The Bureau of Meteorology (BoM) provides climate outlooks of precipitation and temperature for the Australian continent several months in advance. A growing need of similar forecasts for hydrological variables has been expressed by Bureau stakeholders from various sectors, including water resources management, food production and flood and bushfire risk assessments. Here, we present the development and testing of a seasonal forecasting system for soil moisture, evapotranspiration and runoff for Australia using the AWRA landscape water balance model (AWRA-L), and the Australian Community Climate and Earth-System Simulator – Seasonal (ACCESS-S).

AWRA-L (Frost, Ramchurn, and Smith 2016; Viney et al. 2015) was developed by CSIRO and BoM and underpins the Australian Landscape Water Balance service (www.bom.gov.au/water/landscape). The model simulates hydrological fluxes and stores, including runoff, evapotranspiration and soil moisture for three soil layers (0-0.1m, 0.1-1m, 1-6m), on a 5 km grid. The hydrological forecasts were generated by forcing AWRA-L with the interpolated and bias-corrected ACCESS-S climate forecasts (Hudson et al., 2013, Griffiths et al., 2017) for the period 1990-2012. The daily output was aggregated to the monthly scale. Here, we present the results of the assessment of the forecast performance against a historical AWRA-L reference run forced with AWAP data. We applied verification metrics for deterministic and ensemble forecasts that capture the accuracy and reliability of the forecasts (including mean bias, anomaly correlation, CRPS) for mean conditions and percentile-based thresholds. Subsequently, we discuss sources of skill by presenting comparisons with a climatology ensemble forecast.