Rickard Hansen 2008:3, pp. 79. TEK/avd. för matematik och naturvetenskap, 2008.
There is presently no wildfire model developed for Swedish conditions, only a fire danger rating system (FWI) has been developed for Swedish conditions.
The demand for a wildfire model has not been great in the past in Sweden but the climate changes now taking place increases the risk of large and intensive wildfires in Sweden. The need for additional and better tools for sizing-up wildfires will be in great demand in the future.
This pre-study is aimed at:
- Presenting what has been done in the wildfire modeling field during the years and mainly the last twenty years.
- Giving recommendations on the continued work with developing a Swedish wildfire model.
The method that was used was literature and article survey.
The study also looks into the required input data for a wildfire model and the input data available at the moment. This issue is highly crucial as the quality of the output of a wildfire model is depending upon the quality of the input data.
During the study, a primitive wildfire model was constructed and refined in order to get an insight in the complexities and problems with developing an operational model.
The following characterization of wildfire models was used during the study:
- Statistical models: based primarily on statistics from earlier or experimental fires. They do not explicitly consider the controlling physical processes.
- Semi-empirical models: based on physical laws, but enhanced with some empirical factors, often by lumping all physical mechanisms for heat transfer together.
- Physical models: based on physical principles and distinguishing between physical mechanisms for heat transfer.
The statistical models make no attempt to involve physical processes, as they are merely a statistical description of test fires. Thus the lack of a physical basis means that statistical models must be used carefully outside the test conditions.
Semi-empirical models are often based on conservation of energy principles but do not make any difference between conduction, convection and radiation heat transfer.
The semi-empirical model has low computational requirements and includes variables that are generally easy to measure in the field. So despite the issue with limited accuracy, the speed and simplicity of these models make them useful for operational use.
Physical models have the advantage that they are based on known relationships and thus facilitating their scaling. Thus we can expect that physical models would provide the most accurate predictions and have the widest applicability.
But the work on physical models is suffering of for example the lack of understanding of several processes, such as the characterization of the chemical processes taking place during combustion, the resulting flame characteristics and the isolation and quantification of physical processes governing heat transfer.
The input data available today are generally not detailed enough for physical models. As a result, a very detailed physical model will still only give imprecise predictions. As better and more detailed input will be available, the use of physical models will be more justified.
A semi-empirical model is recommended being developed in Sweden. This conclusion is based upon the following factors:
- The accuracy of a semi-empirical model is generally much better than for a statistical model, also the use of a semi-empirical model is much wider than the use of a statistical model.
- The amount of work required for developing a semi-empirical model will not differ much from the amount of work required for a statistical model. In both cases a number of test fires will have to be conducted to define and calibrate a number of fuel models representative of Sweden.
- Presently the performance and application of physical models is not at an acceptable level (due to for example the complexity which they are to model and the computational capabilities of the PC’s of today) for operational use.
The semi-empirical model for Sweden is recommended to be built upon Swedish conditions (i.e. built upon the type of vegetation found in Sweden) instead of trying to retrofit the local Swedish conditions into an existing model. This would most likely give the best output for Swedish conditions.
A system for better input data - weather and fuel data – should be worked on as well. This could for example take advantage of the results of the very promising “Alarm”-project that is being conducted in western part of Sweden.
Regarding the issue on better fuel data, new technology for satellite images or aerial photos and image classification techniques must be monitored as one major problem to be solved is distinguishing between the canopy fuel and the ground fuel.
For more specific conclusions and reflections, please see the analysis and discussion, and conclusions sections of this report.