Hydrological model coupling with ANNs

R. G. Kamp (1, 2) , H. H. G. Savenije (1)

1 Section of Water Resources
2 MX. Systems B.V.

Model coupling in general is necessary but complicated. Scientists develop and improve conceptual models to represent physical processes occurring in nature. The next step is to translate these concepts into a mathematical model and finally into a computer model. Problems may appear if the knowledge, encapsulated in a computer model and software program is needed for another purpose. In integrated water management this is often the case when connections between hydrological, hydraulic or ecological models are required. Coupling is difficult for many reasons, related to data formats, compatibility of scales, ability to modify source codes, etc. Hence, there is a need for an efficient and cost effective approach to model-coupling. One solution for model coupling is the use of Artificial Neural Networks (ANNs). The ANN can be used as a fast and effective model simulator which can connect different models. In this paper ANNs are used to couple four different models: a rainfall runoff model, a river channel routing model, an estuarine salt intrusion model, and an ecological model. The coupling as such has proven to be feasible and efficient. However the salt intrusion model appeared difficult to model accurately in an ANN. The ANN has difficulty to represent both short term (tidal) and long term (hydrological) processes.

Domaines: Interfaces continentales, environnement; Océan, Atmosphère; Sciences de la Terre

R. G. Kamp, H. H. G. Savenije. Hydrological model coupling with ANNs. Hydrology and Earth System Sciences Discussions, 2006, 3 (6), pp.3629-3653. ⟨hal-00298800⟩