author = {M.N.A Mooij, M. Verlaan},

title = {Automatic Calibration of Currents in Shallow Water Flow Models},

organization = {Wessex Insitiute of Technology},

year = {1996},

note = {abstract submited},

annote = {Shallow water flow models are an important tool for forecas-

ting storm surges and predicting the effects of human inter-

ventions in estuaries. Recently these models are being used

increasingly to provide the input for pollutant and sediment

transport models. This change in use has resulted in a demand

for more accurate current predictions.

The accuracy of shallow water flow models is mainly determined

by the accuracy of its inputs, such as depth, friction coeffi-

cient and the tide at the so called open boundary. In order to

obtain accurate model results the model has to calibrated,

which means that the model input is adapted so that the model

results fit the available measurements more closely.

For this purpose an automatic calibration method based on the

adjoint method has been developed to estimate these uncertain

parameters in an efficient way. The parameters are adapted in

a way to increase the similarity between the measured waterle-

vels and the computed waterlevels. Recently, transport predic-

tions have become more important and therefore two different

approaches to improve the model velocities

are proposed.

First of all, the existing method was extended to calibrate,

in a similar way as was used for waterlevel measurements, the

current velocity measurements. Although similar in implementa-

tion, the calibration of waterlevels and currents show diffe-

rent characteristics. For instance the calibration of currents

has a much more local nature.

The second approach adds a term to criterion that penalises

high vorticity near the open boundary. Because the model

boundary is often chosen in a region without much natural

vorticity (no gullies), the vorticity in the model should be

small also. The removal of unrealistically high model vortici-

ty near the boundary will lead to a better performance of

transport models that are based on the currents generated by

the model.

Both methods are illustrated using a model of the Ems-Dollard

estuary in the Netherlands.}

}