Mathematics of Information Technology and Complex Systems


Project Highlights


Team Members

Partner Organizations

Students & PostDocs







MITACS/GEOIDE Conference on

Forest Fire Modelling

June 22-23, 2009
Hinton Training Centre, Alberta




The modelling of forest fire spread involves various physical scales, from the microscopic scales associated with combustion and heat conduction, to macroscopic scales associated with landscape and weather. Various mathematical and statistical methods are needed to understand this multi-level process. In this workshop, we focus on forest fire growth modelling and management. The workshop follows the MITACS/GEOIDE Summer school on Mathematical and Statistical Descriptions of Forest Fire Spread, June 15-19, 2009 at the same venue. It is expected that a number of senior students from this school will stay on site to attend the workshop.

Recent progress under the PROMETHEUS/MITACS project `Forest fire spread in heterogeneous landscape' presents a unique opportunity to review recent results and to identify promising directions for further research. Both mathematical and statistical approaches will be represented as well as fire behaviour prediction (FBP) system issues and other aspects of fire management. Questions around which the workshop will be focussed:

1. How should randomness be incorporated into spread models, and what type of data is or should be available to calibrate such models?

2. How can processes such as spotting and suppression be best incorporated into current models?

We have purposely designed the meeting to be informal. Each day will begin with a plenary talk. Featured speakers are Jim Gould (CSIRO, Australia) and John Dold (U of Manchester, UK). These talks will be followed by a mix of shorter contributed talks by fire scientists and numerous roundtable sessions to encourage interaction amongst the diverse group of scientists attending.

In addition to the MITACS and GEOIDE scientists at the meeting, students and postdoctoral fellows we expect personnel from Alberta Government will lend a substantial presence at the workshop.


The workshop was confirmed as a great success by all attending. Twenty-two participants were present during the two-day program. This number included four Government Scientists from both the Alberta SRD and the CFS and at least 7 students and postdocs interested in fire research. Some of these students had stayed on in Hinton after the previous week's Summer School to attend the workshop.

Jim Gould (CFS-CSIRO Australia) began the meeting with a provocative talk on the challenges presented by the many forms of uncertainty present in today's fire models. Jim's vast experience in both the Canadian and Australian fire landscapes, his broad and scholarly approach to the subject and his ability to 'see both the forest AND the trees' puts him in a unique position to comment on these fundamental and sometimes unsettling challenges to the fire modelling community. His 75 minute talk was followed by a lively question and answer period.

John Dold (Manchester) introduced us to the concept of the 'Byram number' and its use in semi-physical models of unsteady bushfire. An application of Byram's number to the poorly understood and dangerous phenomenon of `eruptive fire' gave us an excellent example of an effective mathematical modelling approach, and its power to steer fire researchers to ask new and critical questions aimed at getting to the root of such complex behavior. John's one hour talk was followed by a lively question and answer period.

Five contributed talks on other aspects of randomness and uncertainty in fire modelling were given during the workshop (see the conference program for details). Six one-hour roundtables were held on the topics of Forest Fire Risk, Stochasticity in Fire Models, Acceleration in Spread Rates, Incorporating Wind in Empirical Spread Models, Forest Carbon Mapping and Distributions and The Fire Spotting Process. It was generally agreed that the format of short talks, interspersed with lively roundtable discussions was highly effective in promoting open exchange of scientific information amongst participants.

Participants and organizers were extremely satisfied with the venue (The Hinton Training Centre) and we hope to be able to use these facilities again for future workshops. We would like to thank the sponsors and staff of the Training Centre for their service and support.

An newspaper article about the Summer School was published in the local newspaper, the Hinton Parklander, much to the delight of both organizers and participants.

The next event planned on this topic is a BIRS workshop awarded for November, 2010. Organizers are Bose/Bourlioux//Braun.


Chris Bose (Victoria), Jim Gould (CFS-CSIRO), Cordy Tymstra (Alberta SRD), Jonathan Lee (Western) Michael Francis (Victoria), Thomas Hillen (Alberta), Anne Bourlioux (Montreal), Reg Kulperger (Western), Hao Yu (Western), Martin Alexander (CFS), Jennifer Beverly (CFS), John Braun (Western), Fangliang He (Alberta), Yanping He (Victoria), Jin Song Chen (Alberta), Yanping Wang (Alberta), Doug Woolford (Simon Fraser), Ed Johnson (Calgary), Kiyoko Miyanishi (Calgary), Mary Ann Jenkins (York), Sean Michaletz (Calgary), John Dold (Manchester UK), Jeff Picka (New Brunswick).


1. US forest fire researchers are moving strongly in the direction of fully physical models of combustion coupled with elaborate numerical fluid dynamic atmosphere models. None of these programs has produced an operational software package for field use to date; they are mostly research-only tools for fire scientists. While improved computing power will undoubtedly bring these complex simulators closer to real time computations, the timeline is not clear. Therefore, for the time being, continued development of the existing empirical spread models is a sensible approach.

2. The current best approach for a more encompassing simulation seems to be a hybrid model, coupling fluid dynamical descriptions of the atmosphere to empirical spread of fire on the ground. There is a whole range of detail that can be considered for the fluid coupling and this is a high priority research direction for our project scientists. This situation also gives another reason to continue with development of empirical models such as PROMETHEUS.

3. Specific modelling challenges included in Item 2 above include: (a) incorporating detailed wind flow into the spread model, and putting in some simple coupling of the local wind with the effects of the convective column. (b) modelling the complex process of mid to long range spotting (200m to 2km). Spotting is a sub-dominant fire spread mode in the boreal forest, but may become more significant in our changing forests, particularly with fire in beetle killed stands. (c) modelling of the unsteady (acceleration) phase of fire growth. A poll taken at the workshop indicated that this last item (acceleration) should be the lowest priority of the three, but as we heard later from John Dold, acceleration in the form of 'eruptive fire' is of high priority for our European counterparts, so perhaps we should reconsider. (d) There was a call, particularly from the experienced operational fire participants at the workshop, to investigate the modelling of suppression efforts on active fire fronts. This last item seems to be a completely unexplored area at the present time.

4. There is great opportunity and much interest to link the mathematical and statistical aspects of this problem. Up until now the former has been more of concern to the MITACS scientists and the latter to the GEOIDE group, but there is a real chance for cross-fertilization between the groups. For example, at the workshop, many of the roundtable discussions touched on the essential component of uncertainty in the (nominally deterministic) details of spread models. Also, as more sophisticated empirical models are developed, they will need finely structured and many-dimensional data on the physical parameters at play, possibly through enhanced remote sensing activity on fire sites. Finally, on the issue of remote sensing and real-time data streams for active fires, there is clearly potential for a big impact IF the data can be produced and analyzed in REAL real-time. There remains a fair degree of scepticism amongst fire operations managers and scientists on this issue. As one notably observed: `three hours is not real-time in the bushfire business'. This area presents a significant challenge, therefore, but one with potentially high value.