Analysis And Comparison Of Methods For Spatial Data And Spatiotemporal Data

In earlier development of spatial data only depends on the space distance, and a series of traditional spatial data analysis methods have been generated by the space distance. These methods not only satisfy people’s perception of things, also accord with the first law of geography, so they are widely used in many fields. But it was discovered that the spatial autocorrelation of the data cannot be ignored in the study, and then the semivariogram which can characterize the spatial autocorrelation was proposed. Kriging method is produced based on the semivariograms. This paper attempts to traditional method combined with kriging method, expand its scope of application, and make a single algorithm is improved. With a deep understanding of the data, the time dimension is taken into account, and an increasing number of space-time data into our field of vision, making the study and analysis of spatial and temporal data become a hot topic. Therefore, this article also outlines the spatiotemporal kriging. The main research contents are as follows:1. This paper discusses the traditional spatial data analysis methods, such as inverse distance weight, adaptive inverse distance weight, trend surface modeling and least squares support vector machines. These methods are summarized and laid a foundation for future research. Because of the shortage of spatial autocorrelation which do not consider in traditional methods, some kinds of kriging methods based on the semivariogram are introduced and this.2. In order to improve the prediction accuracy and expand the range of applications, the traditional method and kriging method are combined in this paper. According to the experiments, the combined methods are more accurate and practical.3. For the study of temporal data, the spatiotemporal krigingis described and applied on real data. The experiments demonstrates the necessary for the introduction of the time dimension and show that selecting the appropriated model of semivariogram and covariance functions can improve prediction accuracy of spatiotemporal kriging for specific data.