dc.description.abstract | Computer modelling offers a sound scientific framework for well-structured analysis and
management of urban drainage systems and flooding. Computer models are tools that are expected
to simulate the behaviour of the modelled real system with a reasonable level of accuracy.
Assurance of accurate representation of reality by a model is obtained through the model
calibration. Model calibration is an essential step in modelling. This report present concepts and
procedures for calibration and verification of urban flood models. The various stages in the
calibration process are presented sequentially. For each stage, a discussion of general concepts is
followed by descriptions of process elements. Finally, examples and experiences regarding
application of the procedures in the CORFU Barcelona Case Study are presented.
Calibration involves not only the adjustment of model parameters but also other activities such as
model structural and functional validation, data checking and preparation, sensitivity analysis and
model verification, that support and fortify the calibration process as a whole. The objective in
calibration is the minimization of differences between model simulated results and observed
measurements. This is normally achieved through a manual iterative parameter adjustment process
but automatic calibration routines are also available, and combination parameter adjustment
methods also exist. The focus of a model calibration exercise is not the same for all types of models.
But regardless of the model type, good modelling practice should involve thorough model
verification before application.
A well-calibrated model can give the assurance that, at least for a range of tested conditions, the
model behaves like the real system, and that the model is an accurate and reliable tool that may be
used for further analysis. However, calibration could also reveal that the model cannot be calibrated
and that the correctness of the model and its suitability as a tool for analysis and management of
real-world systems could not be proven.
The conceptualisation and simplification of real-world systems and associated processes in
modelling inevitably lead to errors and uncertainty. Various modelling components introduce errors
such as the input parameters, the model concept, scheme and corresponding model output, and the
observed response measurements. Ultimately, the quality of the model as quantified by how much
it deviates from reality is an aggregate of the errors that have been brought into it during the
modelling process. Thus, it is important to identify the different error sources in a model and also
account for and quantify them as part of the modelling. | en_GB |