Development and evaluation of the AWRA-R model

[摘要] This report describes the development and validation of the river component of the AWRA modelling system, referred as AWRA-R. The model aims at estimating the fluxes and stores associated with the river system in order to populate the water accounts produced by the Bureau of Meteorology.The report details the content of the database that was used to calibrate and validate the model. The data cover 240 catchments in Australia. It contains GIS layers used to configure AWRA-R, including catchment boundaries and node-link network, and time series data, including streamflow, climate variables and diversions.The report focuses on seven components of AWRA-R: rainfall-runoff, routing, floodplain modelling, surface water-groundwater interactions, irrigation demand modelling and on-farm-storage modelling. For each component, it describes the rationale behind the model structure, the validation procedure and provides recommendations for further developments.Rainfall-runoff and routing components were analysed jointly. The work focused on the comparison between model structures and parameterisation strategies of increasing complexity. The comparison concluded that the choice of the rainfall-runoff model has a critical impact on the model performance. Significant performance improvement was obtained with other models than AWRA-L, especially on catchments with a high aridity index P/PE. As a result it was recommended to continue the development of AWRA-L by comparing it with classical rainfall-runoff models. It was also recommended to consider the use of locally calibrated parameters. Regarding the routing schemes, a linear routing model with one calibrated parameter was recommended. When this parameter was fixed to a default value, the model performance decreased but remained acceptable. This finding indicated that the routing model could be applied to reaches where no data are available for calibration.The floodplain modelling work investigated two methods to estimate river and floodplain volumes. One was based on remotely sensed floodplain volume data and the other on river cross section and water levels data. The methods were compared on six reaches located in the Barwon, Darling and Murumbidgee basins. This analysis recommended a substantial improvement of the remotely sensed data before they can be incorporated in AWRA-R. The use of these data generated large unaccounted volumes, which could not be explained by floodplain losses to groundwater. In addition, large volumes were found during droughts and small volume during floods, which raised some doubts about the accuracy of the remotely sensed data. It was recommended to improve several elements in this dataset including the removal of the noise and the explanation of the large volumes generated outside of flood events. The alternative method based on cross section data at both ends of the reach provided a more promising approach to perform the repartition of volume. It is suggested to improve this method by including an estimate of the groundwater losses from floodplain and refine the topographical description of the floodplains.Regarding the surface water-groundwater interactions, it was found that fitting a reasonable curve to an existing loss function solved some of the problem of unattributed losses to individual daily flow records, usually at the low flow end. It was recommended to fit the groundwater loss parameters at the same time as the rainfall-runoff and river routing fitting is performed, to minimise the annual ungauged loss over the spectrum of observed flows. A powerful functional form was proposed for the loss function based on the Monod hyperbolic function. Finally, rules were proposed to determine when the fitting of the loss function is required. For example, if total losses are very small, it may be reasonable to not include an extra loss term at all.Irrigation demand modelling investigated the relationship between demand of water used in irrigation...

[发布日期] 2012-05-31 [发布机构] CSIRO

[学科分类] 地球科学(综合)