Recent improvements to the Australian Water Resources Assessment Landscape Model (version 7)

The Australian Water Resources Assessment Landscape (AWRA-L) model is a continental gridded, daily time-step, water balance model, developed over the last decade by CSIRO and the Australian Bureau of Meteorology for a range of hydrological applications. The model outputs (including soil moisture, evapotranspiration, runoff and deep drainage; available through www.bom.gov.au/water/landscape) have found wide application for monitoring purposes (e.g. for flood and fire risk, drought monitoring), water reporting (eg. National Water Accounts), and in analysing trends in water balance outputs including streamflow. In addition to these historical/monitoring applications, AWRA-L is being further used for production of 10-day forecasts, seasonal forecasts, and long-term projections of hydrological outputs out to the end of the century.

 

This study details recent development of AWRA-L for improved performance across the water balance for use in monitoring through to long term projections. Changes are implemented across three broad areas: improved static and dynamic inputs, altered conceptual structure (additional urban component and baseflow ephemerality), and altered calibration approach. In particular, a new spatial calibration approach is applied across the nation using over 300 catchments. To do so model pixel output values are compared against spatially distributed satellite data for soil moisture, evapotranspiration (ET), and two new components including fraction of vegetation (F_veg) and terrestrial water storage (TWS). In the previous versions of the model lumped catchment average values of evapotranspiration and soil moisture were used. In addition to comparing to a wide range of national datasets (streamflow observations, flux tower observations, soil moisture network observations, recharge observations), the model performance was compared for drought analysis (reproducing 2-state rainfall-runoff behaviour observed in parts of Australia) and flood analysis (correlating with operationally used flood forecasting parameters). Overall, the modified AWRA-L outperformed previous versions in terms of water balance estimation according to a wide range of validation data. The successful application of the spatial calibration method can potentially pave the path for more frequent application of complex calibration methods for large scale simulations. Furthermore, consideration of a terrestrial water storage component in the objective function highlights the importance of this factor in capturing more accurate simulation of other water balance components, particularly streamflow. The improved streamflow performance demonstrates the enhanced functionality of the model in capturing intermittency and streamflow shifts in seasonally dry and groundwater dependent catchments, further demonstrated in the drought analysis. Finally, the flood study demonstrates the application and value of the model for real time flood-monitoring and forecasting purposes. This study shows the potential of AWRA-L model and associated spatial calibration approach for accurate simulation of water balance variables for use in continental-scale studies.