While there are several fuel types in Australia, only two have fire prediction mechanisms: grass and eucalypt forest. However, the forest fire meter has been openly questioned recently. The difficult fuel types are elevated ones with live fuel, which generate the majority of the most spectacular and the most dangerous fires, but most fuel types receive only estimates or best guesses, and each professional has his or her own method. Is this an acceptable foundation to base an industry on?

Now time has exposed the flaws and the benefits of the system. More than 40 years ago, Alan McArthur set up the framework for Australia's fire protection system and gave us confidence to predict bushfire behavior, to defend against it, to suppress it and to conduct controlled burns. However, since his retirement, little has been done to build on it. McArthur's system is based on assumptions that are sometimes flawed. The fact that the system was originally based on flawed assumptions is less important than the need for a system. It's time to address the deficiencies and keep the system relevant and dynamic.

I previously had developed a wildfire risk assessment and management system that depended on fire behavior inputs. But I found that I was guessing fire behavior in most fuel types, particularly in plantations, so I took time out to examine other literature on fire behavior. It became clear that the long-standing theories aimed to explain fire behavior by incorporating several input variables into each model. As a result, inputs were complex, many interacted, and the prediction process was difficult to understand.

A SIMPLIFIED APPROACH

How can fire behavior knowledge and the prediction process be simplified? Wildfire crosses a landscape under the general influence of definable macro weather conditions, particularly atmospheric dryness and wind speed, but its behavior varies on each micro site according to a number of other influences. If we focus on benchmarking fire behavior in the driest conditions for a describable fuel type using a few key inputs, we could achieve a better understanding of each fuel type and also simplify the prediction process. The additional input variables that influence fire behavior on each micro site can then be incorporated as an independent step.

This three-part benchmark theory can be described as a blend of Rothermel and McArthur approaches, but simplified. This new system allows prediction and analysis of fire behavior in all fuel types based on the following assumptions:

• Wildfire behavior is predictable through behavior of the flame of the head fire and spotting activity. Extreme flame activity within the head fire can be expected, but its location and scale can be explained only in hindsight.
• Benchmark fire behavior in a specified fuel type can be predicted by two primary inputs: atmospheric dryness and wind at fuel level.
• Benchmark fire behavior can be adjusted to site-specific prediction by incorporating other appropriate inputs such as fuel, weather or topography.

Much of the data is derived from quality research and reliable observation, but some falls into the circumstantial evidence category. Data and observation are usually used to either illustrate a principle or a theory. They may not pass the test of statistical significance or experimental design but in the bushfire behavior field where data often is scarce, reasonable data are better than none. Where data and observation are used to build an equation, the sources and the logic and the assumptions are given.

The practical outcome is a set of indicative fire behavior tables for each fuel type, all in similar format and each with common inputs. These have been tested against documented fires, both wildfires and prescribed burns. Their accuracy level can be classified as “operationally useful” for planning and decision purposes. In real fires, subtle changes in a large number of variables can influence on-site fire behavior to the point that prediction may seem to be an exercise in futility. These tables focus on the most significant influences. If observed fire behavior is different from predicted, the tables provide a starting point to search for reasons why.

SYSTEMATIC RESPONSE

Understanding fire behavior in complex fuel types begins with an understanding of single-layer fuels. The confusing thing about forest fire behavior is the combination of a number of fuel layers. The approach I used was to start with the simple and the well-studied, then move to the complex and less studied. If we understand litter-only fire behavior and grass-only fire behavior, we can then move on to multiple layers. If we can understand what happens with dead fuel beds, then we can move on to live-fuel beds.

The Fire Danger Index is directly related to the chances of a fire starting, its behavior and difficulty of suppression. The FDI scale is well-accepted in Australia. It's included in forecasts and is used to determine preparedness schedules and as a basis for fire protection planning. The McArthur Meter uses FDI to predict fire behavior.

McArthur developed fire danger meters for forest and grass fires. They scale from 0 to 100, and are intended to be an indicator of chance of ignition, speed and intensity of fire and difficulty of suppression. FDI for a given day is derived from temperature and relative humidity, which together indicate atmospheric dryness and wind speed.

The forest FDI calculation includes an adjustment to account for seasonal dryness, or the drought index (DI), and recent rainfall, the drought factor (DF). The grassland FDI is adjusted for curing percentage. We can see readily that for a given day, the FDI values for forest and grass can be substantially different.

The forest FDI scale has been consistent since inception, whereas the grassfire scale has varied. For a while, fuel load was incorporated into it, but the grassfire scale also has other limitations:

• The curing adjustment may be suitable for grasslands during late spring months, but it can also tend to reduce FDI for serious weather during that time.
• When curing is 100%, the grass meter does not take moisture content fluctuations due to rainfall into account.
• Grasslands are exposed to full sun, and tend to dry in a different pattern to tree- or shrub-shaded fuels.

The forest FDI scale can apply to both grass and forest, whereas the grass meter is restricted to grasslands. It also has legitimacy in other fuel types including dead litter in long dry periods.

Fundamentally, FDI is a combination of dryness of the atmosphere and wind speed, but caution is required in its interpretation. For most of the values, a given FDI can range from a dry atmosphere with light winds to a moist atmosphere with strong winds. In real life, these differences have a substantial influence on chance of fire ignition, fire behavior and chance of suppression. The meter operates with the assumption that wind speed and atmospheric dryness are of equal and opposite influence. But they are not.

For example, FDI 30 can result from 35°C, 20% relative humidity and 20kph winds or from 30°C, 30% relative humidity and 40kph winds. Equilibrium fuel moisture content varies from 4% to 5%, but it has negligible influence in comparison to doubling the wind speed, yet the meter predicts constant fire behavior. (See “Equilibrium fuel moisture content,” opposite.) FDI above 50 is generally a good indicator of very dry conditions and strong winds, but the same variability factor applies.

On the forest meter, FDI is combined with fuel load to predict the major elements of fire behavior, rate of spread, flame height and spotting distance. As shown above, the weather combination can vary, but the meter predicts specific fire behavior for a given FDI. This assumes that wind speed and atmospheric dryness are of equal and opposite influence on fire behavior. But they are not. Wind speed has a much more powerful influence than atmospheric dryness.

As an indicator of chance of fire ignition, atmospheric dryness is a more relevant measure than wind speed, but FDI does not distinguish its components. As an input for estimating fire behavior in the forest, FDI is inadequate because it fails to respond to changes in influential inputs. FDI is useful as an indicator of fire suppression difficulty, but there are so many other relevant factors it does not consider.

Appropriate use of FDI requires qualification. The best way to minimize the ambiguity of the current FDI system as an indicator of fire danger is to quote weather conditions, particularly wind speed.

An improvement would be to add a double coded index, one code for atmospheric dryness and one for wind speed. For example, if each code had three levels, say 1, 2 and 3 for low, medium and high, a 1-to-3 day would mean low dryness and high wind speed. An “FDI 30 / code 1-to-3” day is a more meaningful indicator to firefighters than simply an FDI 30 day.

SEASONAL WEATHER

The McArthur Meter includes an adjustment for seasonal weather conditions. The intention is valid, but the indicator is not directly relevant to estimating fire behavior in fine dead fuel.

The Byram-Keetch Drought Index is a recognized indicator reflecting the of seasonal dryness of soils, deep forest litter, logs and living vegetation, and it correlates well with moisture content of larger fuels. Soil moisture index is a similar measure that is used in Western Australia and Tasmania. Drought index indicates rainfall deficit in the upper soil layers by calculating the net effect of evapotranspiration, which causes soil moisture depletion, and rainfall. For a given vegetation cover, evapotranspiration rate increases with temperature and atmospheric dryness, but decreases with decreasing rainfall.

In typical summer temperatures of high 20s to low 30s Celsius, DI increases by three to eight millimeters per day. If vegetation cover increases, the evapotranspiration rate also increases. As a result, in rain-free weather, soil moisture decreases exponentially with time to a minimum at wilting point. Conversely, DI increases asymptotically in rain-free weather to a theoretical maximum of 200.

When rain occurs, drought decreases by the quantity that reaches the ground. The DI scale runs from 0 to 200, based on the arbitrary measure that, between saturation and wilting point, soil has a field capacity of 200mm.

Drought index also indicates difficulty of mop up and is a good indicator of internal dryness of large logs. A high drought index means that fires burn deeper into larger fuels, and therefore are more difficult to extinguish and tie up firefighting resources longer. Also, the higher the drought index, the greater the chance of underground root burning fires.

The drought index can be a good indicator of live foliage moisture content. McArthur noted that moisture content of shrubs falls exponentially from around 140% to 80% as drought index increases. McArthur used drought index methods to indicate seasonal severity and measure of fuel availability, eventually linking these to fire behavior.

This relationship between fuel availability and drought index makes sense if first we can assume, that a higher drought index increases the ability of the soil to dry the fuel particle; second, that the fuel particle is in contact with the soil; and third, that low drought index is an indicator of high relative humidity near ground level. In other words, the statement is correct if a high drought index means that litter moisture is lower and therefore more fuel is available for combustion than a low drought index site under the same temperature and relative humidity.

Unfortunately, the assumption is flawed because the drying power of the air has greater influence on surface FMC than the dryness of the soil, as the fuel particles are usually not in direct contact with the soil. Furthermore, even if a low drought index generates a higher relative humidity at ground level, relative humidity is readily changed by a dry northerly wind. Therefore, drought index can't be used to indicate fuel moisture content or fuel availability.

It's incorrect to relate drought index to fire behavior characteristics because behavior is determined by daily weather, whereas the drought index is determined by seasonal weather. Including drought index in fire behavior calculations is not only irrelevant, but it also also lead to significant under-predictions. For example, if seasonal drought index is low, the effect is to reduce calculated fire drought index. But if a spell of dry windy weather dries out fine fuel rapidly, actual fire behavior can exceed predicted fire behavior and catch people by surprise.

RECENT RAINFALL

McArthur's approach to rainfall on eucalypt litter is based on the theory that recent rainfall reduces fuel availability, which therefore reduces rate of spread and flame height. The theory is not strictly correct, because the true effect of recent rain is to increase net fuel bed moisture content, and this is what reduces rate of spread and flame height. Furthermore, as will be examined later, fuel load has a limited influence on rate of spread compared to wind speed. Nevertheless, even though McArthur's foundation theory is incorrect, a similar end result is achieved.

G.M. Byram introduced the term “available fuel” to mean the quantity burned during the flame phase. This is the correct definition and, in fact, fire behavior studies do not make sense unless this is clearly understood. It varies with fuel moisture content and with fire intensity. He distinguished this from total fuel, which is the quantity burned under the driest conditions and the highest intensity fire. Byram said that fuel availability, the quantity burned, is reduced as fuel moisture content increases, and that fuel moisture content is influenced by rainfall, temperature and relative humidity.

McArthur's use of the term fuel availability was slightly different. The relevant fuel quantity in fire behavior was fine fuel available for combustion, but he did not distinguish flame phase from flame and smolder phase. Harry Luke and McArthur said that fuel availability refers to the proportion of fine fuel that will burn in a fire, and that it varies with fuel moisture content, fuel thickness and arrangement.

McArthur strongly linked fuel availability to rate of spread, fire intensity and flame height. He said that fuel availability is reduced by recent rain, meaning that a proportion of the fuel bed remains unburnt. He does not seem to have linked rainfall with its effect on net or average fuel bed moisture content. Yet his studies of fuel bed moisture content as determined by temperature and relative humidity found that fuel moisture content has a strong influence on fire behavior: rate of spread, flame height, spotting, crown fire formation.

McArthur's method can be summarized as follows:

1. Recent rainfall reduces fuel availability,
2. Atmospheric dryness determines fine fuel moisture content, and
3. Fuel availability and moisture content determine fire behavior.

With the influence of decades of estimation for control burning, available fuel has become understood in many quarters to mean total fine fuel load on site. A recent fuel hazard guide proposes that the estimated fuel load of litter, bark and elevated shrub foliage be added together and applied to the McArthur Meter to calculate rate of spread and flame height.

The McArthur Meter presents a systematic method for determining the influence of recent rainfall on fire behavior. It uses the drought index, amount of rainfall and number of days since rain to determine drought factor. This is then used to adjust fire danger index, and the adjusted FDI is combined with fuel load to determine fire behavior.

McArthur concluded from studies in some 400 eucalypt litter fires that the amount of available fuel on the forest floor has a significant influence on spread and fire intensity. He believed that available fuel load determined rate of spread and that recent rainfall caused a reduction in available fuel and would therefore reduce rate of spread and intensity.

He devised a fuel reduction factor to track fuel availability as the fuel dried out after rain. He presents rate of spread as being directly related to the fuel reduction factor. He set FRF to range from 0 to 1 and applied it as follows.

Expected ROS = FRF × maximum ROS for those conditions.

For example, if it has been two days since an area received 25mm of rain, the FRF calculates to be 0.4. This means that 40% of fuel will burn, and this reduces the rate of spread to 40% of the maximum and also leads to lower flame height.

McArthur linked FRF to available fuel load, but an alternative approach seems more realistic. Why did McArthur not observe that rainfall causes fuel bed moisture content to increase and that the fuel moisture content change reduces rate of spread and flame height? Is it not reasonable to view the fuel reduction factor as an indicator that influences fuel moisture content?

McArthur's approach raises two critical questions, but the answers do not appear to support his beliefs.

• What is the evidence that fuel load determines rate of spread? The only available evidence applies to low-intensity fires. When wind becomes a serious influence, fuel load is irrelevant.
• What is the evidence that higher fuel moisture content reduces the fuel quantity consumed during the flame phase? McArthur doesn't present any. R.C. Rothermel and H.E. Anderson's wind tunnel work suggests that the differences in percentage consumed at different fuel moisture contents are insignificant, despite significant differences in rate of spread.

So what then happens when fine fuel is wetted? The logical effect of wetting the fuel is to raise its average fuel moisture content. Yet McArthur does not take this into account, despite several tables and graphs linking fuel moisture content to rate of spread and then to flame height.

When a litter bed is partially wetted, its average fuel moisture content not only rises but may vary throughout its depth. How does it affect fire behavior? Three scenarios apply:

1. If the surface of all the fine fuel particles is wetted, the process of flame generation absorbs heat during evaporation and radiation is absorbed in the steamy flame, which causes the fire to be sluggish and slow. This compares to the effect of curing percentage in grasslands.

2. If the top layer of the litter is wet and the bottom is dry, the fire from the dry fuel may generate enough heat to render the surface damp fuel insignificant.

3. It the bottom layer is wet and the top is dry, the fire will spread unrestricted, leaving the wet layer below to smolder after the flame has passed.

Logically, the FRF system applies to the first scenario. The FRF system could lead to underestimating fire behavior in other scenarios.

If the belief about fuel load and rate of spread did not exist, fire reduction factor could just as easily be interpreted as the effect on average fuel moisture content that increases after rainfall and falls as it progressively dries out.

DANGER OF DROUGHT

Drought factor was described as a measure of fuel availability as affected by recent rainfall and seasonal conditions. The McArthur Meter can be used to predict fire behavior, but the meter does not adjust fuel load availability directly. The meter applies recent rainfall to calculate a drought factor, and then uses temperature and relative humidity to calculate fire danger index. Fuel load is then matched with this DF-modified FDI to determine modified rate of spread.

Using the McArthur Forest Meter to calculate fire behavior, the first step is to determine drought factor from recent rainfall. Because rain gauges are in the open and do not account for absorption by overhead foliage, the effective rainfall falling to the litter bed tends to be over-estimated, which means fire behavior will tend to be under-estimated.

The drying pattern depends on two main variables, the amount of rainfall received and presumably absorbed, and seasonal dryness or the drought index. This implies that daily evaporation is a function of seasonal dryness, which of course it is not. It would be more accurate if the daily drying pattern were linked to daily atmospheric dryness.

Note that for a drought index greater than 100, the drought factor peaks at 10, and for DI 60-100, DF peaks at 9.

The McArthur Meter sets a maximum value for drought factor, depending on the drought index.

DI > 100, maximum DF = 10
100 > DI >60, maximum DF = 9
60 > DI >25, maximum DF = 7
DI < 25, maximum DF = 6

The logic of setting categories and maximum drought factor is faulty for a few reasons. Drought index relates to seasonal soil dryness, whereas drought factor relates to daily fuel moisture content. It assumes the soil has greater control over fuel moisture content than the air. It assumes the fuel particle in contact with the soil. Moreover, even if a low drought index generates high relative humidity at ground level in calm morning air, daily weather readily equalizes relative humidity.

The next step is to see how drought factor affects the fire danger index. The fire danger index is directly proportional to drought factor.

The final step is to determine rate of spread and flame height using fuel load and calculated fire danger index.

Drought factor is directly proportional to fire danger index at a given wind speed. McArthur believed fire danger index was directly proportional to rate of spread for any given fuel load. Combining the two relationships, we see that McArthur believed drought factor was directly proportional to rate of spread at constant wind speed and fuel load.

Unfortunately, this simple relationship falls down because fire danger index is also affected by wind speed. The process is only valid if wind speed remains constant, which does not happen in real life. The true influence of drought factor on fuel availability is thus obfuscated by the influence of wind speed, and this makes the application of drought factor to determine rate of spread on the meter invalid.

To determine flame height, McArthur used fire danger index and available fuel load. However, this method can't distinguish the effect of wind speed from fuel moisture content as affected by recent rain, or temperatures and relative humidity.

Drought factor is designed to relate to changes in fuel moisture content of the forest litter fuel type. It may be useful to understand the drying pattern of forest fuel after substantial rain when the entire litter layer is drenched, but its relevance after less rain is dubious.

It's useful for controlled burning planning. We do not know the drying pattern of grass, or heath or mallee in the open air after rain, but can expect it to be different to sheltered litter. Therefore the inclusion of drought factor in a universal fire danger rating is probably unrealistic, and where fire danger index is used to calculate fire behavior, could result in under-predictions.

A fire prediction system is most critical when dead fuel is driest and under the influence of atmospheric dryness. Then an accurate prediction system reveals the worst fire behavior to be expected. If recent rainfall, or dew formation, or sheltered slopes are relevant on the day, they will tend to ameliorate fire behavior, which means the worst result is an over-estimation of fire behavior, and this could be a valuable safety margin.

Denis O'Bryan graduated from Creswick School of Forestry and Melbourne University and began work as a forester. He has had more than 20 years of experience working in the Victoria government in fire protection at all levels, including hands-on firefighting, fire crew leadership, pre-season planning and preparation, training, statewide fire protection planning, and coordination of statewide firefighting operations. After leaving the government, he trained large numbers of volunteer and professional firefighters and tertiary students in basic and advanced fire courses. O'Bryan has consulted to the Victoria and federal governments in fire policy and planning issues, and has worked interstate. He currently serves as director of Red Eagle, a bushfire protection advisory service that provides objective fire risk assessment and management expertise and training.

Equilibrium fuel moisture content

Equilibrium FMC tables for eucalypt litter beds are based on equilibrium conditions between 12:00 and 16:00 hours in tall eucalypt forest with filtered sun exposure, and apply to surface layer of litter for midsummer in a very dry fuel bed. Fuel moisture content in the morning can be estimated by adding 2-3 % to the chart, and for later in the afternoon by subtracting 1-2 %. This lag is called the hysteresis effect. Assume that danger index and drought factor are constant. An FDI 30 derive can from 35°C, 20% RH and 20kph winds; or from 30°C, 30% RH and 40kph winds, yet the meter will predict the same fire behavior because FDI is 30. An FDI of 70 can derive from 40°C, 5% RH and 20kph; or from 40°C, 10% RH and 30kph or 35°C, 15% RH and 50 kph. A fire controller is entitled to expect that a credible fire behavior prediction table would vary according to the changes in the influential inputs.

The McArthur forest

The reference forest upon which the McArthur Meter is based has the following characteristics:

• A litter layer of the predominant fuel type with full cover and 12 tonnes per hectare of fine fuel;
• A shrub layer with low height and low percentage of cover;
• A canopy layer that is 25 to 30 meters tall, evenly aged and provides 40-50% cover; and
• A tree trunk layer that is flammable.

Anomaly

We have seen that McArthur believed the drought factor was directly proportional to the rate of spread at constant wind speed and fuel load. However, the DF scale is not a linear scale, it is a log scale. By contrast, the relationship between curing percentage (or the percentage of dead fuel) and rate of spread in grassland is very different. Rate of spread increases exponentially as curing percentage increases. (Note also that rate of spread increases exponentially as moisture content decreases in both litter and dead grass fuels.)

A similar process operates on the McArthur Meter

The fuel reduction factor is now repackaged as drought factor. (DF = FRF × 10). DF ranges from 1 to 10. Say DF = 3.5. This means FDI is 35% of potential for given temperature, relative humidity and wind speed. This means rate of spread is 35% of its potential, because the rate of spread is proportional to FDI at constant fuel load.