Theory of model-based geophysical survey and experimental design: Part 2 - Nonlinear problems

In order to design surveys effectively in nonlinear situations, formal design approaches should be developed. Two possible approaches are the Bayesian design and the multimode reduction method. Both are useful in cases where the entropy criterion is simply too computationally expensive to use. The former involves averaging a linearized measure of design quality over the prior model probability distribution, whereas the latter involves a two-step design strategy which is possible only in certain nonlinear problems in which the number of modes can be reduced to one. Among the two approaches, the Bayesian design offers a computationally more tractable alternative.