A new methodology for calibrating the Lowry model

Abstract

A government-funded research project was initiated to investigate the suitability and potential benefit of applying land use forecasting models in Hong Kong. Part of the project involves the development of a new technique for calibrating a Lowry model, using Hong Kong as a case study. The model contains, for each zone, a population and an employment potential that measure the relative attractiveness of the zone in the study area with respect to residents' choice of home locations and business setup (which generates employment opportunities), respectively. The calibration procedure comprises three stages. In the first stage, a genetic algorithm is employed to calibrate the population and employment potentials, together with a coefficient associated with the travel impedance function in the study area. The second stage investigates the relationships between the calibrated potentials and various land use variables, using a multivariate stepwise regression analysis. The significant land use variables are identified and equations relating these variables to the potentials are established. The third stage is model validation. A considerable error is accumulated after the first two stages of the model calibration. Since the applicability of the model is largely dependent on its level of accuracy, fine-tuning the model parameters is required to further minimize the discrepancy between the observed and the modeled distributions of the population and employment. A case study in Hong Kong is presented to illustrate the effectiveness of the proposed methodology. | A government-funded research project was initiated to investigate the suitability and potential benefit of applying land use forecasting models in Hong Kong. Part of the project involves the development of a new technique for calibrating a Lowry model, using Hong Kong as a case study. The model contains, for each zone, a population and an employment potential that measure the relative attractiveness of the zone in the study area with respect to residents' choice of home locations and business setup (which generates employment opportunities), respectively. The calibration procedure comprises three stages. In the first stage, a genetic algorithm is employed to calibrate the population and employment potentials, together with a coefficient associated with the travel impedance function in the study area. The second stage investigates the relationships between the calibrated potentials and various land use variables, using a multivariate stepwise regression analysis. The significant land use variables are identified and equations relating these variables to the potentials are established. The third stage is model validation. A considerable error is accumulated after the first two stages of the model calibration. Since the applicability of the model is largely dependent on its level of accuracy, fine-tuning the model parameters is required to further minimize the discrepancy between the observed and the modeled distributions of the population and employment, A case study in Hong Kong is presented to illustrate the effectiveness of the proposed methodology.