AGU_F05

1
Impacts of Uncertain Flow Data on Rainfall-Runoff
Model Calibration and Discharge Predictions in a
Mobile-Bed River
H43D-1030
Hilary McMillan1, Jim Freer2, Florian
Pappenberger3, Tobias Krueger4 and Martyn Clark1
Contact h.mcmillan_at_niwa.co.nz
1 National Institute of Water and Atmospheric
Research Ltd. (NIWA), New Zealand.
2 School of Geographical Sciences, University of
Bristol, UK
3 European Centre for Medium-Range Weather
Forecasts, Reading, UK
4 University of East Anglia, UK
Why Uncertainty in Flow Data is Important
Case Study Results ...
Rainfall and Flow series are needed to calibrate
hydrological models
Impacts on Model Calibration and Discharge
Predictions
Before
After
A Solution?

• TopNet Model
• Water balance model of sub-basins
• kinematic network routing model
• 7 parameters per sub-basin
• Soil and vegetation parameters from catchment
  maps
• Other parameters set at default constant value

Catchment processes
Incorrect flow data
Model structure/parameterisations are forced to
compensate for poor data
Incorrect model with weaker predictive power
Network routing
What Causes the Uncertainty in Flow Data?
Our data is usually stage data transformed to
flow via a rating curve
Significant Errors can occur
Hydrological Model
Discharge (m3/s)
Discharge (m3/s)
Case Study Wairau River, New Zealand
Stage (m)
Stage (m)

• Discharge information is required to calibrate
  hydrological models for flood warning and water
  resource applications
• All three sources of rating curve error (above)
  are present

Previous rating curves show spread but are not a
surrogate for uncertainty