Analysis of the Spatio-Temporal Characteristics of Nanjing’s Urban Expansion and Its Driving Mechanisms

Abstract

The expansion and evolution of urban areas are the most perceptible manifestations of the transformation of the urban spatial form. This study uses remote sensing images of Nanjing from 2001, 2006, 2011, 2016, and 2021, along with socio-economic data to analyse the spatio-temporal characteristics of the city’s urban expansion. Furthermore, we utilize a binary logistic regression to quantitatively analyse the driving forces in each stage. We find that from 2001 to 2021, Nanjing’s urban area expanded approximately 3.97 times. Notably, the city started moving from a stage of medium-speed development to rapid development in 2006, and then slowed down and returned to medium-speed development in 2011. The urban land mainly expanded in the north, northeast, southeast, and southwest directions in a lopsided cross-shape roughly along the northwest-southeast direction; meanwhile, the city’s centre of gravity continuously moved towards the southeast. Among the driving factors, neighbourhood (distance from planned commercial centres, railways, and highways), topography, and geolocation (distance from the Yangtze River, and elevation) had a greater, albeit inhibitory effect on urban expansion. However, the effects of different socio-economic factors (GDP per capita, resident population, secondary and tertiary industry, etc.) varied across different time periods.

1. Introduction

Urbanisation has become one of the most important themes of development in most countries worldwide [1]. China’s urbanisation has made unprecedented progress since the reform and opening-up policy. Its urban population increased from 18% in 1978 to 50% in 2010, and is expected to reach 80% by 2050 [2,3]. Urbanisation promotes the development of the regional social economy and provides a good living environment for residents [4,5]. However, urbanisation causes drastic changes in land cover/use, creating environmental problems that cannot be ignored [6,7,8]. In addition, the concomitant loss of biodiversity and intensification of the heat island will eventually affect the sustainable development of the city itself [9,10]. Therefore, to promote the optimisation of urban land cover and the sustainable development of cities, it is important to understand the spatiotemporal characteristics of urban expansion and its driving forces.

Urban expansion is the most intuitive activity in the process of urban spatial-form transformation [11,12]. Due to regional differences in natural, social, and economic conditions, the pattern of spatial expansion varies by city and can have different driving forces behind the expansion [13,14]. Exploring the process of urban expansion and driving forces behind each stage can help differentiate the trajectories of urban development, and provide a reasonable reference for the optimisation of urban land-use structure and future urban development [15]. With the development of graphic science, spatial statistics, modern information technology, and other disciplines, scholars have successively studied the process, laws, characteristics, and driving forces of urban expansion and evolution from different perspectives [16,17]. Numerous empirical studies have been conducted on urban agglomerations, such as the Beijing-Tianjin-Hebei region, the Pearl River Delta [18,19], and cities such as Tianjin, Wuhan, and Hefei [20,21]. Scholars have generalised the patterns and laws of the growth of large cities, including spatial diversification, structural polycentricity, and morphological fragmentation [22,23,24].

Generally, urban expansion is influenced by multiple factors, ranging from social and economic factors to the natural environment [25]. The most statistical commonly-used methods in these studies include multiple linear regression, principal component analysis, correlation analysis, and regression trees [26,27]. Most of these analytical methods study the relationship between urban land expansion and socio-economic factors. Considering that spatial factors are non-negligible, such methods are limited when exploring the driving forces [28]. Logistic regression modelling provides a solution to this problem. Through step-by-step regressions of a set of independent variables and a discrete dependent variable, this model can screen out the factors that have a more significant impact on urban expansion and determine the quantitative relationships between them, as well as the relative magnitude of the impact [29,30].

Nanjing is the capital of the Jiangsu Province and is its political, economic, scientific, educational, and cultural centre. Since the beginning of the 21st century, the city has witnessed significant changes, including rapid economic development and urban expansion. As an important growth pole of the Yangtze River Delta city cluster, it plays an important role in the development of the province and the entire region. Considering Nanjing’s importance, this study collects remote sensing images from 2001 to 2021, and quantitatively analyses the process of the city’s urban expansion and its driving forces using logistic regression modelling. Our aim is to provide a reference for decision-makers for Nanjing’s future urban development, city planning, and strategic layout.

2. Materials and Methods

2.1. Study Area Description

Located in the west of the Yangtze Delta, Nanjing ranges from 31°14′–32°37′ N and 118°22′–119°14′ E. It has a subtropical monsoon climate, an annual average temperature is 15.4 °C, and an average annual precipitation of 1090.4 mm. As one of the three core cities in the Yangtze Delta, Nanjing has made considerable progress this century in terms of urbanisation, and its urban land area has expanded rapidly (Figure 1). This study covered 11 administrative districts within Nanjing’s jurisdiction, with a total area of approximately 6587.57 km² and a permanent population of more than 9.31 million people.

2.2. Data Sources

Three Landsat-5 TM and two Landsat-8 OLI images (

Table 1. Information about the images that were used in this study.

Path/Row	Date	Sensor	Cloud (%)
120038	2001.05.14	Landsat-5 TM	1.00
120038	2006.11.04	Landsat-5 TM	0.00
120038	2011.09.15	Landsat-5 TM	0.00
120038	2016.03.28	Landsat-8 OLI	0.72
120038	2021.03.26	Landsat-8 OLI	0.24

), captured during 2001, 2006, 2011, 2016, and 2021, were downloaded from the United States Geological Survey (USGS). With a spatial resolution of 30 × 30 m, these images have six (for TM) or eight (for OLI) spectral bands at visible and shortwave wavelengths. Other data include the vector data of Nanjing administrative boundaries and Nanjing Statistical Yearbook of the relevant years.

Preprocessing of remote sensing images was performed using ENVI 5.3 software, which includes FLAASH atmospheric correction, radiation calibration, band combination, image mosaic, and clipping. The support vector machine (SVM) method was used to obtain the urban areas of each stage for its high accuracy [31]. The land cover type in this study included urban land, forests, grasslands/croplands, bare ground, and water. The SVM classifier was constructed using a radial basis function. The gamma coefficient was set to 1/6 for the Landsat TM images, 1/8 for the Landsat OLI images. The penalty parameter was set to 100. A confused matrix was used to confirm classification accuracy. A total of 187 validation samples of land cover types were randomly chosen from Google Earth in 2001, 163 in 2006, 188 in 2011, 176 in 2016, and 192 in 2021.

2.3. Research Methods

2.3.1. Spatial Expansion Intensity Index

The Urban Expansion Intensity Index (UEII) refers to the percentage of urban land by a space unit in a unit period as a percentage of its total land area [32]. The higher the value, the faster the expansion. At present, the UEII is an important indicator to describe the speed of urban land expansion over a certain period. Its mathematical expression is

$$I=\frac{U_j-U_i}{UT}\times 100$$

(1)

where I is the average annual urban expansion intensity index over the time i to time j, U_i and U_j are the urban land area during time i and time j, U is the total area of the built-up area of urban, and T is the time span of the research. The UEII standard is divided as follows: 0 to 0.28 is slow development; 0.28 to 0.59 is low-speed development; 0.59–1.05 is medium-speed development; 1.05–1.92 is high-speed development; and >1.92 is very high-speed development [33].

2.3.2. Elasticity Coefficient

The elasticity coefficient of urban populations (UPEC) is one of the indicators used to evaluate the rationality of urban expansion. It refers to the coordination relationship between the speed of urban expansion and the speed of population growth [32]. The UPEC was used in this study to assess the characteristics of urban expansion in the Nanjing over time. The calculation formula is

$$R_{\left(i\right)}=\frac{A_{\left(i\right)}}{P_{o{p}_i}}$$

(2)

where $R_{\left(i\right)}$ is the elastic coefficient of urban expansion in period $i$, $A_{\left(i\right)}$ is the average annual growth rate of the urban area in period $i$, and $P_{o{p}_i}$ is the average annual growth rate of the urban population in period $i$. In theory, the urban area growth rate should be synchronised with the population growth rate (UPEC = 1); However, according to the study by the Chinese Urban Planning and Design Institute on the urbanisation process over time, R_(i) is most suitable when it is 1.12 [34].

2.3.3. Urban Land Weighted Mean Centre

Using time-series urban land fraction images, we used the gravity centre to indicate the direction of urban expansion [35]. The gravity centre was designated as the urban land weighted mean centre (ULWMC). The migration of ULWMCs reflects the overall direction and trend of urban expansion during different periods. Its mathematical expression is as follows:

$$X_t={\displaystyle \sum }_{i=1}^n\left(C_{ti}\times X_i\right)/{\displaystyle \sum }_{i=1}^n{C}_{ti}$$

(3)

$$Y_t={\displaystyle \sum }_{i=1}^n\left(C_{ti}\times Y_i\right)/{\displaystyle \sum }_{i=1}^n{C}_{ti}$$

(4)

where $X_t$ and $Y_t$ are the longitude and latitude coordinates of the ULWMC, $C_{ti}$ is the area of the i-th vector spot at year t, and $X_i$, $Y_i$ are the latitude and longitude coordinates of the geometric center of the i-th vector spot.

2.3.4. Standard Deviational Ellipse

Standard deviational ellipse (SDE) is a spatial statistical method for revealing the multi-faceted characteristics of a variable spatial distribution [36]. Through the centre, main axis, auxiliary axis, and azimuth angle, the SDE reflects the centre of gravity, main trend, secondary trend, and main trend direction of an element in a two-dimensional spatial distribution [37]. The ellipse centre of the SDE is the previously (ULWMC). The SDE was calculated as follows:

$${\mathrm{SDE}}_x=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n{\left(x_i-\overline{X}\right)}^2}{n}}$$

(5)

$${\mathrm{SDE}}_y=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n{\left(y_i-\overline{Y}\right)}^2}{n}}$$

(6)

where $x_i$, $y_i$ are the coordinates of the $i$-th subregion, (X, Y) represents the centre of gravity of the region, and $n$ is the total number of subregions.

2.3.5. Spatial Lateral Variation

The study area was divided into 16 regions at an interval of 22.5°, taking into account the centroid of the study area, to identify the spatial lateral variation in the rate of urban expansion for each study period. Then, we calculated the area of each region’s urban land expansion [38,39].

2.3.6. Driving Forces Analysis

Logistic regression has now been widely used in land cover change studies [30]. This method is based on data sampling, diagnosing the possible probabilities of land cover change by stepwise regression, screening the factors that have more significant effects on urban expansion, and determining the quantitative relationships and relative magnitudes of their effects. Here, we overlaid the land cover maps in ArcGIS for different time points (2001, 2006, 2011, 2016, and 2021) to obtain urban change values (i.e., 0, 1 value, where 1 indicates a non-urban area transformed into urban area and 0 indicates still a non-urban area) for each period and used them as the binary dependent variable.

According to research, socioeconomics, neighbourhoods, topography, and geolocation all have an impact on urban expansion [40,41,42,43]. In this study, we selected nine factors in three categories as independent variables (

Table 2. Independent variables and their description.

Categories	Independent Variables	Description
Socioeconomic	GDP per capita	Gross domestic product per capita
	POP density	Population density
	Resident POP	Resident population
	Secondary industry	Secondary industry production value
	Tertiary industry	Tertiary industry production value
Neighbourhood	DPCC	Distance from planned municipal and sub-municipal commercial centres
	DRH	Distance from railways and highways
Topography and geolocation	DYR	Distance from the Yangtze River
	Elevation	Height above sea level

) based on the referenced literature [44,45].

We used the random point creation tool in ArcGIS to generate 10,000 sample points with a uniform distribution of 5000 within the urban and non-urban areas for each study period. The minimum distance between the point is 100 m. After assigning the values corresponding to the dependent and independent variables to the sample points, attribute tables were derived and the data were normalised. We then used IBM SPSS statistics 26 to conduct binary logistic regression to analyse the driving forces of urban expansion in Nanjing and quantified the impact of different driving factors.

3. Results

3.1. Spatio-Temporal Characteristics of Nanjing’s Urban Expansion

3.1.1. Temporal Characteristics

The result of the validations showed that the accuracy of the classification was greater than 0.80 (

Table 3. A summary of the assessment of the urban area classification’s accuracy. OA, overall accuracy.

Date	OA (%)	Kappa Coefficient (%)
2001	91.71	0.88
2006	91.57	0.86
2011	90.63	0.87
2016	89.92	0.86
2021	90.44	0.87

), which fulfilled the requirement for land cover change evaluation.

Nanjing’s urban area has expanded remarkedly from 291 km² in 2001 to 1154 km² in 2021, with an overall UEII of 0.66. The conditions of urban expansion varied in each period (

Table 4. Intensity of urban land expansion across different periods.

Period (Years)	EA (km2)	Annual EA (km2/Year)	UEII
2001–2006	143.14	28.63	0.43
2006–2011	336.86	67.37	1.02
2011–2016	203.62	40.72	0.62
2016–2021	180.10	36.02	0.55
2001–2021	863.72	43.19	0.66

), with the lowest expansion area (EA) occurring from 2001 to 2006 (143 km²) and the highest EA occurring from 2006 to 2011 (336 km²).

The UEII was 0.43 from 2001 to 2006, which was a stage of low-speed development. Notably, from 2006 to 2011, the UEII increased substantially to 1.02 and the EA reached 336.86 km², thereby entering a stage of high-speed development. However, Nanjing’s speed of development subsequently slowed and returned to a medium level. Overall, Nanjing experienced medium-speed urban expansion, with a UEII following an approximately normal distribution over 20 years, but was high from 2006 to 2011.

Within the city, urban expansion in each district and county differed (Figure 2). Jiangning (JN), Liuhe (LH), Pukou (PK), Yuhua (YH) and Qixia (QX) in the periphery expanded considerably, with JN expanding by 261.37 km². Gaochun (GC), Lishui (LS), JN, and PK saw the fastest growth, with GC increasing its urban land area by 12.66 times over 20 years. However, the rate of average expansion (AER) has decreased significantly across all prefecture-level administrative regions in the past 10 years. Notably, Gulou (GL), Qinhuai (QH), and Xuanwu (XW), which are located in the city centre, had low EA and AER due to spatial constraints.

In terms of UPEC (

Table 5. UPEC over different periods.

Period (Years)	Annual Growth Rate of Urban Area (%)	Annual Growth Rate of Population (%)	UPEC
2001–2006	0.0983	0.0821	1.19
2006–2011	0.1551	0.0425	3.64
2011–2016	0.0528	0.0131	4.02
2016–2021	0.0369	0.0382	0.96
2001–2021	0.1483	0.0586	2.53

), the UPEC of Nanjing between 2001 and 2006 was 1.19. For comparison, the China Academy of Urban Planning and Design notes that during the process of urbanisation in China over the years, the UPEC has been 1.12; thus, the expansion of this period is relatively rational [35]. From 2006 to 2016, the expansion of urban areas was too rapid, while the population influx continued to decrease; consequently, the UPEC continued to increase. However, from 2016 to 2021, the UPEC fell back to 0.96 and the intensification of urban expansion increased. Nanjing’s UPEC from 2001 to 2021 was 2.53, indicating that urban expansion was significantly faster than population growth over this 20-year period.

Nanjing’s urban expansion over the past 20 years has clear temporal characteristics. The rate of expansion in each phase is normally distributed as low-fast-medium, peaking in 2006–2011 and then gradually slowing down. However, the city’s expansion was not well coordinated with population growth and land cover should have been intensified.

3.1.2. Spatial Characteristics

The pattern of urban expansion in Nanjing has changed over the past 20 years (Figure 3). To analyse the spatial characteristics, we calculated the urban expansion in each direction using Xinjiekou as the centre of the circle (

Table 6. Urban expansion from each cardinal direction from 2001 to 2021.

	2001–2006 (km2)	2006–2011 (km2)	2011–2016 (km2)	2016–2021 (km2)
N	15.09	28.67	24.68	7.52
NNE	3.45	14.56	14.24	4.27
NE	6.90	20.80	7.89	9.28
NEE	15.10	17.26	5.29	18.85
E	4.43	12.05	13.39	9.44
SEE	0.81	13.47	4.23	3.15
SE	15.24	28.30	5.91	6.77
SSE	17.71	56.46	30.33	40.96
S	19.95	31.03	6.86	21.01
SSW	4.83	11.65	14.00	0.81
SW	21.96	46.08	20.41	13.06
SWW	3.91	8.17	17.95	10.98
W	1.92	10.57	12.08	16.32
NWW	3.58	5.62	7.80	5.72
NW	1.43	11.05	5.76	5.56
NNW	6.84	21.13	12.80	6.40

).

Between 2001 and 2006, urban expansion occurred mainly in the north (N), north-east-east (NEE), south-east-south (SES), and south-west (SW) (Figure 4). During this period, the spatial characteristics of the city evolved from the original clustering pattern to a cross-shaped axial pattern of development. Between 2006 and 2011, urban expansion was mainly concentrated in the N, west (W), and south-south-east (SSE) directions, and the city showed a lopsided cross-shaped pattern of expansion to the south-east (SE), SW, and north-west (NW). Urban expansion from 2011 to 2016 inherited some characteristics of the previous period. In addition to continuing development in the N and south-south-west (SSW) directions, there was further expansion in the SSW in south-west-west (SWW) directions. Finally, from 2016 to 2021, the core city’s expansion came almost to a stop. Meanwhile, sub-cities or new towns, such as GC, continued to expand outward and became new growth points on the periphery of the city, bringing a noticeable trend of urban’s expansion in the SSE direction, resulting in an overall pattern of quasi-T patterns.

Using Equations (5) and (6), we determined the SDE of Nanjing’s urban area for each year (Figure 5). We also calculated the ULWMCs in each year to analyse the expansion of the city based on Equations (3) and (4) (Figure 6).

Nanjing generally keeps expanding in the NW-SE direction, and its centre of gravity also migrates SE by 13,037 m (

Table 7. Urban expansion of every direction from 2001 to 2021.

Year	Longitude	Latitude	Distance (m)	Rotation Angle
2001	118.822	32.031	—	—
2006	118.743	31.989	5119	South by east 66°
2011	118.877	31.984	3253	South by east 82°
2016	118.842	31.962	4101	South by west 52°
2021	118.849	31.945	2052	South by east 68°

). With the successive construction of JN, QX, LH, and PK, and the continual development of sub-cities and new towns, such as Chunhua, Lukou, and Lishui, the short axis of the SDEs lengthened accordingly, leading to a decrease in its eccentricity and it gradually converged toward a “circle”.

Overall, Nanjing has shown relatively obvious cross-shaped axial development in the past 20 years. In terms of the scope of expansion, the city exhibited sprawling expansion with varying speeds and intensities, while gradually expanding outward. However, several new growth points emerged in the city’s suburbs. As the city’s centre of gravity migrated, the pulls and linkages among different districts and counties were significantly enhanced.

3.2. The Driving Forces behind Nanjing’s Urban Expansion

We constructed a binary logistic regression (LOR) model to analyse the driving forces behind the expansion of urban construction land over four periods. The ROC curve was used to test the explanatory power of each model. The results show that the AUCs of the four LOR models range from 0.7 to 0.9 (

Table 8. Summary of regression coefficients, Wald statistics, significance value (Sig.), and AUC of the LOR model at each stage.

Variable	2001–2006			2006–2011			2011–2016			2016–2021
	B	Wald	Sig.	B	Wald	Sig.	B	Wald	Sig.	B	Wald	Sig.
GDP Per capita	0.619	32.882	0.000 **	0.744	67.366	0.000 **	0.187	39.477	0.000 **	−0.098	11.489	0.001 **
POP density	0.561	31.074	0.000 **	0.823	76.797	0.000 **	−0.225	18.099	0.000 **	0.112	11.501	0.001 **
Resident POP	2.288	31.212	0.000 **	−10.792	107.331	0.000**	9.961	1.171	0.027	−2.302	6.278	0.012
Secondary industry	6.528	63.527	0.000 **	2.726	46.236	0.000 **	−0.213	1.601	0.021	2.384	40.539	0.006 **
Tertiary industry	−0.290	24.955	0.000 **	−0.456	3.728	0.053	0.121	2.535	0.012	−6.298	41.845	0.003 **
DPCC	−0.747	528.331	0.000 **	−0.846	721.404	0.000 **	−0.318	142.068	0.000 **	−0.519	318.143	0.000 **
DRH	−0.678	205.987	0.000 **	−0.800	273.533	0.000 **	−0.327	98.871	0.000 **	−0.361	112.673	0.000 **
DYR	−0.649	209.913	0.000 **	−1.033	311.970	0.000 **	−0.769	246.190	0.000 **	−0.837	210.507	0.000 **
Elevation	0.097	2.739	0.098	−0.096	10.850	0.001 **	−0.141	23.098	0.000 **	−0.258	41.334	0.000 **
AUC	0.828			0.837			0.729			0.744

** indicates significant at the 0.01 level.

), indicating that they have good explanatory power for the urban expansion of Nanjing in each period.

Table 8 shows the results of the four phases, with the regression coefficient (B) indicating that for each unit increase or decrease in the value of the explanatory variable, the incidence ratio of land transformation from others to urban either increased or decreased. The Wald statistic (Wald) indicates the relative weight of each explanatory variable in the model. It can be used in event prediction to evaluate the contribution of each explanatory variable. Significance (Sig.) indicated the significance level for each variable. Urban expansion was significantly impacted by the variable when its significance value was less than 0.01. Instead, this indicated that the variable has no appreciable impact on urban expansion. The results indicate that urban expansion was a consequence of the interaction of neighbourhood, socioeconomic, topography and geolocation. Notably, the roles and strengths of various driving factors varied over each time period.

Neighbourhoods had the most significant impact on the expansion of urban areas, with total contributions of 734.318, 994.937, 222.939, and 430.816, across the four periods, respectively. Meanwhile, the DPCC had the maximum contribution in all periods except 2011–2016, indicating that it was the most essential driving force of urban expansion. In addition, the DRH always ranked second or third (Figure 7), indicating that it also played a crucial role in urban expansion. Both factors had an inhibitory effect on urban expansion; the lower the DPCC or DRH, the more likely land was to be converted for urban use.

Topography and geolocation had a slightly lesser impact on urban expansion than on socioeconomic status during the four time periods. Elevation had a smaller influence. Increasing altitude had no significant effect on urban expansion between 2001 and 2006 but gradually starts to affect expansion in the later period. Meanwhile, DYR has a significant inhibitory effect on urban expansion, with its contribution ranking first from 2011 to 2016 and second during the other periods.

Finally, socioeconomic factors played a smaller role than neighbourhood, topography, and geolocation; however, the ranking of each socio-economic factor varied substantially. The most volatile was resident POP (from fourth to tenth), the effect of which on urban expansion shifted from exerting a stimulating effect initially, to exerting an inhibitory effect from 2006 to 2011, and exerting a slightly significant effect from 2016–2021. Increases in GDP per capita and POP density showed a stimulating effect during time all periods. Secondary industry had a stimulating effect on urban expansion in all periods except 2011–2016, when it had few significant effects. Meanwhile, tertiary industry changed from having an inhibitory effect initially, to a non-significant effect in 2006–2016, and finally an inhibitory effect in 2016–2021.

4. Discussion

Since 2001, Nanjing’s urban land has expanded significantly, with the area expanding approximately 3.97 times. Furthermore, the speed, UEII, and UPEC of urban expansion all initially increased and then decreased. On the one hand, the reason for this trend lies in the fact that as the city rapid expanded, its urban area approached the spatial limits of urban development boundaries. On the other hand, there was an essential change in the orientation of Nanjing’s urban construction from the rough expansion mode of purely pursuing “quantitative” expansion (Nanjing Master Plan (1991–2010)) to an intensive growth model pursuing improvement in “quality” (Nanjing Master Plan (2011–2020)). Regarding the spatial characteristics of urban expansion, Nanjing’s urban land has developed in a cross-shaped axial pattern in four directions because its own topography and geolocation put certain constraints on the directions of urban development. Urban expansion occurred in the flat land of the southeast hinterland and along both sides of the Yangtze River, or across the river. New urban land mainly revolved around the original urban land and core areas of the suburbs, showing sprawling expansion at varying speeds and intensities. This indicates that the sprawl of original construction and the core area of the suburbs promoted urban expansion. Furthermore, the SDE and ULWMC show that the overall city’s weighted mean center continued moving southeast, and that the pulls and linkages among districts and counties were significantly enhanced.

Regarding the driving forces of Nanjing’s urban expansion, we find that neighbourhoods (i.e., DPCC and DRH) exerted a dominating influence. The DPCC is essentially a manifestation of government policy, and expresses, to some extent, the government’s control over the direction of urban development. Nanjing has located municipal and sub-municipal commercial centres not only in the mature urban core areas of Xinjiekou, Fuzimiao, and Hunan Road, but also in new towns and suburban areas, such as Hexi, Jiangbei, Xianlin, Dongshan, and Yongyang (Nanjing Commercial Network Plan (2015–2030)). Socio-economic entities gradually spread into socio-economic ”flows” along linear infrastructures (roads, railways), forming new agglomerations of varying intensities at different distances from the planned commercial centres. Nanjing gradually formed a progressive point-axis pattern of urban expansion, which promoted urban expansion to new towns and suburban areas. Thus, neighbourhoods are a major driving force behind urban expansion.

Topography and geolocation put constraints on Nanjing’s urban expansion. Nanjing’s old city is constrained by the topographic conditions in the east (E), south (S), and NW directions. Out of the four main directions of urban expansion, namely the north-east (NE), SE, SW, and N directions, three are circumscribed by the Yangtze River and the Lao Mountain, where the conditions for development and land potential are less suitable than those in the southeast. This may be a major reason for the existing spatial pattern of Nanjing’s urban land development. In addition to the necessity of opting for development along both sides of the Yangtze River for topographical and geolocational reasons, Nanjing, as an important node in the Yangtze River Economic Belt, cannot thrive economically without the advantages of transportation and economic momentum brought by the Yangtze River. Therefore, while analysing the driving forces, land that is closer to the Yangtze River is likely to have a greater intensity of socio-economic activities, and can thus be more easily transformed into urban land. Furthermore, amplification of the inhibitory effect of elevation on the city’s development may be caused by the following: Nanjing’s lower part is a plain formed by the alluvial deposits of the Yangtze Delta and enjoys good natural conditions; however, as the city expands outward, it begins to encounter hilly areas and the constraints of topographic conditions on construction gradually increase.

Socio-economic factors served as a driving force for urban expansion in Nanjing, but their effect was relatively weak. However, it should be noted that different socio-economic factors have different impacts on urban expansion. For example, an increase in GDP per capita contributed to urban expansion in all time periods. Meanwhile, an increase in the resident population initially promoted urban expansion from 2001 to 2006, which is in line with the general law of urban development. However, it inhibited urban expansion from 2006 to 2011, and did not affect urban expansion from 2011 to 2020. This could be related to the degree of coordination between the rates of urban expansion and population growth. When UPEC was rational, the resident population had a stimulating effect. However, when urban expansion was markedly faster than population growth, the latter had an inhibitory effect. As urban development shifted from outward expansion to inward improvement, population growth was no longer necessarily related to urban expansion. Meanwhile, the increase in population density promoted urban expansion from 2001 to 2011, but inhibited it from 2011 to 2016, which may also be related to the change in the mode of urban development from outward expansion to internal upgrading.

Similarly, secondary and tertiary industries played different roles in Nanjing’s urban expansion. For example, the secondary industry showed a stimulating effect in all periods, except for 2011–2016, when it did not have a significant impact. This may be because the secondary industry in Nanjing promoted the formation of industrial and high-tech zones, such as the Jiangning Economic Development Zone, Zhongshan Science and Technology Park, and Gucheng Industrial Park, which mostly led to incremental land cover and played a role in urban expansion. However, with Nanjing’s industrial upgrading (Nanjing 12th Five-Year Plan for Green Urban Development (2011–2015)), numerous industries with high pollution and energy consumption were banned or upgraded. Thus, the secondary industry’s role in promoting urban expansion began to decrease. However, with the establishment of the Jiangbei New Core City, industrial upgrading has achieved some fruitful outcomes, with the high-tech manufacturing industry becoming the core. Thus, the secondary industry again became a major factor in promoting urban expansion during 2016–2020. Meanwhile, the tertiary industry’s effect sharply contrasts with that of the secondary industry, since the former is mainly concentrated in the stock land of the core urban area. Therefore, its role in urban expansion was not significant or even played an inhibitory role, during the four study periods.

Notably, the planning of railways, roads, and commercial centres, which together constitute neighbourhood factors, essentially make up the layouts and measures taken by the government in response to a city’s continuous socio-economic development. These are explicit factors in our model. In contrast, socio-economic factors implicitly drive urban development and push the city to expand in accordance with policies, while also reacting to the city’s changes, and prompting decision-makers to undertake new policies. In our model, they serve as implicit factors when compared to neighbourhood factors. This may be one reason why their effect is not as significant as that of neighbourhood factors. Meanwhile, given the complexity of urban socio-economic activities, there may be complex interactions among our factors. Hence, the strengths and mechanisms of these effects should be further explored.

5. Conclusions

This study analyses the socio-temporal characteristics and evolution of Nanjing’s urban expansion from 2001 to 2021, and examines the driving forces behind this expansion. We integrate environment, socio-economic, and neighbourhoods as potential factors, and use urban land data extracted from five issues of remote sensing images. Methodologically, we utilise the UEII, UPEC, ESA, SDE, ULWMC, and a logistic regression model.

Our results show that firstly, Nanjing’s urban area has expanded continuously from 291.04 km² in 2001 to 1154.76 km² in 2021, expanding by approximately 3.97 times in 20 years. The intensity of urban expansion first increased and then decreased, with urban expansion occurring at a medium speed from 2001 to 2006, and then accelerating from 2006 to 2011, before slowing down again to a medium speed after 2011. However, urban expansion was not fully coordinated with population growth and the efficiency of urban land cover still needs to be improved.

Second, there are spatial differences in Nanjing’s urban expansion. The city expanded unevenly mainly along four directions, N, NE, SW, and SE. Among these, the SE and N directions exhibited the most expansion. The city’s weighted mean centre continued to move to the southeast.

Third, socio-economic, neighbourhood, topographical and geolocational factors all played significant roles in urban expansion. The topography and geolocation placed constraints on urban expansion and the impact of elevation gradually increased. Neighbourhoods are a major driving force. Specifically, the distance from the planned municipal and sub-municipal commercial centres, and infrastructure, such as railways and highways, had an inhibitory effect. However, the effects of different socio-economic factors varied throughout the study periods. For example, GDP per capita showed a stimulating effect in all periods. The resident population and population density sometimes promoted expansion but sometimes inhibited it. Finally, the secondary and tertiary industries also fluctuated between promoting, inhibiting, and having no significant effect on expansion.