Digital Elevation Model is the foundation to analyze terrain in Geographic Information System. When processing the lost sampled points, the traditional spatial interpolation methods are ineffective and impersonal in selection the interpolation function, structures as well as its parameters. This paper applies the technology of Gene Expression Programming to generation of Digital Elevation Model. The contributions include: (1)formalizing the concepts of elevation gene, buffer and cluster, etc.; (2)proving the Depth Theorem and its lemmas; (3)proposing BS strategy, timely introducing good terrain individuals; (4)proposing the algorithms of DCSA, SCPD and ESDA to dynamically search cluster by direction and avoid computing the redundant points; proposing EF-GEP algorithm. The interpolation space is subdivided and the fitting space is reduced. (5)the experiments show that the average generation is 227, the average top fitness is 97.4527 and increased by 7.23%; 3-D Geological model visually express the configuration of geologic body.
