集計的立地・除却確率に基づく都市圏土地利用モデルについて

抄録

To continue our efforts to construct a land use simulation model for a metropolitan area, this article is focused upon modeling of two types of probabilities which play key roles in determining spatial allocations of urban activities.
The location probabilities are the proportions of land occupied by each activity. Such probabilities can be considered either in terms of stocks or of stock increases. Although the latter is desirable in operational models, it is practically difficult to obtain stable results unless long-term data bases are available. Hence, we here employ the former framework to calibrate the model, which is expressed as an aggregate multinomial logit model concerning 34 land consuming activities and the reservation demand. The stock probabilities are then to be transformed into the latter ones through difference operations.
The other types of probabilities are the demolition probabilities, which are defined as the propotions of land to be released by the activity to the one retained by it in the previous period. Although such probabilities are to be considered in terms of land, we instead employ the “locational entities”, either values of production capitals, numbers of employees or housing units, due to lack of credible land use data. Since it is difficult to calibrate models on the activity basis, activities are classified into eight groups to establish binomial logit models.
With those probabilities, the proposed simulation model on metropolitan land use is completed. The entire model consists of three major model blocks, viz., the regional frame, activity and location models, and the results of two separate experimental simulations are also shown to illustrate the performance of the location model, which includes the above probabilities. The one is the total simulations, and the other is the flow simulations with all the variables to be determined in the location model given exogenously. Our study area is the Kanto Region, which is subdivided into 51 zones, and the simulations are made on the annual basiss starting from 1976 and ending in 1980.