The Sakura River flows in a narrow alluvial channel with gentle slope, draining the southeast area of Ibaraki Prefecture (Figs. 1 and 2). The bed consists of arcosic sand and gravel which reflect the geological condition of the upper drainage basin of the river.
The investigation of flood deposits in the upper reaches of the river shows that the flood deposits have reverse grading structure (becoming coarser grain from bottom to top of a bed) and that a sedimentary unit, the lower mud layer and the upper sand layer, is formed during a single flood. It is the purpose of this paper to explain the depositional process of the reverse graded bed from the characteristics of sediment transport in the Sakura. Main results are as follows.
The flood deposits are thick only on the cut-sides of the river channel, which are attacked by the main flow of the flood (Fig. 6). Levees are better developed in these areas. Medium and fine sand make up the thick flood deposits on the natural levee (Fig. 7). It begins with the deposition of a thin silt layer and is followed by a thick sand layer which displays no internal structures. Sand sizes, φ₁ as well as φ₅₀, gradually increase toward upward. Fig. 8 shows the repeated cycles of coarsening upward sequence of flood deposits on the natural levee. The maximum diameter of each flood deposit has the upper limit of about 10(0. 5mm) . The floods which left the thick flood deposits are shown in the time-stage curves of Fig. 9. Reverse grading is also developed among the thin flood deposits on the lower part of the flood plain (Fig. 7). The lowest clayey silt layer with a thickness of a few milimeters suddenly changes the upper sand layer which usually has fine lamination.
Fig. 9 shows the observational results on the hydraulic regimes and the suspended sediment transports in some floods in the Sakura River. It is the common fact to all observations that the maximum concentration of suspended load precedes the peak discharge. Furthermore, it is clear from Fig. 10 that the amount of the finer sediment load of silt-clay and very fine sand is extremely high at the leginning of each flood and rapidly decreases with the rising of waterlevel.
The depositional process of reverse graded bedding in flood deposits can be explained as follows and Fig. 12 schematized them.
1. Deposition of the mud layer (Fig. 12-I)
It is very likely that the lower muddy part of reverse graded bed is deposited during the beginning of the flood when the concentration of finer grained suspended sediments, the socalled “wash load”, is extremely high. Flood water with large quantities of wash load is diverted into the flood plain and the stagnation of flows on the flood plain causes the deposition of this finer sediments. It can be considered that the reason why mud layers on the natural levee are very thin and rather indistinct (Figs. 7 and 8) is that the natural levee is not covered by the inundated water at this stage. Boils along the bank begin to occur in the mean while.
2. Deposition of the sand layer (Fig, 12-II)
The upper sandy part is presumed to be produced by the deposition of suspended sand after the rapid decrease of wash load concentration. The upward coarsening structure in this sand layer may be produced under the condition of gradual increase of flow intensity. The sand make up the natural levee is coarser than that of suspended up to the bankfull level at the center of the channel (Figs. 7, 8 and 11). This fact is explained as follows. Boils over the bank become more and more violent. This suggests that the upward flows (kolks) near the bank become violent with increases of discharge. It is supposed that these kolks cause the coarser sediment grains to suspend near the bank.
3. Preservation of the reverse graded bed (Fig. 12-III)
With the beginning of water level falling, the inundated flow turns back from the flood plain to the channel.

View full abstract

Since 1960's studies of the distribution and the spatial structure of retail establishments in metropolises have been carrledi on in Japan, but the classification of retail locations and the spatial structure of retail trade have scarcely been examined. Particularly, there have been few studies on the locational classification of retail shops and the components of spatial retail structure.
The aim of the present article, therefore, is to examine the classification of the retail locations and the spatial retail structure in the metropolises, making a comparison of Fukuoka City with Sapporo City. Data used in this article are the number of shops which are classi fied into thirty retail groups derived from the Census of Commerce for 1976 (Table 1), and the areal units (statistic units) are based on the census tracts as of 1975. The results are follows.
1. The results of the calculation of localization coefficients for each retail group in bothh cities reveal that every retail group in Sapporo City is a little more localized than in Fukuoka City, but the distributions of the retail groups in both cities show generally similar patterns. and retails groups are roughly classified into four types (Table 2) . The calculation of the specialization coefficient of each retail group by zones (city core area, built-up area, suburbs, and outer zone) in each city (Table 1) suggests that the locational types of retail shops can be classified into three groups and general food retail. As a result of above two analyses in ref erence to the distribution of retail establishments, it has come to be clear that the locational types of the retail groups in both cities are composed of “centripetal retail groups”, “dispersedd retail groups”, “street-oriented retail groups” and “centrifugal retail groups”. The distributionn of “centripetal retail groups” decreases with the distance from the city core area, while “dispersed retail groups” are found uniformly throughout the city. The distribution of “streetoriented retail groups” is concentrated in the area along the main streets in the suburbs and outer zone, and the distribution of “centrifugal retail groups” (supermarkets) increases with. the distance from the city core area.
2. As a result of grouping the retail stores by the R-mode cluster analysis (Figs. 1 and 2) the linkage tree of the retail groups in Fukuoka City is similar to that in Sapporo City, though the termination of clustering in Sapporo City is earlier than in Fukuoka City, and the similarity in distribution between retail groups is more remarkable in Sapporo City than in Fuku oka City. The grouping of the retail stores in both cities is composed of two main groups (sub-cluster) and five isolated groups (motor vehicle retail, bicycle retail including motorcycle and agricultural articles retail, etc.). The two main groups are follows : “centripetal retail groups” (i.e. department store, ladies' and children's dress retail, and glass and ceramics, retail, etc.), and “dispersed retail groups” (i.e. sea food retail, vegetable and fruit retail and hard and kitchen-ware retail, excluding agricultural equipments, etc.).
3. The spatial retail structures in both cities were examined by the use of the factor analysis (principal axis method). As a result, four dominant factors with eigenvalue greater than 1.00 were extracted in Fukuoka City and two in Sapporo City (Table 3). The first and second factors in Fukuoka City are similar to those in Sapporo City, and these two factors which are labelled as “higher central function” and “ubiquitously located function”, respectively are main components which yield the locational differentiation of retail establishments in both cities. On the other hand, however, there are some other retail groups which don't correspond to this factor structure.

View full abstract

The study of innovation diffusion has a proposition that innovation spreads through a system of cities. A recent expansion of this proposition has led to a view that the study of innovation diffusion is a branch of urban systems research analyzing the relationships between elements constituting the system of cities (Van der Knaap, 1980) . However, this proposition is not sufficiently confirmed to support such a view. Therefore, this paper examines an empirical adequacy of the above proposition, taking a case of the electric supply companies established during the years of 1887 to 1898 in Japan. This period is the first stage in the growth of Japanese electric supply industry, and is generally known as the intra-urban supply stage, when the service areas of electric supply companies were limited to the cities concerned and their surroundings. Although thermal power generation was first adopted, the soaring price of coal and the technological development gradually made them adopt water power generation.
The first electric supply company of Japan, or the Tokyo Electric Supply Company, started its business in the year of 1887. The propaganda and support made by the Tokyo Electric Supply Company promoted the earlier opening of electric supply companies in such regional centers as Kobe, Ohsaka, Kyoto, Nagoya, Yokohama, Kumamoto and Sapporo. Later, the electric supply companies spread among 48 cities (Fig. 1). As regards the empirical regularity, the correlation (r=-0. 621^**) is obtained between the year of opening in the city concerned and its population, suggesting the hierarchical effect (Fig. 2). This would have resulted from such factors as market potential and capital availability as well as information flow.
The resultant distribution of electric supply companies as of 1898 shows that more companies were established in Western Japan than in Eastern Japan. Furthermore, if all the nonadopter-cities with more than 20, 000 population are taken into account as potential adoptercities, this pattern should be obvious. Relevant literature suggests that there are two other factors, in addition to the hierarchical effect, which may account for the pattern. The first factor is the type of electricity generation, whether thermal power or water power. The second factor is accessibility to (cotton) spinning companies which had to generate their own electricity for lighting the factories. Based on these discussions, the following two statistical analyses were attempted using relevant variables: (1) a test of significance of difference between means of the variables for adopter-cities and non-adopter-cities both with more than 20, 000 population (Table 2); (2) a forward stepwise regression analysis of the year of opening for all the adopter-cities (Table 3). Those statistical analyses have revealed that “availability of coal” and “existence of (cotton) spinning companies” are not significantly related with the opening of electric supply companies, and that the adopter-city's characteristics and the year of its adoption are explained by “population” which is a surrogate for urban size and “distance from the nearest adopter-city” or “distance from the nearest larger city” which is a surrogate for spatial arrangement of cities.
Then, 36 adopter-cities with more than 20, 000 population out of 48 adopter-cities were chosen to examine the diffusion process with reference to the system of cities. In order to define the system of cities operationally, the spatio-hierarchical position of each city within it was identified by measuring the spatial autocorrelation of population in the interaction space. First of all, the two-dimensional interaction space is recovered by applying KYST-2A (Kruskal et al., 1978) to the data on interurban flows of bank remittance bills among 57 major cities (Fig. 3) .

View full abstract