Spatially explicit inverse modeling for urban planning

Abstract

Urban modeling methods have traditionally followed a forward modeling approach. That is, they use data from today’s situation to forecast or simulate future states of an urban system. In this paper, we propose an inverse modeling approach by which we shift our attention from solely forecasting or simulating future states of an urban system to steering it to a desired state in the future via key variables characterizing the system in the present. We first present a theoretical framework for the use of the inverse approach in urban planning. We test the power of the proposed method using a hedonic house price model in a metropolitan area in Switzerland to investigate the negative effects of densification on house prices. The model is calibrated by mixed geographically weighted regression in order to account for spatial variability of both key variables and model outputs. We show how devaluation of house prices caused by densification can be compensated by different levels of socioeconomic, locational as well as structural variables. We illustrate and discuss how trade-offs between variables may lead to more feasible results from an urban planning perspective. We conclude that the proposed method might be valuable for urban planners for developing implementable spatial plans based on future visions. In particular, the fact that other model specifications than hedonic house price model can also be employed to formulate an inverse model application, allows planners to address other type of problems or externalities from urbanization processes such as urban sprawls, environmental pollution or land uses change.