Biodiversity publications archive

Biodiversity and Fire: The effects and effectiveness of fire management

Proceedings of the conference held 8-9 October 1994, Footscray, Melbourne
Biodiversity Series, Paper No. 8

Biodiversity Unit
Department of the Environment, Sport and Territories, 1996

18. The use of Geographic Information Systems to analyse wildfire threat

Mark Garvey
Risk Management Department, Country Fire Authority, Victoria

18.1 Abstract

Geographic Information Systems cannot prevent the ignition and spread of wildfires, however systems which have the ability to model the likely impact of fire support effective planning to reduce life and property loss.

The Country Fire Authority of Victoria is developing a Wildfire Threat Assessment program that will provide management and other groups with an overall profile of wildfire threat as well as delivering specific information on wildfire behaviour, population and fire statistics.

Key words: wildfire threat, geographic information systems, mitigation strategies, survivability, resource deployment, Victoria

18.2 Introduction

The Country Fire Authority of Victoria (CFA) has a corporate philosophy that 'people and their assets should be able to survive the passage of wildfire' (Country Fire Authority 1988). With a relatively wide range of resources at its disposal the CFA must identify where the threat from wildfire is highest and apply strategies and resources to mitigate that threat to survival.

The analysis of threat is a precursor to more efficient application of resources resulting in a safer wildfire environment for Victorians. Once threat has been characterised and mapped a number of moderating strategies such as deploying firefighting resources more effectively, identifying high threat areas where building construction practices can be modified, and targeting education programs can be applied to increase the survivability (See Figure 18.1). This paper describes the CFA's Wildfire Threat mapping program both in terms of the geographic information systems technology used and the application of the program outputs.

Figure 18.1: As the knowledge of wildfire threat increases so does the ability to select, target and apply resources.

Figure 18.1: As the knowledge of wildfire threat increases so does the ability to select, target and apply resources.

18.3 Growth in use of geographic information systems

Best practice management is increasingly becoming a requirement for all levels of the public and private sector. This is placing an obligation on management to use the best information possible to support corporate goals. Intelligent information is the key to making good decisions and the value of information to an organisation should be placed in the same category as plant, equipment and staff.

Organisations have always had data relating to their operations but only in the last decade have the data, analysis techniques and computers generally been able to efficiently merge disparate data sets into enhanced information sets. The information thus generated has become the basis for superior decision making. A Geographic Information System (GIS) is a computer system which is capable of combining several layers of geographical data to produce outputs in the form of summary statistics and maps. Geographic data are data which can be related to specific locations on the earth's surface. Roads, rivers, soils, vegetation and topography are examples of geographic data that are now capable of being analysed and integrated to produce value added information sets.

The utility of GIS technology lies with the fact that many scenarios can be viewed at the computer screen or as paper plots; each scenario altering a variable to determine the impact of that change on the overall output. When properly constructed and presented, outputs from the GIS (tables, reports and thematic maps) that summarise sometimes complex operations allow groups from varied backgrounds or levels of expertise to gauge the contribution that each component may have on an outcome.

In the past few years there has been enormous growth in the availability of geographic data suitable for computer analysis at a price that is within the reach of most Government Departments. Examples in Victoria include statewide coverage of data such as topography, hydrography, census data, road networks and cadastral boundaries. Data sets derived from satellite imagery and aerial photography such as vegetation, soils and built-up areas can also be added to the list.

Geographic data sets can be considerable in size. Computing requirements for storing large data sets have dropped dramatically. Gigabytes of data can be stored for analysis by workstations and even PC's for a few thousand dollars. The advent of CD-ROM and DAT tape technology has allowed for speedy and cheap storage of massive data sets to occur off the computer.

Software houses have enhanced the sophistication and usefulness of their products to the point that data analyses are capable with the strike of just a few keys or menu bars. In the GIS marketplace software products are being developed to fill niche markets and there now exists a number of capable PC GIS products and even some shareware GIS products.

In summary, there is presently a confluence of data availability, cost effective hardware and software in the GIS arena. It is enabling an increasing number of groups to analyse complex problems and allow for the results to be presented to a much wider audience.

18.4 Geographic information systems in wildfire threat mapping

A number of GIS based applications have been employed in the analysis of wildfire threat over the past ten years throughout Australia. Preplan (Kessell & Good 1985), the South Australian Bushfire Risk Analysis Program (Musto & Stubbs 1985), the Melbourne University Information System For Local Fire Hazard Management (Bishop & Cutler 1986) and the Western Australian Wildfire Threat Model (Muller 1993) have all used GIS concepts to integrate a number of data layers to provide an insight into wildfire threat. The applications all have in common an analysis of fire behaviour variables such as slope and vegetation; however, they also analyse data types unique to their own area.

At the Department of Conservation and Land Management in Western Australia (CALM) Muller (1993) has built a tool for use in the WA forests. He describes the aims of the system as:

Muller's list identifies the essence of the benefits of using GIS techniques to review the impact of wildfire. GISs cannot prevent the ignition and spread of wildfires, however, systems which have the ability to model wildfire behaviour and the likely impact of a wildfire allow effective planning to reduce the loss of life and property.

18.5 Country fire authority wildfire threat model

The CFA's interest in wildfire threat assessment using computer techniques began in the mid 1980's with the commissioning of a project to develop a threat assessment model for structural fires. During the development of this model it was acknowledged that a separate approach was necessary to model wildfire threat.

The need for improved information on the spatial distribution of wildfire threat has led to the development of the Wildfire Threat Model (WTM). A number of techniques already in use elsewhere in Australia (see above) were reviewed for their applicability to Victoria. However, they generally did not include life values or the built environment as their focus and this is a legislated responsibility of the CFA.

The aim of the CFA's WTM is to develop an appropriate database and rationale for assessing the requirements for fire prevention and suppression resources throughout Victoria. It will provide CFA management with a profile of the wildfire threat across the state. As well as giving a generalised classification, the model will deliver specific information on population, dwellings, terrain, wildfire behaviour, and wildfire statistics. This will, in turn, enable objective analysis of the current wildfire brigade structure and identify areas which may require specific fire protection or fire prevention strategies.

To date, the accepted approach to wildfire threat assessment in rural Victoria has been 'Fire Hazard Mapping' (FHM) (Morris & Barber 1980). FHM was designed to be completed at a local level by fire brigade personnel and others having a specific knowledge of an area. It was originally designed for the evaluation of an area's degree of threat in relation to its suitability for urban development. It was applied to an area which had been divided into broadly homogenous sub-areas but of unspecified size. Weightings were applied equally to ten categories. FHM required subjective input to generate data (such as length of fire season, number of fires, aspect and slope). Inconsistencies were thus incorporated into the data. FHM was not used to compare threat between geographically separated areas.

The expression 'Wildfire Threat' has been chosen to avoid confusion with a number of standard engineering terms such as 'Risk' and 'Hazard', both of which are used interchangeably and often with different meanings by many within the fire service. For the purposes of this project 'Wildfire Threat' is a classification system which reflects the potential wildfire behaviour, wildfire history, and the potential exposure of people and dwellings to wildfire within a specified area.

The WTM reflects the CFA's view of wildfire threat and is thus inclined toward the protection of life and property. This view of threat may not necessarily coincide with that of other services or communities where other values vulnerable to fire such as flora and fauna may have precedence.

The WTM and the various products derived from the databases are to be used at a number of scales. At a statewide level, WTM will enable review of wildfire equipment levels and fire brigade structure and possibly assist in the designation of wildfire prone areas. At a regional level the WTM provides input into short and long-term regional planning processes; at it's largest scale the WTM will assist in solving local issues such as the identification of areas unsuitable for development due to high loss potential and in providing an insight into the physical processes of wildfire at community group meetings.

The WTM is comprised of three modules : Fire Behaviour, Damage Potential and Fire Statistics (See Figure 18.2). The Fire Behaviour module combines vegetation data, litter weights, topography, low fuel areas and weather analysis in the GIS to produce a map of potential fire intensity. The Damage Potential module analyses the density of population and dwellings, and the agricultural productivity of a region. The Fire Statistics module examines wildfire statistics over a period of ten years. The three modules are then weighted and combined in a final layer – the Wildfire Threat Index.

Figure 18.2: Flow diagram of the Wildfire Threat Model.

Figure 18.2: Flow diagram of the Wildfire Threat Model.

18.5.1 Fire Behaviour module description

Buildings in wildfires are ignited by three means: embers, radiation and flame contact. Airborne burning embers form the most common cause of building ignition. They are a result of burning vegetation and other combustible materials within the fire, and become airborne due to the convective processes of the fire. They are then transported by strong winds which generally accompany a wildfire. The shower of embers may begin up to half an hour before the arrival of the flame front, and may continue for several hours afterwards. In contrast to the attack by embers, the passage of the fire front and the associated radiation and flames may take only minutes (Standards Association of Australia 1989).

As the intensity of a wildfire increases there is an increased amount of radiative and convective heat produced, which in turn preheats more fuel in the path of the fire to ignition temperature. In addition, the amount of embers produced is increased. More embers leads to increased chances of ignition, and increases in convection lead to more embers becoming airborne, possibly to produce spot fires downwind of the fire. (Standards Association of Australia 1989). Wilson (1984) identified fire intensity as the most important determinant of house survival in wildfires. Accordingly most methods of determining wildfire threat include fire intensity as a component. In summary fire intensity is a measure of the destructive capabilities of fires and is often measured in terms of kilowatts of heat per metre of fire front. By using the equations of Noble et al (1980) and Byram (1958) which describe the McArthur fire models, fire intensity can be calculated spatially in a GIS. Two sets of equations are used in this process. Different methods for calculating fire intensity are applied to forest and grassland fuels.

The fire behaviour module of the WTM combines data layers of vegetation, litter weights, topography and low fuel areas along with weather analyses to produce a map of the potential fire intensity of an area. Output maps are provided at a variety of scales. Generally a standard set of weightings are applied to the model; but specific scenarios are also catered for when mapping is produced for a particular audience such as a shire or community group.

18.5.2 Data for fire intensity calculation

A number of different data are used in the calculation of fire intensity. The sources of the various data are reported in Appendix 1. A generalised description of the data follows.

Vegetation data used in the WTM are derived from an interpreted statewide satellite imagery set and from large scale topographic maps. The following vegetation classes are generally used in the production of fire intensity maps: closed forest, scattered forest, conifer forest, scrub, vineyard, orchards and grasslands. Subsequent studies will investigate the means to include a much finer resolution of vegetation classes.

Litter weight is an important variable in the calculation of fire intensity. Accurate litter weights are normally derived from field sampling which is a time consuming and exacting process. Because the maintenance of an accurate inventory of litter weights on a Victoria wide basis is an almost impossible task, default litter weights are chosen and applied to each fuel type in a map. The fuel weights in the WTM generally represent the maximum litter load possible for that fuel type; however, weights representing a particular litter load have been used in the calculation of specific maps.

The mapping of litter loads is inherently easier in forests than at the urban-rural interface where there is likely to be modification of fuels, substitution of native fuels by exotic species, and more frequent litter weight reduction. Fire intensity could conceivably vary from low to high at the boundary of two properties. Maps including such detailed and site specific fuel regimes have not been included in the program to date.

The fire intensity calculations require as an input the McArthur Fire Danger Index (FDI). The FDI is derived from weather variables such as temperature, wind speed, humidity and rainfall and physical variables such as the curing of grasslands. The FDI is a scale from 0 to 100, with 0 indicating that weather conditions are such that a wildfire will burn with difficulty; 100 indicates that a wildfire will burn with such intensity that it could not be controlled (Luke & McArthur 1978).

To gain a long-term view of the FDI and how it varies between regions across the state, weather data from twenty meteorological stations throughout Victoria have been obtained. A minimum of twenty years of weather records have been analysed. For each reading, the Fire Danger Index (FDI) was calculated using McArthur's Mark 5 grassland and Mark 5 forest meter equations. The FDI data were then statistically summarised and the results used in the calculation of regional scale fire intensity maps. At larger scales specific weather conditions can be applied to predict potential fire intensity.

The Fire Behaviour Index which is carried through in the calculation of the overall Wildfire Threat Index is calculated by categorising data from the Fire Intensity Mapping program.

18.5.3 Damage Potential module

As the CFA is responsible for the protection of life and property in rural Victoria, high potential fire intensity values in isolation do not mean that suppression and prevention resources should be allocated to those areas. The Damage Potential module of the WTM utilises Australian Bureau of Statistics (ABS) data to estimate the density of people, the density of single dwellings and the average agricultural productivity of a region.

Data were obtained from three sources. The Census Collection District (CCD) boundaries were purchased from a private data supplier, whilst the 1991 census data is available from the ABS. The census data was extracted and matched to the appropriate CCD polygon. Comparative agricultural production values on a statistical local area basis and averaged over six years were derived from ABS statistics.

Output from the Damage Potential module is a series of maps depicting the density of people, dwellings and agricultural statistics data layers and is produced at a variety of scales. Specific outputs such as the estimation of numbers of people and houses within a fire district are also composed on request.

18.5.4 Fire Statistics Module

The probability of the occurrence of wildfire varies spatially. This variability has been incorporated into the model by analysing historical fire data.

Using CFA statistics the total number of wildfires and the total number of wildfires greater than 50 hectares over a period of ten years are collated.

Although the fire statistics have been gathered at the brigade level they are aggregated at the fire brigades group level for use in the model (a CFA fire brigades group generally consists of eight to 12 brigades). An internal analysis of CFA fire statistics identified errors (Krusel 1992). This affects the results of the Fire Statistics Index and is the major reason that the fire statistics have been collated at group level. Additionally, it would be preferable to include more data than ten years, however the computerised records are not available prior to 1980.

Group statistics are allocated to all brigades within the group. The Fire Statistics Index is calculated by categorising data from the total number of wildfires and the total number of wildfires greater than fifty hectares.

18.5.5 Wildfire Threat Index

The Wildfire Threat Index is an overall guide to the fire threat across a CFA region and is calculated by combining the Fire Behaviour Index, the Damage Potential Index and the Fire Statistics Index.

For display purposes Wildfire Threat maps are classified into five threat categories. Hydrological features and built up areas are pasted onto the final map. Fire station locations and roads are also plotted.

18.6 Wildfire threat mapping – example of use

CFA users of WTM maps analyse overall wildfire threat categories but also study fire intensity, population, housing and fire statistics maps to understand how each of the base maps contribute to the overall threat. For example, there may be a relatively high density of people or a large number of fires within a particular area; these can be studied and compared across a geographic area. The WTM is to be used as a planning tool both at the strategic or statewide level and at the local level. It will support management decisions regarding the formation of brigades, and the types and amounts
of equipment.

At CFA Regional headquarters, fire threat maps are analysed along with the present fire brigade structure. This is undertaken to identify areas where the threat is high and coverage or performance low. Strategic plans and annual plans would be adjusted to relieve shortcomings in the current structure.

A series of maps was produced to support a CFA case at a Planning Tribunal determining whether development should occur in the Plenty Gorge- approximately 30 kilometres to the north of the Melbourne. The maps consisted of a base potential fire intensity map, and transparent overlays of slope, vegetation, aspect and current administrative boundaries. The tribunal 'found the Plenty Gorge Potential Fire Intensity Map invaluable in assisting it in its deliberations' (Mitchell & Lewin 1994).

The CFA has developed a program titled 'Community Fireguard' for small community groups living in high threat areas. These groups take responsibility for their own fire safety and develop strategies for surviving major wildfires with the support of CFA facilitators. In this situation it is expected that localised fire behaviour maps along with transparent map overlays (such as slope, vegetation and aspect) will be one of a number of information tools that support the program.

The Building Code of Australia has recently been amended to include restrictions on construction methods in designated bushfire prone areas (Australian Uniform Building Regulations Coordination Council 1990). There are currently investigations into the use of GIS to assist in the designation process. It is expected that the data used in the CFA's wildfire threat mapping program will meet at least some of the requirements of the Code.

18.7 The future wildfire threat mapping program

The Wildfire Threat Mapping program is to be completed in it's present form by mid 1995. Maps encompassing all of Victoria will then be compiled to give a statewide view of threat. At that time, a review of the methodology will be undertaken to improve the model as necessary and updating the mapping program will begin again.

A number of areas of review have already been suggested to improve the model and increase the number of data layers available for analysis. These include: the mapping of firefighting resources and estimation of suppression capabilities; the inclusion of curing models to determine the long term 'fuel dryness' patterns in grasslands; an assessment of the means by which the vegetation and litter weights are delineated and characterised.

18.8 Conclusion

The utility of Geographic Information Systems lies in their ability to compress and integrate a number of varied and perhaps complex data sets such that the yields are understandable to a larger audience than would be the case if they were dealing with the raw data and mathematics.

Many fire management groups across Australia are beginning to utilise GIS in the study of fire processes. Many more groups are beginning to realise that the products of GIS improve their comprehension of wildfire and how it may effect them.

The CFA has established a wildfire threat analysis program that is providing a number of GIS based products facilitating more effective and responsible management of wildfire.

18.9 Acknowledgments

I would like to acknowledge the contributions of Mr Stephen Petris who reviewed the text and Mr Gareth Finney who helped with the diagrams. Dr Malcolm Gill of the C.S.I.R.O. Division of Plant Industry kindly provided processed weather data from eight Victorian sites.

18.10 References

Australian Uniform Building Regulations Co-ordination Council Building Code of Australia Part G5, 1990.

Bishop I.D. & Cutler M. 1986, 'Information system for local fire hazard management' School of Environment and Planning University of Melbourne.

Byram G.M. 1959, 'Combustion of forest fuels' in Forest Fire Control and Use, ed. K.P. Davis, McGraw Hill, New York.

Country Fire Authority 1988 Fire Prevention Brochure.

Kessell S.R. &Good R.B. 1985, 'Technological advances in bushfire management and planning' in Symposium – Natural Disasters in Australia. Sydney. October, 1985.

Krusel N. 1992, 'Error analysis of CFA Fire Reports held on computer for seasons 1980-11 and 84-90', Country Fire Authority Internal Report

Luke R.H. & McArthur A.G. 1978, Bushfires in Australia, Australian Government Publishing Service, Canberra.

Mitchell K. & Lewin J. 1994, Planning and Environment Act 1987 Panel Report L46 Diamond Valley Planning Scheme July, 1994.

Morris W. & Barber J.R. 1980 Fire Hazard Mapping Country Fire Authority of Victoria Booklet

Muller. C. 1993, 'Wildfire Threat Analysis – An Effective Decision Support System for Fire Protection Planning' in The Burning Question: Fire in NSW Conference 5-7 August, 1993. Coffs Harbour.

Musto I.P. & Stubbs T.J. 1985, Adelaide Hills Bushfire Hazard Mapping Project, S.A. Department of Environment and Planning.

Noble I.R., Barry G.A.V. & Gill A.M. 1980, 'McArthur's fire danger meters expressed as equations', Australian Journal of Ecology. vol. 5, pp. 201 - 203.

Simmons D. & Adams R. 1986, 'Fuel dynamics in an urban fringe dry sclerophyll forest in Victoria', Australian Journal of Forestry, vol. 49, no. 3, pp. 149-154.

Standards Association of Australia, Public Review Draft Standard – Guide to Development and Construction in Bushfire Prone Areas, Committee BD/64 July 1989.

Wilson A. 1984, 'Assessing the bushfire hazard of houses: a quantitative approach', Rural Fire Research Centre, Technical Paper, no. 6.

18.11 Appendix 1: Availability, use and discussion of digital data used in the Wildfire Threat Model

Digital data suitable for use in the Geographic Information System is obtained from a number of sources. Some data are merely purchases from licensed resellers; a number require licensing agreements for their use.

Most of the information required by the WTM has been obtained. The license for the use of the 1:100,000 map base for all of Victoria has been purchased from the AUStralian Land Information Group (AUSLIG).

Census Data : Census data for all of Victoria from the 1991 census has been purchased from the Australian Bureau of Statistics as have agricultural statistics data.

Weather Data : Weather records were purchased from the Bureau of Meteorology. The data needed much cleaning and correction before it was suitable for analysis.

Vegetation Data : The Victorian Department of Conservation and Natural Resources has generated a digital map of vegetation (tree/no tree) for all Victoria from interpreted satellite imagery at an accuracy of one hectare. This data is used in the model. Further developments of this dataset will include species identification. Improved data will be included as they become available.

Litter Weights : Litter weight is an important variable in the fire intensity equation. Default litter weights have been allocated to vegetation polygons from the work of Simmons & Adams (1986) as well as input from field personnel such as CFA Regional Officers and foresters. The weights reflect the upper limit of litter load for each particular fuel type.

Low Fire Fuel Weight Data : Low fuel weight areas such as 'Built up areas' are extracted from the infrastructure layer of the data supplied by AUSLIG and the Victorian Department of Survey and Mapping (DSM). Once the square black dots which are used to depict individual buildings run into each other and are no longer decipherable, the 'built-up area' code is used. An estimate of the threshold of the building density for the built-up class is four buildings per hectare. Also included in this coverage are other low fire fuel areas such as industrial estates, quarries, sporting ovals, golf courses and parks.

Hydrological Data : The location of water features were extracted from AUSLIG data and DSM.

Contours : Contour maps are obtained from AUSLIG and DSM. The GIS interpolates the contours and generates a cell map where each cell is given a slope value. The slope values are then incorporated into the fire intensity equations.

CFA Brigade Locations : The location of CFA brigades are first checked by CFA regional staff. Once the brigade location data is in the GIS thiessen polygons are generated around each point as an estimate of brigade boundaries. Whilst these lines do not represent the exact location of boundaries it a reasonable estimate of the brigades' operational area. It is not yet feasible to maintain digital files of true brigade boundaries statewide.

Fire Statistics : CFA Fire statistics from the period 1980-90 have been used in the study to date. Subsequent mapping programs will include data from 1990 onwards and also fire statistics from the Victorian Department of Conservation and Natural Resources.