One of the key concerns in geographical and social sciences is to analyze and predict population migration due to its close association with urban planning, industrial upgrade and urban development. Though the most prevailing framework, gravity model, has been applied in its various versions, there is little information available about how city industry structure characteristics function as the invisible distance in the modelling of intercity population migration. Here, we introduce a family of improved gravity models by considering city industry structure proximity, complementarity and diversities. The resulting models predict population migration patterns in good agreement with the flows observed. Our best model (GM_COM) outperforms the benchmark model (GM_O) by 24.6% in terms of MAPE. Further analysis shows the improved models offer several advantages with respect to the base models. They have better prediction accuracies for flows with high intensities and long distances. And the best model demonstrates obvious improvement when flows occur in eastern China. Given the significant improvement of the proposed models, this study broadens existing study by absorbing city industry structure features into the gravity model, and deepens our understanding in the population migration as a function of distance.

intercity population migration,gravity model,city industry structure