of Forest Fuels by Combustible Gas Analysis

Transcription

1 Forest Sci., Vol. 28, No. 2, 1982, pp Copyright 1982, by the Society of American Foresters Characterization of the Thermal Properties of Forest Fuels by Combustible Gas Analysis RONALD A. SUSOTT ABSTRACT. The thermal decompositions of forty-three typical forest fuels have been characterized up to 500øC. The fuels studied include foliage, wood, small stems, and bark. Evolved gas analysis gave quantitative measurements of the oxygen stoichiometry for combustion of volatile pyrolysis products. Oxygen consumption by the volatiles is highly correlated to their calculated heats of combustion. Char heats of combustion were also determined and found to be very similar for the fuels studied. These data were used to partition total heat of combustion into flaming and glowing components. Addition of combustible gas analysis to other thermal analysis techniques provides a more complete comparison of thermal behavior over a wide temperature range. FOR- EST SCI. 28: ADDITIONAL KEY WORDS. Thermal analysis, pyrolysis, charring, ash contents, heat of combustion, biomass, flammability. FIRE BEHAVIOR PREDICTIONS are required for managing fuels in our forests and controlling wildland fires. Such predictions can be calculated from a mathematical model developed by Rothermel (1972) and modified by Albini (1976). Although this model was based on results from a limited variety of laboratory test fuels, predictions are acceptable in numerous field situations (Sneeuwjagt and Frandsen 1977, Andrews 1980). To improve and extend application of the model, we need a better understanding of the combustion properties of a wider variety of common forest fuels. Combustion properties are controlled in part by the detailed thermal reactions of the chemical components of the fuels. Unfortunately, natural fuels are chemically complex and variable, and their decomposition does not lend itself to simple mechanistic or kinetic description. This paper concentrates on characterization of the overall decomposition and volatile generation process for these fuels. Other studies have related fuel chemistry to flammability. The mineral and extractive components are often cited as possible cause for differences in fire behavior. Philpot and Mutch (1971), Vines (1975), and Montgomery (1976) discussed the importance of volatile extractives to fuel flammability. Components that have a high heat of combustion and vaporize or decompose at low temperatures are expected to increase flammability. Shafizadeh and others (1977) and Susott (1980a) showed, however, that not all extractives had those properties. Less volatile material can be extracted as well. Also, the insoluble fractions have thermal properties that are similar to the extractives. Most of the cited studies concluded that a more complete analysis of all components was necessary before relative flammability could be accurately predicted. In this report, quantitative combustible gas analysis (Susott and others 1979) is applied to a variety of common forest fuels. The total heat of combustion obtained The author is Research Chemist, USDA Forest Service, Intermountain Forest and Range Experiment Station, Ogden, Utah, located at Northern Forest Fire Laboratory, Missoula, Mont. Manuscript received 21 May / FoREsT SCIENCE

2 by bomb calorimetry is divided into portions for the volatiles and char by the method described by Susott and others (1975). Gas analysis and calorimetric methods are compared. Fuels are grouped by similarities in their thermal behavior. METHODS Samples.--The forest fuels compared in this study are listed in Table 1. Samples were selected to provide data on a variety of plant components, including foliage, wood, small stems, and bark. Green foliage and stem samples (branches less than 0.25 inch diameter with foliage removed) were frozen when collected and then freeze-dried to less than 10 percent moisture content. Wood, bark, and litter samples were air-dried at room temperature. All samples were ground to pass through a 20-mesh screen. Thermogravimetric analysis of these samples has been reported elsewhere (Susott 1980b). Evolved Gas Analysis (EGA).--Combustible products from thermal decomposition were quantitatively analyzed by the method described by Susott and others (1979) and Susott (1980a). In this method, samples were heated at a constant rate while measuring volatile reaction products through a gas phase titration with oxygen. In a typical experiment, 5 mg of ground sample (smaller than 20-mesh) was heated in nitrogen to 500øC at 20øC/min. Vaporized products were swept into a high-temperature catalytic reactor and burned. The necessary oxygen was generated in an electrolytic cell and a stoichiometric oxygen-to-fuel mass ratio calculated from the cell current. The EGA output has units of mass of oxygen consumed per second normalized by dividing by the dry weight of the sample. Bomb Calorimetry.--Standard methods (ASTM D , 1967) were used to obtain heats of combustion for fuels and their chars. Chars were prepared by heating 100 mg of fuel according to the method used for EGA (above). The heat of combustion for volatiles can be estimated from the conservation of energy equation used by Susott and others (1975) and Rothermel (1976). The general features of the overall pyrolysis reaction are given by: Fuel heart volatiles + char (1) For comparison purposes, the enthalpy of reaction can be assumed to be negligible and the heat of combustion for volatiles can be calculated from: AHøeomb (volatiles) = AHøeomb (fuel) - AHøeomb (char) x fraction char (2) In equation (2), the fraction of char formed by pyrolysis and its enthalpy of combustion are easily measured, and subtraction from the enthalpy of combustion of the original fuel gives a good estimate of the combustion enthalpy for volatile products. The appropriate char fractions and their ash contents are easily measured by thermogravimetric analysis. The char fraction needed in equation (2) is an intrinsic fuel property and is nearly independent of heating rate (Susott 1980b). The heat of combustion is directly related to the combustion stoichiometry. A mean value for a variety of organic compounds is 14.0 _+ 0.6 MJ liberated per kilogram oxygen required for complete combustion (Susott and others 1979). The heat of combustion for volatiles from forest fuels should have a similar relation to oxygen required. This relationship can be evaluated from two of the measurements in this report--the heats of combustion calculated in equation (2), and integrated EGA data. Both measurements apply to the volatiles released by heating to 500øC. Moisture Contents.---All calculations in this report are on a moisture-free basis. Fuel moistures were determined by a standard Karl Fischer method (ASTM E VOLUME 28, NUMBER 2, 1982 / 405

3 Table 1. Forest fuel descriptions. Number Common name Species Part 1 Chamise Adenostoma fasciculatum H. & A. Foliage 2 Greenleaf manzanita Arctostaphylos glandulosa Eastw. Foliage 3 Big sagebrush Arternisia tridentata Nutt. Foliage 4 Gallberry Ilex glabra (L.) Gray Foliage 5 Utah juniper Juniperus osteosperma (Torr.) Little Foliage 6 Lodgepole pine Pinus contorta Dougl. Foliage 7 Pinyon pine Pinus edulis Engelm. Foliage 8 Slash pine Pinus elliottii Engelm. Foliage 9 Western white pine Pinus monticola Dougl. Foliage, dead 10 Ponderosa pine AZ 78 Pinus ponderosa Laws. Foliage 11 Ponderosa pine CA 77 Pinus ponderosa Laws. Foliage 12 Ponderosa pine MT 77 Pinus ponderosa Laws. Foliage 13 Ponderosa pine MT 78 Pinus ponderosa Laws. Foliage 14 Ponderosa pine Pinus ponderosa Laws. Foliage, dead 15 Quaking aspen Populus tremuloides Michx. Foliage 16 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Foliage 17 Black oak Quercus velutina Lam. Foliage, dead 18 White fir Abies concolor (Gord. & Glend.) Wood Linl. 19 Grand fir Abies grandis (Dougl.) Lindl. Wood 20 Excelsior Populus spp. L. Wood, dead 21 Larch Larix occidentalis Nutt. Wood 22 Larch Larix occidentalis Nutt. Wood, lumber 23 Ponderosa pine Pinus ponderosa Laws. Wood 24 Ponderosa pine Pinus ponderosa Laws. Heartwood lumber 25 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Wood 26 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Wood, lumber 27 Big sagebrush Arternisia tridentata Nutt. Stems 28 Utah juniper Juniperus osteosperma (Torr.) Little Stems 29 Pinyon pine Pinus edulis Engelm. Stems 30 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Stems 31 Utah juniper Juniperus osteosperma (Torr.) Little Bark 32 Larch Larix occidentalis Nutt. Bark 33 Ponderosa pine Pinus ponderosa Laws. Bark 34 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Bark 35 White fir Abies concolor (Gord. & Glend.) Wood, rotten Linl. 36 Cheatgrass Bromus tectorurn L. Aerial plant, cured 37 Idaho fescue Festuca idahoensis Elmer Aerial plant, cured 38 Douglas-fir Pseudotsuga menzeisii (Mirb.) Franco Wood, rotten 39 Bracken fern Pteridium aquilinum (L.) Kuhn Fronds, cured 40 Saw-palmetto Serenoa repens (Bartr.) Small Fronds 41 Tundra, interior Hylocomium splendens Hedw. and Top layer of black Pleurozium schreberi (Brid.) Mitt. spruce understory 42 Duff -- "F" layer, dead 43 Cellulose T, 1966). A thermogravimetric analysis method (Susott 1980a, 1980b) was used to convert EGA and char fractions to the dry basis. Ash Contents.--Standard ashing techniques (ASTM D , 1969) were used to measure total ash content. Sample size was increased to 5 or 10 grams to allow subsequent analysis for silica in the ash as needed for another study. 406 / FOREST SCIENCE

4 _7F I I I I I I Piny0n Pine Quaking Aspen 15 F Greenleaf Manzanita 2F Big Sagebrush 3F >- x o TEMPERATURE, øc FIGURE 1. Typical evolved gas analysis (EGA) curves for foliage (F) samples. Numbers in the upper left-hand corners refer to sample description in Table 1. RESULTS AND DISCUSSION QUALITATIVE CHARACTERIZATION OF FUELS BY THEIR EGA CURVES Figure 1 shows EGA curves for typical foliage samples. Curves for sound and rotten wood are given in Figure 2, and stems and bark are shown in Figure 3. (These curves are representative of numerous other samples, as will be discussed later.) Each curve is determined as an average of three replicate runs, presented on an ash-free, dry-weight basis. The ash-free basis allows the thermal behavior of the organic components to be compared without the complication of a variable ash content. Foliage.--The foliage samples in Figure 1 are characterized by numerous, overlapping decompositions. Combustible material is vaporized at fairly high rates from 200 ø to 500øC. Three broad temperature regions 100øC wide can be identified, where components with similar stability decompose. Definite peaks or shoulders are shown in these temperature regions for all foliage samples. Greenleaf manzanita (number 2 in Table 1) in Figure 1 clearly shows the three regions. The region between 200 ø and 300øC is dominated by extractable components (Susott 1980a). The carbohydrate polymers cellulose and hemicellulose dominate VOLUME 28, NUMBER 2, 1982 / 407

5 4 Douglas- Fir, rotten wood Cellulose Ponderosa Pine Lumber White Fir, rotten wood - 24 W 35 0 Excelsior Larch Lumber 20 W 22 W TEMPERATURE, øc F GVRE 2. Typical evolved gas analysis (EGA) curves for sound (W) and rotten (O) wood samples. Numbers in the upper left-hand corners refer to sample descriptions in Table 1. Note that the scale for cellulose is attenuated by a factor of 2. the region between 300 ø and 400øC (Shafizadeh and McGinnis 1971, Philpot 1971). The peak above 400øC, centered at 425øC, is due to more stable components. It is tentatively assigned to lignin (Tang 1967, Browne 1958), or lignin-like components composed of aromatic polymers (Pearce and others 1978). All the pine foliage samples (6 to 14 in Table 1) are qualitatively very similar to pinyon pine (7). Dead or cast needles show very little change from fresh, green needles. Douglas-fir needles (16) and leaves from black oak (17) are also very similar to the pines. Big sagebrush (3) and Utah juniper (5) are the only samples with a sizeable amount of material vaporized below 200øC. This material is mostly terpene hydrocarbons (Susott 1980a) with low boiling point. These samples also have a strong, characteristic odor. Utah juniper (5), gallberry (4), and chamise (1) are similar to manzanita (2) in that the volatile generation rate in the 300 ø to 400øC 408 / FOREST SCIENCE

6 Utah Juniper 31 B 33 B Ponderosa Pine Big Sagebrush Douglas- Fir 27 S 30S TEMPERATURE, øc FIGURE 3. Typical evolved gas analysis (EGA) curves for stem (S) and bark (B) samples. Numbers in the upper left-hand comers refer to sample descriptions in Table 1. region is less dominant. These latter fuels have a fairly constant rate over a wide temperature range. Wood.--The wood samples in Figure 2 are characterized by strong peaks between 300 ø and 400øC. These peaks result from high cellulose and hemicellulose contents. The main peak is due to cellulose, as shown by comparing to the purified cellulose sample (43). Compared to foliage, the decomposition is limited to a much narrower temperature range. The rate of combustible gas vaporization is low below 300øC and also above 400øC. Ponderosa pine wood (23 and 24), however, shows a small peak around 200øC due to pitch or resinous components. Larch wood (21 and 22) also has a minor peak in ths region. The EGA curves for wood, samples 18 to 26, lack an obvious lignin peak. This is consistent with reports that charting, at the expense of combustible volatiles, is the dominant pyrolysis reaction for wood-lignin (Tang 1967, Philpot 1970). Rotten Douglas-fir wood (38), which is more than 50 percent lignin (Susott and others 1975), indicates a wood-lignin decomposition in the 300 ø to 400øC region. Therefore, lignin and cellulose decompositions are not separated in wood. It is possible, however, that lignin components from other fuels such as foliage, stems, and bark, can have widely different thermal stability. VOLUME 28, NUMBER 2, 1982 / 409

7 The two rotten wood samples in Figure 2 had dissimilar curves. Rotten white fir wood (35) has a decomposition rate much like the sound wood samples. The peak rate for rotten Douglas-fir wood (38) was less than half that for the rotten white fir wood. Other measurements (discussed later) also reflect these differences. Microorganisms can attack the cell wall polysaccharides or the lignin of wood, or both, depending on site conditions (Stamm 1964). Blankenhorn and others (1980) discussed the action of brown- and white-rot fungi on aspen wood and also found large variations in the calorimetric properties of the decayed wood. The Douglas-fir sample is probably a good example of brown rot while the white fir sample is more typical of white rot. Because of the numerous possibilities for site conditions, no general statement can be made about the effect of rotting on thermal behavior. Stems and Bark.---The stem samples in Figure 3 have characteristics of both foliage and wood. Big sagebrush stems (27) have a prominent cellulose peak and very little volatile loss above 400øC. This is similar to the wood samples. Utah juniper stems (28), cheatgrass (36), and Idaho fescue (37) are similar to big sagebrush stems. On the other hand, Douglas-fir stems (30) have volatile loss that is similar to most foliage samples. Pinyon pine stems (29) and larch bark (32) are similar to the Douglas-fir stems (30) in Figure 3. The bark sample curves (31 and 33) in Figure 3 illustrate the wide range in thermal behavior for barks. In the 300 ø to 400øC region, ponderosa pine bark (33) decomposed at the lowest rate of all samples studied. The rate for juniper bark (31) was nearly twice as high, very similar to juniper stems (28). The above discussions illustrate several important points relating to the thermal characteristics of forest fuels in general. First, similar thermal behavior is exhibited by widely different fuels. This is not too surprising because the main structural elements, such as the cell wall cellulose and lignins, are common to all plant biomass. The relative proportions of the component classes, such as extractive, cellulosic, and lignin, appear to explain much of the thermal behavior observed in Figures 1, 2, and 3. Detailed description of the molecular structure of all components is not necessary for a qualitative explanation of the curve features. Second, the use of broad, 100øC temperature regions should be adequate for relating thermal behavior to fire behavior. Further resolution is not justified, considering other uncertainties inherent to fire behavior modeling. Nevertheless, more research is needed on the relative flammability characteristics of these fuels before the importance of each region can be assessed. Other factors may be equally important to flammability, such as the maximum rate of volatilization and properties of char. Finally, most of these fuels do not have major amounts of easily vaporized components below 200øC. Two foliage samples, big sagebrush (3) and Utah juniper (5), had significantly more volatiles in this region than other fuels. These materials could cause abnormally high flammability. Other fuels may be similar and could pose severe fire problems. QUANTITATIVE DATA FROM THE EGA CURVES The integration of the EGA curves over time provides a quantitative value for the loss of combustible volatiles up to any temperature of interest. Values for this integral at 100øC increments are listed in Table 2. The major differences in these EGA integrals are discussed in a later section. The ash contents used for conversion to an ash-free basis are also given in Table 2. The stoichiometric amounts of oxygen required for combustion will be discussed in relation to heats of combustion in following sections. 410 / FOREST SCIENCE

8 Table 2. Total combustible gas loss (integrated EGA curves)for forest fuels at 100øC increments. Component and sample number Total oxygen required for combustion b Ash (percent) 200øC 300øC 400øC 500øC Foliage Wood Stems Bark Other I l l a Numbers refer to samples described in Table 1. b Amount of oxygen required for complete combustion of all volatile material lost up to the indicated temperature. Units are grams oxygen per gram of ash-free dry sample. VOLUME 28, NUMBER 2, 1982 / 411

9 CALORIMETRIC MEASUREMENTS Heats of combustion for the fuel samples and chars at 500øC are given in Table 3. The percentage of char yield is also given in Table 3. Table 3 also lists the heat of combustion for char and volatiles per gram of original fuel. The latter were calculated from equation (2). Fuels. The standard heats of combustion (in Table 3) ranged from 17.4 MJ/kg for cellulose (43) to 24.0 MJ/kg for Douglas-fir bark (34). The mean of all samples was 21.4 MJ/kg, with a standard deviation of 1.4 MJ/kg. As pointed out by Rothermel (1976), current fire models use total heats of combustion to calculate reaction intensity in experimental fires. Unfortunately, no distinction is made between the energy available from combustible volatiles and energy available from solid char. The char is expected to burn by glowing rather than by flaming combustion. The relatively minor variations in the total heat of combustion do not explain the wide variations in fire behavior observed for different fuels. As an example, the flaming combustion of ponderosa pine needle litter (14) is much more vigorous than the slow glowing combustion of rotten Douglas-fir wood (38). Table 3 shows little difference between their heats of combustion (22.91 and MJ/kg, respectively). The EGA data in Table 2, however, provide an explanation for this behavior. The pine needle litter generates 1.45 times more combustible volatiles at 500øC than the rotten Douglas-fir wood (1.000 versus g O2/g fuel). On the other hand, the additional char formed by the rotten wood is necessary for smoldering combustion (Ohlemiller and others 1979, Shafizadeh and Bradbury 1979). From this brief example it is apparent that EGA data can be combined with other thermochemical data to help characterize expected fire behavior. Char Yields and Heats of Combustion.---Table 3 indicates the wide variation in char yield from forest fuels. Values ranged from 47 percent for ponderosa pine bark (33) down to 15 percent for excelsior (20), and a low value of 4.8 percent for cellulose (43). The charting reactions are much more dominant in forest fuels than they are for pure cellulose. The samples of wood, and other fuels high in cellulose content, have char yields 8 to 10 percent lower than foliage and other fuel types. Note that cheatgrass (36) and rotten white fir wood (35) have char yields similar to wood samples. The EGA curves for these fuels were also similar to wood. Similarly, Utah juniper bark (31) has a much lower char yield than other bark samples and this difference is reflected in the EGA curves. The high char yield of white pine litter (9) compared to ponderosa pine litter (14) may explain the difference in fire behavior observed for these otherwise similar fuels. Flame sizes for ponderosa pine were four times higher than white pine for similar burning conditions (Rothermel and Anderson 1966). The low yield of combustible volatiles from the latter fuel can be combined with physical characteristics, such as surface area to volume ratio and bulk density, to account for the decreased flame size. Char heats of combustion in Table 3 ranged from 29.9 MJ/kg for Douglas-fir bark (34) to 33.6 MJ/kg for gallberry foliage (4). The mean for all char samples was 32.0 MJ/kg, much higher than the 21.4 MJ/kg mean for the original fuels. The standard deviation was only +0.9 MJ/kg. Rothermel (1976) used a constant value of 29.2 m 1.7 MJ/kg for chars formed at 400øC. As char formation temperature increases, char heat of combustion also tends to increase and becomes equivalent for all fuels. For many thermodynamic calculations, the char heat of combustion can be considered constant for all forest fuels. Char yield is also nearly constant for a given fuel and yield does not depend strongly on heating rate (Susott 1980b). The calculation of heats of combustion for volatiles, by equation (2), is simplified by using the constant char properties cited above. 412 / FOREST SCIENCE

10 TABLE 3. Heats of combustion for fuels, chars, and volatiles. Component and Heat of combustion, MJ/kg b Distribution sample Char yield number (percen0 Fuel Char Volatiles ½ at 500øC Char Foliage Wood Stems Bark Other Numbers refer to samples described in Table 1. All values are on an ash-free, dry basis. Heat of combustion calculated per kilogram of original fuel. VOLUME 28, NUMBER 2, 1982 / 413

11 Volatiles.--The high heat of combustion for char is balanced by a lower value for volatiles, relative to the original forest fuel. A mean value of MJ/ kg (based on weight of volatiles) can be calculated from data in Table 3. Although volatiles such as monoterpenes have a heat of combustion as high as 45 MJ/kg, there are also sizeable amounts of water and carbon dioxide in the volatile mixture. The latter are products of dehydration and decarboxylation reactions leading to charting (Shafizadeh 1968). Clearly, the yield of volatiles must be combined with their heat of combustion in order to develop relations to fire behavior. The heats available to flaming combustion of volatiles are given in Table 3. These heats and the associated amount remaining in char were calculated from equation (2). The mean value for volatiles was 12.7 MJ/kg (based on the weight of original fuel), with a standard deviation of MJ/kg. Thus, on the average, 60 percent of the total heat of combustion is available to flaming. The other 40 percent remains as char and burns by glowing combustion. RELATIONSHIP BETWEEN EGA INTEGRALS AND CALORIMETRIC MEASUREMENTS Figure 4 shows the linear regression between the heat available from volatiles and the integrated data from the EGA curves (in Table 2). Both parameters include all volatiles generated up to 500øC. The correlation coefficient was and the coefficient of determination, r 2, was Clearly, the heat of combus EGA INTEGRAL AT 500øC, G 02/6 FUEL FIGURE 4. Relationship between the heat of combustion and oxygen required for combustion of pyrolysis volatiles. (/ ) foliage, ( ') wood, ([ ) stems, (I) bark, ( ) other. 414 / FOREST SCIENCE

12 tion for volatiles from forest fuels can be estimated from EGA data. An advantage of EGA over standard bomb calorimetry methods is that data are available over the entire temperature range. Also, the rate of heat release can be calculated directly--a nearly impossible calculation for all but extremely precise calorimetry. The value of the EGA curves for qualitative descriptions of the pyrolysis reactions has already been discussed. The slope of the least squares line in Figure 4 was 14.6 MJ/kg oxygen. The intercept was not significantly different from zero at a 0.05 probability level. The constant slope can be used to convert the integral EGA data in Table 2 to heats of combustion for volatiles. Similarly, it converts the EGA curves in Figures 1, 2, and 3 to a potential rate of heat release, a necessary relationship for the modeling of combustion processes. The value of 14.6 MJ/kg found for forest fuel volatiles is in good agreement with the 14.0 _+ 0.5 MJ/kg value calculated for known organic compounds (Susott and others 1979). The slight difference between the two values may arise from simplifying assumptions of equation (2). In particular, char formed from these fuels is very reactive and gains weight when exposed to room air. Much of the weight gain is due to moisture adsorption and is easily corrected by thermogravimetric analysis of the char. Recent studies (Bradbury and Shafizadeh 1980) have shown, however, that highly exothermic oxygen chemisorption could account for some weight gain. This directly affects the heat balance of equation (2), because the measured heat of combustion of char is too low. In other words, some char is oxidized before its heat of combustion is measured. The value calculated for volatiles would then be too high. More research is needed on the chemisorption of forest fuel char to support this hypothesis. GROUPING FOREST FUELS BY THERMAL DECOMPOSITION PROPERTIES The analysis of combustible gases provides very detailed data on the thermal behavior of diverse forest fuels. A simple grouping of fuels with common characteristics would be useful for fire modeling purposes in order to predict similarities in the contributions of fuel chemistry to fire behavior. At the present time, the relationships between pyrolysis characteristics and fire behavior are quite speculative. The detailed relationships of low-temperature volatiles, high-temperature volatiles, and char formation to fire intensity are not known. Future research that compares the fire behavior of diverse forest fuels should determine these relationships. A convenient method for grouping fuels of similar thermal behavior is shown in Figure 5. The heat available from combustion of volatiles at 500øC is plotted against the total heat of combustion of original fuel. Both parameters are sensitive to chemical composition, but the ordinate is sensitive to the overall thermal behavior. The integrated EGA value can also be used for the ordinate because it is proportional to the heat of combustion of volatiles. Both plots were very similar so only one is shown. There is no clear relationship between the two variables in Figure 5 when all fuel samples are considered as a single group. The heat available from volatiles cannot, in general, be predicted from the total heat of combustion of the fuels. It is clear, however, that fuels with similar properties do group together. All the wood samples (numbers 18 to 26), for example, appear in a tight cluster. Other groups can be identified also. The three lines drawn in Figure 5 place fuels with similar properties into distinct groups. From left to right these groups are 1. A wood group, characterized by a low heat of combustion, low char yield, and relatively high amounts of combustible volatiles. VOLUME 28, NUMBER 2, 1982 / 415

13 IJ.I ' ' ' Foliage J Group J Wood 29 i J Group [] i.-j 23,/ 'Z Bark l 2,j 2413/ j Lignin / 9 L s : s Group/ 2 v. -_v 8 e o / / I I I I / A H comb (FUEL), M JIKG FIGURE 5. Grouping of forest fuels by their heat of combustion and the portion of this heat available to flaming. All the volatiles released below 500øC are included in the latter. (/ ) foliage, (V) wood, (r ) stems, (ll) bark, ( ) other. Numbers refer to sample descriptions in Table A foliage group, with a wide range in both heat of combustion and volatile yield, and intermediate char yield. 3. A bark or lignin group, characterized by a high heat of combustion, a high char yield, and a wide range in yield of combustible volatiles. Within each group, the heat available from volatiles increases linearly with total heat of combustion. The increase can be caused by different chemical components and can originate in any of the three temperature regions. Most of the fuel samples studied fall into one of the three groups. Rotten white fir wood (35) and cheatgrass (36) fall neatly into the wood group. This was also apparent from the EGA curves. Pinyon pine stems (29), Douglas-fir stems (30), Utah juniper bark (31), bracken fern (39), saw-palmetto fronds (40) and tundra (41) fall into the foliage group. Rotten Douglas-fir wood (38), duff (partially decomposed pine litter (42)), and three of the bark samples (32, 33, and 34) make up the bark or lignin group. This last group apparently contains aromatic polymers which char when heated. Several of the samples are intermediate, or do not fit the characteristics of any group. Big sagebrush stems (27), Utah juniper stems (28), and Idaho fescue (37) are intermediate between wood and foliage. This was also apparent from the EGA 416 / FOREST SCIENCE

14 A High Medium Low 15 1A-- A4 rj29 11Z 2Z 7Z , 6 27r"] r"1 24 Z V '23 21 J IF18 /,( I I I I ?3 /x H comb (FUEL), M J/KG FIGURE 6. Grouping of forest fuels by the quantity of low-temperature volatiles (below 300øC) as measured by EGA. (A) foliage, (V) wood, (r ) stems, (I) bark, ( ) other. Numbers refer to sample descriptions in Table 1. curves. Excelsior (20), the only woody sample from a hardwood, falls to the left of the wood group. Hardwoods may form a separate group because they are known to have higher holocellulose and lower lignin contents than softwoods (Stamm 1964, p 3-4). The isolated cellulose (43) point is due to an exceptionally low heat of combustion and low char yield. Clearly, the extensive studies of cellulose pyrolysis may not relate to the thermal behavior of natural forest fuels. Furthermore, studies of the effect of fire retardants on cellulose (George and Susott 1971) may not indicate accurately their effect on the majority of natural fuels. GROUPING BY Low-TEMPERATURE VOLATILES Fuels can also be grouped by their tendency to form combustible volatiles in various temperature ranges by using the integrated EGA values in Table 2. Figure 6 is a plot of the combustibles vaporized below 300øC against total heat of combustion of the fuel. There is no apparent dependence of low-temperature volatile generation on the total heat of combustion. Groups can be arbitrarily divided by the horizontal, dashed lines, separating low, medium, and high amounts of low- VOLUME 28, NUMBER 2, 1982 / 417

15 I I I I o 0.4 o c:, 0.3 o 3o[3 4Z c) I.I _,.I "' z L.I,J H comb (FUEL), M J/KG FIGURE 7. Relationship between heat of combustion and high-temperature volatiles (between 400 ø and 500øC). (A) foliage, (V) wood, ([2) stems, (I) bark, (O) other. Numbers refer to sample descriptions in Table I. temperature volatiles. The high group requires above 0.3 g oxygen/g fuel, and the low group below 0.2 g. oxygerdg fuel. The two fuels with high values (sagebrush (3) and juniper (5) foliage) were recognized as being distinctive earlier, by examination of the EGA curves. The impact of the low-temperature volatiles on fire behavior is not well understood and more research in this area would be useful. The placement of these fuels into broad groups should aid the correlation of fire behavior to thermal behavior. It is interesting to note that even at the low level in Figure 6, enough heat is available from pyrolysis volatiles to vaporize water from a fuel with high moisture content. There is also enough heat available to bring adjacent fuel particles to their ignition temperature. This may help explain the ease of burning wet leaves containing volatile oils (Vines 1975). RELATION BETWEEN HIGH-TEMPERATURE VOLATILES AND THE HEAT OF COMBUSTION The EGA data can also be used to test the dependence of volatiles released from 400 ø to 500øC on total heat of combustion. Figure 7 is a plot of these data. The yield of high-temperature volatiles increases roughly linearly with increasing total 418 / FOREST SCIENCE

16 heat of combustion. Outliers in this plot have a high heat of combustion due to either low-temperature volatiles (29 and 32) or components with high char yields (33 and 38). Up to 5 MJ/kg becomes available to flaming from this high-temperature region. This important contribution was not included in previous studies limited to 400øC (Rothermel 1976, Susott and others 1975). This omission could cause a significant underestimation of the heat available to flaming combustion for many fuels. Apparently, fuel components with high thermal stability cause much of the variation in total heat of combustion. The release of vapors above 400øC should be included in models describing the flaming combustion of forest fuels. CONCLUSIONS The EGA curves have been shown to give a detailed qualitative characterization of the thermal behavior of forest fuels. The decomposition of these fuels is extremely complex and extends over a broad temperature range. Fuels with similar properties can be grouped by examination of the EGA curves in three broad temperature regions: below 300øC, 300 ø to 400øC, and 400 ø to 500øC. Measurements of the oxygen stoichiometry and combustion calorimetry support the qualitative assignment of fuels to these groups. Oxygen stoichiometry provides a good estimate of the heat of combustion for volatile products. The heat available to flaming can be partitioned by temperature of release in the EGA curves, and separated from glowing combustion of char. More research on behavior of burning fuel is needed, however, before the above parameters can be related to flammability characteristics. The data provided in this report should provide a sound basis for developing these relationships. LITERATURE CITED ALBINI, F. A Computer-based models of wildland fire behavior: a users manual. USDA Forest Serv, 68 p. Intermt Forest and Range Exp Stn, Ogden, Utah. AMERICAN 8OCIETY FOR TESTING AND MATERIALS Annual Book of ASTM Standards, Part 31. AMERICAN SOCIETY FOR TESTING AND MATERIALS Annual Book of ASTM Standards, Part 19. AMERICAN SOCIETY FOR TESTING AND MATERIALS Annual Book of ASTM Part 16. Standards, ANDREWS, P. L Testing the fire behavior model. In Proc Sixth Conf on Fire and For Meteor (Robert E. Martin and others, eds), p Soc Am Foresters. 304 p. BLANKENHORN, P. R., R. C. BALDWIN, W. MERRILL, JR., and S. P. OxxoNœ Calorimetric analysis of fungal degraded wood. Wood Sci 13: BRADBURY, A. G. W., and F. Shafizadeh Role of oxygen chemisorption in low-temperature ignition of cellulose. Combustion and Flame 37: BROWNœ, F. L Theories of the combustion of wood and its control. USDA Forest Serv, Forest Prod Lab, Rep 2136, 70 p. Madison, Wis. GEORGE, C. W., and R. A. 8USOTT Effects of ammonium phosphate and sulphate on the pyrolysis and combustion of cellulose. USDA Forest Serv Res Pap INT-90, 7 p. Intermt Forest and Range Exp Stn, Ogden, Utah. MONTGOMERY, K. R Ether extractive and flammability of Mediterranean-type shrubs. M S thesis, Calif State Polytech Univ, Pomona. 38 p. OHLEMILLER, T. J., J. BELLAN, and F. ROGERS A model of smoldering combustion applied to flexible polyurethane foams. Combustion and Flame 36: PEARCE, E. M., S.C. LIN, M. S. LIN, and S. N. LœE Relationship between polymer structure and char formation. In Thermal methods in polymer analysis (S. W. Shalaby, ed), p The Franklin Institute Press, Philadelphia, Pa. 204 p. PHILPOT, C. W Influence of mineral content on the pyrolysis of plant materials. Forest Sci 16: VOLUME 28, NUMBER 2, 1982 / 419

17 PHILPOT, C. W The pyrolysis products and thermal characteristics of cottonwood and its components. USDA Forest Serv Res Pap INT-107, 31 p. Interrot Forest and Range Exp Stn, Ogden, Utah. PHILPOT, C. W., and R. W. MUTCH Seasonal trends in moisture content, ether extractives, and energy of ponderosa pine and Douglas-fir needles. USDA Forest Serv Res Pap INT-102, 21 p. Interrot Forest and Range Exp Stn, Ogden, Utah. ROTHERMEL, R. C A mathematical model for predicting fire spread in wildland fuels. USDA Forest Serv Res Pap INT-115, 40 p. Intermt Forest and Range Exp Stn, Ogden, Utah. ROTHERMEL, R. C Forest fires and the chemistry of forest fuels. In Thermal uses and properties of carbohydrates and lignins (Fred Shafizadeh, Kyosti V. Sarkanen, and David A. Tillman, eds), p Academic Press, New York. ROTHERMEL, R. C., and H. E. ANDERSON Flame spread characteristics determined in the laboratory. USDA Forest Serv Res Pap INT-30, 34 p. Interrot Forest and Range Exp Stn, Ogden, Utah. SHAFIZADEH, F Pyrolysis and combustion of cellulosic materials. Advan Carbohyd Chem 23: SHAFIZADEH, F., and G. D. McGINNIS Chemical composition and thermal analysis of cottonwood. Carbohyd Res 16:273,277. SHAFIZADEH, F., P. P.S. CHIN, and W. F. DeGroot Effective heat content of green forest fuels. Forest Sci 23: SHAVlZADEH, F., and A. G. W. B ADBUR¾ Smoldering combustion of cellulosic materials. J Thermal Insulation 2: SNEEUWJAGT, R. J., AND W. H. FRANDSEN Behavior of experimental grass fires vs. predictions based on Rothermel's fire model. Can J Forest Res 2: STAMM, A. J Wood and cellulose science. The Ronald Press Co, New York. 549 p. SUSOTT, R. A. 1980a. Thermal behavior of conifer needle extractives. Forest Sci 26: St SOTT, R. A. 1980b. Effect of heating rate on char yield from forest fuels. USDA Forest Serv Res Note INT-295, 9 p. Interrot Forest and Range Exp Stn, Ogden, Utah. SUSOTT, R. A., F. SHAFIZADEH, and T. W. AANERUD A quantitative thermal analysis technique for combustible gas detection. J Fire and Flammability 10: SUSOTT, R. A., W. F. DEGROOT, and F. SHAFIZADEH Heat content of natural fuels. J Fire and Flammability 6: TANG, W Effect of inorganic salts on pyrolysis of wood, alpha-cellulose, and lignin, determined by dynamic thermogravimetry. USDA Forest Serv Res Pap FPL-82, 30 p. Forest Prod Lab, Madison, Wis. VINES, R. G Bushfire research in CSIRO. Search 6: / FOREST SCIENCE