International Journal of Geo-Information

Article The Socio-Spatial Distribution of Leisure Venues: A Case Study of Bars in Nanjing,

Can Cui 1,2, Jiechen Wang 1,3,*, Zhongjie Wu 1, Jianhua Ni 1 and Tianlu Qian 1

1 Department of Geographic Information Science, Nanjing University, Nanjing 210023, China; [email protected] (C.C.); [email protected] (Z.W.); [email protected] (J.N.); [email protected] (T.Q.) 2 Graduate School of Design, Harvard University, Cambridge, MA 02138, USA 3 Jiangsu Center for Collaborative Innovation in Geographical Information Resource Development and Application, Nanjing 210023, China * Correspondence: [email protected]; Tel.: +86-25-8968-0669

Academic Editors: Mahmoud R. Delavar and Wolfgang Kainz Received: 13 June 2016; Accepted: 19 August 2016; Published: 25 August 2016

Abstract: With the development of service industry and cultural industry, urban leisure and entertainment services have become an important symbol of the city and the driving force of economic and social development. Karaoke, a typical form of urban entertainment, is immensely popular throughout China, and the number of karaoke bars is expected to keep growing in the future. However, little is known about their spatial distribution in the urban space and their association with other location-specific factors. Based on the geospatial entity data and business statistics data, we demonstrate a clustered pattern of 530 karaoke bars in Nanjing by means of point pattern analysis and cluster analysis in GIS. Furthermore, we identify the distribution of population, transportation network, and commercial centers as the three determinants underlying the formation of the pattern.

Keywords: socio-spatial distribution; karaoke bar; leisure venue; China

1. Introduction Leisure industry was listed as one of the “Big Five” engines of economic growth [1]. “Big Entertainment”—everything from bars to video stores to opera houses—are at the forefront of the leisure industry. It has been witnessed that leisure and entertainment sector accounts for an increasing share of gross domestic product in China as well as worldwide, leading to an optimized economic structure. It also becomes a large provider of employment [2] and plays a crucial role in meeting residents’ recreational demand. Among residents’ recreation activities, singing is the most common and popular one. Karaoke bars are the commercial venue providing services for singing. Karaoke, originating from [3], is a form of interactive entertainment in which an amateur singer sings along with recorded music (typically popular song minus the lead vocal) using a microphone. Karaoke soon spread to the rest of Asia and other countries all over the world [4]. Instead of placing karaoke machines in restaurants and hotel rooms in the early times, karaoke bars, with karaoke equipment installed in compartmented rooms and combined with food and other recreation service, emerged as a new business in the market and became enormously popular. The first karaoke bar in Mainland China was opened in 1990 in Hainan Province. Since then, karaoke has become one of the most popular leisure activities in urban China [5], and karaoke bars burgeoned throughout Mainland China. Our previous study shows that there are 87,700 karaoke bars in Mainland China as of 2012 [6], clustering in the eastern and central areas (Figure1).

ISPRS Int. J. Geo-Inf. 2016, 5, 150; doi:10.3390/ijgi5090150 www.mdpi.com/journal/ijgi ISPRSISPRS Int. Int. J. Geo J. Geo-Inf.-Inf. 20162016, 5,, 51,50 150 2 of2 of17 17

FigureFigure 1. 1.TheThe spatial spatial dis distributiontribution of of karaoke karaoke bars bars in in China China (Source (Source:: [6] [6)])..

AlthoughAlthough karaoke karaoke bars, bars, as as a a form form of of entertainment, entertainment, are are widespread widespread in urban in urban China, China, little is little known is knownabout about their spatial their spatial distribution distribution within within cities. Takingcities. Taking the entertainment the entertainment and leisure and leisure industry industry as a whole, as a therewhole, is ath growingere is a growing body of literaturebody of literature investigating investigating its spatial its distribution spatial distribution in urban China.in urban Zhang China. and ZhangWang and [7] summarized Wang [7] summarized that there are that four there distribution are four patternsdistribution of urban patterns commercial of urban entertainment commercial entertainmentvenues: center-oriented venues: center structure,-oriented discrete structure, structure, discrete attached structure, structure, atta andched isolated structure, structure. and isolated Hu and structure.Zhong [8 Hu] conducted and Zhong a longitudinal [8] conducted study a longitudinal and observed study that the and distribution observed that of entertainment the distribution venues of entertainmentin Hankou experienced venues in Hankou a process experienced from clustering a process to dispersing, from clustering and re-clustering. to dispersing, Takingand re- Chengduclustering. as Takinga case study,Chengdu Yang as [ 9a] foundcase study, that the Yang space [9] distribution found that ofthe leisure space industry distribution is closely of leisure related industry to residents’ is closelyrecreational related activity to residents’ range. recreational With the development activity range. of GIS With as the a powerful development tool for of spatialGIS as visualizationa powerful tooland for analysis, spatial morevisualization and more and scholars analysis, carried more outand empirical more scholars studies carried to show out theempirical spatial studies layout andto showstructure the spatial of certain layout industry and structure (e.g., cultural of certain industry, industry financial (e.g., servicecultural industry, industry, automobile financial service service industry,industry, automobile creative industry, service industry, etc.), and creative to investigate industry, its etc.), spatial and evolution to investigate trend, its spatial spatial growth evolution and trediffusionnd, spatial pattern growth [10 and–17]. diffusion pattern [10–17]. InIn addition addition to to the the spatial spatial distribution distribution pattern pattern of of a a certain certain industry, scholarsscholars areare moremore interested interested in inthe the factors factors shaping shaping its spatialits spatial distribution. distribution. The mostThe most critical critical elements elements of success of success for economic for economic activities activiare supply–demandties are supply–demand relationship relationship and location. and Consumptionlocation. Consumption needs of theneeds population of the population are the primary are theand primary direct driving and direct force for driving the industry force development, for the industry including development, leisure and including entertainment leisure industry. and entertainmentThe supply-demand industry. relationship The supply involves-demand the relationship number of potential involves customers, the number macro-economic of potential customers,conditions, macro regional-economic consumption conditions, demand regional and capacity consumption [18]. While demand location and compasses capacity infrastructure,[18]. While locationbusiness compasses competition infrastructure, environment, business and public competition service environment, provision. Through and public an empiricalservice provision. study on Throughcultural an facilities empirical in Shenzhen, study on cultural Wei and facilities his colleagues in Shenzhen, [19] concluded Wei and thathis colleagues local economy, [19] concluded population thatdistribution, local economy, transportation, population and distribution, policies are the transportation, key factors that and influence policies the are distribution the key factors of cultural that influencefacilities. the Transportation, distribution of which cultural links facilities. customers Transportation, and facilities, which is emphasized links customers to be of and great facilities, importance is empto thehasized location to be choice of great of leisureimportance and entertainment to the location facilities choice of [20 leisure]. Zhang and and entertainment Wang [7] pointed facilities out [20] that. Zhangvarious and types Wang of leisure[7] pointed venues out are that attracted various by types each of other, leisure leading venues to anare agglomeration attracted by each pattern other, and leadingusually to located an agglomeration nearby CBDs. pattern and usually located nearby CBDs. However,However, most most of of the the existing existing studies studies take take the the leisure leisure and and entertainment entertainment industry industry as as a awhole whole andand lack lack in in-depth-depth spatial spatial quantitative quantitative analysis. analysis. In Inthis this paper, paper, specifically specifically focusing focusing on on karaoke karaoke bar, bar, whichwhich is is an an immensely immensely widespread widespread leisure leisure activit activityy in in urban urban China, China, we we aim aim to to demonstrate demonstrate their their spatialspatial distribution distribution pattern pattern within within the the city city by by means means of of the the GIS GIS statistical statistical methods, methods, and and to to explore explore the the sociosocio-spatial-spatial factors factors underlying underlying their their distribution, distribution, which which include include population population density, density, transportation, transportation, anandd location location of of commercial commercial centers. centers. The The study study will will shed shed light light on on the the planning planning and and guidance guidance of of leisure leisure

ISPRS Int. J. Geo-Inf. 2016, 5, 150 3 of 17

ISPRS Int. J. Geo-Inf. 2016, 5, 150 3 of 17 and entertainment industry for the government, as well as enrich the empirical research on leisure and entertainmentand entertainment industry. industry for the government, as well as enrich the empirical research on leisure andKaraoke entertainment bars are industry. usually abstracted as points in the urban space. In this paper, the point pattern analysisKaraoke and cluster bars are analysis usually in abst GISracted will be as employedpoints in the to urban illustrate space. whether In this the paper, karaoke the point bars pattern locate in a randomanalysis fashion and cluster or present analysis a certainin GIS will pattern be employed (clustered to illustrate or dispersed). whether Based the karaoke on the literature bars locate review, in wea random hypothesize fashion that or threepresent main a certain factors pattern influence (clustered the distribution or dispersed). of karaoke Based on bars: the literature population review, density, transportationwe hypothesize network, that three and main the factors location influence of commercial the distribution centers. of karaoke To test thebars: hypothesis, population thedensity, density oftransportation karaoke bars’ network, spatial correlation and the location with theof commercial population centers. density To and test transportation the hypothesis,centrality the density will of be karaoke bars’ spatial correlation with the population density and transportation centrality will be measured respectively. To examine the impact of commercial center on the distribution of karaoke measured respectively. To examine the impact of commercial center on the distribution of karaoke bars, Vironoi cells, standing for service region, will be created for each commercial center and karaoke bars, Vironoi cells, standing for service region, will be created for each commercial center and karaoke bars will be quantize for each service region of the commercial centers. bars will be quantize for each service region of the commercial centers. 2. Data 2. Data Nanjing, the capital of Jiangsu Province, is situated in one of the largest economic zones of China, Nanjing, the capital of Jiangsu Province, is situated in one of the largest economic zones of China, the Yangtze River Delta. The total population is about 8.16 million in 2012 (Nanjing Statistics Yearbook the Yangtze River Delta. The total population is about 8.16 million in 2012 (Nanjing Statistics 2013).Yearbook Nanjing 2013) has. Nanjing long beenhas long the been political, the political, economic economic and cultural and cultural center. center. As a pilotAs a pilot city ofcity service of industryservice industry reforms, reforms, the service the sector service accounted sector accounted for 54.38% for of54.38% GDP of in GDP 2013 in (Nanjing 2013 (Nanjing Statistics Statistics Yearbook, 2014).Yearbook, The fast 2014 growing). The fast service growing sector service has become sector the has driving become force the driving of economic force development of economic in Nanjing.development Among in Nanjing. the service Among industry, the service the leisure industry, and entertainmentthe leisure and sectorentertainment accounted sector for 1%accounted of GDP in 2 2013for 1% (Nanjing of GDP Statistics in 2013 (Nanjing Yearbook, Statistics 2014). Yearbook, With a total 2014) land. With area a oftotal 6598 land km area, Nanjing of 6598 km is consists2, Nanjing of 11 districts.is consists Considering of 11 districts. that Considering there are large that rural there and are undeveloped large rural and areas undeveloped within the areas Nanjing within municipal the boundary,Nanjing municipal we only choose boundary, the main we only urban choose areas the as ourmain study urban area areas in thisas our paper, study which area isin surrounded this paper, by thewhich Yangtze is surrounded River and by the the belt Yangtze highway River (shown and the by belt the highway red area (shown in the by left the map red of area Figure in the2). left There map are 4.38of Figure million 2). residents There are in 4.38 the million study residents area, which in the covers study 577.8 area, kmwhich2. covers 577.8 km2.

FigureFigure 2. Nanjing municipal municipal boundary boundary and and the the study study area area..

TheThe basic basic geographicgeographic datadata are drawn from from the the 2012 2012 Nanjing Nanjing City City maps maps published published by byJiangsu Jiangsu ProvinceProvince Surveying Surveying andand MappingMapping Institute. Through Through map map vectorization vectorization and and coordinate coordinate calibration, calibration, thethe boundary boundary of of the the studystudy area,area, boundaries of of sub sub-districts-districts (a (a low low level level administrative administrative division division of a of a district),district), geographical geographical featuresfeatures (river,(river, lake and and mountain) mountain) have have been been obtained. obtained. The The road road network network data data are from OpenStreetMap. The locations of karaoke bars extract from Baidu maps (similar to Google are from OpenStreetMap. The locations of karaoke bars extract from Baidu maps (similar to Google maps, but covering only the Great China region) and Google maps. After data examination (verify maps, but covering only the Great China region) and Google maps. After data examination (verify

ISPRS Int. J. Geo-Inf. 2016, 5, 150 4 of 17 street view images), a point layer of the karaoke bars has been created. Population data are from the 2010 Population Census at sub-district level.

3. Methodology

3.1. Point Pattern Analysis—Ripley’s K and L Functions Point pattern analysis is one of the most fundamental concepts in geography and spatial analysis. It evaluates the pattern, or distribution, of a set of points on a surface [21]. There are three general patterns: random, uniform, and clustered. Ripley’s K-function [22] is a common technique to estimate the presence of spatial dependence among points. Ripley’s K-function is defined as: 1 K(r) = ∑ I(dij < r)/n (1) λ i6=j where λ is the average density of points (generally estimated as n/A, where A is the area of the region containing all points), and I (dij < t) is the number of points within the circle with the ith point as the center and r as the radius. Under the assumption of complete spatial randomness (CSR), K(r) is equal to πr2. For simplication, the K-function is transformed into a standardized version, L-function, which is defined as: r K (r) L = − r (2) (r) π 2 2 If K(r) > πr (or L(r) > 0), this indicates a clustered point pattern. K(r) < πr (or L(r) < 0), on the contrary, implies that the point pattern is regular.

3.2. Cluster Analysis—Nearest Neighbor Hierarchical Clustering Cluster analysis seeks to group a collection of objects into “clusters”, such that those within each cluster are more closely related to one another than objects assigned to different clusters [23]. It is a main task of exploratory data mining, as well as a common technique for statistical data analysis used in many fields, including machine learning, pattern recognition, image processing, information retrieval, and bioinformatics [24]. Hierarchical clustering is a method of cluster analysis, which seeks to build a hierarchy of clusters by connecting “objects” to form “clusters” based on their distance [25]. Hierarchical clustering can be agglomerative (starting with single elements and aggregating them into clusters) or divisive (starting with the complete data set and dividing it into partitions) [26]. One of the simplest agglomerative hierarchical clustering methods is single-linkage clustering, also known as the nearest neighbor technique. The defining feature of this method is that distance between groups is defined as the distance between the closest pair of objects, where only pairs consisting of one object from each group are considered [27]. The distance between clusters X and Y, D(X, Y) is described by the following expression:

D (X, Y) = min d (x, y) (3) x∈X, y∈Y where object x is in cluster X, and object y is in cluster Y. Assuming that there are N objects in region A, the algorithm of nearest neighbor hierarchical clustering performs the following steps:

i. Each object is seen as a cluster, containing one single object. Calculate the N × N distance matrix D contains all distance d(i, j). ii. Define a distance threshold, which is a certain value set manually or can be the expected mean distance of the features given a random pattern: r 1 A D = (4) E 2 N iii. Merge clusters whose distances are within the set distance threshold. iv. Update the distance matrix between the newly formed clusters. ISPRS Int. J. Geo-Inf. 2016, 5, 150 5 of 17

v. Repeat Step 3 to further merge clusters until no more clusters can be merged.

3.3. Spatial Autocorrelation—Bivariate Moran’s I Spatial autocorrelation can be defined as the coincidence of value similarity with locational similarity, which is used to detect patterns of spatial association [28]. Global measure for spatial autocorrelation, Moran’s I, is a single value which applies to the entire study area [29]; while its corresponding local indices, local Moran’s I, also known as LISA (local indicators of spatial association), show the value calculated for each observation unit, allowing researchers to explore local variations in spatial dependency [30]. In addition to univariate Moran’s I, which focuses on the spatial clustering of observations in terms of a single variable, bivariate Moran’s I has been put forward to capture the relationship between two variables, taking the topological relationship between observations into account [31]. It measures strength of the association between two variables varies over the study region. Bivariate Moran’s I between variables k and l for observation unit i is defined as: i = i j Ikl zk ∑ Wijzl (5) j where Wij is the spatial weight matrix between observation unit i and its neighboring units j, i i zk = (xk − xk)/δk is the standardized form of the value of variable k in the observation unit i, j j and zl = (xl − xl)/δl is the standardized form of the value of variable l in the observation unit js. This statistic gives an indication of the degree of linear association between the value of one variable at a given unit i and the average of another variable at neighboring units js. The results of Bivariate Moran’s I is a map with four categories of spatial correlation: two positive spatial correlation, namely spatial clusters (High–High and Low–Low) which relate to values physically surrounded by neighboring units with similar values, and two negative spatial correlation, or called spatial outliers (High–Low and Low–High) which relate to values surrounded by neighboring units with dissimilar values [32].

3.4. Multiple Centrality Assessment—Network Centrality In regional and urban planning, centrality, or “accessibility” and “proximity” (the terms are often used interchangeably), has entered the scene, stressing that “some places are more important than others because they are more central” [33]. The Multiple Centrality Assessment (MCA) is a tool that evaluates the spatial distribution of centrality over geographic systems like urban street system and outputs the results graphically [34]. The first step to operate MCA is to translate the spatial system into a graph. For instance, in a network of intersections and paths, intersections are abstracted into nodes and paths into edges which are defined by two end-nodes. The main characteristics of the MCA are: (1) use a primal approach of the network; (2) anchor all the measures in the real network of street; and (3) define the centrality by a collection of indexes [35]. Depending on what is the notion of “being central”, three indexes are usually used to characterize the shape of a network: closeness, betweenness, and straightness [36].

3.4.1. Closeness Centrality (CC) “Closeness measures the accessibility of a node, in the way that closer a node to the others, more accessible it is” [36]. The closeness index of node i is defined as: N c Ci = (N − 1) / ∑ dij (6) j6=i where dij is the shortest paths from node i to node j, and N is the number of all nodes in the network. Actually, closeness is the reciprocal of the mean of the shortest paths from node i to other nodes. Thus, the more central a node, the lower its average distance to all other nodes. ISPRS Int. J. Geo-Inf. 2016, 5, 150 6 of 17

3.4.2. Betweenness Centrality (CB) ISPRS Int. J. Geo-Inf. 2016, 5, 150 6 of 17 Beyond the closeness index, the interaction between two distant nodes depends on the nodes 3.4.2. Betweenness Centrality (CB) belonging the path which links them. These nodes have a strategic location for the control and the influenceBeyond of the flowsthe closeness [35]. “Betweenness index, the interaction centrality between quantifies two distant the number nodes depends of times on a the node nodes acts as a bridgebelonging along the the shortest path which path links between them. two These other nodes nodes” have [a37 strategic]. The betweenness location for the centrality control and of node the i is: influence of the flows [35]. “Betweenness centrality quantifies the number of times a node acts as a N ( ) bridge along the shortest path betweenB two other1 nodes” [37]. Thenjk betweennessi centrality of node i is: Ci = (7) (N − 1)(N − 2) 푁∑ n j6=k6=i ( )jk 퐵 1 푛푗푘 푖 퐶푖 = ∑ (7) (푁 − 1)(푁 − 2) 푛푗푘 where njk is the total number of shortest paths from nodej≠jk≠toi node k, and njk (i) is the number of those paths that pass through node i. This index measures the volume of flow on a node. where 푛푗푘 is the total number of shortest paths from node j to node k, and 푛푗푘(푖) is the number of those paths that pass through node i. This index measures the volume of flow on a node. 3.4.3. Straightness Centrality (CS)

S 3Starting.4.3. Straightness from the C hypothesisentrality (C that) “the communication between two nodes is better when the path is straight”,Starting the straightness from the hypothesis centrality that assesses “the communication the extent to whichbetween the two shortest nodes pathis better between when nodethe i to otherpath nodes is straight diverts”, from the straightness the straight centrality line linking assesses them the [35 extent]: to which the shortest path between node i to other nodes diverts from the straight line linking them [35]: N Eucl 1 푁 dij CS = 1 푑Eucl (8) i퐶푆 =( − )∑∑ 푖푗 푖 N 1 j6=i dij (8) (푁 − 1) 푑푖푗 푗≠푖 Eucl in which d isEucl the Euclidean distance between node i and node j, and dij is the network distance in whichij 푑푖푗 is the Euclidean distance between node i and node j, and 푑푖푗is the network distance betweenbetween node nodei and i and node nodej. j The. The smaller smaller the the difference between between Euclidean Euclidean distance distance and and network network distance,distance, the morethe more “straight” “straight central” central the the node node is.is.

3.5. Spatial3.5. Spatial Interpolation—Kernel Interpolation—Kernel Density Density Estimation Estimation SpatialSpatial interpolation interpolation is is “the “the procedure procedure ofof estimating the the value value of ofproperties properties at unsampled at unsampled sites sites (or cells(or cells in a raster)in a raster) within within the the area area covered covered by by existing existing observations”observations” [38] [38.]. Kernel Kernel density density estimation estimation is an interpolationis an interpolation technique technique that that evaluates evaluates the the probability probability density density function function of a of random a random variable variable [39]. [39]. “Kernel“Kernel density density estimation estimation involves involves placing placing aa symmetricalsymmetrical surface surface over over each each point, point, evaluating evaluating the the distance from the point to a reference location based on a mathematical function, and summing the distance from the point to a reference location based on a mathematical function, and summing the value of all the surfaces of that reference location” [40] (Figure 3). value of all the surfaces of that reference location” [40] (Figure3).

FigureFigure 3. 3.Kernel Kernel densitydensity estimate for for points points in inplane plane..

TheThe estimated estimated density density of of point points sis is defined defined as:as: N 11 d(s,d(ss,)si) f (sf)(s=) = ∑ K(K( i ) ) (9) (9) r푟 ∑ 푟 r ii=1=1 wherewhereN is N the is the number number of of existing existing observations, observations, h isis the the smoothing smoothing parameter parameter called called the bandwidth, the bandwidth, d(s,sid) (s,si) ( d(s,) s ) is the distance between point s and the existing observation points si, and K( ) is the kernel, d s, si isi the distance between point s and the existing observation points si, and K푟( r ) is the kernel, a weightinga weighting function, function, i.e., i.e. the, the set set of of rules rules whichwhich determines how how much much weight weight a point a point close close to the to the

ISPRS Int. J. Geo-Inf. 2016, 5, 150 7 of 17

centerISPRS should Int. J. Geo be-Inf. given 2016, 5 relative, 150 to one near the edge of the circle. As seen in the equation, the7 results of 17 are influenced by three factors: center should be given relative to one near the edge of the circle. As seen in the equation, the results 1. areThe influenced distance by between three factors: the points. 2. The bandwidth, which is very important to control the degree of smoothing applied to the data. 1. The distance between the points. If the bandwidth is small, we will obtain an under smoothed estimator, with high variability. 2. The bandwidth, which is very important to control the degree of smoothing applied to the data. On the contrary, if it is big, the resulting estimator will be over smooth and farther from the If the bandwidth is small, we will obtain an under smoothed estimator, with high variability. function that we are trying to estimate [41]. On the contrary, if it is big, the resulting estimator will be over smooth and farther from the 3. Thefunction kernel that weighting we are trying function. to estimate Although [41 there]. is a range of kernel functions, empirical studies 3. indicateThe kernel very weighting little difference function. in the Although results betweenthere is a different range of weightingkernel functions, functions. empirical We will studies use the quadraticindicate polynomialvery little difference kernel function in the results in the between following different analysis. weighting functions. We will use the quadratic polynomial kernel function in the following analysis. 4. Spatial Distribution Pattern of Karaoke Bars 4. AsSpatial of the Distribution end of 2012, Pattern there of are Karaoke 530 karaoke Bars bars in the study area. Table1 shows the number and densityAs of ofthe karaoke end of 2012, bars there in each are district.530 karaoke Considering bars in the that study the area. size Table of each 1 shows sub-district the number varies, usingand thedensity number of karaoke of karaoke bars in bars each per distric 1 kmt. Considering2 and per 10,000 that the residents size of each make sub it-district comparable varies, between using sub-districts.the number Within of karaoke the study bars area,per 1 there km2 are,and onper average, 10,000 residents 0.92 karaoke make bars it comparable per 1 km2 and between 1.21 karaoke sub- barsdistricts. serving Within for 10,000 the study residents. area, there Comparing are, on average, each district, 0.92 karaoke Gulou bars and per Qinhuai 1 km2 and are 1.21 the karao twoke districts bars withserving concentration for 10,000 ofresidents. karaoke Comparing bars, respectively, each district, 2.97 andGulou 2.72 and karaoke Qinhuai bars are per the1 two km 2districtsarea, while with in termsconcentration of the size of of karaoke serving bars, population, respectively there, 2.97 are and more 2.72 karaokekaraoke bars per in Jiangning1 km2 area, and while Qixia in terms districts. of Tothe go size beyond of serving the administrative population, there boundaries are more of karaoke sub-districts, bars in Jiangning we divide and the Qixia study districts. area into To gridsgo (300beyond m × 300 the madministrative). Figure4 demonstrates boundaries of thesub number-districts, ofwe karaoke divide the bars study in eacharea into grid. grids Clearly, (300 m karaoke × 300 barsm). concentrate Figure 4 demonstrates at the city center,the number Xinjiekou of karaoke as well bars as in Hunan each grid. Road, Clearly, and just karaoke a few bars of them conce scatterntrate in suburbanat the city areas. center, Xinjiekou as well as Hunan Road, and just a few of them scatter in suburban areas.

TableTable 1. 1The. The number number andand densitydensity of karaoke bars bars in in study study area area.. The Number of The Number of Karaoke The Number of Karaoke Sub-District The Number of The Number of Karaoke The Number of Karaoke Sub-District Karaoke Bars Bars per 1 km2 Area Bars per 10,000 Residents Karaoke Bars Bars per 1 km2 Area Bars per 10,000 Residents Gulou 146 2.97 1.15 JianyeGulou 14625 2.970.54 1.150.62 Jianye 25 0.54 0.62 Qinhuai 134 2.72 1.33 Qinhuai 134 2.72 1.33 XuanwuXuanwu 5757 0.760.76 0.870.87 YuhuataiYuhuatai 2727 0.400.40 0.940.94 JiangningJiangning 6767 0.410.41 2.442.44 QixiaQixia 7474 0.590.59 1.571.57 StudyStudy area area 530530 0.920.92 1.211.21

FigureFigure 4. 4.The The distribution distribution ofof karaokekaraoke bars in the the study study area area of of Nanjing. Nanjing.

ISPRS Int. J. Geo-Inf. 2016, 5, 150 8 of 17 ISPRS Int. J. Geo-Inf. 2016, 5, 150 8 of 17

To determine whether the distribution of karaoke bars exhibits a systematic pattern over the To determine whether the distribution of karaoke bars exhibits a systematic pattern over the study area rather than being randomly distributed, L-function, the standardized version of Ripley’s study area rather than being randomly distributed, L-function, the standardized version of Ripley’s F-function (1976), is used to examine the spatial dependence based on distances of each karaoke bar F-function (1976), is used to examine the spatial dependence based on distances of each karaoke bar from one another. As shown in Figure 5, the blue curve is the observed L value of karaoke bars in the from one another. As shown in Figure5, the blue curve is the observed L value of karaoke bars in study area. It is always larger than the expected L value (green curve) for any given distance the study area. It is always larger than the expected L value (green curve) for any given distance threshold, indicating that the distribution of karaoke bars is in general clustered. However, the extent threshold, indicating that the distribution of karaoke bars is in general clustered. However, the extent of clustering differs. When the distance threshold is set less than 2.5 km, the extent of clustering is of clustering differs. When the distance threshold is set less than 2.5 km, the extent of clustering is positively associated with the distance. The peak of observed L value appears when the distance positively associated with the distance. The peak of observed L value appears when the distance threshold is between 2.5 and 3.5 km, showing that karaoke bars are highly clustered at this distance threshold is between 2.5 and 3.5 km, showing that karaoke bars are highly clustered at this distance scale. Afterwards, the level of clustering decreases with the increase of distance threshold. scale. Afterwards, the level of clustering decreases with the increase of distance threshold.

FigureFigure 5. 5. LL-function-function of of karaoke karaoke bars bars..

AsAs the K K-function-function has has verified verified that the karaoke bars are spatially clustered, we we continue the the analysisanalysis by by making making hierarchical hierarchical clustering clustering to to visually visually illustrate illustrate the the concentrate concentrate areas areas of of karaoke karaoke bars. bars. There are eight identified identified clusters in the study area ( (FigureFigure 66).). EachEach clustercluster hashas moremore thanthan 1010 karaokekaraoke 2 bars,bars, and and the the average area area of the eight clusters is 0.35 km 2.. In In general, general, the the karaoke karaoke bars bars in in the the study study areaarea concentrate in intenselytensely in certain small area areas.s. BasedBased on on the location and characteristics of of the the clusters, we we classify classify these these clusters clusters into into the the followingfollowing categories. categories. (a) (a) Custom Custom-oriented-oriented cluster clusters:s: It It is is the the nature nature of of leisure leisure and and entertainment servicesservices to to distribute distribute at at the the areas areas with more potential cust customs.oms. The The 1st 1st cluster cluster shown shown in in Figure Figure 66 isis locatedlocated at at Hunanlu Hunanlu sub sub-district,-district, which ranks the second in terms terms of of the the population population size size in in main main urban urban areasareas of Nanjing. Besides, Besides, Nanjing Nanjing University, University, East East south University and and some some other institutions of of higherhigher educa educationtion are are located located nearby, nearby, with with a a large large number number of of local local residents residents and and young young consumers consumers;. ((b)b) CBD CBD-oriented-oriented cluster clusters:s: The The business business center center is is a a hub hub of of the the city city of of all all kinds kinds of of commercial commercial activities, activities, includingincluding shopping, shopping, dining, dining, entertainment entertainment,, etc. etc. T Thehe 5 5th,th, 7 7thth and and 6 6thth clusters clusters in in Figure Figure 66 areare locatedlocated atat Confucius Confucius Temple Temple CBD, CBD, Dongshan Dongshan CBD CBD and and Ruijin Ruijin Road Road respectively. respectively. Confucius Confucius Temple Temple CBD CBD is ais traditional a traditional commercial commercial center center in in Nanjing, Nanjing, as as well well as as a a leisure leisure and and tourist tourist center, center, attracting attracting an an immenseimmense number number of of passenger passenger flow flows.s. Thereby, Thereby, karaoke karaoke bars, bars, as a as kind a kind of entertainment of entertainment venues, venues, are alsoare also gathering gathering here. here. The The Dongshan Dongshan CBD CBD is isthe the only only trading, trading, commercial commercial and and leisure leisure center center in in DongshanDongshan satellite satellite city, city, with karaoke bars also clustering here. Ruijin Ruijin Road Road is is proximate proximate to to X Xinjiekouinjiekou CBD,CBD, sharing sharing and and transferring transferring the the functions functions of ofXinjiekou Xinjiekou CBD. CBD. Therefore, Therefore, it is italso is alsoa concentrate a concentrate area ofarea karaoke of karaoke bars. bars; (c) Transportation (c) Transportation-oriented-oriented cluster clusters:s: The The 2nd 2nd and and 6th 6th clusters clusters in in Figure Figure 6 are are distributeddistributed at at Xinjiekou Xinjiekou CBD CBD and and Maigaoqiao Maigaoqiao,, res respectively.pectively. Xinjiekou is located at at the the intersection intersection of of fourfour important important avenues, avenues, having having six six bus bus stops stops with with over over 30 30 bus bus lines lines crossing, crossing, and and is is the the interchange interchange stationstation of of subway subway line line 1 1 and and 2. 2. Maigaoqiao, Maigaoqiao, situated situated in in the the north north of of the the main main urban urban areas areas of of Nanjing, Nanjing, isis the the northern northern originating originating station station of of subway subway line line 1, 1, which which makes makes it it irreplaceable irreplaceable as as a a transportation transportation nodenode especially for the northernnorthern area.area. TheThe traffictraffic advantages advantages of of these these two two places places attract attract karaoke karaoke bars bars to toconcentrate concentrate there; there (d). (d) Business-oriented Business-oriented clusters: clusters: Karaoke Karaoke barsbars areare notnot onlyonly forfor recreation, but also become a popular venue for casual business meetings. The 3rd and 4th clusters are located at

ISPRS Int. J. Geo-Inf. 2016, 5, 150 9 of 17

ISPRS Int. J. Geo-Inf. 2016, 5, 150 9 of 17 become a popular venue for casual business meetings. The 3rd and 4th clusters are located at Hongwu RoadHongwu and Road Zhenghong and Zhenghong Commercial Commercial Pedestrian Pedestrian Street, where Street, various where financial various institutions, financial institutions, consulting firms,consulting and afirms, large numberand a large of corporation number of headquarters corporation areheadquarters clustered. Theare demandclustered. for The causal demand business for venuescausal business results in venues the agglomeration results in the of agglomeration karaoke bars inof thesekaraoke areas. bars in these areas.

Figure 6. Clusters of karaoke bars.

5. Socio Socio-Spatial-Spatial Factors’ Factors’ Influence Influence on on the the D Distributionistribution of of Karaoke Karaoke Bars Bars Karaoke bars, similar as other commercial services, seek a location to obtain as much benefit benefit as possible, taking a variety of spatial factors into account [[42]42],, therefore exhibiting a specificspecific spatial distribution pattern [43] [43].. To To explain the spatial clusters of karaoke bars identifiedidentified in the previous section, analyses analyses were were carried carried out out to quantify to quantify the relationship the relationship among among karaoke karaoke bars, population bars, population density, transportationdensity, transportation network network density, anddensity, commercial and commercial center. center.

5.1. Population D Densityensity As argued in the literature review, consumptionconsumption needs of the population are the primary and direct driving force for the developmentdevelopment of leisureleisure andand entertainmententertainment industry. Chinese cities are undergoing rapid urbanization, and the development of areas w withinithin the cities is is far far away away from from even. even. Population density, density, to to some some extent, extent, reflects reflects the spatial the spatial variations variations in land in price land and price economic and economic structure, whichstructure, may which also exert may influencealso exert oninfluence the distribution on the distribution of karaoke of bars. karaoke bars. The population population data data are are from from the the 2010 2010 Population Population Census, Census, which which only takes onlyresidents takes residents into account into excludingaccount excluding floating populationfloating population (without (without local hukou, local or hukou, reside or less reside than less six months).than six months). The total The number total ofnumber residents of residents is 4.37 million, is 4.37 distributedmillion, distributed in 54 sub-districts in 54 sub within-districts the within study the area. study As seen area. from As seen Figure from7a, 2 theFigure residents 7a, the highly residents concentrate highly concentrate at the old at city the town, old city with town, the density with the over density 32,000 over residents 32,000 residents per km , andper km there2, and is a cleartherepattern is a clear that pattern the density that the decreases density gradually decreases with gradually the increase with ofthe the increase distance of from the distance from the city center. The density of karaoke bars presents a similar pattern, with high values appearing in the city center, but less regularity in the suburb areas (Figure 7b).

ISPRS Int. J. Geo-Inf. 2016, 5, 150 10 of 17

the city center. The density of karaoke bars presents a similar pattern, with high values appearing in

theISPRS city Int. center, J. Geo-Inf. but 2016 less, 5, regularity150 in the suburb areas (Figure7b). 10 of 17 ISPRS Int. J. Geo-Inf. 2016, 5, 150 10 of 17

FigureFigureFigure 7. 7.7. Density DensityDensity comparison comparisoncomparison at atat sub-district sub-districtsub-district level: level:level (: (a(a)a) )Population PopulationPopulation density; density;density; ( (b(bb)) )Karaoke KaraokeKaraoke bar barbar density. density.density .

Figure 8. Spatial correlation between population density and karaoke bars: (a) Moran scatter plot; (b) FigureFigure 8. 8. SpatialSpatial correlation correlation between between population population density density and and karaoke karaoke bars: bars: (a) Moran (a) Moran scatter scatter plot; plot; (b) Spatial clusters and outliers. Spatial(b) Spatial clusters clusters and andoutliers. outliers.

To further assess the spatial association between population density and the density of karaoke ToTo further assess the the spatial association association between between population population density density and and the the density density of of karaoke karaoke bars, we employed Bivariate Moran’s I statistic in GeoDa. A rook weights matrix was created, which bars,bars, we we employed employed BivariateBivariate Moran’s Moran’s I statistic I statistic in GeoDa. in GeoDa. A rook A weights rook weights matrix matrixwas created, was created, which defines a sub-district’s neighbors as those sub-districts with shared borders. The value of Global defineswhich definesa sub-district’s a sub-district’s neighbors neighbors as those as sub- thosedistricts sub-districts with shared with sharedborders. borders. The value The of value Global of BivariateBivariate Moran’s Moran’s I I(the (the slope slope of of the the scatter scatter plot) plot) is is 0.4956, 0.4956, indicating indicating a a positive positive association association between between thethe distribution distribution of of the the population population and and karaoke karaoke bars bars (Figure (Figure 8a). 8a). Four Four quadrants quadrants in in the the scatter scatter plot plot correspondcorrespond to to four four different different types types of of association. association. The The first first and and the the third third quadrants quadrants represent represent spatial spatial clusters,clusters, HH HH (High–High) (High–High) and and LL LL (Low–Low), (Low–Low), respectively. respectively. The The second second and and the the fourth fourth quadrants quadrants representrepresent spatial spatial outliers, outliers, LH LH (Low–High) (Low–High) and and HL HL (High–Low), (High–Low) respectively., respectively Figure. Figure 8b 8b demonstrates demonstrates

ISPRS Int. J. Geo-Inf. 2016, 5, 150 11 of 17

Global Bivariate Moran’s I (the slope of the scatter plot) is 0.4956, indicating a positive association between the distribution of the population and karaoke bars (Figure8a). Four quadrants in the scatter plot correspond to four different types of association. The first and the third quadrants represent spatial clusters, HH (High–High) and LL (Low–Low), respectively. The second and the fourth quadrants represent spatial outliers, LH (Low–High) and HL (High–Low), respectively. Figure8b demonstrates these clusters and outliers spatially. There are 43 sub-districts with positive association (in the first or the third quadrants) and many of them are significant, meaning that karaoke bars are in general distributed at the places where more people reside. As expected, the HH is mainly located in the urban core, and LL is found in the suburban areas, whereas there are fewer sub-districts in which the density of the population and karaoke bars are negatively associated. The only significant spatial outlier (low–high) is found in Xuanwu sub-district. It is a popular tourist area, having Xuanwu Lake and Jiming Temple, attracting numerous tourists, as well as residents from outside of Xuanwu sub-district. Therefore, although the density of local residents is comparatively low, the number of potential customers is abundant, so the distribution of karaoke bars is dense on the contrary.

5.2. Road Network Urban road network constitutes an important factor in urban systems, not only connecting consumers and economic activity venues, but also limiting and influencing the spatial layout of these commercial facilities. Accessibility, which can be transformed into a better visibility and popularity [35], is definitely a crucial factor for location choice of any kinds of commercial activities including karaoke bars. In Multiple Centrality Assessment, accessibility in the urban environment can be measured by the road network’s centrality, based on the assumption that a central area is more accessible. As discussed in the Methodology, three indexes are commonly used to measure “centrality”: closeness, betweenness, and straightness. Before calculating each of the indexes, we firstly need to construct the road network by extracting the ends and intersections of each road to form the nodes and edges of the network. Given the fact that there are different levels of roads with varied capacities, a corresponding weight is assigned to each road level, thereby we have following calculation formulas for each index (Table2).

Table 2. Calculation formulas for centrality indexes.

Centrality Index Calculation Formula N c   Closeness Ci = (N − 1) / ∑ dij × Wj j6=i N n (i) CB = 1 ( jk × W ) Betweenness i (N−1)(N−2) ∑ njk j j6=k6=i N dEucl S = 1 ij Straightness Ci (N−1) ∑ d j6=i ij

Based on the formulas, each road’s closeness, betweenness, and straightness are calculated and illustrated in Figure9. Centrality closeness reflects the proximity of a node to other nodes in the network. As shown by Figure9a, closeness of the study area presents a clear “core–periphery” pattern. The roads with high value of closeness are located at Xinjiekou, along East Zhongshan Road. Centrality betweenness measures the frequency of a node being passed through by the shortest path between two other nodes. The higher the value of betweenness, the more influential is the node acting as a connecting point. In general, the betweenness is comparatively even across the whole study area (Figure9b). Only a few nodes around Zhujiang Road and the subway stop of Daxinggong have higher values of betweenness, indicating large traffic flow. Centrality straightness evaluates the efficiency in communication between a node and other nodes in the network, based on the assumption that “being central means being straight to the others” [34]. In the study area, there are many nodes with higher value of straightness, distributed along the East Zhongshan Road, Zhujiang Road, and South Taiping Road (Figure9c). On the whole, the pattern of straightness is similar to that of closeness. ISPRS Int. J. Geo-Inf. 2016, 5, 150 12 of 17 ISPRS Int. J. Geo-Inf. 2016, 5, 150 12 of 17 ISPRS Int. J. Geo-Inf. 2016, 5, 150 12 of 17

Figure 9. Centrality of road network: (a) Closeness; (b) Betweenness; (c) Straightness. FigureFigure 9. 9.Centrality Centrality of of road road network: network:( (aa)) Closeness;Closeness; (b) Betweenness; (c (c) )Straightness Straightness.. To investigate the spatial association between the centrality of the road nodes and the distributionToTo investigate investigate of karaoke the spatial thebars, spatial we association need association to interpolate between between the the centrality value the centrality of ofeach the discrete road of the nodes point road and onto nodes the a distributioncontinuous and the rasterof karaokedistribution layer bars, by of we karaoke using need kernelbars, to interpolate we density need to the interpolate estimation value of eachthe (KDE) value discrete (seeof each pointSecti discrete ontoon 3.5). apoint continuous As onto the a continuous rasterbandwidth layer substantiallybyraster using kernel layer affects bydensity using the estimationdegree kernel of density smoothing (KDE) (see estimation Sectionapplied (KDE) 3.5to ).the As (seedata, the bandwidth Sweecti experimentedon 3.5). substantially As the with bandwidth a affects series the of bandwidths,degreesubstantially of smoothing including affects applied the 600 degree m, 680 to theof m, smoothing data,and 800 we m. experimentedapplied Afterwards to the, we withdata, chose awe series experimented to use of 680 bandwidths, m aswith the a bandwidth series including of bandwidths, including 600 m, 680 m, and 800 m. Afterwards, we chose to use 680 m as the bandwidth to600 plot m, the 680 estimated m, and 800 density m. Afterwards, of karaoke we bars chose and to the use centralities 680 m as theof the bandwidth road network to plot to the each estimated cell (50 to plot the estimated density of karaoke bars and the centralities of the road network to× each cell (50 mdensity × 50 m) of on karaoke the created bars andraster the layers centralities (Figure of 10). the road network to each cell (50 m 50 m) on the createdm × 50 raster m) on layers the created (Figure raster 10). layers (Figure 10).

FigureFigure 10. 10.Illustration Illustration ofof rasterraster layers with KDEs. KDEs. Figure 10. Illustration of raster layers with KDEs. TheThe matrix matrix in in Figure Figure 10 10 shows shows the the Moran’s Moran’s II betweenbetween each pair of of KDEs. KDEs. Althoug Althoughh the the three three The matrix in Figure 10 shows the Moran’s I between each pair of KDEs. Although the three centralitycentrality indexes indexes are are devised devised from from different different perspectivesperspectives and are are computed computed differently, differently, their their values values centralityareare highly highly indexes correlated correlated are (thedevised (the values values from of of different Moran’s Moran’s perspectives II amongamong them and are are 0.951, 0.951, computed 0.970, 0.970, and anddifferently, 0.919 0.919,, respectively), respectively), their values arecomprehensivelycomprehensively highly correlated reflecting reflecting (the values the the centrality centralitof Moran’sy of of theIthe among roadroad network.themnetwork. are In 0.951, general, general, 0.970, the the and spatial spatial 0.919 distribution, distribution respectively), of of comprehensivelykaraokekaraoke bars bars is is found foundreflecting to to bebe the positivelypositively centralit associated associatedy of the road with with network.the the centrality centrality In general, of the of theroad the road network,spatial network, distribution whereas whereas the of karaoketheextent extent bars of of theis the found association association to be positively varies varies for for associated the the three three with centrality centrality the centrality indexes. indexes. of Among Amongthe road them, them, network, the the straigh straightnesswhereastness the extentcentralitycentrality of isthe theis associationthe most most related, related, varies with with for Moran’s Moran’s the three II ofof 0.527,0.527, centrality and closeness indexes. centrality Amongcentrality them,is is also also relatively the relatively straigh high,tness high, withcentrality Moron’s is the I ofmost 0.473. related, It implies with thatMoran’s whether I of 0.527, the road and is closeness straight to centrality others and is also whether relatively the road high, is

ISPRS Int. J. Geo-Inf. 2016, 5, 150 13 of 17 close to others play a crucial role in the location choice of karaoke bars. However, the betweenness centrality, which measures the traffic flow, is less important, probably because the places with high betweenness may experience traffic congestion and lack of parking spaces.

5.3. Commercial Center Commercial centers act as the main carrier for various kinds of commercial activities, ranging from the buying and selling of goods and services in retail business, wholesale buying and selling, financial establishments, and wide variety of services that are broadly classified as “business”. Commercial centers in a city are an important component of urban structure, having a considerable impact on the choice of the consuming activities. To explore the spatial association between commercial centers and the location of karaoke bars, we firstly partition the urban space into service regions of each commercial center. In geography, Voronoi diagrams are usually used to answer nearest neighbor queries, while in economic geography, Voronoi diagrams can be used to stimulate the division of the service area of commercial centers. When determining the service area of the commercial center, it is essential to consider the influential range of commercial centers, as well as the traffic network. In terms of scale, service radius, targeted customs, etc., there are three different levels of commercial centers in Nanjing: city level, secondary city level, and district level (Table3). These three levels of commercial centers were given weights 1, 0.8, and 0.5, respectively, to indicate their influential range. To take into account the traffic network, Voronoi diagrams were created based on distance along the traffic network instead of the Euclidean distance. Employing the “Voronoi diagram” toolbox in SANET, Voronoi diagrams were created for each commercial center (Figure 11).

Table 3. Commercial centers in Nanjing.

Number of Level List of Commercial Centers Commercial Center City level 2 Xinjiekou, Hexi Hunan Road, Confucius Temple, South Train Station, Secondary city level 5 Dongshan, Xianlin Zhongyangmen, Ruijin Road, Xiaozhuang, Rehe Road, District level 9 Zhongbao, Jiangdong, Andemen, Xiaolingwei, Hedingqiao Source: Nanjing Commerce Bureau.

Table4 summarizes the area of each Voronoi cell of the each commercial center and the number of karaoke bars located in each Voronoi cell. Whether the number of karaoke bars matches the service area of commercial centers implies the influence of commercial centers on the location choice of karaoke bars. The relationship between the karaoke bars and the commercial centers is categorized into three types: (1) Karaoke bars cluster in the commercial centers with higher level and larger influence, such as Xinjiekou, Hunan Road, and Confucius Temple. These areas have substantial influence on the market, attracting consumer groups with relatively fixed consumer habits; (2) In the process of urban development, commercial space is refined into different functional centers to meet the needs of the urban economy. Therefore, being in line with the function and industrial type of commercial centers is also an important factor to be considered when choosing location for business. For instance, the targeted group of the South Train Station commercial center is non-resident population, so the main industries are restaurant, hotel, and other service industry. Besides, the State Council issued the “Entertainment Management Regulations”, in which entertainment venues are restricted to be located in train stations, airports, and other places with a dense crowd. Consequently, there are only four karaoke bars located within the service area of the South Train Station commercial center; (3) The number of karaoke bars is associated with the location and the level of commercial centers. In general, there are more karaoke bars in commercial centers that are located in the urban center than in periphery areas. Furthermore, there are more karaoke bars found in higher level commercial centers. ISPRS Int. J. Geo-Inf. 2016, 5, 150 14 of 17 center than in periphery areas. Furthermore, there are more karaoke bars found in higher level ISPRScommercial Int. J. Geo-Inf. centers.2016 , 5, 150 14 of 17

Table 4. Statistics of karaoke bars in each Voronoi diagram of the commercial centers. Table 4. Statistics of karaoke bars in each Voronoi diagram of the commercial centers. Commercial Area Number of Commercial Area Number of CenterCommercial (kmArea2) KaraokeNumber Bars of CommercialCenter Area(km2) NumberKaraoke of Bars Center (km2) Karaoke Bars Center (km2) Karaoke Bars Xinjiekou 9.76 100 Ruijin Road 19.34 30 Xinjiekou 9.76 100 Ruijin Road 19.34 30 HexiHexi 57.06 57.06 8 8 XiaozhuangXiaozhuang 76.3676.36 46 46 HunanHunan Road Road 9.50 9.50 64 64 ReheRehe Road Road 10.9810.98 33 33 ConfuciusConfucius Temple Temple 10.28 10.28 44 44 ZhongbaoZhongbao 12.4712.47 17 17 SouthSouth Train Train Station Station 26.33 26.33 4 4 JiangdongJiangdong 12.4112.41 26 26 Dongshan 50.47 39 Andemen, 43.00 20 Dongshan 50.47 39 Andemen, 43.00 20 Xianlin 77.39 27 Xiaolingwei 58.50 10 XianlinZhongyangmen 77.39 28.39 27 40 HedingqiaoXiaolingwei 75.5858.50 22 10 Zhongyangmen 28.39 40 Hedingqiao 75.58 22

Figure 11. NetworkNetwork Voronoi Voronoi diagram of commercial centers in Nanjing.

6. Conclusion Conclusionss With thethe economiceconomic development, development, the the advancement advancement of culturalof cultural quality, quality, and and prolonged prolonged leisure leisure time oftime urban of urban residents, residents, a large a large number number of commercial of commercial entertainment entertainment places places have sprunghave sprung up in up urban in urban areas ofareas China. of China. Karaoke Karaoke bar is amongbar is among the most the popular most popular entertainments. entertainments. This paper, This focusing paper, focusing on the karaoke on the barskaraoke in Nanjing, bars in Nanjing, has examined has examined their spatial their distribution spatial distribution in the urban in spacethe urban and theirspace spatial and their association spatial withassociation population with density,population road density, network, road and network, commercial and centers.commercial centers. Overall, karaoke industry is thriving in Nanjing, with 350 karaoke bars distributed in the study area. Spatial Spatial uneven uneven is is evident. evident. Most Most of the of the karaoke karaoke bars bars are found are found in urban in urban core areas. core areas. Throug Throughh point pointpattern pattern analysis, analysis, it has itbeen has verified been verified that the that karaoke the karaoke bars are bars highly are highly clustered clustered at the atdistance the distance range rangeof 2.5– of3.0 2.5–3.0 km. Based km. Based on the on location the location and and characteristics characteristics of ofthe the clusters, clusters, four four types types of of cluster cluster are identified: custom-oriented clustering, CBD-oriented clustering, transportation-oriented clustering, and business-oriented clustering. ISPRS Int. J. Geo-Inf. 2016, 5, 150 15 of 17

The distribution pattern of karaoke bars is closely associated with the location-specific social features. Among them, population density is a significant one. Karaoke bars are in general clustered in the places where more people reside, and these places are usually located in the city center. The accessibility of road network, measured by three “centrality” indexes closeness, betweenness, and straightness, is found to play a crucial role in guiding the location of karaoke bars. To assess the impact of commercial centers on the distribution of karaoke bars, the urban space has been partitioned into service regions of 16 commercial centers using Voronoi diagrams. In general, commercial centers with higher level and better location attract more karaoke bars. The severe uneven distribution of karaoke bars may lead to an imbalance between the recreation demands and supply. Xinjiekou is a major commercial center in the city. This single-center spatial structure may result in highly concentration of commercial services and traffic congestion. Therefore, supporting the development of existing commercial centers in suburbs, such as Xianlin, Dongshan, and Hexi, would decentralized the entertainment service function of central areas, thus easing pressure on the old city space, as well as improving the level of entertainment service in suburbs. In addition to supporting the development of commercial center, taking advantage of the subway to locate karaoke bars is another strategy. Subway, as an important travel mode, induces business activities, human resources, information and other resources gathering along the subway lines, forming a huge passenger, business, and capital flows. In the study area, there are three subway lines. There are several subway stops in the suburbs that have already demonstrated aggregation effects for business activities. Entertainment and leisure industry may make full use of the function and resource of subway stops and locate their venues nearby the subway stops in the suburbs, which will not only reduce the competition pressure, but also diminish the regional differences.

Acknowledgments: This project was supported by the National Natural Science Foundation of China [41571377] and the Program for New Century Excellent Talents in University [NCET-13-0280]. Author Contributions: Can Cui and Jiechen Wang conceived the research and wrote the paper; Zhongjie Wu, Jianhua Ni and Tianlu.Qian collected the data, designed and performed the data analysis, and prepared the figures and tables. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Molitor, G.T. The next 1,000 years: The ‘big five’ engines of economic growth. Futurist 1999, 33, 13. 2. Godbey, G. Leisure and Leisure Services in the 21st Century; Venture Publishing Inc.: State College, PA, USA, 1997. 3. Fung, A. Consuming karaoke in China: Modernities and cultural contradiction. Chin. Sociol. Anthropol. 2009, 42, 39–55. [CrossRef] 4. April, T.P.L. The Role of Karaoke Box Servicescapes Play in Customer Satisfaction. Ph.D. Thesis, Baptist University, Hong Kong, China, 2005. 5. CE.cn. 1991: Karaoke Started a New Epoch of Entertainment (January 9). Available online: http://views.ce. cn/fun/corpus/ce/zsrs/200901/09/t20090109_17917725.shtml (accessed on 23 October 2015). 6. Wu, Z.J.; Rui, Y.K.; Jia, J.; Ni, J.H.; Wang, J.C. Analysis on the spatial distribution and expansion characteristics of urban entertainment and leisure venues: A case Study on KTV. Geo-Inf. Sci. 2015, 17, 590–597. (In Chinese) 7. Zhang, B.; Wang, X.Z. The study of spatial structures of the urban commercial entertainment places. Sci. Geogr. Sin. 2007, 27, 853–858. (In Chinese) 8. Hu, J.X.; Zhong, A.P. Spatial cluster and urban development of popular cultural and entertainment in Hankou. Wuhan Univ. J. (Hum. Sci.) 2012, 65, 25–32. (In Chinese) 9. Yang, G.L. Study on the influence of residents’ recreational activities on leisure industry. Hum. Geogr. 2003, 18, 18–22. (In Chinese) 10. Frank, L.; Glanz, K.; McCarron, M.; Sallis, J.; Sealens, B.; Chapman, J. The spatial distribution of food outlet type and quality around schools in differing built environment and demographic contexts. Berkeley Plan. J. 2006, 19, 79–95. ISPRS Int. J. Geo-Inf. 2016, 5, 150 16 of 17

11. Hou, G.L.; Huang, Z.F.; Zhao, Z.X. Research on the formed mechanism and spatial structure of urban business recreation district. Hum. Geogr. 2002, 17, 12–16. (In Chinese) 12. Liu, H.; Shen, Y.M.; Deng, X.L. Study on the spatial pattern and structure of financial service industry in Beijing. Hum. Geogr. 2013, 29, 61–68. (In Chinese) 13. Shearmur, R. The clustering and spatial distribution of economic activities in eight Canadian cities. Int. J. Entrep. Innov. Manag. 2007, 7, 223–250. [CrossRef] 14. Wall, G.; Dudycha, D.; Hutchinson, J. Point pattern analyses of accommodation in Toronto. Ann. Tourism Res. 1985, 12, 603–618. [CrossRef] 15. Xue, D.Q.; Liu, H.; Ma, B.B. Characteristics of spatial distribution of cultural industries in urban area of Xi’an city, China. Sci. Geogr. Sin. 2011, 31, 775–780. (In Chinese) 16. Yao, L.; Zhang, M.; Wang, F. Characteristics of evolution and differences of the spatial distribution among different types of creative industries in Nanjing. Hum. Geogr. 2013, 29, 42–48. (In Chinese) 17. Zhen, F.; Yu, Y.; Wang, X.; Zhao, L. The spatial agglomeration characteristics of automotive service industry: A case study of Nanjing. Sci. Geogr. Sin. 2012, 32, 1200–1208. (In Chinese) 18. Ryder, A. The changing nature of adult entertainment districts: Between a rock and a hard place or going from strength to strength? Urban Stud. 2004, 41, 1659–1686. [CrossRef] 19. Wei, Z.C.; Zhen, F.; Shan, L.; Mou, S.J.; Ming, L.B. Study on the spatio-temporal distribution patterns of cultural facilities in Shenzhen. Urban Stud. 2007, 14, 8–13. (In Chinese) 20. Liu, Y.H.; Bai, G.R. Spatial structure of public facilities for recreations and leisure activities: A case study of metropolitan Shanghai. Hum. Geogr. 2006, 21, 6–9. (In Chinese) 21. Gatrell, A.C.; Bailey, T.C.; Bailey, P.J.; Rowlingson, B.S. Spatial point pattern analysis and its application in geographical epidemiology. Trans. Inst. Br. Geogr. 1996, 21, 256–274. [CrossRef] 22. Ripley, B.D. The second-order analysis of stationary point processes. J. Appl. Probab. 1976, 13, 255–266. [CrossRef] 23. Everitt, B.S.; Landau, S.; Leese, M.; Stahl, D. Cluster Analysis, 5th ed.; John Wiley & Sons.: Chichester, UK, 2011. 24. Strickland, J. Predictive Analytics Using R; Lulu. Inc.: Colorado Springs, CO, USA, 2015. 25. Rokach, L.; Maimon, O. Clustering methods. In Data Mining and Knowledge Discovery Handbook; Maimon, O., Rokach, L., Eds.; Springer: New York, NY, USA, 2005; pp. 321–352. 26. Strickland, J. Operations Research Using Open-Source Tools; Lulu. Inc.: Colorado Springs, CO, USA, 2015. 27. Huang, Y.; Zhang, Y.; Ji, X.; Wang, Z.; Wang, S. A data distribution strategy for scalable main-memory database’. In Advances in Web and Network Technologies, and Information Management; Chang, K.C., Wang, W., Chen, L., Ellis, C.A., Hsu, C.-H., Tsoi, A.C., Wang, H., Lin, X., Yang, Y., Xu, J., Eds.; Springer: Berlin, Germany, 2009; pp. 13–24. 28. Anselin, L.; Bera, A.K. Spatial dependence in linear regression models with an introduction to spatial econometrics. In Handbook of Applied Economic Statistics; Ullah, A., Giles, D.E.A., Eds.; CRC Press: New York, NY, USA, 1998; pp. 237–290. 29. Moran, P.A. Notes on continuous stochastic phenomena. Biometrika 1950, 37, 17–23. [CrossRef][PubMed] 30. Anselin, L. Local indicators of spatial association-LISA. Geogr. Anal. 1995, 27, 93–115. [CrossRef] 31. Lee, S.I. Developing a bivariate spatial association measure: An integration of Pearson’s r and Moran’s I. J. Geogr. Syst. 2001, 3, 369–385. [CrossRef] 32. Matkan, A.A.; Shahri, M.; Mirzaie, M. Bivariate Moran’s I and LISA to explore the crash risky locations in urban areas. In Proceedings of the Conference of Network-Association of European Researchers on Urbanisation in the South, Enschede, The Netherlands, 12–14 September 2013. 33. Wilson, A.G. Complex Spatial Systems: The Modelling Foundations of Urban and Regional Analysis; Routledge: London, UK, 2000. 34. Porta, S.; Crucitti, P.; Latora, V. Multiple centrality assessment in Parma: A network analysis of paths and open spaces. Urban Des. Int. 2008, 13, 41–50. [CrossRef] 35. Produit, T. A Novel GIS Method to Determine an Urban Centrality Index Applied to the Barcelona Metropolitan Area. Master’s Thesis, Politecnico di Milano, Human Space Lab, Milano, Italy, 2009. 36. Crucitti, P.; Latora, V.; Porta, S. Centrality in networks of urban streets. CHAOS 2006, 16, 015113. [CrossRef] [PubMed] ISPRS Int. J. Geo-Inf. 2016, 5, 150 17 of 17

37. Ji, S.; Li, W.; Srivatsa, M.; He, J.S.; Beyah, R. Structure based data da-anonymization of social networks. In Information Security; Chow, S.S.M., Camenisch, J., Hui, L.C.K., Yiu, S.M., Eds.; Springer: Basel, Switzerland, 2014; pp. 237–254. 38. Waters, N.M. Spatial Interpolation I, Unit 40. 1989. Available online: http://ibis.geog.ubc.ca/courses/klink/ gis.notes/ncgia/u40.html (accessed on 1 February 2016). 39. Bowman, A.W.; Azzalini, A. Applied Smoothing Techniques for Data Analysis: The Kernel Approach with S-Plus Illustrations; Oxford University Press: Oxford, UK, 1997. 40. Levine, N. CrimeStat II: Spatial Modeling, Part III. 2002. Available online: https://www.icpsr.umich.edu/ CrimeStat/files/CrimeStatChapter.8.pdf (accessed on 5 March 2016). 41. Guidoum, A.C. Kernel estimator and bandwidth selection for density and its derivatives. 2015. Available online: https://cran.r-project.org/web/packages/kedd/vignettes/kedd.pdf (accessed on 26 February 2016). 42. Sadahiro, Y. A PDF-based analysis of the spatial structure of retailing. GeoJournal 2001, 52, 237–252. [CrossRef] 43. Li, Z.Z. Coupling analysis on business location and population distribution: The case study of Baiyin city. J. Lanzhou Univ. (Soc. Sci.) 1993, 2, 22–27. (In Chinese)

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).