KNAW

Research

Integration of generic automated Calibration Tools in FEWS

Pagina-navigatie:


Update Research data


Title Integration of generic automated Calibration Tools in FEWS
Period 01 / 2004 - unknown
Status Completed
Research number OND1307341

Abstract

In practical case studies, aggregated and/or integrated numerical models are commonly used for the simulation and prediction of hydraulic, hydrological, ecological, etc. processes. These models usually contain one or more application-dependent parameters or coefficients, even in cases where these models are based on sound physical principles. Such uncertain model parameters often represent non-measurable quantities and they must be estimated using calibration, i.e. they are adjusted until the model's response agrees as closely and consistently as possible with the corresponding observations. Examples of efficient automated and generic calibration techniques which have been developed and applied at Delft Hydraulics in the past include the DUD-method (an efficient derivative-free, though local minimisation method, based on least square criteria) and SCE (Shuffled Complex Evolution; a global minimisation technique for more arbitrary calibration criteria, but at the cost of much more computation time).
In the present project, WL / Delft Hydraulics' Flood Early Warning environment (FEWS) was expanded with the addition of generic and automated calibration tools. An important issue involved in this development is that these tools must be applicable for virtually any numerical simulation model that is (or will be) linked to the FEWS environment, such as, for example, hydraulic models (SOBEK-Rural, SOBEK-River, Delft3D, ISIS, etc.) and rainfall-runoff models (e.g. HBV-96 and MCMR). Because of its proven skill in the past, the DUD-method was the first calibration method that was made operational within the FEWS-environment. Using dialogue boxes, the user can identify the specific model application and select the model parameters for calibration. In the same flexible way, the calibration run options are set and results are presented. The implementation and performance were successfully tested by means of a calibration of runoff parameters in an MCMR model of the Severn catchment (Midlands, UK), dealing with the prediction of the river flow at Rhos-Y-Pentref. In a second application DUD was applied for the estimation of bed roughness parameters in a hydraulic SOBEK-Rural model included in a model suite for the Tanshui basin in Taiwan.

Related organisations

Related people

Project leader Dr. H.F.P. van den Boogaard
Project leader Dr. A.H. Weerts

Related research (upper level)

Classification

A12000 Surfacewater and groundwater
D11000 Mathematics
D15600 Hydrospheric sciences
D16200 Software, algorithms, control systems
D22400 Ecology

Go to page top
Go back to contents
Go back to site navigation