The Application of DBSCAN Algorithm to Improve Variogram Estimation and Interpretation in Irregularly-Sampled Fields

Abstract

The empirical variogram is a measure of spatial data correlation in geostatistical modeling and simulations. Typically, the empirical variogram is estimated for some defined lag intervals by applying method of moments on an underlying variogram cloud. Depending on the distribution of pair-wise lag values, the variogram cloud of an irregularly-sampled field may exhibit clusteredness. Issues of noisy, uninterpretable and inconsistent empirical variogram plots are commonly encountered in cases of irregularly-sampled fields with clustered variogram clouds. An insightful diagnosis of these problems and a practical solution are the subject of this paper. This research establishes the fact that these problems are caused by the neglect of variogram cloud cluster configurations when defining lag intervals. It is here shown that such neglect hinders the optimal use of spatial correlation Authors and Affiliations

Induced Seismicity and Detailed Fracture Mapping as Tools for Evaluating HDR Reservoir Volume

“…One of the most commonly used density-based clustering algorithms is density-based spatial clustering of applications with noise (DBSCAN). This algorithm was proposed in 1996 [9] and is still widely used today [43][44][45][46]. One of the advantages of this method is the ability to extract clusters of arbitrary shape in dense regions and recognize noise points.…”