Model-Oriented Data Analysis with Applications to Lake and Soil Water Simulation

摘要

The general aim of all water resources modelling is to cover the broadestpossible range of empirical facts with the smallest possible set of hypotheses. The study, which summarizes eight earlier publications, undertook to investigate practical techniques to attain the goal. The aims of the study were to develop methods for structural characterization, optimal observational design and for choice of the best model structure from among rival candidates. The problems were analyzed using the theory of optimal experimental design. A method to detect and identify non-linear dependency among model parameters was suggested. The method is based on the logarithmic transformation of parameter space and the principal component analysis of the Fisher information matrix. The technique can also be used to approximate the admissible level of the observational error. The non-linear D-optimal design of observations was studied.