Abstract: Concrete arch dam, as a highly statically indeterminate structure, has strong adaptability and integrity. Traditional statistical models for concrete arch dam mainly focus on the deformation of a single point rather than the interaction among different points. In this article, the spatial autocorrelation of errors of different points at the same instance was mined by using spatial econometric method. Furthermore, the errors were modified by using the spatial regression model, and hence a spatial error model for concrete arch dam was established. As an example, the deformation sequences of 34 monitoring points in Xiaowan arch dam were studied and the results show that the error panel data are of strong positive correlation, and the spatial error model is superior to traditional statistical models.

Key words: concrete arch dam, spatial panel data, spatial error model, spatial autocorrelation, deformation forecast