The relative importance of fuels and weather on fire behavior in subalpine forests

W.C. Bessie

INTRODUCTION

Forest fire ecologists have found it difficult to determine the relative importance of fuels or weather on fire behavior (e.g., intensity and crown fire occurrence). On the one hand fuel accumulation (Heinselman 1973, Romme and Knight 1981, Lotan et al. 1985), fuel chemistry (Mutch 1970), crown/understory fuel structure (Despain and Sellers 1977), and community development or aging (Habeck and Mutch 1973, Agee and Huff 1987, Clark 1988, Despain 1990) are often thought to be central in determining fire behavior. On the other hand, weather variability is also thought to be central in determining fire behavior (e.g., Schroeder et al. 1964, Anderson 1968, Finklin 1973, Newark 1975, Alexander et al. 1983, Harrington et al. 1983, Street 1985, Flannigan and Harrington 1986, Stocks and Flannigan 1987, Harrington and Flannigan 1987, Fryer and Johnson 1988, Swetnam and Betancourt 1990, Johnson and Larsen 1991, Johnson and Wowchuk 1993, Swetnam 1993).

The purpose here is to assess the relative roles played by weather (fuel moisture content and wind speed) and fuel (fuel loads, fuel depth, mass density, heat of combustion and surface-area-to-volume ratio) in forest fire behavior. Fuel data collected from 47 subalpine forest stands and weather data from the southern Canadian Rocky Mountains were used to model fire intensity with Rothermel's (1972) surface fire intensity model and crown fire initiation with Van Wagner's (1977) crown fire initiation model. Both models were reorganized by collecting all weather and fuel components into single fuel and weather variables that allowed the determination of the relative contribution of fuel load/crown structure variation and weather variation on surface intensity and crown fire initiation. These variables also allowed quantitative examination of weather-caused variation in fire behavior among years and fuel-caused variation in fire behavior among stands of varying composition and ages.

Fire behavior modelling objectively determines the role of fuels in fire behavior based on the mechanisms of fire behavior. This approach differs from simply accepting the intuitive assumption that increased fuel loads always lead to increased fire behavior potential (e.g., Heinselman 1973). Furthermore, the inclusion of a full range of weather conditions in the fire behavior modelling differs from past approaches in which all fires were modelled under identical weather conditions (Agee and Huff 1987, Keane et al. 1990).

Our objective is to answer the following three questions:

1) What are the relative roles of fuel and weather in each of surface fire intensity and crown fire initiation?

2) How strongly does fire behavior variation relate to weather variation, when the effects of fuels are excluded from the predictions?

3) How does the potential for high intensity or crown fires relate to fuel variation among stand ages when the effects of weather are excluded from the predictions?

Surface fire intensity model

Rothermel's (1972) model predicts rate of spread (metres per second) and intensity (kilowatts per metre) for surface forest fires. This model is the basis of the National Fire Danger Rating System of the U.S. Forest Service (Bradshaw et al. 1984). Fire spread is idealized as a series of particle ignitions at a rate determined by the interaction of fuels, weather, and slope. Rate of spread (R, metres per second) is determined by dividing the forward heat flux (kilowatts per square metre) by the fuel heat sink (kilojoules per cubic metre):

R = Flux/Sink = [I.sub.R][Xi](1 + [[Phi].sub.w] + [[Phi].sub.s])/[[Rho].sub.b][Epsilon][Q.sub.ig], (1)

where

[I.sub.R] = reaction intensity (kW/[m.sup.2])

[Xi] = propagating flux ratio (dimensionless)

[[Phi].sub.w] = wind coefficient (dimensionless)

[[Phi].sub.s] = slope coefficient (dimensionless)

[[Rho].sub.b] = fuel bulk density (kg/[m.sup.3])

[Epsilon] = effective heating number (dimensionless)

[Q.sub.ig] = heat of pre-ignition (kJ/kg)

Reaction intensity is the total rate of heat release per unit area (kilowatts per square metre):

[I.sub.R] = [[Eta].sub.m][[Eta].sub.s][Gamma]h[W.sub.n] (2)

[[Eta].sub.m] = moisture damping function (dimensionless; The moisture of extinction used in calculating [[Eta].sub.m] was 0.3.)

[[Eta].sub.s] = mineral damping function (dimensionless)

[Gamma] = optimum combustion rate ([s.sup.-1])

h = heat of combustion (kJ/kg)

[W.sub.n] = net fuel load (kg/[m.sup.2]).

Intensity (I in kilowatts per metre) is the product of rate of spread, reaction intensity, and flame residence time (Albini 1976):

I = [I.sub.R][Rt.sub.r], (3)

where

[I.sub.R] = reaction intensity (kW/[m.sup.2])

R = rate of spread (m/s)

[t.sub.r] = flame residence time (s).

Substitution and rearrangement results in the following intensity equation:

I = [([[Eta].sub.m][[Eta].sub.s][Gamma]h[W.sub.n]).sup.2][Xi](1 + [[Phi].sub.w] + [[Phi].sub.s])[t.sub.r]/[[Rho].sub.b][Epsilon][Q.sub.ig]. (4)

There are two general groups of components: fuel components (variables derived from fuel load, surface-to-volume ratio, fuel depth, mass density, heat of combustion), and environmental components (variables determined from wind speed, moisture content, and slope). These two groups of components were separated to define a fuel variable ([F.sub.v]) with the dimensions of kilowatts per metre and weather variable ([E.sub.v]) that when [greater than] 1 magnifies the rate in [F.sub.v] and when [less than] 1 reduces the rate in [F.sub.v]. Thus:

I = [E.sub.v] [multiplied by] [F.sub.v]. (5)

The fuel variable ([F.sub.v]) can be described as the intensity at which a fire would burn if [E.sub.v] = 1, which occurs when there is no wind and fuels are completely dry. The fuel variable included the constant heat of preignition of dry fuels ([Q.sub.d]: 581 kJ/kg) such that the fuel variable had the same units as intensity (kilowatts per metre):

[F.sub.v] = [([[Eta].sub.s][Gamma]h[W.sub.n]).sup.2][Xi][t.sub.r]/[[Rho].sub. b][Epsilon][Q.sub.d]. (6)

The weather variable (dimensionless) included the effects of fuel moisture content and windspeed (and slope, which is held constant at zero slope) on surface fire intensity:

[E.sub.v] = [([[Eta].sub.m]).sup.2](1 + [[Phi].sub.w] + [[Phi].sub.s])[Q.sub.d]/[Q.sub.ig]. (7)

Notice that [Q.sub.d], for dimensional purposes, appears in both [F.sub.v] and [E.sub.v] so that it cancels when multiplied. For [E.sub.v] to equal 1, the effect of fuel moisture must balance the effects of wind and slope.

Crown fire initiation model

Crown fire initiation involves two steps. First, crown fuel ignition from a burning surface fire must occur. Second, rate of spread must allow direct crown-to-crown fire spread. The first step alone results in a passive crown fire with the surface fire largely affecting the spread rate. If both steps are achieved, the fire becomes an active crown fire with the burning crowns largely affecting the spread rate; the flaming front is a wall of flames (Van Wagner 1977). Passive crown fire intensity will be somewhat higher than surface fire intensity due to additional consumption of crown fuels; active crown fire intensity is 10-100 times higher than surface fire intensity due to increased fuel consumption and higher spread rates (Van Wagner 1980). This model of crown fire initiation is incorporated in the Canadian Forest Fire Behavior Prediction System (Fire Danger Group 1992).

Crown fuel ignition occurs when the surface fire intensity exceeds the critical surface intensity ([I.sub.p]: kilowatts per metre):

[I.sub.p] = [(0.01z[Q.sub.c]).sup.3/2], (8)

where

z = lower crown base height (m)

[Q.sub.c] = heat of crown foliage ignition (kJ/kg).

The value 0.01 is an empirical constant of complex dimensions (cf. Van Wagner 1977). The heat of foliage ignition ([Q.sub.c]) is assumed to be constant at 3060 kJ/kg, assuming foliage at 100% moisture (dry mass). A minimum critical intensity value is set at 200 kW/m. Eq. 8 states that crowning could occur when intensity is very small and crown height is zero. This is, however, not very reasonable, so we have chosen arbitrarily the 1 m crown height which gives an intensity of [approximately equal to]200 kW/m. Fire intensity predictions are then compared to the critical intensity to predict passive crown fire initiation.

A passive crown (fire initiation) fuel variable ([C.sub.p]) is defined as:

[C.sub.p] = [F.sub.v]/[I.sub.p], (9)

where

[C.sub.p] = passive crown fuel variable (dimensionless)

[F.sub.v] = surface fuel variable (kW/m)

[I.sub.p] = critical intensity (kW/m).

When [C.sub.p] [greater than] 1, surface fuels produce enough heat to cause crowning.

Crown-to-crown fire spread (active crown fire initiation) is achieved when crown fire rate of spread exceeds the critical rate of spread ([R.sub.a]: m/s):

[R.sub.a] = [S.sub.0]/[[Rho].sub.c], (10)

where

[S.sub.0] = critical mass flow rate (kg [multiplied by] [m.sup.-2] [multiplied by] [s.sup.-1])

[[Rho].sub.c] = crown fuel bulk density (kg/[m.sup.3]).

A critical mass flow rate for conifer forests was empirically determined to be 0.05 kg [multiplied by] [m.sup.-2] [multiplied by] [s.sup.-1] (Van Wagner 1977). The critical rate of spread is compared to an estimate of crown fire rate of spread to determine whether the crown fire should be classed as active.

An active crown (fire initiation) fuel variable ([C.sub.a]) is defined by first converting the critical crown rate of spread ([R.sub.a]) to a surface rate of spread value. This conversion uses the empirical relationship (Rothermel 1991) between predicted surface fire rate of spread (R, Eq. 1) and active crown fire rate of spread ([R.sub.c]):

[R.sub.c]/3.34 = R. (11)

Since we have no model to predict crown fire rate of spread for comparison to [R.sub.a] (to determine crowning), we make the assumption that [R.sub.a] can be substituted for [R.sub.c] to give an approximate surface fire critical rate of spread that can be compared to R from Rothermel's (1972) model. Using this reasoning, the critical surface rate of spread ([R.sub.a]/3.34) is multiplied by heat of combustion (h) and the net surface fuel load ([W.sub.n]) to give a second critical intensity value that pertains to active crowning:

[Mathematical Expression Omitted],

where

[I.sub.a] = critical intensity for active crowning (kW/m)

[R.sub.a] = critical rate of spread (m/s)

h = low heat of combustion (12 700 kJ/kg)

[W.sub.n] = net surface fuel load (kg/[m.sup.-2]) from Rothermel's model.

Note that both [I.sub.a] and [I.sub.p] pertain to surface fire conditions and therefore only results from Rothermel's model are required to determine when crowning will occur. Finally, substitution of [I.sub.a] for [I.sub.p] in Eq. 9 gives the active crown fuel variable ([C.sub.a]).

[C.sub.a] = [F.sub.v]/[I.sub.a], (13)

following Van Wagner's (1977) criteria for active crowning.

The initiation of passive or active crown fires (I/[I.sub.p] or I/[I.sub.a]) is the appropriate crown fuel variable ([C.sub.p] or [C.sub.a]) multiplied by the weather variable ([E.sub.v], Eq. 6):

[C.sub.p][E.sub.v] = I/[I.sub.p]; [C.sub.a][E.sub.v] = I/[I.sub.a]. (14)

When I/[I.sub.p] [less than] 1, the fire is a surface fire. When I/[I.sub.p] [greater than or equal to] 1 but I/[I.sub.a] [less than] 1, the fire is a passive crown fire. When I/[I.sub.p] [greater than or equal to] 1 and I/[I.sub.a] [greater than or equal to] 1, the fire is an active crown fire.

METHODS

Study area

Forty-seven stands were sampled to obtain estimates of surface and crown fuel properties for input to the fire models. Stands are representative of fuel types that cover [greater than] 95% of the study area. Stand locations and characteristics (elevation, age, aspect, density, basal area, slope, and overstory tree percentages) are listed in Table 1. Forty-four stands were located in the Kanariaskis Valley of southwest Alberta situated in the front and main ranges of the southern Canadian Rocky Mountains, [approximately equal to]100 km west of Calgary, Alberta. Three additional young stands (22 yr) were sampled in the Vermillion Pass area of Banff and Kootenay National Parks (Table 1). All sampled stands were upland sub-alpine conifer forests and represent the range of principal vegetation types. These forests consist of Pinus contorta Loud. var latifolia Engelm., Picea engelmannii [TABULAR DATA FOR TABLE 1 OMITTED] Parry ex. Engelm., Abies lasiocarpa (Hook.) Nutt., and occasionally Populus tremuloides Michx. Most conifer stands in the Kananaskis Valley originate from stand-replacing fire disturbances with a fire cycle of 90 yr (Johnson and Larsen 1991), which result in even-aged fire cohorts of Pinus contorta or Picea engelmannii trees (Johnson and Fryer 1989). The main lightning fire season (area burned) is July and August. In the last 100 yr, fires [greater than] 400 ha have almost always been caused by lightning (E. A. Johnson and D. R. Wowchuk, personal communication).

Fuel inventory and calculations

Surface fuels were measured by methods of McRae et al. (1979) and Brown et al. (1982) along an equilateral triangle with each side 30 m long (90 m total). Downed wood fuels were measured by the line intersect technique (Van Wagner 1968). Shrub, herb, litter, and moss fuels were measured in quadrats evenly spaced along the transect. Shrubs (by species) had basal diameter measured in nine (1 [m.sup.2]) quadrats at 10-m intervals. Litter (L layer) and moss depth, herbaceous plant cover, height, and percentage dead were measured in 27 (0.06 [m.sup.2]) quadrats at 3 m intervals. Fuel loads were calculated from equations determined from fuel samples collected in the Kananaskis Valley (Appendix 1).

The surface fuel components were divided into the following classes: litter, moss/lichen, live herbs, live shrubs, 1-h time-lag fuels (dead wood, dead shrub wood, dead herbs), 10- and 100-h fuels (dead wood and dead shrub wood). Each fuel class was assigned a standard surface-area-to-volume ratio (cf. Brown and Bevins 1986) and a fuel depth based on field estimates (see Appendix).

Crown fuels were measured in three 100-[m.sup.2] quadrats per stand by measuring each tree's diameter at breast height (dbh). Crown foliage biomass was predicted from dbh using published empirical equations (see Appendix 1). Crown top and base heights were measured on a 2 m wide strip transect until three dominant and three subdominant trees per species were measured. Mean crown heights (top and base) were determined by weighting tree heights of dominant and subdominant trees by population densities in the three quadrats. Fuel loads were divided by crown depth to give the crown bulk density.

Weather data source and calculations

Daily noon weather data (temperature, precipitation, wind speed, and relative humidity) for 1954-1988 were obtained from the Atmospheric Environment Service for Banff, Alberta and represented subalpine fire-weather for the southern Canadian Rockies (cf. Johnson and Wowchuk 1993). July and August were chosen since these are the largest area burned months (Johnson and Wowchuk 1993). Herb, shrub, and dead wood moisture contents were predicted for each stand from National Fire Danger Rating System equations (Bradshaw et al. 1984). The conversion from moisture contents to [Q.sub.ig] and [[Eta].sub.m] (Eqs. 1 and 2) is accomplished via equations from Rothermel (1972). Litter moisture was predicted from the Fine Fuel Moisture Code (Van Wagner 1987). Lichen moisture was predicted from equations by Pech (1989). As no equations were available for moss, the lichen moisture was used as an approximation since moss is also a passive (i.e., no cuticle) fuel.

Surface fire intensity analyses

1. Relative importance of fuel and weather. - Surface fire intensity predictions from Rothermel's (1972) model (Eq. 4) were made using the fuel loads from 47 forest stands and the weather representing each day in July and August for the 35-yr weather record. Log transformation of the rearranged model (Eq. 5) would give log(I) = log([F.sub.v]) + log([E.sub.v]). However, in order to determine the relative importance of the fuel ([F.sub.v]) and weather ([E.sub.v]) variables, the following model was used:

log(I) = [b[prime].sub.1]log([F.sub.v]) + [b[prime].sub.2]log([E.sub.v]), (15)

where

I = intensity

[F.sub.v] = fuel variable

[E.sub.v] = weather variable

[b[prime].sub.1] = standardized partial regression coefficients.

In this model, the sum of squares of the logarithm of intensity was partitioned between the two variables such that the standardized partial regression coefficients [b[prime].sub.1] = [b.sub.1]([SS.sub.log[F.sub.v]]/[SS.sub.logI]) and [b[prime].sub.2] = [b.sub.2]([SS.sub.log[E.sub.v]]/[SS.sub.logI]) provide an estimate of the proportional change in variation in intensity produced by a unit change in variance of either the fuel or weather variable where [b.sub.i] is the partial regression coefficient (see Snedecor and Cochran 1967:398 ff.). The relative size of these standardized partial regression coefficients indicates how strongly each variable (fuel or weather) influences surface fire intensity.

2. Comparison between large and small fire years. - The amount of variation in fire intensity prediction in relation to large and small area burned years was tested using the following ANOVA model:

log(I) = [b.sub.0] + [b.sub.1] group + [b.sub.2] year(group) (16)

where

I = intensity predictions from Rothermel's model

group = large or small area burned years (1 or 0)

year(group) = year nested within groups.

Large area burned years had [greater than] 50-100 ha of forested area burned by lightning fires in the combined areas of Kootenay, Yoho, and Banff National Parks, and the Bow-Crow Forest of Alberta for the period 1954-1988 (Johnson and Wowchuk 1993). The hypothesis tested was that large area burned years have higher fire intensity predictions throughout the entire July-August period than do small area burned years.

Crown fire initiation analyses

1. Relative importance of fuels and weather. - The relative importance of the crown fuel and weather variables in crown fire initiation was determined using crown fire initiation predictions from the 47 sampled stands and 1968 summer weather. The year 1968 was chosen because it had the largest areas burned and thus represented a wide range of fire weather conditions from fire-free to actual crown fire conditions. Fire type (crown vs. surface fire) was related to the crown fuel variables ([C.sub.p]: passive crown fuel variable, and [C.sub.a]: active crown fuel variable) and weather variable ([E.sub.v]) using multiple logistic regression:

Type = exp[[b.sub.0] + [b.sub.1]([C.sub.x]) + [b.sub.2]([E.sub.v])]/1 + exp[[b.sub.0] + [b.sub.1]([C.sub.x]) + [b.sub.2]([E.sub.v])], (17)

where

Type = crown fire type probability

[C.sub.x] = passive or active crown fuel variable

[E.sub.v] = weather variable

The initiation of passive crown fires from surface fires was related to [C.sub.p] and [E.sub.v]. The initiation of active crown fires was examined in relation to [C.sub.a] and [E.sub.v] in two ways: first by excluding all surface fire predictions from the [TABULAR DATA FOR TABLE 2 OMITTED] data, and second by including them. The first test examined only the transition from passive to active crown fire, whereas the second test examined the entire transition from surface to passive to active crown fire as a single step. The relative importance of crown fuel and weather variables was determined from the standardized partial logistic regression coefficients that indicated the relative effect of each variable on each initiation examined. The weather variable is not a time series in the equation since the fire fuel moisture and wind speed are not highly autocorrelated from one day to the next.

2. Crown fire initiation: graphical model. - A logarithmic contour plot of the crown fire initiation relationship of Eq. 13 was defined with the logarithm of the weather variable on the abscissa and crown fuel variable on the ordinate. This graph, by definition, had a crown fire initiation line running along the diagonal from upper left to lower right, which represented the point of passive or active crown fire initiation for each combination of the crown fuel and weather variables. Upper and lower boundary values of the passive and active crown fuel variables ([C.sub.p] and [C.sub.a]) were plotted on the ordinate and interpolated through the transition line to predict the weather variable ranges at which crowning will: (a) occur in no stands; (b) occur in some stands; or (c) occur in all stands. The values chosen to represent the upper and lower boundaries of stand conditions were the [10.sup.th] and [90.sup.th] percentiles of each of the variables [C.sub.p] and [C.sub.a]. These percentiles were chosen so that extremely high or low conditions that do not commonly occur in forest stands would not bias the results away from the central tendency of conditions. The percentages of days in those types were then compared between large and small area burned years.

Weather variable frequencies in large and small fire years

The weather variable ([E.sub.v]) was predicted for all stands for July and August 1954-1988 for large and small area burned years. The weather variable values were fitted to a negative exponential density distribution of the form: percent frequency = [Alpha] exp(-[Beta][E.sub.v]). The distributions of large and small area burned years were then compared using a heterogeneous slope t test (Zar 1984). This analysis tested whether there were differences in the frequency distribution of weather variable values between large and small area burned years.

Fuel variables vs. stand age

Regression analysis was used to determine whether there was an increase in any of the three fuel variables with time-since-fire age, as would be expected if the forest's potential fire intensity was changing with forest age. The regression model used was a log-log transformed model, since this stabilized the variances across the range of ages and best met the linear model assumption of regression. The fuel variables were tested in two sets: (a) all stands; and (b) all stands over 25 yr of age. This examined whether the young stands had a significant influence on the observed trends. Twenty-five years was chosen since this is the approximate age at which crown closure occurred in the observed stands.

RESULTS

Surface fire intensity analysis

1. Relative importance of fuel and weather. - The standardized partial regression coefficients give the relative importance of the weather and fuel variability by estimating the change in intensity (as the proportion of the variation in intensity) produced by a unit change in variance of the weather or fuel variable. The weather variable had a standardized partial regression coefficient of 0.91 and explained 83% of the regression sum of squares ([SS.sub.log[E.sub.v]]/[SS.sub.reg]) (Table 2). The fuel variable's standardized coefficient was 0.39 and this variable explained 15% of the model sum of squares. Therefore variation in surface fire intensity was determined more by the weather variable than the fuel variable.

2. Comparison between large and small area burned years. - Mean fire intensity predictions for all days in large area burned years were different from all days in small area burned years (Table 3, P [less than] 0.001), although there was also significant variation between the means for individual years within each group (P [less than] 0.001). The mean intensities ([+ or -] standard deviations) were 41.4 [+ or -] 17.26 kW/m in small area burned years, and 105.8 [+ or -] 15.16 kW/m in large area burned years. Means for each individual year, with large and small area burned categories indicated, are shown in Fig. 1. The mean intensity predictions do not indicate how fires would burn, since they included low intensity days as well as high days. Instead the means indicated that large area burned years generally had more days when high intensity fires could have occurred than small area burned years.

TABLE 3. Analysis of variance comparison of intensity predictions between large and small area burned years. Intensity was predicted from Rothermel's model for 47 forest stands and 35 yr of summer weather (1 July-31 August) data. Years were nested within the large and small area burned categories to account for variation in individual years. Intensity was log-transformed (log intensity + 1).

Source df MS F P

Group(*)                1     17 728.1   2246.5   [less than]0.001
Year(group)(**)        33       1087.7    137.8   [less than]0.001
Error             101 955          7.8

* Group = Large or small area burned years.

** Year(group) = Years nested within groups.

Crown fire initiation analyses

1. Relative importance of fuel and weather. - Crown fire initiation (determined from the models) was also primarily influenced by weather rather than fuels (Table 4). In the passive crown fire initiation test, standardized slopes were - 1.80 for the weather variable and -0.38 for the passive crown fuel variable. In the first active crown fire initiation test (for only those stands that achieved passive crowning), standardized slopes were -4.51 for the active crown fuel variable and -2.95 for the weather variable. However, most stands that achieved passive crown fire criteria also achieved active crown fire status. For the entire transition from surface to active crown fires, the standardized coefficients were -1.85 for the weather variable and -0.39 for the fuel variable. Thus, the first step (crown ignition) is mainly dependent on weather, and the second step (crown spread) slightly more dependent on fuels. However, the entire transition is mainly dependent on weather.

TABLE 4. Multiple logistic regression results for crown fire initiation vs. crown fuel and weather variables for 47 stands and weather data from 1968 for (a) passive crown fires vs. surface fires, (b) active crown fires vs. passive crown fires only, (c) active crown fires vs. surface or passive crown fires.

Std. coef. Wald [[Chi].sup.2] p

a) Passive Initiation: n = 2914,(*) -2LL(**) [[Chi].sup.2] = 1684, P [less than] 0.0001

Intercept       0.0             893.87          0.0001
[C.sub.p]      -0.38            196.24          0.0001
[E.sub.v]      -1.80            611.20          0.0001

b) Active Initiation: n = 992, -2LL [[Chi].sup.2] = 842, P = 0.0001

Intercept       0.0             106.26          0.0001
[C.sub.a]      -4.51            107.87          0.0001
[E.sub.v]      -2.95             75.50          0.0001

c) Active Initiation: n = 2914, -2LL [[Chi].sub.2] = 1715, P [less than] 0.0001

Intercept       0.0             701.85          0.0001
[C.sub.a]      -0.39            136.80          0.0001
[E.sub.v]      -1.85            606.15          0.0001

* n = number of samples: days when passive or active crown fire initiation was possible, multiplied by the 47 tree stands.

** -2LL = negative two log likelihood.

2. Crown fire initiation: graphical model. - The crown fire initiation graph [ILLUSTRATION FOR FIGURE 2 OMITTED] demonstrated three weather variable ranges: (a) the range for which the models predict only surface fires among the sampled forest stands; (b) the range for which the models predict crown fires for every forest stand; (c) the transitional range between the two. Fig. 2A shows the weather variable ranges that correspond to the 10th and 90th percentiles of the passive crown fuel values; Fig. 2B shows the ranges for the 10th and 90th percentiles of the active crown fuel values. The surface fire weather variable range was [greater than]0 to 1, since the passive crown transition occurred in all stands [greater than]1. The transition weather range was 1 to 10, and the crown fire range was [greater than]10. The ranges were very similar for both active and passive fuel values. The crown fire weather range shows that, once these weather values are achieved, all stands will support crown fires. Thus, fuels play no further role in determining whether or not a stand will crown. Fuels only play a significant role within the transition range. Although not shown on the graphs, the weather value of 0 is important. It defined conditions in which fire intensity would be zero, and thus no fires would be able to occur.

Weather variable frequencies in large and small fire years

For all years combined the weather variable was found to decline exponentially (% frequency = 0.433[([E.sub.v]).sup.-0.566], [r.sup.2] = 0.98, P [less than] 0.01). The negative exponential relationship meant that most days had low weather variable values (from 0-1), while there were very few extreme weather ([greater than]10) days. Large and small area burned years had significantly different weather variable distributions ([ILLUSTRATION FOR FIGURE 3 OMITTED], t = 2.99, P [less than] 0.05). Both large and small area burned years had similar numbers of days of low to moderate weather values, but high to extreme values were progressively more common in large area burned years.

Based on the crown fire initiation graph and the frequency of weather values, for the average summer 18.5% of days would allow no fires ([E.sub.v] = 0), 45.6% of days would allow only surface fires (0 [less than] [E.sub.v] [less than] 1), 35.7% of days would support a mix of surface and crown types (1 [less than or equal to] [E.sub.v] [less than] 10) (days in the transition weather range), while only 0.2% of days would support crown fires in all stands ([E.sub.v] [greater than] 10). Although the fewest days are accounted for in the crown fire weather ([E.sub.v] [greater than] 10) range, this is where the greatest difference between large and small area burned years occurs. Large area burned years had three times as many days in the crown fire range compared to small area burned years. Thus, the weather variable distribution was shifted in large area burned years such that a few days in those years are much more likely to sustain high intensity/crown fires.

Fuel variables vs. stand age

Regression coefficients between each fuel variable and stand age are presented in Table 5 for all stands and for only stands [greater than]25 yr of age. The surface fuel variable ([F.sub.v]) was unrelated to age for stands [greater than]25 yr [ILLUSTRATION FOR FIGURE 4A OMITTED]. However, among all stands there was a positive relationship (Table 5) with age, which explains 18% of the regression sum of squares. The difference between the two results shows that there is an initial increase in the fuel variable, but this does not continue after 25 yr. There is, however, substantial scatter of points in the older stands that reflects the very wide range of surface fuel differences in the stands. This scatter was unrelated to any of the other stand features listed in Table 1.

The passive crown fuel variable was unrelated to age if all ages were examined or if stands [less than]25 yr of age were removed (Table 5; [ILLUSTRATION FOR FIGURE 4B OMITTED]). However, the active crown fuel variable was very low in the stands [less than]25 yr of age [ILLUSTRATION FOR FIGURE C OMITTED], so that a significant relationship to age was seen if these were included, but no relationship [TABULAR DATA FOR TABLE 5 OMITTED] was found in stands [greater than]25 yr. The five stands [less than]25 yr of age differed from the others in having very low crown fuel loads and thus very high critical rates of spread (and low active crown fuel variable values). Thus, passive crowning is unrelated to age, while active crowning is less likely in stands [less than]25 yr of age.

DISCUSSION

Predicted forest fire behavior in the subalpine forests of the southern Canadian Rocky Mountains is more strongly related to weather variation (over time) than to fuel variation (among stands). The weather variable set the lower threshold below which surface fire cannot occur (i.e., when the value of the weather variable was zero, conditions were too wet for burning to occur in the models). The weather also set the upper threshold beyond which only crown fires occur. In comparison, the fuel variable was always [greater than]zero and therefore could not be controlling whether or not fires could burn. Of course, fuels play an important role in fire behavior - this is not disputed. However, the range of fuels among sampled stands in this study was overwhelmed by the variation in weather, such that in a relative sense, fuel moisture and wind speed were the main features controlling variation in fire behavior.

Under extreme weather conditions, the crown fire model predicts that all stands will exhibit crown fires, regardless of the range of fuel conditions. Only in moderate weather will differences in crowning behavior due to fuels be achieved. Furthermore, most large fires occur in years when there are elevated weather variable values. Since fires in those years account for [greater than]99% of the area burned, it is unlikely that fuel variation is strongly influencing fire behavior over the entire forested landscape.

Two basic reasons explain why weather is more important than fuel in fire behavior. First, weather is more strongly associated with the mechanisms of fire behavior than fuel: wind speed strongly affects heat transfer rates by increasing both radiative heat transfer (due to flame tilt) and convective heat transfer; fuel combustibility and heat loss rates depend strongly on fuel moisture content; fuel primarily acts as the burn substrate, providing the heat of combustion for the fire (Rothermel 1972). Second, as shown by the values of the standardized partial regression coefficients for the weather (0.91) and fuel (0.39) variables, the weather variable explained 83% of the variance in intensity while the fuel variable explained only 15%. The explanation for the differences in these standardized coefficients is the fact that fire weather is much more variable over time than fuels are among stands. Surface fuel moisture contents may range from [approximately equal to]5 to [greater than]100% (based on dry mass). Wind speed may range from 0 to [greater than]100 km/h. These changes can occur over very short time scales, hours to weeks, due to very fast drying rates of fine fuels and highly variable wind speeds. Thus, the range of weather effects on fire behavior potential is quite great. High variability and strong ties to the fire behavior mechanisms in the models combine so that the weather variable (and thus variation in intensity due to weather) ranges over four orders of magnitude (i.e., [10.sup.4]).

Fuel variation within stands occurs over such long time scales (years to decades) that it may be ignored at any individual time period (see Van Wagner 1965). The only important variation in fuel loads exists among stands of differing forest type and history (stand age, developmental stage, past disturbance, etc.). However, the amount of variation among stands is still generally low. For the stands in this study, total burnable fuel loads ranged from 0.5 to 3.6 kg/[m.sup.2], which amounts to only a sevenfold difference among stands. These fuel loads are not substantially different from ones in similar Pinus contorta forests (e.g., Kiil 1968, Muraro 1971, Lawson 1973, Fischer 1981, Romme 1982, Brown and Bevins 1986). Often larger biomass components such as the duff layer and large logs are observed to vary over a much greater range than the smaller fuels (Fischer 1981, Romme 1982, Lotan et al. 1985, Spies et al. 1988), but large fuel components play a lesser role in the active spread rate and intensity at the fire front than fine dead or cured fuels (Rothermel 1972, Brown 1983). However, it can be argued that large fuels (particularly the duff layer) do play a much greater role in high intensity (crown) fires. Rothermel (1993) points out that Albini's (1976) burnout model, which could increase reaction intensity 2-3 fold within the 1st min of combustion at the flaming front, is one method to incorporate this effect. This higher reaction intensity could increase the role of larger fuels in surface/transition weather [ILLUSTRATION FOR FIGURE 2 OMITTED] conditions such that high fuel loading stands more easily achieve passive or active crown fires. Since duff is not incorporated in this study, we have a conservative estimate of high-intensity fires.

The lower variability of fuel loads and weaker relationships to intensity in the models combined so that there was less than two orders of magnitude between the highest and lowest fuel variables among stands in this study. Therefore, fuel variation among stands may result in up to two orders of magnitude variation in the fire intensities. Although this may seem like a large amount of variation attributed to fuels, it is comparable to 1% of the variation in intensity due to the weather.

Why are fuels thought to be so important to fire behavior? The first reason is the belief that fire behavior is mainly an attribute of the plant community. Mutch (1970) gave a clear statement of this hypothesis. Thus, as plant communities change due to growth of individuals, senescence, etc., the fire behavior is thought to change concurrently. This appears to be the main reason behind the notion that older forests are more likely to sustain a high-intensity fire than a younger forest (e.g., Heinselman 1973). It has also led to the notion that fire suppression in natural reserves such as Yellowstone National Park has resulted in unnatural fuel accumulation that will (or did) result in more catastrophic fires (e.g., Romme and Despain 1989). These arguments lack a fire behavior mechanism by which community change should affect fires. Fire behavior depends on only certain fuel structural components that change over time, not on all fuels and all structural changes.

The second reason that some ecologists believe that fuels are so important comes from a misunderstanding of the pattern of fire occurrence. Studies must distinguish between fires burning during marginal or extreme fire weather (small and large area burned years). In marginal fire weather, the role of fuel differences among forest stands will appear to be more important. For example, a crown fire burning in a spruce woodland with a low crown base may change to a surface fire in a lodgepole pine forest because of the pines' elevated crown base (e.g., Despain and Sellers 1977). Under these conditions, different fuel types can show differences in fire behavior and area burned. However, the total area burned (the ecologically meaningful variable) by fires in marginal weather will be small because the fires will be more easily extinguished (by rain or by man) and have lower rates of spread than fires during extreme weather conditions. In extreme weather, areas burned will be much larger and the role of fuel much smaller.

CONCLUSION

Rothermel's (1972) fire behavior model, along with empirical fuel and weather data for 35 yr, predicts that variation in surface fire intensity in subalpine forest stands in the southern Canadian Rockies would be primarily determined by weather (fuel moisture and wind speed) rather than by variation due to fuels. Fires that occur in low to moderate weather conditions remain small surface fires or intermittent crown fires, while those in extreme conditions often become large crown fires. In these weather conditions, crown fires will initiate from all surface fires largely independent of fuel conditions or stand type. Extreme weather occurs most commonly in large area burned years compared to small area burned years.

Fire behavior should be directly related to regional patterns of weather that influence fuel moisture contents and wind speeds, rather than ecosystem properties that affect fuel loads and structure. Thus, fire behavior should not vary strongly with stand age or with species composition types. Fire behavior research should be directed toward understanding weather phenomena and their relationship to fire behavior events as well as to regional fire patterns.

The lack of any strong relationships between fuel variables (and thus fire behavior) and stand age, and the overall larger importance of weather than fuel in fire behavior, support the assumption of independence of the hazard of burning to stand age found in the negative exponential model of fire frequencies (Johnson et al. 1990b, Masters 1990, Johnson and Larsen 1991).

The lack of a significant fuel importance is also consistent with the relationship of large area burned years in the southern Canadian Rockies with mid-tropospheric anomalies, i.e., surface-blocking higher pressure systems (Johnson and Wowchuk 1993).

ACKNOWLEDGMENTS

We should like to thank S. Bridge, C. Clark, S. L. Gutsell, L. D. Harder, K. Miyanishi, C. Nash, R. C. Rothermel, C. E. Van Wagner, J. Weir and P. H. Zedler for critical and helpful comments on the research or manuscript. Alberta Forest Service and Parks Canada supplied weather and fire records. This paper was supported by a Natural Sciences and Engineering Research Council operating grant to E. A. Johnson.

LITERATURE CITED

Agee, J. K., and M. H. Huff. 1987. Fuel succession in a western hemlock-Douglas fir forest. Canadian Journal of Forest Research 17:697-704.

Albini, F. A. 1976. Estimating wildfire behavior and effects. U.S. Forest Service General Technical Report INT-30.

Alexander, M. E., B. Janz, and D. Quintillo. 1983. Analysis of extreme wildfire behavior in east-central Alberta: a case study. Pages 38-46 in Proceedings of the Seventh National Conference on Fire and Forest Meteorology, Fort Collins, Colorado. American Meteorological Society, Boston, Massachusetts, USA.

Anderson, H. F. 1968. Sundance fire: an analysis of fire phenomena. U.S. Forest Service Paper INT-56.

ASTM. 1988. Standard test methods for specific gravity of wood and wood-base materials. Pages 357-368 in Annual book of standards. American Society for Testing and Materials, Philadelphia, Pennsylvania, USA.

Bradshaw, L. S., J. E. Deeming, R. E. Burgan, and J. D. Cohen. 1984. The 1978 National Fire Danger Rating System: technical documentation. U.S. Forest Service General Technical Report INT-169.

Brown J. K. 1970. Ratios of surface area to volume for common fine fuels. Forest Science 16:101-105.

-----. 1978. Weight and density of crowns of northern Rocky Mountain conifers. U.S. Forest Service Research Paper INT-197.

-----. 1983. The unnatural fuel buildup issue. Pages 127-128 in Proceedings of the Symposium and Workshop on Wilderness Fire, 15-18 November 1983. U.S. Forest Service General Technical Report INT-182.

Brown, J. K., and C. D. Bevins. 1986. Surface fuel loadings and predicted fire behavior for vegetation types of the northern Rocky Mountains. U.S. Forest Service Research Note INT-358.

Brown, J. K., R. D. Oberheu, and C. M. Johnston. 1982. Handbook for inventorying surface fuels in the interior West. U.S. Forest Service General Technical Report INT-129.

Clark, J. S. 1988. Effect of climate change on fire regimes in northwestern Minnesota. Nature 334:233-235.

Despain, D. G. 1990. Yellowstone vegetation: consequences of environment in a natural setting. Roberts Reinhart, Boulder, Colorado, USA.

Despain, D. G., and R. E. Sellers. 1977. Natural fires in Yellowstone National Park. Western Wildlands 4:20-24.

Finklin, A. I. 1973. Meteorological factors in the Sundance fire run. U.S. Forest Service General Technical Report INT-6.

Fire Danger Group. 1992. Development and structure of the Canadian Forest Fire Behavior Prediction System. Forestry Canada Information Report ST-X-3.

Fischer, W. C. 1981. Photo guide for appraising downed woody fuels in Montana forests: lodgepole pine and engelmann spruce-subalpine fir cover types. U.S. Forest Service General Technical Report INT-98.

Flannigan, M. D., and J. B. Harrington. 1986. A study of the relation of meteorological variables to monthly provincial area burned by wildfire in Canada (1953-1980). Journal of Applied Meteorology 27:441-452.

Fryer. G. I., and E. A. Johnson. 1988. Reconstructing fire behavior and effects in a subalpine forest. Journal of Applied Ecology 25:1063-1072.

Habeck, J. R., and R. W. Mutch. 1973. Fire-dependent forests in the northern Rocky Mountains. Quaternary Research 3: 408-424.

Harrington, J. B., and M. D. Flannigan. 1987. Drought persistence at forested Canadian stations. Pages 204-206 in Proceedings of the Ninth Conference on Fire and Forest Meteorology, San Diego, California. American Meteorological Society, Boston, Massachusetts, USA.

Harrington, J. B., M. D. Flannigan, and C. E. Van Wagner. 1983. The study of the relation of components of the Fire Weather Index to monthly provincial area burned by wildfire in Canada 1953-1980. Petawawa National Forestry Institute, Canadian Forestry Service, Department of Environment Information Report PI-X-25.

Heinselman, M. L. 1973. Fire in the virgin forests of the Boundary Waters Canoe Area, Minnesota. Quaternary Research 3:329-382.

Johnson, A. F., P. M. Woodard, and S. J. Titus. 1990a. Lodgepole pine and white spruce crown fuel weights predicted from diameter at breast height. Forestry Chronicle 66:596-599.

Johnson, E. A., and G. I. Fryer. 1989. Population dynamics in lodgepole pine-Engelmann spruce forests. Ecology 70: 1335-1345.

Johnson, E. A., G. I. Fryer, and M. J. Heathcott. 1990b. The influence of man and climate on frequency of fire in the interior wet belt forest, British Columbia. Journal of Ecology 78:403-412.

Johnson, E. A., and C. P. S. Larsen. 1991. Climatically induced change in fire frequency in the southern Canadian Rockies. Ecology 72:194-201.

Johnson, E. A., and C. E. Van Wagner. 1985. The theory and use of two fire history models. Canadian Journal of Forest Research 15:214-220.

Johnson, E. A., and D. R. Wowchuk. 1993. Wildfires in the southern Canadian Rocky Mountains and their relationship to mid-tropospheric anomalies. Canadian Journal of Forest Research 23:1213-1222.

Keane, R. E., S. F. Arno, and J. K. Brown. 1990. Simulating cumulative fire effects in ponderosa pine/Douglas-fir forests. Ecology 71:189-203.

Kiil, A. D. 1968. Weight of the fuel complex in 70-year-old lodgepole pine stands of different densities. Ministry of Forestry and Rural Development, Forestry Branch, Departmental Publication 1228.

Lawson, B. D. 1973. Fire behavior in lodgepole pine stands related to the Canadian Fire Weather Index. Canadian Forestry Service, Department of Environment Information Report BC-X-76.

Lotan, J. E., J. K. Brown, and L. F. Neuenschwander. 1985. Role of fire in lodgepole pine forests. Pages 133-152 in D. M. Baumgartner. G. Krebill, J. T. Arnott, and G. F. Weetman, editors. Lodgepole pine - the species and its management. Symposium Proceedings. Washington State University, Pullman, Washington, USA.

Masters, A. M. 1990. Changes in forest fire frequency in Kootenay National Park, Canadian Rockies. Canadian Journal of Botany 68:1763-1767.

McRae, D. G., M. E. Alexander, and B. J. Stocks. 1979. Measurement and description of fuels and fire behavior on prescribed burns: a handbook. Canadian Forestry Service Report O-X-287.

Muraro, S. J. 1971. The lodgepole pine fuel complex. Canadian Forestry Service, Department of Environment Information Report BC-X-53.

Mutch, R. W. 1970. Wildland fires and ecosystems - a hypothesis. Ecology 51:1046-1051.

Newark, M. J. 1975. The relationship between forest tire occurrence and 500 millibar longwave ridging. Atmosphere 13:26-33.

Pech, G. 1989. A model to predict the moisture content of reindeer lichen. Forest Science 35:1014-1028.

Romme, W. H. 1982. Fire and landscape diversity in sub-alpine forests of Yellowstone National Park. Ecological Monographs 52:199-221.

Romme, W. H., and D. G. Despain. 1989. Historic perspective on the Yellowstone fires of 1988: a reconstruction of prehistoric fire history reveals that comparable fires occurred in the early 1700's. BioScience 39:695-699.

Romme, W. H., and D. H. Knight. 1981. Fire frequency and subalpine forest succession along a topographic gradient in Wyoming. Ecology 62:319-326.

Rothermel, R. C. 1972. A mathematical model for predicting fire spread in wildland fuels. U.S. Forest Service Research Paper INT-115.

-----. 1991. Predicting behavior and size of crown fires in the Northern Rocky Mountains. U.S. Forest Service Research Paper INT-438.

-----. 1993. Some fire behavior modelling concepts for fire management systems. Pages 71-81 in Proceedings of the Twentieth National Conference on Fire and Forest Meteorology, Jekyll Island, Georgia. American Meteorological Society, Boston, Massachusetts, USA.

Schroeder, M. J., M. Glovinsky, V. H. Hendricks, et al. 1964. Synoptic weather types associated with critical fire weather. U.S. Forest Service Pacific Southwest Forest and Range Experiment Station, Berkeley, California, USA.

Simard, A. J., D. A. Haines, R. W. Blank, and J. S. Frost. 1983. The Mack Lake Fire. U.S. Forest Service General Technical Report NC-83.

Singh, T. 1982. Biomass equations for ten major tree species of the Prairie Provinces, Environment Canada, Canadian Forestry Service, Information Report NOR-X-242.

Snedecor, G. W., and W. G. Cochran. 1967. Statistical methods. Sixth edition. Iowa State University Press. Ames, Iowa, USA.

Spies, T. A., J. F. Franklin, and T. B. Thomas. 1988. Coarse woody debris in Douglas-fir forests of western Oregon and Washington. Ecology 69:1689-1702.

Stocks, B. J., and M. D. Flannigan. 1987. Analysis of the behavior and associated weather for a 1986 northeastern Ontario wildfire: Red Lake #7. Pages 94-100 in Proceedings of the Ninth Conference on Fire and Forest Meteorology, San Diego, California. American Meteorological Society, Boston, Massachusetts, USA.

Street, R. B. 1985. Drought and synoptic fire climatology of the boreal forest region of the Canadian prairie provinces. In Proceedings of the Eighth National Conference on Fire and Forest Meteorology, Detroit, Michigan. American Meteorological Society, Boston, Massachusetts, USA.

Swetnam, T. W. 1993. Fire history and climate in Giant Sequoia groves. Science 262:885-889.

Swetnam, T. W., and J. L. Betancourt. 1990. Fire-Southern Oscillation relations in the southwestern United States. Science 249:1017-1020.

Sylvester, T. W., and R. W. Wein. 1981. Fuel characteristics of Arctic plant species and simulated plant community flammability from Rothermel's model. Canadian Journal of Botany 59:898-907.

Van Wagner, C. E. 1965. Describing forest fires - old ways and new. Forestry Chronicle 41:301-305.

-----. 1968. The line intercept method in forest fuel sampling. Forest Science 14:20-26.

-----. 1977. Conditions for the start and spread of crown fires. Canadian Journal of Forest Research 7:23-34.

-----. 1980. Fire behavior in northern conifer forests and shrublands. Pages 45-80 in R. W. Wein, and D. A. MacLean, editors. Fire in northern circumpolar ecosystems. Academic Press, New York, New York, USA.

-----. 1987. Development and structure of the Canadian Forest Fire Weather Index System. Canadian Forestry Service Technical Report 35.

Zar, J. H. 1984. Biostatistical analysis. Second edition. Prentice-Hall, Englewood Cliffs, New Jersey, USA.

APPENDIX

Methods and equations for calculating surface fuels (downed wood fuels, shrubs, herbs, litter, and moss) and crown fuels, together with the fuel characteristics used in the surface fire intensity and crown fire initiation models. Fuel load calculations for downed wood inventoried by the line intersect method was done using the following equation (Van Wagner 1968).

W = [Rho][[Pi].sup.2] [Sigma] [d.sup.2]/8L,

where

W = fuel load (kg/[m.sup.2])

[Rho] = mass density (kg/[m.sup.3])

d = diameter (m)

L = transect length (m)

The equation required wood mass density and quadratic mean diameters for each size class and species of wood. Wood samples were collected throughout Kananaskis and measured for diameter and mass density (cf. ASTM 1988). Diameters were squared, means obtained, and then transformed back to linear values. Mean diameters, mass densities, and sample length over which the size class is measured are presented in Table A1. Species sampled were pine (Pinus contorta), spruce (Picea engelmannii), fir (Abies lasiocarpa), aspen (Populus tremuloides).

Fuel load calculations for shrubs were made by estimating biomass from the following empirical equations calculated from samples collected in Kananaskis. All equations predict dry mass in grams from diameter in centimetres (Table A2). Fuel load was then determined by dividing the mass by the quadrat area, which was, in this inventory, 1 [m.sup.2]. For live shrubs, log(mass) = log(a) + b log(diameter); dead branch mass was transformed as log(mass + 1).

Fuel load calculations for herbaceous classes used in the inventory predicted fuel load directly from cover, height, and percent dead. All equations were of the form: Load = a + b1(% cover) + b2(height[cm]) + b3(cover-height) + b4(% dead). Note that the maximum % dead value permitted for prostrate shrubs was 20.

Fuel load calculations for litter and moss were made from depth, based on samples collected throughout the Kananaskis Valley. The litter depth was taken to the fermentation layer. Moss depth was taken to the extent of living material.

Wood fuel loads were divided among the 1, 10, and 100 h time-lag size classes (0-0.63 cm, 0.64-2.54 cm. and 2.557.62 cm diameter, cf. Bradshaw et al. 1984) based on conversion factors determined from wood diameters collected in Kananaskis (Table 4a). Dead shrub wood was also divided into the three time-lag size classes based on an expected distribution of stem sizes (Table 4b) for stems of a certain basal diameter (cf. Brown et al. 1982). Each fuel class was then assigned a characteristic surface-area-to-volume ratio (cf. Sylvester and Wein 1981, Brown and Bevins 1986), and a fuel depth based on field estimates (Table 5).

Crown foliage fuel loads were predicted from previously published equations. As a number of choices were available, the selection criteria for a useable equation was that it covered the appropriate range of tree sizes, provided a good fit to the data used to make the equation, and was applicable to the study area, Table A7 lists the equations chosen and their authors. All equations predict foliage load from dbh, with the exception of the equations for small trees, which are based on height.

[TABULAR DATA FOR TABLE A1 OMITTED]

[TABULAR DATA FOR TABLE A2 OMITTED]

[TABULAR DATA FOR TABLE A3 OMITTED]

TABLE A4. Litter and moss equations for Kananaskis, Alberta which predict fuel load (kg/[m.sup.2]) from D, depth (cm).

Type(*)           Equation(2)        MSE     N   [r.sup.2]

TABLE A6. Standardized values of surface area to volume ratio ([cm.sup.-1]) and fuel layer depth for each surface fuel category.

                   Surface                     Source
Fuel             area/volume     Depth      (for surface
component       ([cm.sup.-1])     (cm)      area/volume)(*)

COPYRIGHT 1995 Ecological Society of America
COPYRIGHT 2004 Gale Group