.56	0.90	1.08	1.39	−0.34
5	−0.28	1.17	0.10	0.24	1.24	65	0.04	−0.04	0.19	0.01	0.24	125	−0.15	−0.04	−0.02	0.68	0.26	185	−0.16	−0.24	0.13	0.73	−0.39	245	−0.48	−0.96	0.65	0.71	−1.27
6	1.16	3.40	0.45	−0.97	0.87	66	−0.36	0.75	−0.31	−0.14	1.27	126	−0.08	1.19	0.29	0.51	0.58	186	−0.19	0.34	0.26	0.76	0.21	246	−0.12	−0.25	0.09	0.00	0.10
7	−1.64	−0.26	1.79	2.38	−0.81	67	−0.05	−0.47	0.46	0.23	−0.35	127	−0.31	−0.43	0.42	0.68	−0.44	187	0.14	−0.10	−0.11	−0.02	0.61	247	0.45	−0.37	−0.06	0.09	−0.57
8	−0.05	0.09	0.09	0.04	0.00	68	0.67	0.29	0.12	0.68	−0.39	128	−0.52	1.10	0.18	0.68	1.06	188	−0.31	−4.09	−0.25	−1.13	−3.20	248	−0.69	0.49	0.57	0.43	0.30
9	0.19	0.35	−0.38	−0.25	0.52	69	−0.89	0.01	1.26	0.62	−0.22	129	−0.14	0.82	0.06	0.38	0.66	189	−0.17	−0.23	−0.02	−0.03	0.01	249	−0.20	0.31	0.58	0.73	−0.01
10	0.69	0.70	0.06	0.44	0.12	70	0.20	−0.02	−0.76	−0.28	0.32	130	−0.74	−1.11	−0.11	0.02	−0.02	190	0.03	−0.15	0.27	0.41	0.13	250	0.00	0.87	0.22	0.42	0.67
11	0.42	−0.46	0.26	−0.03	−0.78	71	−1.11	−0.28	0.28	0.41	0.37	131	0.25	−0.06	0.07	0.18	0.00	191	−0.12	−0.22	0.06	−0.25	0.22	251	−0.16	−0.13	0.12	0.21	−0.05
12	1.22	0.08	−0.61	−0.69	0.70	72	0.60	−0.55	−0.94	−0.71	−0.11	132	−0.24	−0.76	0.10	0.43	−1.02	192	−0.52	−0.76	−0.90	−0.08	−0.16	252	−0.15	−0.72	−0.04	−0.21	−0.30
13	0.67	1.75	0.51	−0.40	−0.40	73	−0.01	0.52	−0.32	−0.25	0.77	133	−0.85	0.76	0.51	1.20	0.04	193	−0.29	−0.29	0.63	0.98	−0.26	253	−0.16	−0.02	0.17	0.27	0.07
14	0.40	0.08	−0.95	−0.33	0.23	74	−0.19	−0.28	0.36	0.64	−0.27	134	−0.25	0.81	−0.37	−0.21	1.42	194	−1.03	0.99	0.38	1.52	0.69	254	−1.13	1.06	1.31	0.27	0.64
15	0.08	0.93	−0.23	1.14	−0.12	75	−0.31	−3.44	0.09	−0.07	−1.23	135	0.27	−0.20	−0.34	0.07	−0.53	195	−0.23	0.09	0.07	0.46	0.11	255	−1.84	−1.49	1.81	3.31	−2.02
16	0.18	−0.04	0.11	0.24	0.00	76	−0.61	0.88	−0.20	0.08	1.49	136	0.16	−0.53	1.29	1.37	−1.30	196	−0.73	0.69	0.50	0.67	0.18	256	−0.01	−0.40	0.23	0.22	−0.25
17	0.03	0.04	−0.15	−0.01	0.19	77	−0.71	−1.58	1.07	1.05	−2.11	137	0.21	0.05	0.53	0.43	−0.24	197	1.22	1.38	1.23	2.41	−2.32	257	0.40	−0.31	−0.67	−0.24	0.77
18	0.52	2.72	0.71	3.58	−1.73	78	−0.08	−0.48	0.05	0.09	−0.32	138	−0.11	−0.24	0.11	−0.40	0.33	198	−0.09	−0.47	0.09	0.12	−0.32	258	0.36	−0.71	−0.68	−0.31	0.44
19	−0.25	0.98	0.06	0.36	0.98	79	−0.18	−0.55	−0.04	−0.11	−0.16	139	−0.18	0.13	0.08	0.33	0.44	199	−0.31	1.12	1.02	1.07	0.26	259	0.04	−0.16	0.26	0.44	−0.04
20	−0.01	−0.26	0.60	0.91	−0.61	80	−1.08	−0.11	0.15	0.91	−0.45	140	−0.73	0.27	0.69	0.74	−0.59	200	−0.59	1.18	0.59	1.11	0.78	260	−0.50	1.13	0.81	0.44	0.72
21	−1.58	1.59	1.29	1.70	0.25	81	0.12	−0.16	0.11	−0.32	0.21	141	0.38	−0.12	0.17	−0.02	0.43	201	−0.29	0.18	0.39	0.53	−0.05	261	−0.07	−0.48	0.02	0.02	−0.30
22	0.56	−0.11	−0.51	−0.39	0.57	82	−0.12	−0.55	0.21	0.56	−0.54	142	−1.02	−0.45	0.56	2.07	−2.08	202	−0.38	−3.32	−0.13	−0.68	−2.61	262	−0.10	−0.06	0.13	0.28	0.10
23	−0.19	1.66	0.37	1.54	0.11	83	0.13	1.69	1.41	0.38	−0.48	143	0.44	−0.67	−0.50	−0.53	−0.17	203	0.75	0.54	−0.22	−0.23	0.66	263	−0.01	−0.39	0.03	0.48	−0.47
24	−0.43	−0.30	0.03	−0.16	−0.17	84	−0.28	−0.50	−0.37	−0.43	0.12	144	0.03	−0.13	−0.02	0.01	0.04	204	0.34	0.54	0.08	0.11	0.68	264	−0.16	−0.19	0.14	−0.07	0.26
25	−0.48	−1.55	0.03	−0.23	−0.75	85	−1.37	2.12	0.60	1.13	1.95	145	−0.19	0.08	0.12	0.18	0.20	205	0.05	−0.43	−0.23	−0.67	−0.10	265	−0.16	−0.27	0.02	−0.39	0.16
26	−0.99	1.60	0.49	0.74	1.65	86	−0.06	−0.48	0.01	0.50	−0.50	146	0.24	−0.43	0.34	0.60	−0.73	206	0.19	−0.11	−0.07	0.24	0.14	266	0.79	0.98	1.46	3.29	−2.92
27	0.24	−0.48	0.48	0.75	−0.84	87	−1.01	−1.67	−0.45	−0.10	−1.48	147	−0.33	0.34	0.15	0.04	0.42	207	−0.45	−0.01	0.22	0.13	0.19	267	0.12	−0.18	−0.38	−0.29	0.07
28	0.03	−0.03	−0.41	0.06	0.43	88	−0.13	−0.28	−0.04	−0.37	0.13	148	−0.02	−0.45	0.28	0.21	−0.32	208	−0.43	−0.44	−0.52	0.00	0.27	268	−0.13	1.30	0.06	−0.09	1.45
29	−0.27	0.25	0.14	0.12	0.33	89	1.02	−0.38	−0.62	−0.60	0.18	149	−0.21	−0.07	−0.04	−0.30	0.43	209	−0.52	0.88	0.79	0.60	0.11	269	−0.20	0.76	−0.49	−0.33	1.20
30	−0.55	2.75	0.42	−0.28	2.71	90	0.28	0.34	−0.04	0.02	0.70	150	−0.88	−0.18	0.53	0.49	−0.08	210	−1.26	−0.50	0.30	1.13	−0.96	270	−1.95	−0.79	1.46	3.96	−3.22
31	0.63	−0.14	−0.53	−0.39	−0.11	91	0.58	−0.53	−1.33	−0.99	0.25	151	0.22	−0.68	−0.41	−0.28	0.20	211	0.12	−1.56	−0.59	−0.46	−0.28	271	−0.03	0.10	−0.09	0.02	0.31
32	−0.48	0.30	0.19	0.67	0.39	92	−0.34	−0.87	0.22	1.36	−1.71	152	−0.13	0.28	−0.14	0.00	0.47	212	−1.23	3.63	1.21	3.01	−0.01	272	0.20	−0.29	0.02	0.23	−0.17
33	−0.09	0.10	−0.15	−0.03	0.14	93	−0.15	−0.25	0.17	1.18	−0.36	153	0.36	−0.18	−0.37	0.02	−0.26	213	−0.50	2.96	0.86	3.74	−1.46	273	−0.05	−0.32	0.16	0.00	−0.02
34	1.05	−0.15	−0.56	−0.58	0.50	94	−0.47	−0.86	0.07	1.01	−1.38	154	−0.08	−4.06	−0.23	−1.19	−3.02	214	0.56	0.11	−0.93	−0.78	0.35	274	−0.07	−0.10	0.40	0.59	0.06
35	−0.06	−0.29	0.40	0.52	−0.25	95	−0.36	−0.32	0.56	0.96	−0.27	155	−0.07	0.29	−0.10	0.14	0.35	215	−0.52	0.87	0.16	0.77	0.69	275	−0.73	−0.02	0.61	0.74	−0.70
36	−2.68	−0.03	2.87	4.38	−0.78	96	−0.11	−0.47	0.15	0.13	−0.29	156	−0.22	0.08	0.13	0.17	0.25	216	0.03	0.63	−0.11	0.04	0.65	276	−0.54	1.15	0.41	0.79	0.62
37	0.07	−0.14	0.25	0.39	−0.14	97	0.05	−0.30	−0.11	0.09	−0.29	157	−0.49	−1.35	−0.79	−0.22	−0.54	217	−0.39	0.65	0.17	0.30	0.80	277	0.07	−0.10	−0.12	−0.06	0.02
38	−0.40	1.73	0.17	0.29	1.95	98	−0.05	−0.02	0.06	0.11	0.08	158	0.21	−0.12	−0.06	0.42	−0.28	218	−0.91	−0.31	0.89	1.75	−0.66	278	−0.49	−1.07	0.24	−0.01	−0.69
39	0.10	−0.34	−0.49	−0.94	−0.07	99	−1.30	−0.26	0.65	1.44	−1.23	159	−0.60	−0.53	0.83	1.07	−0.62	219	0.73	−0.42	−0.72	−0.70	−0.28	279	0.66	2.06	0.11	1.03	0.59
40	−0.17	1.15	0.62	0.77	0.37	100	−0.59	0.34	0.24	0.51	0.42	160	−0.28	1.02	0.15	0.17	1.11	220	−0.39	−1.39	−0.80	0.17	−0.65	280	0.54	0.28	0.11	0.45	−0.37
41	−0.43	−0.71	−0.10	−0.09	−0.05	101	−0.30	−0.05	0.69	0.49	−0.52	161	−0.28	0.29	0.17	0.29	0.11	221	−0.81	−0.47	0.97	0.17	−0.73	281	0.13	−0.45	0.78	0.66	−0.67
42	−0.19	1.15	0.04	0.20	1.21	102	−0.10	−0.49	0.07	0.13	−0.33	162	0.20	0.02	−0.30	0.05	0.04	222	−1.12	0.20	0.83	1.09	−0.65	282	−0.28	0.14	0.38	0.53	−0.04
43	0.21	0.00	0.05	0.18	0.06	103	−0.46	−2.30	−0.48	−0.40	−1.61	163	−0.04	−0.26	0.12	−0.11	0.08	223	−1.18	−0.25	1.13	1.73	−0.81	283	0.05	−0.22	0.11	−0.34	0.24
44	−0.34	0.20	0.17	0.13	0.29	104	0.52	0.26	0.11	0.45	−0.41	164	0.26	0.59	0.47	0.90	−0.16	224	−0.32	0.03	0.16	0.92	0.45	284	0.03	0.94	1.23	−0.25	0.62
45	−1.17	−0.10	0.53	1.21	−0.77	105	−0.25	−0.34	0.88	1.30	−0.19	165	−0.50	−2.72	0.45	0.35	−1.48	225	−0.20	−0.67	0.29	0.68	−0.78	285	−0.12	−0.37	0.22	0.57	−0.29
46	0.31	0.03	−0.29	0.10	−0.36	106	−0.22	−0.22	0.44	0.36	−0.45	166	0.34	−0.07	−0.04	0.10	0.02	226	1.31	1.80	0.97	1.34	−1.38	286	−0.01	0.04	0.86	0.81	0.18
47	−0.20	−0.30	0.72	1.00	−0.26	107	0.19	−0.31	0.10	0.63	−0.65	167	−0.06	−0.47	0.14	0.12	−0.31	227	−0.07	1.35	0.93	1.00	0.26	287	−0.18	−0.90	−0.20	−0.15	0.13
48	−1.62	2.41	1.54	1.31	1.42	108	0.73	0.43	0.16	0.01	0.52	168	0.07	−0.37	−0.53	−1.13	0.05	228	−0.53	−2.51	−0.02	−0.35	−1.84	288	−0.43	−0.26	0.23	0.45	−0.51
49	0.45	1.64	0.59	0.22	−0.33	109	−0.32	0.06	0.21	0.53	0.52	169	0.07	0.62	−0.58	−0.48	1.00	229	0.54	−0.58	−0.60	−0.63	−0.36	289	−0.23	−0.29	0.76	1.09	−0.20
50	−0.34	−0.79	0.12	1.09	−1.34	110	−0.51	0.78	0.26	0.61	0.77	170	0.06	−0.14	0.10	0.06	0.01	230	0.28	0.52	1.30	2.81	−1.96	290	0.80	1.14	−0.02	0.17	0.58
51	−1.05	−1.41	0.84	0.49	−0.78	111	−0.13	−0.03	0.74	0.34	−0.42	171	0.20	−0.51	0.82	0.71	−0.96	231	−0.10	0.29	−0.02	0.19	0.54	291	−0.57	−0.24	−0.09	1.63	0.99
52	−0.11	0.12	−0.20	−0.09	0.28	112	−0.02	1.71	1.11	−1.10	1.82	172	0.03	0.11	−0.12	0.04	0.24	232	−0.20	−0.61	0.03	0.90	−0.97	292	−0.45	−2.37	−0.60	−0.47	−1.46
53	−0.14	−0.23	0.84	1.04	−0.12	113	0.11	−0.38	0.07	0.57	−0.40	173	−1.68	3.15	1.05	0.96	2.18	233	0.68	0.58	0.20	0.32	0.28	293	0.55	−0.85	0.07	−0.16	0.16
54	0.28	0.10	0.38	0.86	−0.49	114	−0.04	−0.55	0.05	0.31	−0.47	174	−1.14	0.23	0.92	1.15	−0.60	234	−0.52	0.01	0.57	1.28	−0.18	294	−0.27	−0.31	0.78	1.19	−0.23
55	−0.27	0.75	0.06	0.47	0.66	115	0.11	−0.20	−0.20	0.02	−0.02	175	0.16	−0.17	0.73	0.71	−0.06	235	0.68	0.33	0.17	0.59	−0.22	295	0.02	−0.24	0.06	−0.17	0.01
56	0.18	0.02	−0.49	−0.03	−0.14	116	−0.33	−0.30	−0.48	−0.21	−0.30	176	−1.09	−0.75	1.40	2.61	−0.89	236	0.43	−0.06	−0.15	0.00	0.03	296	0.34	0.24	0.19	0.05	0.52
57	−0.91	0.63	0.38	1.40	0.39	117	0.51	0.23	0.11	0.44	−0.44	177	−0.40	−0.52	0.78	1.16	−0.70	237	0.00	1.59	0.45	0.48	0.99	297	−0.30	0.15	0.14	0.10	0.32
58	−0.45	−0.13	−0.17	0.20	−0.55	118	−0.17	−1.65	0.15	0.26	−1.01	178	−0.38	−0.17	0.22	0.42	0.05	238	−0.35	−0.66	0.07	0.59	−0.79	298	−1.27	−0.02	0.80	1.34	−1.02
59	0.28	0.10	−0.07	0.07	−0.16	119	−0.65	−0.56	0.73	1.00	−0.54	179	0.18	−0.01	0.66	1.02	−0.62	239	−0.16	0.05	−0.09	0.59	−0.08	299	−0.17	0.02	0.11	0.19	0.19
60	0.78	0.60	−0.90	−0.81	0.82	120	0.16	0.32	−0.01	0.36	0.35	180	−0.89	0.45	0.60	1.36	−0.41	240	0.04	−0.21	0.15	−0.37	0.33	300	−0.62	−0.03	−0.42	0.29	1.10

) were obtained, respectively. The GWR model demonstrated robust performance, with R² values of 0.77, 0.78, 0.79, and 0.80 for the four years, respectively, accompanied by progressively decreasing AIC values (1182.45, 1155.30, 1121.33, and 1120.78). All regression coefficients were statistically significant at the α = 0.05 level (p < 0.05). Spatial distribution maps of these coefficients for each driving factor were generated (Figure 8), which visually present the GWR model results, illustrating the relationships between various land-use landscape pattern indices and carbon emissions across China’s 300 cities from 2005–2020. These results clearly demonstrate significant regional variations in how different landscape pattern indices influence carbon emissions.

For PD, most cities presented negative correlations, suggesting that increased fragmentation is associated with reduced emissions. Strong negative correlations clustered in the Bohai Rim urban agglomeration, mid-Yellow River Basin, and Chengdu-Chongqing urban cluster, whereas positive correlations predominated in the northwestern, northeastern, and southwestern regions, as well as the Huai River Basin. Temporally, negative correlation areas expanded from 2005 to 2020, particularly in northeastern regions, which transitioned from positive to negative relationships, whereas northwestern areas showed strengthening positive correlations and southwestern regions demonstrated weakening positive trends.

LPI generally showed weak negative correlations with emissions, with stronger negative associations concentrated near the Henan–Shandong–Anhui border and Shanghai–Jiangsu–Zhejiang coastal areas, indicating that greater patch adjacency facilitates emission reduction. Positive correlations clustered in central Inner Mongolia, northern Hebei, and the Beijing–Tianjin region, with another positive zone in the southeast. Over time, northern positive areas expanded into Gansu and Liaoning, whereas some southeastern positive areas converted to negative correlations.

PAFRAC was predominantly positively correlated with emissions, suggesting that more complex patch shapes are associated with higher emissions. The negative correlations were concentrated in Anhui–Jiangsu–Shanghai and northeastern China, whereas strong positive correlations appeared along the Bohai Rim to the mid-Yellow River corridor and southwestern regions. During the 2005–2020 period, negative correlation zones decreased, particularly in the northeast, with expanding positive correlations along the Bohai–Yellow River axis.

IJI was positively correlated in most cities, indicating that more aggregated patch distributions (lower IJI) were correlated with reduced emissions. Strong positive correlations centered on the Loess Plateau core and the Beijing–Tianjin–Hebei region expanded over time. The negative correlation clusters in Northeast China, the Yangtze River Delta core, and the Hunan–Jiangxi regions gradually diminished, with many northeastern cities transitioning to positive correlations.

MSIEI generally demonstrated weak negative correlations, suggesting that more even patch-type distributions are associated with lower emissions. In 2005, strong positive correlations clustered in the Mongolia Plateau–North China Plain transition zone and the Huang–Huai–Hai Plain junction of Jiangsu–Anhui–Shandong. By 2020, these positive areas had expanded across Inner Mongolia and Northeast China but weakened in intensity, whereas southern positive correlation zones had shrunk. Overall, the negative correlation intensities gradually weakened from 2005 to 2020.

3.5. Heterogeneity of the Impacts Under Different Levels of Economic Development

To analyze the spatial heterogeneity characteristics of how land-use landscape patterns influence carbon emissions in detail, as revealed by the GWR model, this study established a core–periphery analytical framework based on economic development levels to systematically compare these impacts. Using the annual median per capita GDP as the threshold, we categorized the study area into developed cities (DCs) and underdeveloped cities (UCs). Kernel density estimation was applied to analyze the GWR coefficients for each group, revealing distinct patterns between the economically core and peripheral regions. This classification system provides a scientific basis for formulating regionally differentiated low-carbon land-use strategies.

As shown in Figure 9, both commonalities and differences exist in how land-use morphology affects carbon emissions between developed cities (red curves) and underdeveloped cities (green curves). All the GWR coefficient distribution diagrams exhibit multimodal characteristics, indicating significant variations in the relationships between landscape pattern indices and carbon emissions across different cities. The distributions for underdeveloped cities are typically more concentrated with higher peaks, suggesting more consistent impacts of landscape patterns on emissions. In contrast, developed cities have broader distributions, reflecting greater variability in how landscape patterns influence emissions, likely due to more complex influencing factors.

For PD, most years presented coefficient peaks slightly left of zero for both city types, indicating generally stable negative PD-emission relationships. However, DCs exhibited more dispersed peak distributions with stronger leftward shifts, suggesting more pronounced negative impacts of fragmentation on emissions in developed urban contexts. This negative effect may be significantly associated with the intensity of low-carbon pilot policies.

LPI analysis revealed coefficient peaks to the left of zero for UCs across most years, reflecting consistent negative relationships. DCs showed even stronger leftward peak positions, confirming that greater patch adjacency more effectively reduces emissions in developed cities. Notably, from 2005 to 2020, DCs’ peaks transitioned from being lower to higher than UCs’ peaks, with distributions evolving from multimodal to unimodal, indicating increasing consensus about the negative emission effects of LPI in developed areas. After 2015, the emission reduction effect of DCs’ LPI showed marked enhancement, which coincides with the implementation timeline of the carbon emission trading pilot program covering industrial land.

PAFRAC coefficients for both city types peaked right of zero, establishing stable positive shape complexity–emission relationships. However, DCs’ peaks declined relative to UCs’ over time, suggesting growing variability in how shape complexity affects emissions in developed cities amid more diverse influencing factors. This suggests that mandatory spatial regulations can partially offset the negative impacts of extensive development.

The IJI index in UCs consistently exhibited a minor peak slightly to the right of zero, indicating a stable positive correlation between landscape aggregation and carbon emissions. In contrast, DCs not only demonstrated rightward peak shifting, but also displayed pronounced multimodal distribution characteristics and greater dispersion. This reveals significantly higher variability in how patch aggregation patterns influence carbon emissions within DCs. These findings reflect the gradient disparities in carbon governance resulting from unbalanced policies such as environmental protection inspections.

MSIEI coefficients for UCs exhibited multiple-peaked or zero-centered distributions, suggesting context-dependent neutral relationships. This corresponds to the following two distinct policy response patterns: (1) for cities that participated early in ecological compensation programs, where increased evenness resulted from ecological restoration, MSIEI showed negative correlations with carbon emissions; (2) for cities relying on traditional land finance, where evenness reflected disorderly urban expansion, leading to mixed cropland and built-up areas, MSIEI demonstrated positive correlations with emissions. DCs’ distributions predominantly peaked near or left of zero, establishing generally negative evenness–emission relationships where greater patch-type uniformity is associated with lower emissions.

4. Discussion

4.1. Impact of Land-Use Changes on Landscape Patterns and Carbon Emissions

The findings of this study reveal significant transformations in China’s land-use patterns and their cascading effects on landscape configurations and carbon emissions. The observed decline in cropland area, coupled with increases in forest and water body coverage, aligns with the outcomes of national ecological restoration initiatives such as the Grain for Green Program. Simultaneously, rapid urbanization has driven substantial expansion of built-up areas, corroborating regional-scale land-use change patterns documented in prior studies [77]. These land-use transitions have fundamentally reshaped ecosystem functions, structural attributes, and landscape pattern dynamics.

Analysis of landscape pattern indices demonstrates that most cities experienced increased fragmentation (PD) and diversity (MSIEI), alongside reduced contiguity (LPI), aggregation (IJI), and complexity (PAFRAC). These trends are consistent with our initial hypothesis that urban expansion would fragment natural and agricultural landscapes, while standardized development models would simplify morphological complexity. The mechanisms driving these changes include dispersed construction land development, mixed urban–rural fringe patterns, and the adoption of grid-based urban planning frameworks. Notably, Shandong and Jiangsu provinces exhibited divergent trends—decreased fragmentation and enhanced aggregation—attributable to early implementation of spatial governance policies like the “Three Zones and Three Lines” strategy. This deviation underscores the moderating role of stringent land-use regulations, supporting our hypothesis that planning interventions can alter landscape–carbon relationships.

The temporal evolution of carbon emissions further validates our research framework. The rapid emission growth from 2005–2015, driven by urban sprawl and industrial expansion, mirrors findings from studies linking built-up area dynamics to carbon output [39,78]. The subsequent stabilization and decline post-2015 reflect the efficacy of China’s energy-saving policies and industrial upgrades, consistent with national-scale assessments of decarbonization efforts [79]. Spatial disparities in emission hotspots—persistent in eastern coastal hubs but emerging in central/western regions—align with the literature on industrial relocation effects, confirming the spatial non-stationarity hypothesized in our study.

4.2. General Characteristics and Spatiotemporal Heterogeneity of the Relationship Between Landscape Patterns and Carbon Emissions

Our analysis reveals several key patterns in the landscape–carbon emission relationship. Overall, in most cities, patch fragmentation (PD), contiguity (LPI), aggregation (IJI), and diversity (MSIEI) are negatively correlated with carbon emissions, whereas only patch shape complexity (PAFRAC) is positively correlated. These findings largely confirm our initial hypotheses regarding landscape–carbon relationships. Our core findings align with international research on urban form and sustainability. The benefits of compact development resonate with studies conducted in metropolitan areas across Europe and North America [80,81]. Landscapes with high fragmentation and high adjacency often reflect scientifically planned urban spatial organization. Under such configurations, although the number of patches may be greater, optimized functional zoning and infrastructure networks can shorten commuting distances and improve energy efficiency. This finding is consistent with research concluding that compact urban forms can reduce carbon emissions [82]. Complex patch shapes typically indicate irregular boundaries and disordered development, which can lead to excessively extended infrastructure and inefficient public service coverage. In contrast, regularly shaped clustered development is more conducive to implementing low-carbon solutions, such as centralized heating and power supply. This finding aligns with studies suggesting that increased landscape shape complexity may increase carbon emissions [79,80]. The increase in evenness among different patch types can effectively reduce carbon emissions. Similar conclusions have been documented in studies focusing on ecological service zones [83,84]. Evenly distributed green spaces and service facilities can effectively mitigate urban heat island effects and reduce air conditioning energy consumption. Uniform layouts also prevent the overconcentration of energy facilities (e.g., substations), reducing the carbon emissions generated during energy transmission.

However, the exceptional cases observed in certain regions demonstrate that these general patterns are subject to critical boundary conditions. This finding also aligns with our initial predictions regarding the spatial heterogeneity of these effects. Unlike the weakening positive correlations observed in other regions, the association between patch fragmentation and carbon emissions exhibits a strengthening trend in western and northern China. This contrasts with findings from comparable studies on BRICS nations [85], and it is closely tied to the region’s relatively underdeveloped and extensive economic growth model. In the core Yangtze River Delta and Northeast China, PAFRAC and carbon emissions show a negative correlation, contrary to most other cities. Similar findings have been reported in studies of developed cities in Europe and North America [86,87]. This suggests that when morphological regularization and industrial upgrading are implemented simultaneously, conventional positive correlations can be overcome, forming a low-carbon spatial restructuring paradigm guided by planning. The core Loess Plateau and Beijing–Tianjin–Hebei regions show a unique pattern; as patch aggregation increases, so do carbon emissions. This contrasts with most other cities in China. The trend is especially evident in the Loess Plateau. Here, concentrated coal mining areas and monoculture farming zones boost industrial scale efficiency. However, they also strengthen spatial dependence on high-carbon industries. This phenomenon supports Batty’s “spatial path dependence” theory [88].

4.3. Moderating Effect of Economic Development Level

The findings reveal a nonlinear moderating effect of economic development mediated through spatial governance capacity. External factors complicate the use of land-use strategies for sustainable development, particularly given regional disparities caused by uneven economic growth. In developed cities, the more dispersed distribution of GWR coefficients indicates their carbon emissions are influenced by multiple factors, including technology and planning policies. In contrast, less-developed cities rely predominantly on singular land-use characteristics. These results provide strong support for our original hypothesis that developed cities possess superior regulatory capacity for achieving carbon mitigation effects.

Differences in economic development levels first manifest as gradient differences in spatial governance capacity. Developed cities have established comprehensive spatial planning systems. By altering industry–space interactions and infrastructure efficiency, they reshape the carbon regulation pathways of landscape patterns and adjust the intensity of their impact on emissions. The study demonstrates that well-established planning systems can effectively translate landscape advantages into tangible emission reductions, as evidenced by existing research. For instance, certain cities have achieved “fragmented yet efficient” development patterns through functional mixing strategies [89,90]. In developed cities, stronger planning control capabilities and technical support systems enable these areas to fully leverage the carbon regulation potential of landscape patterns. On the one hand, mixed-use industrial land and optimized infrastructure transform patch fragmentation into functional diversity advantages, achieving “fragmented yet efficient” low-carbon development. On the other hand, mature planning systems facilitate scientific layouts of land-use types, ensuring that increased evenness translates into carbon sequestration and emission reduction. For example, functional mixing in central urban areas reduces transportation-related emissions, whereas ecological protection zones maintain appropriate separation. This finding corroborates the “institutional threshold theory” proposed by another scholar [87].

In contrast, underdeveloped cities face constraints such as inadequate planning implementation capacity and homogeneous industrial structures. These challenges make it difficult to overcome the disconnect between form and function in landscape pattern optimization. For example, increased land-use patch fragmentation in some cities stems mainly from the disorderly expansion of construction land rather than functional reorganization. This phenomenon is equally prevalent in developing regions such as India and Sub-Saharan Africa [91,92]. These differences confirm that the carbon reduction potential of landscape optimization can only be effectively realized when a region’s institutional development level (including planning systems, governance capacity, and policy coordination) reaches a critical threshold.

4.4. Research Limitations and Future Directions

This study has several limitations that should be acknowledged. At the data level, the use of 1 km resolution remote sensing datasets may affect the precise calculation of landscape pattern indices, particularly in identifying small patches. Additionally, while focusing on city-level relationships between land-use landscape patterns and carbon emissions, the reliance on statistical yearbooks for emission data fails to fully capture intracity spatial heterogeneity. This study employed per capita GDP as the economic criterion for the core–periphery framework. While this indicator effectively reflects regional disparities in economic development, it carries inherent limitations that warrant discussion. First, the per capita GDP metric may obscure heterogeneity in urban economic structures. For instance, resource-dependent cities and innovation-driven cities could demonstrate fundamentally distinct landscape–carbon relationships despite having similar per capita GDP levels. Second, as a purely economic output measure, it fails to fully capture critical dimensions like governance capacity and institutional quality that substantially influence carbon reduction outcomes. This oversight might lead to biased estimations of economic moderation mechanisms. Alternative approaches using composite indicators—incorporating factors like industrial structure and Human Development Index—could potentially identify more nuanced urban development typologies. Such multidimensional metrics might reveal more complex pathways through which economic factors moderate the landscape–carbon relationship. These methodological differences ultimately shape policy recommendations. The per capita GDP framework tends to favor growth-oriented emission reduction strategies, whereas multidimensional approaches could lead to more integrated spatial governance solutions.

This study reveals fundamental advances with interconnected theoretical and practical implications. The spatial non-stationarity of landscape–carbon relationships challenges the universality of compact city paradigms. The findings demonstrate that the emission effects of fragmentation, aggregation, and other landscape metrics vary systematically according to geographic location and economic–institutional contexts. The results highlight the need to develop regionally tailored land-use strategies that account for local economic conditions, planning capacity, and administrative characteristics. The study demonstrates the spatial heterogeneity and non-stationarity of landscape pattern effects on carbon emissions, as well as their modulation by economic development, thereby supporting and refining key landscape ecology theories.

These significant findings have important implications for future research directions. First, the spatial non-stationarity of landscape effects suggests that the GWR model may systematically misestimate carbon impacts. Advanced machine learning algorithms (e.g., random forest and XGBoost) could be employed to better clarify the complex interaction mechanisms among climate conditions, industrial policies, and landscape patterns. Second, the economic moderation effects reveal an institutional threshold phenomenon. Future research should systematically examine how sensitive the findings are to alternative economic indicators, which would significantly enhance the robustness of the conclusions. Third, predictive modeling approaches such as CA-Markov or system dynamics models could simulate long-term carbon emission responses under various low-carbon scenarios, including different ecological protection redline configurations and industrial upgrade pathways. Finally, integrating high-resolution remote sensing data (e.g., Sentinel-2) with multisource datasets, such as nighttime light and POI (points of interest) data, could significantly improve the accuracy of landscape index calculations. Further development in these aspects will contribute to establishing a comprehensive “pattern-process-policy” analytical framework. This innovative approach will pioneer new pathways for understanding the complex interplay between landscape dynamics and carbon emissions, ultimately enabling more precise decision-support systems for developing regionally tailored sustainable land-use planning strategies.

5. Conclusions

This study establishes an analytical framework encompassing regional land-use landscape pattern analysis, spatiotemporal correlations between landscape patterns and carbon emissions, and core–periphery economic disparities across 300 Chinese cities. By calculating city-level land-use landscape pattern indices (2005–2020) and applying geographically weighted regression (GWR) models, we systematically examined regional variations in how landscape patterns influence carbon emissions. Cities were categorized into developed and underdeveloped groups based on median per capita GDP to analyze the moderating effects of economic development. Key findings include the following:

The implementation of Grain-for-Green policies and rapid urbanization has generally increased landscape fragmentation (PD) and diversity (MSIEI) while reducing contiguity (LPI), aggregation (IJI), and complexity (PAFRAC). This results from dispersed urban expansion, mixed urban–rural development, and standardized planning. However, the Shandong and Jiangsu Provinces demonstrated inverse trends due to stringent spatial governance, highlighting effective regulatory interventions.

Carbon emissions exhibited the following phased spatiotemporal patterns: rapid growth during early urban/industrial expansion (2005–2015), followed by stabilization and gradual decline after 2015 through energy-saving policies, industrial upgrades, renewable energy adoption, and ecological conservation, aligning with China’s ecological civilization initiatives.

Landscape–emission relationships showed marked spatial heterogeneity, challenging conventional assumptions such as “compactness ensures low carbon”. Coordinated landscape optimization with industrial/ecological upgrades can establish innovative low-carbon paradigms; however, excessive concentrations may increase emission intensity through path dependence. These findings underscore the need for regionally tailored land-use strategies that consider economic conditions, planning capacity, and administrative characteristics.

Economic development significantly moderates landscape–carbon relationships through spatial governance capacity. Developed cities achieve emission reduction via intensive planning, whereas underdeveloped cities face “quantity-over-quality” expansion patterns. This dichotomy reveals the critical interplay between regional development, industrial policies, and ecological protection during urbanization. Future strategies must synchronize landscape optimization with industrial transition and ecological protection.

A comparative analysis with case studies from other regions in China reveals that the impacts of landscape patterns on carbon emissions exhibit significant variations across different research scales. This necessitates spatially differentiated solutions to support China’s carbon peak and neutrality goals. For instance, Eastern regions should further advance governance networks for functional mixing. Western regions need to establish “ecology-renewable energy” integrated zones to balance conservation and development. Central resource-intensive regions must innovate mechanisms for industrial land transformation to break carbon lock-in pathways.

This study advances the understanding of the complex “landscape pattern-carbon emission-economy” nexus, offering evidence-based insights for regional sustainable development.