A comparison of two seasonal rainfall forecasting systems for Australia

Abstract

Operational El Nino-Southern Oscillation (ENSO)-based statistical forecasting of seasonal rainfall has been undertaken in Australia now for the best part of two decades. This article compares the performance in recent years (1997 to 2009) of the two major governmental programs currently forecasting seasonal rainfall across Australia, through the verification of independent forecasts. These programs are run by the Australian Bureau of Meteorology and the Queensland Government's Department of Environment and Resource Management (and its predecessors). In this study, the Queensland Government's Southern Oscillation Index (SOI) phase-based forecasts are not verified directly. Instead, the SOI-phase system methodology is reconstructed using the Bureau's monthly rainfall analyses and those reconstructed forecasts verified. Verification techniques employed include linear error in probability space (LEPS2) skill scores, per cent consistent rates, and reliability statistics. Over recent years, the Bureau forecasts and the SOI-phase forecasts have performed comparably when measured by the per cent consistent statistic, although with successes and failures in different parts of the country. Both sets of forecasts show widespread forecast skill in excess of climatology, the Bureau forecasts performing better in the west of the country and the SOI-phase forecasts performing better in the centre and east. The generally more emphatic SOI-phase forecasts translated into higher LEPS2 skill scores, but the Bureau forecasts showed greater reliability.