Why may the dynamical and statistical model outlooks be different?

When comparing outlooks from the new dynamical model and the previous statistical model, it is expected that there will be differences, just as there would be between any two models which generate outputs using different techniques.

The dynamical climate model uses the physics of the oceans, atmosphere, land and ice and the multiple complex interactions between them to estimate the most likely average climate state for several months ahead. As it uses physics, a dynamical model is suitable for use in areas experiencing long-term trends.

In contrast, the statistical model compares past sea surface temperatures in the Pacific and Indian oceans to corresponding Australian rainfall and temperatures, and then uses these historical relationships and current observations to make a three-month rainfall or temperature outlook. In other words, the statistical model assumes that the past represents the future.

Apart from the differences in the models themselves, other factors may also create some difference in the outlooks. For instance, the dynamical model creates outlooks of above or below average relative to the period from 1981 to 2010, whereas the statistical model generates outlooks relative to the period 1950 to 1999. The 1981 to 2010 period has been recognised by many agencies around the world as being representative of recent climate, and hence may better align with the recent climate experienced by most users. Rainfall and temperature maps averaged over the 1981 to 2010 period are provided to assist with interpretation of the outlooks.

Averaged over all seasons, the accuracy of the dynamical model is better than the statistical model. It is recommended that users refer to the skill for the specific time of year when using the outlooks. Dynamical models will continue to evolve and improve over time as developments continue with the science, the observations, the models and in computing power.