Research on the Correlation between the Dynamic Distribution Patterns of Urban Population Density and Land Use Morphology Based on Human–Land Big Data: A Case Study of the Shanghai Central Urban Area

Abstract

The dynamic distribution of urban population density and the interaction with land use elements involve mutual constraints and guidance. However, in the existing research on the relationship between urban population density and land use, the discussion on the distribution patterns of urban population density typically spans long time periods and uses large spatial units, lacking analysis of the dynamic changes in population density within high granularity land parcels over a day. In studies related to the urban built environment, the complex relationships between different-dimensional land use elements and the dynamic distribution of population density also need further exploration. To address these bottlenecks, this study takes Shanghai’s central urban area as an example. Based on 24 h mobile signaling data on weekdays, weekends, and typical holidays, as well as urban land use data, clustering algorithms are used to summarize patterns of dynamic population density distribution. Pearson correlation analysis is then employed to study the correlation between dynamic population density distribution patterns and different land use elements. The results indicate that various urban land use factors such as locational centrality, functional diversity, transportation accessibility, compactness, and landscape quality have different impacts on the dynamic distribution of population density in spatial units, and the dynamic distribution patterns of population density in different land use types also vary. This research contributes to guiding the optimization of spatial quality and formulating planning and management measures that more effectively match construction intensity with population activity density.

1. Introduction

Based on the United Nations’ World Population Prospects report, the global population is projected to increase from 6.1 billion in 2000 to 8.5 billion by 2030, representing a growth rate of 40%. Cities are the epicenters of population concentration, and the continuous increase in population density has become a common phenomenon in many urban areas. For instance, the population in the administrative regions of cities like Chongqing, Shanghai, and Beijing in China; New Delhi in India; and Tokyo in Japan has surpassed 10 million. However, with the development of information technology, people’s ways of life, and communication are constantly changing and innovating [1]. The spatial mobility of various factors in cities, including human flow, has been greatly enhanced. Not only is the flow speed faster, but the cyclical fluctuations are also shorter [2]. This has a positive impact on enhancing urban vitality and promoting economic development [3]. However, this high-frequency dynamic factor flow can also pose more severe challenges to high-density cities, causing the population density in a certain area to exceed the carrying capacity threshold of the space momentarily, leading to conflicts between people and land use. For example, the stampede incident in Litaeyon district, Seoul, South Korea, in October 2022, triggered a crisis in the management of public space safety in high-density areas [4]. The increasing per capita exposure to pollution and pollution costs can also reduce environmental efficiency [5]. Furthermore, infectious diseases can spread more widely and on a larger scale in high-density environments because of increased population mobility and contact frequency, affecting the safety and health of urban residents [6]. Therefore, understanding the dynamic distribution pattern of high temporal and spatially accurate population density and exploring its correlation with land use patterns are the main bases for scientifically formulating land use policies to meet the activity needs of the population. This information is also a scientific reference for implementing sustainable urban development [7,8].

The dynamic distribution of population density is mutually mapped with land use structure [9], exhibiting a high degree of spatiotemporal heterogeneity [10]. As human beings are the agents of behavior, their behavioral needs and motivations are the driving forces behind the dynamic distribution of urban population density [11]. Land use, as the carrier of behavior, with its functions, attributes, and morphological characteristics, will have differential attractive effects on the agents of behavior, thereby influencing the dynamic distribution of urban population density [12,13]. Exploring the interaction between the dynamic distribution of population density and land use patterns requires not only analyzing which land use factors have an impact on the dynamic distribution of population density but also delving into the differences and relationships in the influences among different dimensions of land use factors.

In the existing research on the correlation between the dynamic distribution of urban population density and land use factors, many scholars have utilized big data such as social networks [14], nighttime lights [15], etc., to summarize the spatiotemporal distribution characteristics of urban population density and analyze the impact of relevant land use. Scholars such as García-Palomares, J. C. et al. used Twitter data to analyze the relationship between dynamic human behavior and land use, finding significant differences in the characteristics of urban population density changes under different land use functions [16]. Shi, Y and others proposed a spatiotemporal correlation mechanism between urban population density and public service facilities, suggesting that the spatiotemporal dynamic correlation is influenced by residents’ daily activities at different times of the day [17]. As research progresses, more and more scholars are expanding their analyses from a single dimension of land use functions to the level of morphological elements such as building form and road network density, exploring the impact of different dimensional land use morphological elements on the dynamic distribution of urban population density. Liu, X et al. divided a city into grids of 1 × 1.1 km to study the impact of morphological elements such as road network density and various service facility densities on the dynamic distribution of population density [18]. Meng, Y. and Xing, H. evaluated the relationship between landscape features and urban population density through regression analysis, finding that the spatiotemporal distribution of urban population density is variable and highly dependent on the multi-level features of the landscape [19]. Ye, Y. et al. focused on morphological factors of density and type, exploring the relationship between urban form and urban vitality through regression models [20]. Other scholars have also used further spatial morphological analysis tools to characterize the architectural form and network configuration elements of cities [21]. For example, Berghauser Pont, M. et al. quantitatively described streets, plots, and buildings to gain a more intuitive and comprehensive understanding of the geometric structure of land use and spatial form [22]. Paraskevopoulos Y and Bakogiannis E explored the relationships among functional density, functional diversity, population density, and functional mixing patterns [23].

The above research indicates that the location, function, transportation, and other factors of land have a significant impact on the dynamic distribution characteristics of urban population density, showing a strong correlation. However, there are also some shortcomings in the current research. First, there is a lack of research on the high-precision dynamic distribution patterns of urban population density, with limited studies on the dynamic distribution characteristics of population density at the block and parcel spatial units and their correlations with land use patterns. Second, there is still a lack of empirical research that proves the differences and interaction relationships of the impacts of different dimensions of land use factors on the dynamic distribution of urban population density.

Faced with the above bottlenecks, the rapid development of mobile information and big data technologies has enabled urban population data sources to be quickly updated, providing a data foundation for conducting high-frequency research on the dynamic distribution of urban population density [24]. Accurate measurements of the built environment also provide a variable foundation for analyzing the complex relationships between land use patterns and urban population density. Compared with traditional population survey data and large-scale regional population density studies, studying the spatial distribution and composition characteristics of real-time urban population density from a dynamic perspective and analyzing the impact of high-precision land use patterns on urban population density are of great significance for understanding the differentiated demands of residents for urban resource spatiotemporal allocation and for the fine management of urban space.

Therefore, this article aims to use high-resolution spatial unit land use data and continuous hourly mobile signaling data as an example to measure and summarize the spatiotemporal dynamic distribution of population density in the central urban area of Shanghai. By selecting different dimensional land use feature indicators and analyzing their correlation with the dynamic distribution of population density, this article seeks to explore the mechanisms through which land use patterns influence the dynamic distribution of urban population density and analyze the interaction relationships among different dimensional land use factors. The goal is to analyze the characteristics and key influencing factors of the dynamic distribution of urban population density through big data analysis, providing a reference for relevant urban planning and land use policy formulation.

The remaining structure of this article is as follows: in the second section, the location selection, data acquisition and preprocessing, indicator selection, and measurement methods will be introduced; the third section will discuss the analysis results; the fourth section will present the findings, innovations, and limitations of this study; and the fifth section will provide a summary.

2. Data and Methodology

2.1. Study Site

Taking into account the data availability and the universality and effectiveness of the research case, the study area for this research is the central urban area of Shanghai. As one of the four direct-controlled municipalities in China, Shanghai is not only a transportation and technology center in China but also a world-renowned financial center. In recent years, Shanghai has experienced rapid urbanization, with a rapid increase in the level of urban construction. The building area grew from 93.591 million square meters in 2010 to 154.805 million square meters in 2022, an increase of nearly 2/3. Not only have the various functions become concentrated, but the land use patterns for residential, facilities, transportation, and other purposes have also become more complex. Additionally, Shanghai is also one of the cities with the highest population concentration globally. As of 2022, the permanent population in Shanghai is 24.8943 million, with a population density reaching 23,000 people per square kilometer. The Shanghai urban area, especially the central urban area, is characterized by high population density, frequent population mobility, comprehensive urban functions, and complex land use patterns. According to statistics, the central urban area of Shanghai covers an area of approximately 664 square kilometers and has a population of around 11.32 million. Despite accounting for only 10% of the city’s total area, it is home to 49.2% of the city’s population, making its dynamic population density distribution characteristics typical (Figure 1). Furthermore, the central urban area of Shanghai features a complex mix of functions, buildings, facilities, and transportation infrastructure, making its land use patterns intricate. Therefore, using the central urban area of Shanghai as an example not only facilitates the identification of the dynamic distribution characteristics and trends in population density but also provides a clearer reflection of the general rules governing the correlation between population density and land use patterns in major global central cities.

2.2. Data

• Mobile signaling data

Mobile signaling data are primarily obtained by extracting information exchanges and timestamps between mobile base stations and mobile phone terminals to determine the spatial location and status information of mobile phone users. The mobile dataset used in this study includes the user data from China Mobile, a mobile communication operator in Shanghai covering the entire city. The dataset includes mobile base station data and individual signaling data. The base station data mainly consist of base station codes and base station latitude and longitude coordinates. The individual signaling data have been anonymized to protect privacy and include anonymous encrypted mobile terminal IDs, signaling occurrence times, base station codes to which a mobile phone was connected during signaling events, and the latitude and longitude coordinates of the mobile base stations.

Table 1. Description of individual signaling data.

Dataset	Data Layer	Contents of Subdataset
Individual signaling data	ID	Anonymous encrypted mobile terminal ID
	Station	Base station code to which a mobile phone was connected during signaling events
	TIME	Signaling occurrence time, reflecting the temporal information of population spatial dynamics
	Longitude	Latitude and longitude coordinates of a mobile base station, reflecting the spatial information of population spatial dynamics
	Latitude

provides an example of individual signaling data. The spatial scope of this dataset covers the Shanghai metropolitan area (16 districts and 1 county), and the time range is 9 working days in 2013, including 4 typical workdays (Wednesday or Thursday), 4 typical weekends (Sunday), and 1 typical holiday. Regarding the impact of weather on people’s travel activities, the weather conditions for each sampling day are clear skies, with maximum daily temperatures of 15 °C in spring, 35 °C in summer, 25 °C in autumn, and 5 °C in winter.

Within the specified days mentioned above, approximately 20 million unique mobile identification numbers were recorded in the mobile signaling data from 9578 base stations in the Shanghai metropolitan area. To avoid the impact of users appearing in the service areas of multiple base stations at the same time, this study considers the base station where the user accumulates the longest stay time within each hour as the user’s primary service base station during that time period. The data are aggregated on an hourly basis, with a daily average of 500–900 million signaling records.

2.

Land use data

The land use data used in this study were provided by the Shanghai Urban Planning and Natural Resources Bureau and were compiled in 2014. The data include 4926 block areas, 13,222 plot areas, 152,947 building areas, and a total road centerline length of 3183 km. The data contain information at five levels, including plots, block areas, and others. The information content for each level is shown in

Table 2. Description of land use data.

Dataset	Data Layer	Contents of Subdataset
Land use data	Land parcel level	Function (land use type), boundary, and area of developed land parcels
	Street block level	Boundary and area of street blocks
	Building level	Planar profile of buildings, number of stories, land area of buildings, function of buildings
	Road transport level	Central line and boundary line of roads, rail transport stations, public transport stops
	Natural environment level	Rivers, water systems, and mountains

.

2.3. Methodology

To determine the correlation between population dynamics and land use under different urban population density patterns, we first cleaned and processed the mobile signaling data from the central urban area of Shanghai. We quantitatively analyzed the overall characteristics of population density and daily fluctuation amplitudes in different spatial units, summarizing their patterns. Next, we calculated the indicators of land use data in the central urban area of Shanghai to obtain quantitative results of land use factors. Finally, through correlation analysis between daily average population density in classified spaces and land use feature indicators, we analyzed the impact of various land use feature indicators such as location, function, transportation, scale, intensity, and landscape on different types of spaces. We also analyzed their mechanisms of action (Figure 2).

2.3.1. Measurement of the Dynamic Distribution of Population Density

This study selected blocks as the basic spatial units for the dynamic distribution of population density. Blocks are composed of several plots, and population density data on the block scale can be obtained by aggregating plot data. Therefore, this study utilized a spatial correlation method based on Thiessen polygons and three-dimensional land use to calculate the density of mobile phone users within each Thiessen polygon. Subsequently, the number of mobile phone users was distributed according to the proportion of the total area of the three-dimensional activity space [25].

To mitigate the impact of daily errors, this study used the average daily population density $\overline{D}d$ over 9 specified days and the hourly average population density $\overline{D}{h}_i$ as quantitative indicators of population density. For the 9 specified days, population density was calculated at the hourly level. This study primarily used the hourly average population density $\overline{D}h$ (i.e., the population density for each hour of the day averaged over a 24 h period) for the quantitative evaluation of dynamic population density. The specific calculation formula is as follows:

$$\overline{D}{h}_i={\displaystyle \frac{\sum _{j=1}^9\left(P_{ij}\right)}{S}}\times {\displaystyle \frac{1}{9}}\overline{D}d={\displaystyle \frac{\sum _{j=1}^9\left(\sum _{i=1}^{24}{P}_{ij}\right)}{S}}\times {\displaystyle \frac{1}{9}}$$

In the above formula, j represents the number of specified days for obtaining mobile signaling data, i represents the 24 h (from 1 to 24, 0 being midnight) within a day, and $P_{ij}$ represents the total population count (i.e., total number of mobile users) for the ith hour of the jth day.

By plotting the hourly population density $\overline{D}{h}_i$ against the time periods, with each hour’s population density as the vertical axis, a fluctuation curve of population density over a day was obtained (Figure 3a). To make the fluctuation patterns in population density at different scales and ranges comparable, this study introduced a curve of daily fluctuation amplitudes of population density to measure the time series of dynamic population density distribution (Figure 3b). The curve of daily fluctuation amplitudes in population density also used the 24 h periods as the horizontal axis, but the vertical axis represents the percentage of hourly average population density to daily average population density for each time period, eliminating the influence of differences in spatial population density magnitudes.

2.3.2. Clustering of Dynamic Population Density Distribution Patterns

The curve of daily fluctuation amplitudes in population density reflects patterns and regularities in human activities. In this study, the daily fluctuation amplitude curve of population density for spatial units was chosen as the time series for analysis, and spatial units with similar sequence characteristics were clustered. The clustering method involved an improved CORT method that combines distance measurement between time series and time series autocorrelation [26]. The similarity in the temporal evolution of two time series was measured by the first-order autocorrelation coefficient. The first-order correlation coefficient is defined as shown in Equation (1), where t represents the moment in time. By measuring the similarity, we can obtain a similarity matrix (2).

$$CORT\left(X,Y\right)={\displaystyle \frac{\sum _{t=1}^{T-1}(X_{t+1}-X_t\left)\right(Y_{t+1}-Y_t)}{\sqrt{{\sum _{t=1}^{T-1}(X_{t+1}-X_t)}^2}\sqrt{{\sum _{t=1}^{T-1}(Y_{t+1}-Y_t)}^2}}}$$

(1)

$$\mathrm{V}=\left[\begin{array}{c}\begin{array}{c}{V}_1\\ V_2\end{array}\\ ⋮\\ \begin{array}{c}{V}_m\end{array}\end{array}\right]=\left[\begin{array}{ccc}{L}_1& L_2& \dots \end{array}\begin{array}{c}\hspace{1em}{L}_m\end{array}\right]=\left[\begin{array}{ccc}CORT\left(\mathrm{1,1}\right)& \cdots & CORT\left(1,m\right)\\ ⋮& \ddots & ⋮\\ CORT\left(m,1\right)& \cdots & CORT\left(m,m\right)\end{array}\right]$$

(2)

Subsequently, the similarity matrix was clustered using the K-Medoid method, which is an improvement of the K-means algorithm and has greater robustness against noise and outliers [27]. Cluster analysis identifies spatial units with similar daily fluctuation amplitude curves of population density. The number of clusters is determined using two metrics, the Silhouette coefficient [28] and the Dunn index [29], to consider the comprehensive results of the clustering metrics and the size of the original data. In this study, the number of clusters, denoted as K, was set to 8. The main characteristics of the primary clusters will be discussed in Section 3.1.

2.3.3. Measurement Methods for Land Use Factors

In studies of urban population density, researchers recognize the role of external environmental conditions in space. They often choose macroscopic physical environmental characteristics such as location, land use, and transportation as the main influencing factors for analysis [30]. In terms of location, population activities exhibit differences in various urban regions [31]. The central areas of cities have a strong ability to attract population aggregation in surrounding areas because of the dense public facilities and the service functions of the tertiary industry. The closer the spatial relationship between a unit space and the central urban area and the stronger its locational centrality, the greater the spatial attractiveness for the population and their activities [32]. This reflects the influence of locational factors on the distribution of population density [33,34]. In general, a mix of diverse land uses and building functions can meet the diverse needs of the population and promote an increase in population density. Good road conditions and public transportation systems directly affect the aggregation and dispersal capacity of the population in a particular area.

In specific spatial studies of urban population density, in addition to macro-level influencing factors, the characteristics and quality of small-scale spaces have a significant impact on population density. Factors such as construction scale and ecological environment directly influence the temporal and spatial outcomes of the dynamic distribution of urban population density [35]. The humanized spatial scale directly affects people’s use of spatial places and their comfort in activities within the space. It enables spaces to attract and accommodate people for longer periods, thereby stimulating more activities to occur within the space [36]. The importance of a good ecological environment and landscape in shaping modern urban spaces is self-evident. Green spaces, parks, lakes, and waterfront areas are all important components of urban public spaces. In high-quality landscape environments, there are abundant and diverse population activities, which not only attract people but also bring about spillover effects that lead to population aggregation in surrounding areas [37].

Based on the physical environmental conditions and internal spatial characteristics, this article categorized the impact of land use morphology into six aspects with 12 factors as follows: location centrality, functional diversity, transportation accessibility, scale intensification, form compactness, and landscape quality. Each factor was quantitatively measured to establish a quantitative evaluation system for land use factors (

Table 3. Quantitative evaluation system for land use factors.

Dimension		Measure	Computational Formula	Remarks
A	Location centrality	City center location index (A1)	$CL{I}_n={\displaystyle \sum _{k=1}^K}{\displaystyle \frac{\sqrt{S_k{G}_k}}{{D_{nk}}^2}}$	In the formula, K is the number of city centers; Gk is the total building area of public service facilities in the k city center; Sk is the total area of developed land in the k city center; and Dnk is the straight-line distance from the n spatial unit to the k city center.
		District center location index (A2)	${DLI}_n={\displaystyle \frac{\sqrt{S_c{G}_c}}{{{\left[Min\right(D}_{nc}\left)\right]}^2}}$	In the formula, c is the district center to which the straight-line distance from the n spatial unit is the smallest; Gc is the total building area of public service facilities in district center c; Sc is the total area of developed lands in the district center c; and Min(Dnc) is the straight-line distance from the n spatial unit to district center c.
B	Functional diversity	Land use diversity (B1)	${LDI}_n=-{\displaystyle \sum _{u=1}^U}\left({\displaystyle \frac{p_u}{{\sum }_{u=1}^U{p}_u}}\times \mathit{ln}({\displaystyle \frac{p_u}{\sum _{u=1}^U{p}_u}})\right)$	In the formula, pu is the total area of the u type of developed land in the n spatial unit and U is the total number of land development types in the spatial unit.
		Building functional diversity (B2)	${FDI}_n=1-{\displaystyle \sum _{u=1}^U}{\left({\displaystyle \frac{q_u}{{\sum }_{u=1}^U{q}_u}}\right)}^2$	In the formula, qu is the total building area of the u type of developed land in the n spatial unit and U is the total number of land development types in the spatial unit.
C	Transportation accessibility	Road network accessibility (C1)	${RAI}_n={\displaystyle \frac{L}{S}}$	In the formula, L is the total central line length of roads in the spatial unit and S is the total area of developed land in the spatial unit.
		Rail-transit convenience (C2)	${RTI}_n=D_0-Min\left(D_{nr}\right)$	In the formula, Dnr is the straight-line distance from the n spatial unit to the rail transport station; Min(Dnr) is the minimum of the Dnr measurements; and D0 is a constant, indicating the minimum distance.
D	Scale intensification	Plot ratio index (D1)	${PRI}_n={\displaystyle \frac{Q}{S}}={\displaystyle \frac{\sum _{u=1}^U{q}_u}{\sum _{u=1}^U{p}_u}}$	In the formula, Q is the total building area in the n spatial unit; S is the total area of developed land in the n spatial unit; and U is the total number of land development types.
		Building density (D2)	${BDI}_n={\displaystyle \frac{M}{S}}$	In the formula, M is the total land area of buildings in the spatial unit and S is the area of the spatial unit.
E	Form compactness	Space fragment index (E1)	${SFI}_n=\left(F-1\right)\times {\displaystyle \frac{Min\left(p_u\right)}{S}}$	In the formula, F is the total number of plots in the nth spatial unit; Min(pu) is the area of the smallest plot in the spatial unit; and S is the total area of the n spatial unit.
		Plane shape index (E2)	${PSI}_n={\displaystyle \frac{\sqrt{\pi S}}{V}}$	In the formula, S is the total area of developed lands in the spatial unit and V is the length of the planar boundary of the spatial unit.
F	Landscape quality	Green coverage index (F1)	${GCI}_n={\displaystyle \frac{E_{500\mathrm{m}}}{S_{500\mathrm{m}}}}$	In the formula, E is the area of green space in a 500 m buffer zone and S is the area of the 500 m buffer zone in the n spatial unit.
		Space hydrophilic index (F2)	${SHI}_n=D_0-Min\left(D_{nw}\right)$	In the formula, Dnw is the straight-line distance from the n spatial unit to the w water system in the vicinity; Min(Dnw) is the minimum of the Dnw measurements; and D0 is a constant, indicating the minimum distance.

).

2.3.4. Methodology for the Correlation between Dynamic Distribution of Population Density and Land Use Factors

Pearson correlation coefficient was used in this study to characterize the degree of correlation between the dynamic distribution characteristics of population density and the morphological features of land use. The Pearson correlation coefficient between two variables X and Y is defined as the covariance in the two variables divided by the product of their standard deviations. It can also be equivalently defined as:

$${\mathrm{r}}_{xy}={\displaystyle \frac{\sum _{i=1}^n\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)}{\sqrt{\sum _{i=1}^n{\left(x_i-\overline{x}\right)}^2{\left(y_i-\overline{y}\right)}^2}}}$$

where $\overline{x}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{x}_i$,$\overline{y}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{y}_i$. The coefficient ${\mathrm{r}}_{xy}$ ranges from −1 to 1. If the variables are directly related, the sign of the correlation coefficient is positive; if the variables are negatively related, the sign of the correlation coefficient is negative. If ${\mathrm{r}}_{xy}$ is 0, then x and y are considered uncorrelated. The magnitude of the correlation coefficient can be used to determine the degree of correlation between the two variables. When 0 < |r| ≤ 0.1, it is considered a weak correlation; when 0.1 < |r| ≤ 0.3, it is a low correlation; when 0.3 < |r| ≤ 0.5, it is a moderate correlation; when 0.5 < |r| ≤ 0.8, it is a high correlation; when 0.8 < |r| < 1, it is a significant correlation; when |r| = 1, it is perfectly linearly correlated. In this study, the Pearson correlation coefficient was calculated using SPSS 19.0 software.

3. Results

3.1. Characteristics and Patterns of the Dynamic Distribution of Population Density

The results (Figure 4) show that in the central urban area of Shanghai, the average hourly population density is 164.0 people per hectare, which is nearly 4.5 times the city’s overall population density and 7.9 times the population density in the outer suburbs. Throughout a typical day, the underestimated and peak values of population density are 153.4 people per hectare at 5 a.m. and 174.3 people per hectare at 6 p.m., with a range of 20.9 people per hectare.

The 24 h period was divided into five time intervals, including nighttime, daytime, evening, early commuting, and late commuting, to explore the differences in and fluctuation characteristics of population density during these five time periods (Figure 5). During the nighttime from 0 a.m. to 6 a.m., the population density fluctuates smoothly within the lowest range. From 6 a.m. to 9 a.m. during the early commuting period, the population density sharply increases because of early commuting activities such as work and school. From 9 a.m. to 5 p.m. during the daytime period, the population density gradually rises and fluctuates within a higher range. During the late commuting period from 5 p.m. to 8 p.m., the population density starts to decrease gradually. Finally, from 8 p.m. to before midnight, the population density rapidly declines until it reaches the low-density range of nighttime.

By conducting cluster analysis on the daily fluctuations in population density in each block unit, this study identified the five most common and significant types of fluctuation patterns in the central urban area of Shanghai (as shown in

Table 4. Modes of the population density dynamic distribution.

Types of Daily Fluctuation Curve of Crowd Activities	Features	Distribution	Land Use Patterns
Type-1	It describes a situation where there is a short-term, extremely high peak value during a day, leading to strong fluctuations with high and low oscillations.		Large commercial facility land, public facility land, business office land, and residential land adjacent to them.
Type-2	It maintains a relatively stable high-intensity state during the day, with only transient increases and decreases in population density occurring during the morning and evening commuting periods, respectively.		Mainly composed of large industrial land and commercial consulting land, with a small amount of residential land and commercial land.
Type-3	It maintains a relatively stable low-intensity state during the day, with only transient decreases and increases in population density occurring during the morning and evening commuting periods, respectively.		The central urban area is mainly composed of small residential communities and under-construction sites; the peripheral areas are primarily focused on large residential land functions.
Type-4	The population density remains relatively low and stable throughout the day in each time period.		Mainly composed of agricultural and forestry land, suburban green space, and non-construction land; partially consists of industrial land and logistics warehousing land.
Type-5	There is a peak inflection point in population density during the morning and evening commuting periods.		Primarily consists of large-scale residential land and educational and research land.

). These five types of fluctuations, when coupled with land use functions, were found to occupy 86.7% of the total land area in the central urban area of Shanghai. It is important to note that the extraction of these five types primarily focuses on the overall and significant temporal fluctuations in population density, overlooking the local variations and differences in the fluctuation curves and merging some unique cases of specific fluctuations.

3.2. Measurement Results of Land Use Characteristics

This article conducted a quantitative analysis of the land use characteristics in the central urban area of Shanghai across the following six aspects: location, function, transportation, scale, form, and landscape. The quantitative analysis results are presented in Figure 6 below.

The areas most influenced by the radiation of the two municipal central districts are concentrated within the People’s Square central district and the Lujiazui central district. The centralness values of the blocks in this area reach above 600, with centralness decreasing sharply from the inner central district towards the outer ring layers, reaching the weakest centralness value at the periphery of the central city, which is less than 0.2. The uneven distribution of the district-level central districts results in a significant east–west disparity in district centrality. The results of land use diversity are consistent with the diversity of building functions. Areas with lower functional diversity are mainly distributed in peripheral regions such as Taopu Town, Zhenxin Street, Zhangjiang Town, Jinqiao Town, Gaodong Town, and Gaoqiao Town. Additionally, Xinjiangwan City and Songnan Town also exhibit lower functional diversity. Areas with the strongest road accessibility are concentrated in the region west of the Huangpu River and within the Inner Ring Road. Areas with lower road accessibility are mainly located in the northwest and southeast peripheries of the central city, generally less than 0.003. For convenient rail transit access, areas with higher values are concentrated within the Inner Ring Elevated Road. Because of the dense distribution of rail stations, these highly accessible block units show a continuous and clustered trend. The plot ratio decreases overall from the inner central city towards the outer areas, with low plot ratio areas mainly distributed in the outer periphery of the central city. Areas with high building density aggregation are mainly located along the west bank of the Huangpu River in the Bund, Yuyuan, Xiaodongmen Street, and Tilanqiao Street areas. In terms of fragmentation and compactness, areas with different degrees of fragmentation are relatively dispersed in the central city. Areas with high compactness values are evenly distributed in the central city, while areas with low compactness values are mainly linearly distributed downstream of the Huangpu River and are more concentrated in the southeast and northwest corners of the outer periphery of the central city. Areas with high green space coverage are patchily distributed within the central city, especially along the east bank of the upper Huangpu River and the west bank of the lower Huangpu River where riverside green spaces are concentrated. Regarding spatial waterfront accessibility, the high waterfront accessibility spaces in the northern half of the central city exhibit a linear network distribution, while areas with lower waterfront accessibility within the water system network are distributed in a perforated block-like pattern.

3.3. Contrast of Differences in the Correlation between Population Density Dynamic Distribution Patterns and Land Use Factors

The quantified evaluation results of the 12 spatial factors for each block unit in the central urban area of Shanghai were standardized. Through classification statistics, the average quantitative values of spatial characteristic indicators for the five population density distribution types in all blocks were obtained. These results are presented in

Table 5. Mean normalized measurements of land use factors for the modes of population density fluctuation.

Factors of Urban Form		Type-1	Type-2	Type-3	Type-4	Type-5
Radar chart of land use influencing factors
Location centrality (A)	City center location index (A1)	8.05	1.23	0.91	0.34	2.05
	District center location index (A2)	7.56	2.04	0.49	0.38	2.48
Functional diversity (B)	Land use diversity (B1)	4.53	2.89	0.82	1.37	2.37
	Building functional diversity (B2)	4.40	2.76	0.86	1.57	2.44
Transportation accessibility (C)	Road network accessibility (C1)	4.07	3.14	2.52	1.57	3.32
	Rail transit convenience (C2)	4.60	3.16	2.84	1.66	3.62
Scale intensification (D)	Plot ratio index (D1)	5.17	2.94	0.93	0.62	2.91
	Building density (D2)	4.00	2.40	1.36	1.00	2.53
Form compactness (E)	Space fragment index (E1)	2.93	2.28	1.40	1.55	2.17
	Plane shape index (E2)	2.13	2.10	2.01	1.81	2.11
Landscape quality (F)	Green coverage index (F1)	2.37	2.25	3.08	2.72	1.56
	Space hydrophilic index (F2)	1.92	2.00	2.04	2.09	2.01

and Figure 7.

The Type-1 spatial pattern exhibits the highest values in factors such as location centrality, functional diversity, scale intensification, transportation accessibility, and form compactness among all types. The quantified factors of each type have a significant impact on their population densities, with location centrality being the primary factor influencing population density.

The Type-2 spatial pattern ranks just below the Type-1 spatial pattern in the quantified results of functional diversity and form compactness factors. The influence of locational centrality in unit space shows a significant gap compared with the Type-1 spatial pattern. The impact intensity of quantified factors on population density for each type is relatively balanced, with transportation accessibility having a relatively significant influence.

The Type-3 spatial pattern ranks at the lowest level in the quantified results of the land use diversity (B1) and building functional diversity (B2) factors while showing the highest values in the quantified results of the green coverage index (F1) and space hydrophilic index (F2) factors among all types. For this unit space, the quantity and quality of public spaces are important factors influencing its population density.

The Type-4 spatial pattern exhibits the lowest values among the five types in indicators such as location centrality, transportation accessibility, and scale intensification. Additionally, the quantified value of the plane shape index (E2) is slightly lower than the other types. It can be inferred that land use factors have a relatively minor impact on this type.

The Type-5 spatial pattern ranks just below Type-1 in the quantified results of location centrality and transportation accessibility land use factors. Apart from the green coverage (F1) and space hydrophilic index (F2) factors, which show minimal correlation quantified results, the standardized quantified values of other land use impact factors are at a moderately high level among all types. For this unit space, besides landscape factors, other factors have a significant impact on its spatial population density.

4. Discussion

With the rapid development of information and communication technology, collecting vast amounts of population activity data has become easier, providing valuable resources for studying the dynamic distribution of population density and its relationship with land use. In this study, mobile signaling data were used to identify the characteristics of dynamic population density distribution, revealing patterns and regularities in population density dynamics. By investigating the impact of different dimensions of land use morphology on the dynamic distribution of population density, new insights and recommendations are provided for optimizing urban land use and formulating zoning policies.

In the existing studies, scholars generally agree that the scale of a space is determined by the physical dimensions within the space, such as the area of functional land use, building area, building height, etc. This scale determines the space’s capacity to accommodate people, with intensively scaled environments having a stronger capacity to accommodate populations and their activities [38,39]. Our research also confirms this point, as the intensity of spatial scale and transportation accessibility are the primary influencing factors affecting the dynamic distribution of urban population density. This also indicates the consistency between human activities and urban spatial structure, emphasizing the importance of incorporating human activities when discussing urban spatial centrality [40]. Furthermore, we found that although the plot ratio index (D1) has a decisive impact on Type-1 land use, Type-4 land use, and Type-5 land use, for large residential land uses mostly located on the outskirts of the city, rail transit convenience (C2) has a stronger promoting effect on population density. This to some extent indicates that enhancing the connectivity between spaces and their surroundings will help improve the convenience of space utilization for different populations [41,42]. The interaction and coordination between functions and space demonstrate that integrated and converging road networks and convenient public transportation systems are prerequisites for promoting population aggregation [43,44]. The study also found that location centrality has a significantly higher impact on population density in urban fringe areas compared with urban central areas. From the perspective of increasing population density, functional diversity has a much greater impact on non-residential land uses than on predominantly residential Type-3 and Type-5 land uses. Regarding landscape quality factors, it was observed that their impact on Type-3 land, primarily residential in inner-city areas, is much greater than on Type-5 land, mainly residential and educational research-oriented, located in outer-city areas and along rail transit lines. Compared with the other factors, form compactness has a relatively weak impact on all types of land uses.

The research results can provide a basis for policymakers to formulate policies from the perspective of population spatiotemporal behavior. In the traditional urban planning and management paradigm, the allocation and management of urban public services and infrastructure are based on static population indicators. However, in the increasingly complex, dynamic, and diverse activities of urban residents, this model no longer matches the needs of urban residents. Revealing the dynamic distribution patterns of population density and identifying the core land use factors that determine its distribution are of significant importance for achieving efficient resource allocation.

This study provides two reference approaches for policy formulation and urban governance. Firstly, it proposes a method to measure the characteristics of population density fluctuations, identifying peak periods and fluctuation patterns in population density in different land uses and dynamically monitoring trends and changes in urban population density. This could provide a possible technical path for the prevention of public safety issues and urban governance. Secondly, the urban spatial classification method based on the dynamic distribution patterns in population density and the study of land use impact factors based on the types of dynamic population density distribution can provide a new classification and guidance control approach for evaluating and optimizing existing urban spaces. It can offer zoning guidance for future planning revisions and provide new decision support for facility construction, site selection, and land use layout optimization.

This study also has certain limitations. Firstly, using base stations as positioning accuracy units inevitably introduces spatial errors. It cannot analyze individual-level behavioral characteristics such as inflow, stay, and outflow of individuals in unit spaces. Additionally, mobile signaling data can only reflect overall population density fluctuation information characterized by population quantity, without considering the fluctuation in population density among different demographic groups such as gender, age groups, social classes, which may result from the preference of different groups for various spaces. Secondly, it is important to clarify that Pearson correlation analysis is a valuable tool for identifying statistical associations between variables, but it cannot establish causal relationships on its own. This study primarily explores the dynamic distribution characteristics of population density in the central urban area of Shanghai as a whole. However, it lacks an in-depth exploration of the spatiotemporal characteristics and differentiation patterns of population behavior. Although it initially reveals the differential impact of different dimensions of land use factors on the dynamic distribution of population density, further research is needed to explore the coupling relationships among land use factors.

5. Conclusions

This paper quantitatively measures population density fluctuations and land use characteristic indicators. It represents the temporal fluctuations in population density through the daily fluctuation curve and fluctuation amplitude curve of population density. By using the CORT method to obtain the similarity matrix of the dynamic distribution of population density, a clustering algorithm is employed to cluster the population density fluctuation curves of spatial units in the central urban area of Shanghai and extract five typical patterns. For the identified five typical patterns, correlation analysis is conducted between daily average population density and land use characteristic indicators. This study analyzes the differential impact of different land use factors on the dynamic distribution of population density and combines specific land use functions and locations to analyze their mechanisms.

The research findings indicate that Type-1 and Type-2 are mainly located near city and district centers, with a focus on commercial and business land uses, exhibiting strong agglomeration characteristics closely related to resident commuting behavior. Type-3 land use has a slightly more dispersed spatial distribution compared with Type-5, with a tendency towards residential functions. In contrast to the previous four types, the population density of Type-4 land use is mostly determined by land use functions, primarily industrial warehousing and non-construction land. The primary influencing factors for Type-1 are the plot ratio index (D1), building density (D2), and road network accessibility (C1). Increasing road network density and enhancing development intensity can effectively promote the dynamic distribution intensity of this type of space in spatial planning and environmental improvement. Type-2 still faces challenges such as poor location centrality and lack of quality landscape resources, which are the main spatial factors weakening its spatial intensity. The primary influencing factors for Type-3 include road network accessibility (C1), rail transit convenience (C2), and the plot ratio index (D1). High levels of transportation accessibility and quality landscape resources are the main spatial characteristics of this type, leading to low intensity during the day, stable high intensity at night, and sharp fluctuations during commuting hours due to a focus on residential functions and insufficient production and service functions. The primary influencing factors for Type-4 include land use diversity (B1), building function diversity (B2), the plot ratio index (D1), and building density (D2). The primary influencing factor for Type-5 is the plot ratio index (D1), with high transportation accessibility as the dominant advantage. The formation of this type is mainly due to the spatial functional composition dominated by residential functions and the transportation conditions near urban rail transit line stations, allowing people to gather and disperse quickly during commuting hours. By analyzing the patterns and relationships in Shanghai, this study provides a framework and reference for urban planners and policymakers in other global central cities to optimize spatial quality, formulate planning measures, and manage construction intensity based on population activity density.

In future research, we will reference existing studies to address the limitations of mobile signaling data in reflecting individual social attributes, psychological perceptual information, etc. Future research could utilize data sources such as Location-Based Services (LBS) data, social network data, etc., to combine behavioral patterns, individual characteristics, and spatial–temporal distribution characteristics of population density, establishing a more comprehensive framework for analyzing population density spatiotemporal features. Additionally, future research could incorporate implicit morphological indicators such as land economic indicators and land policy indicators on top of explicit land use morphological factors. This would help establish a more comprehensive and integrated evaluation index system for land use impact, allowing for a deeper and more systematic analysis of the coupling relationships between land use factors and the dynamic distribution of population density.