EMLK2D: a computer program for spatial estimation using empirical maximum likelihood kriging

Abstract

The authors describe a Fortran-90 program for empirical maximum likelihood kriging. More efficient estimates are obtained by solving the estimation problem in the ‘Gaussian domain’ (i.e., using the normal scores of the experimental data), where the simple kriging estimate is equivalent to the maximum likelihood estimate and to the conditional expectation. The transform to normality is done using the empirical cumulative probability distribution function. A Bayesian approach is adopted to ensure a conditionally unbiased estimate, which is obtained as the mean of the posterior distribution. The posterior distribution also provides a complete specification of the probability of the variable and thus provides the basis for a more realistic evaluation of uncertainty by various methods: inverting Gaussian confidence intervals, confidence intervals measured from the posterior distribution, variance measured from the posterior distribution or intervals obtained using the likelihood ratio statistic. A detailed case study is used to demonstrate the use of the program.