The Benefits of Ensemble Prediction for Forecasting an Extreme Event: The Queensland Floods

In February 2019, northern Queensland experienced an extreme weather system with cold temperatures, high rainfall, and flooding. The event had serious impacts on agriculture including the mortality of over 620,000 head of cattle and maximum inundation of around 200,000 km2 of land surface.

Researchers at the UK Meteorology Office (Met Office) and the Bureau of Meteorology used a set of several forecasts (ensembles), to investigate the predictability of this event. Ensemble forecasts are used to understand forecast uncertainty and the likelihood of possible events. This makes them valuable for identifying extreme events. The ensembles used in this study are based on those from the Met Office’s numerical weather prediction (NWP) system. There were 25 members in each ensemble, and they were used to investigate the multiweek predictability of the event, in terms of precipitation and temperature. The researchers also investigated forecast improvements gained by increasing resolution, using a newer version with better representation of physics, or using a model where the dynamics of the atmosphere and ocean are connected.

The NWP had reasonable skill in predicting the event, with the ensembles providing a strong indication of an event at the limits of the observed record. Using a model with interactions between the atmosphere and the ocean improved the prediction of the trajectory of the low-pressure system. Thus, atmospheric-only models had greater biases in the distribution of rainfall across northern Queensland. In contrast, increasing the resolution and improving the physics of the atmospheric model did not improve the forecast skill for this event. Further repeated analyses of this technique over many cases with probabilistic evaluation is required before this method can be used for regular forecasting.