10.4122/1.1000000624
Falk, Anne Katrine
Anne Katrine
Falk
akf@dhi.dk
Butts, Michael B.
Michael B.
Butts
mib@dhi.dk
Madsen, Henrik
Henrik
Madsen
hem@dhi.dk
Hartnack, Johan
Johan
Hartnack
jnh@dhi.dk
Falk, Anne Katrine
Anne Katrine
Falk
akf@dhi.dk
DATA ASSIMILATION TO IMPROVE FORECAST QUALITY OF RIVER BASIN MODELS
XVI International Conference on Computational Methods in Water Resources
2006
2006
Ideally, real-time flood management decisions must be based on an understanding of
the uncertainties and associated risks. It is therefore central for effective flood
management tools to provide reliable estimates of the forecast uncertainty. Only by
quantifying the inherent uncertainties involved in flood forecasting can effective
real-time flood management and warning be carried out. Forecast uncertainty requires
the estimation of the uncertainties associated with both the hydrological model
inputs (e.g. precipitation observations and forecasts), model structure,
parameterisation and calibration, and methodologies that predict how the
uncertainties from different sources propagate through the hydrological and
hydraulic system.
Within the EU 5th framework project FLOODRELIEF, an ensemble-based approach has been
developed to address the issue of handling and quantifying forecasting and modelling
uncertainties. A general stochastic framework for flood forecast modelling is
presented based on the Ensemble Kalman Filter (Evensen, 1994). The Kalman filter
provides a natural framework for determining how the different sources of
uncertainty propagate through the hydrological and hydraulic models and to reduce
forecast uncertainty via data assimilation of real-time observations. An evaluation
of this framework is presented for several case studies including the US NWS study
catchment, the Blue river basin and the Welland and Glen River Basin in the UK.
Two methods for introducing uncertainties into the model are compared:
1. Stochastic errors are added to the runoff calculated by the catchment model.
Only states in the river channel model are updated
2. Stochastic errors are added to the input to the catchment model (e.g.
precipitation and evaporation). States in both the catchment model and in the river
channel model are updated.
In particular, an investigation of the value of these two approaches for rapidly
responding river basins versus more slowly responding systems is presented. As
expected it is observed that updating in both the catchment model and the river
channel model has a longer lasting effect on the forecast than updating in the river
channel alone. Finally the results of this evaluation highlight the fact that one of
the major outstanding problems in estimating the forecast uncertainty is the
characterisation of the sources of uncertainty.
References:
Evensen, G. (1994), Sequential data assimilation with a nonlinear quasi-geostrophic
model using Monte Carlo methods to forecast error statistics, J. Geophysical
Research, vol. 99, no. C5, pp. 10143-10162.