Differences in Urban Vibrancy Enhancement among Different Mixed Land Use Types: Evidence from Shenzhen, China

Abstract

Mixed land use has the advantages of promoting the economic and intensive utilization of land and improving the efficiency of land use, which can help alleviate the current urban problems and promote the sustainable development of cities. Existing studies have usually used quantitative indicators to reflect complex and diverse mixed land use situations, and the conclusions obtained usually cannot provide a basis for functional selection in mixed land use practices. Therefore, this study took Shenzhen as the study area to explore whether there are differences in the urban vibrancy enhancement among different mixed land use types. First, the block-scale mixed land use dataset of the study area was constructed. Second, the spatial distribution characteristics of the main functional types and urban vibrancy in the study area were explored. Finally, the impact of mixed land use types on urban vibrancy was explored by using a multiple linear regression model and setting land use type as the dummy variable. The results show that the number of mixed-function blocks in Shenzhen is relatively small, and the mixed land use degree still needs to be improved. Among the 12 main land use types in the study area, those containing industrial land are usually clustered in the northern industrial area of Shenzhen, those containing public or commercial service land are usually clustered in the city center, and those containing residential land are widely distributed in the study area. From the perspective of urban vibrancy, there is a phenomenon of “jobs–housing mismatch” in Shenzhen, as well as a problem of low urban vibrancy in the peripheral areas of the city. In addition, the urban vibrancy intensity of mixed land use types including residential or commercial land is higher, such as “administration+residential”, “residential+commercial”, “industrial+residential+commercial”, and “administration+residential+commercial” land, which includes residential or commercial land, is stronger, while. However, the urban vibrancy stability of mixed land use types including industrial land is higher, such as “industrial+residential” and “industrial+administration” land. The results of this study can provide a basis for future mixed land use practices in terms of land use type selection. For the urban central areas and subcenters in urban peripheral areas, mixed land use types such as “administration+residential”, “residential+commercial”, and “administration+residential+commercial” can be selected to enhance the urban vibrancy stability of the area. For industrial parks in urban peripheral areas, mixed land use types such as “industrial+residential”, “industrial+commercial”, “industrial+administration+residential”, and “administration+residential+commercial” can be selected to enhance the urban vibrancy intensity of the area.

1. Introduction

Since the post-industrial era, many big cities around the world have begun to decline, and urban vibrancy has gradually received attention. Vibrancy can promote the generation of positive emotional experiences, creativity, resilience, and prosperity, and it is a fundamental element in creating a livable and active living environment and improving the quality of urban life. Vibrancy contributes to create an attractive, competitive, resilient, and prosperous urban environment, which is the driving force and key engine of urban development, as well as an important indicator of the sustainability of urban development [1,2,3]. Creating sustainable urban vibrancy is crucial for urban development. A low level of urban vibrancy will lead to the ineffective utilization of urban infrastructure and resources, stagnation of economic development, and loss of talent, information, and capital, forming a series of problems such as monotonous and undifferentiated urban space and even “ghost cities”, which seriously hinder the healthy and orderly development of cities [4,5]. Therefore, improving urban vibrancy is essential for the health monitoring, compact development, innovative growth and people-centered urbanization of cities [6].

Jane Jacobs, an American scholar, argued that functional urban zoning simplifies and purifies urban activities and severs the organic connections between urban functions, which lead to the loss of diversity and richness of urban life [7]. She argued that functional mixing is an effective way to stimulate urban vitality and help create safe and livable urban spaces [8]. Mixed land use can significantly reduce commute trips and travel distances by co-locating multiple urban functions [9,10], as well as promote non-motorized travel and relieve traffic pressure [11,12,13]. Moderate mixed land use can improve the attractiveness and livability of residential areas [14], which helps to provide more housing and increase the investment value of housing [15,16]. In addition, mixed land use can prolong the physical activity time of residents [17], reduce the risk of death for the elderly [18], reduce the likelihood of community burglaries [19], and promote increases in racial diversity and income equality [20]. Mixed land use can not only enhance urban vibrancy but also contribute to sustainable regional development [21].

Since the reform and opening up, China has experienced rapid urbanization, with the urbanization rate increasing from 17.92% in 1978 to 65.22% in 2022. During the period of rapid urbanization, China’s economy and society developed rapidly, leading to large-scale urban construction and the development of a large number of single-function blocks within cities, resulting in problems such as traffic congestion, environmental pollution, soaring housing prices, and a decline in urban quality and vitality [22,23]. After a profound reflection on traditional urbanization practices, the concept of “people-centered new urbanization” has been proposed to shift the focus of China’s urbanization from quantity to quality. Recognizing that mixed land use is an important way to achieve enhancements in urban vibrancy and high-quality development, both national and local governments in China have issued policy documents containing relevant contents to promote the implementation of mixed land use [24]. However, China’s current planning and management system still does not provide a clear definition of mixed land use, nor does it provide a corresponding management model, making mixed land use in Chinese cities highly autonomous, and thus it has high research value [25]. Shenzhen is one of the earliest cities in China to promote the practice of mixed land use. Since 2010, it has successively issued relevant policy documents. Therefore, it was particularly important to take Shenzhen as the study area to explore its current urban vibrancy and mixed land use status, as well as the selection of mixed land use types with the goal of improving urban vibrancy.

Existing studies have shown that mixed land use has a strong correlation with urban vibrancy [26]. It is a positive predictor of urban vibrancy and plays a leading role in promoting urban vibrancy [27]. In terms of specific urban areas, mixed land use is significantly and positively correlated with the urban vibrancy of areas such as neighborhood spaces [28], subway stations [29], and waterfront areas [30] in a city. When exploring the impact of mixed land use on urban vibrancy, indicators such as the entropy index [31,32], dissimilarity index [33], and balance index [34] have usually been used to quantify mixed land use and to serve as the independent variables of a model. However, the conclusions obtained in this way can only reflect the impact of changes in the mixed land use degree and cannot provide a reference for mixed land use practices in terms of land use type selection. Traditional land use data are difficult to obtain and slow to update, while the emerging geospatial big data have the advantages of being highly available, are quickly updated, and provide accurate spatial location information [35]. In the research field of urban land use mapping, scholars have begun to utilize emerging data and technologies for urban land use mapping [36]. However, in this research field, the land use types of the mapping units are mainly determined based on the dominant land use types within the units, ignoring the microstructural information within the mapping units and the widespread phenomenon of mixed land use in a city [37]. In the research field of urban functional zone identification, although scholars have considered the widespread phenomenon of mixed land use in cities, there is still no consensus on the standard proportion of the single functions and mixed functions. It is common to use 50% as the proportion standard to distinguish the two [38], but the proportion standard still lacks basis and is not consistent with reality. In addition, existing studies have typically used methods such as the proportion [39], frequency density [40], or kernel density [41] of each Point of Interest (POI) type within the research unit for functional identification when constructing urban land use status data. However, POI data are point data obtained after abstracting the spatial range, and the spatial range of different POI types varies greatly, so utilizing this method for urban function identification affects the accuracy of the research results [42].

Based on the shortcomings of the existing studies mentioned above, the specific objectives of this paper were as follows: (1) determining the definition and proportion standard of mixed land use, and constructing the land use data of the study area based on this definition and proportion standard; (2) improving the method of constructing land use status data to improve the accuracy of the research results; (3) analyzing the current situation of mixed land use in Shenzhen and exploring the spatial distribution characteristics of the main mixed land use types and urban vibrancy; (4) examining the effects of mixed land use types on the intensity and stability of urban vitality. And, based on this, we aimed to put forward suggestions for the selection of land use types in mixed land use practices. Therefore, this study first determined the definition and proportion standards of mixed land use by referring to the existing literature, relevant policy documents, and the current mixed land use situation in Shenzhen. By utilizing multi-source data, including Area of Interest (AOI) data, POI data, Sentinel-2 remote sensing data, road data, building footprint data, and Baidu heat map data, and by improving the method based on the number of facilities in the method based on the area of the facilities, the land use status data within the built-up area of Shenzhen were constructed. The spatial distribution characteristics of the main land use types and the urban vibrancy in the study area were explored by using kernel density analysis. Finally, a multiple linear regression model was used to explore the effects of different mixed land use types on urban vibrancy intensity and stability by setting qualitative functional types as dummy variables. The analytical framework of this study is shown in Figure 1.

2. Theoretical Framework

Urban vibrancy, also known as urban vitality, was first proposed by Jane Jacobs in her book The Death and Life of Great American Cities. She argued that the interaction between human activities and urban space constitutes the diversity of urban life, which is the main manifestation of urban vibrancy [7]. Lynch [43] defined urban vibrancy as the ability of an urban system to maintain its internal survival, growth, and development. Maas [44] argued that urban vibrancy stems from a variety of unique commercial and recreational opportunities, as well as a dense, socially heterogeneous pedestrian populations, which can be broken down into the people who persist on the streets and in the public spaces, their activities and opportunities, and the environments in which these activities take place. Montgomery [45] argued that urban vibrancy can be defined by the presence of people and their activities in urban space, as well as the degree to which a place feels alive or vibrant. Urban vibrancy is a human-centered rather than a material-centered concept, and its essence is the interaction between people and their activities and urban space. Human activity is a key element and the most important manifestation of urban vibrancy, so indicators for measuring urban vibrancy should directly reflect the intensity of human activity [46]. A growing body of research is taking human activities as a proxy of urban vibrancy, for example, by using data reflecting human activities such as restaurant comments [47], mobile phone location request data [48], Baidu heat maps [49], Tencent location big data [50], Facebook check-in data [51,52], and Weibo check-in data [53] as the basic data for research. Among them, location-based spatiotemporal big data such as Baidu heat maps can be used to obtain real-time user location information and has the advantages of wide user coverage and detailed spatiotemporal resolution, so can objectively reflect human activities in urban space, and has become a key tool for urban vibrancy research. Urban vibrancy is mainly divided into two dimensions: spatial and temporal. The spatial dimension usually includes the distribution, scale, intensity, and diversity of spatial aggregation of people and their activities within a specific range of space. The temporal dimension includes the duration, the time points of peaks and valleys, the fluctuations, and the frequency of crowd activities. This study mainly focused on the intensity characteristics in the spatial dimension and the fluctuation characteristics in the temporal dimension, which are named urban vibrancy intensity and urban vibrancy stability, respectively.

Mixed land use can satisfy differentiated needs through different land functions and promote the agglomeration of differentiated people in the region [54]. It can improve urban vibrancy by diversifying urban activities, increasing the intensity of crowd gathering, and prolonging a high level of activities at night [55]. Urban residents usually travel to places with a certain function, so the spatial distribution of functions in a city can affect the spatial distribution of urban residents. In terms of urban vibrancy intensity, mixed land use, by co-locating multiple urban functions in a certain area, can effectively reduce the cross-regional travel of urban residents and promote their nearby activities, which can improve the concentration degree of people in the region, thereby increasing the urban vibrancy intensity. In terms of urban vibrancy stability, different urban functions in mixed land use usually attract people to gather during specific time periods. For example, workplaces usually attract people during the day, residential places usually attract people at night, and transportation places usually attract people during the commuting hours. Therefore, mixed land use can help reduce the variation in population density between different time periods throughout the day in a region, thereby improving urban vibrancy stability. Based on the positive effects of mixed land use on urban vibrancy, this paper explores the differences in urban vibrancy enhancement among different mixed land use types and puts forward corresponding suggestions on the selection of mixed land use types for different urban areas according to the research results.

3. Materials and Methods

3.1. Study Area

Shenzhen is located in the south of Guangdong province, China, and is a typical coastal city surrounded by low-lying mountains and hills. The total area of Shenzhen is 1997.47 km², with nine administrative districts and one functional district under its jurisdiction; the administrative districts include Futian, Luohu, Nanshan, Yantian, Baoan, Longgang, Longhua, Pingshan, and Guangming; and the functional district is Dapeng New District, as shown in Figure 2. At the end of 2021, the gross domestic product (GDP) was CNY 3066.49 billion, and the permanent population was 17.68 million [56]. Shenzhen is one of the cities with the greatest economic growth in China and is one of the first cities in China to promote the practice of mixed land use. Since 2010, a series of relevant policy documents have been issued to promote the implementation of mixed land use, such as the Shenzhen Urban Planning Standards and Guidelines, which provide detailed explanations of the definition, encouraged types, and encouraged spatial areas of mixed land use.

3.2. Data Sources and Processing

3.2.1. Data Sources

The information on the multi-source big data used in this study is shown in

Table 1. Information of various datasets.

Data Type	Source	Accessed Date	Received Date	Purpose
Road data	Amap (https://lbs.amap.com)	13 May 2021	13 May 2021	To obtain the boundaries of blocks
POI data	Amap (https://lbs.amap.com)	19 March 2021	25 March 2021	To determine the land use types of blocks
Baidu heatmap	Baidu Maps (https://lbsyun.baidu.com/)	19 May 2021	19 May 2021	To measure the urban vibrancy of blocks
AOI data	Baidu Maps (https://lbsyun.baidu.com/)	14 September 2021	17 September 2021	To determine the land use types of blocks
Sentinel-2 remote sensing data	GEE platform (https://developers.google.cn/earth-engine)	5 September 2021	5 September 2021	To extract impervious layers in the study area
Building footprint data	Tianditu (https://www.tianditu.gov.cn)	14 November 2021	23 November 2021	To determine the building density and plot ratio of blocks

. Baidu heat maps are dynamic, continuous, and easy to identify. Based on the spatial positioning information of hundreds of millions of user groups, they can estimate the relative distribution of population density in a region. Therefore, this study quantified the urban vibrancy based on these data. Compared with POI data, AOI data contain not only the name, address, and type information of various facilities but also the spatial range information of various facilities.

The main gap among these six types of data lies in their different geographical coordinate systems. The geographical coordinate system of Sentinel-2 remote sensing data and building footprint data is WGS84, the geographical coordinate system of road data and POI data is GCJ02, while the geographical coordinate system of AOI data and Baidu heat map data is BD09, making it difficult to overlay and analyze these data under the same coordinate system. To solve this problem, the geographical coordinate systems of these data were unified under the WGS84 coordinate system in this study and then converted to the projection coordinate system.

3.2.2. Construction of Land Use Data

Due to the slow update speed and difficulty of obtaining land use data from government departments, this study used multi-source big data, which are more quickly updated and are more available, to construct land use status data of the study area. Since there were overlaps in the dynamics and state of land use types, the changes in land use types between different years were relatively small. Therefore, we only mapped the land use status of the study area in 2021 and conducted subsequent analysis and research based on this. The framework for constructing land use data is shown in Figure 3.

(1)

Definition and proportion standard of mixed land use

This study defined mixed land use as two or more urban land use types within the same block unit. In this study, six major urban land use types were considered, and information on their names, codes, and scope are shown in

Table 2. Abbreviation, code, and scope of the six major land use types.

Land Use Type	Abbreviation and Code	Scope of Land Use Type
Industrial land	Industrial (I)	Land that is dominated by activities such as production, manufacturing, and finishing of products, and is supported by activities such as research and development, design, testing, and management.
Administration and public services land	Administration (A)	Land for administration, culture, education, scientific research, sports, medical and healthcare, social welfare, public security, religion, and land of a special nature.
Transportation land	Transportation (T)	Land for regional transportation, urban roads, rail transit, transportation facilities, etc.
Residential land	Residential (R)	Land for residential buildings and corresponding supporting service facilities.
Green space and public square land	Green (G)	Land for public open spaces, such as parks, green spaces, and squares.
Commercial and service land	Commercial (C)	Land used for various commercial sales, service activities, and accommodating various activities, such as offices, hotels, and amusement activities.

. According to the planning documents of Shenzhen and Shanghai [57,58], this study took 10% as the threshold to distinguish the ancillary functions and main functions within a block and 70% as the threshold to distinguish the dominant functions and main functions within a block. The functional importance corresponding to the different area proportions within a block unit is shown in

Table 3. The functional importance corresponding to the different area proportions within a block unit.

Area Proportion	Functional Importance
≥70%	dominant function
10~70%	main function
<10%	auxiliary function

. The functional types and land use types corresponding to the different area share conditions within a block unit are shown in

Table 4. The functional types and land use types corresponding to the different area share conditions within a block unit.

Area Proportion	Functional Type	Land Use Type
one of Px ≥ 70%	Single function	X (Px ≥ 70%)
Px are all < 70%	Mixed functions	X1 + X2 +… $(10\%<P_{x_1}<70\%, 10\%<P_{x_2}$ < 70%, …)

Note: P_x is the area proportion of a certain land use type x in the block. This paper refers to the expression of mixed land use in the Shenzhen Urban Planning Standards and Guidelines [57] and uses “+” to connect the codes of different land use types in mixed land use. Due to the complexity of mixed land use, this paper does not rank the land use type according to their area proportion when expressing the mixed land use types.

. Mixed land use is mainly divided into two dimensions: the horizontal dimension and the vertical dimension, and this study mainly considered mixed land use in the horizontal dimension above ground.

(2)

Urban block extraction

Since this study focused on urban mixed land use, the blocks within the built-up area of the city were taken as the basic research unit. The process of urban block extraction was mainly conducted to extract the spatial units surrounded by roads (SUSRs) within the city’s built up area, and the extracted SUSRs were used as urban blocks. In this study, the spatial range of the impervious layer in the study area was considered as the spatial range of the built-up area. When the spatial range of the extracted SUSRs was closest to the spatial range of the impervious layer, the spatial range of the SUSRs was the spatial range of the urban blocks. Therefore, it was necessary to calculate the area proportion of impervious layer in each SUSR and compare the overall spatial range of the SUSRs with different impervious layer area proportion thresholds with the spatial range of the impervious layer. When the two were closest, the overall spatial range of the SUSRs at that impervious layer area proportion threshold was the spatial range of the urban blocks.

Step 1: Extraction of SUSRs. We selected the road types that can reflect the basic structure and morphology of the city in the road data and then performed topological processing on the road data to make it closed and without redundancy. Since the widths of roads of different grades are different, it was necessary to estimate the actual width of the roads of different grades by using the distance-measuring tools on electronic maps (https://amap.com/, accessed on 23 May 2021) and to construct the buffer zones of the corresponding widths. The SUSRs could be obtained by erasing the surface data of the study area using the road buffer zones and then splitting the erased data. The SUSRs in the study area are shown in Figure 4.

Step 2: Extraction of the impervious layer within the study area. Supervised classification was performed on Sentinel-2 remote sensing data to extract impervious layers in Shenzhen.

Step 3: Extraction of urban blocks. The overall spatial range of the SUSRs under different impervious layer area proportion thresholds were compared with the spatial range of the impervious layer in the study area (Figure 5). It was found that the spatial range of the two was closest when the impervious layer area proportion within the SUSR was greater than 50%. The spatial range of the urban blocks extracted by using this threshold is shown in Figure 6.

(3)

Block land use type identification

AOI data and POI data contain the type information of facilities, which can be used to identify the land use types of blocks. POI data are the point data obtained after abstracting the spatial range of facilities. The land use type identification results obtained by methods based on the number of different types of POIs in the research unit, such as the proportion, kernel density, and frequency density, contain significant errors. For example, a block is a residential community containing multiple living facilities. In the POI data, a residential community with a large spatial range is represented as only one point, while living facilities with a small spatial range are represented as multiple points. If calculated based on the number of facilities, the land use type of the block would be commercial and service land, but it is actually residential land. Therefore, this study developed a method for identifying block land use types based on facility footprints. The specific identification method was as follows:

Step 1: The AOI data and POI data were reclassified according to the range of the six major urban land use types shown in Table 1.

Step 2: The area of the different AOI types in each block unit was calculated by overlay analysis of the AOI data containing spatial range information of the facilities and block units.

Step 3: The POI data were then used to perform functional identification on the remaining area identified using the AOI data. Since the POI data did not contain the spatial range information of the facilities, this study estimated the area proportion of the different POI types in each block by setting the area coefficients for the different POI types. The method for determining the area coefficient of each POI type was as follows: Firstly, POI types with similar footprints were grouped into seven categories. Secondly, samples of POIs with different footprint types were selected, the footprint of each sample was estimated by using the distance measurement tools on the electronic map (https://amap.com/, accessed on 10 October 2021), and the average area of each POI footprint type was calculated. Finally, the area coefficient of the POI type with the largest footprint was set to 100, and the area coefficient of the other POI types was determined according to the area ratio of the POI type with the largest footprint. The POI area coefficients in this study had a total of seven values, including 0.1, 0.5, 2, 10, 20, 80, and 100.

Step 4: The identification results obtained using the AOI data and POI data were summarized to calculate the area proportion of the different land use types in each block unit, and the land use type of each block was determined based on this result. Due to the lack of AOI and POI data in some blocks, it was impossible to identify their land use types, and the land use type of such blocks was defined as “null” in this study. The land use status data of Shenzhen constructed in this study are shown in Figure 7.

(4)

Accuracy verification

In this study, a sample of 200 blocks was randomly selected to compare the constructed land use data with the real land use conditions on an electronic map (https://www.amap.com/, accessed on 19 October 2021). The accuracy of the land use status data constructed in this study was tested using the method of conformity verification, and the conformity degree was divided into four levels: completely inconsistent, relatively inconsistent, relatively consistent, and completely consistent, with corresponding conformity scores of 0, 1, 2, and 3 for each level. The formula for calculating the overall conformity degree of the samples is as follows:

$$c={\displaystyle {\sum }_{i=1}^n{x}_i}/{\displaystyle {\sum }_{i=1}^n{X}_i}\times 100\%$$

(1)

In the formula, c is the overall conformity degree, i is a certain sample, n is the number of samples, x_i is the actual score of the conformity degree of sample i, and X_i is the highest score of the conformity degree of sample i. By calculating the overall conformity degree of the sample, the result obtained was 86.67%, which indicated that the land use status data constructed in this study had a high conformity degree with the real land use situation, and the data could be used for further analysis and research.

3.2.3. Land Use Type Selection

There were 48 land use types in the study area, and the bar chart and cumulative curve for the proportion of each land use type are shown in Figure 8. It can be seen from the figure that the first 12 land use types of blocks accounted for 85.04% of the total number of blocks, making them the main land use types in Shenzhen. The rest of the land use types accounted for less than 15% of the total number of blocks. Therefore, this study conducted further analysis on the 12 main land use types in the study area only, specially including I, A, R, C, I+R, I+C, A+R, R+C, I+A+R, I+A+C, I+R+C, and A+R+C.

3.3. Methods

3.3.1. Quantitative Measurement of Urban Vibrancy

As with mixed land use, this paper mainly considers the urban vibrancy of the horizontal dimension of the aboveground portion. This study utilized Baidu heat map data, which can reflect the relative spatial distribution of urban population density in real time, to reflect the intensity characteristics of population activities (urban vibrancy intensity) and the fluctuation characteristics of population activity intensity over time (urban vibrancy stability) by calculating the average and standard deviation of heat values.

(1)

Urban vibrancy intensity

The urban vibrancy intensity in this study was based on the Baidu heat map obtained at 12 time points (1:00, 3:00, …, 23:00) throughout the day, calculated by averaging the heat values of the 12 time points of each block unit.

$$\overline{H_i}={\displaystyle {\sum }_{i=1}^n{H}_{ij}/n}$$

(2)

In Equation (2), $\overline{H_i}$ is the average of heat value; n is the number of time points to obtain the Baidu heat map in a day. In this study, the value of n was 12. i is a certain time point in a day; j is a certain block unit in the study area; and H_ij is the heat value of unit j at time i. The higher the average heat value, the higher the concentration of people in the area and the stronger the urban vibrancy intensity.

(2)

Urban vibrancy stability

The urban vibrancy stability in this study was also based on Baidu heat maps, obtained by calculating the standard deviation of the heat values of each block unit at 12 time points (1:00, 3:00, …, 23:00) throughout the day.

$$H_i^{\prime }=\sqrt{{\displaystyle {\sum }_{i=1}^n{(H_{ij}-\overline{H_i})}^2}/n}$$

(3)

In Equation (3), H_i’ is the standard deviation of the heat value; n is the number of hours calculated in a day, where the value of n is 12; H_ij is the heat value of unit j at time i; and $\overline{H_i}$ is the average of heat value throughout the day. The lower the standard deviation of the heat value, the lower the dispersion degree of the heat value at different times throughout the day and the higher the urban vibrancy stability.

3.3.2. Kernel Density Analysis

Kernel density analysis is an important tool for the study of spatial disequilibrium. The calculation formula is as follows [59]:

$$f(x)={\displaystyle {\sum }_{i=1}^{n}K((x-x_i)/h)}/nh$$

(4)

In Equation (4), f(x) is the kernel density estimation function at position x; n is the number of POIs whose path distance from position x is less than or equal to h; h is the path distance attenuation threshold, that is, the bandwidth; K(x) is the spatial weight kernel function, and the Gaussian kernel function was adopted in this study; and x − x_i represents the distance between point x and point x_i.

3.3.3. Regression Analysis

Due to the numerous factors influencing urban vibrancy, in addition to mixed land use, there are many other built environment factors that can affect urban vibrancy. Therefore, it was necessary to control the other possible influencing factors in addition to mixed land use types to accurately explore the differences in the impact of different mixed land use types on urban vibrancy.

(1)

Selection and quantification of variables

Existing studies have shown that in terms of land use, mixed land use and moderate land use intensity can increase urban vibrancy [60,61,62]; in terms of location, proximity to urban centers or regional centers usually enhances urban vibrancy [63]; in terms of facilities, living convenience such as facility density is a key factor promoting urban vibrancy [64], and traffic convenience such as road density and proximity to transportation stations is also a main factor influencing urban vibrancy [65,66]; in terms of environment, vegetation coverage can also have a significant impact on urban vibrancy [67]. Therefore, in this paper, nine independent variables, including land use type, building density, plot ratio, distance from city center, distance from subway station, distance from bus station, vegetation coverage, facility density, and road density, were selected from four aspects, namely, mixed land use characteristics, land use intensity characteristics, location characteristics, and environmental and facility characteristics, to investigate the influence of mixed land use type and other built environment factors on the intensity and stability of urban vibrancy. In addition, the dependent variables of the models were urban vibrancy intensity and urban vibrancy stability. The quantification methods and descriptive statistical results of each variable are listed in

Table 5. Variables and descriptive statistics.

Variable Name	Description of Variable (Units)	Min.	Mean	Max.
Urban vibrancy intensity	Average of heat value within the block (–)	0.00	2.82	6.05
Urban vibrancy stability	Standard deviation of heat value within the block (–)	0.00	1.49	2.83
Building density	Ratio of building footprint within a block to the area of the block (%)	0.00	0.63	10.21
Plot ratio	Ratio of gross floor area within a block to the area of the block (–)	0.00	13.56	331.57
Facility density	Number of POI points per unit area within a block (pcs/m2)	0.00	0.00	0.04
Road density	Length of roads per unit area within a block (m/m2)	0.00	0.03	0.14
Vegetation coverage	Vegetation coverage calculated by using the dimidiate pixel model (%)	0.00	0.32	0.84
City center	Shortest straight-line distance between the centroid of the block and the city center (m)	174.03	19,569.50	44,864.79
Subway station	Straight-line distance between the centroid of the block and the nearest subway station (m)	31.37	1848.85	25,204.02
Bus station	Straight-line distance between the centroid of the block and the nearest bus stop (m)	7.10	194.27	863.90
Mixed land use type	Quantify by setting dummy variables with values of 0 or 1	0.00	—	1.00

.

(2)

Multiple linear regression model

This study used a multiple linear regression model to conduct preliminary research on the effect of mixed land use type on urban vibrancy. As the mixed land use type was a qualitative factor, it needed to be quantified by setting dummy variables. The expression of the model is as follows:

$$y={\beta }_0+{\beta }_1{x}_1+\cdots +{\beta }_m{x}_m+{\delta }_1{d}_1+\cdots +{\delta }_n{d}_n+\epsilon$$

(5)

In Equation (4), y is urban vibrancy intensity or stability; β₀ is the intercept term; β₁, β₂, …, β_m are the regression coefficients of each continuous variable; x is a continuous variable; m is the number of continuous variables (the value of m used in this study was 8); δ₁, δ₂, …δ_n are the regression coefficients of each dummy variable; d is a dummy variable; n is the number of dummy variables; and ε is the random error term. Since a total of 12 mixed land use types were considered in this study, the value of n was 12. When a mixed land use type was a reference group, its d value was 0, and when a mixed land use type was a comparison type, its d value was 1. The sample value of the dummy variables was either 0 or 1. If the sample was of a certain mixed land use type, its value was 1; otherwise, it was 0. For δ > 0, the model results could be interpreted as follows: when other influencing factors remained unchanged, the urban vibrancy intensity of the comparison mixed land use type could increase by δ units or the urban vibrancy stability of the comparison mixed land use type could decrease by δ units compared with the reference mixed land use type. In this study, 12 main land use types in the study area were analyzed, and a total of 12 models were constructed with each land use type as a base group type. Since the urban vibrancy intensity or urban vibrancy stability was the dependent variable of the model, a total of 24 models were constructed in this study.

4. Results

4.1. Spatial Distribution Characteristics of Mixed Land Use and Urban Vibrancy

4.1.1. Overall Characteristics of Mixed Land Use

A total of 5048 urban blocks were extracted in this study. The number of single-function blocks, mixed-function blocks, and no-data blocks accounted for 59.39%, 36.63%, and 3.98%, respectively, while their areas accounted for 45.48%, 53.77%, and 0.75%, respectively. In terms of quantity, the number of single-function blocks was greater than the number of mixed-function blocks, but in terms of area, the area of mixed-function blocks was greater than the area of single-function blocks, indicating that mixed-function blocks were mostly distributed in larger blocks. Overall, the mixed land use degree of Shenzhen was still very limited and needed to be improved. From the perspective of spatial distribution, single-function and mixed-function blocks in Shenzhen were inlaid and interleaved (Figure 9). Of these, the spatial distribution of single-function blocks was more concentrated, mainly distributed in the urban center and the northern industrial areas, while the spatial distribution of mixed-function blocks was relatively dispersed.

4.1.2. Spatial Distribution Characteristics of Main Mixed Land Use Types

As can be seen from Figure 10a, the blocks of land use type I of Shenzhen were distributed mainly in areas outside the city center and were concentrated in the northern industrial area of the city. The blocks of land use type A were concentrated mainly in Futian District (Figure 10b). This was because Futian District was the administrative center of Shenzhen City, and there were many public service facilities, such as primary and secondary schools, hospitals, and libraries, in the area. The blocks of land use type R were concentrated in Futian District, in the west of Luohu District, and in the south of Nanshan District in the city center, as well as in the regional centers of Baoan, Longhua, and Longgang Districts outside the city center (Figure 10c). Shenzhen is a mega city with a huge demand for housing, resulting in a lot of single-function residential areas in the city. The blocks of land use type C were clustered in Futian District (Figure 10d), because Futian District is located in the core of the central business district, and there are more commercial and service facilities gathered here.

The blocks of land use type I+R were clustered in the northern urban peripheral areas (Figure 10e), because a large number of urban villages are distributed in this area. Urban villages are typical informal settlements in Shenzhen [68]; they contain not only residential land but also village-owned industrial zones, thus forming a mixture of industrial and residential land. The blocks of land use type I+C were clustered in the eastern part of Futian District and the central part of Nanshan District (Figure 10f). The formation of this cluster was mainly due to the fact that there are many science and technology parks and business service facilities in this region. The blocks of land use type A+R were concentrated mainly in Futian District (Figure 10g), because Futian District has both rich educational resources and a greater number of residential areas. The blocks of land use type R+C were clustered in Futian District and in the west of Luohu District (Figure 10h), because these areas are not only the central business district of Shenzhen but also have a large number of residential districts and a small number of urban villages.

The blocks of land use type I+A+R were clustered in the north of Baoan District and the north of Pingshan District (Figure 10i), because not only are there urban villages that include residential land and industrial land but also administration and public service facilities, such as schools, hospitals, and sports centers, in this region. The blocks of land use type I+A+C were clustered in the northwest of Longgang District, the east of Futian District, and the west of Luohu District (Figure 10j), because there are many science and technology parks, commercial and service facilities, and educational resources in this area. The blocks of land use type I+R+C were clustered in the east of Futian District and the west of Luohu District (Figure 10k), because industrial areas located there and many commercial and service facilities and residential areas. The blocks of land use type A+R+C were concentrated in Luohu District, Futian District, and Baoan District (Figure 10l), because there are large numbers of financial institutions, residential communities, educational resources, administrative institutions, and other facilities in these areas.

4.1.3. Spatial Distribution Characteristics of Urban Vibrancy

The spatial distribution of the urban vibrancy intensity and stability in Shenzhen are shown in Figure 11. As can be seen from the figure, the blocks with higher average of heat value and standard deviation of heat value were distributed in Futian District, Luohu District, and the southern part of Nanshan District in the city center, as well as the Qianhai Bay area of Baoan District and the south of Longhua District outside the city center, while the blocks with lower average of heat value and standard deviation of heat value were mainly distributed in the urban periphery. The main reason for this phenomenon is the mismatch between the living spaces and the working spaces in the city. As the city center is an area with a concentrated distribution of places to work, study, and reside, the city center attracts lots of people to gather there, resulting in a high average heat value and a high urban vibrancy intensity for the blocks in that area. However, at night, people go to their living places, resulting in a significant decrease in the concentration of people in the area, and thus the urban vibrancy stability of the urban center is low. The peripheral areas of the city were mainly industrial areas, which were characterized by the waste of land and resources, environmental degradation, and poor-quality urban infrastructure [69], resulting in a low concentration of people and their activities at all times throughout the day. Therefore, the urban vibrancy intensity of these areas was low, while the urban vibrancy stability was high.

4.2. Impact of Mixed Land Use Types on Urban Vibrancy

4.2.1. Impact of Mixed Land Use Types on Urban Vibrancy Intensity

The model results of the impact of mixed land use types on urban vibrancy intensity are shown in

Table A1. The values of ${\beta }_1–{\beta }_8$ and adjusted R² in Models 1–12.

β1–β8	M1–M12
Building density	−0.018
Plot ratio	0.016
Facility density	0.291 ***
Road density	0.187 ***
Vegetation coverage	−0.150 ***
City center	−0.204 ***
Subway station	−0.154 ***
Bus station	−0.111 ***
Adjusted R2	0.551

Note: *** significant at 0.1% level.

and

Table A2. The values of ${\delta }_1–{\delta }_{12}$ in Models 1–12.

δ1–δ12	M1	M2	M3	M4	M5	M6	M7	M8	M9	M10	M11	M12
I	—	−0.105 ***	−0.252 ***	−0.090 ***	−0.106 ***	−0.072 **	−0.209 ***	−0.291 ***	−0.099 **	−0.092 **	−0.253 ***	−0.299 ***
A	0.053 ***	—	−0.074 ***	0.008	0.000	0.017	−0.052 **	−0.094 ***	0.003	0.007	−0.075 ***	−0.098 ***
R	0.260 ***	0.151 ***	—	0.167 ***	0.151 ***	0.186 ***	0.045	−0.041	0.158 ***	0.166 ***	−0.002	−0.048
C	0.059 ***	−0.010	−0.105 ***	—	−0.010	0.012	−0.077 ***	−0.131 ***	−0.006	−0.001	−0.106 ***	−0.136 ***
I+R	0.062 ***	0.000	−0.085 ***	0.009	—	0.020	−0.059 **	−0.107 ***	0.004	0.008	−0.085 ***	−0.112 ***
I+C	0.031 **	−0.014	−0.077 ***	−0.008	−0.015	—	−0.059 ***	−0.094 ***	−0.012	−0.009	−0.078 ***	−0.097 ***
A+R	0.079 ***	0.039 **	−0.016	0.045 ***	0.039 **	0.052 ***	—	−0.031 *	0.041 **	0.044 **	−0.017	−0.034 *
R+C	0.110 ***	0.070 ***	0.015	0.076 ***	0.070 ***	0.083 ***	0.031 *	—	0.072 ***	0.075 ***	0.014	−0.003
I+A+R	0.037 **	−0.002	−0.056 ***	0.003	−0.003	0.010	−0.040 **	−0.071 ***	—	0.003	−0.057 ***	−0.074 ***
I+A+C	0.033 **	−0.005	−0.058 ***	0.001	−0.005	0.007	−0.042 **	−0.072 ***	−0.003	—	−0.058 ***	−0.075 ***
I+R+C	0.169 ***	0.098 ***	0.001	0.108 ***	0.098 ***	0.121 ***	0.030	−0.025	0.102 ***	0.107 ***	—	−0.030
A+R+C	0.107 ***	0.069 ***	0.017	0.074 ***	0.069 ***	0.081 ***	0.032 *	0.003	0.071 ***	0.074 ***	0.016	—

Note: * significant at 5% level, ** significant at 1% level, *** significant at 0.1% level.

. The variance inflation factors of each independent variable in Models 1–12 were all less than 10, indicating that there was no multicollinearity among the variables in the models. The adjusted R² value of the model was 0.551, which could explain 55.1% of the variation in the urban vibrancy intensity. Since the model was significant overall, the results could be utilized for further analysis. In order to comprehensively reflect the urban vibrancy intensity relationship among the different mixed land use types, this study visualized the standardized regression coefficients of each dummy variable in Models 1–12 (M 1–12) by using a matrix heatmap (Figure 12).

As can be seen from Figure 10, the urban vibrancy intensity of land use type I was significantly lower than that of the other 11 land use types, indicating that the urban vibrancy intensity of land use type I was the lowest. This is due to the fact that the industrial land in Shenzhen is mainly distributed in the peripheral areas of the city, so spatial agglomerations of people do not easily form. The urban vibrancy intensities of land use types A, C, I+R, I+C, I+A+R, and I+A+C were significantly higher than that of land use type I and significantly lower than those of land use types R, A+R, R+C, I+R+C, and A+R+C, indicating that the urban vibrancy intensity of these six land use types was at a medium–low level. All of the mixed functions in this group contained industrial land, indicating that mixing with industrial land may result in lower urban vibrancy intensity. The urban vibrancy intensity of land use types R, A+R and I+R+C was significantly higher than that of land use types I, A, C, I+R, I+C, I+A+R, and I+A+C, indicating that the urban vibrancy intensity of these land use types was at a medium–high level. All of these land use types included residential land, and the residential land had a large number of people gathering at night. The urban vibrancy intensity of land use types R+C and A+R+C was the highest, which was significantly higher than that of land use types I, A, R, C, I+R, I+C, A+R, I+A+R, I+A+C, and I+R+C, indicating that the urban vibrancy intensity of these two land use types was the highest. Both land use type R+C and A+R+C contained residential and commercial and service land, which usually attract a large number of people. Therefore, the mixing of the two land use types resulted in a higher urban vibrancy intensity.

In summary, mixing with industrial land usually led to a decrease in urban vibrancy intensity, while mixing with residential land or commercial and service land usually led to an increase in urban vibrancy intensity. Notably, the simultaneous mixing of residential land and commercial and service land led to a significant increase in urban vibrancy intensity.

4.2.2. Impact of Mixed Land Use Types on Urban Vibrancy Stability

The model results of the impact of mixed land use types on urban vibrancy stability are shown in

Table A3. The values of ${\beta }_1–{\beta }_8$ and adjusted R² in Models 13–24.

β1–β8	M13–M24
Building density	−0.062 ***
Plot ratio	0.093 ***
Facility density	0.112 ***
Road density	0.151 ***
Vegetation coverage	−0.049 ***
City center	−0.197 ***
Subway station	−0.222 ***
Bus station	−0.201 ***
Adjusted R2	0.410

Note: *** significant at 0.1% level.

and

Table A4. The values of ${\delta }_1–{\delta }_{12}$ in Models 13–24.

δ1–δ12	M13	M14	M15	M16	M17	M18	M19	M20	M21	M22	M23	M24
I	—	−0.105 ***	−0.097 ***	−0.165 ***	−0.041	−0.197 ***	−0.124 ***	−0.120 **	−0.028	−0.154 ***	−0.160 ***	−0.192 ***
A	0.053 ***	—	0.004	−0.030 *	0.032 *	−0.047 *	−0.010	−0.008	0.039	−0.025	−0.028	−0.044 *
R	0.101 ***	−0.008	—	−0.070 **	0.059 *	−0.103 **	−0.027	−0.024	0.071 *	−0.059	−0.064 **	−0.098 **
C	0.107 ***	0.039 *	0.044 ***	—	0.081 ***	−0.021	0.027	0.029	0.089 ***	0.007	0.003	−0.018
I+R	0.024	−0.037 *	−0.033 *	−0.072 ***	—	−0.091	−0.048 *	−0.046 *	0.007	−0.066 **	−0.069 ***	−0.088 ***
I+C	0.085 ***	0.040 *	0.043 **	0.014	0.067 ***	—	0.031	0.033	0.072 ***	0.018	0.016	0.002
A+R	0.047 ***	0.007	0.010	−0.016	0.031 *	−0.028	—	0.001	0.036 *	−0.012	−0.014	−0.026
R+C	0.045 **	0.006	0.009	−0.017	0.030 *	−0.029	−0.001	—	0.035 *	−0.013	−0.015	−0.027
I+A+R	0.010	−0.028	−0.025 *	−0.050 ***	−0.005	−0.062 ***	−0.035 *	−0.034 *	—	−0.046 **	−0.048 ***	−0.060 **
I+A+C	0.056 ***	0.018	0.020	−0.004	0.041 **	−0.015	0.011	0.012	0.045 **	—	−0.002	−0.014
I+R+C	0.106 ***	0.036	0.041 **	−0.003	0.079 ***	−0.025	0.024	0.026	0.087 ***	0.004	—	−0.022
A+R+C	0.069 ***	0.031 *	0.034 **	0.010	0.054 ***	−0.002	0.024	0.026	0.058 **	0.013	0.012	—

Note: * significant at 5% level, ** significant at 1% level, *** significant at 0.1% level.

. The variance inflation factors of each independent variable in Models 12–24 were all less than 10, indicating that there was no multicollinearity among the variables. As the adjusted R² value of the model was 0.410, and the model was significant overall, so the model results could be used for further analysis. The standardized coefficients of each dummy variable in Models 13–24 (M 13–24) are visualized by using a matrix heatmap in Figure 13.

As can be seen from Figure 11, the urban vibrancy stability of land use type I was significantly higher than that of the other 11 land use types, indicating that land use type I had the highest urban vibrancy stability. This was due to the fact that the urban vibrancy intensity of industrial land was lower at all times throughout the day. The urban vibrancy stability of land use types I+R and I+A+R was significantly higher than that of land use types A, R, C, I+C, A+R, R+C, I+A+C, I+R+C, and A+R+C, indicating that the urban vibrancy stability of land use types I+R and I+A+R was at medium–high level. Both land use types I+R and I+A+R contained industrial land, and mixing with industrial land improved the urban vibrancy stability. The urban vibrancy stability of land use types A, R, A+R, R+C, I+A+C, and I+R+C was significantly lower than that of land use types I, I+R, and I+A+R, indicating that the urban vibrancy stability of these six land use types was at medium–low level. These land use types contained either administration and public service land or residential land, which led to a large number of people gathering during the day or night, respectively, with a small number of people gathering during the rest of the day. The urban vibrancy stability of land use types C, I+C, and A+R+C was the lowest, which was significantly lower than that of land use types I, A, R, I+R, and I+A+R, indicating that the urban vibrancy stability of these three land use types was the lowest. This is because land use types C, I+C and A+R+C all included commercial and service land, and the spatial service range of commercial and service land was relatively large, which resulted in higher urban vibrancy intensity during the day and greater variation in urban vibrancy intensity at different times throughout the day.

In summary, mixing with industrial land usually increased urban vibrancy stability, while mixing with commercial and service land usually decreased urban vibrancy stability. However, when mixing with administration and public service land or residential land, the change in urban vibrancy stability was not obvious.

5. Discussion

5.1. Validation of Results

By comparing the results of this study with those of existing research, the accuracy of the research findings in this study could be indirectly verified. In terms of the spatial distribution characteristics of mixed land use in the study area, Zhao et al. [70] found that the mixed land use degree in the urban periphery of Shenzhen was relatively low. In terms of the spatial distribution characteristics of the urban vibrancy intensity in the study area, Li et al. [71] found that the urban vibrancy intensity was higher in the central areas of Shenzhen, such as Futian, Luohu, and Nanshan. In terms of the spatial distribution characteristics of the urban vibrancy stability in the study area, Li and Zhao [72] found that the urban vibrancy of urban center areas changed more dramatically at different times throughout the day. In terms of the impact of mixed land use on urban vibrancy, most of the existing studies have shown that mixed land use had a positive effect on the enhancement in urban vibrancy [50,73]. Existing studies have usually quantified mixed land use in terms of diversity to explore its impacts, and policy recommendations based on such research results have typically been to increase or decrease the diversity of mixed land use. However, such research findings and policy recommendations have limited reference value for mixed land use practices. A study on which mixed land use types yield higher benefits would provide a more valuable reference for future mixed land use practices [74]. Therefore, this study investigated the differences in the urban vibrancy enhancement of different mixed land use types. Doan et al. [75] found that the area of commercial and retail uses within a research unit positively affected urban vibrancy. Chen and Huang [48] found that the area share of commercial land in the research unit had a higher impact on urban vibrancy than that of residential, industrial, transportation, and public service land. In addition, Rehman and Asghar [76] found that the mixing residential land and commercial land improved urban vibrancy. In summary, the results of this study are basically consistent with those of existing studies.

5.2. Policy Implications

In terms of national policies, China’s current national-level policy documents containing relevant policies to promote the implementation of mixed land use are mainly focused on promoting mixed land use in areas such as transportation stations, important waterfront areas, and industrial parks. However, at the international level, some countries have issued relevant policy documents to promote mixed land use in urban centers in order to facilitate the transformation and upgrading of traditional central business districts (CBDs). The traditional CBD concentrates workspaces for business, finance, and administration in the city, and this study confirms that although the urban vibrancy intensity of the traditional CBD is high, its urban vibrancy stability is low. Based on the results of this study, in the future, when formulating national-level policy documents related to the promotion of the implementation of mixed land use, it is necessary to encourage mixed land use in urban centers to alleviate the phenomenon of job–housing mismatch, improve the livability of cities, and promote the sustainable development of cities.

In terms of the mixed land use policies in Shenzhen, the mixed land use degree in Shenzhen was still relatively low, so it is necessary to further promote mixed-function development in the city. In particular, “top-down” government-led and “bottom-up” collective-led mixed land use development should be combined to effectively balance economic and public interests, as well as promote the maximization of mixed land use benefits [77].

In terms of urban functional layout, for the urban center area, due to its high level of urban vibrancy intensity and low level of urban vibrancy stability, the primary goal of this region is to improve stability. According to the results in this paper, the urban center area of Shenzhen is still dominated by single-function blocks, which is also an important reason for the low level of urban vibrancy stability in the area and the existence of the phenomenon of “jobs–housing mismatch” in the city. In the future, the urban center of Shenzhen should be transformed from the single-function land use pattern of traditional CBDs to a mixed-function land use pattern in order to solve the problem of “pendulum commuting” on weekdays and “hollowing out” of the city center at night and weekends and promote the “jobs–housing balance” in Shenzhen. The residential space in the urban central area can be moderately increased to balance the urban vibrancy intensity and stability, and the specific mixed land use types that can be laid out include A+R, R+C, and A+R+C.

For the peripheral areas of the city, which had lower urban vibrancy intensity and higher urban vibrancy stability, the main focus should be on increasing urban vibrancy intensity. Shenzhen is a city of satellites, comprising many subcenters [78]. For the subcenter areas in the peripheral areas of the city, the urban vibrancy intensity can be promoted by adding workspaces for public or commercial services or living spaces, and the specific mixed land use types that can be selected include A+R, R+C, and A+R+C. For industrial parks in urban peripheral areas, functions such as administration and public services, as well as residential and commercial services can be laid out to promote city–industry integration and the transformation of traditional industrial land into new industrial land, ultimately realizing the enhancement in the urban vibrancy intensity of the area. For specific land use types, the mixed functions that can be laid out include land use types I+R, I+C, I+R+C, I+A+R, and I+A+C.

In Shenzhen, a two-tier regulatory regime is emerging, with planned unit development in commercial and residential zones, while following a bottom-up incremental path in industrial zones. Therefore, in mixed land use practices, in addition to selecting specific land use types based on the recommendations in this paper, it is also necessary to comprehensively consider the local bottom-up real estate dynamics and top-down planning process.

5.3. Contributions and Deficiencies

Compared with existing research, the contributions and innovations of this study are as follows: (1) By referring to the existing literature and relevant policy documents, and combining the results with the current situation of mixed land use in Shenzhen, the definition and proportion standard of mixed land use were determined, which can provide reference for future research on urban land use mapping and mixed land use. (2) The accuracy of the block land use type identification results was improved by shifting the perspective from being based on the number of facilities to being based on the area of facilities. (3) The analysis of the spatial distribution characteristics of urban vibrancy and mixed land use in Shenzhen helped to understand the current status of both in the study area. (4) The differences in urban vibrancy enhancement among the different mixed land use types were explored, filling the gap in the research exploring the impact of qualitative mixed land use types on urban vibrancy, and the research results can provide a reference for the selection of land use types for future mixed land use practices aiming at improving urban vibrancy.

Although this study innovatively explored the differences in urban vibrancy enhancement among different mixed land use types, there are still some shortcomings. Due to the large number of mixed land use types in the study area, totaling 48, and the significant difference in sample size between different land use types, this study only studied the 12 land use types with a large sample size and did not cover the land use types with a small sample size. In future research, in addition to the 12 land use types studied in this paper, other land use types can be further studied to enrich the theoretical basis of mixed land use practices.

6. Conclusions

In this study, multi-source data were used to construct the land use status data of Shenzhen. Based on the proportion of each land use type in the study area, 12 main land use types in Shenzhen were selected, and the spatial distribution characteristics of each land use type were explored by using kernel density analysis. In addition, the spatial distribution characteristics of urban vibrancy intensity and stability were explored in this study. Finally, a multiple linear regression model was utilized to explore the effects of mixed land use type on urban vibrancy intensity and stability by setting land use type as a dummy variable. The conclusions of this study are as follows:

(1)

The consistency between the land use status data of Shenzhen constructed using the method proposed in this study and the real land use situation was as high as 86.67%, so could be used for further analysis.

(2)

The mixed land use types containing industrial land were mainly concentrated in the northern industrial area of Shenzhen, while the mixed land use types containing residential, commercial and services, or administrative and public services land were mainly concentrated in the city center. The spatial distribution trends in urban vibrancy intensity and stability were inconsistent in most areas of Shenzhen, indicating the problem of “jobs–housing mismatch” and low urban vibrancy in the peripheral areas of the city.

(3)

There were differences in the urban vibrancy enhancement among the different mixed land use types. The urban vibrancy intensity of the mixed land use types that included residential or commercial land related to residents’ daily living conditions or consumption was stronger. The urban vibrancy stability of the mixed land use types containing industrial land related to residents’ daily work was stronger. For the urban center area, the urban vibrancy stability can be improved by reasonably improving housing supply to encourage residents to live nearby, appropriately increasing the layout of pollution-free industries, or evacuating some workplaces outside the urban central area. For the peripheral area of the city, providing residents with more living and consumption space can promote the enhancement in urban vibrancy intensity.