ESTIMATING THE UNCERTAINTY OF KRIGING ESTIMATES: A PRACTICAL REVIEW AND THE PROPOSAL OF TWO NOVEL-APPROACHES

Estimando a incerteza das estimativas de krigagem: uma revisão prática e a proposta de duas novas abordagens

Abstract

This paper presents two novel kriging-based methods to measure the uncertainty associated with geostatistical estimates. Both approaches combine the Kriging Variance (KV), which is a good summary of the spatial configuration of the data, with a second component which locally measures the data dispersion. The first proposed method, referred to as the Combined Index (CI), merges KV with the local conditional distribution estimated by median Indicator Kriging. The other method is named Combined Variance II (CVII) and it is an adaptation of the originally proposed Combined Variance, where the equation used to merge KV and the Interpolation Variance was replaced. The two proposed methods are compared to other approaches available in the literature in a case study using the widely known Walker Lake dataset. The results indicate that both proposed methods outperformed the other tested solutions. The CVII index outperformed its original formulation in all the tested scenarios.