Report on Early Warning for Fire and Other Environmental Hazards: II. HAZARD ASSESSMENT AS THE BASIS OF RISK ANALYSIS
Report on Early Warning for
Fire and Other Environmental Hazards
II. HAZARD ASSESSMENT AS THE BASIS OF RISK ANALYSIS
Fire Danger Rating (Fire Risk Assessment)
Use of Satellite Data to Help Assess Fire Potential
Active Fire Detection by Satellite Sensors
Climate-change and Fire Risk Modelling
Towards A Global Wildland Fire Information System
Early warning systems for fire and smoke management for local, regional, and global application require early warning information at various levels. Information on current weather and vegetation dryness conditions provides the starting point of any predictive assessment. From this information the probability of risk of wildfire starts and prediction of the possibility of current fire behaviour and fire impacts can be derived. Short- to long-range fire weather forecasts allow the assessment of fire risk and severity within the forecasting period. Advanced space borne remote sensing technologies allow fire weather forecasts and vegetation dryness assessment covering large areas (local to global), at economic levels and with accuracy which otherwise cannot be met by ground-based collection and dissemination of information. Remote sensing provides also capabilities for detecting new wildfire starts, monitoring ongoing active wildfires, and, in conjunction with fire-weather forecasts, providing an early warning tool for escalating, extreme wildfire events.
Fire Danger Rating (Fire Risk Assessment)
Fire danger rating systems have been devised by fire authorities to provide early warning of conditions conducive to the onset and development of extreme wildfire events. The factors that predispose a particular location to extreme wildfire threat change over time scales that are measured in decades, years, months, days and hours. The concept of fire danger involves both tangible and intangible factors, physical processes and hazard events. By definition:
“Fire danger” is a general term used to express an assessment of both constant and variable fire danger factors affecting the inception, spread, intensity and difficulty of control of fires and the impact they cause (e.g., Chandler et al., 1983).
The constant factors in this definition are those which do not change rapidly with time but vary with location e.g. slope, fuel, resource values, etc. The variable factors are those which change rapidly with time and can influence extensive areas at one time and these are primarily the weather variables which affect fire behaviour. All the potentials referred to in the definition must be present. If there is absolutely no chance of ignition – there is no fire danger. If fuels are absent or cannot burn – there is no fire danger. If fires can start and spread but there are no values at risk as may be perceived for remote areas managed for ecological diversity, there is no fire danger for values at risk.
Fire danger rating systems produce qualitative and/or numerical indices of fire potential that can be used for guides in a variety of fire management activities including early warning of fire threat. Different systems of widely varying complexity have been developed throughout the world which reflect both the severity of the fire climate and the needs of fire management. The simplest systems use only temperature and relative humidity to provide an index of the potential for fire starts (e.g. see the Angstrom index, cf. Chandler et al., 1983). Fire danger rating systems of intermediate complexity combine measures of drought and weather as applied to a standard fuel type to predict the speed of a fire or its difficulty of suppression (e.g., McArthur 1966, 1967; Sneeuwjagt and Peet, 1985). The most complex systems have been developed in Canada (Forestry Canada Fire Danger Group, 1992) and the United States (Deeming et al., 1978) which combine measures of fuel, topography, weather and risk of ignition (both lightning and human-caused) to provide indices of fire occurrence or fire behaviour which can be used either separately or combined to produce a single index of fire load.
While a single fire danger index may be useful to provide early warning of wildfire activity over broad areas it is impossible to communicate a complete picture of the daily fire danger with a single index. Therefore, it is necessary to break fire danger rating into its major components to appreciate where early warning systems for single factors fall into the overall picture of fire danger rating. These fall into three broad categories of changes in fuel load; changes in fuel availability or combustion and changes in weather variables that influence fire spread and intensity.
Early warning of fire precursors
Changes in fuel load
In all fire danger rating systems fuel load is assumed to be constant although specific fuel characteristics may be formulated for specific forest or other vegetation types as in the Canadian fire danger rating system or for specific fuel models i.e. combinations of vegetation and fuel with similar characteristics as in the U.S. National Fire Danger Rating System. These fuel models may overlook major shifts in total fuel loads which may be changing over periods of decades or even centuries. Fuel changes start immediately after the cessation of cultural or agricultural burning. This change usually runs in parallel with increased suppression efficiency whereby small fires under moderate fire danger conditions are suppressed early in their life. In this scenario fire authorities and the general public may be lulled into a false sense of security because the potential for high-intensity forest fires is not manifest except under rare events of extreme weather. In places this may be complicated by the introduction of exotic forest species (e.g. the establishment of eucalypt forests on formerly oak woodland savannahs in central California) and a shift of the population from living in relatively low-fuel areas which were maintained either by frequent burning through cultural or agricultural practices, or though frequent low-intensity wildfires.
Thus, the first element of early warning for a potential fire risk is a major shift in the total forest fuel complex towards denser forests with a large build up of surface debris and a change in vulnerability of the population by living more intimately with these fuels. Over the last 20 years this change has occurred in the urban/forest intermix associated with most of the centres of population located in forest regions of many of the more developed countries.
The seasonal change in fuel availability as fuels dry out during the onset of the fire danger period sets the stage for severe wildfires. Under drought conditions more of the total fuel complex is available for combustion. Deep litter beds and even organic soils may dry out and become combustible. Large fuels such as downed logs and branches may burn completely. Drought stress on living vegetation not only reduces the moisture content of the green foliage but also dried plant matter such as leaves and bark can be shed adding to the total load of the surface fuel. Under extreme drought conditions normally moist areas such as swamps and creek lines dry out and are no longer a barrier to the spread of fires as might be expected in a normal fire season. Long-term moisture deficiency in itself cannot be used to forecast critical fire situations because if the smaller fine fuels are wet or green, serious fires will not occur at any time of the year. However, most devastating fires occur when severe fire weather variables are combined with extreme drought.
There are a number of bookkeeping methods of monitoring the seasonal development of drought. The Keetch-Byram (1968) Drought Index is a number representing the net effect of evapotranspiration and precipitation in producing a cumulative measure of moisture deficiency in the deep duff and soil layers. It is a continuous index which can be related to the changes in fuel availability mentioned above and the occurrence of severe fires. The Index has proved to be a useful early warning tool and is now incorporated into the US National Fire Danger Rating System (Pyne et al., 1996) and the Australian Forest Fire Danger Rating System (McArthur, 1967).
There are a number of similar drought indices used elsewhere in the world. For example, the drought code component of the Canadian Fire Weather Index System (Forestry Canada Fire Danger Group, 1992), the Australian Mount Soil Dryness Index (Mount, 1972) and the Drought Index used in France (Orieux, 1974, cited from Chandler et al., 1983).
Although drought indices can be built into a broader fire danger rating system they are most effective as an early warning system when they are maintained separately and charted to illustrate the progressive moisture deficit for a specific location. This allows the fire manager to compare the current season with historical records of past seasons. The fire manager can also make associations between level of drought index and levels of fire activity which are specific to the region. This overcomes the problems caused by variation of both forest and soil type which can mask the recognition of severe drought when a drought index is applied across broad areas.
Regular charting of bookkeeping-type systems such as the Keetch-Byram Drought Index or the Mount Soil Dryness Index are particularly useful in monitoring the effects of below-average rainfall during the normal wet or winter season. Moisture deficits from the previous dry season may be carried over winter. As the next fire season develops, high levels of drought may occur early in the season when, under the normal seasonal pattern, large and intense fires rarely occur. In some parts of the world there are indices which indicate the changes in the global circulation patterns which may provide warning as much as 6 to 9 months in advance of extremely dry conditions. One of these is the Southern Oscillation Index which records the difference in atmospheric pressure between Darwin in the north of the country and Melbourne in southern Australia which can be related to the El Niño events in the southern Pacific Ocean. When the Southern Oscillation Index is strongly positive wetter than normal conditions are expected in south-eastern Australia; when the index is strongly negative drought conditions are forecast for the south-east of Australia.
Early warning of fire behaviour
The fire spread component of fire danger rating systems is designed to combine the weather elements affecting fire behaviour and provide a prediction of how fires will change hourly during the day. Most indices use 24 hour precipitation, and daily extremes or hourly measurements of temperature, relative humidity, and wind speed to predict the rate of spread of forest fires. In some systems, notably the U.S. National Fire Danger Rating System and the Canadian Fire Weather Index System, indices of fire spread are combined with a long-term measure of drought to provide an index of the total severity of the fire. This is termed a Burning Index in the United States system or a Fire Weather Index in the Canadian system.
In some systems the risk of ignition from either lightning activity or human activities is calculated to form an index of fire occurrence which can be combined with a Burning Index to give an overall Fire Load Index (e.g. Deeming et al., 1978). These are rarely used in the U.S.A. today (Pyne et al., 1996). The risk of ignition by lightning is calculated separately and areas with historical records of high human-caused ignitions are mapped as a constant fire danger variable and are used in concert with a burning index to calculate fire threat in a wildfire threat analysis system.
Fire spread indices are essentially weather processors (Andrews, 1991) and the data required to provide early warning of severe fire conditions, depends primarily on the ability to provide adequate space and time forecasts of the weather. The synoptic systems which are likely to produce severe fire weather are generally well known but the ability to predict their onset depends largely on the regularity of movement and formation of atmospheric pressure systems. In Australia the genesis of severe fire weather synoptic systems has, at times, been recognised up to three days in advance; more often less than 24 hours warning is available before the severity of fire weather variables can be determined. Extended and long range forecasts contain greater uncertainty, and there is less confidence in fire severity forecasts at these time scales. Even so, these forecasts are useful in fire management in that the forecasts can be used to develop contingency plans, that is, developing options, but not implementing them until the forecasts are more certain.
As improved fire behaviour models for specific fuel types are developed there is an increasing need to separate the functions of fire danger and fire spread (Cheney, 1991). A regional fire weather index based on either fire spread or suppression difficulty in a standard fuel type and uniform topography is required to provide public warnings, setting fire restrictions, and establishing levels of readiness for fire suppression. At a local level, fire spread models which predict the development and spread of a fire across the landscape through different topography and through a number of fuel types are required for suppression planning and tactical operations. However, these systems can be confusing on a broader scale by providing too much detail. They may be influenced by atypical variations of critical factors at the measuring site and may lose the broad-scale appreciation of regional fire danger that is required for early warning purposes.
Use of Satellite Data to Help Assess Fire Potential
The amount of living vegetation, and its moisture content, has a strong effect on the propagation and severity of wildland fires. The direct observation of vegetation greenness is therefore essential for any early warning system. Current assessment of living vegetation moisture relies on various methods of manual sampling. While these measurements are quite accurate, they are difficult to obtain over broad areas, so they fail to portray changes in the pattern of vegetation greenness and moisture across the landscape.
The current polar orbiting meteorological satellites provide the potential for delivering greenness information and other parameters needed for fire management and fire impact assessment at daily global coverage at coarse spatial resolution (cf. Following section on Active Fire Detection by Satellite Sensors; see also Kendall et al., 1997). This is achieved using wide angle scanning radiometers with large instantaneous fields of view, e.g. the NOAA Advanced Very High Resolution Radiometer (AVHRR) instrument which measures reflected and emitted radiation in multiple channels including visible, near-infrared, middle-infrared, and thermal (Kidwell, 1991). Because of its availability, spatial resolution, spectral characteristics, and low cost, NOAA AVHRR has become the most widely used satellite data set for regional fire detection and monitoring. Currently, AVHRR data are used for vegetation analyses and in the detection and characterization of active flaming fires, smoke plumes, and burn scars.
Since 1989 the utility of using the Normalized Difference Vegetation Index (NDVI) to monitor seasonal changes in the quantity and moisture of living vegetation has been investigated (Tucker, 1977, 1980; Tucker and Sellers, 1986; Holben, 1986; Tucker and Choudhury, 1987; Goward et al., 1990). Daily AVHRR data are composited into weekly images to remove most of the cloud and other deleterious effects, and an NDVI image of the continental U.S. is computed by the U.S. Geological Survey’s Earth Resources Observation Systems Data Center (EDC). These weekly images are obtained via the Internet and further processed into images that relate to fire potential (Burgan and Hartford, 1993; Burgan et al., 1996) and that are more easily interpreted by fire managers.
Vegetation greenness information: An early warning indicator
Four separate images are derived from the NDVI data — Visual Greenness, Relative Greenness, Departure from Average Greenness, and Live Shrub Moisture.
Visual greenness is simply NDVI rescaled to values ranging from 0 to 100, with low numbers indicating little green vegetation. Relative greenness maps portray how green each 1 km square pixel is in relation to the historical range of NDVI observations for that pixel. The Departure from Average Greenness maps portray how green the vegetation is compared to the average NDVI value determined from historical data for the same week of the year. Use of this map, along with the Visual and Relative Greenness maps, can give fire managers a good indication of relative differences in vegetation condition across the nation and how that might affect fire potential.
Live Shrub Moisture: The National Fire Danger Rating System (NFDR) used by the United States requires live shrub and herbaceous vegetation moisture as inputs to the mathematical fire model (Burgan and Hartford, 1996). For this reason, and to help fire managers estimate live shrub moistures across the landscape, Relative Greenness is used in an algorithm to produce live shrub moistures ranging from 50 to 250 percent.
These maps may be viewed at http://www.fs.fed.us/land/wfas/welcome.html.
Development of fire hazard maps
Improvement in the spatial definition of fire potential requires use of a fire danger fuel model map to portray the spatial distribution of fuel types. In the U.S.A. the Geological Surveys Earth Resources Observation Systems Data Center (EDC) used a series of eight monthly composites of NDVI data for 1990 to produce a 159 class vegetation map of the continental U.S. at 1 km resolution (Loveland et al., 1991). Data from 2560 fuel observation plots randomly scattered across the U.S. permitted the development of a 1 km resolution fuel model map from the original vegetation map. This fuel model map is now being used in two systems to provide broadscale fire danger maps.
Integration of satellite data into fire danger estimates
The state of Oklahoma in the United States provides a good example for early warning of wildfires. The state operates an automated weather station network that consists of 111 remote stations at an average spacing of 30 km. Observations are relayed to a central computer every 15 minutes. Cooperative work between the Intermountain Fire Sciences Laboratory (U.S. Forest Service) and the Oklahoma State University resulted in development of a fire danger rating system that produces map outputs (Carlson et al., 1996). The satellite-derived NFDR fuel model map is used to define the fuel model for each 1 km pixel, and the weekly Relative Greenness maps are used to calculate live fuel moisture input for the fire danger calculations. This results in a fire danger map showing a smooth transition of fire danger across the state. These maps may be viewed at http://radar.metr.ou.edu/agwx/fire/data.html.
A goal of fire researchers in the U.S. is to expand the techniques provided for Oklahoma to other states and nations. An alternative method of estimating fire potential has been developed (Burgan and others, in prep.) using just the 1 km resolution fire danger fuel model map, relative greenness, and interpolated moisture for dead fuels about 1.25 cm in diameter. This map was found to be highly correlated with fire occurrences for California and Nevada for the years 1990 to 1995 (Klaver et al., 1997). It is now being, or will be, further tested by Spain, Chile, Argentina, and Mexico as part of an effort between the Intermountain Fire Sciences Laboratory and the EDC, sponsored by the Pan American Institute for Geography and History. The Fire Potential Map is updated daily and can be seen at http://www.fs.fed.us/land/wfas/welcome.html under “experimental products”.
While these examples, and many other published papers (Chuvieco, 1995), indicate the usefulness of current satellite data for fire management purposes, it is obvious that satellite data will become ever more useful and accurate. Instruments that will be flown on the “Mission to Planet Earth” hold great promise for several fire management requirements, such as fire detection, fuel mapping, monitoring seasonal greening and curing.
Fire Weather Forecasts
Improved fire weather forecasts are needed at a variety of time and space scales. At large space and time scales, accurate fire weather forecasts have potential for long range planning to allocate scarce resources. At smaller time and space scales, accurate fire weather forecasts have potential use in alerting, staging and planning the deployment of fire suppression crews and equipment. At the smallest time and space scales, accurate fire weather forecasts can be helpful in fighting fires as well as determining optimal periods for setting prescribed silvicultural fires (Fosberg and Fujioka, 1987; Roads et al., 1991, 1997).
Current U.S. fire weather forecasts are prepared from short-range weather forecasts (1-2 days) by the Eta model of the National Center for Environmental Prediction (NCEP), other model output statistics, and human judgment. These fire weather forecasts include information about precipitation, wind, humidity, and temperature.
To test whether even longer range forecasts focused on fire weather products would be useful, an experimental modelling system, developed at the U.S. National Center for Environmental Prediction (NCEP) for making short-range global to regional weather forecasts, is currently being developed at the Scripps Experimental Climate Prediction Center (ECPC). Although this system is currently focused on making and disseminating experimental global to regional fire weather forecasts focused for Southern California, it could be easily transported and applied anywhere else in the world.
Global to regional fire-weather forecasts
At the largest space and time scales, a modelling system utilizes NCEP’s MRF or GSM (global spectral model; see Kalnay et al. 1996). A high resolution regional spectral model (RSM; see Juang and Kanamitsu, 1994) is nested within the global model by first integrating the GSM which provides initial and low spatial resolution model parameters as well as lateral boundary conditions for the RSM. The RSM then predicts regional variations influenced more by the higher resolution orography and other land distributions within a limited but high resolution domain (Kalnay et al., 1996).
Global to regional forecasts of the fire weather index and precipitation are currently displayed on the world-wide web site of the ECPC at http://meteoral.ucsd.edu/ecpc/special/globaltoregional/.
Due to bandwidth limitations of the Internet, only the complete initial and 72-hour forecasts for the global model are transferred four times daily (at 0000, 0600, 1200, 1800 hrs. UCT). From these global initial and boundary conditions, regional forecasts at 25 km resolution are then made and also displayed.
New features are under development. Besides beginning development of longer-range monthly global to regional forecasts, the current fire weather forecasting methodology will be validated. Experimental global to regional forecasts for other regions are also under development. Provision of additional output of corresponding land surface variables such as snow, soil and vegetation moisture are now being extracted and may soon be provided as part of the forecasts. These additional variables are needed to transform fire weather indices into fire danger indices, which include vegetation stresses.
Active Fire Detection by Satellite Sensors
The middle-infrared and thermal AVHRR bands of the NOAA polar-orbiting satellites have been used for identifying fires. Several techniques are currently used to detect active fires at regional scales using multi-spectral satellite data. A comprehensive validation of AVHRR active fire detection techniques through a range of atmospheric and surface conditions has not yet been performed. A number of studies, however, have provided some level of validation.
Limitations in AVHRR fire detection
Even in full configuration, with two NOAA satellites in operation, the AVHRR data provides only a limited sampling of the diurnal cycle. The orbital characteristics of the satellites result in two daytime and two nighttime orbits per location. The afternoon overpass provides the best coverage in terms of fire detection and monitoring in tropical and subtropical regions (Justice and Dowty, 1994). In addition, the afternoon overpass enables detection of the full range of parameters described (i.e. vegetation state, active fires, burn scars, smoke).
Perhaps the most fundamental problem to AVHRR fire detection is that analysis is limited to relatively cloud-free areas. This can be a serious issue in tropical and sub-tropical regions. Cloud cover can cause an underestimation in the extent and frequency of burning, and limits the ability to track vegetation parameters. This issue is not limited to the NOAA satellite system. Dense clouds will prevent detection of the surface by all visible and infrared sensors. A satisfactory methodology for estimating the amount of burning missed through cloud obscuration has yet to be developed.
Due to characteristics of the NOAA meteorological satellites described it is possible to collect near real-time information to support fire management activities.
Automatic fire alerts
A prototype software has been developed in Finland for automatic detection of forest fires using NOAA AVHRR data. Image data are received by the Finnish Meteorological Institute. From each received NOAA AVHRR scene a sub-scene covering as much as possible of the monitoring area is extracted (approximately 1150 square km).
The processing includes: detection and marking of image lines affected by reception errors, image rectification, detection of “hot spots”, elimination of false alarms, and generation of alert messages by e-mail and telefax.
A fully automatic system has been developed to detect forest fires using data from NOAA AVHRR. The prototype system has been developed in Finland and tested in four experiments in 1994-1997 in Finland and its neighbouring countries Estonia, Latvia, Russian Carelia, Sweden and Norway. For each detected fire, a telefax including data on the location of the fire, the observation time and a map showing the location, is sent directly to the local fire authorities. Nearly all detected fires were forest fires or prescribed burnings.
The screening of false alarms is an essential technique in fire detection if the results are to be used in fire control. Effective screening enables fully automatic detection of forest fires, especially if known sources of error like steel factories are eliminated. In the experiments in 1994-96, most of the detected fires that were in areas where verification was possible, were real fires. This shows that space borne detection of forest fires has potential for fire control purposes.
Atmospheric Pollution Warning
The drought and fire episodes in Southeast Asia between 1992 and 1994 and again now in 1997 (September-October) resulted in severe atmospheric pollution. The regional smog events of 1991 and 1994 triggered a series of regional measures towards cooperation in fire and smoke management. In 1992 and 1995 regional workshops on “Transboundary Haze Pollution” were held in Balikpapan (Indonesia) and Kuala Lumpur (Malaysia). This was followed by the establishment of a “Haze Technical Task Force” during the Sixth Meeting of the ASEAN Senior Officials on the Environment (ASOEN) (September 1995). The task force is chaired by Indonesia and comprises senior officials from Brunei Darussalam, Indonesia, Malaysia, and Singapore. The objectives of the work of the task force is to operationalize and implement the measures recommended in the ASEAN Cooperation Plan on Transboundary Pollution relating to atmospheric pollution, including particularly the problem of fire and smoke (ASEAN, 1995a,b; Goldammer, 1997a,b).
First regional cooperation plans include the use of satellite data to predict smoke pollution from wildfires based on detection of active fires and smoke plumes and the forecast of air mass trajectories. In addition, some Southeast Asian countries have developed an air quality index for early warning of smoke-generated health and visibility problems.
In Singapore air quality is monitored by 15 permanent stations and reported using the Pollutant Standard Index (PSI), a set of criteria devised by the U.S. Environmental Protection Agency (EPA). The PSI value of 100 equals legal air quality standard (or limit) and is based on risk to human health (primary standard) or non-human health (animals, plants; secondary standard). Under this system, the levels of key pollutants like sulphur dioxide (SO2), nitrogen dioxide (NO2), carbon monoxide (CO), ozone (O3) and respirable suspended particles (PM10) are used to come up with a single index, the PSI. The PSI is a health-related index, averaged over a 24-hour period, on a scale of 0-500.
Another potentially useful tool for analysing fire-generated smoke sources, as detected or monitored by space borne sensors, is the rose-diagram technique (Brivio et al., 1997). In conjunction with trajectory analysis this spatial analysis technique allows to establish the relationships between smoke pollution and the potential sources, e.g. wildfires vs. industrial pollution.
Climate-change and Fire Risk Modelling
The Intergovernmental Panel on Climate Change (IPCC) has recently concluded that “the observed increase in global mean temperature over the last century (0.3-0.6° C) is unlikely to be entirely due to natural causes, and that a pattern of climate response to human activities is identifiable in the climatological record” (IPCC, 1995). In Canada and Russia, for instance, this pattern of observed changes has taken the form of major winter and spring warming in west-central and northwestern Canada and virtually all of Siberia over the past three decades, resulting in temperature increases of 2-3° C over this period (Environment Canada, 1995).
Numerous General Circulation Models (GCMs) project a global mean temperature increase of 0.8-3.5° C by 2100 AD, a change much more rapid than any experienced in the past 10,000 years. Most significant temperature changes are projected at higher latitudes and over land. While GCM projections vary, in general summer temperatures are expected to rise 4-6° C over much of Canada and Russia with a doubling of atmospheric carbon dioxide. In addition, changes in the regional and temporal patterns and intensity of precipitation are expected, increasing the tendency for extreme droughts associated with an increase of fire risk and severity.
In the lower latitudes and in coastal regions the expected changes in temperatures, precipitation and dry season length will be less pronounced than in higher latitudes and in continental regions. However, the manifold interactions between changing climate and human-caused disturbances of ecosystems may result in change of fire regimes in the densely populated regions of the tropics and subtropics (Goldammer and Price, 1997; see para.4.6.3).
Modelling climate change and forest fire potential in boreal forests
Despite their coarse spatial and temporal resolution, GCMs provide the best means currently available to project future climate and forest fire danger on a broad scale. However, Regional Climate Models (RCMs) currently under development (e.g. Caya et al., 1995), with much higher resolution, will permit more accurate regional-scale climate projections. In recent years GCM outputs have been used to estimate the magnitude of future fire problems. Flannigan and Van Wagner (1991) used results from three early GCMs to compare seasonal fire weather severity under a 2xCO2 climate with historical climate records, and determined that fire danger would increase by nearly 50% across Canada with climate warming. Wotton and Flannigan (1993) used the Canadian GCM to predict that fire season length across Canada would increase by 30 days in a 2xCO2 climate. An increase in lightning frequency across the northern hemisphere is also expected under a doubled CO2 scenario (Fosberg et al., 1990; Price and Rind, 1994). In a recent study (Fosberg et al., 1996) used the Canadian GCM, along with recent weather data, to evaluate the relative occurrence of extreme fire danger across Canada and Russia, and showed a significant increase in the geographical expanse of the worst fire danger conditions in both countries under a warming climate.
In a recent study (Stocks et al., 1997), Canadian and Russian fire weather data from the 1980’s were used, the warmest decade on record in Canada (Gullet and Skinner, 1992), in conjunction with outputs from four recent GCMs, to compare the spatial distribution of current seasonal levels of fire weather severity across both countries with those projected under a 2xCO2 climate.
Daily May – August weather data was collected for the 1980’s for 224 Russian and 191 Canadian climate stations. Local noon measurements of temperature, relative humidity, windspeed and precipitation were used to calculate the component codes and indices of the Canadian Fire Weather (FWI) System (Van Wagner, 1987) for each station. Daily FWI values were then converted to Daily Severity Rating (DSR) values using a technique developed by Williams (1959) and modified by Van Wagner (1970). This severity rating technique permits the integration of fire severity over periods of various lengths, from daily (DSR) through monthly (MSR) to seasonal (SSR) values. In this analysis both MSR and SSR values are used. The FWI System provides an assessment of relative fire potential based solely on weather observations, and does not take forest type into consideration.
The following shows an example of possible conclusions: The monthly progression of modelled MSR under a 2xCO2 climate indicates an earlier start to the fire season, with significant increases in the geographical extent of extreme fire danger in May. The month of June shows the most significant increase, however, with virtually all of Siberia and western Canada under extreme fire danger conditions during that period. A more modest increase is observed in July and August. The seasonal pattern changes indicate an earlier annual start of high to extreme fire severity, and a later end to the fire season across Canada and Russia as a whole, although there are important regional variances from this pattern.
Changes in the area in each fire danger class are perhaps more important than absolute value changes in MSR. Dramatic changes in the areal extent of high to extreme fire danger in both countries under a doubled CO2 climate were observed. In general, there is a decrease in moderate MSR and SSR levels, and a significant increase in the area experiencing high to extreme MSR and SSR levels under a warmer climate. This is particularly true in June and July, but increases in the area under extreme fire danger (and therefore greatest fire potential) are common to all months. Significantly, two to three-fold increases are projected for Russia during the June-July period.
Although hampered somewhat by coarse spatial and temporal resolution, the four GCMs utilized in this study show similar increases in fire danger levels across much of west-central Canada and Siberia under a warmer climate. While shifts in forest types associated with climate change were not considered in this analysis, these increases in fire danger alone will almost certainly translate into increased fire activity, and, as fire management agencies currently operate with little or no margin for error, into large increases in area burned. The result will be more frequent and severe fires, shorter fire return intervals, a skewing of forest age class distribution towards younger stands, and a resultant decrease in the carbon storage of northern forests (cf. Kurz et al., 1995).
A warmer climate, in combination with severe economic constraints and infrastructure downsizing, will decrease the effectiveness, and thus the area protected, by fire management agencies. This then means that a new reality in forest fire impacts is on the horizon. There is a strong need to continue modelling future climates, using higher-resolution models as they become available, so that future development of long-range early warning systems and fire management planning can be accomplished in the most informed manner possible.
Assessing impacts of climate change and human population growth on forest fire potential in the tropics
With growing population pressure and accelerating change of land use in tropical vegetation — i.e., conversion of tropical forested ecosystems into farming and pastoral ecosystems — fire is being used increasingly. While certain tropical dry forests and savannas have been adapted to anthropogenic fire use for millennia and show typical features of sustainable fire ecosystems, the opening and fragmentation of tropical evergreen forests has increased the risk of wildfires that will have destructive impacts on biodiversity and sustainability of these forest ecosystems.
An assessment of potential impacts of climate change on fire regimes in the tropics based on GCMs and a GCM-derived lightning model (Goldammer and Price, 1997) recently concluded that there is a high degree of certainty that land use and climate features under conditions of a 2xCO2 atmosphere will influence tropical fire regimes.
In this respect, tropical closed evergreen forests will become increasingly subjected to high wildfire risk because of land-use changes (opening and fragmentation of closed forest by logging and conversion), increasing fire sources (use of fire as land clearing tool), and climate change (prolongation of dry seasons, increasing occurrence of extreme droughts, increase of lightning as fire source). Tropical dry forests and savannas in regions with predicted reduction of average total annual precipitation and average prolongation of dry seasons will be subjected to higher fire risk. However, the reduction of net primary production (NPP) and the increasing impacts of farming and grazing systems will lead to formation of open and sparse vegetation cover with restricted capability to support the spread of fires (discontinuity of fuelbed).
Tropical dry forests and savannas in regions with a predicted increase of average total annual precipitation and average reduction of dry season length will be subjected to higher fire risk due to the fact that increased NPP will lead to the build-up of more continuous fuelbeds that may carry more frequent and larger-sized wildfires.
Long-range forecasting of fire potential: conclusions
The models and assumptions described in this section clearly exceed the time horizon of early warning systems. However, the Working Group strongly suggests that relevant follow-up processes, in conjunction with other international activities, programmes and agreements, will consider this extended time horizon. The disaster management community needs to be prepared for managing situations which, in the near future, may require the development of innovative technologies and the preparedness of administrations to accomplish tasks that may differ from today’s situation. While warning of potential disaster implies a high level of confidence, a second level, or alert level, with lower level of confidence is useful from the standpoint of strategic or contingency planning. This alert level is intended to convey the message that the potential for disaster has increased, but that actions would be limited to planning.
Towards A Global Wildland Fire Information System
A demonstration concept
One demonstration project is the Canadian Wildland Fire Information System (CWFIS), developed by the Canadian Forest Service. The CWFIS is a hazard-specific national system envisioned as a prototype system that is adaptable to other countries. Establishing and linking a number of compatible national systems could provide the nucleus of a global fire information network. Following the conceptual design of CWFIS, future early warning systems would have three goals:
- Facilitate information sharing among all agencies through a national network.
- Facilitate inter-agency sharing of resources by providing national fire information.
- Facilitate the application of fire research results through an interoperable platform.
The CWFIS incorporates several functions: weather observations, weather forecasts, fire danger, fire behaviour, fire activity, resource status, situation reports, decision support systems, technology transfer, and information exchange.
The system automatically downloads weather observations from a national satellite network. Although Canadian weather data are not mapped, exported systems (ASEAN, Florida) provide this capability. Data needed for daily fire-danger calculations are extracted from a larger set of hourly weather observations. Most countries operate national weather observing networks. The World Meteorological Organization maintains a global network of synoptic weather stations which is accessible through satellite downlinks. Nationally, research is underway to produce automated spot fire-weather forecasts using a Regional Atmospheric Modelling System (RAMS). When operational, users will be able to submit coordinates for a specific fire and obtain computer- generated hourly forecasts for that location.
At global and national scales, forecasts are important because large-scale mobilization requires one or more days to accomplish. The CWFIS accesses 3 days of numeric forecast data generated by the Canadian Meteorological Centre (CMC). In Florida, a Regional Atmospheric Modelling System (RAMS) is used to forecast weather on a finer scale than that available nationally. Many countries operate similar national weather forecasting systems. Alternatively, the CMC (or other major national agencies) can generate a numeric weather forecast for any region on earth (see also Fire Weather Forecasts section, above).
Weather data are transformed into components of the Canadian Forest Fire Danger-Rating System. Station data are converted to national contour maps with an ARC/INFO GIS processor. The maps are converted to GIF images and stored on a World-Wide Web server. Daily maps overwrite those from previous days and date indices are automatically updated. Fire-danger maps are retained for seven days to provide backup.
Digital fuel and topographic databases enable calculating absolute fire behaviour potential such as rate of spread, head-fire intensity, fuel consumption, and fire type. The CWFIS uses a 16-class satellite-derived land-cover classification to approximate a national fuel map which is not directly available. The fire-behaviour maps are in a cell format, reflecting the underlying fuel database. Satellite-based land cover classifications should be derivable for most countries. The system provides seven days of history, current observations, and three days of forecasts.
Fundamental to any fire information system is compiling and disseminating fire statistics such as number of fires and area burned. Although this currently requires manual reporting, tabulation, and graphing, it could be automated by having data entered directly into a remote database. A project has been proposed to develop an automated national satellite monitoring and mapping system for fires >200 ha. This system would transmit large-fire maps and associated statistics directly to the CWFIS for distribution via the web server (see also Global Fire Monitoring, below).
It is important to continuously monitor the disposition of suppression resources. This includes the location and status of individual resources as well as potential availability for inter-agency mobilization. Manual systems are in place for monitoring resource status at agency and national levels; this information could be displayed by the CWFIS.
It is useful to provide public information on the status of individual fires on the world-wide web. Providing an alternate media access point reduces the workload of public information officials during fire emergencies. Nationally, an overall synopsis of the current situation and prognosis for the near future is useful for senior executives, policy analysts, and governments. Reports are prepared manually and distributed through the Internet.
Decision support systems
Decision-support systems (DSS) are often used for complex tasks, such as resource prepositioning, detection route planning, fire prioritization, and dispatch. Most agencies in Canada operate such systems.
A web-based fire information system provides an interoperable platform to inform users about scientific results and technological developments. It also allows users to test and evaluate new systems. Accessibility through the web allows system developers to focus on underlying technology while avoiding system-specific idiosyncrasies. The CWFIS accomplishes this through a link to the Canadian Forest Service Fire Research Network, where emerging technologies such as hourly and seasonal fire growth models can be tested.
The most important aspect of the CWFIS may be its use as an example and a platform that enables fire management agencies to exchange information among themselves. The CWFIS also provides a national node that links individual fire agencies to the global fire community and vice versa. Similar national nodes in other countries could be linked readily to form a global forest fire information network. For example, FireNet (Australia) has proven invaluable as the principle server for a global fire community discussion group.
Canadian experience has shown that exchanging information among fire agencies is a precursor to developing mutual understanding. This, in turn, fosters agreements to exchange resources as no agency or nation can be an island unto itself in fire management. Prior inter-agency and intergovernmental agreements are the key to avoiding bureaucratic delays that can preclude effective resource exchanges. The process begins slowly and increases gradually as mutual trust develops among agencies. Implementing resource exchanges also fosters common standards for equipment and training; exchanging people fosters technology transfer. The overall result is enhanced fire management effectiveness and efficiency among all participants.
Global fire monitoring
It is currently technically feasible technically to use the described earth observation and information systems to collect, analyse and share information on wildfire throughout the world on a daily basis. The Monitoring of Tropical Vegetation Unit of the Space Applications Institute at the EC Joint Research Centre has been working on a global fire dataset based on the NOAA AVHRR products (Malingreau and Grégoire, 1996; Grégoire et al., 1996). The “Global Fire Product”, in its first phase, is generating a dataset for the 21 months of global daily coverage from April 1992 to December 1993. Because of the significance of the dataset for global change studies, the latest state-of-the art report was produced under the umbrella of the International Geosphere-Biosphere Programme Data and Information System (IGBP-DIS) (Malingreau and Justice, 1997).
Malingreau (1996) recently proposed the creation of the World Fire Web in which a network of centres with facilities to receive and process fire observation data from satellites, will be connected via the World Wide Web (WWW). Through the World Fire Web scientists, managers, and policy makers can have instant access to local, regional and world data; they can exchange experience, methods and trouble-shoot with each other. The World Fire Web, in conjunction with the space borne evaluation of vegetation dryness and fire-weather forecasts can provide a powerful early warning and disaster preparedness and management tool.