There are times when a single stationary specification of anisotropy will not capture the complexities of a geologic site. In this situation, the anisotropy can be varied locally. The directions of continuity or even the range of the variograms can change depending on the area within the domain being modeled. There are algorithms that permit Kriging with a locally varying anisotropy model. The common approach is to generate different Kriging matrices at each location using the local direction of anisotropy from the anisotropy model. As each location is visited, the direction of anisotropy is changed according to the anisotropy at the estimation location and a new Kriging matrix is produced. This assumes that the anisotropy is constant between the location being estimated and all near-by data. The anisotropic distance does not "track" through the anisotropy model to calculate the most relevant anisotropic distance. This paper presents an algorithm to calculate the anisotropic distance using a grid of local directions of continuity. Although CPU intensive, the algorithm calculates the true anisotropic distance. This algorithm is integrated into Kriging and examples are provided.
展开▼
