Long range impact-based forecasts for drought early warning in Australia

Early warning of drought is critical for national water security and farm business outcomes. The Drought Early Warning Service (DEWS) project is developing indicators to measure and forecast the extent and severity of drought impacts in the Australian water and agricultural sectors. Through a multi-disciplinary approach, long-range climate forecasts are combined with agricultural, hydrological and agro-economic models to generate impact-based drought indicators, translating climate data into specific impacts such as crop yields, pasture growth, farm business outcomes and inflows.

In this work, we develop a high-resolution spatial forecasting system to generate forecasts of rainfall, temperature, solar radiation, vapour pressure and evaporation on a 5km grid across Australia, at a daily time step out to 18 months ahead. Forecasts are derived from the Bureau of Meteorology’s ACCESS-S2 climate model. Forecast calibration and downscaling are implemented using Bayesian joint probability modelling and an empirical disaggregation approach, which seamlessly extends forecasts beyond the 7-month range of the climate model.  The ensemble forecasts are targeted at multiple observational datasets to drive a suite of models: APSIM, AussieGrass, FarmPredict and FoGSS.

The outputs of the models become indicators that will be presented as percentiles for a defined historical reference period. The definition of an appropriate reference period is a challenging problem given significant changes in both temperature and rainfall patterns across Australia over the last century. Drought indicators will be published online via the prototype Climate Services for Agriculture (CSA) platform currently under development. We discuss the impacts of the forecasting system and adoption for other use cases such as hydropower, which require long-range forecasts of inflows to assess sustainability.