Geostatistical estimation of coal quality variables by using covariance matching constrained kriging


ERTUNÇ G., TERCAN A. E., HİNDİSTAN M. A., ÜNVER B., ÜNAL S., ATALAY F., ...More

INTERNATIONAL JOURNAL OF COAL GEOLOGY, vol.112, pp.14-25, 2013 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 112
  • Publication Date: 2013
  • Doi Number: 10.1016/j.coal.2012.11.014
  • Journal Name: INTERNATIONAL JOURNAL OF COAL GEOLOGY
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.14-25
  • Keywords: Covariance matching constrained kriging, Variogram, Block kriging, Block model
  • Hacettepe University Affiliated: Yes

Abstract

Covariance matching constrained kriging is a technique originally developed for estimating linear or nonlinear functional of the variable of interest. Practically this may be regarded as a hybrid system that considers local accuracy property in kriging and preservation of spatial variability in stochastic simulation. The method estimates the unknown vector by adding the constraint to match the variance–covariance matrix of estimated values with the variance–covariance matrix of actual values under unbiasedness constraint.

This study applies the method to a certain section of a lignite deposit which is subjected to severe tectonic activity. Lower heating value, ash content, and moisture content are estimated by covariance matching constrained kriging and ordinary kriging and also simulated by sequential Gaussian simulation. Comparisons between these methods show that covariance matching constrained kriging reproduces the spatial variability better than ordinary kriging and gives more accurate estimates than conditional simulation. Covariance matching constrained kriging can be used when spatial variability of the estimations is important as well as local accuracy.

Covariance matching constrained kriging is a technique originally developed for estimating linear or nonlinear functional of the variable of interest. Practically this may be regarded as a hybrid system that considers local accuracy property in kriging and preservation of spatial variability in stochastic simulation. The method estimates the unknown vector by adding the constraint to match the variance-covariance matrix of estimated values with the variance-covariance matrix of actual values under unbiasedness constraint.