Geostatistcs has been an important tool for spatial analysis recently, it has been applied many fields of natural science, such as geography, ecology, soil science and so on. The paper discusses the progress in the theory method of Geostatistics. The whole framework has been pointed out, including estimation, local uncertainty, stochastic simulation and multipoint simulation. Especially the new methods, for example Sequential Gaussian Simulation, LU Simulation, P-field simulation, Annealing Simulation are described. The paper analyzes the difference between multipoint simulation with the traditional geostatistics methods. That is the breakthrough of two-point methodvariogram to training image to describe the spatial structure of objects. Meanwhile, the paper makes a comment on the software of Geostatistics. The need of graphic user's interface, high quality image input and output, together with the absence of spatial analysis function, caused the combination of Geostatistics and GIS is an inevitable trend.