Predicting fire thunderstorms
Global monitoring of the atmosphere began in 1978. For most of this era, however, bushfires were not on the agenda. Volcanoes were an initial focus from 1982, after a British Airways 747 flew into a volcanic ash cloud, stalling all four engines. Before this, atmospheric scientists had been intensively watching for the effects of nuclear weapons tests, as these show up clearly in the clean stratosphere. And yet, cases of stratospheric aerosol injections were observed that could not be explained.
By running weather models backwards, researchers deduced that large bushfires in Canada were the source of these aerosol injections. Now the search was on for a fire event to confirm this. In 2001, the Chisholm fire in Alberta, Canada, formed a fire thunderstorm, also known as pyrocumulonimbus, or ‘pyroCb’ for short. Several papers written about this fire raised key questions, and another detailed case study was sought.
On 8 January 2003, a dry lightning storm lit scores of fires across the Australian Alps. The only previous consideration of a fire thunderstorm was the Berringa fire in Victoria in 1995, which addressed the threat to firefighters if the smoke plume collapsed. The Berringa fire made firefighters and fire managers think about what was going on over our heads – that we had to monitor the fire plumes, not just the flames.
When the January 2003 alpine fires formed violent pyro-convection, nobody was prepared. We had no predictive capability. These events are still the subject of scientific studies – I will go as far as saying that they are the most scientifically important bushfires ever.
From 1978 to 2001, Australia recorded two minor pyroCbs. Since then, we have 56 on record to November 2016, including some of the most intense events globally. PyroCbs have long been a problem in the forests of the United States, Canada, eastern Russia and Mongolia. They became a problem in Australia in 2001, in western Russia in 2010 and in Europe in 2017. The Black Saturday fires in 2009 had the record for the most intense pyro-convection until August 2017, when the Chezacut fire erupted in British Columbia, Canada.
How pyroCbs develop
Australia has entered an ‘era of violent pyro-convection’. Bushfires are modelled and predicted based on the assumption of steady-state spread. This includes weather, terrain and vegetation inputs to predict a fire’s behaviour. For any inputs, the model gives a unique prediction of what the fire will be doing. This is the basis for all fire service preparedness, fuel management and community protection.
PyroCbs are now known to occur when a fire forms ‘deep flaming’ under an unstable atmosphere. Deep flaming is the depth of the active fire, as in how far back from the fire front strong heat is released. This can be some kilometres, and is not to be confused with the fire front.
A Canberra-based research group (comprised of myself, Associate Professor Jason Sharples of the University of NSW, Mike Fromm from the US Naval Research Laboratory and Rene Servranckx from Environment Canada) is developing a list of conditions under which deep flaming can occur. Two of these occur during steady-state fire spread—high rate of spread and a wind change. Other conditions are now grouped under dynamic fire spread, and include the following.
- Dense spotting creates deep flaming as the spotfires merge.
- Vorticity-driven lateral spread (VLS, or fire channelling) is the most effective source of deep flaming known.
- Eruptive growth is a concept that emerged from a fire in London’s Kings Cross Underground railway station.
- Professor Domingos Viegas from the University of Coimbra in Portugal has shown how the flame attachment involved can very easily lead to fire crew burn-overs.
- Sebastien Lahaye at the University of NSW in Canberra has recently extended our knowledge of dynamic burn-over causes.
- Interior ignition is a new concept resulting from staggered flammability of different fuels.
- A final cause of deep flaming is inappropriate use of a drip torch on a bad fire day.
The unstable atmosphere is a difficult concept for fire agencies to handle. The Haines Index and its continuous variants are the main tools that have been used for decades, but they do not pick up the key elements needed. A better tool must be developed or found. Researchers in the Bureau of Meteorology, through the Bushfire and Natural Hazards CRC – including Dr Mika Peace, Dr Jeff Kepert and their colleagues – are looking closely at the instability above fires.
An extreme bushfire is defined as a fire that, on one or more occasions over its duration, will form deep flaming in an atmospheric environment conducive to the fire coupling with the atmosphere. The plume then punches through the cloud base (termed a ‘blow-up’). Our group developed a process model called the Blow-Up Fire Outlook (BUFO) which takes fire behaviour analysts through a series of questions that mostly seek to anticipate deep flaming, with raised fire danger a prerequisite condition. The answers determine which question is next, or whether the analyst loops back to the beginning to wait for conditions to change before starting the questions again.
Note that vegetation, which forms fuel for the fire, does not currently have a role in the BUFO model. It is therefore an open research topic. Similar coupled fire atmosphere events occur over the vast range of fuel types found in alpine ash, Siberian steppe, Albertan boreal forests, or the Great Victoria Desert.
The BUFO model has been formally tested in NSW and the ACT. I conducted the trial with formal oversight from Associate Professor Jason Sharples and Laurence McCoy (NSW Rural Fire Service). Over three fire seasons we obtained enough data to confirm the model, with several blow-up events predicted and incident management teams alerted. No pyroCbs occurred in the trial domain, but we did informally anticipate some pyroCbs elsewhere in Australia. The formal statistical results were sufficient for the model to be declared successful, and it is now operational.
Our group is now seeking to expand its implementation into jurisdictions beyond the trial area. The model is most useful south of the Tropic of Capricorn.
The Sir Ivan fire
A key part of the BUFO is its ability to distinguish the one or two fires that have the potential to develop a pyroCb on a day of widespread raised fire danger with many fires burning. On 11 February 2017, a bad fire day was predicted in north-east NSW, with even worse conditions forecast on the following day. Although many fires were burning, the BUFO model successfully predicted that most would not blow-up to a pyroCB, and that only the Sir Ivan fire had the potential to do so, through VLS. An alert was issued for this fire late on 11 February, and by mid-afternoon the next day it had formed a pyroCb just as a trough-line passed, producing peak instability. Other fires nearby did not blow-up.
The Sir Ivan fire is the only fire, globally, for which a formal operational forecast of a blow-up and subsequent pyroCb has occurred. While not wishing for more pyroCbs, our goal is to anticipate their formation. We now require discussions with fire services to see how the model might be implemented elsewhere.
It has long been thought that with temperatures on the rise, the impact of climate change on bushfires would involve turning up the dial. Now it is clear that there is a big switch as well—and that in Australia, this switch was flicked to ‘on’ in 2001. Blow-up pyroCb fires are the cause of much of bushfire impact on the Australian community, and they are poorly handled by the primary fire prediction tools in use.
For more information, go to www.bnhcrc.com.au
- PyroCbs last for about two to three hours.
- As the day warms up, a slight inversion at the cloudbase is eventually breached and free convection can then reach the top of the troposphere, or even into the lower stratosphere.
- PyroCbs often exceed 12 km in height, and their cloudtops can be glaciated—a fire can form ice.
- When a fire blow-up occurs, the best approach to protecting the community is to evacuate everyone in a footprint ahead of it. Doing this ahead of the event on the basis of forecasts should be the emergency management sector’s goal.