Abstract
The study of spatial patterns of urban activities has been approached from two main points of view, factorial ecology and urban economic geography, whose major empasises are quite different. Factorial ecology pays particular attension to the spatial distribution of socio-economic characteristics of urban residents based on a nighttime urban structure. Urban economic geography, on the other hand, analyzes the areal arrangement of economic activities within a daytime urban structure.
In their approaches to spatial analysis and pattern specification, two defects are common to both points of view:
1) Quasi-subjective identification of spatial patterns, and
2) failure to integrate diverse aspects of urban activities according to their time-serial occurence.
The present article is a search for synthetic explanation of spatial structure of diverse urban activities, by applying spatial autocorrelation concepts to obtain pattern identification and integration. The research is divided into three major phases, outlined below:
The dynamic process of spatial patterns produced by urban activities occurs, and should be investigated, at various time. In this paper, the annual time series and the 24-hour cycle constitute the two major time spans. The annual time series is used in reference to changes in spatial patterns of economic activities, and the 24-hour cycle refers to the inter-transformation of spatial patterns of quotidian urban activities, appropriately represented by daytime, nighttime and intermediate population characteristics.
b) Spatial autocorrelation concepts, defined as spatial interdependency based on contiguity, are very useful in undertaking exact spatial pattern analysis. The coefficient I, devised by Moran and improved by Cliff and Ord, is used for test of pattern. I is evaluated from I=(n/W)∑(2)WijZiZj/n∑iZi2(i≠j)……(1) Spatial interdependency, which is significant as underyling spatial structure, is expressed in the weight matrix.
c) Combining the two above analysis, we can proceed to an explanation of spatial-temporal processes of patterns. They are modeled on the independent variables of spatial and/or spatial-temporal lag components, which are equivalent to the received influences of each cell from the neighbors and evaluated from the most valid weight matrix in the test of spatial autocorrelation.
The data were collected in Osaka City, the second largest economic metropolis in Japan, a city that shows a remarkable transformation of spatial patterns of urban activities within 24 hours. All the data were compiled into 1, 010 meshes covering whole city, each mesh measuring 500m×500m.
1) Spatial Patterns of Daytime Activities
The principal component analysis consists of 26 economic activity variables for 1975.From the results obtained, the first component can be identified as an economic activity agglomerated in the central zone of the city. (Tab. 1) The second component is activity densely distributed in the inner city zone. The two components show clear centered and ring-like patterns respectively. (Fig. 2 and 3) Both patterns were examined by testing spatial autocorrelation based on isotropic weight (QUEENG). How-ever, to improve the power of the test, an attempt was made to construct a center-to-periphery type of weight, which is properly named as ‘diffusional’. My newly devised ‘diffusional’ weight confirms more clearly the underlying spatial structure of both patterns, which have a strong directional bias. (Tab. 3. a, b).
Statistical explanation of the observed patterns is attempted by using the spatial lag components calculated from the most effective ‘diffusional’weight matrix.
