Conditional Simulation: Practical Application for Sampling Design Optimization

Abstract

Detailed spatial models generated by conditional simulation provide a powerful tool for case-specific optimization of sampling designs. The entire process of sampling, estimation, and decision can be simulated on such a model by a Monte-Carlo approach. Optimization can be based on economic functions or on decision quality constraints rather than simple minimization of estimation variance. Efficient algorithms and 32-bit desktop computers make simulation feasible for routine use. A design solution based on conditional simulation will approach the true optimum only to the degree that the simulations accurately reflect the relevant real-world characteristics. The quality of a simulation depends on a number of factors including the number of conditioning data, the accuracy of the variogram model, and the use of data transformations. The method is illustrated with two examples — one based on the well-known Walker Lake model and the other on an actual case study involving remediation of contaminated soils. For practical purposes, the method appears to be accurate, precise, and robust.