Science, technology, and human factors in fire danger rating: the Canadian experience.

Additional keywords: Canada; Canadian Forest Fire Danger Rating System; fire behavior; technology transfer; wildland fire research.

Forest fire danger rating schemes underlie all contemporary fire management systems. These systems are the principal means by which scientific knowledge of fire potential is synthesized and integrated with operational experience into practical fire management applications. Many forest fire danger rating systems have been developed throughout the world (Lin 2000; San-Miguel-Ayanz et al. 2003), although no universal or leading system has emerged. Fire management agencies developing new systems often look to use or adapt such well-developed existing systems as the US National Fire Danger Rating System (Deeming et al. 1977), the McArthur fire danger rating meters used in Australia (Luke and McArthur 1978) or the Canadian Forest Fire Danger Rating System (CFFDRS) (Stocks et al. 1989). The CFFDRS, for example, has been fully implemented in parts of the USA and in New Zealand, and components of the system have been used in countries such as Fiji, Argentina, Mexico, Indonesia, and Malaysia. The CFFDRS is accepted outside of Canada likely because it is relatively simple to use, robust in a variety of environments, and has strong interpretive products (i.e. posters, look-up tables, electronic data processing and display systems) that are useful in a variety of situations.

In Canada, the CFFDRS is the principal source of fire intelligence for all forest fire management agencies. It is used to support fire management decision making at strategic and tactical levels, from fire prevention to firefighter safety. The development and implementation of the CFFDRS was challenging. Canada is a large country; although boreal forests are the predominant forest-cover type, fire environments range from coastal rain forests and taiga to grassland and semi-desert. Land management objectives, fire load and fire management capacities vary across the country. As well, the federal government has a limited role in operational forest and fire management, and must work with provincial and territorial agencies to implement national programs in a kind of cooperative federalism (e.g. Wilson 2000).

The objective of the present paper was to address the question ‘What makes for an effective national wildland fire danger rating system?’ This was done by reviewing the purpose and principles of fire danger rating, and by examining the development and implementation of the CFFDRS as a case study. Four key scientific, technical and human/institutional elements or strategies were identified that can inform the further development of fire danger rating in Canada, the use of the CFFDRS in other countries, and the development of other fire danger rating systems.

In the 5th century BC, military strategist Sun Tzu wrote that ‘one who knows the enemy and knows himself will not be in danger in a hundred battles’.¹ In contemporary Canadian fire management a forest fire may be considered a threat to forest resources, property, and public safety, or an important ecological process, depending on where and when it occurs. However, Sun Tzu’s principle remains valid: a fire-intelligence system is an important element of any effective fire organization (Barrows 1969). A major component of any forest fire intelligence and management system is a fire danger rating system.

Fire danger refers to an assessment of both fixed and variable factors of the fire environment (i.e. fuels, weather, and topography) that determine the ease of ignition, rate of spread, difficulty of control, and impact of wildland fires (Merrill and Alexander 1987). Fire danger rating began in the temperate and boreal forests of North America, where fire danger typically varies throughout the fire season as weather systems bring rain, and are followed by drying periods of varying frequency, intensity, and duration. In times of greater fire danger, there may be many ignitions across large areas in short periods of time, resulting in peaks in fire management service demand. In Canada, at least, fire seasons are usually of limited duration, beginning in spring after snowmelt and ending in late autumn when snow cover resumes.

Wildland fire management is principally an economic activity. Fire danger rating systems were first developed for regions where the fire environment (weather, fuels, and topography) varies in space and time, and where fire management resources are costly and limited (Beall 1967). Their purpose is to provide a way to efficiently allocate an appropriate level of resources across a region or country from day to day or place to place, on the basis of existing and forecasted fire danger levels. The process of systematically evaluating and integrating individual and combined effects of factors influencing fire potential is referred to as ‘fire danger rating’. Systems that rate fire danger provide for one or more qualitative and numerical indices of ignition potential and probable fire behavior (Countryman 1966).

Fire danger measures must provide useful information to support fire management decisions. The primary role of fire danger rating systems in North America is to enable fire managers to properly judge levels of preparedness needed and corresponding suppression resources required to keep wildfire losses or adverse impacts to a minimum. To the Canadian public, fire danger indicator road signs are still the most visible evidence of the existence of a fire danger rating system (Fig. 1). However, fire danger rating system outputs are used in a variety of fire management activities. These include: prevention planning (e.g. informing the public of impending fire danger, regulating access and risk associated with public and industrial use of forest and rural areas); preparedness planning (i.e. level of readiness and pre-positioning of suppression resources); detection planning (e.g. lookout staffing and aircraft scheduling and routing); initial-attack dispatching (e.g. prioritizing of targets for air tankers and ground crews); formulating suppression plans on active wildfires (including short-range predictions of fire spread and behavior); evaluating fire behavior potential and guidelines for safe work practices for firefighters; escaped fire situation analysis (including long-range projections of fire growth and behavior); prescribed-fire planning and execution (which includes smoke management); fire and fuel management modeling and planning; and fire behavior training.

Fire management objectives change over time and space; we can distinguish at least four developmental stages in fire management programs in Canada:

1. Unregulated use of wildland fire by rural and aboriginal peoples as a part of traditional land management practices;

2. Government agencies begin to control fires to prevent unwanted damage to timber or other state and private resources as development and competition for resources increases, and to reduce people-caused fires by instituting fire laws, education programs, and suppressing traditional practices;

3. Government agencies and private concerns attempt to control all wildfires and restrict the use of prescribed fire to fire managers; wildland fire is institutionalized;

4. Realization that it is not possible or ecologically or economically desirable to control all fires. Wildfires may be allowed to burn in some areas where they play a natural role, and prescribed fire may be used to manage fuels and maintain ecological integrity.

As fire management objectives change, fire danger rating systems must also evolve to support more complex decision making. For example, a more thorough understanding of the complexities involved in forecasting fire behavior (Alexander and Thomas 2004) is needed to decide between suppressing a fire in a natural area or letting it burn freely.

A fire danger scheme, like a smoke detector, should sound an alarm before fire danger and difficulty of control reach extreme levels, thus allowing fire managers time to prepare and take preventative action. Nelson (1955) outlined five general principles for the measurement of fire danger:

1. A fairly simple method of measuring key variables such as fuel moisture, wind, and rain, and a way of integrating these variables into numerical values;

2. Close adherence to standards for fire weather station location and instrumentation that are established for the particular system in use;

3. Careful training of fire weather observers;

4. Periodic and thorough inspection of fire weather stations; and

5. Continuity of fire weather and fire danger records.

Fire danger systems must be based on fire danger factors that are easy to measure accurately, and give consistent measures of fire danger from place to place and time to time, with approximately the same antecedent environmental conditions. Fire danger rating systems must also integrate a large number of fire danger parameters in simple, easy-to-use, and yet soundly based systems. As Van Wagner (1970) notes:

Forest fire danger rating is a fascinating but exasperating branch of forest research. The goal is easily stated: Make an index such that any given index value will always represent the same fire behaviour, no matter what weather history leads up to it. The trouble is, one quickly outruns the available practical knowledge and theory. A liberal dose of philosophy is therefore required as well.

Thus, fire danger rating, as with other aspects of forestry, involves both science and professional judgment. Van Wagner (1971) further noted that ‘if it is complicated, then the complexity should be buried out of sight, as in prepared tables or computer programs’.

Early cognitive science research suggested that the span of memory for a particular type of information is about seven items or factors (Miller 1956). While span of memory varies among individuals, the use of systems that integrate many significant factors affecting fire danger into a few indices is sound. A stronger argument is the advantage of substituting an objective method for the opinions of individuals in assessing risk and allocating resources, particularly as fire danger across a region or country is beyond any one individual’s direct experience. Some advantages and disadvantages of using a fire danger rating system are listed in Table 1.

Table 1.  Comparison of fire management with and without the use of a fire danger rating system (from Van Wilgen and Burgan 1984)

Fire danger rating systems attempt to simulate reality, but often fall short of it. In meeting the objective of simplified relationships, minor factors are neglected and systems are usually based on single sets of idealized conditions. If certain physical fundamentals are observed, this permits approximations that are close enough for many purposes – but they are approximations only. Consequently, fire danger rating systems tend to be applied beyond their field of usefulness. To avoid this tendency, the assumptions on which the systems are based and the range of conditions under which the systems are valid need to be defined carefully and checked frequently (Brown and Davis 1973). Most importantly, fire danger values must be correlated with empirical measures of fire occurrence or severity through experimentation and analysis of historical records (Harvey et al. 1986; Viegas et al. 1999; Cruz et al. 2003).

The use of fire danger systems can result in two types of error, which are illustrated for two conditions, ‘High’ and ‘Low’ fire danger levels in Table 2:

Table 2.  Types of fire danger rating errors

• Type I errors (errors of commission or false positives) occur when the system ‘sounds an alarm’, but no real potential for serious fires exists: fire danger is overestimated;

• Type II errors (errors of omission or false negatives) occur when serious fires take place prior to the system ‘sounding an alarm’ or when the system ‘sounds no alarm at all’: fire danger is underestimated.

Type I errors can result in fire management agencies overcommitting expensive resources, thereby incurring excessive pre-suppression costs. Type II errors can result in fire management agencies failing to anticipate increasing fire danger and being unprepared, leading to potential initial attack failures and incurring excessive suppression and damage costs. It is usually most important to minimize Type II errors or false negatives. However, it is also important that the system not be on high alert all the time – a fire danger system must isolate with certainty those days each season on which extreme fire danger conditions occur (Andrews 1987). In addition to weaknesses in fire danger systems or input data, human factors such as cognitive fixation – the failure to revise situation assessments as new information is received (De Keyser and Woods 1990) – may result in both Type I and II errors of interpretation.

In as much as fire danger rating systems are a form of media in McLuhan’s (1994) sense of media as any extension of human powers or senses, media theory provides a useful framework for understanding the evolution and impact of fire danger rating systems. McLuhan (1994) held that ‘The “message” of any medium or technology is the change of scale or pace or pattern that it introduces into human affairs’. Innis (1972) suggested that media are not only human constructs that attempt to fill perceived communication needs, but new media can influence culture and society, and in turn effect the development of further technology. McLuhan and McLuhan (1988) articulated four laws of media. Every new media extends or enhances some human ability or faculty; obsolesces an older form; builds on an older form; and reverses its properties when pushed to its limits. In this sense, new fire danger systems extend our ability to process and perceive fire danger in space and time beyond individual experience; they may displace daily weather observations as primary measures of fire potential; they build on the concept of fuel moisture integrators in the form of mathematical models; and pushed to the extreme they may reverse or amputate experiential-based judgment. While fire danger rating systems necessarily grow and change in response to changes in fire management, they may also influence how fire management is practised, and in turn affect the development of new technologies. Successful media are readily assimilated by and influence a culture – the fire management organizational culture – in the case of fire danger rating systems.

Systematic fire management began in North America in the early 1900s with the recognition that fire protection was essential to the development of a forest industry. Early efforts at fire danger rating attempted to describe the fire problem in relation to the moisture content of critical fuels, with emphasis on fine fuels important to ignition and early spread. Different approaches to estimating fuel moisture have been used over the years. These include: direct measurement; fuel moisture analogs such as fuel moisture sticks, duff hygrometers, and evaporation devices; correlation of fuel moisture with weather elements such as relative humidity; and cumulating the effects of current weather and past weather to describe the rate of change in fuel moisture in various fuels with different response times or timelags.

Research into forest-fuel flammability began in Canada in the mid-1920s (Williams 1964; Beall 1990). Fire danger rating pioneers James G. Wright and Herbert W. Beall produced their first set of fire danger tables in 1933. These were based on a ‘Tracer Index’ that related moisture content of needle litter and top-layer duff in red and white pine stands to behavior of small-scale test fires (Wright 1933). This early work led to the development of a number of regional fire danger rating tables by the 1960s (Williams 1964).

In the 1950s and early 1960s, an expanding forest industry spurred the development of federal forest research centres across Canada, and renewed efforts in fire research. In 1965, a fire danger working group, representing Canadian Forest Service (CFS) fire researchers from across the country, was formed to guide development of what is now the CFFDRS. In 1967, a modular national fire danger rating system was proposed to replace a variety of regional systems (Muraro 1968). It has two primary subsystems (Fig. 2a) – the Canadian Forest Fire Weather Index (FWI) System and the Canadian Forest Fire Behavior Prediction (FBP) System (Stocks et al. 1989). Elements of an Accessory Fuel Moisture System (e.g. Lawson et al. 1996; Wotton et al. 2005) and a Fire Occurrence Prediction System, including models for predicting flaming and smoldering ignition potential (e.g. Lawson and Dalrymple 1996a; Lawson et al. 1997a; Otway 2005) have been developed, but they have not been implemented on a national basis.

A provisional version of the FWI System was field tested at selected locations during the 1969 fire season, and the first edition was issued in 1970. The FWI System consists of six components (Fig. 2b). The Fine Fuel Moisture Code (FFMC) represents the moisture content of surface litter, which is key to ignition and fire spread, and is derived from Wright and Beall’s Tracer Index. The Duff Moisture Code (DMC) and Drought Code (DC) represent the moisture content of shallow and deep organic layers, respectively. These organic layers are important to surface fire intensity, crowning potential, and difficulty of control in temperate and boreal coniferous forests. Three fire behavior indexes – the Initial Spread Index (ISI), Buildup Index (BUI), and FWI (patterned after Byram’s [1959] concept of fire intensity) were developed and scaled in relation to fire behavior observations of small- to moderate-scale test fires in a standard or reference fuel type (i.e. a mature jack pine or lodgepole pine stand), but can be correlated with fire behavior in other fuel types (Van Wagner 1987a).

The three moisture codes of the FWI System do not represent the moisture content of all components of all fuel complexes in Canada. Fuel-specific moisture content models have also been developed for rapidly wetting and drying reindeer lichen (Cladonia spp.) (Pech 1989) and slowly wetting and drying downed woody fuels (Van Wagner 1987b), but they are not implemented on a national basis. The DMC and DC components are calibrated against organic layers that differ distinctly from the standard pine type (Lawson and Dalrymple 1996b; Lawson et al. 1997b; Wilmore 2001; Otway 2005).

All FWI System values are calculated from four simple weather observations (Turner and Lawson 1978) – temperature, relative humidity, wind speed, and 24-h accumulated precipitation – as recorded at noon local standard time (LST), but the values represent conditions at ~1600 LST, which is the peak fire danger period (Van Wagner 1987a). Other schemes have been developed to model or calculate diurnal variation in the FFMC and, in turn, the ISI and FWI components at any time of the day (e.g. Lawson et al. 1996; Beck and Armitage 2004).

As soon as the FWI System was released in 1970, work began on a system to quantitatively predict primary behavior characteristics (McAlpine et al. 1990). Experimental fires were carried out in major Canadian fuel types, including coniferous and deciduous forests and logging slash (Alexander and Quintilio 1990; Stocks et al. 2004), over a range of burning conditions. These data, combined with observations from selected wildfires, were correlated with FWI System components. Primary relationships between the ISI and rate of spread, and the BUI and fuel consumption were modeled. Experimental fire data from Australia were used for grasslands (Cheney and Sullivan 1997). An interim edition of this FBP System was released in 1984 (Lawson et al. 1985), and the first complete edition of the FBP System (Fig. 2a) was published 8 years later (Forestry Canada Fire Danger Group 1992).

Lin (2000) notes that most fire danger rating systems in use in the world do not account well for human factors affecting fire occurrence. Efforts have been made in some Canadian provinces to develop prediction systems for both human-caused and lightning-caused fire occurrence (Martell et al. 1989; Todd and Kourtz 1991; Kourtz and Todd 1992; Vega-Garcia et al. 1995; Anderson 2002; Wotton et al. 2003; Wotton and Martell 2005); however, they have not been integrated into a national system.

A number of studies relate FWI System components to daily, monthly, and annual fire occurrence and area burned across Canada. For example, Harrington et al. (1983) found that certain FWI System components recorded at individual fire weather stations explained from 0 to 43% of the variance in monthly area burned by Canadian provinces. Anderson (2002) found that a model of lightning-caused fire occurrence incorporating FWI System values and number of lightning strikes correctly predicted the number of lightning-caused fire starts on 56% of days over a 5-year period in Saskatchewan. This model correctly predicted low fire activity days 90% of the time.

The CFFDRS was completed largely due to the dedicated efforts of a small group of researchers in the CFS Fire Danger Group who, for more than two decades, continued a line of work that began 40 years earlier. The fact that the group represented different regions of the country no doubt contributed to the national flavor of the CFFDRS.

Although Canadian and American fire researchers have communicated frequently over many decades (Beall 1990) and share some common views (Alexander and Andrews 1989), there was little formal collaboration in the pursuit of what has amounted to two very different approaches to modeling and predicting fire danger and fire behavior. In April 1992, researchers from the CFS and US Forest Service met in Missoula, Montana, to discuss development of a common North American fire danger rating system. While this was not resolved, the discussions led to a formal Canada–USA–Australia cooperative agreement on wildland fire research (McCaw and Alexander 1994), and in part to the International Crown Fire Modeling Experiment, which took place in Canada’s Northwest Territories from 1995 to 2001 (Stocks et al. 2004). Such cooperative work may lead to the development of common fire behavior models in North America, if not entire systems.

The present FWI System was designed in the 1970s before computers and electronic communications technology were widely available. It was therefore designed so that simple observations from manual fire weather stations could be used and FWI System values could be determined from look-up tables (Canadian Forestry Service 1984).

One of the first ways in which the FWI System was implemented was in fire danger class schemes. Fire danger levels were based on correlations between FWI System values and fire activity developed from analyses of historical fire weather and occurrence data (e.g. Turner 1973). Fire danger levels must also be determined in relation to fire management objectives, while the kinds of interpretations depend on values put at risk by fire and on the fire management resources or tools available. In British Columbia, for example, fire danger classes were defined in terms of FWI and BUI values for three danger regions with different fire climates (BCMF 1983). The fire danger classes were incorporated into fire prevention regulations that required forest operators to take preventative measures such as the provision of watchmen and equipment, and to abide by industrial forest closures. Guidelines for resource levels, patrols, and alert status were also tied to fire danger class (BCMF 1983). Fire danger classes were an important means for guiding objective systematic decision making but, as with any tool, they required interpretation. Fire danger classes are still used in British Columbia in the form of the Forest Fire Prevention and Suppression Regulations.² Effectively communicating fire danger to the public and other interest groups is an important part of fire prevention programs (Dawson 1991). However, at present there is no common fire prevention message in Canada.

After its release in 1992, the first complete edition of the FBP System was implemented in several provincial fire management systems, as well as in commercial software that is primarily used by provincial and regional fire centres and by specialists. In 1994, there were several near-entrapments of firefighters in British Columbia. A safety review of the incidents recommended that fire behavior, fire potential, and safe work practices be emphasized in all training and that a firefighter safety awareness campaign be developed (BCMF 1994). In 1995, the CFS field tested a prototype field guide to the FBP System to assist in this effort (Taylor et al. 1997). Reflecting on Van Wagner’s (1971) advice, the guide was designed to be simple to use and formatted so that significant thresholds could be readily identified (e.g. surface- to crown-fire transition). Popularly known as the ‘red book’, the field guide is used at the tactical level and in training courses by all fire management agencies in Canada, enabling operational staff to make fire-behavior predictions that previously were limited to specialists. Although FBP System software is available for small computers such as palmtops, computers are not yet widely used in field operations in Canada (i.e. ‘on the fireline’). In the late 1990s, the British Columbia Ministry of Forests adopted a policy of issuing fire-behavior warnings and removing crews from direct attack at the head of the fire when fire intensities are forecasted to exceed 4000 kW/m (Beck et al. 2002), a threshold identified in the field guide.

The development of remote automatic weather stations and communications technology in the 1980s and 1990s permitted collection of weather data from isolated locations in almost real-time on a provincial and even national basis. The effect of this automation could not have been anticipated and has been fortuitous. Automation has permitted the development of sophisticated fire management systems such as the Spatial Fire Management System to process, interpolate, and display fire weather data in concert with fuels and topographic data using geographic information system-based engines (Lee et al. 2002). Internet-based map displays such as the Canadian Wildland Fire Information System (Lee 1995) have also been developed to distribute text and map products to fire centres and other users across the country (http://cwfis.cfs.nrcan.gc.ca/en/index_e.php, accessed 15 September 2005), and around the world. An example of such a map display for south-east Asia is shown in Fig. 3. Systems have also been developed to guide daily and seasonal resourcing levels (e.g. Anderson and Lee 1991; Beck 2004), based in part on CFFDRS outputs and in some cases spatial statistical analyses. Computerized fire management decision support systems are very effective for communicating spatial variation in fire danger, allowing for the use of spatial analysis or visual interpretation in resource allocation decisions. They are a major reason for the expanded use and application of the CFFDRS at strategic levels.

Thus, there continues to be a place for a wide range of tools, from networks of automatic weather stations linked with sophisticated computer systems to basic weather instruments and look-up tables, to provide information on fire potential for users with different needs, capabilities, and resources.

Fire danger rating systems must also be cost effective for fire management agencies to use. According to a Government of Canada review of the CFFDRS undertaken in 1987 and 1988, for the period 1971–1982, at least $CAN 750 million in benefits (i.e. in terms of a reduction in firefighting expenditures) could be directly attributable to the system for Canada as a whole. The benefit-to-cost ratio was ~3 : 1 (Moore and Newstead 1992).

Although the CFS has led forest fire danger research in Canada, the cooperation of provincial and territorial fire management agencies has been critical to successful development and implementation of the CFFDRS. Fire management agencies lead implementation of the CFFDRS through training and operational practices. Although good working relationships between researchers and operational staff are crucial, formal working groups and cooperative agreements between federal and provincial agencies have also been important in setting policy and direction. These groups included, most importantly, the Canadian Committee on Forest Fire Control (CCFFC), later renamed the Canadian Committee on Forest Fire Management (CCFFM), which operated between 1952 and 1997 under the auspices of the National Research Council of Canada. Its membership included representatives of fire management agencies, universities and technical schools, the forest industry, and the CFS. Throughout the committee’s 45 years, various subcommittees and task groups on fire research, terms definition, equipment, training, education, and communications addressed important common issues. While the CCFFC/CCFFM did not provide direct funding for fire research, it was an important vehicle for communication between the fire research community and fire management agencies across the country during the development and implementation phases of the CFFDRS.

In 1981, the Canadian Interagency Forest Fire Centre (CIFFC) was formed to manage the exchange of resources across Canada under the Mutual Aid and Resource Sharing (MARS) Agreement. Members represent federal, provincial, and territorial fire management agencies. The centre is one-third funded by the federal government, with the balance shared among the provinces. When the CCFFM was disbanded in 1996, the CIFFC’s role expanded. Six CIFFC working groups – aviation, resource management, fire equipment, fire science and technology (S&T), national training, and forest and fire meteorology – formed to address common problems and issues, and resource sharing on a nation-wide basis. The three latter groups have been increasingly active in the continuing development, implementation, and application of fire danger rating in Canada.

The CIFFC S&T working group exchanges information on developments in fire science and technology, and shares funding of research and development projects of common interest. They helped acquire funding to develop PROMETHEUS, a wildland fire growth model (Tymstra 2002). The CIFFC Forest and Fire Meteorology Working Group is developing a common approach to fire weather data standards and sharing. The CIFFC National Training Working Group establishes training standards for staff exchanged under the MARS Agreement, oversees implementation of national courses, and shares resources to develop new national courses on such topics as fire behavior, firefighter safety, and fire weather (Thorburn et al. 2000, 2003; St John and Alexander 2004). Starting in the mid-1990s, two national training courses – Advanced Wildland Fire Behavior and Wildland Fire Behavior Specialist (Alexander and Van Nest 1995) – were developed and delivered by CIFFC members. These national courses, as well as other, regional courses (e.g. de Groot 1989), have contributed to the degree of acceptance and implementation of the CFFDRS in Canada.

Other cooperative groups formed in the 1960s–1980s, which helped bring fire management and fire research staff together at regional levels. These included the Northwest Fire Council (Alaska, Yukon, British Columbia, Washington, and Oregon), the Interior West Fire Council (Northwest Territories, Alberta, Saskatchewan, Montana, Idaho, Colorado, Utah, Wyoming, Kansas, Nebraska, North Dakota, and South Dakota), the Great Lakes Forest Fire Compact (Manitoba, Ontario, Michigan, Minnesota, and Wisconsin), and the Northeast Fire Protection Compact (Quebec, New Brunswick, Nova Scotia, Maine, Massachusetts, Connecticut, Vermont, New York, and Rhode Island). Such groups sponsored annual and biennial meetings for operational staff to discuss common problems and innovations in fire management and research (e.g. Lawson et al. 1989; Alexander and Bisgrove 1990).

The CFFDRS has been implemented in whole or in part in a number of countries: these experiences provide useful lessons regarding the transfer of fire danger rating knowledge from one environment to another. New Zealand and Fiji adopted the FWI System for use in exotic pine plantations in 1980 (Valentine 1978) and 1988 (Alexander 1989), respectively. New Zealand adopted the FWI System in part because of its simplicity, the similarity of climate and topographic condition to British Columbia, and the fact that decision aids for prescribed burning of logging slash tied to the FWI System existed (Muraro 1975). For many years, little work was done on local adaptation (except for modification of the daylength influence on drying at lower latitudes) or technology transfer, and so the system’s full benefits were not realized (Fogarty et al. 1998). Subsequently, the FWI System was used for fire danger rating in a range of vegetation types, including shrublands for which it was not designed. In 1992, a research program began to develop new models for indigenous fuel types (mainly shrublands), and to improve technology and information transfer (e.g. Fogarty 1994, 1996; Alexander and Fogarty 2002). Alexander (1994) developed fire danger classification schemes for both grasslands and exotic pine forests based on FBP System models. An evaluation of the latter indicated that the initial decision to adopt the FWI System remained valid (Pearce and Alexander 1994).

From 1981 to 1991, the Ontario Ministry of Natural Resources carried out a project to develop a model forest fire management system, including a fire danger rating system, in north-eastern China (Lynham and Stocks 1990; Thomas 1990; White and Rush 1990). This included establishing a network of Canadian-made fire weather stations, producing fuel-type maps, and implementing both the FWI System and a computer system to prepare daily strategic plans. Many of the implementation issues encountered by Canadian staff in China derived from cultural differences. Decisions related to fire suppression were often made at a very senior level. Also, the use of Canadian-made electronics and fire weather instruments posed maintenance problems.

The Alaska wildland fire community adopted the CFFDRS in 1990 (Alexander and Cole 1995), mainly because of the similarity between fuel types in Alaska and northern Canada. Since then, other northern USA states, such as Minnesota and Michigan, have adopted portions of the system. Despite similar environments, adjustments for particular conditions may need to be considered. For example, in Alaska, the effect of permafrost on wetting and drying processes with respect to the DMC and DC requires investigation (Wilmore 2001; Jandt et al. 2005), and local interpretive guidelines tied to FWI System component values still need to be developed (Alexander and Cole 2001).

The CFFDRS has been used, in part, in the development of computerized fire management decision-support systems in Mexico, Florida (Brenner et al. 1998), and south-east Asia, largely because of its simplicity and its strong interpretive products (Lee et al. 2002). In 1999, the CFS began a 5-year project to implement a danger rating system in south-east Asia based on the CFFDRS. The CFFDRS was adapted to serve as an early warning system for haze events, because of their serious negative impacts on regional health and economy (Field et al. 2004). Through early warning, strategies to reduce open burning during critical periods were developed with local agencies. The CFFDRS was adapted using airport visibility data to calibrate the DC as a smoke-potential indicator, using grass ignition and hotspot data to calibrate the FFMC as an ignition indicator, and using the FBP System O-1 fuel type rate of spread model to indicate difficulty of controlling grassland fires (de Groot et al. 2005, 2006).

In Argentina, the National Fire Management Plan began to implement the FWI System in three pilot areas in 2000. In a review of the project, Taylor (2001) suggested that institutional mechanisms be created and strengthened to enhance communication and cooperation between national and provincial agencies before implementing the system nationally. Taylor (2001) also noted that countries implementing the CFFDRS would probably go through similar steps to develop a fire danger rating system as Canada has gone through, although perhaps over a shorter time period.

Several European countries, including Sweden (Granstrom and Schimmel 1998), Portugal, and Spain, have also adopted portions of the CFFDRS. Viegas et al. (1999) found that FWI System components were correlated with fire activity in southern Portugal, Spain, France, and Italy, even though the vegetation, fuel types, and dry Mediterranean climate differ from those in Canada. Not surprisingly, the FWI was found to be a good predictor of ignition in pine and spruce forests in Finland (Tanskanen et al. 2005), which are more similar to the Canadian boreal forest.

Although forest fire danger rating systems have their origins some 80 years ago they continue to be one of the most important means by which scientific knowledge of wildland fire is applied in fire management. It is important to understand what underlies their successful and unsuccessful application. Experience with fire danger rating in Canada and elsewhere (e.g. Peet 1980) reveals important lessons. Developing a national fire danger rating system and implementing it across a number of fire management agencies is a long and complex process, with scientific, technological, human, and organizational/institutional dimensions. Four elements need to be integrated in national fire danger rating systems (Fig. 2a).

Fuels wet and dry according to the same physical processes regardless of location, and fires respond to variation in the same physical influences of fuel, weather, and topography. Although the CFFDRS was designed for boreal and temperate forest fuel and weather conditions, in other regions where conditions are different, elements of existing fire danger systems, such as the CFFDRS, can often be successfully applied. The key to adapting an existing fire danger rating system is to identify moisture models that represent the fuels that are important to fire behavior in the new environment, and to develop new relationships or modify existing ones as necessary. This process begins with identification of fuel types of most concern, and fuel elements within them that are important to ignition, spread, difficulty of control, and impact of wildfires. Seasonality of fuel conditions and diurnal variations should also be investigated and addressed (McArthur 1971).

Developing a national fire danger rating system is a lengthy process: countries adopting such systems will likely go through many of the same stages to develop fire danger rating systems as Canada has gone through. It is important for adopting countries carrying out their own research to test and adapt their systems, to develop local interpretive products, and to participate in the international fire research community. This requires a long-term commitment to fire research and training by fire management and science agencies. A dedicated professional staff must remain committed to the work over a long period. Timelines and expectations must be realistic. Canada has a long and rich history of fire danger rating research. This foundation of knowledge and experience has been an important factor in getting Canadian fire management organizations to accept and use fire danger rating. Nevertheless, there was a lag time of about 5–10 years, from time of introduction of research products to full implementation in operational regulations, guidelines, procedures, and training courses.

National fire danger rating systems must continue to evolve. Improvements occur as research and operational experience continue, as fire management organizations become more sophisticated, as problems and opportunities change, and as technology advances. The current CFFDRS, for example, represents the fifth generation of fire danger rating methods developed in Canada by the CFS (Stocks et al. 1989).

Fire danger indexes, such as the FWI System components, are usually calculated from one weather observation per day to represent a large administrative area that may include many different fuel types and variable topography. This is difficult to do well. Two of the most significant general assumptions concerning any fire danger rating system are that: (1) the observations made at a particular fire weather station, and the simulated fuel moistures calculated from these observation, are representative of the area being assessed (i.e. they ignore spatial variations in fire danger); and (2) the time of day on which the index is based is valid (i.e. it ignores temporal variations in fire danger due to diurnal wetting, drying, and wind speed).

Spatial fire management systems have been developed to interpolate between weather stations and to integrate spatial variability in fuel, weather, and terrain data. These computer systems are effective tools for interpolating and portraying fire danger over large spatial scales, and they allow for powerful spatial analysis. Nevertheless, the accuracy of these systems may be limited by data inputs such as local variability in rainfall amounts and wind speed in complex terrain. High-resolution numerical weather models have promise for forecasting winds in complex terrain. Stull et al. (2004) ran four models in nested grids of 104, 36, 12, and 4 km for southern British Columbia and 2 km for the Vancouver area on a daily basis. However, high-resolution forecast models have not as yet been explicitly included in spatial fire management systems in Canada.

Useful information can also be obtained in the field using simple weather instruments and look-up tables. Use of electronic and computer technology requires ongoing training and maintenance, which may not be available or appropriate in all situations. It is often possible and cost effective to ‘sense’ weather conditions between fire weather stations or for particular locations by using remote sensing (including precipitation radar), installing additional rain gauge sites, and using inexpensive fire weather instruments for measuring temperature, relative humidity, and rain, and the Beaufort wind scale for estimating wind velocity.

Fire danger information will not be used if it is only disseminated from the top down – if fire managers in remote, rural locations perceive it to be coming from a central or regional office some distance away, it may be questioned. End users must feel ownership and have input into determining fire danger at the local level – by collecting weather data or by determining weather station location, for example. Effective means are also needed to communicate fire danger to the public and other interest groups.

As a form of media, fire danger rating systems should be appropriate not only for the fire environment but also for the human environment – the organizational and institutional structures, culture, economic circumstances, and technological capability – that influence the capability and decision making of the implementing fire management agencies. The importance of developing suitable technology transfer plans cannot be overemphasized (Kiil et al. 1986; de Groot 1989; Fogarty et al. 1998).

Fire danger rating systems will be useful only where fire managers have authority to make pro-active decisions about resource allocation based on fire danger and values at risk. The transfer of North American fire danger systems will be most successful in organizations with organizational structures similar to those found in North American fire management agencies: a centralized command structure with well-trained professional firefighters empowered to make operational decisions (e.g. an Incident Command System), and a systematic allocation of resources depending on fire danger and values at risk. A fire danger rating system is of little use where decisions are reactive or where operational resources are allocated at a political level.

A national system should be sufficiently flexible that it can be implemented at varying degrees of intensity and complexity in different regions, depending on local importance of the fire problem and capacity of fire management agencies. Various interpretive media, from posters (e.g. Alexander 1995; Cole and Alexander 1995; Stocks and Hartley 1995) and look-up tables (Canadian Forestry Service 1984; Taylor et al. 1997) to computer displays (Lee et al. 2002), are useful in providing fire danger information depending on the background, literacy, numeracy, and culture of fire management staff. The CFFDRS supports decision making by providing real-time fire danger information that is useful at multiple operational levels. System developers often encounter contradictory responses to their attempts to satisfy user demands: models and systems aren’t accurate enough, v. models and systems are too complicated (Rothermel 1987). Presumably, simple but reliable decision aids are needed at the field level.

In North America, fire management organizations have adapted their operations to changing technology. When new technologies are introduced, it takes time for organizational practices to change and implement the technology effectively. Implementation issues surrounding new fire danger applications range from rejecting what is not ‘home grown’ to too-ready acceptance of ‘black box’ predictions. While developing some fire management applications of fire danger rating systems or indices can be easy, in other cases applications come about as a result of long operational experience (e.g. Melton 1989, 1996).

No fire danger rating decision aid can replace an individual’s local knowledge of the fire environment and influence on fire behavior, fire-suppression and prescribed-burning experience, or skill and common sense. A fire danger rating system is simply a decision aid that is available to fire managers who choose to use it. It is not a solution for all fire management problems and decisions and should not amputate good judgment. In this regard, Nelson (1955) made a very apropos statement:

I do not want to leave the impression that I think a good system of [fire] danger measurement is the answer to all fire control and management problems. It can be a guide, and a very useful one, but it can never take the place of cool, calculating, and experienced judgement.

Nelson’s observation is confirmed by recent research that reveals the importance of experience and intuition in decision making in emergency situations (Klein 1998).

To fully realize a national system, formal institutional mechanisms are often needed to coordinate implementation among agencies and between regions. Cooperative mechanisms also facilitate communication between researchers and operations staff (Alexander 2003). Researchers must understand the skills, capabilities, and needs of fire managers, and managers and operations staff must have realistic expectations for research products and time frames. Formal agreements and working groups composed of federal and provincial research and operational staff play important roles in setting policy and direction, as well as in setting standards and sharing information and resources. Perhaps most importantly, common vision and a sense of common cause between fire scientists and fire managers is needed to successfully realize a national fire danger rating system.

Technological innovation in communications and computing has greatly changed the volume and speed by which fire danger information is gathered and disseminated in Canada, as elsewhere, and has allowed for the incorporation of increasingly sophisticated scientific models in fire management systems. However, as the volume and complexity of available fire danger information increases, a better understanding of the influence of cognitive and other human factors on the use of such information in decision making is needed (Weick and Sutcliffe 2001). Indeed, fire danger rating systems exist within and to serve the larger fire management environment. Thus it is important to create or support the human and institutional infrastructure needed to make a fire danger system work in a particular organizational and cultural environment, as well as the science and technology infrastructure appropriate for the fire environment (Fig. 2a).

Many thanks to Brad Hawkes, Bill de Groot, Brad Stennes, Brian Stocks, Peter Engelfield, and three anonymous reviewers for reviewing an earlier draft of this paper and to Monique Keiran for editorial help. Thanks also go to María del Carmen Dentoni and staff of the Plan Nacional del Manejo del Fuego of Argentina for their hospitality and for conversations about the CFFDRS, which provided impetus for this work.