The introduction of grid-based forecasting a decade ago revolutionized forecast production in the Australian Bureau of Meteorology.  The role of the forecaster in production shifted from author of text to curator of a gridded database using the Graphical Forecast Editor (GFE) software.  Graphical and text products could then be generated for any location in Australia, out to 7 days, using largely automated processes.  The efficient capture of forecast decisions in the gridded database, together with automated product generation, enabled a huge expansion in the forecast service which the Bureau provided to the Australian community.

Having a gridded forecast database also facilitates objective forecast verification.  We have developed tools to perform such verification and have applied them to forecasts of precipitation, temperature, wind and fire danger.  This has shown where the forecaster adds most value in their curation of the gridded database and has provided a strong case for more forecaster reliance on outputs from automated forecast systems, particularly in the outer days of the forecast period.  Our tools have allowed us to perform hindcast experiments and evaluation of the results has supported changes to forecaster tools in GFE and to the automated forecast systems.

Verification has highlighted several 'service tensions' where a single-valued parameter in the database is used to underpin different services with competing requirements.  To date, the forecaster has played a role in balancing these requirements, for instance to optimize warning areas during significant weather events.   To allow greater reliance on automated forecasts we need to resolve such tensions more comprehensively, by redefinition of services to use the forecast parameters consistently, or by introduction of new forecast parameters to better meet different service requirements.

In this presentation, we will review learnings from four years of work in this area.