Efficient computational methods for iterative cokriging

Cokriging is a powerful tool for geostatistical parameter identification. The unknown parameter field, e.g. hydraulic conductivity, is considered a stationary random space function, which then is conditioned on observations of dependent quantities, such as hydraulic head and the arrival time of conservative tracers. Discretizing the parameter field by many element-, cell- or point-related values, the underlying problem is underdetermined since only few measurements are available. The problem of under-determination is overcome by the introduction of a priori knowledge, allowing a rigorous uncertainty analysis in the Bayesian framework. Cokriging is, however, often restricted by its computational costs. We show how to increase the computational efficiency of iterative cokriging by using a combination of both well-known and newly developed mathematical methods.