[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Next Article in Journal
Lessons Learned from the Last Moments Captured of Traditional Small-Scale Land Use in a European Fen Meadow
Previous Article in Journal
What Kind of Institutional Configuration Incentivizes Farmers’ Behavior in Ecological Value Co-Creation of Cultivated Land?
Previous Article in Special Issue
Characteristics of Urban–Rural Integration at the County-Scale Interface: The Case of Linqu County, China
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 13
Issue 12
10.3390/land13122154
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Enhancing Detailed Planning from Functional Mix Perspective with Spatial Analysis and Multiscale Geographically Weighted Regression: A Case Study in Shanghai Central Region

by
Liu Liu
1,2,
Huang Huang
1,3,* and
Jiali Yang
1
1
Department of Urban Planning, College of Architecture and Urban Planning, Tongji University, Shanghai 200092, China
2
Key Laboratory of Spatial Intelligent Planning Technology, Ministry of Natural Resources, Tongji University, Shanghai 200092, China
3
Key Laboratory of Ecology and Energy Saving Study of Dense Habitat, Ministry of Education, Tongji University, Shanghai 200092, China
*
Author to whom correspondence should be addressed.
Land 2024, 13(12), 2154; https://doi.org/10.3390/land13122154
Submission received: 20 October 2024 / Revised: 5 December 2024 / Accepted: 6 December 2024 / Published: 11 December 2024
(This article belongs to the Special Issue New Technologies and Methods in Spatial Planning, 2nd Edition)
Browse Figures
Figure 1
<p>Workflow of investigating effect of urban functional mix and POI composition at different grid scales.</p> "> Figure 2
<p>Location map of Shanghai city and its central region.</p> "> Figure 3
<p>Experimental data of the research area. (<b>a</b>) Administrative boundary of central region. (<b>b</b>) WorldPop population data. (<b>c</b>) POI data.</p> "> Figure 4
<p>Results of urban functional mix at different grid scales.</p> "> Figure 5
<p>Moran’s I results of functional mix at different grid scales.</p> "> Figure 6
<p>Hotspot analysis results of functional mix at different grid scales.</p> "> Figure 7
<p>Population aggregation patterns at different grid scales.</p> "> Figure 8
<p>Contribution of functional mix and POI-related variables to population configuration at different grid scales. (<b>a</b>) at 300 m scale; (<b>b</b>) at 500 m scale; (<b>c</b>) at 700 m scale.</p> "> Figure 9
<p>MGWR results of different POI types on functional mix at three grid scales. (<b>a</b>) at 300 m scale; (<b>b</b>) at 500 m scale; (<b>c</b>) at 700 m scale.</p> ">
Versions Notes

Abstract

:
Detailed spatial planning serves as statutory guidance for regulating specific spatial functions, including public services, living conditions, and production spaces. It emphasizes meeting the comprehensive needs of the local population, making it crucial to understand the relationship between population distribution and the mix of various city functions, particularly in the era of urban regeneration. Therefore, this study utilized point-of-interest (POI) data representing land functions and population data to investigate these relationships via spatial analysis and Multiscale Geographically Weighted Regression (MGWR). Applied to the central urban area of Shanghai, the study reveals that the level of mixed land use and various functionalities affect population distribution at different adaptive scales. We also found a higher degree of functional mix does not always meet population needs. Although generally there is a positive correlation between functional mix and population distribution, they are not always closely bonded. The proposed method provides an efficient workflow for identifying the applicable scale of various functions to increase functional mix and attract the population, which can provide real-time evidence supporting detailed planning. Test results also reveal the less-considered space along the boundaries of administrative districts. We also found developing tools for detailed planning is an urgent need to facilitate cross-boundary cooperation and development, especially in the context of urban regeneration where they always are overlooked at the detailed planning level. By using open-sourced POI and population data, our proposed workflow can be easily applied to other cities or regions, enhancing their practical value for similar research contexts.

1. Introduction

Detailed planning (which may be referred to by different names), acting as a part of the legal framework within the planning system, directly affects the achievement of planning objectives [1,2,3,4,5,6]. It addresses various aspects such as the environment, living conditions, and production spaces. In the ‘Notice of the Ministry of Natural Resources on Strengthening the Work of Territorial Spatial Detailed Planning’ issued in China in 2023 [7], the focus is placed on developing plans based on the resident population, tailored to meet the comprehensive development needs of the population in the post-pandemic era. These plans should optimize functional layouts according to local conditions and gradually establish a multi-centered, clustered, and networked spatial structure to enhance the balance, accessibility, and convenience of urban services. It is also essential to address the gaps in public services such as employment, education, healthcare, and elderly care within close proximity. Additionally, the layout of pedestrian pathways and public leisure spaces should be improved, and the level of infrastructure related to ecology, safety, and digitalization should be enhanced. In the era of stock land planning, detailed planning must integrate land policies, such as revitalizing inefficient land use, coordinating above-ground and underground spaces, and promoting the development and utilization of underground areas. Mixed-use land development and integrated spatial utilization should be encouraged, guiding the transformation of single-function industrial parks into integrated industrial communities. This approach will improve the efficiency of land use and enhance the overall spatial value [8,9,10,11]. Therefore, it is crucial to explore effective approaches to understand the relationship between population distribution and mixed land use.
Jane Jacobs has emphasized the importance of a diverse mix of land uses for developing vibrant urban environments [12]. Mixed-use land emerged in the late 20th century as a key concept and planning strategy to enhance economic vitality, pursue social equity, and improve environmental quality [13,14,15]. It was also identified to be closely related to housing varieties and density [16,17], which directly affects residential population distribution. Under the influence of New Urbanism, it has become a core principle for achieving sustainable development [16]. Early research primarily focused on the types of mixed-use land at a macro level and the mix ratios of different land uses, as well as the functional compatibility at the neighborhood scale [13]. Discussions often centered on the compatibility of functions such as commercial, entertainment, public services, institutional, and residential uses and their corresponding mixed-use land [13,18]. Key factors in the study of mixed-use land include land use priority, functional accessibility, and land use attributes, which can guide relevant practices [19,20].
Previous studies have demonstrated that simple area-based land use combination indices are insufficient to capture the complex differences arising from subtle variations in land use combinations [21,22]. Consequently, new mixed-use indices have been introduced in studies across various spatial scales (e.g., city, district, neighborhood) and land use classifications [19]. Among these factors influencing the development of mixed-use land, scale, density and mix degree have been identified as the most critical ones [16].
Along with the advances in data collection methods and digital technologies [18,22,23,24,25,26], researchers have extended to new data sources to study land-use mix. POI (point of interest) data are increasingly utilized to measure the mixed functionalities of land, particularly to balance the relationship between mixed-use land and the accessibility of public services. In addition, different studies have focused on urban land-use mix at various spatial scales [25], POI data are more commonly applied at micro- and mesoscales (e.g., street, etc.). Related studies express the multi-purpose nature of land parcels through measuring the balance of POI data within spatial units. For instance, urban functional mix degree can be calculated based on POI data with information entropy [27].
The measurement of mixed functionalities using POI data provides a more accurate, detailed, and comprehensive description of spatial diversity at the meso- and microscales. POI data can reflect changes in the usage of public service and facility in a timely manner. Additionally, research based on POI data is used to identify various factors, such as population changes and socioeconomic activities, associated with urban functional areas. Some researchers mapped different cities to reveal distinct ways of functional mix changes at neighborhood scales and below [28].
There is a spectrum of studies on POI-based functional mix or land-use mix. By using a regression model, an empirical study shows land-use mix positively associates with the number of patents in a 2 km × 2 km grid model of Shanghai [29], which attempts to reflect the spatial effect of land-use mix on technological innovation. Based on machine learning methods, other researchers [30] mapped the population distribution of Guangzhou city in a grid model with multi-source spatial data, including POI and land use data. It results in a clear correlation between land use and population. Moreover, a study [31] has explored the clustering patterns of population and land use in Wuhan by examining factors such as population density, floor area ratio, commercial POIs, road networks, and built-up areas. The density of commercial POIs highly correlates with population density and decreases rapidly with growing distance from the city center.
In addition, multiscale geographically weighted regression (MGWR) [32] is a promising spatial statistical method to investigate geographic phenomena sensitive to spatial scales and can provide insights on spatial heterogeneity and relevant local variations, which also can benefit detailed studies of functional mix. MGWR can capture local spatial heterogeneity by investigating the variations in different variables at distinct spatial scales. Based on POI and OpenStreetMap data, some researchers have analyzed the distribution and aggregation features of land-use mix. A MGWR (multiscale geographically weighted regression) model is employed to investigate the influence of mixed land use on housing prices within the research area [33]. Another study applies MGWR to investigating and comparing the POI influence on urban traffic crashes [34].
Previous studies show that land-use mix significantly influences population distribution. Relevant research is typically conducted in spatial units such as administrative parcels or road-delimited units. In contrast, research on functional mix supported by POI data often employs grid cells. In a suburb district of Shanghai, a study reviewed the optimal spatial scale for plans with a group of selected factors regarding land-use mix on population distribution [25]. The study presents the fact that different grid scales can unintentionally lead to different analysis outcomes. Scholars have used grid models of 200 m × 200 m, 500 m × 500 m, and 1000 m × 1000 m to explore the impact of different grid scales on the identification of functional areas in Xicheng District, Beijing city. In this case, kernel density analysis has been utilized to examine the spatial structure of Xicheng District, and it highlights 200 m × 200 m and 500 m × 500 m grids are most suitable for identifying single-function and mixed-function areas [35]. Most existing studies focus on neighborhood scales, and select different grid scales such as 200 m, 300 m, 500 m and 1000 m [23,35]. However, there is still a lack of sufficient analysis and demonstration on the reasonable grid scales for land-use mix. Therefore, further exploration of key factors such as proper grid scales and related thresholds is required in current research on land functional mix supported by POI data.
To investigate the relationship between POI functional mix and population distribution, our primary research hypothesis is that there is a significant correlation between the degree of POI functional mix and the population in central metropolitan regions. Specifically, we assume:
  • The spatial heterogeneity of these relationships over different locations can be interpreted (at least partially) with the local POI density and the functional mix degree.
  • The correlation between POI functional mix and population distribution can vary across different grid scales.
  • POI types contribute differently to the functional mix degree of each spatial unit (grid), which indirectly shapes the population distribution.
By using spatial analysis and MGWR, we aim to validate the above hypotheses and obtain insights for detailed planning in the context of spatial planning, especially for adjusting the functional mix degree reflected by POI types and balancing their influence on population distribution at various scales. We devise a methodology to investigate the influence of functional mix in urban central areas. It supports us in analyzing the correlation between functional mix and population and evaluating the influence of different POI types on functional mix across various grid scales. Accordingly, a straightforward and simple-to-use workflow is proposed to elicit the appropriate scales for functional mix analysis, confirm the pivotal POI types for population with their effective scales, and pinpoint the POIs’ inconsistent influence on functional mix and population distribution. Our tests are conducted in the central districts of Shanghai city. By using POI data and open data of population, we compare the influence of functional mix calculations at different grid scales and confirm the appropriate grid scales for functional mix analysis are 700 m and below. Furthermore, we compare the impact of distinct POI types on functional mix and population at proper grid scales and find that functional mix and population may not always be closely linked. Thus, we can develop crucial recommendations on POI configuration for achieving detailed planning objectives. This study offers a new insight into revealing the dual influence of POI data on functional mix and population distribution, which also enriches the theory of detailed planning on driving force analysis for population.

2. Methodology

2.1. Overview

Before entering more details, we need to clarify several notions as follows.
  • POI functional composition. It refers to the combination of different functional types of POIs in a given geographic region. The types and distribution of different POI types can reflect the functional diversity of a region [36,37]. It also depends on the data source and the number of POI types considered in a specific study.
  • Functional mix. In this paper, we consider functional mix an indicator of the evenness on different public facilities and/or related functionalities in a given geographic region [28]. We represent the functional mix with information entropy [38,39]. A high entropy value indicates the functionalities within the region are more diverse and balanced; while low entropy values imply insufficient functionalities in the region.
  • Spatial heterogeneity. This term refers to variation across different locations for a spatial feature [40,41]. For instance, here we use it to reveal the differences in the influence of functional mix degree (independent variable) on population distribution (dependent variable) over different locations.
Based on POI data of a study area, the three terms support us to analyze the composition of functional types, calculate the evenness of functional types and reveal the spatial heterogeneity in their influence on a spatial phenomenon or variable. In addition, they lay a foundation for eliciting relevant urban planning suggestions.
Specifically, this study follows a macro-to-micro workflow to develop a methodology for investigating the impact of functional mix on spatial variables in central urban areas. By taking population distribution as an example, this study is divided into two primary objectives: analyzing the correlation between functional mix and population distribution, and evaluating the influence of different POI types on functional mix across different grid scales.
Basic spatial statistical methods such as Moran’s I and Getis–Ord Gi* are employed to detect spatial autocorrelation within the data, helping to determine whether there is a clustering or dispersal effect in population distribution or functional mix within the study area. For grid scales with significant spatial autocorrelation, this study further investigates the relationship between functional mix and population at these scales.
Subsequently, MGWR (multiscale geographically weighted regression) is used to explore spatial heterogeneity, revealing the relationship between functional mix and population distribution across different regions, as well as the local contributions of different POI types to functional mix. This approach enhances our understanding of how POI types via functional mix may produce varying effects on population distribution across different areas.
Methodologically, an analytical workflow is proposed for revealing composition-sensitive effects of functional mix to population configuration. Moreover, we investigate the potential relocation of POI types across different spatial scales.
‘Different POI types’ in Figure 1 can represent different datasets that may contain distinct combinations of facilities or functionalities. Here, we select information entropy to gauge functional mix degrees, which can reflect the evenness between different functionalities with respect to POI data. The conventional spatial statistics methods of Moran’s I and Getis–Ord Gi* are applied to functional mix and population data, which supports us to conduct exploratory data analysis. The primary purpose is to confirm and narrow down the search scope of the optimal spatial scales for the following analyses.
As one may obtain a group of candidates of optimal scales, we check each scale with the same process. First, MGWR analysis is applied to investigating the correlations between population distribution and functional mix/functionality of single POI types. As a result, we can obtain the spatial heterogeneity of the contribution of different independent variables on population. Based on that, we can ensure their key spatial scales where they are essential for the population variable. Meanwhile, we can compare and select the pivotal variables for a given scale based on their performance.
Following that, we can examine the correlations between functional mix and functionality of single POI types by using MGWR. In this procedure, one can explore the scale effects of distinct POI types on the functional mix degree, i.e., to observe the key POI type(s) at a given scale. Also, we can investigate the dual effect of a POI type on both the population and the functional mix at a spatial scale, which can support us to elicit the potential significant facilities for specific regions and provide suggestions for future planning measures.
This workflow focuses on the straightforward correlations between functional mix and its POI composition and population distribution. This design concentrates on the independent impact of spatial functional configuration, and excludes those of other factors involving transportation and economics. This measure promotes the results to be lucidly interpreted. Meanwhile, policymakers and urban planners can adjust POI types and density to optimize functional configuration and then introduce or reduce potential population. This study provides an in-depth analysis of how these POI-related factors influence population distribution across different spatial scales. Some of these variables play key roles in optimizing functional configuration in urban planning.

2.2. Data Preparation

This study focuses on the central urban area of a city. As mentioned above, we require POI data to measure urban functional mix degree in a specific central area and aim to investigate its relationship with population. Detailed planning primarily relies on updating stock land and adjusting mixed land use to meet the diverse living and working needs of residents. This requires effective analytical methods to understand the relationship between population distribution and mixed land use, serving as a foundation to support updates in the detailed plan. In other words, it is about determining what type of functions to introduce and where to enhance an area’s attractiveness to the population.
We selected Shanghai’s central urban area as the target for this exploratory method because there is no new land available (Figure 2). Specifically, the central region of Shanghai city involves dense population, a group of important transportation hubs, and urban landmarks with rich mixed functions. In addition, the open data of Shanghai is well accessible, so it is selected as the research area in this paper.
We adopt the central region of Shanghai, including seven administrative districts: Huangpu, Jing’an, Xuhui, Yangpu, Hongkou, Putuo, and Changning (Figure 3a). To clearly define the study area, we use the complete boundaries of these administrative districts, excluding the Pudong area from the ‘central urban area’. The boundaries of the seven districts are the administrative map data. Various types of POI data (including commercial, residential, educational, medical, etc.) were collected from Amap’s 2020 POI dataset. To align with the daily-life needs of urban residents, we selected the most relevant functional use types, such as residential locations, commercial service facilities, and public service. The original POI data were classified, and irrelevant subcategories like ‘disease prevention institutions’, ‘adult education’ and ‘photo printing shops’ were removed. They finally resulted in seven types: residence, restaurant, shop, workplace (including office buildings/industrial parks), amenity (involving life services), school, and healthcare (including hospitals/clinics) (Figure 3c), with 56,480 entries in total.
Moreover, the WorldPop dataset was adopted as the population data and its spatial units are geographic grids. For the study area, we extracted 2020 population data at a 100 m × 100 m resolution (Figure 3b) to reflect the spatial distribution of urban population. These high-resolution data were aggregated to create population data for other lower resolutions, ranging from 300 m to 2000 m grids.

2.3. Exploratory Spatial Data Analysis on Functional Mix

We employed the commonly used information entropy model to measure urban functional mix degree. The formula for land-use mix structure based on information entropy is given in [42]:
M = − ∑ i = 1 n p i log ( p i )
Equation (1) involves the number of use categories and the area evenness of distinct land uses. We also adopt it to calculate functional mix degree regarding POIs. More specifically, n represents the total number of POI categories and i represents each POI type, and p i represents the proportion of the i-th type of POIs. A large value of M indicates a relatively balanced distribution of POIs in the given spatial unit (i.e., a high degree of functional mix). After spatially joining grid units with their contained POIs, we compute M with the POI categories and the number of POIs in each grid cell. For a specific spatial scale (e.g., 300 m × 300 m grids), the M value reflects the evenness of different POI numbers within the cell. As M could vary in grid cells at different spatial scales, we set grid models with distinct scales to find out the differences in functional mix degrees.
More importantly, we attempt to unveil the spatial correlation between urban functional mix and population at a given scale. Moran’s I is a spatial statistic that identifies spatial autocorrelation, and it is applied to our exploratory analysis. With global Moran’s I, we can examine whether both functional mix and population data present significant autocorrelation of their distribution within the research area, which can extract the spatial characteristics of both variables. The related formula of global Moran’s I is given in [43]:
I = ∑ i ∑ j ≠ i w i j ( x i − x ¯ ) ( x j − x ¯ ) S 2 ∑ i ∑ j ≠ i w i j
S 2 = 1 n ∑ i ( x i − x ¯ ) 2
where x i is the functional mix degree/population value in grid unit i, and x ¯ is the mean of x.  w i j is an element of the spatial weight matrix representing the spatial relationship between grid i and j. n represents the total number of grid cells contained in the study area. S 2 is used to measure the overall dispersion of x i . A significant p-value indicates that a high positive Moran’s I index represents obvious spatial correlation, while a low negative Moran’s I value represents distinguishable spatial difference.
After considering the spatial autocorrelation of M and N, we can employ the ‘Hotspot Analysis’ (Getis–Ord GI*) to investigate their clustering patterns within grid cells at a given spatial scale and to identify statistically significant high-value spots. Its formula is provided in [44]:
G i * = ∑ j = 1 n w i , j x j − X ¯ ∑ j = 1 n w i , j S n ∑ j = 1 n w i , j 2 − ∑ j = 1 n w i , j 2 n − 1
X ¯ = ∑ j = 1 n x j n , S = ∑ j = 1 n x j 2 n − X ¯ 2
where x j is the attribute value (population number/functional mix degree) of grid j, and n is the total number of grid cells and w i , j is an element of the spatial weight matrix. S is used to measure the degree of dispersion of variable x, while G i * corresponds to z-value. A statistically significant positive z-value indicates more intense clustering of high values (hotspots), while a significant negative z-value relates to more intense clustering of low values (cold spots). Hotspot analysis can support us to identify statistically significant clustering patterns of M and N at a specific spatial scale. In this way, the clustering upper bound of spatial scales for functional mix can be confirmed. That is, the search scope of scales for obvious functional mix influence can be narrowed down.

2.4. MGWR Analysis for Functional Mix and Urban Population

MGWR analysis can help us to reveal the spatial non-stationarity of relationships between variables, and statistically provide their impact and changes in different regions [32]. It also increases the explanatory power through the multi-scale perspective by manifesting more detailed spatial heterogeneity. On the spatial scales derived from the exploratory spatial data analysis, MGWR analysis (Figure 1) can be employed to analyze the spatial heterogeneity of functional mix and POI functional composition on the population of the research area. For the correlation between population (Y, dependent variable) and functional mix degree/the grid density of various POIs (X, independent variables), we need to compare the variations in X in the correlations across distinct grid scales and pinpoint their most effective spatial scales. On the basis of these effective scales, further discussions would be led to reveal the pivotal variables for population with their specific grid scales.
More specifically, by comparing the results of global regression (Ordinary Least Squares, OLS) with MGWR, one shall cautiously determine the most active independent variables with statistical significance (identified by p-values). We also can observe the changes of different dependent variables across the given grid scales. Bandwidth and variable coefficients of each independent variable in MGWR results need to be compared as well. For instance, if the STD of coefficients is relatively large and variable bandwidth is limited, we consider it includes significant heterogeneity. Also, the mean of variable coefficients can reflect its overall influence on the dependent variable (population) over the research area. By combining the heterogeneity and obvious impact of these variables, one can measure and summarize the influence range of a variable X on Y. In addition, the visualization of MGWR coefficients can illustrate the variables’ influence on population and promote the organization of planning suggestions.
MGWR is also applied to gauging the non-stationarity of the influence of POI composition to functional mix. Based on the selected POI types, MGWR variable coefficients can reflect different contribution of POI types, which can support us to identify POI impact on the evenness of functionalities (i.e., functional mix) with their effective scales in a central urban area. Similarly, visualizing these coefficients can depict the correlation of each POI type with functional mix and locate the uneven regions at all the selected scales. These regions could be of concern for public service/facility configuration.
Furthermore, we need to compare the correlations of a POI type with both functional mix and population at different spatial scales. This measure can provide insights into whether certain patterns are consistent. For instance, if a POI type positively relates to the both dependent variables at all scales, the POI type could promote the increase in population via functional mix. In contrast, a POI type positively relates to the population and negatively relates to functional mix at a specific scale, which may imply the POI type has no contribution to functionality evenness, but it still involves positive influence on population. We name such results as ‘dual effects’ of POI types on functional mix and population and summarize active POI types with their effective spatial scales. Consequently, advice on urban planning in the research area can be drafted according to multiscale heterogeneity of functional mix and POI functional composition in the population.

3. Tests and Results

3.1. Exploratory Spatial Data Analysis

As mentioned above, we aim to adopt various statistical methods to assist decision-making in detailed planning on population according to functional mix. We introduce and create an easy-to-use and interpretable workflow that can provide objective-oriented results based on any available dataset of POIs and provide interpretation and suggestions for the key POI-related variables of a research region. In the following part, we will present the tests conducted in the research area and their results.
Functional mix degree may vary across distinct spatial scales. By preparing seven grid models (from 300 m to 2000 m), we calculated functional mix degrees for each grid scale by applying Equation (1) to investigate the differences of functional mix degrees (see Figure 4). An obvious fact is that functional mix degrees are more even at a lower resolution grid model, such as 2000 m × 2000 m. In the grid models with resolution larger than 1000 m, the public service facilities represented by these POI types are close to evenly distributed. Therefore, the appropriate spatial scale should be grids below 1000 m to distinguish functional mix degrees in a research area.
On the seven grid scales, the clustering patterns reveal the differences between urban functional mix. Figure 5 shows that 2000 m is a threshold for detecting significant autocorrelation since the value is almost 0. From the grid scale of 700 m to 1000 m, the Moran’s I index of functional mix declines sharply, which indicates autocorrelation is more pronounced at smaller scales. In contrast, the Moran’s I index of 300 m or 500 m scales are 0.528 and 0.535, respectively. Both scales can reflect the heterogeneity of urban functional mix within the corresponding grid cells.
By applying the ‘hotspot analysis’ method, Figure 6 shows the clustering patterns of functional mix at the seven grid scales. As the grid size increases, the high-value clusters of functional mix significantly shrink. Combining the results in Figure 5 and Figure 6, we consider the scale 700 m the threshold for investigating the heterogeneity of functional mix. It is also consistent with the results of the Moran’s I index of functional mix (Figure 5). Thus, grid models with resolution lower than 700 m can better reflect the differences in functional mix and their clustering patterns. It confirms the applications in the literature that a neighborhood level is appropriate for analyzing mixed-use land with POIs, which may be limited to the service scope of public facilities. For instance, restaurants and shops are often densely concentrated in small regions and they naturally form a local functional mix. In this work, we select the grid scales from 300 m to 700 m to conduct the following tests with MGWR analysis.
In contrast, the ‘hotspot analysis’ results of population (Figure 7) imply that population clustering is stable across the seven spatial scales. Even at the largest grid scale of 2000 m, the hotspots of population remain obvious in the core part of the research area. It could result from other factors beyond functional mix, such as housing density and employment opportunities.
The results of this exploratory analysis underscore the different scale effects on the clustering of functional mix and population. These findings suggest that the factors driving the distribution of functional mix and population may not align perfectly across spatial scales. Specifically, functional mix degree may correlate with the distribution pattern of population at certain small scales. As the impact of functional mix is primarily observed at small scales, it is more suitable for micro-level planning and design. Therefore, the most appropriate grid scale for analyzing the relationship between functional mix and population distribution should be lower than 700 m. In the following subsection, we will investigate their correlations only on 300 m, 500 m and 700 m scales.

3.2. Correlation of Functional Mix with Population

3.2.1. Overall Interpretation of MGWR Results

We select the seven types of POIs (see Figure 3c) and the corresponding functional mix degree as independent variables (X) and add the population in the research area of Shanghai as dependent variable (Y). By applying both global regression (i.e., Ordinary Least Squares, OLS) and MGWR, we can confirm the performance of MGWR in our tests. These results are organized in Table 1.
Compared with global regression, MGWR has a better performance on R 2 . R 2 reflects the model accounts for the variance degree in the dependent variable, i.e., goodness of fit. R 2 values of the MGWR are 0.945, 0.924, and 0.902 from 300 m to 700 m, significantly improved from those of global regression ( R 2 = 0.362, 0.472, 0.508). The global regression results can reveal the overall influence of the X variables. This model shows that the variables mix degree, residence and workplace have significantly positive correlations (see p-values in Table 1) with the dependent variable (population), while shop, school, amenity, and healthcare are insignificant globally. A special case is restaurant: it is significant at the 300 m and 500 m scales but not at the scale of 700 m. At the microscales of 300 m and 500 m, restaurants may primarily cater to surrounding local residents.
The global regression model failed to capture the complete spatial heterogeneity, while the MGWR (multiscale geographically weighted regression) model significantly improves its explanatory power for population. At the scale of 300 m, mix degree, residence, workplace and amenity show significant heterogeneity in the population. It is reflected in their bandwidths at different scales, such as mix degree (46, 43, 44), residence (44, 43, 43), and workplace (44, 45, 44) at scales of 300 m, 500 m, and 700 m. Specifically, the ‘variable coefficients’ in Table 1 represent the strength of each independent variable’s impact on Y at different locations and show the correlation direction (positive or negative). ‘Bandwidth’ can capture the heterogeneous impact of these independent variables on population distribution at different grid scales.
For the MGWR results, the larger the absolute value of a variable’s standardized coefficient, the greater its impact on the dependent variable (population in our case). These coefficients in Table 1 are mean values and they reflect the overall impact of these variables on population distribution at different scales. We rank the importance of these coefficients to the population. The overall positive correlation between ‘mix degree’ (0.232)/ ‘restaurant’ (0.152), and population distribution reaches a peak at the scale of 500 m, while ‘residence’ and ‘workplace’ reach their greatest influence at the 700 m scale (0.183 and 0.168, respectively). In addition, ‘residence’ and ‘workplace’ consistently maintain high positive correlations with population distribution across all the scales from 300 m to 700 m.
Table 2 lists the mean and standard deviation (STD) values of the estimated coefficients for independent variables. There is a significant and persistent heterogeneity of ‘mix degree’/‘residence’/‘workplace’ from 300 m to 700 m (see ‘bandwidth’ in Table 1 and ‘STD’ in Table 2), and they positively correlated with population. ‘Restaurant’ is homogeneous at 300 and 500 m but heterogeneous at 700 m (bandwidth = 44, STD = 0.18). In contrast, ‘amenity’ contains significant spatial heterogeneity at the scales of 300 m and 500 m (STD = 0.132, 0.138), while its influence becomes uniform at 700 m. This implies its influence gradually becomes homogeneous along with scale-up.
In general, the consistent heterogeneity and large coefficients of ‘mix degree’ and ‘workplace’ at the three scales indicate they are key factors for population distribution on a micro-level. Although ‘restaurant’ has a homogeneous influence on the 500 m scale (bandwidth = 1162, STD = 0.002), it contains the highest positive correlation (mean = 0.152) with population distribution. Thus, the density of restaurants on this scale could better measure the living population. ‘Restaurant’ also shows a significant heterogeneity at the 700 m scale (STD = 0.180 in Table 2).
The bandwidths of ‘residence’ at the three scales are 44, 43, and 43 (Table 1), and their STDs are 0.176, 0.255, and 0.316, respectively (see Table 2). This result presents its evident spatial heterogeneity from 300 to 700 m scales. But it significantly correlates with population distribution (mean = 0.183) only at the 700 m scale (Table 2).
At the urban microscales (300–700 m), the coefficients of ‘shop’, ‘school’, and ‘healthcare’ are approximately 0, and their STDs are relatively small as well. It indicates the three variables involve a limited and stable impact on the population. At the 700 m scale, the STD of ‘healthcare’ is slightly higher (0.054). It can be interpreted that the impact of POI types such as shop, school, and healthcare on population may exist at a macro-scale, possibly due to the following reasons: (1) places with high density of shops may lead to more commercial activities than just residence; (2) the shopping, school, and healthcare areas may be separated from main residential areas.
The above results demonstrate that an appropriate spatial scale is crucial for accurately revealing the relationships between population and functional mix and other independent variables of POIs.

3.2.2. Spatial Pattern of MGWR Coefficients

Based on Table 1 and Table 2, we summarize the salient independent variables with spatial heterogeneity and list their influence in terms of their mean coefficients in Table 3. In general, ‘amenity’ shows obvious heterogeneity on population at 300 m and 500 m, though its influence is limited. ‘Restaurant’ has no clear heterogeneity at 500 m while it shows obvious influence on population. Its heterogeneity occurs at 700 m with a limited influence. For the other variables with the scales in Table 3, they present both heterogeneity and salient impact on population.
Here, we compare the distribution of variable coefficients with significant heterogeneity in the research area. The spatial pattern of MGWR-estimated coefficients for the heterogeneous variables at each scale are shown in Figure 8. Although the overall positive correlation indicates ‘mix degree’ may generally boost the population at 500 m and 700 m, its heterogeneity suggests different regions may require different strategies to enhance functional mix to optimize population. The results of ‘functional mix’ at the three scales show its impact on population is positively correlated in most regions (Figure 8), and the significant effect is in the central parts of Huangpu, Jing’an, Hongkou, Changning, and Minhang.
Compared with the ‘residence’ coefficients at the three scales, the positive correlation of ‘residence’ to population concentrates in Putuo, Changning, Minhang, and Yangpu districts. These regions can be regarded as marginal parts of the research area. The positive correlation between ‘workplace’ and population is similar to that of ‘residence’, with a greater positive impact in peripheral areas. But ‘workplace’ covers a larger scope. As mentioned before, ‘residence’ has the greatest positive correlation with population at 700 m. At this scale, its negative correlation covers the central parts of all districts representing the locations of significant commercial vitality (see Figure 8). At the scale 300 m, the range of ‘amenity’ with positive coefficient also approximates that of ‘residence’ but differs at the scale of 500 m. In general, the positive correlation of ‘amenity’ concentrates in the city centers.The positive correlation of ‘restaurant’ at the 700 m scale covers a larger range of the central region.

3.3. Contribution of POIs to Functional Mix

As mentioned before, we calculated functional mix with information entropy, and the essence is the balance between different functionalities reflected by POI composition. The high values of functional mix concentrate in the city centers from scales 300 m to 700 m (Figure 6), while peripheral regions involve low-value clusters of functional mix in the research area. Obviously, the city centers with high entropy contain various types of functionalites and they counter a relatively balanced distribution. In contrast, in the peripheral regions, the proportion of a certain type of functionality may be high, while other functionalities are limited. Then, it can result in degenerating the accessibility of services.
To analyze the impact of different POI types on functional mix and to understand the relationships and their contributions, this study also applied MGWR to assess how various POI types correlate with functional mix. By comparing the impact of different POI types on population and functional mix, we can estimate in which type and direction that functional mix degree may need to be adjusted. In this case, we identify which POI types have a significant impact on functional mix across the three grid scales. For the research area, Figure 9 visualizes the MGWR coefficients of different POI types at the three scales to illustrate their variations.
The heterogeneity and coefficient mean of ‘shop’ reach the peak at the scale 500 m (bandwidth = 110, coefficient = 0.235, STD = 0.247). Under the same criterion, ‘restaurant’ includes the most heterogeneity and negative correlation at 700 m (bandwidth = 47, coefficient = −0.450, STD = 0.212). The similar cases include ‘amenity’ at 300 m (bandwidth = 46, coefficient = 0.357, STD = 0.379) and ‘healthcare’ at 500 m (bandwidth = 110, coefficient = 0.198, STD = 0.139). The above POI types and the related scales link to potential adjustment measures for specific local areas, i.e., assisting urban planners to identify areas that require differentiated policies on facilities at the micro-level.
From scales 300 m to 700 m, the influence of ‘residence’ on the population gradually increases from the central to the periphery regions. It positively lifts the degree of functional mix in the peripheral areas.Based on this trend, the dual effect of ‘residence’ on the increase in functional mix and population growth occurs in the peripheral regions. This result implies that residences outside the central area could effectively drive a new population growth.
In addition, there is significant heterogeneity of ‘residence’ on the functional mix across the three scales. Except for certain small parts, it positively correlates with functional mix degree in all other locations. This implies the current configuration of ‘residence’ is beneficial for improving functional mix in most cases, and the effect is more legible in the peripheral areas of the research area.
‘Workplace’ significantly and positively correlates with population and it shows heterogeneity across the three scales. Also, it contains an overall homogeneous positive correlation with the functional mix, which implies ‘workplace’ POIs can improve the degree of functional mix globally in the research area, though with a limited impact (Figure 9).
‘Amenity’ shows a uniform and weak positive correlation with functional mix at the 500 m and 700 m scales, which can be considered an equilibrium state. On the contrary, there is a significant heterogeneity at the 300 m scale, and ‘amenity’ greatly contributes with high values to functional mix in the peripheral parts of the research area. According to its correlation with population (Figure 8), there is an overlap between the two types of positive correlation ranges at the 300 m scale.
The impact of ‘school’ on the population is not significant (approximately 0). But, ‘school’ is positively globally correlated with the functional mix and shows significant high values in the peripheral regions. Though schools, as public service facilities, have played a role in improving the degree of functional mix, they may not be a proper indicator to reflect population variations at a microscale.
Similarly, the correlation of ‘shop’ and ‘healthcare’ with population is not significant (approximately 0). Thus, ‘shop’ and ‘healthcare’ are not suggested as the indicators for population at a microscale. But, they show a homogeneous positive correlation with functional mix at scales of 300 m and 700 m. At the scale 500 m, the high correlation between ‘shop’ and functional mix occurs in the northwest, south, and northeast parts. In addition, a high correlation between ‘healthcare’ and functional mix occurred in the south and north of the research area. It suggests 500 m may be the optimal grid scale to observe significant increase in functional mix in these areas, which could result from the differences in the functional positioning of the related administrative districts.
In certain cases, the spatial distribution of ‘restaurant’ facilities may differ remarkably from other POI types. For example, restaurants may gather in specific commercial regions or shopping centers, while ‘healthcare’ or ‘educational’ facilities distribute in a more sparse way in different districts.
Specifically, ‘restaurant’ shows a homogeneous negative correlation with functional mix at scales of 300 m and 500 m (Figure 9). Its heterogeneity surges at 700 m, and the negative correlation is strengthened in some parts. By comparing Figure 8 and Figure 9, we find ‘restaurant’ positively correlates with the population in the central part of research area at scale 700 m, while it shows a negative correlation with the functional mix in similar regions. Although increase in ‘restaurant’ numbers at a microscale is not conducive to raising the degree of functional mix, it could lead to population growth in the majority of the research area at the scales of 500m and 700 m. This comparison reminds us that some urban areas with a single dominant POI type may not always result in a population decrease. By focusing on an appropriate spatial scale, we could better comprehend its role via functional mix.

4. Discussion

Drawing on the application of spatial analysis and MGWR in the case study region (i.e., the central urban area of Shanghai), we need to further discuss some critical points regarding revealing population distribution and land mix use as follows.
(1) Effective analysis of mixed land use methods and the data representing various land functions requires an adaptive scale. Mixed land is encouraged in detailed planning in the spatial planning system in China. It can be regulated in the form of ‘large plot’ in the unit planning of 1–5 km2 [11]. It is also often discussed from the perspective of mixed functions in one land plot and within one building (or building group) [9]. Similarly, mixed land use in blocks and buildings is an essential part of urban regeneration and population distribution in France [45]. However, current discussions mainly focus on the spatial unit of one block, which may limit the effects of mixed land use and miss the most appropriate scale for increasing the degree of mixed use of functions.
Thus, an appropriate scale is crucial to ensure that analysis outcomes can support detailed planning. For example, the cross-referenced analysis method proposed in this research is applicable to grid scales ranging from 300 to 700 m, aligning with the spatial unit used in detailed planning within the territorial spatial planning system. The analysis of functional mix, indicated by points of interest (POIs), suggests an optimal spatial scale threshold of less than 1 km. This scale is most suitable for assessing neighborhood functional diversity and comparing differences between areas. Conversely, at larger scales within city center grid models (greater than 1 km), the functional mix of POIs is high and values tend to converge, suggesting a more balanced and stable mix of functionalities. Furthermore, by examining the spatial impact and changes in different POI variables on population distribution and functional mix, the results show that the density of single-type POIs correlates with population configuration and functional mix patterns, with significant spatial differences at microscales. Therefore, this method has both advantages and limitations in supporting spatial planning.
(2) Understanding which types of functional mixes meet people’s needs and the most efficient scale for enhancing land-use mix is crucial for achieving detailed planning objectives. This is often discussed in the form of various ‘best recipes’ for mixed functions in one plot or vertical mixture in buildings [9,10]. To allow more flexibility in spatial planning, some cities (e.g., Shenzhen) avoid regulating the proportions of the land mixture in one plot but only regulates the main use of the land and leaves the mixed use to the market [46]. This study shows the necessity to build awareness of various functions and the most appropriate spatial scale for their positive influence on mixed-use land
In our study, selected land use functions showed distinct contributions to land-use mix and population distribution. For instance, ‘residence’ and ‘workplace’ are POI types with significant dual effects on population and functional mix in the study area. In contrast, ‘school’, ‘shop’ and ‘healthcare’ positively impact functional mix but have a relatively small influence on the population at scales of 300 m, 500 m and 700 m. Specifically, ‘restaurant’ negatively correlates with functional mix globally but positively correlates with most population at the 700 m scale. At the 300 m scale, ‘amenity’ shows a significant dual effect, correlating with both population and functional mix. These results provide a foundation for detailed facility configuration in the future, allowing detailed planning to identify the most needed functions and their impact areas.
It is worth noting that our method, applied in Shanghai, tested only a limited but relevant selection of POIs, though it is adaptable to various types of POIs representing different land uses and functionlities. By tailoring POI selections to different regions and development goals, this approach can be particularly useful for identifying the most relevant POIs at a specific scale supporting detailed planning.
(3) Land-use mix and population distribution have a positive relationship, yet they are not always closely linked. Some functions that enhance land-use mix are highly related to population distribution [16] and vice versa. Land-use mix generally positively affects population distribution, as observed in most cases within the study region. Take ‘amenity’ as an example, increasing it at various scales may enhance land-use mix but could have less impact on population distribution except the 300 m scale. We also found that certain land-use functions, such as restaurants, may contribute little to land-use mix or regional development from a quantitative perspective, yet they play a crucial role in attracting people to visit and settle. In other words, even if a land use function has minimal impact on improving land-use mix—perhaps due to an overabundance—it may still significantly boost population distribution. Therefore, it is critical to identify these types of POIs and land functions, along with their influencing scales, to find out its mechanism and support detailed planning.
(4) Land-use mix and population distribution are strongly correlated at district borders. Currently, mixed land use is primarily guided by detailed planning within specific spatial units [9,11,46]. For instance, in Shanghai, the typical unit spans approximately 3 km2, while in Beijing it ranges from 1 to 3 km2. In Guangdong, the spatial units generally fall between 1 and 5 km2, whereas Shenzhen employs smaller units, typically between 1 and 2 km2 [11]. Despite detailed plans encompassing urban development areas, functions along boundaries (particularly those between districts shaped by administrative divisions) are often neglected.
In the study areas, the analysis shows that ‘amenity’ (at 300 m) and ‘residence’ that play important roles in land-use mix, are also highly positively related to population distribution around district borders. This may relate to how detailed planning is conducted and implemented. Typically, detailed planning is developed by local governments and primarily focuses on areas within administrative boundaries. As a result, functions such as ‘amenity’ and ‘residence’ may not locate at the boundaries. The land along these borders, which administratively belongs to different districts, may fall only within the service radius. This leads to land use along the boundaries being less coordinated. We argue that despite higher-level plans guide detailed planning, there is often less consideration at administrative borders. However, development dynamics should not be limited by these invisible lines, especially in urban centers. Therefore, we suggest that detailed planning should explore approaches to develop plans that extend beyond district boundaries.

5. Conclusions

As stock land use and urban regeneration become new themes and approaches in spatial planning, a critical issue for detailed planning is analyzing the most effective ways to optimize spatial use. Increasing mixed land use is one option to achieve this. This raises key questions: What is the appropriate scale for discussing land-use mix in detailed planning? Where should new functions be introduced to enhance mixed land use? And what specific functions should be introduced? This study tried to respond to these questions from the perspective of population distribution.
To understand how the relevant relationship between land-use mix represented by functional mix and population distribution varies across different spatial scales, this study reveals the range and intensity of the functional mix’s impact at each scale by proposing an analytical method combining spatial analysis and MGWR. It lays a foundation for developing more refined and region-specific urban planning strategies. It applied POI data and population data, which are open-sourced, the research workflow can be easily applied to other cities or regions, which increases its practical value for similar research scenarios. Our methods regarding POIs are consistent since POI data are readily available, and the analysis results could be robust due to its simplicity. This study provides a direct quantitative methodology and analytical basis for these efforts. We argue that the proposed methodology can properly support straightforward analysis on planning suggestions especially on revealing important driving forces of functional mix on population for detailed planning.
Applying the methods in the central area of Shanghai, the experimental results from the study area highlight four critical points that support detailed planning. First, effective analysis of mixed land use methods and data representing various land functions requires an adaptive scale. In the case of Shanghai, discussions on the contribution of mixed-use land, amenities are most effective at a 300-meter scale, while residences show a strong impact from the scale of 300 to 700 m. Moreover, the optimal scale for mixed land use may vary when applied to other cities, highlighting the limitations of current detailed planning. These plans often regulate or promote mixed land use merely at the plot level within standardized spatial scales between 1 and 5 km2. An appropriate scale is essential to ensure that analysis outcomes support detailed planning. This leads to our second point: understanding which types of functional mixes meet people’s needs and identifying the most efficient scale for enhancing land-use mix are crucial for achieving detailed planning objectives. Besides considering scales, it is important to note that we cannot assume that a higher degree of function mix automatically better suits population needs. While land-use mix and population have a positive relationship, they are not always closely linked. As mentioned before, current detailed planning within spatial planning system is often conducted within standard spatial units, and it tends to overlook boundary functions. Thus, there is an urgent need to develop planning tools that facilitate cross-boundary cooperation and development, especially regarding stock land planning.
Although this study has demonstrated its feasibility, it still contains some limitations. The adopted POI and population data are static, and they cannot reflect the dynamic evolutionary relationship between functional mix and population distribution. To further reveal the long-term evolutionary trends, it would be useful to collect more time-series data and analyze them to compare the dynamic influence of functional mix on population over time. In addition, from the perspective of refined governance, population mobility data can be further incorporated to consider short-term population change patterns. When one analyzes the scale impact of public facilities and services, the proposed model may also need further development to account for POI size differences in the next phase.

Author Contributions

Conceptualization, L.L. and H.H.; methodology, L.L. and H.H.; validation, L.L. and H.H.; formal analysis, L.L. and H.H.; investigation, H.H. and L.L.; resources, H.H.; data curation, L.L.; writing—original draft preparation, L.L. and H.H.; writing—review and editing, L.L., H.H. and J.Y.; visualization, L.L.; supervision, L.L. and H.H.; project administration, H.H.; funding acquisition, H.H. All authors have read and agreed to the published version of the manuscript.

Funding

The research has been funded by the National Key Research and Development Plan (Grant No. 2022YFC3800801), National Natural Science Foundation of China (NSFC, Grant No. 52208075), and the Shanghai Pujiang Program (Grant No. 2021PJC113). It was also supported by the Fundamental Research Funds for the Central Universities.

Data Availability Statement

The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
MGWRMultiscale Geographically Weighted Regression
POIPoint of Interest
OLSOrdinary Least Squares
STDStandard Deviation

References

  1. Zhao, M. On the construction logic and implementation agenda of the territory development planning system of China. In Urban Planning Forum; Tongji University: Shanghai, China, 2019; Volume 4. [Google Scholar]
  2. Zhao, G.; Li, C. Thoughts on the reform of detailed planning within the territory development planning system. In Urban Planning Forum; Tongji University: Shanghai, China, 2019; Volume 4, pp. 37–46. [Google Scholar]
  3. Pan, H.; Zhao, M. On the reform course, basic contents and main features of territory planning system of China. Urban Rural. Plan. 2019, 5, 4–10. [Google Scholar]
  4. Huang, M.; Zhao, Y.; Gao, J.; Wang, Y. Planning and rules: Discussion on the development direction of regulatory detailed planning. City Plan. Rev. 2020, 44, 52–57, 87. [Google Scholar]
  5. Han, W.; Qiu, Z.; Wang, M.; Zhang, Z.; Wan, C.; Yang, X. The Enlightenment of Zoning Practice in the U.S. Under the National Land and Spatial Planning Framework to Chinese Regulatory Detailed Planning Reform. Urban Plan. Int. 2020, 35, 89–95. [Google Scholar]
  6. Gao, J.; Zhao, M. Origin, evolution, and transformation of regulatory detailed planning in the new era: Based on the paradigm of historical institutionalism. City Plan. Rev. 2021, 45, 72–104. [Google Scholar]
  7. Ministry of Natural Resources of China. Notice of the Ministry of Natural Resources on Strengthening the Work of Territorial Spatial Detailed Planning. 2023. Available online: https://www.gov.cn/zhengce/zhengceku/2023-03/25/content_5748273.htm (accessed on 5 December 2024).
  8. Lin, L.; Guo, J. Exploration on Compilation of the Shanghai Regulatory Detailed Planning with Implementation Orientation: A Case Study of the Open Area Hub Area, Lin’gang New Area. Shanghai Urban Plan. Rev. 2023, 01, 47–52. [Google Scholar]
  9. Tang, S.; Zhang, J.; He, H.; Wang, Y.; Chen, J. Mixed-use development of land and institutional innovation based on the integration of planning, construction, and management. City Plan. Rev. 2023, 47, 4–14. [Google Scholar] [CrossRef]
  10. Chen, D.; Zhou, D.; Hu, J.; Jin, H. The Transformation of Space Production Logic and the Response of Compilation and Management in Regulatory Planning. Planners 2024, 06, 72–80. [Google Scholar]
  11. Hu, R.; Cai, Y.; Liu, Z. Research on the Transformation of Local Planning Units in Territorial Spatial Detailed Planing. Planners 2024, 7, 1–8. [Google Scholar]
  12. Jacobs, J. The Death and Life of Great American Cities; Random House: New York, NY, USA, 1961. [Google Scholar]
  13. Grant, J. Mixed use in theory and practice: Canadian experience with implementing a planning principle. J. Am. Plan. Assoc. 2002, 68, 71–84. [Google Scholar] [CrossRef]
  14. Grant, J. Planning the Good Community: New Urbanism in Theory and Practice; Routledge: Brighton, UK, 2005. [Google Scholar]
  15. Ghosh, P.A.; Raval, P.M. Reasoning the social benefits of mixed land-use and population density in an Indian city. J. Eng. Res. 2022, 10, 17. [Google Scholar] [CrossRef]
  16. Raman, R.; Roy, U.K. Taxonomy of urban mixed land use planning. Land Use Policy 2019, 88, 104102. [Google Scholar] [CrossRef]
  17. O’Driscoll, C.; Crowley, F.; Doran, J.; McCarthy, N. Land-use mixing in Irish cities: Implications for sustainable development. Land Use Policy 2023, 128, 106615. [Google Scholar] [CrossRef]
  18. Tian, L.; Liang, Y.; Zhang, B. Measuring residential and industrial land use mix in the peri-urban areas of China. Land Use Policy 2017, 69, 427–438. [Google Scholar] [CrossRef]
  19. Jaroszewicz, J.; Denis, M.; Fijałkowska, A.; Graszka, O.; Pluto-Kossakowska, J.; Krzysztofowicz, S. Spatially explicit mixed-use indicators to measure life quality across the city—A conceptual framework and case study: Piaseczno—A medium sized city in the peri-urban zone of Warsaw, Poland. Cities 2023, 137, 104296. [Google Scholar] [CrossRef]
  20. Motieyan, H.; Azmoodeh, M. Mixed-use distribution index: A novel bilevel measure to address urban land-use mix pattern (A case study in Tehran, Iran). Land Use Policy 2021, 109, 105724. [Google Scholar] [CrossRef]
  21. Manaugh, K.; Kreider, T. What is mixed use? Presenting an interaction method for measuring land use mix. J. Transp. Land Use 2013, 6, 63–72. [Google Scholar] [CrossRef]
  22. Zhuo, Y.; Zheng, H.; Wu, C.; Xu, Z.; Li, G.; Yu, Z. Compatibility mix degree index: A novel measure to characterize urban land use mix pattern. Comput. Environ. Urban Syst. 2019, 75, 49–60. [Google Scholar] [CrossRef]
  23. Liu, Y.; Zhan, Z.; Zhu, D.; Chai, Y.; Ma, X.; Wu, L. Incorporating multi-source big geo-data to sense spatial heterogeneity patterns in an urban space. Geomat. Inf. Sci. Wuhan Univ. 2018, 43, 327–335. [Google Scholar]
  24. Ma, W.; Jiang, G.; Zhou, T.; Zhang, R. Mixed land uses and community decline: Opportunities and challenges for mitigating residential vacancy in peri-urban villages of China. Front. Environ. Sci. 2022, 10, 887988. [Google Scholar] [CrossRef]
  25. Liu, L.; Huang, H.; Qi, J. Towards a Multi-Scale Effect of Land Mixed Use on Resident Population—A Novel Explanatory Framework of Interactive Spatial Factors. Land 2024, 13, 331. [Google Scholar] [CrossRef]
  26. Bordoloi, R.; Mote, A.; Sarkar, P.P.; Mallikarjuna, C. Quantification of land use diversity in the context of mixed land use. Procedia-Soc. Behav. Sci. 2013, 104, 563–572. [Google Scholar] [CrossRef]
  27. Xue, B.; Xiao, X.; Li, J.; Zhao, B.; Fu, B. Multi-source data-driven identification of urban functional areas: A case of Shenyang, China. Chin. Geogr. Sci. 2023, 33, 21–35. [Google Scholar] [CrossRef]
  28. Dovey, K.; Pafka, E. What is functional mix? An assemblage approach. Plan. Theory Pract. 2017, 18, 249–267. [Google Scholar] [CrossRef]
  29. Jiang, H.; Xiong, W. The Impact of Land-Use Mix on Technological Innovation: Evidence from a Grid-Cell-Level Analysis of Shanghai, China. Land 2024, 13, 462. [Google Scholar] [CrossRef]
  30. Zhao, G.; Yang, M. Urban Population Distribution Mapping with Multisource Geospatial Data Based on Zonal Strategy. ISPRS Int. J. Geo-Inf. 2020, 9, 654. [Google Scholar] [CrossRef]
  31. Li, Z.; Jiao, L.; Zhang, B.; Xu, G.; Liu, J. Understanding the pattern and mechanism of spatial concentration of urban land use, population and economic activities: A case study in Wuhan, China. Geo-Spat. Inf. Sci. 2021, 24, 678–694. [Google Scholar] [CrossRef]
  32. Oshan, T.M.; Li, Z.; Kang, W.; Wolf, L.J.; Fotheringham, A.S. mgwr: A Python Implementation of Multiscale Geographically Weighted Regression for Investigating Process Spatial Heterogeneity and Scale. ISPRS Int. J. Geo-Inf. 2019, 8, 269. [Google Scholar] [CrossRef]
  33. Yang, H.; Fu, M.; Wang, L.; Tang, F. Mixed Land Use Evaluation and Its Impact on Housing Prices in Beijing Based on Multi-Source Big Data. Land 2021, 10, 1103. [Google Scholar] [CrossRef]
  34. Qu, X.; Zhu, X.; Xiao, X.; Wu, H.; Guo, B.; Li, D. Exploring the Influences of Point-of-Interest on Traffic Crashes during Weekdays and Weekends via Multi-Scale Geographically Weighted Regression. ISPRS Int. J. Geo-Inf. 2021, 10, 791. [Google Scholar] [CrossRef]
  35. Luo, S.; Liu, Y.; Du, M.; Gao, S.; Wang, P.; Liu, X. The influence of spatial grid division on the layout analysis of urban functional areas. ISPRS Int. J. Geo-Inf. 2021, 10, 189. [Google Scholar] [CrossRef]
  36. Huang, W.; Cui, L.; Chen, M.; Zhang, D.; Yao, Y. Estimating urban functional distributions with semantics preserved POI embedding. Int. J. Geogr. Inf. Sci. 2022, 36, 1905–1930. [Google Scholar] [CrossRef]
  37. Chin, W.C.B.; Fu, Y.; Lim, K.H.; Schroepfer, T.; Cheah, L. Identifying urban functional zones by analysing the spatial distribution of amenities. Environ. Plan. B Urban Anal. City Sci. 2024, 51, 1274–1289. [Google Scholar]
  38. Li, Y.; Liu, C.; Li, Y. Identification of Urban Functional Areas and Their Mixing Degree Using Point of Interest Analyses. Land 2022, 11, 996. [Google Scholar] [CrossRef]
  39. Xia, X.; Lin, K.; Ding, Y.; Dong, X.; Sun, H.; Hu, B. Research on the coupling coordination relationships between urban function mixing degree and urbanization development level based on information entropy. Int. J. Environ. Res. Public Health 2021, 18, 242. [Google Scholar] [CrossRef]
  40. Goodchild, M.F. The validity and usefulness of laws in geographic information science and geography. Ann. Assoc. Am. Geogr. 2004, 94, 300–303. [Google Scholar] [CrossRef]
  41. Jiang, B. Geospatial Analysis Requires a Different Way of Thinking: The Problem of Spatial Heterogeneity. In Trends in Spatial Analysis and Modelling: Decision-Support and Planning Strategies; Behnisch, M., Meinel, G., Eds.; Springer International Publishing: Cham, Switzerland, 2018; pp. 23–40. [Google Scholar] [CrossRef]
  42. Chen, Y.G.; Liu, M.H. The basic laws of the Shannon entropy values of urban land-use composition. Hum. Geogr. 2001, 16, 20–24. [Google Scholar]
  43. Zhou, X.; Lin, H. Moran’s I. In Encyclopedia of GIS; Shekhar, S., Xiong, H., Eds.; Springer: Boston, MA, USA, 2008; p. 725. [Google Scholar] [CrossRef]
  44. Getis, A.; Ord, J.K. The Analysis of Spatial Association by Use of Distance Statistics. Geogr. Anal. 1992, 24, 189–206. [Google Scholar] [CrossRef]
  45. Yang, C.; Tang, M. 15-minute city: Exploration and enlightenment of urban renewal in France in the post-epidemic era (Self-translation). Beijing Plan. Constr. 2023, 01, 38–43. [Google Scholar]
  46. Zhang, J.; He, H.M. Reflections on the policy of mixed land use to meet the needs of scientific and technological innovation. China Land 2021, 10, 4–7. [Google Scholar] [CrossRef]
Land 13 02154 g001
Figure 1. Workflow of investigating effect of urban functional mix and POI composition at different grid scales.
Figure 1. Workflow of investigating effect of urban functional mix and POI composition at different grid scales.
Land 13 02154 g001
Land 13 02154 g002
Figure 2. Location map of Shanghai city and its central region.
Figure 2. Location map of Shanghai city and its central region.
Land 13 02154 g002
Land 13 02154 g003
Figure 3. Experimental data of the research area. (a) Administrative boundary of central region. (b) WorldPop population data. (c) POI data.
Figure 3. Experimental data of the research area. (a) Administrative boundary of central region. (b) WorldPop population data. (c) POI data.
Land 13 02154 g003
Land 13 02154 g004
Figure 4. Results of urban functional mix at different grid scales.
Figure 4. Results of urban functional mix at different grid scales.
Land 13 02154 g004
Land 13 02154 g005
Figure 5. Moran’s I results of functional mix at different grid scales.
Figure 5. Moran’s I results of functional mix at different grid scales.
Land 13 02154 g005
Land 13 02154 g006
Figure 6. Hotspot analysis results of functional mix at different grid scales.
Figure 6. Hotspot analysis results of functional mix at different grid scales.
Land 13 02154 g006
Land 13 02154 g007
Figure 7. Population aggregation patterns at different grid scales.
Figure 7. Population aggregation patterns at different grid scales.
Land 13 02154 g007
Land 13 02154 g008
Figure 8. Contribution of functional mix and POI-related variables to population configuration at different grid scales. (a) at 300 m scale; (b) at 500 m scale; (c) at 700 m scale.
Figure 8. Contribution of functional mix and POI-related variables to population configuration at different grid scales. (a) at 300 m scale; (b) at 500 m scale; (c) at 700 m scale.
Land 13 02154 g008
Land 13 02154 g009
Figure 9. MGWR results of different POI types on functional mix at three grid scales. (a) at 300 m scale; (b) at 500 m scale; (c) at 700 m scale.
Figure 9. MGWR results of different POI types on functional mix at three grid scales. (a) at 300 m scale; (b) at 500 m scale; (c) at 700 m scale.
Land 13 02154 g009
Table 1. Comparison of global regression and MGWR results on population for different grid sizes.
Table 1. Comparison of global regression and MGWR results on population for different grid sizes.
StatisticGlobal RegressionMGWR
700 m 500 m 300 m 700 m 500 m 300 m
Model Information
Number of Observations5931163316359311633163
R 2 0.5080.4720.3620.9020.9240.945
Variable Coefficients
Mix Degree0.2470.3230.3390.1780.2320.093
Residence0.2150.2120.1800.1830.0640.038
Restaurant0.0850.1220.1200.0780.1520.064
Shop0.0430.007−0.024−0.016−0.003−0.008
School−0.022−0.026−0.019−0.031−0.012−0.012
Workplace0.2380.1960.1580.1680.1370.106
Amenity0.1140.0490.0210.061−0.0130.022
Healthcare0.0120.014−0.0020.026−0.012−0.009
p-values (Global Regression)
Mix Degree0.0000.0000.000N/AN/AN/A
Residence0.0000.0000.000N/AN/AN/A
Restaurant0.1900.0040.000N/AN/AN/A
Shop0.3580.8320.199N/AN/AN/A
School0.5320.2940.210N/AN/AN/A
Workplace0.0000.0000.000N/AN/AN/A
Amenity0.0570.2180.330N/AN/AN/A
Healthcare0.7600.6060.885N/AN/AN/A
Bandwidth (MGWR)
Mix DegreeN/AN/AN/A44.00043.00046.000
ResidenceN/AN/AN/A43.00043.00044.000
RestaurantN/AN/AN/A44.0001162.0003161.000
ShopN/AN/AN/A592.0001162.0003161.000
SchoolN/AN/AN/A592.0001046.0003161.000
WorkplaceN/AN/AN/A44.00045.00044.000
AmenityN/AN/AN/A592.00045.00044.000
HealthcareN/AN/AN/A233.0001148.0003161.000
Table 2. Parameter coefficient estimates at three grid scales.
Table 2. Parameter coefficient estimates at three grid scales.
VariableScaleMeanSTD
300 m0.0930.132
Mix Degree500 m0.2320.265
700 m0.1780.286
300 m0.0640.002
Restaurant500 m0.1520.002
700 m0.0780.180
300 m−0.0080.001
Shop500 m−0.0030.002
700 m−0.0160.002
300 m−0.0120.002
School500 m−0.0120.009
700 m−0.0310.003
300 m0.1060.193
Workplace500 m0.1370.240
700 m0.1680.253
300 m0.0220.132
Amenity500 m−0.0130.138
700 m0.0610.001
300 m−0.0090.001
Healthcare500 m−0.0120.004
700 m0.0260.054
300 m0.0380.176
Residence500 m0.0640.255
700 m0.1830.316
Table 3. Heterogeneity and maximum impact of important variables by grid scale.
Table 3. Heterogeneity and maximum impact of important variables by grid scale.
VariableGrid ScaleHeterogeneitySalient Impact (Mean > 0.1)
Amenity300 m✓-
500 m✓-
Workplace300 m✓✓
500 m✓✓
700 m✓✓
Mix Degree500 m✓✓
700 m✓✓
Restaurant500 m-✓
700 m✓-
Residence700 m✓✓
Note: ‘Heterogeneity’ refers to the variability in a variable’s influence across different scales. ‘Salient Impact (Mean > 0.1)’ stands for the mean value of a variable coefficient greater than 0.1, representing a notable influence.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Liu, L.; Huang, H.; Yang, J. Enhancing Detailed Planning from Functional Mix Perspective with Spatial Analysis and Multiscale Geographically Weighted Regression: A Case Study in Shanghai Central Region. Land 2024, 13, 2154. https://doi.org/10.3390/land13122154

AMA Style

Liu L, Huang H, Yang J. Enhancing Detailed Planning from Functional Mix Perspective with Spatial Analysis and Multiscale Geographically Weighted Regression: A Case Study in Shanghai Central Region. Land. 2024; 13(12):2154. https://doi.org/10.3390/land13122154

Chicago/Turabian Style

Liu, Liu, Huang Huang, and Jiali Yang. 2024. "Enhancing Detailed Planning from Functional Mix Perspective with Spatial Analysis and Multiscale Geographically Weighted Regression: A Case Study in Shanghai Central Region" Land 13, no. 12: 2154. https://doi.org/10.3390/land13122154

APA Style

Liu, L., Huang, H., & Yang, J. (2024). Enhancing Detailed Planning from Functional Mix Perspective with Spatial Analysis and Multiscale Geographically Weighted Regression: A Case Study in Shanghai Central Region. Land, 13(12), 2154. https://doi.org/10.3390/land13122154

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop