Research on the Spatial Structure of Xinjiang Port Cities Based on Multi-Source Geographic Big Data—A Case of Central Kashi City

Abstract

Exploring urban spatial structure plays an important role in promoting urban development, but there is a lack of research on the urban spatial structure of Xinjiang ports. This paper takes the central urban area of Kashi City as the study area and integrates points of interest (POI) data with nighttime light (NTL) data using the Open Street Map (OSM) road network to perform kernel density analysis, two-factor combination mapping, and partition identification. It identifies the spatial structural characteristics of the central urban area and divides it into different functional subdivisions. This research shows that ① the overall distributions of nighttime luminance values and POI kernel density are similar, and the overall distribution pattern gradually weakens from the city centre to the surrounding area. High-value areas are distributed in groups, presenting the spatial structure characteristics of one main area and two subareas. ② The fusion of POI data with OSM road network data identifies urban single functional zones and mixed functional zones and divides different functional zones in a more detailed way, with higher accuracy in identifying functional zones. ③ The coupling of POI and nighttime light remote sensing can better characterise the spatial features of the urban structure, such as large-scale homogeneous areas, urban fringe areas, suburbs and township centres, etc. The fusion of POI and the OSM road network can better characterise single and mixed land use types of urban land use and improve the part of the city that cannot be characterised by POI and night light. The results of this study are conducive to the realisation of rational and functional zoning in Kashi City and provide a reference for promoting urban human–land coordination and sustainable development.

1. Introduction

Urban spatial structure refers to the distribution and combination of different functional areas formed in the process of urban development because of various functions and their different forms of material expression [1]. The study of urban spatial structure provides a comprehensive understanding of the layout and characteristics of a city’s current functional zoning, so as to better grasp the operating conditions of the entire city. It also, through analysis, can predict the possible changes and adjustment direction in the future functional areas of the city to provide a scientific basis for urban planning and avoid the problems caused by over-concentration or dispersal of functional areas [2]. Therefore, it is of great significance to further explore urban spatial structure and achieve a rational layout of urban functional areas, which will promote the coordination and sustainable development of the urban human land area.

Points of interest (POI) and nighttime light (NTL) data are the main emerging spatial data sources for the study of urban spatial structure, which have rich data volume, fast updating speeds, and lower acquisition costs compared with traditional data sources [3]. POI data are location data with specific spatial significance, usually including names, addresses, categories, coordinates, and other information that have been applied in the fields of urban industrial spatial layout [4,5], the identification of urban functional zones [6,7], built-up area boundary discrimination [8], and urban mixed-use land use research [9]. NTL data can reflect NTL brightness changes on the surface and analyse the spatial distribution characteristics of regional economic and human activities. It has been widely used in the study of urban agglomeration expansion [10,11], the spatial structure of the urban system [12,13], urban population estimation [14], and urban land expansion and activity change [15].

Meanwhile, some scholars have also carried out research on the fusion application of multiple data, but a discussion about the spatial coupling relationship of multiple geographic big data sources is still relatively lacking. In the current research, the study of urban spatial structure with the help of multi-source geographic big data is mainly explored in qualitative and quantitative dimensions. Yao Yao [16] et al. analysed the distribution of urban land use types in the Pearl River Delta by combining POI with deep learning models. Du Shouji [17] and others fused remote sensing imagery and open-society data to identify large-scale urban functional areas in Beijing. Cai Jixuan [18] et al. combined nighttime light data and social media check-in data to identify polycentric city structures. Li [19] et al. analysed the spatial structural characteristics of Dongguan based on big data about taxi trips. Yu Bingchen [20] and others studied the urban spatial structure of Sanya City from the perspective of the South China Sea harbour. Wang Yuqian [21] et al. explored the urban spatial structure of Beijing using the following types of data: POI, NTL, and microblogging check-ins. Chen Bin [22] and others explored the urban spatial structure of the main urban area of Wuhan using POI and NTL data. Liang Lifeng [23] et al. used POI, NTL, and mobile phone positioning data to study the urban spatial structure of the main urban area of Dongguan City. Chi Jiao [24] et al. identified urban functional areas within the fifth ring road of Beijing by performing kernel density clustering analysis on POI data. Kang Yuhao [25] et al. conducted a density analysis of POI data for functional area identification in Wuhan. Wang Junjue [26] et al. used kernel density fusion to analyse the OSM road network and POI data for a functional zoning study in Shanghai. Ding Yanwen [27] et al. used kernel density estimation algorithm fusion with the OSM road network and POI data for functional area identification in the Nanjing City centre. In summary, in the current research, the study of urban spatial structure with the help of multi-source geographic big data has mainly been explored from qualitative and quantitative dimensions, and the combination of the two types of analyses is rare. Compared with the current body of literature, this paper combines qualitative and quantitative analyses to identify the comprehensive spatial structure of the city of Kashi and classify different functional zones. Combining the two methods can help us better understand the spatial and functional structure of a complex city, aid urban planners to better site and plan the different functional zones of a city based on POI, and enrich the existing literature on the study of urban spatial structure.

As one of the core cities of Xinjiang, Kashi is the gateway and hub of China’s opening to Central and West Asia, as well as an important pivot in the core area of the “One Belt, One Road” economic zone, with clear advantages in terms of location. With the promotion and implementation of national policies such as the “China–Pakistan Economic Corridor” and the “One Belt, One Road” initiative, the development of Kashi City has ushered in unprecedented opportunities for development. Its internal spatial structure is also closely linked to the development of the Silk Road Economic Belt [28]. Therefore, this study selected the central city of Kashi as the study area and investigated the structural characteristics of its internal spatial elements from both qualitative and quantitative aspects, which resulted in a functional area division. This not only has reference value for the future urban planning and development of Kashi City but is also significant for the coordination and sustainable development of Xinjiang’s port cities in terms of space–human–land coordination and sustainable development.

2. Study Area and Data Sources

2.1. Overview of the Study Area

Kashi is located in the southwestern part of China’s Xinjiang Uygur Autonomous Region and the western edge of the Tarim Basin, with a geographic location between 73°20′ and 79°57′ east longitude and 35°20′ and 40°18′ north latitude. With the Pamir Plateau to the west, the Tianshan Mountains to the north, and the Taklamakan Desert to the east, the terrain is undulating and relatively flat, and the climate is hot and dry, with a continental arid climate. Kashi City is a regional economic centre of gravity, trade and logistics centre and cultural centre facing Central, West and South Asia, as well as an important trade town on the ancient Silk Road and an important window for opening up to the West in China. The development of Kashi City has a significant role in promoting the construction of the “China–Pakistan Economic Corridor” and “One Belt, One Road”, so this paper takes Kashi City as the study area. The study area was chosen as the central city of Kashi because of the low distribution of NTL data and POI data in the townships under Kashi (Figure 1).

2.2. Data Sources

(1)

POI data. POI are special places or interesting locations marked on a map. They usually include a variety of commercial, cultural, recreational, and social service locations. The POI data used in this paper comes from the API interface provided by the 2022 Kashi Gaode Map, which provides a total of 14 major categories of POI data in Kashi, totalling 14,751 pieces of data, specifically including food and beverage services, shopping services, science, education and cultural services, scenic spots, public facilities, companies and enterprises, transport facility services, business and residential, living services, sports and leisure services, healthcare services, government organisations and social groups, and accommodation services. In order to facilitate the spatial coupling analysis with the nighttime lighting data and OSM road network data at a later stage, the toponymic addresses, ATMs, public toilets, entrances, and exits, which are not very relevant to this study, were deleted to obtain the required study data.

(2)

Nighttime lighting data. NTL data refer to the collection and recording of data on the quantity, brightness, and distribution of nighttime lighting in a city or region, which is important for comparing the level of development of different regions, monitoring urban sprawl, and improving urban planning. The remote sensing data source for nighttime lighting in this paper is the NPP—VIIRS “Nighttime Lights Data set” published by the National Earth System Science Data Centre (NESDC) at a “global resolution of 500 m”, which was the global NPP-VIIRS long time series night light data obtained by scholars Chen Zuoqi et al. [29] based on cross-sensor calibration, with a spatial resolution of about 500 m.

(3)

OSM road network data. The OSM data in this article was sourced from the official Open Street Map website and covers geographic information data on roads and transport networks around the world. OSM roads have very high positioning accuracy and topological relationships and contain basic spatial information such as latitude and longitude as well as attribute information such as road name, road type, maximum travelling speed, and one-way streets. The irregular grid constituted by the road network is the basic unit to undertake the socio-economic functions of urban management and urban planning, and the final required research unit is generated by extending, deleting, topology checking, and conducting other operations on the 0SM road network data and topologising the processed road network data into surfaces. Detailed data sources are shown in

Table 1. Data resource.

Data Name	Resource
POI	Gaode Open Platform (https://lbs.amap.com/tools/picker, accessed on 25 December 2023)
NPP/VIIRS NTL data	National Earth System Science Data Centre—Yangtze River Delta Sub-centre (http://geodata.nnu.edu.cn/, accessed on 27 December 2023)
OSM	Open Street Map (https://www.openstreetmap.org, accessed on 10 January 2024)

.

3. Research Methodology

The experiment mainly uses ArcGIS 10.7 software for analysis, main research methods include kernel density analysis, two-factor mapping, and partition identification. Firstly, POI data, NTL data, and OSM road network data were pre-processed. Then, a Tyson polygon grid was created based on the administrative boundaries, grid analysis was performed on the POI data and the nightlight data, and the spatial coupling of the nightlight data and the data from the results of the POI kernel density analysis was visualised using a two-factor mapping method. Finally, using the smallest unit divided by OSM, different functional zones in the central city of Kashi were delineated by quantitatively identifying different types of POI data.

3.1. Kernel Density Analysis

Kernel density analysis is a statistical method for estimating the density of the distribution of data points in space, characterised by the ability to produce high-quality density estimates for point data that are unaffected by raster size or location. Kernel density analysis is a local map algebra operation that creates network image elements from points, defines a neighbourhood of a specific shape and size (e.g., a region such as a circle, ring, or rectangle) around a specific raster location point, and then counts and calculates the number of all point elements within this neighbourhood. By analysing the distribution of the number of these points, and based on the neighbourhood’s area estimates, the average density of the point elements in the region is determined, which is then used to identify the geographical distribution of hotspots by analysing the frequency of each location [30]. This approach has been widely used in the field of geographic research, including the psychology of crime in geography [31], the spatial distribution of industries [32], and the distribution of landscape spatial features [33].

$$\mathrm{f}(\mathrm{S})={\displaystyle \frac{1}{R}}\sum _{i=1}^{n}K{\displaystyle \frac{S-S_i}{R}}$$

(1)

where f(S) is the estimated probability density at point S; K denotes the kernel function; n denotes the number of sample points; $S-S_i$ denotes the distance from a certain location point S to a known point $S_i$; and R denotes the bandwidth. The larger the R value, the better it can represent the macro spatial variation in the region and, vice versa, it can only represent the micro spatial variation [34]. In this paper, considering the coupling with night lighting data, after several experiments (500, 1000, 1200 m), 1000 m was selected as the search radius for POI kernel density analysis and 200 m was selected as the spatial image element. Finally, the results of kernel density analysis of POI data were obtained.

3.2. Raster Data Grid Analysis

Data grid analysis is a method of dividing geographic space into regular grid cells and performing statistics and analyses on each cell. Data gridding can unify a variety of different raster data standards, which helps to compare and analyse different types of data. There are various types of data meshing methods to choose from, but the ortho-hexagonal shape has a closer resemblance to a circle than other shapes, as well as richer topological relationships [35]. In this study, the most commonly used Tyson polygonal grid was used for raster data meshing, and 1178 Tyson polygonal grids with an area of 750 × 500 were established in the central city of Kashi as the scope. Then, the regular grid and the processed night light remote sensing data and POI kernel density analysis results were superimposed to obtain the average values of the inner night light remote sensing intensity and POI kernel density of each grid and link the two averages to the same regular grid data, which was convenient for the subsequent analysis of the coupling relationship.

3.3. Two-Factor Mapping

Two-factor combination mapping is a combination of two different influencing factors with high and low values, which can lead to a variety of combination situations, while intermediate transitions can be formed among different combinations to produce more combination forms. To better visualise these combinations, different hues can be selected and combined to form a colour scheme. The advantage of this method is that the gradients among different combinations can be clearly shown with a certain degree of differentiation. This paper adopted a 3 × 3 hierarchy, which contains a total of nine different combinatorial relationships between the two factors. Based on the quantitative and spatial distributions of the different relationship types, we explored the spatial coupling relationship between the night light remote sensing and POI data, and further analysed the spatial differences and the relationship among spatial combinations within the central urban area.

3.4. Delineation of Land Use Units Based on OSM

Firstly, motorways, national highways, trunk roads, primary roads, and secondary roads were screened from the OSM road network to form an irregular grid that served as the natural boundary of the urban area. The road network was then topologically treated, including extending roads by 100 m to deal with the unconnected road network and pruning overhanging roads and separate sections of the network. According to the “Urban Road Engineering Design Code” and combined with the actual situation in the urban area of Kashi City, the road was divided into four levels, of which the highway and national highway was the first level, the first-level road was the second level, the second-level road was the third level, and the third-level road was the fourth level. Corresponding 40 m, 20 m, 10 m, and 5 m buffers were generated for different classes of roads to create road space [36]. Finally, the road space was removed from the study area, separate parcel units were generated, and smaller parcel units were deleted, resulting in the final land use units for the study area.

3.5. Identification of Urban Functional Areas

Based on the number of each POI within each land use unit, the frequency density and category ratio of each type of POI were calculated to determine the functional nature of the land use unit and thus, the division of functional areas [37]. The formula is:

$$F_i={{\displaystyle \frac{{\mathrm{n}}_{\mathrm{i}}}{N}}}_{\mathrm{i}}(\mathrm{i}=1,2,\cdots ,6)$$

(2)

$$C_{\mathrm{i}}={\displaystyle \frac{F_{\mathrm{i}}}{{\displaystyle \sum _{\mathrm{i}=1}^6}{F}_{\mathrm{i}}}}\times 100\%, \mathrm{i}=1,2,\cdots ,6$$

(3)

In Equation (2), $\mathrm{i}$ denotes the POI type; ${\mathrm{n}}_{\mathrm{i}}$ denotes the number of type i POI in the cell; $N_i$ denotes the total number of POIs of type i; and $F_i$ denotes the frequency density. In Equation (3), $C_i$ denotes the ratio of the frequency density of the type i POI to the frequency density of all types of POI in the cell.

Urban functional structures are complex and diverse, including single-function areas and areas with a mixture of multiple functions [38]. The functional nature of a delineation cell is judged based on the category proportions of different POI within the cell as follows: if the category proportion of a certain POI is more than 50%, the cell is judged as a single-function area; if the category proportions of two or more POI are between 20% and 50%, the cell is judged as a mixed area, with the specific mixing type depending on the two types that account for the highest percentage of POI within the cell; and if all POI have a category proportion of 0, then they are judged as no-data areas [39].

4. Analysis of Spatial Integration Results

4.1. Analysis of NTL and POI Integration

A two-factor mapping method was used to visualise the spatial coupling relationship between POI and nighttime lighting data (Figure 2). Symbol counting statistics were performed on both data sources to analyse the same (low–low, medium–medium, high–high) areas and dissimilar (low–middle, low–high, medium–low, medium–high, high–low, high–medium) regions to identify the spatial structure of the central city of Kashi.

4.1.1. Overall Distribution Characteristics of NTL and POI Values

The overall distribution patterns of NTL and POI density values were similar (Figure 3), with both gradually weakening from the city centre to the peripheral fringe areas. High NTL values were concentrated in the core area of the old city and medium values were in the east and north of the city in a small area. This exhibited a spatial structure with one primary and two secondary characteristics.

The spatial distribution of NTL values at night exhibited a pattern of “high in the centre and low all around”. The high-value zone was located in the old city of Kashi, which is the centre of politics, economy, culture, and education, as well as business services. Development units in the new eastern part of the city had light intensity values that were second only to the main core area and were small centres with a new concentration of schools, commercial housing, and services. The other small centre area was an economic development zone in the north of the city, which is the concentrated distribution place of industrial parks and trade and logistics parks, as well as the location of the Kashi Free-Trade Zone and Comprehensive Bonded Zone. The two small centres and the main core area of the old city had a spatial structure of one main area and two secondary areas.

The overall distribution of POI density values also gradually weakened from the city centre to the peripheral areas. High-value clusters were concentrated in the core area of the old city where the commercial and service centres of Kashi are distributed. POI density values for the east and north city districts were much lower than that of the old city centre. They were coupled with the distribution of economic and technological development zones, industrial parks and trade and logistics parks, which were farther away from the main urban area but had small concentrated and contiguous distributions. This was because there was less distribution of retailing, commercial services, and other industries in these development units. There was also a single type of land use and less foot traffic, which resulted in low POI density. In the southern part of the Old Town, a small distribution of point-like high-value zones occurred that was coupled with the distribution of the Central Asia Trade City and the Shuguang International Logistics Centre. POI density was lower in the peripheral areas outside the central city, where there were many villages and towns, low urban land use development, no industrial, commercial, or service centres, and a concentration of cultivated land, forest land, reservoirs, and ecological reserves.

4.1.2. The POI and NTL Coupling Relationship Is the Same

NTL and POI data were similar in the same area of spatial coupling (see Figure 4). High–high areas were mainly distributed in the core area of the old city, with the old city of Kashi, the pedestrian street shopping centre around the border, and the Tangcheng International Commercial Centre forming the centre of the central distribution. Medium–medium areas were mainly distributed in the periphery of the old city in the form of a ring, located around the passenger station and dense residential areas. Low–low areas were distributed at the edge of the central city. They were primarily farmland, forests, waters, and ecological protection areas located in the townships and streets under the city. Low–low areas were distributed around the edge of the central city. They were primarily arable land, forest land, waters, and ecological protection zones, located in the townships and streets under the city of Kashi that were not suitable for urban use development or construction. Overall, the POI data were spatially coupled to the NTL data in the same region with a high degree of coupling.

4.1.3. POI Is Lower Than NTL

In addition to the areas where the coupling was the same, there were also some areas with different couplings. These were small in size but equally important for analysing the differences between the two different data sources and interpreting the spatial characteristics of the city. In order to explore these regions in a more detailed way, this paper extracted these spatially coupled regions individually for mapping. Figure 5 shows the distribution map of the regions where NTL values are higher than the POI kernel density values.

The main types of uncoupled relationships for nighttime lighting above the POI kernel density values included low–medium, low–high, and medium–high areas. Low–medium areas were concentrated in the urban fringe, while low–high areas were mainly distributed in the urban periphery in a semi-ring shape. This is the location of Kashi International Airport and the newly developed construction area in Dongcheng, which corresponds to sample points (1) and (2) in Figure 5, respectively. Medium–high areas were mainly located in the main urban area, in Kashi Old Town, Tangcheng International Commercial Pedestrian Street, and Lily Court Commercial Street in Dongcheng, corresponding to sample points (3) and (4) in Figure 5. These areas are the demonstration areas of the nighttime economy that Kashi City has been building, and they are places where the citizens of Kashi City gather for consumption in the evening. They also belong to the high-value areas where nighttime lights are centrally distributed. These areas are economically active, and the construction of nighttime lighting infrastructure is also relatively complete. They are all large, homogenous areas with clear nighttime lighting characteristics.

NTL could characterise the intensity of urban lights at night, but the specific spatial characteristics of the area were difficult to characterise. POI data, expressed as point entities, were also limited in their ability to characterise larger homogeneous areas such as economic development zones, airports, and the east city district New Construction Area, which were also expressed with a small amount of point data. The spatial characteristics of the larger homogeneous region were better characterised by a coupled dissimilarity analysis of the NTL intensity and POI kernel density values.

4.1.4. POI Is Higher Than NTL

The distribution of regions with POI densities that were greater than NTL is shown in Figure 6, including medium–low, high–low, and high–medium coupling relationships. The distributions of these areas were very small, mainly scattered in the old city. This was related to the level of economic development of the city. The old city had a higher level of economic development and urbanisation compared with the new city, which had more commercial centre distribution. Sample point (5) in Figure 6 is a high–medium coupling area located in the New World Department Store shopping centre on West Field Avenue and the surrounding residential areas. Sample point (6) is a medium–low coupling area located in the old city at the articulation transition with the east city and the north city. It is concentrated in the centre of Kashi Shuguang Agricultural Expo City, Kashi Yazhong Mechanical and Electrical Markets, Central Asia Trade City, and Shuguang International Building Material Market near the South Road of Century Avenue. Compared with the new city, the old city of Kashi is the most concentrated place for business activities and human flow. The POI density values were higher than the NTL values.

4.2. Analysis of POI and OSM Road Network Convergence

Based on the smallest unit of the OSM road network delineation, the ratio value of land use types within the unit was calculated to identify single-function and mixed-function zones in the central city. In order to better interpret and compare the specific types of functional zones, this paper compared the identification results of the functional zones of the site with the Golde map.

4.2.1. Single Functional Area Identification Analysis

The single function areas were identified, and a total of six types of sites were identified, including commercial service facility sites, residential sites, and public administration and public service facilities (Figure 7). The land use functional area for commercial service facilities was mainly scattered in the Old Town, with a small distribution in the Eastern Town, mainly located in the vicinity of Lily Court Commercial Street. The residential land use function area was widely distributed, mostly concentrated and contiguous along both sides of the road. The functional area of land for public administration and public service facilities was mainly located in the east city district, which was related to the fact that this district has built many new schools, hospitals, stadiums, and other scientific, educational, and cultural facilities in recent years. Green space and plaza sites were mainly located in the city centre and east city. Green squares in the city center were concentrated in People’s Park, Kashgar Ancient City and East Lake, the green squares in east city were mainly distributed in Dongcheng Park and XiaoYarang Wetland Park. Industrial land was mostly located in the periphery of the city, mainly in the industrial park near Kashi Airport, Shenzhen Industrial Park, and the Comprehensive Bonded Zone, and the International Automobile City to the south of the central city. Land for roads and transport facilities was mainly located at the airport and railway station sites in Kashi.

A count of the number of units in a single functional area showed that the highest number was for residential land use, with 109 units. Public land was generally spread over a large area and was compactly distributed along roads. The road network was divided into the smallest units so that there was less mixing of public land with other land. This was followed by the Public Administration and Public Service Facilities (PAPS) land unit with 95 units. Residential land was widely dispersed and mixed with other land types and was numerous and scattered. The number of functional area units of green space and square facilities land was the lowest, at only 40, which indicated that green spaces and squares in the urban area of Kashi City were relatively lacking in urban planning and construction.

In this paper, based on the Gaode map, the following typical areas were selected to verify the functional area identification results. Areas A, B, and C were identified as public service land and residential land (Figure 8a), which was consistent with the results of the Gaode map and the Kashi City planning map (Figure 8b). A was the University of Kashi, which was the public service land, and B and C were residential areas, which showed that the identification results of this paper’s method were consistent with the actual situation.

4.2.2. Mixed Functional Area Identification Analysis

A total of 10 mixed functional areas were identified, as shown in Figure 9. As can be seen in Figure 9, the mixed functional areas were mainly concentrated within the Old Town, with a small number in the East Town, but there was a lack of mixed types of green space plaza uses with other functional areas. Among them, the mixed -function zones distributed near the scenic parts of Kashi Ancient City were mostly “commercial–public” mixed zones because there were many restaurants, department stores, hotels and lodgings, photography companies, and pharmacies, clinics, and hospitals in the vicinity, so the degree of commercial–public mixing was relatively high. The two shopping malls, New World Department Store and Ming Sheng International, were mostly in the vicinity of “commercial–transportation” mixed zones. This indicated that there were many car parks and other transportation facilities near the commercial zones, and the degree of transportation access was relatively good. The mixed functional area on the East Side was mainly distributed with a mix of residential and other land uses. The East Side is a new development unit that was developed and built with a large number of residential areas and ancillary shops and restaurants; therefore, it was mostly a mixed area of residential and commercial facility land uses. Technical statistics on the number of units in each type of mixed-use area showed that the “residential–public” mixed-use areas, with a maximum of 33 units, were the main type of mixed-use land in the region. This was followed by “residential–commercial” and “commercial–public” sites, with 31 and 25 units. “Industrial–transport” and “commercial–industrial” sites were the least numerous, with only 12 and 6 units, respectively. The statistics showed that the land for commercial and service facilities in the urban centre of Kashi City was mostly mixed with residential land and land for public service facilities, and industrial land was mostly mixed with residential land.

In this paper, areas D, E, and F were identified as Mixed Residential and Public Facility Land Use, Mixed Commercial Facility Land Use and Public Facility Land Use, and Mixed Residential and Industrial Facility Land Use, respectively. When compared with the Gaode map (Figure 10b), it can be seen that the Hong sheng yuan residential area and Kashi 19th Middle School were in area D, which was residential–public mixed land use. There were residential areas, such as the Old Thermal Power Plant and Kashi Huijin Fruit and Kashi Cotton Development Co., Ltd. (Kashi, China), in area E, which was residential–industrial mixed land use.

Area F was divided into Hengchang commercial street, Sunshine Primary School, and other residential areas, as well as the periphery of some shops, hotels, etc. The distribution of functional areas was more complex. This paper identified area F as a mixed commercial–public area. Compared with the Gaode map (Figure 10b), it can be seen that in addition to the distribution of residential districts in area F, there were also commercial service facilities such as food and beverage outlets, hotels, shopping malls and banks, and schools. Because public facility land was more widely recognised than residential land, it was more reasonable to identify area F as a mixed commercial–public area. This showed that the method used in this paper was more detailed than the planning map because it used the road network to divide the units, and it better identified the mixed functional areas.

5. Discussion

In this study, POI data, night light remote sensing data, and OSM road network data were used to characterise the urban spatial structure of Kashi, a typical inland port city, and to delineate its functional zones. The results of this study showed that both POI fused with night light data and POI fused with OSM road network data could better characterise the spatial structure of the central urban area and identify different functional districts. By integrating POI with luminous remote sensing, the areas where the two differed better characterised the spatial features of the urban structure, such as large-scale homogeneous areas, urban fringe areas, economic development zones, township centres, etc. [20]. The integration of POI with the OSM road network better characterised the single and mixed land use types of urban land use, refining the part of POI that could not be characterised by night lighting [40]. Compared with the results of traditional research, this paper combined qualitative and quantitative analyses to identify the comprehensive spatial structure of the City of Kashi and delineate the functional zoning, which addresses the research gaps in the existing research literature.

There are still some deficiencies in this study, which need to be continuously improved. Firstly, in terms of data application, POI data were used to identify functional areas in the central city by fusing them with nighttime lighting data and the OSM road network. The spatial fusion of the three types of data was limited to the urban spatial structure existing in Kashi and failed to consider the functional structure of the peripheral areas of the city. In the future, high-resolution remote sensing images and other geographic big data can be added to improve the accuracy of functional area identification. Secondly, the choice of the year in which the data were obtained may have impacted the results of this study, especially because Kashi has been developing very rapidly in recent years. Only one year was examined in this study, which is relatively short for investigating the “evolution” of the spatial structure of a city. In the future, multitemporal and multivariate data will be obtained for integration and comparative analyses in subsequent studies.

Kashi, as a key central city for the future development of Xinjiang, brings together multiple advantages such as policy inclination, economic vitality, transport hubs, and population concentration [41], the concentration of which will promote the rapid development of each functional area. In the face of the expansion of basic industry and the growth of commercial and residential demand, the urban infrastructure and transportation network will continue to improve, and various elements will be concentrated in the central urban area, which requires more intensive and effective use of urban land [42]. In addition, the policy support and guidance of urban development also affects the urban spatial structure, such as the economic and technological development zone in the North City and the new construction zone in the East City, which promotes the rationalisation and differentiated development of the functional layout of the two emerging regions through scientific planning guidance. This not only ensures the complementary symbiosis of the functional zones but also promotes the optimisation and upgrading of the urban spatial structure. In the rapid expansion of the city, with the development and construction of more different functional areas, it is important to take advantage of the complementary strengths of the functional areas to improve the efficiency of urban land use and to achieve coordinated and sustainable development of the city in terms of people and land.

6. Conclusions

In this paper, by analysing the spatial integration relationship between POI data and NTL and OSM road network data in the central urban area of Kashi City, we explored the spatial distribution characteristics of the central urban area. We discussed the spatial distribution of the different areas and their relationship with the spatial structure of the city and delineated different functional sub-districts. The following key findings were drawn:

(1)

The spatial coupling relationship between POI and NTL data was good and had high consistency. The overall spatial distributions of the two types of data in the central urban area of Kashi were consistent, and the percentage of areas with the same spatial coupling relationship was 75.47%. Both data types better characterise the spatial structure of the central urban area with one main body and two sub-branches.

(2)

The spatial distribution characteristics of POI and night lighting data were somewhat different. POI data were less distributed in economic development zones, new towns, airports, and other areas, while they were distributed in places with concentrated commercial and human flows. NTL data were more intense in areas with better road infrastructure, such as urban core areas, economic development zones, airports, etc., and had a significant “spillover” effect.

(3)

The fusion of POI data with OSM road network data enabled the quantitative identification of urban single-function and mixed-function areas. Compared with the urban planning map for a single type of functional area, the functional area identification results of central Kashi City were more accurate and better identified mixed land use and integrated land use types. They delineated the different land use units, and the identification of the spatial structure characteristics of the central city was more detailed.

In summary, this study identified the spatial structure of Kashi City using POI data fused with nighttime lighting data and OSM road network data and divided different land use units by combining qualitative and quantitative analyses. The combination of the two methods can help us understand the spatial functional structure of a complex city, assist urban planners in planning for different urban functional zones based on POI, and also provide important decision-making support for specific practices, like adjusting the urban spatial structure and other land use siting.