IWRA World Water Congress 2008 Montpellier France
1. Water availability, use and management
Author(s): A.J. Saul
J.B. Boxall & P. Willems
(1)Professor, Department of Civil and Structural Engineering, University of
Sheffield, Sir Frederick Mappin Building, Mappin Street, Sheffield, S1 3JD. UK.
Email:
n.lan@sheffield.ac.uk
(2)Professor, Department of Civil Engineering - Hydraulics Labor
Keyword(s): Rainfall-runoff, calibration, modelling
Article: PDFAbstractLumped
conceptual rainfall runoff models have been widely used in hydrology for many years. Those models are usually able
to describe most important processes in a catchment through a set of solvable equations. Thus, in many cases they
are preferably over fully physically-based models since they have such advantages: basic physically-based and
simplicity. However, many parameters of those models can not always be directly measured due to the fact that the
conceptual models are lumped on catchment scales. Even though the model structure can be very detailed, the
modelled results are possibly meaningless if the model parameters are poorly specified. The usefulness of the
hydrological model relies on how well the model is calibrated.
Three lumped conceptual rainfall-runoff models are
presented and compared in this paper: NAM (DHI), FEH (UK), and TVM (a simplified model developed by the
author). If the NAM and TVM models are representative for continuous modelling, then the FEH model is event-
based type. These three different models have been tested for the Bradford catchment in UK. For each model,
simulations are conducted on a seasonal basis (summer and winter) with a time step of hourly or quarter hourly for
model calibration. A longer time series of data, 5 years, are used for model validation. The procedure of model
calibration and validation is presented. It is on the basic of combination of automated and manual calibration to make
use of advantages of the automatic calibration technique and the experience of hydrologists. Model validation is
performed together with statistical analysis of the observed extreme values. The results suggest that the semi-
automated calibration method enables the hydrologist to take advantages of the powerful automated method to
obtain good estimates of the parameters which produce simulated results matching observed data well, then refine
the estimates based on their experiences and take into account factors that may not be easy to include in the
automated procedure.
It has also been shown in the study that overall the NAM and TVM models perform well
under given conditions of the study site although they has a tendency to slightly overestimate interflows, while the
FEH model presents a poor result in term of continuous modelling due to model assumptions for baseflow.
However, there is no big discrepancy between the NAM and FEH model when the selected events are
modelled.
It can be concluded that overall the hydrological models represented in the paper give reasonable
results in term of accuracy. The selection of models for particular catchments should be based on data availability,
project objective and model structure.