WAFARi: A new modelling system for Seasonal Streamflow Forecasting service of the Bureau of Meteorology, Australia

The Bureau of Meteorology (the Bureau) is responsible for compiling and disseminating comprehensive water resource information for Australia through the Commonwealth Water Act 2007. One of its roles is to provide regular forecasts of streamflows and water availability. To perform this role, the Bureau launched a new Seasonal Streamflow Forecasting (SSF) service in December 2010. The service delivers three-month probabilistic forecasts of total streamflow volumes at single locations or total inflows into a water storage, issued monthly. To ensure the timely and reliable delivery of the SSF service, a new modelling system named WAFARi (Water Availability Forecasts of Australian Rivers) was developed. Using the CSIRO's statistical Bayesian Joint Probability (BJP) system as its kernel, WAFARi is equipped with many tools to support the entire workflow for the streamflow forecasts. These tools range from data management into a central database, to web publication through an operational server. WAFARi stores and manages data in self-descriptive files with fully annotated metadata and provides graphical tools that generate highly tailored visual products for the SSF service. Some of the tools were designed to exploit the parallel computing power of the massive Bureau cluster to accelerate the computationally intensive analysis of BJP. All of these tools are available through the powerful scripting environment of Python, which allows users to inspect intermediate model outcomes at any step and flexibly control their behaviours, either in an interactive command shell or with script files. This paper highlights the major features of the system and explains its application to the SSF service.