Member's Area - Login/Register

Intercomparison of forecasting methods for flood warning in the river Cam catchment

Paper Topic: 
General
 
Volume: 
 
Issue: 
 

Pages :
89 - 97

Authors: 
Lekkas D.F., Maxey R. T. and Wheater H.S.
Paper ID: 
289
Paper Status: 
Published
Abstract: 

Currently flood warning in the catchment of the River Cam in Cambridgeshire relies on the issuing of
alerts when the river level at the monitoring station at Byron's Pool, just upstream of Cambridge, reaches
certain pre-determined levels. Warnings are shown to be fairly accurate, but there is very little lead
time between the trigger being exceeded and the commencement of flooding. At present there is no
method used that can forecast in advance when the trigger is likely to be reached.
Three conceptually different methods of forecasting if and when the trigger at Byron's Pool will be
exceeded are presented. The first of these is a simple additive model, in which flows from the three tributaries
that are gauged are summed to give a combined flow. The second method involves the derivation
and application of two transfer function models capable of transforming river levels on the
upstream tributaries to a level at the trigger site. These models are applied both with and without realtime
updating techniques. The third method involves the calibration and application of a lumped rainfall-
runoff model of the whole catchment to Byron's Pool. Two different calibration periods are used,
and the results compared.
The results indicate that the simple additive model, while being better than no model at all, is very inaccurate,
and fails to replicate the hydrograph shape and timing, most likely because of the influence of
an ungauged tributary. The transfer function models perform well, especially when real-time updating
is used. The rainfall-runoff model performs less well, struggling to reproduce the hydrograph shape.
Ôhe main conclusions are that for this site a hierarchy of models may be appropriate, with rainfallrunoff
models providing an early indication of flooding, and transfer function routing models with
updating providing a more accurate forecast, with the additive model as a back up. The importance of
obtaining more data, including validation of ratings, and the future gauging of the ungauged tributary,
is noted throughout this investigation.

Keywords: 
Flood warning, forecasting, Additive model, Transfer Functions, NAM model