Inference of 2D and 3D Locally Varying Anisotropy Fields

There has been a lot of interest to incorporate non-stationarity into geostatistical modeling. While first order non-stationarity can be easily considered using data transformations or considering a locally varying mean, the association of second order stationarity is less straightforward and few practical techniques exist. Current methodologies such as MPS with local pattern orientation, local kriging search orientation and kriging with LVA all deal with a varying anisotropy field. However, these techniques can be employed provided a map of local orientations and magnitude of anisotropy are exhaustively defined, called an LVA field map. The focus of this paper is to highlight a host of methodologies that can generate the required LVA map from various data sources.