Let the data build the model: optimising the structure of hydrochemical models using advanced methods of non-linear data analysis

摘要

In spite of being widely applied to understand and to predict freshwater quality better, hydrochemical models suffer from severe limitations. On the one hand, it is assumed that the predominant hydrological and hydrochemical processes and the corresponding appropriate model structures are well known, and the work remaining is to link different processes in numerical models and to calibrate the model parameters. On the other hand, there is increasing doubt that the bottom-up way of model development,starting with, for example, the Darcy equation for water flux, should be the most efficient way to model catchment scale biogeochemical and hydrochemical processes.