Do initial parameter values matter?
If you are doing traditional parameter estimation where you formulate a well-posed inverse problem, and if your model is not too nonlinear then initial parameter values may not matter too much. In this case there will only be one objective function minimum, this corresponding to a unique set of parameter values. PEST will probably find this minimum irrespective of initial parameter values. However if your model is very nonlinear, and there are multiple objective function minima, then your initial parameter values should be chosen more carefully. Chose parameter values that you consider to be realistic on the basis of your expert judgement. If doing highly parameterized inversion the same rule applies. Choose initial parameter values that are your preferred parameter values on the basis of expert knowledge or site characterization information. Then let the parameter estimation process introduce perturbations to this preferred parameter set, with Tikhonov regularization ensuring that the
