Spatial Analysis of Population Distribution by Employment Sectors in Tokyo Metropolitan Area

Transcription

1 1 Spatial Analysis of Population Distribution by Employment Sectors in Tokyo Metropolitan Area MONZUR, Tawhid Abstract This research is focused on analyzing the urban spatial structure of the Tokyo metropolitan area by using Exploratory Spatial Data Analysis (ESDA) techniques and aims to observe the effectiveness and acceptability of ESDA in identifying the spatial clustering of population distribution by employment sectors in such a large area. The ESDA techniques that have been selected for this research are Global Moran s I and Local Moran s I. The population density data has been selected for this analysis which is extracted from the 2000 population census database. The population data has been manipulated and tabulated by using Excel statistical software. ArcGIS 10.1and GeoDa software have been used to project and analyze the manipulated data. Population number has been converted to population density to exclude the influence of the census unit area sizes. A null hypothesis (The population distribution in Tokyo is showing no spatial clustering pattern) has been tested through the analyses. The analysis showed strong evidence against the null hypothesis which means the population distribution in Tokyo is showing a spatial clustering pattern. Both Global Moran s I and Local Moran s I analyses identified specific spatial pattern types of the population distribution in Tokyo. The Local Moran s I further divided all the population density neighboring values into five clustering groups (HH, LL, LH, HL, NS). The results show that most of the values are located in the HH clustering group which means that the density of population in one observed area is very high and are also surrounded by the areas with high density of population. The analysis

2 2 also located the outliers (HL and LH types) which are high populous areas surrounded by low populous areas or vice versa. Besides, the statistical significance tests through 95% to 99.9% level of confidence identified the exact locations of the spatial clustering of population density in Tokyo. The ESDA analysis pointed out specific locations of clustering types that cannot be observed only from the normal distribution map. Different significance level tests pinpoint the exact locations of the clustering of high population concentration besides the outliers of the area. Moreover, this study proved that huge data sets can be analyzed through the ESDA techniques. Keywords: GIS, Tokyo, spatial structure, population, employment Introduction Cities can be regarded as the complex of socio-economic and physical spaces. Understanding the spatial structure or the framework of land use and socio-economic patterns of the complex yields insights about economy-wide growth processes and hence provides the knowledge and tools for policymakers in city management. Studying the spatial structure of urban areas, especially that of large urban areas is becoming more important than ever because of the current pace of urban growth. Many researchers from diverse fields are trying to study/understand and identify the spatial pattern of the urban areas. This research has selected Tokyo; firstly, it is the heart of the country, sharing only 8.5% of the land in Japan but yields 40% of the national GDP. Secondly, the urban management plan of Tokyo is praised because of its uniqueness and effectiveness (Hein, 2010). Thirdly, the urban development process of Tokyo experienced multiple political, economic and environmental upheavals; but still carries the fame of sustainable urban form

3 3 attainment (Cho, 2011; Morita et al., 2012). Henceforth, understating Tokyo could not only contribute to the urban and regional planning of this region, but also will provide a role model for the emerging Asian cities as well. Asia Pacific countries are facing the problem of proper urban planning. Besides, most of the metropolitan areas located in the Asia pacific are beset with which intensifies social, environmental and economic issues which is referred as the The developing country phenomena [Bugliarello, 1999]. New emerging cities in Asia pacific regions are more complex in nature and difficult to study by using the conventional methods of urban studies. Advancement in spatial econometrics, Exploratory spatial data analysis (ESDA) is used for urban structure analysis is well performed for the cities includes Los Angeles USA, Ile-de-france France, Singapore, Jakarta Metropolitan Area (JMA) Indonesia, Alexandria Egypt, Rome Italy, Athens Greece, Hermosillo Mexico and Dijon France. By further research, using the ESDA method the spatial structure of Tokyo can be unveiled and it will contribute to a better urban policy-making for the region and serve as a reference to the emerging megacities of Asian which need proper urban planning and policy as well (Uchiyama and Okabe, 2012) This research aims to explore the methods to understand the spatial structure of Tokyo. It uses Geographic Information System (GIS) and applies Exploratory Spatial Data Analysis (ESDA) techniques for the purpose. 2. URBAN SPATIAL STRUCTURE STUDIES FOR TOKYO Spatial structure of Tokyo has been studied from different perspectives. It has been analysed based on historical review; categorization of the important events; location and amenities preference; employment distribution and inter-urban population distribution/migration (for example, see, Ichikawa, 1994; Watanabe, 1972, Watanabe et al., 1980; Kikuchi and Obara, 2004; Sorensen, 2001a, 2001b; Okata and Murayama, 2011; Pernice, 2007; Tonuma,1998; Hein,

4 4 2010). Tokyo has a good transportation network which connects the urban and suburban locations and enables city population to commute from long distance. Decentralization as stated in the paper of An (2008) is one of the main governmental policies which has been emphasized for reducing the concentration within CBD (Central Business District) areas. To get rid of the negative urban externalities, government plans seemed successful for the development of the CBD areas but some research criticize that concentration outside the CBD areas get worse which became one of the prominent issues the government of Tokyo has to deal with (Sorensen, 2001a, 2004). Several urban policies have been implemented but population concentration surpasses all the regulations. Tokyo spatial structural studies can be found in the papers where the main focus was personal income, transportation cost, land price fluctuation and land use change for agricultural purpose which has influenced the spatial pattern of Tokyo (Fujita and Kashiwadani, 1989; Zheng, 1990, 1991; Inoue et al., 2007; Kikuchi and Obara, 2004). Moreover, A Global Moran s I technique based analysis has been performed in Tokyo to find out the urban land use pattern (Zhao and Murayama, 2005, 2006). However, it needs further research because Global Moran s I failed to identify the local level distribution. A Local Moran s I with k-order neighbours used to analyse the distribution of elderly people in Ichikawa City (Murayama, 2011), where it is only a small part of Tokyo metropolitan area. Moreover, a Global Moran s I analysis on ranking the world megacities based on the spatial distribution of population needs further research which referred Tokyo as category A means high concentration of population within the Tokyo central areas or not spread out (Uchiyama and Okabe, 2012). A grid cell based analysis also on the population distribution results seem incomplete and rather restricted from the side it showed only the normal

5 5 distribution of the population and its dependency over land use changes to understand the urban growth of Tokyo (Bagan and Yamagata, 2012). Articles related to Tokyo spatial structure are large in number but most of which focus on modelling the factors that affect the distribution and lack of a visual understanding. Only very small number of research tried to understand the spatial structure itself, however, they failed to study the whole metropolitan region. This research, though only focusing on the population distribution, will explore the methods for understanding the spatial structure of the entire Tokyo metropolitan area by applying ESDA techniques on GIS to give a clearer view of the urban spatial structure. By further research, the spatial structure of Tokyo can be unveiled and it will contribute to a better urban policy-making for the region and serve as a reference to the emerging megacities of Asian which need proper urban planning and policy as well (Uchiyama and Okabe, 2012). 3. RESEARCH APPROACH AND METHODOLOGY The research aims to analyze the spatial distribution of Tokyo metropolitan area in 2000 by using the ESDA techniques. The GISA (Global Indicator of Local Association) and LISA (Local Indicator of spatial Association), the two branches of ESDA techniques has been used for studying the spatial structure of many western cities which lacks in the case of Tokyo from the sense to understand the spatial association and spatial heterogeneity. In this case, the ESDA techniques have been implemented to demonstrate the acceptability as well as performance in identifying the urban structure of Tokyo.

6 6 3.1 TOKYO METROPOLITAN AREA This research focuses on the Tokyo metropolitan area, including the 23 inner words and the neighboring prefectures: Saitama, Kanagawa, Chiba, Gumma, Ibaraki, and Tochigi. It is an area of 32,236km2 and is referred as the largest mega metropolitan region in the world. Tokyo shares 8.5% of the land in Japan and has a population of 42.6 million mostly concentrates around Tokyo Metripolitan Areas. It is the economic hub of Japan with a high rate of population concentration. Since the fact that concentration of the population distribution is high up to 70km from the central business district (CBD), some parts of the Gumma and Tochigi prefectures are excluded from the analysis (figure 3.1 and 3.2).

7 7 Figure 3.1. Population Density of Tokyo, 2000 (Natural Breaks)

8 8 Figure 3.2. Population density map up to 70 km range, DATA This research uses population census data which is carried out by the Ministry of Internal Affairs and Communications (MIC) every 5 years. The population census dataset contains more than one hundred of variables, but for the research purpose, only the population data has been

9 9 manipulated by using the Excel statistical software. The boundaries of census area units are provided in an ArcGIS shape file as also can been seen from Figure 3.1. Since the areas of census units (polygons) are different, this study further divided the study area into 200 meter by 200 meter cells with population evenly allocated to each cell according to the total population number within the polygon. The figure shows the distribution of the population in Tokyo after the treatment. It contains 2, 46,458 cells in total. From the figure 3.2 and 3.3, it can be observed that after treating the density dataset the locations of the high density are clearly visible. Figure 3.3. Population density Map (per ha), 2000

10 10 For the tabulation and manipulation of the extracted data, Excel statistical software has been used. For the selected research, ESDA techniques have been used. The two types of spatial statistical methods of ESDA techniques have been used for this research study. From the Global spatial statistical methods, Global Moran s I has been selected. From the local spatial statistical methods, Local Indicator of Spatial Association (LISA) which includes Local Moran s I toolset has been selected for this research. The spatial statistical toolsets have been analysed by using the ArcGIS 10.1 and GeoDa software. 4. EMPIRICAL RESULTS 4.1 GLOBAL MORAN S I The Global Moran s I or also called the global spatial autocorrelation measures the linearity of the observed value to its neighbouring values which determines in identifying the spatial distribution pattern (Dispersed, Random and Clustered) of a certain phenomenon. The dispersed spatial pattern means that each value from its neighbouring values is located far from each other in a uniformed manner. The Random spatial pattern means the distribution of the values is homogenous or independent in nature. The Clustered spatial pattern means most of the values are concentrated to nearby locations or adjacent together. The Global Moran s I can be written as follows (Viton, 2010): ( ) Where, S = number of observations

11 11 = sum over all i and j of = spatial weight between i and j. = weight * cross product terms. The Global Moran s I Index for the Tokyo analysis based on the Rooks contiguity (Getis, 2009) is and this significantly implies that the population density in Tokyo is showing a strong spatial autocorrelation. More specifically, the Global Moran s I Index proves that the distribution of the population in Tokyo tend to have a clustered spatial pattern. Moreover, The Z-score is as high as while higher than can show enough that there is a 99% chance that the data is taking a clustered pattern (Figure 4.1). The Moran s I Index for the 10 selected employment sectors by population distribution shows a significant strong spatial autocorrelation (Table1). More specifically, the Global Moran s I Indexes prove that the distribution of the population by employment sectors in Tokyo tend to have a clustered spatial pattern. Moreover, The Z-scores for each employment sectors are quite high while higher than can show enough that there is a 99% chance that the data is taking a clustered pattern. Table 1: Moran s I Index and Z score for each employment sectors Employment Sectors Moran s I Index Z-score Agriculture Construction Maker Finance

12 12 Infrastructure Office Real State Retail Service Traffic Figure 4.1: Global Moran s I output Global Moran s I analysis gives an overall result of what kind of pattern the data is imposing. However, Global Moran s I analysis alone cannot provide enough evidence to identify the specific patterns of spatial differences. The output fails to provide evidence against whether the high density locations and low density locations are separately located or

13 13 not. To find out the high and low density locations and also the dissimilar locations, LISA techniques have been used which is further discussed in next section. 4.2 LOCAL MORAN S I Global Moran s I statistical method, found strong evidence that a clustered pattern is existing in Tokyo in terms of population density but fails to identify what type of clustering and where exactly the spatial association exists. On the other hand, Local Moran s I, one of the techniques of LISA calculates each location separately. After that, the separated values are added up to get the Global Moran s I. The Local Moran s I identifies the specific locations that are having significant positive or negative autocorrelation. The local Moran s I can be written as follows (Anselin, 1995)- Where, is the variance. The Local Moran s I analysis divides all the values into five separate groups which can give a clear understanding of where the similar and dissimilar values are having a statistical significant clustering. Local Moran I not only identifies the high and low values but also address the outliers of the values. Insignificant (i.e. no significant clustering is existed); high-high group (i.e. high values are clustering with other neighbouring high values and are above the mean); Low-Low group (i.e. low values are clustering with other low values and below the mean); high-low group (i.e. high values clustering with low values and are below the mean) & low-high group (i.e. low values are clustering with high values and are below the mean). The first 2 groups (HH and LL) signify a positive spatial

14 14 autocorrelation or in other words spatial clustering (Baumont, et al, 2004; Anselin, 1995; Griffith & Wong, 2007). The 3rd and 4th group (HL and LH) signifies a negative spatial autocorrelation or also referred as the outliers or dissimilar values. Figure 4.2 shows the 5 types of clustering in Tokyo. The insignificant clustering group or pattern contains the highest number of 1,60,234 areas (cells), the high-high clustering group or pattern contains about 21,637 cells, low-low clustering group or pattern contains 59,460 cells, low-high clustering group or pattern contains 166 cells and high-low clustering group or pattern contains 4 cells. The outcomes are also statistically significant at 0.05 confidence level. Figure 4.2: The 5 types of Clustering The Local Moran s I analysis found out the specific locations of the high density areas or hotspots and low population density areas or coldspots in Tokyo. Besides, the analysis also identified the outliers or dissimilar locations adjacent to high population density areas. The Local Moran s I identified clustering patterns in local level.

15 15 The Local Moran s I analysis (Figure 4.3) found out the specific locations of the high density areas or hotspots and low population density areas or coldspots in Tokyo. Besides, the analysis also identified the outliers or dissimilar locations adjacent to high population density areas. The Local Moran s I identified clustering patterns in local level. Figure 4.3: The 5 types of Clustering in 0.05, 0.01 and significance level.

16 16 5. CONCLUSION This paper contributes to the urban spatial structure studies through using the ESDA statistical techniques in identifying and understanding the population distribution pattern of Tokyo. The outcomes of the research prove that the method is applicable in understanding the spatial structure of such a large region. The ESDA techniques revealed that the distribution of the population in Tokyo is clustered which means existence of spatial heterogeneity. They also specified the locations with high and low density as well as identified the locations of dissimilar areas. The clustering type can further help to analyse the centers and sub centers of the Tokyo area as well as employment density. Also from the analysis it can be observed that the population density is monocentric but it still needs further analysis. Several papers have studied the spatial structure of Tokyo by using the ESDA techniques restricted to the global level rather than local. Local Moran s I identified different clustering types which is necessary to understand the spatial structure of a city. For this research we have used a grid 200X200m, but it could be different if the range is expanded. For the further studies the size of the grids will be expanded to 500 up to 2000 meters to see the difference. Moreover, the research has used only the population density as for the preliminary research which cannot draw a conclusion about an area s spatial structure. In further research the other population and employment variables will be manipulated and analysed to see why the HH type of clustering is monocentric and what made the dissimilar clustering types exist within the HH clustering types. Further research also aims to use distribution of employment, transportation and population up to 2013 to see the change in spatial pattern. The research study can be a blueprint for the development of the urban policy for the Asia pacific region. The methods and findings of the research study can help the newly emerged and growing urban cities in the developing countries in the Asia

17 17 pacific region to come up with better and effective urban planning by specifying and identifying the exact urban problems responsible for the urban decay. Moreover, the research study can assist in improving the ongoing urban problems by pinpoint the intensity of the urban problems. Nevertheless, the research study can help the TMR to be more sustainable for the side of urban management as well as can be a role model for the other emerging cities around the globe. REFERENCES Anselin, L. (1995). Local indicators of spatial association LISA. Geographical analysis, 27(2), An, S. K. (2008). Recentralization of Central Tokyo and Planning Responses. Journal of Regional Development Studies, Bagan, H., & Yamagata, Y. (2012). Landsat analysis of urban growth: How Tokyo became the world's largest megacity during the last 40 years. Remote sensing of Enviornment, 127, Baumont, C., Ertur, C., & Gallo, J. l. (2004). Spatial Analysis of Employment and population Density: The Case of the Agglomeration of Dijon Geographical Analysis, 36(2), Bugliarello,G.(1999). Megacities and the developing World. ( G, Bugliarello,Ed.) The bridge, 29(4), Cho, S. (2011). Urban transformation of Seoul and Tokyo by logal redevelopment project. ITU A/Z, 8(1), Fujita, M., & Kashiwadani, M. (1989). Testing the Efficiency of Urban Spatial Growth: A Case Study of Tokyo. Journal of Urban Economics, 25, Getis, A. (2009). Spatial Weights Matrices. Geographical Analysis, 41(4), Griffith, D. A. (1992). What is spatial autocorrelation? Reflections on the past 25 years of spatial statistics. In: Espace géographique, Hein, C. (2010). Shaping Tokyo: Land Development and Planning Practice in the Early Modern Japanese Metropolis. Journal of Urban History, 36(4), Ichikawa, H. (1994). the evolutionary process of urban form in Edo/Tokyo to The town planning review, 65(2),

18 18 Inoue, R., Kigoshi, N., & Shimizu, E. (2007). Visualization of spatial distribution and Temporal change of land prices for residential use in Tokyo 23 wards using Spatio-Temporal Kriging. 10th international conference on computers in urban planning and urban management, 63, pp Tokyo. Kikuchi, T., & Obara, N. (2004). Saptio-temporal changes of urban fringe in Tokyo metropolitan area. Geographical reports of Tokyo metropolitan university, Morita, T., Nakagawa, Y., Morimoto, A., Maruyama, M., & Hosokawa, Y. (2012). Changes and Issues in Green Space Planning in the Tokyo Metropolitan Area: Focusing on the "Capital Region Plan". International Journal of Geomate, 2(1), Murayama, Y. U. (2011). Testing local spatial Autocorrelation using k-order neighbours. In &. R. Yuji Murayama, Spatial Analysis and Modeling in Geographical Transformation Process: GIS based application (pp ). London: Springer. Okata, J., & Murayama, A. (2011). Tokyo s Urban Growth, Urban Form and Sustainability. In Megacities, Pernice, R. (2007). Urban Sprawl in Postwar Japan and the Vision of the City based on the Urban Theories of the Metabolists. Journal of Asian Architecture and building Engineering, 6(2), Sorensen, A. (2001a). Building suburbs in Japan: continuous unplanned change on the urban frindge. TPR, 72(3), Sorensen, A. (2001b). Subcentres and Satellite Cities: Tokyo s 20th Century Experience of Planned Polycentrism. International Planning Studies. International Planning Studies, 6(1), Sorensen, A. (2004). Major issues of land management for sustainable urban regions in Japan. Towards Sustainable Cities. Towards sustainable cities, Tonuma, K. (1998). Tokyo: Policies toward the 21st century. Ekistics, Uchiyama, Y., & Okabe, A. (2012). Categorization of 48 Mega-Regions by Spatial Patterns of Population Distribution: The Relationship between Spatial Patterns and Population Change. 48th ISOCARP congress. Viton, P. A. (2010). Notes on Spatial Econometric Models. City and Regional Planning, 870(3), Watanabe, Y. (1972). Some aspects of recent Japanese metropolitan growth. Geographical reports of Tokyo metropolitan university,

19 19 Watanabe, Y., Takeuchi, K., Nakabayashi, I., & Kobayashi, A. (1980). Urban growth and landscape change in the Tokyo metropolitan area. Geographical Reports of Tokyo metropolitan University, Zheng, X.-p. (1990). The spatial Structure of Hierarchical inter-urban system: equilibrium and optimum. Journal of Regional Science, 30(3), Zheng, X.-P. (1991). Metropolitan Spatial Structure and its Determinants: A Case-study of Tokyo. Urban Studies, 28(1),