Spatiotemporal Variability in Municipal Solid Waste Production and the Determinants in Hefei’s Core Urban Districts

Abstract

Precision in discerning the spatiotemporal dynamics of municipal solid waste (MSW) production and its drivers is pivotal for informing the seasonal management and recycling of urban waste streams. This investigation zeroed in on Hefei’s central urban zone, deploying a nuanced principal component analysis and geographically and temporally weighted regression (PCA-GTWR) to quantify the sway of the environmental, economic, and living standard variables on the MSW generation patterns. The methodology unfolded across four main phases: (1) leveraging nocturnal light data to approximate the MSW output; (2) employing spatial autocorrelation to probe the variable trends and spatial interdependencies of the waste generation; (3) harnessing principal component analysis to pinpoint critical determinants and preprocess these as inputs for the GTWR model; (4) mapping the GTWR outcomes to elucidate the differential impacts of various factors on the waste production patterns. Key findings reveal a distinctively polycentric MSW distribution, with high-density areas anchored in the urban core and diminishing intensities beyond the secondary periphery. The trio of socioeconomic variables, residents’ living standard variables, and natural variables emerge as pivotal, with the PCA-GTWR offering a vivid spatial delineation of their effects. Notably, socioeconomic growth exerts a pronounced positive influence in more affluent quarters, residential standards bear greater relevance in burgeoning urban sections than in the established core, and environmental influences wield the least sway, ebbing and flowing with the seasons. These insights demystify the undercurrents shaping the MSW production in urban China, serving as a strategic compass for waste minimization initiatives and policy formulation.

1. Introduction

In the throes of an unprecedented urban metamorphosis, China has witnessed a phenomenal escalation in both the magnitude and velocity of its urbanization processes [1]. This has catalyzed massive urban infrastructural advancements while concurrently aggregating considerable populace densities. Nevertheless, the rapid urban sprawl has precipitated an acute tension between the production of MSW and the imperatives of environmental sustainability. An emergent “garbage siege” scenario has beset approximately two-thirds of Chinese cities [2,3]. Notably, Hefei—a vital sub-central city within the burgeoning Yangtze River Delta economic region—generated a staggering 1.8623 million tons of MSW in the year 2019 alone. This statistic is accompanied by an annual growth rate of approximately 12.22%, accentuating already extant environmental apprehensions [4,5]. Hence, the imperative to rigorously scrutinize the present status and influential determinants of the urban MSW emissions becomes indisputable, as it provides an indispensable foundation for decoding MSW production dynamics and informing sustainable waste management paradigms in urban ecosystems.

Academics both within China and on the global stage have undertaken multifaceted explorations into the influencing variables of MSW [6,7,8]. For instance, Xiao et al. formulated a model governed by multiple regression analyses to discern the principal modalities driving the MSW production across various communities, singling out economic conditions, household structures, and lifestyle behaviors [9]. Similarly, Xu et al. employed path analyses to quantify the relative impacts of a range of metrics—such as the urban scale, per capita disposable income, rate of urbanization, and GDP—on MSW outputs [10]. Wang et al. constructed a hierarchical factorial analysis system and utilized multiple regression techniques to pinpoint key determinants like the total residential area and the enumeration of sanitation machinery [11]. Zhao et al. opted for a more expansive lens by integrating variables like the urban populace, GDP, urban built-up areas, and per capita income and expenditure, and developed a panel data regression model that highlighted provincial disparities in MSW production influencers [12]. Cheng et al. applied the geographic weighted regression (GWR) model to quantitatively assess the impact amplitude of specific variables and to contrast the explanatory powers of the ordinary-least-squares (OLS) and GWR models [13]. Li et al. leveraged geographical detectors to conclude that while socioeconomic parameters exerted the most profound influence on takeaway packaging waste, environmental factors were comparatively inconsequential [14]. Despite these comprehensive examinations, a conspicuous lacuna exists in the current corpus of research. Primarily, existing studies are typically confined to an annual temporal scale, thereby sidelining the potential intricacies attributed to seasonal fluctuations. Additionally, the prevailing scope of inquiry predominantly concentrates on provincial and supra-regional echelons, resulting in an underrepresentation of municipal and sub-regional investigations. A further limitation is the static analytical frameworks predominantly deployed, signaling an unmet need for more dynamic spatial–temporal analyses. To ameliorate these gaps, the present study adopts the PCA-GTWR methodology [15,16,17]. This innovative approach simultaneously mitigates the multicollinearity among independent variables and elucidates the complex spatiotemporal interplays between various influential determinants. Consequently, it promises a more nuanced understanding of the multidimensional forces shaping urban MSW production.

Leveraging nighttime remote sensing light data, the present study endeavors to estimate the quarterly production of MSW in the study area and construct a PCA-GTWR model for exploring the multifaceted determinants of MSW production. Nighttime remote sensing light data encompass the luminosity emanating from structures, vehicular traffic, and maritime vessels. They serve as an objective representation of the human activity intensity and boast the advantage of continuous and consistent data. These data have been extensively applied in fields such as population distribution, land-use analysis, and air pollution assessment [18,19,20,21,22].

From a spatiotemporal vantage point, this investigation systematically demystifies the patterns of MSW production in Hefei’s core urban district locality. Furthermore, the study examines the intricate mechanisms through which natural condition variables, socioeconomic variables, and residents’ living standard variables modulate these patterns. The empirically derived insights offer instrumental guidelines for regulatory policies geared towards optimized domestic waste production and resource utilization strategies.

2. Overview of the Study Area and Data Processing

2.1. Study Area

Hefei’s core urban districts, delineated in Figure 1, manifest a subtropical, monsoonal, humid climatic profile, punctuated by an annual average precipitation of approximately 1000 mm and a mean temperature registering at 15.7 °C. With the geographical sprawl enveloping an estimated 1312.48 km², this segment incorporates the administrative districts of Yaohai, Shushan, Luyang, and Baohe. Despite constituting a mere 10% of Hefei’s total landmass, this region serves as the epicenter of the population density within the city. The year 2022 marked a significant milestone in Hefei’s economic trajectory, with the GDP attaining a staggering CNY 1.2 trillion. Concomitant with this rapid economic ascendancy is an elevated standard of living, which has precipitated a consistent augmentation in the urban MSW production over the past decade. Specifically, the sector experienced an average growth rate of approximately 11.15%. By the closure of the year 2021, Hefei’s MSW production culminated in 18.897 million tons, thereby contributing to 28.60% of the total waste production in Anhui Province. This exponential surge in waste production engenders formidable challenges for its benign and efficacious management, thereby heightening the urgency for efficacious mitigation strategies. Based on the urban development rate of Hefei and the characteristics of the domestic waste generation, the residential area is selected as the main unit of MSW generation.

2.2. Data Sources

The empirical foundation for this study is constituted by a diverse set of data, encompassing variables such as nighttime remote sensing light data, meteorological indicators from the Hefei metropolitan area, population density parameters, per capita GDP figures, per capita residential area measurements, per capita disposable income, total retail sales of consumer goods, and household natural gas consumption metrics. The nighttime remote sensing light data are from the Global Nighttime Light Database (GNLD) and are predicated upon VIIRS/DNB imagery data that are generated by the United States National Oceanic and Atmospheric Administration (NOAA). This dataset, characterized by a 500 m spatial resolution, has been subjected to noise-reduction algorithms and covers the time span from 2015 to 2022 for the Hefei metropolitan area. The data for the years 2015–2021 represent the annual average luminance, whereas the 2022 dataset comprises monthly averages. Meteorological data specific to the Hefei region were sourced from the National Meteorological Science Data Sharing Service Platform. This dataset offers raster data ranging from January to December 2022, and it also features a 500 m spatial resolution. Data pertaining to the population density were gleaned from heat maps available on Baidu Maps, predicated on Location-Based Service (LBS) big data analytics. This dataset spans from January to December 2022 and maintains a 500 m spatial resolution. Data regarding the per capita residential usable area were procured through Python web-scraping techniques from the Lianjia website and encompass granular details such as community names, geographic coordinates, and regional classifications. Metrics on the household natural gas consumption were accumulated through field surveys. Lastly, statistics on the per capita GDP, per capita disposable income, and total retail sales of consumer goods were extracted from official statistical bulletins for the four-quarterly of 2022, pertaining to the Baohe, Shushan, Yaohai, and Luyang districts of Hefei (See Supplementary Materials).

2.3. Estimation of MSW Production Based on Nighttime Remote Sensing Light Data

Nighttime remote sensing light data serve as an objective surrogate for delineating the spatial distribution of human nocturnal activities [23]. A robust correlation exists between the production of MSW and nighttime luminosity, substantiated by extant empirical studies that have utilized nocturnal light data for the estimation of domestic waste output [24]. The annual data for the domestic waste production pertinent to this study were culled from the relevant HEFEI STATISTICAL YEARBOOK.

In an endeavor to ascertain the spatial distribution of the domestic waste production, a regression analysis was conducted correlating the Total Nighttime Luminosity (TNL) raster values from the years 2015 through to 2022 for Hefei’s central urban zone with the metrics of the domestic waste production. Both the post-noise-reduction nighttime luminosity data for the study locus and the associated domestic waste production metrics are tabulated in

Table 1. TNL and MSW production in the study area.

Year	TNL	MSW Production/104 t
2015	105,290	129.16
2016	220,976	144.3
2017	360,009	153.74
2018	506,492	166.31
2019	669,388	186.23
2020	839,997	188.97
2021	1,029,704	204.43
2022	1,255,252	216.93

.

Utilizing Python 3.0, a regression analysis was conducted on data spanning various years, the TNL, and the MSW produced. The results (Equation (1)) demonstrate a significant correlation between the production of MSW and both the year and the TNL. This correlation was statistically significant at the 0.01 level, with an R² value of 0.98.

$$f(x,y)=6.23x^2+109.55x-3.02\times {10}^{-11}{y}^2+1.28\times {10}^{-4}y+0.75$$

(1)

In the equation, f(x,y) represents the production of MSW in ten thousand metric tons, where x is the year and y is the TNL. Leveraging this formula, predictions were made regarding the production of MSW for each grid in the four quarters of 2022 using nighttime light data. The predicted value (W_i) for the spatial distribution of MSW within a specific grid is as follows:

$$W_i=W_t\times (N{L}_i/TNL)$$

(2)

In the given equation, W_i symbolizes the quantum of waste generated in the grid (i), quantified in metric tons per quarter (t/quarter). W_t designates the projected MSW production for the specified quarter, ascertained via regression analysis, also articulated in t/quarter. NL_i denotes the mean nocturnal luminance within the grid (i) over the duration of the quarter. TNL is an acronym for the aggregate nocturnal luminance across the central urban grid for the quarter in question. Each grid encompasses a spatial area of 0.25 km². Figure 2 shows the estimated output of the MSW in Hefei’s core urban districts. Subsequent to the data collection, geospatial tools within ArcGIS 10.7 were employed. The “Feature to Point” utility was utilized to transmute the residential spatial configurations into discrete point entities. Thereafter, the “Extract Value to Point” utility was leveraged to approximate the production of MSW accumulation within individual grids. By superimposing this projected waste production over the residential point features, the aggregate waste generated across the four quarters within the residential area was computationally ascertained.

2.4. Spatialization of Social Statistical Data

To fulfill the analytical and modeling prerequisites, the vectorization of the aggregated grid and statistical samples was imperative. Citing the scholarly contributions of Yan et al., the Geographically and Contextually Area Weighted Index (GCAWI) methodology was employed to spatially extrapolate metrics such as per capita GDP figures, the retail commerce of consumer goods, and disposable income statistics [25]. Employing the collated quarterly data—which include variables such as territorial dimensions, aggregate GDP, total disposable income, collective retail sales of consumer goods, and population density—a foundational socioeconomic spatial distribution map was constructed. This map utilized residential domains as the basic analytical units. Grids were instantiated, with each grid constituting approximately 2% of the overarching administrative territorial expanse. Subsequently, the per capita socioeconomic spatial values for each grid were computationally determined. The formula deployed for this calculation is articulated in the subsequent section:

$$Per{X}_i=\frac{X_i}{POPden\times Area}$$

(3)

In the stipulated equation, PerX_i (where i = 1, 2, 3) embodies the spatialized socioeconomic metrics pertinent to each grid. X_i serves as an indicator for the cumulative socioeconomic data of the designated administrative district. POP_den is the parameter representing the demographic density within the grid, and Area signifies the spatial extent of the individual grid unit. Consequently, if the per capita spatialized socioeconomic data for grid units A and B are denoted as PerX_Ai and PerX_Bi, respectively, then the amalgamated per capita spatialized socioeconomic metric (PerX_Ci) for the composite region C = A + B can be calculated as follows:

$$Per{X}_{Ci}=\frac{Per{X}_{Ai}\times POPde{n}_A\times Are{a}_A+Per{X}_{Bi}\times POPde{n}_B\times Are{a}_B}{POPde{n}_A\times Are{a}_A+POPde{n}_B\times Are{a}_B}$$

(4)

Utilizing Equations (3) and (4), the triad of statistical categories can be spatially integrated into the vectorized topographical representation of the central urban districts, thereby emulating the geographic diffusion of the data.

3. Research Methods

3.1. Spatial Autocorrelation

Antecedent to the implementation of GTWR modeling, an examination of the variable spatial correlation was requisite. Spatial autocorrelation can be dichotomized into two primary categories: global Moran’s I and local Moran’s I. For the purview of this study, the global Moran’s I was selected as the evaluative metric for ascertaining the spatial autocorrelation of the domestic waste emissions within Hefei’s central urban agglomeration. The computational framework is articulated as follows [26]:

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

(5)

In the governing equation, n is the number of units; m is the quantity of proximal units for feature i; w_ij functions as the spatial weight coefficient, where w_ij = 1 if features i and j are contiguously situated; otherwise, w_ij = 0; x_j signify the empirically observed quantities of waste production on features i and j; x denotes the mean waste production for the overall study region. The value range of the Moran’s I index oscillates between −1 and 1. If Moran’s I > 0, it insinuates a positive spatial correlation; a larger value is indicative of a more pronounced positive correlation and intensified spatial clustering. If Moran’s I = 0, it implies the absence of spatial correlation, suggesting that the spatial units are distributed in a stochastic manner. If Moran’s I < 0, it signifies a negative spatial correlation; a smaller value underscores a more robust negative correlation and amplified spatial heterogeneity.

Local Moran’s I constitutes a specialized form of correlation analysis that examines the interrelationships between a focal geographic entity and its contiguous regions across multiple spatial scales. In contrast to its global Moran’s I, local Moran’s I furnishes a nuanced depiction of the entity’s intrinsic spatial attributes. The computational model employed for this assessment adheres to the mathematical formalism delineated in Equation (6) [27]:

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

(6)

where the notational conventions and parameter definitions are congruent with those established in Equation (5).

3.2. PCA-GTWR Model

The PCA-GTWR model amalgamates the methodological strengths of principal component analysis and the GTWR model [28]. Prior to the deployment of the GTWR model for predictive analytics, principal component analysis is invoked to rectify the issue of multicollinearity among the explanatory variables. Subsequent to this, a set of synthesized indicators are formulated. Indicators that collectively account for a cumulative contribution ratio surpassing 85% are subsequently incorporated as predictor variables within the GTWR model analytical schema. The PCA-GTWR model thereby mitigates the analytical complications engendered by multicollinearity, while concurrently elucidating the spatially contingent influences exerted by an array of determinants. The operational sequence of this hybrid model is articulated as follows, with reference to the existing methodologies cataloged in [29]:

(1)

The extraction of principal components. Principal component analysis serves as a computational mechanism to transmute an array of correlated indicators into a novel set of orthogonal composite indicators. Principal components (PCs) manifesting eigenvalues exceeding the unity threshold are earmarked for inclusion. The factor loadings on these PCs are subsequently calculated. Those indicators that either exhibit the highest loadings or lie within 90% of the maximal loading are retained in a minimal-requisite dataset. In instances in which multiple indicators subsumed under a single PC demonstrate significant inter-correlations, the indicator displaying the superior normative value is retained. In cases lacking notable correlation, all the constitutive indicators are retained. The underlying mathematical formulation is operationalized through the following computational formula:

$$N_{ik}=\sqrt{({\sum }_{i=1}^k{{{\lambda }_{ik}u}_{ik}}^2)}$$

(7)

where N_ik encapsulates the cumulative loading of the indicator (i) on the initial k principal components (PCs) that have eigenvalues surpassing unity. Herein, u_ik symbolizes the loading of the indicator (i) on the principal k component, and λ_k signifies the eigenvalue corresponding to the principal k component;

(2)

The GTWR model. An extension of the GWR model, the GTWR model is employed to incorporate both spatial and temporal aspects [30]. It furnishes a more nuanced understanding of data by accounting for the non-stationarity inherent in spatiotemporal relationships. Mathematically, the model can be represented as follows:

$$y_{ti}=a_0(u_i,v_i,t_i)+{\displaystyle \sum _{j=1}^n{a}_j(u_i,v_i,t_i)}{x}_{tij}+b_{ti}$$

(8)

where y_ti is the explained variable, n is the number of units, x_tij represents the explanatory variable, a₀ is the regression constant, a_j denotes the regression parameter, b_ti represents the residual error, and (u_i, v_i, t_i) designate the spatial and temporal coordinates for the unit i.

4. Results

4.1. Accuracy Verification of the Estimated MSW Production

To ensure the reliability of the estimated MSW production values, a comparison was made with the MSW production of Hefei City based on statistical data. As indicated in Figure 3, the estimated MSW production showed an RMSE of 1.7702 × 10⁴ t in relation to the statistical data, and an MRE of 0.0001%. It can be observed that the accuracy of the MSW production derived from nighttime remote sensing light data is commendable. Therefore, it can be utilized for studies on the spatiotemporal characteristics and underlying mechanisms of the MSW production in Hefei’s core urban districts.

4.2. MSW Production Estimation from Nighttime Remote Sensing Data

Utilizing nighttime remote sensing light data, the MSW production was estimated, and a trend graph representing the evolution of this production in the central urban area across four quarters was constructed through Origin2023 software, as illustrated in Figure 4.

Figure 4 elucidates the hierarchical waste production across the four predominant administrative districts: Baohe District (14.04 × 10⁴ t) > Shushan District (11.06 × 10⁴ t) > Yaohai District (6.85 × 10⁴ t) > Luyang District (3.89 × 10⁴ t). Notably, Baohe District manifests a diminishing trajectory in its MSW production, yet it witnessed an augmentation over the initial three quarters, culminating in the third quarter with an apex of 3.68 × 10⁴ t (averaging a growth of 1.90%). This zenith was followed by a dip, reaching its nadir in the fourth quarter at 3.23 × 10⁴ t. Contrarily, Shushan District portrays a vacillating ascent in its waste production, bottoming in the third quarter at 2.26 × 10⁴ t and surging to its pinnacle in the fourth quarter at 3.03 × 10⁴ t. Yaohai District, during the observation period, demonstrated a relatively static trend, predominantly descending, with its zenith in the first quarter marking 18.8 thousand tons. Luyang District, analogously to Yaohai, presented a decelerating trajectory, peaking in the first quarter at 1.08 × 10⁴ t. Potential rationales for these variances in the waste production distribution encompass the population density, flourishing economy, and expansive urban sprawl in Baohe District. Conversely, Shushan District, a domicile to an avant-garde high-tech industrial park and an economic development zone, emerges as Hefei’s most rapidly burgeoning economic precinct, thus registering elevated waste yields. In contrast, both the Yaohai and Luyang districts, encapsulating Hefei’s archaic urban core and characterized by a constricted urban expanse, have manifested a languid developmental pace, consequently leading to diminished waste output.

4.3. Spatial Autocorrelation Analysis of MSW Production

4.3.1. Global Spatial Autocorrelation

Leveraging statistical data pertaining to the production of MSW in residential vicinities, this study employed the global Moran’s I index to ascertain the extent of the spatial clustering. Utilizing ArcGIS for spatial autocorrelation analysis, the index was computed, and the outcomes are tabulated in

Table 2. Global autocorrelation results for MSW production in the central urban districts.

Project	Quarter 1	Quarter 2	Quarter 3	Quarter 4
Moran’s I Index	0.2675	0.2687	0.2633	0.2639
Z-value	28.1005	28.2674	27.6868	27.6965
p-value	0.001	0.001	0.001	0.001

. During the interval spanning the first and second quarters of the fiscal year 2022, the global Moran’s I index for the core urban districts was discernibly positive and statistically significant at the p < 0.01 level. This evidence corroborates the existence of the salient spatial correlation and spatial agglomeration of MSW production. Intriguingly, the index manifested its zenith in the second quarter, thus implying an intensified tendency for waste to be geographically concentrated during this temporal phase. Consequently, at a confidence interval of 99%, there exists a non-random, spatially correlated pattern in the municipal waste distribution throughout the quarters under scrutiny.

4.3.2. Local Spatial Autocorrelation

Employing the spatial clustering and anomaly analysis capabilities of ArcGIS, this study delved into the local spatial clustering traits of the MSW generation in Hefei’s core urban districts, culminating in the generation of a LISA cluster map (Figure 5). Concurrently, statistical analysis was conducted on the high–high (HH), low–low (LL), and outlier cluster areas (

Table 3. Statistics of MSW production and agglomeration area in Hefei’s core urban districts in the four quarters.

Quarter	Percentage of HH Type	Percentage of LH Type	Percentage of LL Type	Nonsignificant Type
First Quarter	22.68%	5.05%	24.01%	48.26%
Second Quarter	21.56%	3.04%	27.35%	48.05%
Third Quarter	19.70%	1.95%	33.42%	44.93%
Fourth Quarter	23.27%	5.05%	24.35%	47.33%

). Figure 5 illustrates the significant local spatial clustering in the core urban district MSW production, primarily dominated by HH-type and LL-type clustering. This pattern establishes a pronounced “core-periphery” configuration, demonstrating notable patterns of concentration and dispersion. In the initial quarter, HH-type clustering accounted for 22.68% of the area, creating a contiguous high-value zone encircling the city’s second ring road, notably in the central and northeastern segments of the urban core. The LL-type clustering extended slightly beyond the HH-type clusters, with two notable LL-type clusters situated in the northern part of Luyang District and in the heart of Baohe District. During this period, the HH-type clusters in central Baohe District were comparatively smaller. Moving into the second quarter, the contiguous HH-type cluster bordering the city’s second ring road saw a marginal reduction in size, whereas the central Baohe District’s HH-type cluster experienced growth, thereby expanding the differential between these two regions. The third quarter witnessed the LL-type area attaining its annual zenith at 33.42%, with LL-type clusters emerging inside the first ring road. Conversely, the HH-type cluster area shrank to its minimum, although the central Baohe District’s HH-type clusters persisted in their expansion. By the final quarter, the HH-type area’s proportion escalated to a peak of 23.27%, mirroring the first quarter’s local spatial clustering phenomena of MSW production, forming an HH-type belt along the second ring road, while the LL-type areas were predominantly situated beyond the city’s second ring road (the peripheral zones). This pronounced central spatial structure of MSW production in Hefei’s core urban districts underscores the critical role of geographic location as an influencer of MSW production.

4.4. Examination of Factors Affecting the Production of MSW

4.4.1. Principal Component Analysis

A set of pre-selected indicators showing substantial correlation with the production of MSW were subjected to principal component analysis through SPSS26. The principal objective was to eliminate multicollinearity and reduce data redundancy, thereby identifying a parsimonious set of predictors suitable for the construction of the GTWR model. Examination of the results of the sphericity hypothesis test (

Table 4. KMO and Bartlett’s test.

KMO Sampling Suitability Measure		0.69
Bartlett’s test of sphericity	Approximate chi-square	600.85
	Degrees of freedom	49
	Significance	0.00

) led to the rejection of the null hypothesis, thus intimating that the eight indicators under scrutiny are not statistically independent. Concurrently, the KMO statistic approached unity, suggesting the appropriateness of principal component analysis for data reduction in this context. A judicious number of principal components were subsequently extracted to supplant the original variables in the ensuing analytical stages. Within the eigenvector contribution table (refer to

Table 5. Principal component analysis results.

Principal Components	Characteristic Roots	Contribution %	Cumulative Contribution %	Dominant Factor	Dimension
PC1	3.80	54.37	54.37	GDP	Socioeconomic factors
PC2	1.07	15.16	69.37	Urban disposable income	Living standard factors
PC3	1.01	14.98	84.35	Temperature	Natural factors

), three principal components were discerned to have eigenvalues surpassing the threshold of 1. The foremost component exhibited an eigenvalue of 3.80 and accounted for 54.21% of the total variance, primarily indicative of socioeconomic variables. The second component, with an eigenvalue and contribution rate of 1.07 and 15.16%, respectively, underscores facets of the residents’ living standard variables. The third principal component, boasting an eigenvalue of 1.01 and accounting for 14.98% of the variance, predominantly characterizes natural variables. The cumulative contribution of these three principal components attains an impressive 84.35%, adequately encapsulating the preponderance of information contained in the original dataset. Consequently, these initial three principal components were selected as the input variables for the GTWR model.

4.4.2. Analysis of Model Results

To explore the spatiotemporal variation in the key factors affecting the MSW output, an initial global regression analysis using ordinary least squares was conducted. Stepwise regression refined the selection of explanatory variables, enhancing the model’s R² and confirming statistical significance through T-tests, with no detected multicollinearity, as detailed in

Table 6. OLS regression estimates.

Variable	Estimate	Standard Error	T-Statistic	p-Value	VIF
Natural variables	0.14	0.05	5.32	0.002	1.81
Socioeconomic variables	0.78	0.14	11.87	0.000	1.23
Residents’ living standard variables	−0.32	0.11	7.22	0.001	1.96
R2	0.4132
Adjusted R2	0.4157
RSS	1535.24
AIC	1179.20
Sigma	3.580

. This foundational regression provides a benchmark for subsequent comparisons with results from both the GWR model and the more nuanced GTWR model, realized through STATA 17.0 and ArcGIS software.

Global linear regression unveiled that socioeconomic progress and improvements in living standards yielded positive coefficient estimates, inferring a beneficial influence on the MSW generation when viewed on a global scale with all other factors remaining static. In contrast, the coefficient for natural factors is negative, signaling an inverse relationship with the MSW generation under similar constant conditions. With an R² of 0.4157, the model accounts for nearly half of the data’s variance, suggesting that further refinement through spatial econometric modeling could enhance its interpretative power.

Delving into the spatiotemporal discrepancies of the explanatory variables on the MSW production within the urban core, enhanced analytical depth was achieved through the GWR and GTWR models. These models demonstrated a marked increase in explanatory capacity over the global linear regression, with the R² values ascending to 0.5073 and 0.7031, respectively. Moreover, GTWR outperformed GWR in terms of the model fit and offered a lower Akaike Information Criterion (AIC). From the GTWR’s estimations and the spatial distribution of the coefficients, as presented in

Table 7. Estimates from GWR and GTWR models.

Variable	GWR				GTWR
	Minimum	Median	Maximum	Mean	Minimum	Median	Maximum	Mean
Natural variables	−0.014	0.14	0.233	0.135	−1.10	0	1.17	−0.01
Socioeconomic variables	0.504	0.77	0.886	0.74	−4.60	0.57	5.87	0.61
Residents’ living standard variables	−1.026	−0.39	−0.269	−0.41	−7.32	−1.57	2.90	−1.84
R2	0.5076				0.7046
Adjusted R2	0.5073				0.7031
RSS	187.99				155.62
AIC	390.7				378.3
Sigma	1.5703				1.4207

, we deduce that socioeconomic elements are the foremost positive influencers of the MSW emissions in Hefei’s core urban districts, the living standard of residents is the variable with the highest negative impact, while the impact of natural factors is minimal. A socioeconomic regression coefficient of 0.61 signifies that every 10% increment in the socioeconomic metrics correlates with a 6.1% rise in the MSW production. The living standard coefficient is pegged at −1.84, implying that a 10% enhancement in living conditions could slash the MSW production by 18.4%. Natural factors, with a regression coefficient of just −0.01, exert a negligible effect on the MSW production.

To account for the spatiotemporal variations in the influence exerted by the explanatory variables, the GTWR model was employed. Utilizing principal component analysis (PCA), the three principal components were extracted and juxtaposed against the regression coefficient values across various quarters, as visualized through Origin software (Figure 6).

Figure 6a delineates the complex interplay of the natural variables on the MSW production, illustrating a temporal fluctuation in their effects over the course of a year. Spatially, the impacts manifest divergently across distinct administrative districts. Shushan District is characterized by a linear escalation in waste production, Luyang District displays a W-shaped progression, while both the Baohe and Yaohai districts exhibit an inverted U-shaped trend. Notably, an evolutionary pivot from negative to positive impact is discernible in the Shushan, Luyang, and Baohe districts, whereas Yaohai maintains a persistently positive trajectory. The oscillatory behavior may be partially attributed to the urban heat island effect, which exacerbates ambient temperatures and consequently intensifies energy consumption, thereby catalyzing MSW production. The unique topography of Hefei City engenders a ventilation corridor, oriented from southeast to northwest, that is particularly salient during the spring and summer seasons. This geographical feature tempers the heat island effects in the Baohe and Shushan districts, which are strategically located on the windward side. Luyang District, conversely, finds itself on the leeward side, culminating in heat accumulation that exacerbates the heat island phenomena in both the Shushan and Luyang districts.

As discerned from Figure 6b, a robust positive correlation manifests between the socioeconomic variables and the production of MSW. The Baohe and Yaohai districts, which encompass the more antiquated sections of Hefei City, exhibit a conspicuous peak in the socioeconomic indicators during the fourth quarter. This elevation can potentially be ascribed to the suboptimal implementation of waste-sorting initiatives and a less-than-optimal level of waste reduction awareness in these particular regions. The ensuing rise in MSW production is further amplified by a surge in retail sales of social commodities, especially as the Spring Festival approaches. Conversely, the annual trends in the Shushan and Luyang districts reveal a diminution in the impact of the socioeconomic variables on waste production, suggesting a possible deceleration in the socioeconomic vitality of these administrative sectors.

Figure 6c reveals a uniformly negative influence of the residents’ living standard variables on the MSW production across all four quarters. Diverse temporal patterns are evidenced among the districts: the Shushan and Luyang districts manifest a U-shaped curve, Baohe District delineates a declining trend, while Yaohai District unveils a W-shaped trajectory. These variegated patterns can be predominantly attributed to escalating environmental consciousness and technological advancements in waste reduction. As the standard of living ascends in these districts, residents increasingly incline towards sustainable consumption patterns, waste recycling, and similar eco-friendly practices, thereby engendering a decline in overall waste production and wastage.

Employing ArcGIS software in conjunction with the natural breakpoint method allows for the cartographic representation of the spatial heterogeneity in the average regression coefficients for the explanatory variables under consideration. According to the model’s outcomes, it becomes unequivocally apparent that the influences of the natural variables, socioeconomic variables, and residents’ living standard variables on the MSW exhibit marked spatial heterogeneity, as visualized in Figure 7. This spatial diversity underscores the necessity for context-specific waste management strategies, tailored to the unique environmental, social, and economic conditions prevailing in each administrative district.

In Figure 7a, the spatial heterogeneity of the natural variables affecting the municipal solid waste production across Hefei’s urban districts is delineated. Notably, the districts of Luyang and Yaohai and the southern sector of Baohe exhibit more pronounced positive impacts compared to the other regions. The overall spatial arrangement reveals a stratified pattern, characterized by elevated levels of influence in the northern sectors and diminishing effects toward the south. As for the geographic disparities concerning the natural factors’ influence on waste production, zones experiencing a positive effect predominantly cluster at the intersection of Hefei’s four principal administrative districts, situated in the city’s core. Encompassing the Binhu region, these positive-influence zones constitute approximately 41% of the central urban expanse. Conversely, the areas subjected to negative influence are largely relegated to the peripheries beyond the city’s second ring road, southward, covering an estimated 59% of the city’s total land area. Collectively, the spatial variance in the influence of natural factors on the MSW production in Hefei’s urban milieu is considerable. Zones conducive to this influence are principally concentrated within the city’s central district as well as in the Binhu New District. However, the geographical scope of their influence remains circumscribed. The genesis of the positive impact in these focal regions may be attributed to their role as a nexus for commerce and services in Hefei, which is consequently marked by elevated levels of human activity. Accelerated urbanization has precipitated a pronounced urban heat island effect within this sector, typified by higher annual mean temperatures and reduced average precipitation. The Binhu New District, in proximity to Chaohu Lake, confronts a unique hydrothermal dynamic: the expansive water surface of the lake facilitates algal bloom, which, in turn, elevates the water temperature. This adversely affects the lake’s efficacy in ameliorating the urban heat island effect, culminating in elevated annual temperatures in areas adjacent to Chaohu Lake relative to other regions. Such climatic conditions, exacerbated by diminished rainfall, may incentivize residents to frequent dining establishments or opt for disposable tableware, consequently augmenting the production of municipal waste.

In Figure 7b, the cartographic representation elucidates the intricate spatial nexus between the socioeconomic variables and household waste production. The influence of economic development on the MSW production is markedly positive, particularly within the confines of the city’s second ring, centering around pivotal arteries such as Xiaoyaojin Street, Baogong Street, and Sanli’an Street. Additionally, the High-Tech Industrial Development Zone in Shushan District also manifests this positive correlation. The overarching spatial configuration evinces a “centrifugal dispersion” pattern. Areas exhibiting a positive influence on MSW production are expansive, encompassing approximately 47% of the central urban landscape. Conversely, isolated pockets of negative impact are observed predominantly in the Baohe and Luyang districts. The economically affluent zones within the second ring are highly accessible and consistently outperform other areas in the province in terms of GDP. This economic robustness serves a dual purpose: it both furnishes the material substrate to accommodate an expanding permanent or transient populace and epitomizes enhanced urban construction standards. Both the burgeoning resident population and escalating infrastructure development concomitantly culminate in an upsurge in household waste production.

As discerned from Figure 7c, the interrelationship between the residents’ living standard variables and the MSW exhibits spatial variance contingent on the urban topography. Zones evincing a positive correlation include the High-Tech Industrial Development Zone in Shushan, Baohe Economic Development Zone, and Binhu New District. In stark contrast, other sectors manifest a negative correlation, intimating that, in the city center, living standards are increasingly becoming uncoupled from waste production rates. A plausible explanation for this disassociation is a noteworthy metamorphosis in consumption paradigms as living standards escalate. Enhanced emphasis on sectors such as entertainment and healthcare implies that the per capita consumer expenditure is inversely proportional to the MSW production. This observation is congruent with the theoretical underpinnings of the Environmental Kuznets Curve (EKC), which hypothesizes an initial increase in environmental degradation with economic growth, followed by a subsequent decline, thereby tracing an inverted “U” trajectory [31]. Coinciding with Hefei’s recent endeavors in waste segregation and industrial reformation, there has been an increased adoption of cleaner industrial practices. Concurrently, public expenditure on eco-friendly consumption has escalated, engendering a relative attenuation in MSW production. Special economic zones function as the linchpins of Hefei’s economic vitality, magnetizing a substantial influx of migrants. Consequently, the living standards within these zones harbor significant potential for growth, thereby sustaining a direct correlation with the household waste production.

5. Discussion

Utilizing the PCA-GTWR model, we have discerned that the regression coefficients for the three pivotal explanatory variables demonstrate spatial non-stationarity in influencing MSW production. The analysis shows that socioeconomic elements are the foremost positive influencers of the MSW emissions in Hefei’s core urban districts, the living standard of residents is the variable with the highest negative impact, while the impact of natural factors is minimal. The influence of natural elements on the MSW production in the core urban districts exhibits temporal and regional variations, generally increasing and then diminishing throughout the year, with significant positive effects centered in the urban heart and Binhu’s burgeoning district, possibly linked to the urban heat island effect [32].

Socioeconomic variables predominantly bolster MSW production in all four of Hefei’s main administrative districts, emanating from the urban nucleus and the high-tech development sector. This reflects Qin et al.’s research, indicating a stronger correlation between economic growth and MSW production in economically advanced city centers [33]. Because urban MSW primarily arises from consumption within the urban logistics chain, the vigor of economic transactions directly dictates material consumption, naturally correlating the economic levels with the activity intensity. Nonetheless, as urbanization progresses, the frequent inter-city resource flow and industrial transference from the center to less developed outlying areas have begun to influence the MSW production. Population size, a crucial socioeconomic metric, also propels urban MSW production. Zhang et al. categorize Hefei’s central urban population into densely populated and accessible areas (the old city and Shushan, and Baohe districts), moderately dense and accessible areas (Binhu’s burgeoning district, Luyang, and high-tech zones), and average-density and -accessibility areas (Yaohai District) [34]. Recently, Hefei has augmented public services in densely populated zones to offload functions to lower-density districts, easing the population density. However, this study suggests that optimizing the population distribution paradoxically spikes the MSW output at the city’s core. Consequently, Hefei is compelled to expedite policies to curb rapid population growth and enhance public amenities in less dense districts, drawing inhabitants to these new urban MSW production and management hubs.

Living standards predominantly impact the MSW production negatively, with the degree of this impact varying by district and over time. In the city’s core, living standards have a smaller, sometimes even inverse, effect on the MSW production compared to those in newly established districts, echoing Liao et al.’s findings that this relationship adheres to the Kuznets curve [35]. These central areas appear to have passed the waste production inflection point and are witnessing a downturn. Contrastingly, in burgeoning zones like the Shushan High-Tech Zone, Shushan Economic Development Zone, and Baohe Binhu’s burgeoning district, the standard of living continues to significantly propel MSW generation. The Shushan High-Tech Zone, as a pioneering national high-tech hub, stands as a key technological incubator and economic catalyst in Hefei, with its GDP skyrocketing from CNY 13.84 billion to CNY 110.13 billion from 2008 to 2020. Per the Lewis model, labor shifts from low-productivity agriculture to high-productivity urban industries in developing countries, a transition expedited by urbanization [36]. This urbanization mobilizes surplus rural labor into city production roles, bolstering the service sector with inexpensive labor, refining the industrial framework, and raising urban income levels while simultaneously intensifying the MSW generation. Therefore, Hefei needs to strengthen ecological civilization education, promote sustainable consumption, and advocate for conservation to curb waste. Concurrently, optimizing the industrial structure and advancing production technology are crucial for a healthy industrial evolution.

The control of MSW production and disposal holds considerable theoretical and practical relevance for urban sustainability. Thus, it is essential to leverage advanced techniques to increase the research precision. This study attempts to combine classical methodologies with contemporary GIS technology to analyze urban MSW, yet further investigation is needed to delve into the spatial characteristics of the factors identified via PCA-GTWR to extend the timeframe of social data collection for richer data sources, and to perfect the spatial representation methods for granular social data. It is anticipated that with deeper research and the evolution of big data technologies, the study of the spatial dynamics of urban MSW management will gain greater diversity and accuracy.

6. Conclusions

Investigating the spatiotemporal dynamics and determinants of MSW production from an urban core perspective offers vital insights for crafting targeted waste minimization policies. This research harnesses the PCA-GTWR model to assess the spatiotemporal variability of the MSW production influencers, successfully mapping out the distribution and driving factors of the MSW in Hefei’s core urban districts. The salient findings include the following:

(1)

The MSW production is characterized by an imbalanced distribution. Global Moran’s I analysis indicates a significant positive spatial correlation within Hefei’s core urban districts. Zones exhibiting “H-H”-cluster attributes predominantly reside in the northeast, whereas “L-L” clusters are situated in the southwest. The MSW production is ranked as follows: Baohe District (140,400 tons) surpasses Shushan District (110,600 tons), which, in turn, exceeds Yaohai District (68,500 tons) and Luyang District (38,900 tons);

(2)

Through principal component analysis, eight MSW-related indicators were distilled into three independent principal components, mitigating the multicollinearity challenge within the GTWR model due to the high correlation between the variables. A comparative assessment of the OLS, GWR, and GTWR models demonstrates the superior explanatory power of GTWR in delineating the variables impacting urban MSW production;

(3)

Employing the PCA-GTWR model to scrutinize the variable impact on the MSW production yields that the natural variable influence on the MSW production in Hefei typically rises and then falls over successive quarters, with the city center and Binhu’s burgeoning district experiencing the most pronounced positive effects. Socioeconomic variables exert a compelling facilitative role on the MSW production throughout Hefei’s quartet of administrative regions, radiating from the central core and the high-tech zone. Contrarily, living standards predominantly exert a dampening effect on the MSW, with a lessened or even inverse impact in the established city center relative to the burgeoning districts.