In December 2007, Kangaroo Island was set ablaze by numerous dry lightning strikes. Our research into the event has been conducted in two parts. The initial phase was a case study investigating the interactions between the local meteorology and observed fire behaviour. Subsequent focus has been on simulating the fires using the coupled atmospheric-fire behaviour model WRF-fire. The findings highlight (1) the importance of including instability and spatial information in fire weather forecasts and (2) the potential for coupled atmospheric-fire models to con- tribute towards current bushfire research and future forecast operations.
The case study
Analysis of the observations and numerical weather prediction models identified three days of unusual fire behaviour. On the first day, a fire located in a local sea breeze front convergence zone produced a spectacular convection column that developed by realisation of potential instability. On the second day, unprecedented fire behaviour was observed in relatively benign conditions due to convective plume entrainment of dry air aloft, enhanced by topographic interactions in very dry, open structured fuels. The third day saw dramatic transition in fire ground conditions due to break down of the nocturnal and marine boundary layers in the late morning by turbulent processes, resulting in sudden changes in fire behaviour.
The circumstances leading to the unusual fire behaviour(s) highlight limitations in the current Australian approach to forecasting fire weather conditions. Point forecasts of meteorological parameters provide an incomplete picture, as they neglect to describe temporal and spatial variations that may have an important influence on fire behaviour. Additionally, forecast ratings are contingent on precise pre- dictions of near surface wind, temperature and humidity, omitting detail on three dimensional atmospheric evolution and, in particular, any interaction of the fire with the atmosphere. These limitations are widely recognised by the Australian fire management community, and in view of recent bushfire events, development of new fire danger rating system has been proposed. Fire behaviour and fire spread models such as WRF-fire have the potential to contribute to developing a rating system underpinned by contemporary scientific evidence.
The WRF-fire simulations
The coupled atmospheric-fire behaviour model WRF-fire has been used to simulate days of the Kangaroo Island fires when unusual fire behaviour occurred. WRF-fire comprises the Weather Research and Forecasting model coupled with a fire behaviour model. The fire model is based on the Rothermel equations for fuel combustion. A level-set method is employed for propagation of the fire front across the model terrain, again following Rothermel's parameterisation for fire spread.
The aims of the current study have been to (1) test the ability for WRF-fire to be run on a real event, using available (spatial and other) Australian data (2) explore the capabilities of WRF-fire, in particular by assessing the models skill in capturing aspects of fire behaviour observed during the case study and (3) provide evidence to contribute towards the discussion on improving fire weather forecasting methods in Australia. These aims have been addressed by (1) a successful (albeit with potential for improved verification) run of the Kangaroo Island bushfires. This success was largely achieved due to the comprehensive and freely available code and documentation for both WRF and WRF-fire, as well as support from developers. (2) Verification shows that WRF-fire captures some, but not all, of the phenomenon observed (limitations include differences in inferred height of convection columns and accuracy of fire spread through terrain). (3) Simulations show that WRF-fire may be run in near-real time and, through available visualisation tools, produce output that may be interpreted by a range of users.
Our study shows that WRF-fire has the capacity to contribute towards the science required to develop better techniques for anticipating fire behaviour in Australia. As development of coupled fire models continues and necessary improvements such as dynamic fuel dryness and refined fire-spread parameterisations are included, model fidelity will continue to improve. In the short term, WRF-fire has demonstrated the ability to provide valuable insights into fire behaviour. In the longer term, it is reasonable to anticipate that output from real-time coupled simulations of actual bushfire events may, given appropriate communication channels, be distributed in real time to enable decisions pertinent to saving lives and property.