Abstract: Based on in-depth analysis of the temporal and spatial characteristics of the geographical entity, by abstracting and standardizing the geographical entity,this paper puts forword a reasoning prediction method according to the characteristics of complex geographical entity. The method focused on the predictive inference theory, model method and procedure of RBF network learning method based on temporal information of geographic entities. Through the abstract description and standardization of geographical entity object, hypersurface temporal information fusion is established. Using the strong nonlinear fitting ability of RBF neural network, a prediction and inference model is constructed with integration of a complex geographical entity with time, space and attribute information. And The method, by predicting the content of the dissloved oxygen (DO) in Lijiang River in Yangshuo section as an example, is proved to be feasible. This method provides an effective way for the intelligent decision and reasoning of spatiotemporal data processing.

Key words: spatiotemporal data, hyper surface, RBF, data reasoning