The Australian Water Resource Assessment Landscape (AWRA-L) model was regionally calibrated at 16 unregulated catchments using two calibration schemes: one biased towards high streamflow and another towards low streamflow. Two parameter sets obtained from the two model calibrations are used as starting points to generate 10,000 parameter sets used for uncertainty analysis (Section 2.6. 1.5).
Both model calibrations result in predictions that perform well across a wide range of streamflow conditions. The high-streamflow calibration predicts reasonable hydrological response variables related to high-flow characteristics (indicated by very narrow interquartile ranges in model bias, and the median bias approaching zero), including P99 (streamflow at the 99th percentile), flood days, annual flow and interquartile range. In contrast, the low-streamflow calibration predicts reasonable hydrological response variables for low-streamflow metrics including P01 (streamflow at the first percentile), low-flow days, low-flow spells, and length of the longest low-flow spell. Neither calibration scheme predicts zero-flow days well, but this is unsurprising since most streams are perennial.
The regional calibration procedure employed in BA is characterised by minimal degradation in prediction performance between calibration catchments and other parts of the modelling domain. The good performance of the model in calibration therefore provides confidence that it will also apply well to each receptor location where there are no streamflow observations.
Product Finalisation date
- 220.127.116.11 Methods
- 18.104.22.168 Review of existing models
- 22.214.171.124 Model development
- 126.96.36.199 Calibration
- 188.8.131.52 Uncertainty
- 184.108.40.206 Prediction
- Currency of scientific results
- Contributors to the Technical Programme
- About this technical product