Satellite-Observed Mismatch in Urban Growth and Population Dynamics: Implications for Sustainable Regional Planning in Guangdong Province

Abstract

Understanding spatiotemporal mismatches between urban expansion and population dynamics is essential for guiding sustainable development in rapidly urbanizing regions. Using multi-source nighttime light (NTL) images and global settlement layers, this study investigates the settlement growth pattern and potential spatiotemporal mismatch with population distribution in Guangdong, China, from 1995 to 2019 at a 5-year interval. Specifically, population spatialization in urban and rural areas is separately mapped by adopting a population-based thresholding method, achieving strong agreement with the census record. Our analysis reveals distinct expansion patterns and mismatch conditions across Guangdong’s Core, Belt, and District subzones. The Core and District subzones primarily experienced infilling and edge-expansion urban growth, while the Belt subzone exhibited more dispersed spatial patterns. Notably, only 5 of 21 prefectures exhibited faster population growth than urban expansion, likely due to sustained migration driven by economic opportunities and advanced urbanization. Quantitatively, both urban expansion and population growth followed a Core, Belt, District order. Spatially, population-dominated areas were primarily clustered within 10 km of urban centers, while the District subzone extensively displayed overfilled settlements, indicating low-efficient land use. Temporally, urban growth relative to population in the Core subzone turned from slower pre-2000 to faster post-2000, followed by gradual deceleration, while the Belt subzone maintained balanced growth throughout the study period. The District subzone sustained faster urban growth from 2000 to 2019. Findings of the study provide an important reference for scientific urban planning and sustainable regional development, not only in Guangzhou but other rapidly urbanizing regions globally.

1. Introduction

Urban expansion, encompassing both the physical growth of built-up areas and the increasing anthropogenic dispersion of population, is widely studied due to its profound ecological and socio-environmental impacts. Environmentally, expansion of built-up land disrupts essential biological processes [1,2] and degrades urban environments [3,4]. Socially, over-expansion can lead to considerable housing vacancy [5,6] while demographic oversaturation depletes urban sustainability [7,8]. Therefore, a comprehensive understanding of urban expansion patterns and the balance with population is essential for regional sustainable development.

Spatiotemporal characteristics are critical in urbanization research, as the location of urban land change reflects transformation in urban spatial structure, which is intrinsically related to key morphological indicators such as land-use efficiency and compactness [9,10,11,12], both of which affect urban energy consumption with significant sustainability implications [13,14,15]. Moreover, temporal information facilitates recognition of urban revolution patterns and enables a more robust identification of driving factors through temporal correlation analysis [16].

Numerous methods have been developed to extract spatiotemporal characteristics of urban land expansion. Spatially, the type of urban growth is decided based on land cover percentages within moving windows [17], an approach known to be sensitive to window size selection. The boundary sharing rate (BSR) was introduced [18] to distinguish between infilling, edge-expansion, and outlying growth. The land expansion index (LEI) was designed utilizing buffer zone analysis [19], and subsequent studies have incorporated vertical information [20]. New indices considering centrality and aggregation of land patches were utilized to characterize urban expansion structure and form [21]. Temporally, the multi-order landscape expansion index (MLEI) is proposed for the analysis of urban expansion dynamics based on multi-temporal data. Temporal effect is also utilized on the basis of geographic weighted regression to deal with both spatial and temporal heterogeneity and influence of non-stationary factors [22].

While existing studies have significantly advanced our understanding of urban expansion patterns, current methods primarily focus on quantifying individual urban patches and lack the capacity to assess their correlations and balance with other urban elements. Urban expansion inherently includes the physical extension of built-up land and the spatial concentration of population [23], necessitating an approach that can comprehensively evaluate sustainability across spatial, temporal, and quantitative dimensions. Buffer zone analysis, a fundamental and widely applied technique in geospatial research [24], provides a robust framework for such assessments. By generating concentric rings around urban centers at fixed intervals, this method enables systematic examination of land-use change patterns. Its strength lies in capturing both spatial consistency and distance-decay effects in urban development processes [25,26,27], making it particularly valuable for analyzing the multidimensional aspects of urban growth.

Recent global analyses have revealed significant differences in growth rate between urban land and population [28,29,30]. China, in particular, has undergone unprecedented urban expansion in recent decades [31]. Marked socioeconomic development and environmental conditions have resulted in significant spatial mismatches between urban growth and population concentration [32,33]. This issue is especially pronounced in Guangdong province, where regional inequalities exist between the Pearl River Delta and other areas. Therefore, a scientific reference regarding urban-population consistency is urgently required for rational urban planning and sustainable regional development.

Data accuracy is vital for investigating urban-population consistency; census data are acknowledged for their authenticity in population quantity, yet the limitation in spatiotemporal resolution restricts their broader applications. Data from numeral sources are utilized to characterize fine-scale population distribution, including nighttime light (NTL), building footprints, and mobile phone-based data. Due to the finer resolution and strong correlation with social activity characteristics [34], NTL data has been effectively employed to map population dynamics across diverse regions and scales [35,36,37,38].

This study aims to comprehensively demonstrate the spatiotemporal alignment between urban land and population in Guangdong Province, China, from 1995 to 2019 at 5-year intervals. Utilizing global settlement data and population maps derived from NTL data, we employ buffer zone analysis to (1) classify urban expansion based on directional patterns, and (2) analyze the mismatch between urban growth and population dynamics in quantitative, spatial, and temporal dimensions. Our analysis of urban-population consistency in Guangdong’s unequal regional context reveals new urbanization perspectives, providing actionable insights for sustainable development strategies.

2. Study Area and Datasets

2.1. Study Area

Guangdong Province, situated along China’s southern coastline, is the nation’s most populous province with 127.06 million residents in 2023 [39] and a leading economic hub renowned for its manufacturing and commercial sectors. The province comprises 21 prefecture-level cities exhibiting substantial developmental disparities. Following Guangdong’s 2019 coordinated development strategy named “One Core, One Belt, One District” [40], we categorize these prefectures into three distinct subzones based on economic status and municipal functions: (1) Core subzone: the most economically advanced region, containing 9 prefectures around the Pearl River Delta, serving as Guangdong’s economic and technological center. (2) Belt subzone: seven coastal prefectures spanning Guangdong’s southern and eastern coastline, serving as a crucial coastal economic belt with advantages in commercial and industrial activities, including port logistics and manufacturing. (3) District subzone: five less-developed prefectures primarily located in the northern part of Guangdong, with a greater emphasis on ecological conservation and resource provision. The significant regional heterogeneity among these prefectures makes Guangdong an ideal case study for examining the complex dynamics between urban expansion and population distribution. The geographic location and subzone division of Guangdong province are presented in Figure 1.

2.2. Datasets

2.2.1. Settlements Data

Diurnal satellite images offer higher radiometric resolution and contexture information when depicting dynamics of urban expansion, with a focus on change of resident-related urban land, two components of World Settlement Footprint (WSF) suite [41], which is a global settlement dataset built upon Landsat and Sentinel satellite images, are employed to examine the spatiotemporal dynamics of habitable urban land in Guangdong over study period.

• WSF-Evolution:

WSF-Evolution is a raster layer dataset that documents the annual global settlements expansion from 1985 to 2015 at 30 m resolution. The distribution of the settlements is identified by applying morphological filtering on spectral indices extracted from Landsat satellites and classified using a machine learning technique. Each pixel of WSF-Evolution is assigned a four-digit value indicating the earliest detection year of the corresponding settlement, enabling the temporal analysis of urban expansion patterns.

2.

WSF2019:

WSF2019 outlines the global settlements distribution in 2019 at 10 m resolution using the complementary spectral information from Sentinel satellites.

2.2.2. Nighttime Light Imagery

Distribution of population is difficult to obtain as settlement land from diurnal satellite images, however, the utilization of electricity provides the chance for population spatialization [42]. To establish a continuous NTL time series over the study period, NTL images from DMSP-OLS and VIIRS-DNB are integrated and processed.

DMSP-OLS (Defense Meteorological Satellite Program—Operational Linescan System) NTL images were collected by the U.S. Air Force from 1992 to 2013 from a weather forecasting project, which is composed of six satellites, coded F10, F12, F14, F15, F16, and F18, respectively. The onboard OLS enables the satellites to effectively capture low-intensity ground radiance during nighttime. DMSP-OLS has a global coverage from −65° to 75° in latitude and a spatial resolution resampled to 30-arcseconds with 6-bit quantification digital values that range from 0 to 63. We employ the annual stable light product of DMSP-OLS Version 4, which is available on Google Earth Engine (GEE) [43]. The necessary calibration for quality control of DMSP images was accomplished by the Earth Observation Group (EOG) from Colorado School of Mines.

VIIRS-DNB (Visible Infrared Imaging Radiometer Suite—Day/Night Band) images, launched aboard the Suomi NPP satellite, offer significant advancements over DMSP-OLS with a spatial resolution of 15-arcseconds and a wider dynamic range with seven orders of magnitude. The VIIRS-DNB Composite Version 1 is also available on GEE, with necessary preprocessing accomplished by EOG. The VIIRS images were additionally filtered to exclude data impacted by stray light, lightning, lunar illumination, and cloud cover.

2.2.3. Supplementary Datasets

Official census records from 1990 to 2019 and prefectural-level administrative boundaries for Guangdong are employed as supplementary data; the global impervious surface layer (GISA) [44] is utilized for visualization and determination of urban centers.

3. Methodology

The methodology of this work mainly consisted of three parts: (1) the construction of extended NTL time series. (2) NTL-based population spatialization, and (3) consistency analysis using buffer zone. The first step includes calibration of DMSP-OLS images and transformation of VIIRS-DNB images, and the second step includes population-based NTL thresholding and separate population mapping. The detailed procedure of the methodology is summarized in Figure 2.

3.1. Construction of Consistent NTL Time Series

To enable population spatialization using NTL data, we processed DMSP and VIIRS datasets to create a consistent time series with comparable intensity values. Particularly, images before 2015 were obtained from DMSP datasets and images after 2015 were acquired from VIIRS datasets. The extended NTL series was generated through two key steps: (1) calibration of DMSP images, and (2) DMSP-like transformation of VIIRS images. Calibration methodology for the two datasets was, respectively, based on the framework of [45,46], where a detailed calibration process can be referred to. Specifically, radiance adjustment using vegetation index was not performed because we found the adjustment compressed the range of pixels’ digital number (DN), making NTL classification threshold inaccurate. Following calibration of both NTL datasets, the two adjusted datasets with identical spatial resolution and value distributions were integrated to generate a consistent, DMSP-like NTL time series for subsequent population spatialization.

3.2. Separate Population Spatialization

NTL data have become a vital tool for fine-scale population spatialization, owing to their strong correlation with human activity and comprehensive spatiotemporal coverage. While numerous gridded population datasets have been created using NTL imagery and ancillary data across multiple scales [47], research emphasized persistent limitations in distribution accuracy—particularly in rural areas [37,48,49]. These discrepancies arise from variations in anthropogenic activity, insufficient proxy representation, and the absence of rural-specific modeling approaches. Such uncertainties substantially reduce data reliability in areas with pronounced rural-urban disparities. To enhance population estimation accuracy in unevenly developed regions, we applied two key assumptions from [48] that assume the following: (1) a province’s urbanization level positively correlates with its urban population, and (2) urban population estimates are more reliable than rural ones. Leveraging region-specific census data, we developed an inverse mapping relationship between urban population and urban land area using the Ratio of Urban to Total Population (RUTP). This ratio links classification thresholds to maximum DN values—for example, if urban residents represent 60% of the total population, the threshold becomes 63 × 0.6 ≈ 38, with pixels exceeding this value classified as urban. The threshold was initialized using 1995 census data and dynamically adjusted for subsequent years by incorporating annual RUTP increments multiplied by the elapsed time. This adaptive approach, which mirrors urban land change dynamics, proves equally effective for rural areas. Following annual threshold determination, urban and rural zones were systematically delineated.

For urban and rural areas in each study year, the total DN value in each study year is aggregated for both areas using extended NTL time series, then a series of regression functions of DN against census population are conducted with Equation (1):

$$\sum {Pop}_{ij}=f_{ij}\left(\sum {DN}_{ij}\right)$$

(1)

where Pop and DN are the total population from the census record and the aggregated DN value in each area, respectively. $f_{ij}$ is the regression function, where the indices $i$ and $j$, respectively, denote spatial and temporal range, with $i\in \left\{\mathrm{0,1}\right\}$ represents the area classification (0 for rural, 1 for urban), while $j\in \left\{\mathrm{1995,2000,2005,2010,2015,2019}\right\}$ represents the study year. Population density at the pixel level was derived by applying year-specific regression functions to DN values, followed by mosaicking to generate a seamless provincial-scale population map for Guangdong. This method preserves spatial consistency while accurately capturing urban–rural population distribution patterns.

3.3. Growth Dynamics Estimation

The analysis of settlement expansion and population growth dynamics was conducted using a systematic spatial framework. Settlement growth was identified through temporal filtering of pixel values, with WSF-Evolution data refined using higher-resolution WSF2019 masks to enhance accuracy. Population distribution patterns were examined within a series of concentric buffer zones generated around each prefecture’s urban center, all processing implemented on the Google Earth Engine (GEE) platform. After obtaining the distribution of settlement and concentration of population on the pixel level, their growth dynamics in different directions over the study period can be analyzed using buffer zone analysis. The concentric rings are built from the urban center with an interval of 1 km and a radius of 40 km for each Guangdong prefecture. Specifically, the 40 km radius was selected based on both the literature review and empirical evidence. For the radius selection, ref. [50] estimated that the radius of Guangzhou’s main urban area reached 39.6 km. Similarly, ref. [51] calculated a 37.11 km urban radius for the Guangzhou–Foshan agglomeration. This value is theoretically applicable to either city, given the proximity of the two urbans’ centers. Our empirical observations further confirm that a 40 km radius effectively covers the majority of urban land, while smaller radii would exclude peri-urban zones. The 1 km interval selection was informed by both global urban analysis standards [52,53] and regional characteristics, particularly Guangzhou’s documented 1 km/year expansion rate during 1990–2009 [54]. This granularity maintains compatibility with the spatial resolution of DMSP-NTL data while preserving necessary detail for urban growth analysis. Urban centers are manually selected with the highest NTL intensity and earliest construction year by visualizing the NTL image and GISA layer. Figure 3 demonstrates the procedure of center selection for Guangzhou. The selection of an urban center for another prefecture follows the same method. The buffer zones were further segmented into four cardinal sectors (north, south, east, west) to examine directional spatial development patterns. Administrative boundaries and water bodies were systematically excluded during analysis to ensure geographic relevance, with all spatial operations performed on GEE. An example of rings to the north from the clip of the GEE interface is presented in Figure 4.

Following the determination of buffer parameter settings, concentric rings are created, by filtering the pixel value of settlement data and choosing the corresponding population map for a specific year, the total quantity of population and settlement over the study period can be acquired through aggregation within a defined spatial extent. Then, the growth dynamics of urban and population within each ring are available and we define the periodic growth rate in Equation (2):

$${PGR}_r={\displaystyle \frac{S_{t+5}-S_t}{S_t}}$$

(2)

where ${PGR}_r$ refers to the expansion rate for variables in region $r$ during each 5-year interval, $r$ can be urban or rural areas of each prefecture. For population, $S_t$ and $S_{t+5}$, respectively, represent the sum of pixel values within the study region in the current and next temporal stage. For settlements, $S_t$ and $S_{t+5}$, respectively, represent the number of valid pixels in the current and next temporal stage. Using $\mathrm{P}\mathrm{G}\mathrm{R}$, the pattern of growth between settlements and population can be compared.

4. Results

4.1. Consistent NTL Time Series for Population Spatialization

As presented by Figure 5, major drawbacks of original NTL images can be observed: (1) Inconsistency between and instability within DMSP satellites. The former can be found from the discontinuity between solid lines of DMSP’s satellites and the latter is displayed by fluctuation of lines, especially F15 and F16. (2) Temporal incompatibility. Notable difference in total DN value exists between images taken by different satellites during the same year, as presented by the distance between lines with intersection over the x-axis. (3) Incomparability between images from different sensors, as displayed by the total pixel value of the original VIIRS images.

After calibration and transformation, the total pixel value of NTL images before and after calibration is, respectively, shown in solid and dashed lines. Compared with the discontinuity of the DN sum from original images, the trend of the DN sum from calibrated time series shows a consistent and logical increase. The natural connection between two dashed lines indicates that images from different datasets are now compatible. Specifically, using the methodology proposed in [55], calibration coefficients for some DMSP-OLS images are optimized with better calibration effects, updated coefficients with corresponding evaluation indicators are listed in

Table 1. Calibration coefficients for DMSP-OLS images. Data in black represents optimized coefficients with better calibration effects, data in blue represents kept coefficients from reference work.

Satellite	${\mathsf{\alpha }}_0$	${\mathsf{\alpha }}_1$	${\mathsf{\alpha }}_2$	${\mathbf{R}}^2$	$\mathbf{R}\mathbf{M}\mathbf{S}\mathbf{E}$
F101992	1.6462	0.7887	0.0018	0.9812	1.6784
F101993	−0.6608	1.104	−0.0031	0.9953	0.793
F101994	1.3087	0.8922	−0.0004	0.984	1.595
F121994	1.4318	0.6273	0.0047	0.9478	2.9837
F121995	0.6495	0.7977	0.001	0.9608	2.5888
F121996	1.6255	0.727	0.0027	0.9716	2.1967
F121997	0.4176	0.7686	0.0028	0.969	2.179
F121998	1.0554	0.5449	0.0055	0.9736	2.7661
F121999	1.5706	0.4731	0.0063	0.9613	3.3543
F141997	0.1617	1.1521	−0.0029	0.9871	1.2277
F141998	1.9157	0.8932	−0.0005	0.9808	1.7504
F141999	0.6451	0.9543	−0.0006	0.9948	0.8545
F142000	1.7528	0.7779	0.0011	0.9695	2.2903
F142001	0.8781	0.7903	0.0013	0.9666	2.3549
F142002	1.3648	0.6628	0.0040	0.9821	2.2809
F142003	0.8566	0.775	0.0012	0.9559	2.7934
F152000	0.4816	0.5850	0.0051	0.9780	2.5347
F152001	0.5699	0.5790	0.0058	0.9806	2.3729
F152002	1.1773	0.4723	0.0070	0.9774	2.5632
F152004	1.3358	0.7435	0.0032	0.9803	1.7718
F152005	0.1758	0.8458	0.0016	0.9822	1.6403
F152006	0.4979	0.8196	0.0022	0.9846	1.5266
F152007	1.6631	0.7507	0.0027	0.9804	1.7479
F162004	0.9376	0.6658	0.0042	0.9838	2.1760
F162005	0.1366	0.88	0.0005	0.9802	1.7188
F162006	0.8544	0.6405	0.0045	0.9334	3.3025
F162007	1.3989	0.4562	0.0069	0.9664	3.1223
F162008	1.2588	0.5202	0.0059	0.9686	3.0233
F162009	1.9049	0.6641	0.0035	0.9571	2.7242
F182010	2.1357	0.1869	0.0104	0.9628	3.2883
F182011	2.2331	0.3531	0.0072	0.9228	4.7330
F182012	2.0651	0.2699	0.0088	0.9524	3.7276

in black, while kept coefficients are in blue. Coefficients for the transformation of VIIRS images are listed in

Table 2. Transformation coefficients for VIIRS-DNB images.

Year	$\mathbf{a}$	$\mathbf{b}$	$\mathbf{c}$	$\mathbf{d}$	${\mathbf{R}}^2$	$\mathbf{R}\mathbf{M}\mathbf{S}\mathbf{E}$
2013	−1.07	58.07	4.36	0.64	0.9423	2.7306

.

In addition to addressing intra-dataset inconsistencies, we further validated our provincial-scale NTL calibration using an established harmonized NTL dataset with a global perspective [56], to verify the generalization ability of our proposed method beyond Guangdong province. Two correlation indices, the coefficients of determination (${\mathrm{R}}^2$) and root mean square error (RMSE), are calculated for both our consistent NTL time series and harmonized global NTL dataset before 2015, using the original DMSP images as the calibration standard. As listed in

Table 3. Correlation coefficients between original DMSP images and two sets of integrated NTL data in each study year.

Data Source	Indicator	1995 (F12)	2000 (F14)	2000 (F15)	2005 (F15)	2005 (F16)	2010 (F18)
Harmonized global NTL	R2	1	0.9711	0.9744	0.9446	0.9479	0.9844
	RMSE	0	2.7896	2.9767	4.4510	4.2501	2.7301
Our NTL time series	R2	0.9681	0.9761	0.9451	0.9776	0.9774	0.9039
	RMSE	2.9783	2.5367	4.3642	2.8309	2.7984	6.7820

, the employed dataset showed higher ${\mathrm{R}}^2$ and lower RMSE for image F121995, F152000, and F182010, and our NTL time series shows the same statistical advantage for image F142000, F152005, and F162005. The contrast displays a comparable calibration result with global NTL harmonization from 1995 to 2010.

Given that VIIRS image DN distributions were transformed to match DMSP patterns, direct comparisons between original VIIRS composites and calibrated images are methodologically unsound. To effectively execute validation, we presume the reference dataset images as ground truth and consider our NTL time series as calibrated derivatives. Correlation analysis of corresponding image pairs in

Table 4. Correlation coefficients between two sets of NTL dataset.

Indicator	2015	2019
R2	0.9584	0.9308
RMSE	2.8880	4.2858

reveals the determination coefficients over 93% for both study years, with particularly strong agreement in 2015 (0.96). This consistent high correlation with benchmark data throughout the study period confirms the validity of our NTL time series and the generalization of the analysis framework worldwide.

4.2. Population Spatialization Based on Consistent NTL Time Series

The population spatialization for Guangdong was generated by applying our separate mapping strategy across all study years. As NTL data is often criticized for being unable to fully characterize population distribution, our validation against census records confirms the result’s reliability. As shown in

Table 5. Estimated population count and estimation accuracy at regional scale in each study year. Populations are presented in unit of one million.

	Year	1995	2000	2005	2010	2015	2019
Source
urban-census		37.4	47.6	55.8	69.1	81.2	90.7
urban-estimation		32.8	45.4	57.5	68.6	80.8	93.2
Accuracy		87.7%	95.38%	96.95%	99.28%	99.51%	97.24%
rural-census		36.4	38.9	36.1	35.2	35.6	34.2
rural-estimation		38.7	38.1	36.7	35.8	31.7	28.1
Accuracy		93.68%	97.94%	98.34%	98.3%	89.04%	82.16%

, strong agreement is demonstrated during intermediate study periods (2000–2010), with both areas achieving accuracy over 95% relative to census data. This performance demonstrates the effectiveness of our approach for local-scale population mapping.

In addition to regional-scale validation, we conducted a provincial-scale comparison to evaluate the performance of our separate mapping strategy and its consistency with widely used population datasets, including the Gridded Population of World Version 4.11 (GPW-v4), the Global Human Settlement Layer (GHSL), LandScan, and WorldPop. Key results listed in

Table 6. Population aggregation from different sources at provincial scale. Populations are presented in unit of one million. ‘NA’ indicating data unavailability for the corresponding year and ‘2020’ indicating the population count in the nearest available year of 2020.

	Year	1995	2000	2005	2010	2015	2019
Data
Census		73.8	86.5	91.9	104.4	116.8	124.9
Separate Mapping Sum		75.5	83.5	94.2	104.3	112.5	121.3
Direct Mapping		78.8	86.8	96.05	104.1	110.6	116.9
GPW-v4		NA	84.55	92.63	101.74	112.02	123.66 (2020)
GHSL		65.57	86.78	96.1	104.16	110.62	116.88 (2020)
WorldPop		NA	73.55	88.51	88.47	97.41	105.39
Land Scan		NA	69.41	88.06	89.72	105.18	106.93

reveal that the result of direct mapping is less accurate than the employed separate mapping strategy, yet it performs better than LandScan and WorldPop. The high quantity consistency with census data achieved by our result confirms its comparability with a well-established dataset benchmark at the provincial scale. Discrepancies among reference datasets likely reflect differences in their underlying modeling frameworks. The consistent accuracy across local and provincial scales validates the reliability of our population spatialization.

The coefficients for linear regression between total DN value and the sum of population at different scales in each year are in

Table 7. Coefficients for linear regression between total DN value and sum of population at different scales. Row name under each year represent extents of population spatialization.

	${\mathsf{\alpha }}_3$	${\mathsf{\alpha }}_2$	${\mathsf{\alpha }}_1$	${\mathsf{\alpha }}_0$
1995
Province	−0.7130	1.3298	−0.0542	0.0009
Urban	−902.6327	71.6581	−1.8382	0.0156
Rural	−0.2138	1.0154	−0.0370	0.0011
2000
Province	−1.6127	1.6356	−0.0758	0.0013
Urban	−764.8120	63.9415	−1.7042	0.0150
Rural	−0.4262	0.9565	−0.0290	0.0009
2005
Province	−2.0530	1.8518	−0.0962	0.0016
Urban	−789.2343	68.3244	−1.8744	0.0168
Rural	−0.6158	1.1000	−0.0514	0.0013
2010
Province	−4.5706	2.3133	−0.1228	0.0020
Urban	−551.1269	50.8325	−1.4678	0.0138
Rural	−1.8341	1.2675	−0.0619	0.0013
2015
Province	−9.9781	3.2537	−0.1700	0.0025
Urban	−305.5835	31.0363	−0.9716	0.0098
Rural	−6.0543	2.2434	−0.1337	0.0026
2019
Province	−13.0996	3.6596	−0.1893	0.0027
Urban	−199.5136	22.3208	−0.7493	0.0079
Rural	−5.8873	1.8782	−0.1044	0.0020

.

4.3. Growth Dynamics in Different Directions at Buffer Zone Scale

Figure 6, Figure 7 and Figure 8 demonstrate the dynamics of settlement expansion during each temporal stage in different directions. Each figure illustrates expansion dynamics of a subzone with $\mathrm{n}\times 4$ subplots representing four directions of $\mathrm{n}$ prefectures, marked by ‘W’, ‘N’, ‘E’, and ‘S’ in representation of west, north, east, and south, respectively. In each subplot, the y-axis represents the cumulative settlement increment and the x-axis represents the number of concentric rings where expansion occurred. As the ring number denotes distance to urban center at 1 km interval, the expansion pattern of settlements in every direction can be determined by distribution of evident growth on x-axis, for example, the growth concentrated in left-side of x-axis indicates expansion primarily clustered around urban center, growth concentrated in right-side of x-axis indicates expansion mainly occurred in urban fringe, and an even distribution indicates expansion is scattered within urban land, the above expansion pattern are commonly interpreted as infilling, edge expansion and dispersion, respectively.

The dynamics of settlement expansion of the Core subzone are depicted in Figure 6. For this subzone, Guangzhou, Zhuhai, and Jiangmen show the infilling expansion pattern, growth of settlement is primarily concentrated within 15 km of the urban center. Settlement expansion is evenly distributed across the buffer radius in Shenzhen, Foshan, Zhongshan, and Zhaoqing, demonstrating the dispersion pattern. Settlements in Dongguan and Huizhou show a pattern of edge expansion, with settlement growth mainly occurring in eastern and southern directions, away from the urban center; however, minor expansion also appeared near the urban center in those two prefectures. Quantitatively, Guangzhou, Shenzhen, Foshan, and Dongguan exhibited the largest magnitude of total settlement growth in this subzone. Temporally, settlement increment in Guangzhou is stable in each temporal stage, indicating a stable and robust urban expansion. The same pattern is observed in Foshan, Shenzhen, and Huizhou, but their magnitude after 2005 is relatively minor. For the rest of the prefectures, settlement growth gradually decreased after 2000.

Figure 7 shows the dynamics of settlement expansion in the Belt subzone. As most urban centers are located along coastal line, many positions with settlement expansion are spaced by the sea from the urban center, leading to an inconsecutive distribution on the x-axis, thus, different from other subzones, dispersion becomes the major expansion type. Quantitatively, the expansion magnitude is lower than the Core subzone. Jieyang and Shantou have the largest increment, particularly, expansion in Jieyang mainly took place in the southern direction, while expansion in Shantou occurred in four directions. Temporally, except for the stable settlement expansion in southern Jieyang and western Shantou, expansion in other prefectures is stable yet minor, indicating a steady expansion with a small increment in this subzone.

Figure 8 shows the dynamics of settlements expansion in the District subzone. Based on the distribution of expansion with distance to the urban center, expansion type in this subzone followed the same pattern as the Core subzone. Expansion types in Heyuan and Shaoguan belong to the Infilling, Qingyuan has the dispersion expansion across buffer rings, Meizohu and Yunfu have an obvious edge expansion, but the infilling also occurred in western and northern directions. Quantitatively, the magnitude of expansion in all prefectures is lower than the other two subzones; settlement expansion in eastern and southern directions takes up the majority proportion in most prefectures except Meizhou, indicating the geographical tendency of urban expansion in recent years. Temporally, a considerable magnitude of expansion can be found in Heyuan and Qingyuan post-2000, a phenomenon uncommon in Guangdong province, indicating urban construction exceeding population growth has been ongoing in this subzone since 2000.

4.4. Total Growth at Administrative Scale

Figure 9 demonstrates the total growth rate of settlement and population during the study period for each prefecture. A noticeable difference exists between the two variables. For the Core subzone, the faster growth of population over that of settlements is observed in Guangzhou, Shenzhen, Dongguan, and Zhuhai, all of which have a higher urbanization degree and decent economic condition. Considering the fast urbanization process in China over the past decades, this trend is most likely attributed to sustained population inflow driven by economic opportunities and limited urban expansion. For the Belt subzone, while having a similar growth rate of settlements, the growth rate of population is lower than the Core subzone, a mark of demographic difference. Specifically, Shantou is the only prefecture outside the Core subzone with a higher population growth rate. For the District subzone group, despite the subzone’s economically and demographically constrained condition, settlement growth rates remain remarkably high. Especially in Heyuan, Shaoguan, and Qingyuan. Overall, except Guangzhou, Shenzhen, Dongguan, Zhuhai, and Shantou, the growth rate of urban land prevails over that of population in the majority of Guangdong province.

Growth rate highlights the relativity of growth and can be affected by base number, e.g., growth rate can be large when the original quantity is small and vice versa; thus, growth dynamics of settlement and population are compared in absolute quantity for a more explicit contrast. As demonstrated by Figure 10, the most obvious pattern is that most prefectures have a larger total quantity for population than settlements, showing a converse quantitative trend against Figure 9. However, for Foshan and Zhongshan, the growth rate of settlement is still larger. Temporally, for most prefectures, the greatest population increase happened between 1995 and 2000, followed by a deceleration. The settlements, however, experienced greater expansion after 2000 in the Core subzone, such as Foshan, Dongguan, and Zhongshan. In the District subzone, the temporal increment after 2000 is greater than that during 1995 to 2000, while in the Belt subzone, urban expansion increments are similar between 1995 and 2000, and 2000 and 2005, followed by a gradual decrease.

4.5. Consistency Between Settlement and Population at Buffer Zone Scale

Total growth rate and temporal increment revealed distinct growth pattern between urban land and population at provincial scale; however, to obtain deeper comprehension on urbanization consistency, it is vital to find out the specific spatial range and temporal stage with growth mismatch, thus, the incremental dynamics of both variables were compared at buffer zone scale through a symmetrical display on x-axis to intuitively exhibit their spatiotemporal alignment.

Figure 11 demonstrates a mismatch in urban growth and population dynamics of the Core subzone. In whole, an evident mismatch in quantity can be found in Guangzhou and Shenzhen. Spatially, population-dominant land can be found concentrated within a radius of 10 km around the urban center of most prefectures, especially Guangzhou, Zhuhai, Huizhou, and Jiangmen. Temporally, population increment is greater than settlement during 1995 to 2000, while in the next 5-year span, growth of settlement turned to be the greater than population, e.g., Guangzhou, Foshan, Dongguan, and Zhongshan, the temporal mismatch implies the awareness of urban planning to improve accommodation capacity after the surge of population.

Figure 12 demonstrates the mismatch in the Belt subzone. Overall, the quantity mismatch is observed in a few prefectures, such as Jieyang and Shantou. Spatially, prefectures in the first row, which are located in the west wing of Guangdong, have less growth for settlement and population after 1995 compared to the rest of the prefectures that are located in the east wing, except for population growth in Maoming after 2010. Temporally, the increment of population is larger than urban land after 1995 in Jieyang and Shantou, where the most evident mismatch took place, yet in the western wing, the increment of settlement is greater after 2005, especially in Zhanjiang and Yangjiang.

Figure 13 demonstrates the mismatch in the District subzone. Overall, the quantity mismatch between population and settlement is limited. Evident mismatch occurred within a range of 10 km away from the urban center across the subzone, while as the distance extends, settlements exhibit significantly faster increase after 2000, especially in Heyuan, Shaoguan, and Qingyuan. The faster growth of population only occurred in the central regions of Heyuan and Qingyuan from 1995 to 2015. This faster increase of settlement over population implies as population growth slows due to migration, the faster urban land expansion may result in extensive housing vacancy and low land use efficiency.

5. Discussion

This study examines the spatiotemporal alignment between urban land expansion and population dynamics in Guangdong, China, within the framework of global urbanization and Sustainable Development Goals (SDGs). While previous research has separately investigated urban land change and demographic shifts [57,58,59,60], our work provides a comprehensive analysis of their coupled relationship through detailed spatiotemporal comparisons.

Using buffer zone analysis with 5-year intervals (1995–2019), we reveal distinct development patterns across Guangdong’s three subzones that reflect regional economic disparities. The Core subzone, functioning as Guangdong’s industrial and manufacturing hub, exhibited the most rapid urban growth through infilling and edge-expansion patterns, with urban land expansion consistently outpacing population growth post-2000. This suggests capital-driven development preceded demographic changes. In contrast, the Belt subzone showed moderate, dispersed growth aligned with its export-oriented coastal economy, while the District subzone demonstrated limited infilling and edge expansion with evidence of settlement overfill—likely constrained by ecological priorities and weaker economic foundations. At the provincial scale, 76% of prefectures followed China’s characteristic urbanization pattern of urban expansion exceeding population growth [61,62,63]. Only five prefectures exhibited faster population growth, attributed to sustained population influxes drawn by advanced urbanization and robust economic growth. To achieve balanced urban-population growth at the provincial scale, policymakers must address both unilateral migration patterns and uncontrolled urban sprawl through policy interventions. At a finer scale, population-dominated growth was mainly concentrated within 10 km around urban centers across Guangdong, consistent with findings in the Yangtze River Delta [57]. The population concentration in urban centers suggests that improvement in public transportation can increase the commuting capacity of the urban zone.

Despite the valuable insights gained into the urbanization mismatch using NTL, limitations still exist. First, light saturation in bright city cores, especially for DMSP, distorts radiance values, masking intra-urban density variations. Second, NTL data struggle to capture populations in rural and peri-urban regions due to insufficient light coverage, resulting in systematic underestimation where artificial lighting is sparse or absent. This bias is exacerbated in developing regions where electrification rates may be low. Third, environmental factors like cloud cover, lunar illumination, and seasonal changes like vegetation, introduce temporal inconsistencies, requiring extensive preprocessing. Although calibration and fusion with ancillary data can mitigate some issues, these inherent constraints highlight the need for multi-source validation and context-specific adjustments, and ground-truth demographic data for validation remains essential.

6. Conclusions

This study investigated urban land expansion and its spatiotemporal consistency with population distribution in Guangdong Province, China, from 1995 to 2019. Utilizing global settlement layers and population distribution maps derived from nighttime light (NTL) data, we evaluated the growth of habitable urban land and population, and analyzed their spatiotemporal consistency using buffer zone analysis at 5-year intervals. A key methodological innovation was the use of separate urban–rural mapping approaches to minimize discrepancies in population spatialization compared to traditional census data.

Our analysis of Guangdong’s three subzones—Core, Belt, and District—revealed distinct patterns of urban land expansion influenced by development policies and economic conditions. Overall, the Core and District subzones exhibited infilling and edge expansion, while the Belt subzone showed dispersed growth. Notably, only 5 out of 21 prefectures exhibited faster population growth than urban expansion, likely due to sustained migration driven by economic opportunities and advanced urbanization. Quantitatively, both urban expansion and population growth followed a Core, Belt, District order. Spatially, population-dominated areas were primarily clustered within 10 km of most prefectures’ centers, while the District subzone extensively displayed overfilled settlements, indicating low efficient land use. Temporally, urban growth relative to population in the Core subzone turned from slower pre-2000 to faster post-2000, followed by gradual deceleration, while the Belt subzone maintained balanced growth throughout the study period, the District subzone sustained faster urban growth from 2000 to 2019. These findings provide valuable insights into the complex dynamics of urban land expansion and its relationship with population distribution.

The study highlights significant spatial, temporal, and quantitative discrepancies between these two key indicators of urbanization across Guangdong Province. These discrepancies underscore the importance of considering both urban land expansion and population distribution in the formulation of sustainable urban development strategies and policies. Furthermore, the enhanced methodology developed for population spatialization in this study, which minimizes discrepancies with census data by employing separate urban-rural mapping approaches, can be readily applied to other rapidly developing regions grappling with similar regional imbalances.

7. Future Works

Urbanization processes reduce environmental stability and create social imbalances through uneven demographic distribution, making a thorough understanding of urban growth patterns and their potential mismatch with population dynamics crucial for sustainable regional planning. Using Guangdong, China as a case study, we integrate NTL-based population spatialization with global settlement layers to categorize urban growth patterns across subzones and examine spatiotemporal settlement-population relationships at multiple scales. Validation against census records and benchmark datasets confirms our method’s reliability and generalizability. Our satellite-based framework provides a comprehensive, multidimensional (spatial, temporal, and quantitative) assessment of urban-population mismatches, offering precise references for sustainable development planning.

While advancing current understanding, this study has several limitations that suggest directions for improvement. First, data timeliness is constrained by settlement datasets that lag behind current development conditions. Although newer products like Global Built-up Surface (GBS) offer extended temporal coverage (1975–2030), their post-2020 extrapolated data raises reliability concerns. Higher-authenticity alternatives include ESA WorldCover and Google Dynamic World, though their varying spatial resolutions require harmonization.

Second, while census validation confirms strong NTL-population linear correlations in our study area, local conditions can create nonlinear relationships. Enhanced accuracy could be achieved through the following: (1) integrating supplementary geospatial data (land cover, building footprints, POIs) to better represent population distribution heterogeneity, and (2) employing advanced modeling techniques. Machine learning approaches like random forest and XGBoost improve variable interpretation [64,65], while neural networks excel at processing high-dimensional data and capturing complex nonlinear patterns with inherent noise resistance [66]. In the future, more attention will be paid to the adaptation of supplementary data and complex techniques to provide more detailed population support for the achievement of SDGs.

Third, while buffer zone analysis has been extensively used in urbanization studies, the criterion for radius and ring interval selection lacks consensus, and the sensitivity of results to different parameter selection is scantily discussed. Most research chooses 1 km as an interval for simplicity and determines ring numbers based on the reach of the study region. Ref. [67] leveraged a smaller buffer interval in accordance with the resolution of land use data, coinciding with our parameter selection policy. Refs. [68,69] chose different intervals for Shanghai and eight satellite cities; however, the impacts of different parameters are not analyzed. Future works can test the effect of different parameters on results with the above factors considered.

Like Guangzhou, many regions in the world are experiencing inconsistent urbanization; our methodology can be applied to provide scientific guidance for scientific urbanization planning and achieving sustainable regional development. The reliability of our framework can be validated by local census data or land-use survey information, ensuring agreement with realistic conditions.