Temporal and Spatial Attractiveness Characteristics of Wuhan Urban Riverside from the Perspective of Traveling

Abstract

Improving the attractiveness of urban waterfronts has become an important objective to promote economic development and improve the environmental quality. However, few studies have focused on the inherent characteristics of urban waterfront attractiveness. In this study, mobile phone signaling data and the TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) were used to construct the attractiveness evaluation system of the riverside in Wuhan. The OLS (ordinary least squares) regression model was used to analyze the relationship between the POI (point of interest) and the attractiveness of river waterfronts. Furthermore, the high-or-low-value aggregation classification of research units was performed according to attractiveness and the POI indicators to reveal the influencing factors of the attractiveness of the Wuhan urban riverside. Results showed the following. (1) The high-value distribution of attractiveness of the river waterfronts in Wuhan presented regional aggregation characteristics, and the attractiveness of economically developed areas was high. (2) Consumer POIs (CPOIs) and outdoor recreation POIs (RPOIs) had a positive effect on the attractiveness of the riverside in Wuhan, while housing POIs (HPOIs), public service POIs (OPOIs), and the high degree of POI mixing had a negative impact on the attractiveness of the urban riverside. (3) The high–high agglomeration spaces were mainly concentrated in the economically developed areas of the city center, which are mainly open spaces where urban cultural activities are held, while the low–low agglomeration spaces were mostly gathered in the suburban areas. The spatial distribution of the high–low agglomeration spaces, which are mainly green open spaces, was relatively fragmented, while the low–high agglomeration spaces, which are mainly freight terminals, linear walks, and residential areas, were near the city center.

1. Introduction

Urban waterfronts are the parts of a city that are adjacent to water bodies (rivers, lakes, oceans, bays, creeks, and so on) [1,2] that carry social, ecological, and economic attributes such as urban ecology [3], culture [4], economy [5], and politics [6]. From the early 18th century to the 19th century, urban waterfronts functioned as ports, warehouses, and factories [1]. With the rapid development of the economy, the attractiveness of waterfronts has gradually been recognized by planners and citizens and used for public purposes [1]. Since the 1980s, urban development has brought many environmental and socio-economic problems to urban waterfronts [7,8,9,10], so waterfront areas have gradually become the focus of urban planning and construction interventions [11]. Since the 21st century, the revitalization of waterfront areas has usually been accompanied by the optimization of urban functions and an important way for cities to improve their competitiveness [12]. In recent years, the research on the renewal of urban waterfronts has become a development trend in cities around the world [13]. Increasing studies are considering the renewal of urban waterfronts as an important approach to boosting the vitality of cities [14,15,16]. The renewal planning of waterfronts is increasingly becoming an important part of sustainable urban development strategies [17].

Although the issue of urban waterfront renewal has been given increasing attention, problems remain such as insufficient facility functions, low environment quality, and insufficient utilization of the waterfront, which in turn lead to low attractiveness [16,18]. Therefore, exploring the attractive characteristics of urban waterfronts and proposing effective methods to enhance them are crucial to the future construction of urban waterfronts. Unlike the vitality measured by density [19], attractiveness is difficult to define and quantify [20]. Biernacka, M. and Kronenberg, J. et al. proposed that a place is attractive when people are willing to spend their time there and it meets their personal needs, expectations, and preferences [21,22]. With the progress of research, scholars gradually began to pay attention to urban spatial attractiveness. In the spatial scale, urban spatial attractiveness can be divided into three types. At the macro-level, studies have concentrated on the economic attractiveness between cities or regions [23,24,25], tourism attractiveness [20,26], and the characteristics of population migration [27,28,29,30], exploring socioeconomic strategies for urban development. At the meso-level, studies have tended to focus on factors influencing attractiveness (e.g., accessibility [21,22,31], facilities and vegetation distribution [32,33,34,35], spatial comfort [36] and biodiversity [37,38]), crowd perception features [35,39], and spatiotemporal distribution features [40,41]. At the micro-level, digital techniques such as eye tracking [42] and 3D simulation [43] are used to explore the dynamic characteristics of the visual attractions of urban streets from the pedestrian perspective [44]. Although studies on attractiveness are extensive, they are mainly on the regional scale and focus more on the attractiveness of urban green open spaces than urban waterfronts.

In terms of research methods, early studies on the attractiveness of urban waterfronts mostly explored people’s preferences and usage of public spaces through questionnaires, interviews, and field research [31,32,33,45,46,47,48]. However, the rise in big data and the development of related technologies in recent years have brought new exploration methods for quantitative research on urban spatial attractiveness [49]. Banet, K. et al. used bicycle travel trajectory data to explore the characteristics of user dwell time, revealing the spatial pattern of urban attractiveness [40]. Cai, L. et al. proposed the concept and model of the hotspot attractiveness indicator to explore the spatiotemporal distribution of hotspots and determine the degree of attractiveness to residents [41]. Huang, H.L. et al. used taxi trajectory data to establish a time-dependent attraction function to represent the attractiveness of a destination to passengers from the perspective of dynamic demand [50]. Among all of the related big data, mobile phone signaling data have the advantages of large coverage and sufficient samples and contain the traveling information of users in different periods [51,52,53,54]. Therefore, mobile phone signaling data can accurately describe the traveling patterns of users and help in the study of the spatiotemporal characteristics of urban spatial attractiveness from the perspective of traveling [55,56].

In general, this study established indicators from the spatial and temporal dimensions to measure the attractiveness of the riverside area in Wuhan, adopting the TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) to calculate the total attractiveness score, then used the OLS (ordinary least squares) model to study the relationship between the POI (point of interest) and attractiveness before finally proposing suggestions for attractiveness improvement and the promotion of urban waterfronts. However, it cannot be neglect that mobile phone signaling data only provide the time and location of users but does not engage with any of the other variables that assist in determining an individual or community’s attractiveness to a location. Therefore, this work only offers an additional method but not as a standalone approach for determining the attractiveness of the urban waterfront. The remainder of this paper is organized as follows. Section 2 presents the materials including an introduction of the study area and data. In Section 3, the methodology is discussed including an introduction of the research framework and mathematical models. Section 4 discusses the results including the attractiveness distribution characteristics, the total attractiveness score, and the regression model and classification results. Section 5 and Section 6 present the discussion and the conclusions, respectively.

2. Materials

Materials are the basis of research. In this part, we introduce the study areas, the data sources, and the pre-processing progress.

2.1. Study Area

Wuhan (113°41′–115°05′ E, 29°58′–31°22′ N) is the capital city of Hubei Province, China [57]. This city is traversed by the Changjiang River and its tributary, the Han River, and is geographically divided into three towns: Wuchang, Hankou, and Hanyang [58]. The Third Ring Road is the dividing line between Wuhan and its suburbs. The areas within the Third Ring Road are highly developed and economically prosperous, while those outside the Third Ring Road are relatively underdeveloped [59]. This study takes the core area within the Third Ring Road of the Changjiang River in Wuhan as the study area. As shown in Figure 1, the study area is the core area of the Changjiang River in Wuhan, China, with a length of 23.74 km. To study the attractiveness characteristics of the riversides, we divided 46 research units according to the road network, with an average area of 0.36 km² and a total area of approximately 16.6 km².

Table 1. The research unit information.

ID	Name	Category	District	ID	Name	Category	District
1	River Landscape Park	Landscape park	Hankou	24	Hongshan Beach Park	Landscape park	Wuchang
2	Hankou River Beach Park Phase IV	Landscape park	Hankou	25	Yangsigang Yangtze River Bridge Park	Landscape park	Wuchang
3	Hankou River Beach Park Phase III	Landscape park	Hankou	26	BAPU Street embankment Beach	Landscape park	Wuchang
4	Hankou River Beach Park Phase III	Landscape park	Hankou	27	Changjiang Zidu residential community	Residential community	Wuchang
5	Fengfan Square, Hankou River Beach Park Phase II	Cultural spot, landscape park	Hankou	28	Under building	/	Wuchang
6	Zhiyin Dock, Hankou River Beach Park Phase II	Cultural spot, landscape park	Hankou	29	Meihuayuan residential community	Residential community	Wuchang
7	Hankou River Beach Park Phase I	Landscape park	Hankou	30	Jiefanglu residential community	Residential community	Wuchang
8	Culture Plaza of Hankou River Beach Park, Hankou River Beach Park Phase I	Landscape park	Hankou	31	Zhonghualu Dock	Transportation hub	Wuchang
9	Wuhan Harbor, Wuhan Science and Technology Museum	Transportation hub, public building	Hankou	32	Dadikou Square, Wuchang Beach Park	Landscape park	Wuchang
10	Wuhanguan Dock	Cultural spot	Hankou	33	Wuchang Beach Park	Landscape park	Wuchang
11	Temple of the Dragon King	Cultural spot, landscape park	Hankou	34	Wuchang Beach Park	Landscape park	Wuchang
12	Nananzui Park	Landscape park	Hanyang	35	Moon Bay Square of Wuchang River Beach Park	Landscape park	Wuchang
13	Qingchuan Pavilion, Dayu Square Park	Cultural spot, landscape park	Hanyang	36	Moon Bay Dock	Transportation hub	Wuchang
14	Chaozong Park	Landscape park	Hanyang	37	Wuchang Beach Park	Landscape park	Wuchang
15	Yingwuzhou Culture Plaza	Cultural spot	Hanyang	38	Wuchang Beach Park	Landscape park	Wuchang
16	Hanyang beach Park	Landscape park	Hanyang	39	Wuchang Beach Park	Landscape park	Wuchang
17	Under building	/	Hanyang	40	Qingshan River Beach Park Phase I	Landscape park	Wuchang
18	Under building	/	Hanyang	41	Qingshan River Beach Park Phase I	Landscape park	Wuchang
19	Under building	/	Hanyang	42	Qingshan River Beach Park Phase I	Landscape park	Wuchang
20	Wuhan International EXPO Center	Public building	Hanyang	43	Qingshan River Beach Park Phase I	Landscape park	Wuchang
21	Yangsigang bridge Beach Park	Landscape park	Hanyang	44	Qingshan River Beach Park Phase II	Landscape park	Wuchang
22	Hongshan Beach Park	Landscape park	Wuchang	45	Jieteng jianjiu Music and Sports Park	Cultural spot, landscape park	Wuchang
23	Hongshan Beach Park	Landscape park	Wuchang	46	Jieteng jianjiu Music and Sports Park	Cultural spot, landscape park	Wuchang

lists the names, categories, and districts of different research units in Figure 1. The table reveals that most of the research units belong to the landscape park category, while the rest are cultural spots, residential areas, public buildings, and transportation hubs.

2.2. Data Sources and Pre-Processing

As for the data, we obtained mobile phone signaling and POI data as the research data sources and then pre-processed them for further research.

2.2.1. Mobile Phone Signaling Data

The mobile phone signaling data used in this study comes from the Smart Steps digital platform (http://www.smartsteps.com/, accessed on 11 August 2022) of China Unicom, which is one of the three major telecommunications operators in China. This study selected the mobile phone signaling data of the riverside area within the Third Ring Road of Wuhan in June 2021 including the average and cumulative values of data on weekdays and weekends such as the user density, user travel distance, user OD data, the number of users arriving in 24 h, and the average dwell time and dwell frequency of the users. As shown in

Table 2. The mobile phone base station information of the Wuhan riverside area.

Length	Num	Length	Num	Length	Num
250 m	73,469	2000 m	4299	16,000 m	135
500 m	64,206	4000 m	2154	32,000 m	37
1000 m	17,177	8000 m	539	96,000 m	4
Total number: 162,020

, the study area had 162,020 mobile phone base stations with different coverage accuracies, and the closer they were to the central city, the higher their accuracy. A total of 73,469 base stations have a coverage of 250 m.

Additionally, we declare that the collection and processing of the data was undertaken by China Unicom from whom we obtained our dataset for this study. In the process of collection, it was not allowed to query user details including user ID and specific location, etc. The final returned data only included the statistical cumulative number of users in a certain period of time or in a certain region. Therefore, in this study, the use of mobile signaling data fully complies with the ethical and legal standards related to user privacy.

2.2.2. POI Data

In this study, the POI data within the 200 m buffer area of the research unit along the river was selected, and 37,101 data items (including location and type of each POI) were obtained. According to the POI characteristics, the data were divided into four categories, namely, consumer POI (CPOI), outdoor recreation POI (RPOI), public service POI (OPOI), and housing POI (HPOI). Among them, CPOI includes catering and shopping services, RPOI includes scenic spots, HPOI includes residential and living services, and OPOI includes public service facilities. Through data preprocessing, a distribution map of the POI points in the study area was obtained (as shown in Figure 2). The figure shows that the POI points are mainly distributed in the riverside area of Hankou, followed by the riverside area of Wuchang.

3. Methodology

This part mainly discusses how to use certain methods for research. The research framework, the calculation method of attractiveness indicators, TOPSIS, and the OLS regression model are discussed.

3.1. Research Framework

This study aimed to explore the spatiotemporal characteristics of attractiveness on the urban riverside area of Wuhan and then design a framework to study attractiveness from the perspective of traveling combined with POI data. As shown in Figure 3, this study used mobile phone signaling data to establish indicators from the time and space dimensions, exploring the spatiotemporal characteristics of the attractiveness distribution. First, the research units were divided by the road network and the Changjiang River, and the mobile phone signaling data were cleaned and processed to obtain information such as the numbers of users on the weekends and working days in each research unit, the origin and destination of users, the number of hourly visit of users, and the average number of users dwelling in each research unit, their dwelling frequency, and stay duration. Then, we established the spatial dimension indicators (i.e., the spatial density, diversity, and distance) and the time dimension indicators (i.e., temporal stability, dwell time, and dwell frequency) to obtain the spatiotemporal distribution characteristics of waterfront attractiveness. After normalizing and standardizing the data, the TOPSIS was used to comprehensively evaluate different indicators, and the total score of waterfront attractiveness was obtained. Finally, using the OLS regression model, the correlation between the POI density, POI mixing degree, and waterfront attractiveness was calculated, and a strategy for improving the attractiveness of the waterfront in Wuhan is proposed.

3.2. Calculation of Attractiveness Indicators

On the basis of the above research framework, this study constructed an attractiveness indicator system based on the time and space dimensions. In existing studies, density, dwell time, dwell frequency, and travel distance are common indicators for measuring attractiveness [33,40,41,48], and the mobile phone signaling data used in this study also contained these variables. However, to study the spatiotemporal characteristics of attractiveness more comprehensively, this study attempted to combine the characteristics of mobile phone signaling data to provide abundant observation indicators of attractiveness. Referring to the research of Liu, Song et al. [60], this study added spatial diversity and time diversity to measure attractiveness. Each indicator is discussed in detail below, and the calculation method is explained in

Table 3. The spatiotemporal attractiveness indicators.

Dimension	Indicator	Calculation Formula	Description
Spatial	Space Density	$Space Density=N_i/S_i$	$N_i$ is the number of users of each research unit, $N_i$ is the area of each research unit.
	Space Distance	$Space Distance=L_i/N_i$	$L_i$ is the sum of user travel distance of each research unit, $N_i$ is the number of users of each research unit.
	Space Diversity	$Space Diversity=N_i$	$N_i$ is the number of origins of users in research unit i.
Temporal	Dwell Time	$Dwell Time=T_i/N_i$	$T_i$ is the sum of the user Dwell Time per study unit, $N_i$ is the number of users of each research unit.
	Dwell Frequency	$Dwell Frequency=F_i/N_i$	$F_i$ is the sum of the user Dwell Time per study unit, $N_i$ is the number of users of each research unit.
	Time Diversity	$Time Diversity={({\displaystyle {\displaystyle \sum }_{j=1}^N}{P}_{i.j}^2)}^{-1}$	$P_{i.j}$ refers to the proportion of the number of users arriving in research unit i to the total number of users arriving in 24 h a day during period j.

.

• Space Density

The space density indicator measures the density of users in a research unit. The larger the value, the larger the ability of a certain unit to gather a larger number of people.

2.

Space Distance

The space distance indicator measures the travel distance of users in the research unit. The larger the value, the stronger the ability of a certain unit to attract people from farther places.

3.

Space Diversity

The space diversity indicator measures the number of user origins in the research unit. The larger the value, the more diverse the regions of origin of the people that can be attracted by a certain unit.

4.

Dwell Time

Dwell time measures the average dwell time of users in the research unit. The larger the value, the greater the ability of a certain unit to attract people into staying for a long time.

5.

Dwell Frequency

Dwell frequency measures the average dwell frequency of users in the research unit. The larger the value, the greater the ability of a certain unit in attracting people to visit several times.

6.

Dwell Diversity

The time diversity indicator measures the diversity of user arrival times in the research unit, and the larger the value, the stronger the ability of a certain unit to attract people at different times of the day.

3.3. TOPSIS

Given the diversity of the spatiotemporal attractiveness indicators, their quantitative evaluation is a typical multi-attribute decision-making problem. Thus, this study adopted the TOPSIS method to comprehensively evaluate attractiveness. On the basis of the similarity of the studied alternatives to the ideal solution, the optimal solution in the TOPSIS model should have the shortest distance from the positive ideal solution and the farthest distance from the negative ideal solution. TOPSIS has been proven to be a reasonable and feasible performance evaluation method [61]. When conducting TOPSIS analysis, determining whether the indicator is a gain (bigger is better) or cost type (smaller is better) is critical [62]. The basic assumptions are outlined as follows. For each study unit, the higher the values of its space density, space distance, space diversity, dwell time, dwell frequency, and time diversity indicators, the more attractive it is. Therefore, all selected indicators were classified as a gain type.

$$E_j=-\mathit{ln}{\left(n\right)}^{-1}{{\displaystyle \sum }}_{i=1}^n{p}_{ij}\mathit{ln}{p}_{ij},w_j=\frac{1-E_j}{k-\sum E_j}\left(j=1, 2, \dots , m\right)$$

(1)

According to the indicators obtained above, a standardized evaluation matrix was established, and the entropy weight method was used to determine the weights of various evaluation indicators. In Formula (1), $E_j$ is the information entropy; k is the number of indicators; and $w_j$ is the weight of each indicator.

$$D_i^{+}=\sqrt{{\displaystyle \sum }_{j=1}^m{w}_j{(Z_j^{+}-Z_{ij})}^2},D_i^{-}=\sqrt{{\displaystyle \sum }_{j=1}^m{w}_j{(Z_j^{-}-Z_{ij})}^2},$$

(2)

$$C_i=\frac{D_i^{-}}{D_i^{+}+D_i^{-}}$$

(3)

Afterward, as shown in Formulas (2) and (3), we calculated the sum of the distance ($D_i^{+}$ and $D_i^{-}$) of the attractiveness indicator of each research unit to the positive ideal solution $Z_j^{+}$ and the negative ideal solution $Z_j^{-}$. Finally, we obtained the relative approach degree $C_i$ of each research unit; the closer the value is to 1, the better the evaluation object. Thus, the standardized value was used in this study as the total score of attractiveness.

3.4. OLS Regression Model

After a comprehensive evaluation of various indicators of attractiveness, this study used the OLS regression model to study the relationship between the total attractiveness score and POI (including POI density and mixing degree of the different POI types). The OLS regression model is a common and effective statistical model for exploring the relationship between variables.

$$y={\beta }_0+{\displaystyle \sum }_{j=1}^m{\beta }_j{x}_j+\epsilon$$

(4)

In Formula (4), $y$ represents the dependent variable; $x_j$ represents the j-th POI indicator; ${\beta }_j$ represents the coefficient of each independent variable; and $\epsilon$ represents the residual.

$$Simpson={({\displaystyle \sum }_{j=1}^N{P}_{i.j}^2)}^{-1}$$

(5)

In addition, for the POI mixing degree, as shown in Formula (5), this study adopted the Simpson value, where $P_{i.j}$ is the ratio of the number of j-type POIs to all of the POIs in each research unit.

4. Results

This section presents the spatiotemporal distribution characteristics of attractiveness, the TOPSIS evaluation result, the OLS regression result, and the type analysis.

4.1. Spatiotemporal Distribution Characteristics of Attractiveness

4.1.1. Space Density

Space density is a measure of user density and represents the ability of a research unit to gather crowds. As shown in Figure 4, the high-density areas in the riverside area of Wuhan were mainly concentrated in the central area of the city and were mostly cultural spots and landscape parks such as the Temple of Dragon King, the Nananzui Park, and the Culture Plaza of the Hankou River Beach Park. The units with low spatial density were mainly concentrated in the urban fringes, with only a small scattered distribution. Overall, the space density during weekends was much lower than that during weekdays, and the relative value of the space density of the landscape park research units on weekends was decreased.

In general, the space density of the riverside area of Wuhan presented a significant centripetal distribution, and many people gathered in the cultural spot and landscape park category research units, but the space density of the landscape park category research units decreased on weekends.

4.1.2. Space Distance

The space distance indicator is a measure of the distance traveled by a user, indicating the ability of a research unit to attract people from distant places. As shown in Figure 5, the research units with a high space distance in the riverside area of Wuhan had regional aggregation characteristics, especially the transportation hub such as the research units where the Qingchuan and Zhonghualu Docks are located, which had the highest space distance value, followed by the research unit where high-quality landscape parks and residential areas such as Hankou River Beach Park Phase IV and Baishazhou Bridge River Beach were located. The space distance distributions during weekends and weekdays were similar, but the space distance of the landscape parks and cultural spots during weekends was higher than that during the weekdays, indicating that people were willing to access the city’s riverside area for activities on the weekends.

In general, the distribution of spatial distance in the waterfront area of Wuhan presented regional aggregation characteristics; waterway transportation hubs had the highest spatial distance value, and people were willing to come to the landscape parks and cultural spots in the riverside area from farther places for activities on the weekends.

4.1.3. Space Diversity

The space diversity indicator presents the number of user origins, indicating the ability of a research unit to attract people from different regions. As shown in Figure 6, the distribution pattern of the units with high spatial diversity in the Wuhan riverside showed regional aggregation characteristics, indicating that the units were mainly distributed in the research units where high-quality landscape parks and residential areas such as the Er’qi Yangtze River Bridge, Fengfan Square, the Culture Plaza of Hankou River Beach Park, and the Temple of the Dragon King are located. Overall, the space diversity during weekdays was higher than that during the weekends, especially the space diversity of the residential category research units, which decreased the most on weekends, indicating that the source of people visiting riverside residential areas during weekdays was relatively more diverse than that during the weekends.

In general, the research units with high space diversity in the riverside of Wuhan showed regional aggregation characteristics in different areas. High-quality landscape parks and residential areas can attract people from different areas, and the source of people visiting the riverside residential areas on the weekdays was relatively more diverse than that during the weekends.

4.1.4. Dwell Time

The dwell time measures the average dwell time of users and represents the ability of a research unit to attract people to stay for a long time. As shown in Figure 7, the units with a high dwell time in the riverside area of Wuhan were mainly marginally distributed, and most were landscape parks such as Hankou River Beach Park Phase IV, the Baishazhou Bridge River Beach Park, and the Qingshan River Beach Park, and the research units where residential areas or public buildings such as the Wuhan International EXPO Center, are located. The distributions of dwell time during the weekdays and weekends were basically the same, and the dwell time during the weekends was slightly higher than during the weekdays.

In general, the research units with a high dwell time in the Wuhan riverside were mostly marginally distributed, and the research units where the landscape parks, residential areas, and public buildings were located had long dwell times.

4.1.5. Dwell Frequency

Dwelling frequency measures the average dwell frequency of users, indicating the ability of a research unit to attract crowds to visit multiple times. As shown in Figure 8, the units with a high dwell frequency in the riverside area of Wuhan were mainly scattered in distribution, and most of them were cultural spots and well-built landscape parks such as the research units where the Temple of Dragon King, the Nananzui Park, the Culture Plaza of Hankou River Beach Park, and the Wuchang River Beach Park are located. In addition, the research unit where the residential area is located also had a high dwell frequency. The areas with low dwell frequency were mainly distributed in the urban fringe area, and a few were scattered. The dwell frequency during the weekdays was generally higher than that during the weekends, and the dwell frequency of some landscape parks was relatively higher during the weekends than that during the weekdays.

In general, the research units in Wuhan riverside with high dwell frequency presented a scattered distribution. Cultural spots and well-built landscape parks can attract people to visit and stay many times.

4.1.6. Time Diversity

The time diversity indicator measures the diversity of the user arrival times and represents the ability of a research unit to attract crowds at different times of the day. As shown in Figure 9, the distribution of units with high time diversity had regional aggregation characteristics. Most of the units were mainly where landscape parks such as Hankou River Beach Park Phase IV, the Baishazhou Bridge River Beach Park, Nananzui Park, and the Temple of Dragon King, are located, public buildings such as the Wuhan International EXPO Center, and residential areas. The spatial distributions of time diversity during the weekdays and weekends were similar, but the time diversity of the research unit where the transportation hub is located was lower during the weekends than the weekdays, and the time diversity of a few research units where landscape parks are located was relatively higher during the weekends than the weekdays.

In general, the distribution pattern of time diversity in the Wuhan riverside showed regional aggregation characteristics. Cultural spots and well-built landscape parks could attract crowds in highly diverse times of the day, while the time diversity of the transportation hubs was relatively lower on the weekends than on the weekdays.

This part analyzes the spatiotemporal attractiveness characteristics of the riverside area of Wuhan from the spatial and temporal dimensions. The specific findings are as follows:

• The difference between the spatial dimensions on the weekends and weekdays was larger than that between the time dimensions.
• The distribution of high values of various indicators was highly varied, presenting four main types: centripetal aggregation, marginal distribution, regional aggregation, and scattered distribution. Among the indicators, space density showed a centripetal distribution, dwell time a marginal distribution, space distance, space diversity, and time diversity a regional aggregation, and dwell frequency a scattered distribution.
• A significant correlation existed between the type of research unit and the high value of the indicator. The research units with a high space density were mostly cultural spots and landscape parks; the research units with a high space distance were mostly transportation hubs, cultural spots, and landscape parks; the research units with high space diversity and dwell frequency were landscape parks and residential areas; and the research units with high dwell time and time diversity were landscape parks, public buildings, and residential area.

4.2. TOPSIS Evaluation Result

4.2.1. Weight Calculation Result

After standardizing the indicators, the entropy weight of the TOPSIS method was used to calculate the spatiotemporal attractiveness indicators of the Wuhan riverside area on the weekends and work days. Then, we obtained information such as the weight and relative approach degree of each indicator. As shown in

Table 4. The weight distribution of the attractiveness indicators of the riverside area of Wuhan.

Dimension	Indicator	Work Day Average	Work Day Standard Deviation	Work Day Weight	Rest Day Average	Rest Day Standard Deviation	Rest Day Weight
Spatial	Space Density Space Diversity	0.069 0.201	0.148 0.156	40.31% 10.21%	0.07 0.159	0.149 0.145	37.23% 10.28%
	Space Distance	0.619	0.298	7.02%	0.5	0.315	10.67%
Temporal	Dwell Time Dwell Frequency	0.551 0.14	0.203 0.217	3.70% 35.61%	0.518 0.145	0.22 0.219	4.26% 32.43%
	Time Diversity	0.756	0.237	3.15%	0.7	0.294	5.13%

, overall, the weight of the spatial indicators was slightly larger than that of the time indicators. Among the indicators, space density had the largest weight (40.31% on weekdays, 37.23% on weekends), followed by dwell frequency (35.61% on weekdays and 32.43% on weekends), and both played a decisive role in the total attractiveness score. In addition, the weights of space density and dwell frequency on working days were slightly higher than those on the weekends.

In general, the indicator of the spatial dimension had a higher weight than that of the time dimension, and the space density and dwell frequency were the most important indicators for determining the total attractiveness score.

4.2.2. Overall Attractiveness Evaluation Result

After calculating the relative approach degree of each research unit according to the TOPSIS model, we normalized it to obtain the total attractiveness score of each research unit as the total attractiveness evaluation result for working days and weekends in the riverside area within the Third Ring Road of Wuhan. As shown in Figure 10, the highly attractive areas of the waterfront showed regional aggregation characteristics. The research units where high-quality landscape parks and cultural spots are located were the most attractive, followed by the research units where urban public buildings and some residential areas are located. The difference in the distribution of attractiveness between weekends and weekdays was small, but the research units where landscape parks such as the Wuchang and Baishazhou Bridge River Beach Parks are located were relatively more attractive on the weekends than on the weekdays. Research units where transport hubs such as the Zhonghualu and Qingchuan Docks are located were relatively less attractive on the weekends than on the weekdays.

In general, the distribution of attractiveness in the riverside area of Wuhan presents regional aggregation characteristics. Most of the areas with high attractiveness were research units where high-quality landscape parks and cultural spots are located, and the research units where few landscape parks are located were relatively more attractive on the weekends, and those where the transport hubs are located were less attractive on the weekends than on the weekdays.

4.3. OLS Regression Result

On the basis of the discussion in Section 4.1 and Section 4.2, to continue studying the relationship between different types of POI and POI mixing degrees and attractiveness, the OLS regression model was used to analyze the attractiveness during the weekends and weekdays and the density of different types of POI and the Simpson value of POI (representing POI mixing degree) were studied. As shown in

Table 5. The OLS regression results.

Variables	Beta (Weekday)	t (Weekday)	Beta (Weekend)	t (Weekend)	VIF
CPOI	1.361	7.466 ***	1.424	7.183 ***	4.811
RPOI	0.291	2.716 **	0.344	2.951 **	1.666
HPOI	−0.758	−5.099 ***	−0.721	−4.457 ***	3.198
OPOI	−0.379	−2.231 **	−0.598	−3.237 **	4.173
Simpson	−0.212	−2.268 **	−0.181	−1.788 *	1.259
Adjusted R2: 0.689			Adjusted R2: 0.632

* Significant at the 0.1 level, ** Significant at the 0.05 level, *** Significant at the 0.001 level.

, the independent variables passed the collinearity test (VIF < 7.5) and the significance test. The adjusted R² indicates that the independent variables had 68.9% and 63.2% explanation levels for attractiveness on the weekdays and weekends.

The beta values in Table 5 show that the CPOI (1.361 on the weekdays and 1.424 on the weekends) had the strongest positive correlation with waterfront attractiveness on the weekdays and weekends, followed by the RPOI (0.291 on the weekdays, 0.344 on the weekends). This proves that consumer POI and outdoor recreation POI have a positive effect on the attractiveness of the Wuhan riverside. However, HPOI (−0.758 on the weekdays, −0.721 on the weekends) and OPOI (−0.379 on the weekdays, −0.598 on the weekends) were negatively correlated with waterfront attractiveness, suggesting that residential-related POI and public service POI have a negative effect on the attractiveness of urban waterfronts. Particularly, we noted a negative correlation between the Simpson value (−0.212 on the weekdays, −0.181 on the weekends) and attractiveness, indicating that areas with a high POI mixing degree in Wuhan riverside areas are more likely to have low attractiveness.

Combining previous conclusions, we can speculate that POI is not the main factor that attracts people to visit urban waterfronts. Certain types of POI may reduce attractiveness, and areas with a high degree of POI mixing may not necessarily have high attractiveness. However, in some waterfront areas, the pleasant open space landscape may be highly appealing to the crowd.

4.4. Type Analysis

To clearly explore the characteristics of waterfront attractiveness from the perspective of POI, we classified 46 research units from three perspectives: attractiveness scores, POI density, and POI mixing degree. Combined with the results of the OLS regression, the POIs were divided into positive classes (CPOI and RPOI) and negative classes (HPOI and OPOI), and the positive (positive POI density) and negative values (negative POI density) of each study unit were calculated. Finally, the four indicators of each research unit, namely, the attractiveness, positive, negative, and Simpson values, were normalized, and their averages were calculated. Then, the ones above the average were assigned to the high score H class, and those below the average were assigned to the low score L class. For example, if a research unit’s attractiveness score was above average, but all the other indicators were below average, the unit was classified as “HLLL”.

According to the number and type of research units in each category, we determined four categories of research units: “HHHH”, “HLLL”, “LHHH”, and “LLLL”. As shown in Figure 11, the types of distribution on the weekdays and weekends were basically the same.

• The “HHHH” category was mainly distributed in the central area of the city such as the unit where the Zhiyinhao Dock is located, which is a famous urban cultural tourism area in Wuhan. This category of area often had strong centrality and attractiveness and belonged to the relatively economically developed region, where the POI variety and quantity were also very high.
• The “LLLL” category was mainly distributed in the urban fringe area and closer to the suburbs than the “LLLH” category. This category belonged to a relatively underdeveloped area with low POI density, POI mix, and attractiveness.
• The “HLLL” category was scattered and mainly located in well-constructed landscape parks in the riverside area. This category’s high attraction was mainly its well-designed and high-quality open spaces, which provide opportunities for the surrounding crowd to commune with nature. People were attracted by the scenery and environment here rather than the variety of POIs.
• In the “LHHH” category, research units were mainly distributed in the central area of the city. This type of area was low in attractiveness, but the density and mix of the POIs were high. The surrounding areas were mostly residential areas with complete infrastructure, and the part along the river is dominated by cargo terminals and linear walks. Thus, the area is unfavorable for people to stay, and the landscape and activities of the place are relatively monotonous, resulting in low attractiveness to the crowd.

Through type analysis, we can obtain an in-depth understanding of the close relationship between the attractiveness and POI factors of the riverside area of Wuhan. Overall, the attractiveness, the POI density, and the POI mixing degree were generally higher for research units near the city center, while the opposite was true for those further away from the city center. However, the spatial quality and public cultural diversity of the research units where landscape parks and cultural spots are located were the decisive factors of their attractiveness, and the density or abundance of POI did not largely determine their attractiveness. Therefore, we believe that for urban riverside areas with low attractiveness, the space quality should be improved first to increase the landscape value or accommodate many public activities, subsequently increasing the consumer and outdoor recreation POIs.

5. Discussion

This study established indicators from the spatial and temporal dimensions to measure the attractiveness of the riverside areas within the Wuhan Third Ring Road and used the TOPSIS method to calculate the total attractiveness indicator. Finally, we used the OLS model to study the relationship between the POI and attractiveness. Our research revealed the following.

• The high-value distribution of attractiveness of the river waterfronts in Wuhan presented regional aggregation characteristics, and the attractiveness of the economically developed areas was high.
• CPOIs and outdoor RPOIs had a positive effect on the attractiveness of the riverside in Wuhan, while HPOIs, OPOIs, and the high degree of POI mixing had a negative impact on the attractiveness of urban riverside.
• The high–high agglomeration spaces were mainly concentrated in the economically developed areas of the city center and were mainly open spaces where urban cultural activities are held, while the low–low agglomeration spaces were mostly gathered in the suburban areas. The spatial distribution of the high–low agglomeration spaces, which are mainly green open spaces, was relatively fragmented, while the low–high clusters, which are mainly freight terminals, linear walks, and residential areas, were near the city center.

Through the above results, we found that common places with greater economic investment in the waterfront were more attractive to the users, but sometimes, the results showed the opposite in different land categories. For landscape parks and cultural spots with low economic investment but high attractiveness, the spatial quality and public cultural diversity were the decisive factors of their attractiveness; for residential areas with high economic invest but low attractiveness, the monotony and low quality of riverside space led to their low attractiveness. In general, we believe that the attractiveness of the riverside area is affected by economic and natural factors, and there are different strategies to enhance the attractiveness for different land types.

6. Conclusions

6.1. Research Innovation

On the basis of the concept of attractiveness and the existing research, this research studied the attractiveness of waterfront areas through quantitative analysis methods and presented the spatiotemporal distribution characteristics of the riverside areas within the Wuhan Third Ring Road from the perspective of traveling. Using mobile phone signaling data, the characteristics of people’s travel in the city’s riverside area were studied at the urban scale. The TOPSIS model was used to construct the attractiveness indicator system, and the OLS regression model was used to explore the relationship between the POI and attractiveness. Finally, the research units were classified and analyzed for a deeper understanding of the Wuhan riverside’s attractiveness.

Our research results have a certain reference value for urban planning and design and to research personnel in designing, constructing, and transforming the riverside area of Wuhan and further enhancing the attractiveness of waterfront open spaces.

6.2. Future Construction Suggestions

This study revealed that the attractiveness of the riverside area of Wuhan presented regional aggregation characteristics, and the attractiveness of the economically developed regions was high. The areas with low attractiveness were generally characterized by insufficient facilities, poor service, and monotonous space design. Thus, this study presents recommendations for enhancing the attractiveness of waterfronts.

First, in future urban master planning, the balanced development of various regions should be given more attention, especially the enhancement of the attractiveness of economically underdeveloped riverside areas.

Second, the construction quality of landscape parks must be improved and the openness of waterfront spaces must be increased to accommodate many urban public cultural activities.

Finally, consumer and outdoor recreation POIs should be increased to enhance the attractiveness of the city’s riverside areas.

6.3. Deficiencies and Limitations

This study also had some limitations. First, the mobile singling data were provided by China Unicom Company, while users of other mobile companies were not included in the study. Second, the mobile phone signaling data only provided the time and location and did not engage in any of the other variables that assisted in determining an individual or community’s attractiveness to a location. It only shows the objective observation of human stay, but does not present a subjective preference of attractiveness. Therefore, this work only offers an additional method for determining the ‘attractiveness’ of a public space, but not as a stand lone approach. Third, although we found that economic investment and natural factors are important influencing factors on attractiveness, we have not drawn any specific conclusions on how these two factors affect the attractiveness and the differences between them.

Therefore, in the future, we hope to add multisource data (such as social media data, wearable device data, Wi-Fi data, traffic data, etc.) that present the individuals’ perception and behaviors to make up for the limitations of signaling data as well as integrate more urban economic and natural indicators to establish a model to explore their relationship with attractiveness. In addition, studying the urban attractiveness for certain groups of people from different aspects (such as age, gender, occupation, etc.) will also be beneficial.