Very high resolution model forecasts of fire weather for the January 1994 Fires

Wildfires from 27 December 1993 to 14 January 1994 occurred along most of the southeast coast of Australia. More than 800,000 hectares of land were affected, 205 houses destroyed and three people perished. The most extensive fires were near Sydney and there were many smaller but intense bushfires in suburbs, even close to the city centre.

The fire weather leading up to the worst of the fire days in the Sydney area, namely January7-8, has already been simulated with the mesoscale atmospheric prediction model developed by one of the authors (LML), successfully capturing vital detailed wind speed and direction, relative humidity and temperature patterns both near the surface and above the boundary layer. The simulations also produced skilful predictions of the Forest Fire Danger Index

In this study, use is made of much higher resolution orography and the assimilation of additional observational data which was available at the time but not used by the operational model. The extra data allowed a refinement of the predictions of wind speed and direction, relative humidity and temperature patterns for the period of the fire episode. These improvements are needed to simulate finer mesoscale detail, for example, of the wind fields, over the Sydney basin and surrounding areas. Specifically, to address these improvements further work is being carried out, examining two additional factors: the effects of increased horizontal resolution down to one kilometer or less, which will allow the use of far more detailed orography; and the incorporation of a considerable amount of additional data in the initial fields in the form of a re-analysis process. For example, there is a substantial automatic weather station (AWS) network which is still not included in the present analysis scheme. Work is under way in this area and results from numerical experiments carried out with the improved version of the analysis/forecast system will be presented at the Conference.