Mundéfidru-mpd .umr—d 4A» This is to certify that the thesis entitled AN APPRAISAL OE UPLAND FOREST FUELS AND POTENTIAL FIRE BEHAVIOR FOR A PORTION OF THE BOUNDARY WATERS CANOE AREA presented by Peter Jon Roussopoulos has been accepted towards fulfillment of the requirements for Ph . D . degree in Forestry n ‘ Majo rofessor Date QCthez 5. 1228 0-7639 ovmuz PINES ARE 25¢ PER DAY . man rm Return to book drop to remove this checkout from your record. Men 1:“, ' j .l.‘ o ‘ ,-' I a! ' ' AN APPRAISAL OF UPLAND FOREST FUELS AND POTENTIAL FIRE BEHAVIOR FOR A PORTION OF THE BOUNDARY WATERS CANOE AREA BY Peter Jon Roussopoulos A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Forestry 1978 ABSTRACT AN APPRAISAL OF UPLAND FOREST FUELS AND POTENTIAL FIRE BEHAVIOR FOR A PORTION OF THE BOUNDARY WATERS CANOE AREA BY Peter Jon Roussopoulos Forest flammability conditions were quantitatively appraised within a 40,000 hectare portion of the Boundary Waters Canoe Area (BWCA), where a pilot study has been proposed to determine the feasi- bility of using naturally occurring wildfire to restore and maintain pristine wilderness environments. A broad-scale inventory of upland forest fuels was conducted within the study area during the summer of 1976. An assortment of vegetation and detritus sampling techniques was used to quantify the living and dead fuel components at each inventory site. Amounts of both surface and aerial fuel components were recorded by species, size class, condition, and height above the ground in increments of 30.5 cm. Total fuel loading ranged from 3.8 to 17.2 kg/mz, with most of the variation attributable to humus and large diameter downed dead- wood. Eighty percent of the inventory samples were within aspen- birch communities. Repeated burning in the late 19th century is thought to be responsible for the abnormally high representation of aspen-birch types within the study area. The remaining 20% of the samples fell within a variety of conifer and mixed deciduous-conifer community types. Peter Jon Roussopoulos Little of the dispersion in fuel loading estimates could be attributed to differences among forest community types. Variation was greater within these types than it was among them. For each inventory sample, potential fire behavior was predicted under standard weather conditions using available fire behavior models. A cluster analysis, performed on transformations of the fire behavior predictions, identified four major groups of samples on the basis of fuel flammability. The largest cluster contained 81% of the inventory samples, reflecting the apparent homogeneity of upland stands in the study area. Unfortunately, there was no clear relation- ship between the fuel type classification and recognized vegetation types, while a multiple discriminant analysis required 42 descriptor variables to correctly classify 91% of the inventory samples on the basis of observed stand and site characteristics alone. Results suggest that a single fuel model should be used to represent the entire study area. Nomographs were constructed for the two most representative sample clusters to display predicted surface fire intensities and spread rates, as well as threshold conditions for vertical fire development into tree crowns. Predicted occurrence of long-distance spotting, using one nomograph, compared favorably with actual condi- tions on three project fires in or near the study area, suggesting that the nomographs may be operationally useful for assessing the threat of fires escaping the study area. Further analysis is recommended before extending these results to the entire BWCA. ACKNOWLEDGMENTS I am deeply indebted to the many individuals whose effort and cooperation have made this study possible. First, I would like to ac- knowledge the U. S. Forest Service for providing the opportunity and resources for this research. In particular, I am grateful to Project Leaders Von J. Johnson and Stanley N. Hirsch for their administrative support and personal encouragement throughout the study. To my colleagues at the North Central Forest Experiment Station I offer thanks for their interest, cooperation, and expert counsel. Special thanks go to Robert M. Loomis, Richard W. Blank, and William A. Main. Their technical contributions and assistance proved invaluable in all phases of the study. I also would like to thank the many field and laboratory assistants whose long hours and diligent effort provided the database for this analysis. Appreciation is also extended to the Superior National Forest for the cooperative work environment it provided. Fire Management Officer Clifford E. Crosby, Jr. and District Rangers Ray C. Chase, A. Earl Niewald, and Wayne A. Smetanka deserve a personal thanks for logistic support in the field and constructive criticism of study results. I am especially grateful to Dr. Gary Schneider for his guidance, encouragement, and patience throughout my academic program, and also to Dr. Victor J. Rudolph, who accepted me as his own upon Dr. Schneider's departure. May I also offer thanks to Dr. Erik D. Goodman, Dr. James B. ’L’L Hart, Jr., and Dr. Peter G. Murphy for their friendship and assistance, and for serving on my graduate committee. Finally, and most importantly, I wish to thank my wife Cori, son Andrew, and daughter Nicole for their patient understanding and en- couragement despite the many family plans and opportunities foregone. {it TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES INTRODUCTION The BWCA Fire Management Proposal Problem Formulation OBJECTIVES THE BWCA AND PILOT STUDY AREA Physiography and Soils Climate and Weather Vegetation Upland Plant Communities Lowland Plant Communities Fire Activity and Fire Management FUEL INVENTORY AND FUEL MODEL DEVELOPMENT Inventory Methods Inventory Sample Design On-Site Procedure General Site Description Downed Deadwood Minor Vegetation Organic Mantle Mosses and Lichens iv 10 12 14 17 19 20 25 25 29 31 32 33 33 35 36 37 37 Shrubs and Seedlings Crown Fuels Auxiliary Fuel Characterization and Inventory Data Analysis Downed Deadwood Minor Vegetation Dry Weight Conversion Component Biomass Vertical Distribution Organic Mantle Mosses and Lichens Shrubs and Tree Seedlings Methods Analysis Results Component Weights Size Distribution Vertical Distribution Crown Fuels Weight of Living and Dead Crown Components Size Class Distribution Vertical Distribution Inventory Results Floristic Composition Fuelbed Properties CHEMICAL AND PHYSICAL FUEL PROPERTIES Chemical Properties Methods Page 38 38 39 4O 50 50 55 56 58 58 6O 60 61 62 62 64 68 69 69 71 72 79 82 87 97 97 98 Page Results 98 Physical Properties 103 FUEL TYPE CLASSIFICATION 107 Fuel Appraisal 110 Surface Fire Behavior 110 Torching and Crowning Requirements 115 Fire Behavior Ordination and Classification 118 Classification Methods 119 Classification Results 122 Discriminant Analysis and Relationship to Cover Type 126 APPLICATION AND VALIDATION OF RESULTS 133 Fire Behavior Prediction 133 Validation 140 Frequency and Distribution of Spotting Conditions 143 SUMMARY AND CONCLUSIONS 147 APPENDIX 155 LITERATURE CITED 156 m’ Table Table Table Table Table Table Table Table Table Table Table l. 10. 11. LIST OF TABLES Area of virgin landscape by plant communities in the Boundary Waters Canoe Area, Minnesota, January, 1973. Quadratic mean diameters (qi-) and particle specific gravities (01-) for downed deadwood materials by species and Size class. Measured vertical slash distributions by particle size class on nine Michigan clearcuts, and the derived beta distribution parameter Yj' Seasonal average dry weight conversion factors (C.F.), standard errors (S.E.), number of sample collection days, and species composite membership for some grasses, forbs, and small woody plants collected near the Kawishiwi Field Laboratory. Consensus estimates of percentage biomass distributions among three height strata for taller minor vegetation species. Sample size and regression coefficients for estimating component dry-weights of shrubs and small trees (<2.5cm dbh). Regressions through origin (y = bx) for height and crown length, and linear regressions (y = a + bx) for basal stem diameter versus stem diameter (cm) at 15 cm above ground level. Regression statistics for estimating fractional weight contributions of woody components by size class and condition (live or dead) for 17 species of northern Minnesota shrubs and small trees. Dry weight regression equations and sources for live and dead crown components. Statistical estimators and sources for live crown dry weight by size classes. Statistical estimators and sources for dead crown dry weight by size classes. vii Page 22 43 46 52 57 63 65 67 7O 73 76 Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table l2. l3. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. Land survey sections selected for sampling and distribution of subsample units among management zones and community types. Summary data for vegetative surveys in the study area and the BWCA in total. Relative species dominance or density by stratum for the study area and the BWCA in total. Summary of fuel loading means (E), ranges, and stand— ard deviations (s) by fuel component and community type. Summary of surface fuel loadings (below 1.8 m) by size class, condition, and community type. Summary of crown fuel loadings (above 1.8 m) by size class, condition, and community type. Summary of total ash content determinations by fuel component, size class, and condition. Summary of silica-free ash content determinations by fuel component, size class, and condition. Summary of low heat of combustion (cal/gm) determin- ations by fuel component, size class, and condition. Fuel particle densities (p) and surface area-to- volume ratios (0) for selected foliar materials. Fuel moisture and windspeed conditions used in appraising potential fire behavior. Entry sequence for the 42 stand and site variables in a stepwise discriminant analysis of fuel types. Distribution of sample plots among cover types and fuel types. Summary of fuel properties for BWCA fuel types I and III. Weighting coefficients for dead fuel moisture cal- culation. Comparison of predicted versus actual spotting occurrence on three project fires in or near the study area. viii Page 81 85 86 88 9O 92 99 100 101 106 116 128 130 132 137 142 Page Table 28. Probability that any day is a "spotting day" by 145 month, and conditional probability that any spotting day (day 0) is followed by exactly n additional con- secutive spotting days. ix Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure 10. LIST OF FIGURES Geographic location of the BWCA and fire management pilot study area. Bedrock geology of the fire management pilot study area. Logging history of the fire management pilot study area. Stand origin map for fire management pilot study area. Standard sample plot and orientation. Relationship between predicted and measured vertical slash distribution for size class I in the Roscommon 1 sample area. Predicted versus actual cumulative percentage points for vertical slash distributions on nine Michigan clearcuts. Time series of measured dry weight correction factors for composite groups of grasses, forbs and small woody plants collected near the Kawishiwi Field Laboratory during 1976. Locations of land survey sections randomly selected for fuel inventory. Probability distribution histograms for a.) age of dominant overstory component, and b.) tree basal area at fuel inventory plots. Vertical distribution of fuel materials for two con- trasting stands. Principal components ordination of fuel inventory units in the pilot study area. a.) Location of subsampling units in the plane of the first two components. b.) Location of subsampling units in the plane of the first and third components. Page 15 23 28 34 48 49 54 8O 83 95 123 Page Figure 13. Clustering dendrogram showing the hierarchical 125 relationships among the inventoried plots and the level of dissimilarity (distance) chosen to define fuel types. Figure 14a. Fire behavior nomograph for fuel type I. 134 Figure 14b. Fire behavior nomograph for fuel type III. 135 INTRODUCTION The extent to which unplanned forest fires should be used as a tool to manage wilderness vegetation has recently become an important and controversial issue among conservationists, ecologists, land managers, and forest product manufacturers. All have identified a need for better understanding of fire's role in establishing and maintaining natural ecosystems. ‘Accordingly, a vast body of literature, including several symposia and at least one full book (Kozlowski and Ahlgren 1974), has been devoted to this topic. One common thread woven through much of this literature is the recognition that fire has been an important ecological factor for as long as terrestrial plant communi- ties have existed. Over fifty years of aggressive fire suppression, however, have minimized its influence, often facilitating gradual fuel accumulations and notable successional changes over relatively broad areas. Because wilderness managers are charged with preserving or maintaining a "natural landscape, one in which man is only a temporary occasional visitor (Heinselman 1965), they have become concerned over observed changes brought about by their own management activities. Many have advocated that fire be allowed to play a more natural role in wilderness environments. Although the realization that fire is needed to maintain certain ecosystems is not new (Soper 1919, Maissurow 1935), two simultaneous developments can be given much of the credit for its recent revival. First, mounting evidence of fire's historical role in North American forests, and of adverse or potentially adverse ecological responses to past fire suppression activities has provided motivation for public and professional concern. Second, the advent of quantitative models for predicting fire behavior (Rothermel 1972, Albini 1976) and improved methods for inventorying forest fuels (Van Wagner 1968, Brown 1971, Brown and Roussopoulos 1974, Sando and Wick 1972) has helped make the restoration of fire to wilderness a realizable objective. Potential answers to some of the "Hows?" as well as the "Whys?" of fire management are now beginning to appear. Though certainly not without controversy, the net result has been a marked liberalization of fire suppression policies among public land management agencies. Several attempts to reintroduce fire to natural ecosystems have recently been evidenced in the literature (Agee 1974, Aldrich and Mutch 1972, Butts 1976, Daniels 1974, Devet 1976, Gunzel 1974, Kilgore and Briggs 1972, Kilgore 1975, Loope and Wood 1976, Sellers and Despain 1976). At present, at least 20 land units within national parks, monuments, or national forest wilderness areas total- ling roughly two million hectares are being managed with provisions for "supervised" wildfire activity under prescribed conditions.l/ Further- more, a 1978 revision to the USDA Forest Service Fire Management Policy (FSM 5100-5130) promises to extend and tailor similar programs to a great variety of additional wildlands throughout the United States. 1/ Personal communication, November 7, 1977. Richard J. Barney, Intermountain Forest and Range Experiment Station, USDA Forest Service, Missoula, Montana. The BWCA Fire Management Proposal 1/ A planr- is now being considered to restore fire, on a pilot basis, to a portion of the Boundary Waters Canoe Area (BWCA), a lake- melwilderness occupying more than 400,000 hectares in northeastern Minnesota (Figure l). The historical role of fire as an environmental factor in the BWCA has been well documented by several investigators (Heinselman 1969, 1970, 1973; Swain 1973; Wright 1969, 1974), while evidence of successional change in absence of disturbance has been discussed by Ohmann and Ream (1971b), Grigal and Ohmann (1975), and Ahlgren (1974, 1976). The proposed fire management pilot study is intended to determine the feasibility of using natural fire ignitions in the BWCA to restore and maintain a pristine wilderness condition. To initially reduce the complexity of the program and ease the transition from the BWCA's traditional fire exclusion policy, a contiguous 40,000 hectare pilot study area--comp1ete1y bounded by established canoe routes--was selected (Figure 1). Inside this area, the plan will allow fire, within given prescription guidelines, to resume a more natural ecological role. Criteria for a candidate "prescribed natural fire" include: 1. The fire must be lightning-caused. 2. It must be within the pilot study area. 1/ Crosby, Clifford E. 1978. BWCA Pilot Study on Prescribed Natural Fire. Unpublished draft plan on file at the Superior National Forest Supervisor's Office, Duluth, Minnesota. .mmum xonum uoawm ucwfimmmgfi 952% 9.8 flozm 9.3 No coaumooa OHSAMHmooo .H wusmwm a... .3.» 1'03 3.. a 13.3.5 :5... E II!- li - I . 5.... .5 :5 51.5.... ll :3 =3: .3!!- ll as 4(2th $0.6; 7.62cm dia. lOOO hr. Inventory Methods Sources of organic materials that may become fuel for forest fires include the L, F, and H layers of the forest floor, mosses and lichens that grow on the forest floor, herbaceous plants, downed deadwood of various sizes lying above the litter layer, understory shrubs and tree reproduction, and tree crown materials--either living or dead. Inven- tory techniques differ for the various fuel components, due both to physical differences and efficiency considerations, and to differences 32 in fire model sensitivity to sampling errors. In the interest of sampling effeciency, a double-sampling or allometric inventory procedure was used for all fuel components. Only relatively easily quantified dimensional characteristics of vegetation and detritus mat- erials were measured in the field--and pre-established theoretical or statistical relationships were utilized to convert measurements to estimates of the needed loadings by size class, height, etc. For example, to estimate litter loadings, only the vertical depth of the litter layer was measured at each inventory site. Supplementary bulk density determinations (Loomis 1977) facilitated conversion of these depths to fuel loadings in grams per square meter. Without such measures, fuel inventories would necessarily involve tedious extractive sampling methods, and in remote areas this would be economically pro- hibitive. Inventory Sample Design Due to the unavailability of a suitable cover type map for the study area at the onset, prestratification of the inventory sample by vegetative type was not feasible. Hence, a two-stage random sampling scheme was chosen for this survey to minimize travel time. The primary sampling units were individual land survey sections (or portions thereof) containing 64 hectares or more of land area. Sampled sections were chosen randomly with equal probability. Ten secondary sampling units were located in each chosen section. Secondary sampling units consisted of two subsamples, the first located randomly within the section, rejecting points falling in water or on lowland sites, and the second spaced 80 meters away at a randomly 33 chosen azimuth from the first. Again, a rejection technique was used to assure that only upland sites were sampled. The plot design of each subsample is shown in Figure 5. It consists of a central point with four 30.5 meter transects radiating outward at right angles to one another, each ending in a 2 x 2 meter square quadrat. One of two orientations for the transects was assigned with equal proba- bility, the first corresponding to the cardinal directions (orientation 1) and the second being a 45° clockwise rotation of the first (orienta- tion 2). On—Site Procedure Details of the plot inventory process are described below for each of the recognized fuel components. General Site Description A general description of each plot location was recorded upon arrival at each designated sample point. Cover types (Society of American Foresters 1954), Grigal-Ohmann Community Type (Grigal and Ohmann 1975), and physiographic characteristics of the site were described. Any natural or man-caused disturbance to the site was also noted, estimating the time lapsed since the disturbance occurred. An increment boring was taken on a dominant of the cover-type species for site index determination using locally derived curves. Slope, aspect, soil texture, elevation above the nearest body of water, and percent of crown closure were also recorded. These measurements provide a mechanism for investigating relationships between fuel flammability and more conventional land classification criteria. 34 .coflumucwwho odd uoam meEMm ohmocmum .m ounmflh P--1 0') .— . :- TIII' .IIIIY a“ E m.om . uwuamo soda umucmu uoam \\.n I a; xv ’(\ 'P q. N u IIL :ofiuoucwfiuo u coaumuamwuo 35 Downed Deadwood Dead twigs, limbs, and boles of woody plants that have fallen to the forest floor can provide a highly significant source of fuel material. These fuels were sampled using the planar intersect technique (Van Wagner 1968, Brown 1971, Brown and Roussopoulos 1974), observing tally—rules given by Brown (1974). Each of the transects radiating from the sample point (Figure 5) was used to define a conceptual 30.5 meter sample plane that extended vertically from the litter surface to 2 meters above that level. The ground-level edge of each plane was delineated by stretching a measuring tape between two chaining pins placed at the ends of each transect. Individual dead and down woody particles intersecting designated seg- ments of each sample plane were tallied by species and size class. Transect tally zones, expressed in distance from the central sample point, were established separately for each size class as follows: Diameter Diameter Range Transect Tally Class (cm) Zone (m) I O-.64 28.0-30.5 II .64-2.54 24.4-30.5 III 2.54-7.62 18.3-30.5 IV >7.62 0-30.5 The actual diameter of each particle at the point of intersection with the sample plane determined its size class. When particle diameters were borderline, they were classified using a go-no-go gauge with Openings of 0.64cm, 2.54cm, and a length of 7.62cm, similar to that described by Brown (1974). Diameters of class IV material intersecting any portion of the 30.5 meter transect were measured with a diameter 36 tape or caliper to the nearest 0.25cm and recorded as rotten or sound. All other classes were simply dot-tallied if they intersected within their designated tally zone. The number of particle intersections over 30.5cm above ground level was recorded separately by size class, as was the height of the highest particle intersected. These data provide a means of represent- ing the vertical distribution of deadwood fuel materials--facilitating estimation of a characteristic packing ratio. Finally, the topographic slope along each sample transect was measured with a clinometer. Auxiliary data needs and procedures for analyzing the planar intersect data are discussed in a subsequent section. Minor Vegetation Grasses, forbs, and small woody plants such as twinflower and blue- berry, as well as shrub and tree reproduction under 30.5cm tall were clipped and weighed using a ranked-set sampling procedure (McIntyre 1952). The 2 x 2 meter plot at the distal end of each transect was divided into four sub-plots; one square meter each. These sub-plots were ranked visually in terms of total qualifying plant biomass. On the first transect, the sub-plot rated highest was clipped at ground level. Using a spring scale accurate to 1 gram, green weights were obtained individually for all species contributing at least two grams to the total sample weight. Species contributing less than this were pooled and weighed as a composite miscellaneous (other) category. On the second transect, the sub-plot rated second in biomass was clipped and weighed as above. This process was continued on sub-plots three 37 and four so that the ranking of the clipped sub-plot corresponded to the number of the transect to which it belonged. Periodically, a subsample of each clipped species was weighed, bagged, labeled, and returned to the laboratory for moisture content determination. Moisture content estimates were applied to convert from green to dry biomass (gm/m2). Organic Mantle Forest floor loadings were determined by measuring the individual depths of the L layer (the surface layer of the forest floor consisting of freshly fallen leaves, needles, twigs, stems, bark, and fruits) F layer (partially decomposed litter with portions of plant structures still recognizable) and H layer (well decomposed organic matter of unrecognizable origin). These depths were measured to the nearest 0.25 cm at the center of each squaredmeter quadrat and averaged for each transect arm. Bulk densities from a concurrent study (Loomis 1977) were applied to convert average depths to weight estimates. Mosses and Lichens Ground cover plants were characterized by occular estimates of "percent cover" and average height for each of the three unclipped square-meter plots on each transect. These estimates were made separately for each of four groups represented: l. Dicranum spp. 2. Cladonia spp. 3.‘ Plurosium schreberi 4. "other" Again, conversions from volume to loadings were accomplished using bulk density values obtained from supplemental samples. 38 Shrubs and Seedlings Woody shrubs and tree seedlings greater than 30.5cm tall but less than 2.54cm dbh were sampled by a "ranked-set, double-sampling" procedure using regression analysis. The square-meter sub-plots at the end of each transect were ranked occularly according to shrub and seedling biomass and sub-plots were chosen as for herbaceous material. Within each sampled sub-plot, stem diameters of all shrubs and seedlings were measured 15cm above ground level and tallied by species and condition (living or dead). A supplementary study near the study area provided regression sta- tistics for estimating the amount and distribution of shrub and seedling biomass from the relatively simple basal diameter measurements made in the actual field inventory. Crown Fuels Finally, tree crowns were described by a technique similar to that of Sando and Wick (1972). The distal end of each transect served as a "sampling point" for a 20- factor Bitterlich sample (Bitterlich 1947). Trees qualifying for sampling or "in" trees were described by the following independent variables: a. Species b. D.B.H. c. Total height d. Height to base of live crown e. Height to the point where unpruned branches predominate the bole of the tree f. Live crown width 9. Average width of the cylinder comprised of dead branches below the live crown 39 h. Generalized geometric shape of the live crown in the following categories: (1) Cylinder (2) Cone (3) Spheroid (4) Hemispheroid (5) Truncated Spheroid (6) Paraboloid of Rotation (7) Inverse Paraboloid i. Condition of tree (living or dead) j. Number of dead branches below the live crown by diameter class at the point of exit from the bole. A double-sampling approach termed "dimensional analysis" by Whittaker and Woodwell (1968, 1971) or "allometry" by Kira and Shidei (1967) was applied to estimate total and component tree biomass for each species and condition category from the above descriptors. Further- more, the simple models of crown geometry allowed estimation of the volume within the canopy that is occupied by tree crowns as well as the vertical distribution of biomass within the canopy. Auxiliary Fuel Characterization and Inventory Data Analysis The indirect fuel inventory approach used in this study made it necessary to secure information on the relationships between measured fuel and vegetation properties and fuel loadings (gm/m2) by condition and size class, as well as vertical distributions of these materials. In this section, the various sources and uses of this information in 4O analyzing the raw inventory data are discussed. Again the fuel component classes are treated separately. Downed Deadwood Theoretical development of the planar intersect sampling technique has been discussed thoroughly by Van Wagner(l968), Brown and Roussopoulos (1974), and DeVries (1973). When cylindrical fuel particles are oriented randomly within a horizontal plane, the established dead- wood fuel loading (1) is computed as: i 1 i 4.1 8£ where: L = estimated fuel loading (gm/m2) di = diameter of the ith intersected particle (m) pi = density of the ith intersected particle (gm/m3) k = horizontal length of the sample plane (m) n = number of intersected particles When the plane is positioned along a slope, the horizontal length is computed as: 1 = 90(1 + ¢2)'1/2 4.2 where: 2' = slope length of the sample plane (m) ¢ = topographic slope tangent along plane Adapting Equation 4.1 for simple counts of intersected particles by species and size class, rather than actual diameter measurements, produces Equation 4.3: 41 n qijnijpij L.. = 4.3 ij 8%. 3 where: . - . .th . qij = quadratic mean diameter of particles from the i speCies .th . and 3 diameter class (m) nij = the number of intersected particles of the ijth species-size class combination pij = the density of particles from the ijth species-size class combination (gm/m3) ij = the horizontal length of the tally zone for the jth diameter class . . ..th . . . Lij = the estimated loading of the 13 speCies-Size class combina- tion (gm/m2) If qij is in centimeters, pij in gm/cm3 (specific gravity}, and l'j in meters, the combined Equations 4.2 and 4.3 become: = 2 Lij 0.01234 q ijnijpij/ll + ¢2 4.4 2'. 3 The values of l'j were 2.5, 6.1, 12.2, and 30.5 meters, respectively for size classes I, II, III, and IV. Values for ¢ and nij (or 2di2 for class IV particles) were determined at the inventory plots, while qij and pij are constants to be established by species and size class. 1/ Data from earlier fuel studies on the Superior National Forest— Roussopoulos, P. J. 1971. Results of four prescribed burns at Virginia, MN. Paper presented at USDA Forest Service R-9 Fire Control Meeting, Combined Air Officers and Fire Staff, Theodosia Springs, Missouri, April 26-30, 1971. 42 (Roussopoulos and Johnson 1973, Brown and RouSSOpoulos 1974) provided the required estimates (Table 2). In addition to loading by species and size class, the vertical distribution of this loading is useful in appraising fire behavior potential (Sando and Wick 1972). Measurements of fuelbed depth are the conventional means of determining fuel packing ratios, but highly skewed vertical loading distributions have made depth measurements both difficult and questionable (Brown 1970a, Albini 1975). In this study, an attempt is made to represent the vertical distribution of fuel loadings as a first step in establishing a characteristic fuel packing ratio. Desired properties for a theoretical probability function, to adequately represent the vertical distribution of downed deadwood, are that it be bounded at both ends and be capable of representing high positive skewness, since most deadwood particles are on or near the ground. Its parameters must also be easily evaluated from relatively simple on—site measurements. A special case of the beta distribution satisfies these requirements. Over the finite interval (0, l), the beta probability density function (Hahn and Shapiro 1967) is given as: 1"(Y+n) 1‘ (WP (n) 0, elsewhere where: c is a random variable \1 y and n are beta distribution parameters P(y) = { xy_1 e-x dx, or F(y) = (y-l)!, when y is an integer 0 43 Table 2. Quadratic mean diameters (qi.) and particle specific gravities (oij) for downed deadwood materials by species and size class. 1/ Diameter Class (cm) 0-.64 .64-2.54 2.54-7.62 >7.62 Species qij(cm) pij qij(CM) oij qij(Cm) pij pij Balsam fir .180 .690 1.118 .405 3.937 .360 .360 Jack pine .371 .550 1.311 .590 4.549 .490 .430 Red pine .450 .610 1.250 .529 4.031 .490 .440 White pine .279 .550 1.067 .509 4.064 .490 .350 Black spruce .203 .674 1.245 .647 3.759 .486 .400 Northern White cedar .203 .560 1.194 .491 3.759 .429 .310 Trembling aspen .376 .590 1.290 .440 4.318 .400 .380 Paper birch .320 .690 1.039 .537 4.318 .550 .550 Red maple ' .424 5650 1.013 .576 4.039 .550 .500 Other .424 .650 1.179 .576 4.267 .550 .500 Values obtained from both published (Roussopoulos and Johnson 1973, Brown and Roussopoulos 1974) and unpublished data sources (Roussopoulos 1971, 92, cit.; Roussopoulos 1973, 22, cit.). 3/ Values obtained from Wood Handbook (U.S.D.A. 1955). \ 44 When y < l and n > 1, this function displays the desired "reverse J" shape (Hahn and Shapiro 1967). The cumulative beta distribution is: O, x < 0, x = T(1+n) Y-l _ n-l F(x) F(Y)F(n) t (1 t) dt, 0 g x S 1, 4.6 l, x > 1 0 where: t is simply a dummy variable of integration For a special case of this function, when n=2, equation 4.6 can be simplified considerably: X F(x) = W tY-l(l-t) dt, 0 _<_ x s 1 0 Y tY+1 t=x = y(y+l) -—-- -—- , 0 S x S 1 Y Y+1 t=0 = xY[1+Y(1-x)}, O S x s 1 4.7 In this form, the function can easily be solved numerically for y given just one value of F(x) at a known fractional height within the fuel bed (x). The proportion of deadwood intersections below a height of 30.5 cm (pj) and the height in centimeters of the highest inter- sected particle (hj), determined by particle size class at each inventory plot, provide the required information. For the jth size class, the value of Yj can be found by fixed-point iteration as follows: For n = 0, l, 2, ..... , until convergence criteria are satisfied, compute: 45 Y- J(n) 30.5 30.5 . =-——- 1+. 1--———— -.+. . YJ hj Y3m) hj 9] Yam) 4 8 For this a lication iteration was terminated if , - , < ‘6. pp ' Y3(n+1) Y3(n) 10 Convergence was normally achieved in less than 10 iterations. With estimates of Yj determined in this manner, the deadwood fuel loading for each size class was distributed vertically up to a height of hj according Yj F(x) = l 1 + 1 - J‘— 4 9 j h. Yj h. ' 3 3 where: x = height above the litter layer (cm) F(x)j = the estimated proportion of the total jth size class loading occurring below a height of x cm. The above procedure was validated using unpublished data on the ver- tical distribution of cutting slash in nine Michigan clearcut areas (Table 3). These data were obtained using the planar intersect sampling method, but tallying intersections by height stratum as well as species and size class. Height strata were defined as 0-.30m, .30-.61m, .61-.9lm, .91e1.22m, 1.22-1.52m, 1.52—1.83m, and 1.83-2.13m. Represented in the sample were four clearcuts in the jack pine-oak type in central Lower Peninsula Michigan (Roscommon 1—4), three clearcuts in the black spruce- northern white cedar type in Upper Peninsula Michigan (Shingleton 1-3), and two mixed—oak harvests in the west-central lower Peninsula (White Cloud 1 & 2). Table 3 shows the measured cumulative distribution per- centage points by size class for each of these areas, along with the derived estimate of Yj using Equation 4.8. 46 Table 3. Measured vertical slash distributions by particle size class on nine Michigan clearcuts, l/ and the derived beta distribution parameter Yj. Fraction of Slash Volume Below Datum Height Size Datum Height (m) Sample Area Class .30 .61 .91 1.22 1.52 1.83 2.13 Yj Roscommon 1 I .871 .956 .988 .997 1.000 .16111 II .846 .932 .973 .990 .998 1.000 .16577 III .898 .978 .985 .993 1.000 .12719 IV 1.000 *----- Roscommon 2 I .929 .972 .995 1.000 .11032 II .916 .979 .989 .998 1.000 ' .10492 III .898 .983 .994 1.000 .15828 Iv .958 .992 1.000 .09523 Roscommon 3 I .781 .928 .954 .973 .979 1.000 .23936 II .804 .911 .958 .982 .993 1.000 .21341 III .779 .919 1.000 .48112 IV .950 .971 1.000 .11157 Roscommon 4 I .851 .920 .954 .989 .998 1.000 .16036 II .734 .900 .954 1.000 .42172 III .917 .988 .988 1.000 .12976 - IV .412 .570 .844 .976 .976 1.000 .77237 Shingleton 1 I .811 .889 .958 .984 1.000 .23940 II .628 .885 .965 .988 1.000 .49642 III .829 .926 .974 .989 1.000 .21476 IV .647 .856 .933 .987 1.000 .46875 Shingleton 2 I .741 .829 .924 .981 .989 1.000 .28644 II .558 .767 .919 .972 .998 1.000 .53003 III .588. .809 .958 .979 1.000 .55922 IV .334 .572 .978 1.000 1.27362 Shingleton 3 I .380 .526 .705 .863 .942 .997 1.000 .75299 II .362 .547 .769 .851 .940 1.000 .87187 III .320 .586 .786 .961 1.000 1.09928 IV .225 .542 .879 .976 .993 .997 1.000 1.11123 White Cloud 1 I .974 .987 .987 .987 1.000 .03236 II .943 .986 1.000 .12736 III .923 .962 .962 1.000 .11982 IV .835 .976 1.000 .35985 White Cloud 2 I .880 .960 1.000 .26386 II .674 .859 .989 1.000 .52364 III .649 .865 .973 .973 .973 1.000 .40329 IV .825 .907 .907 1.000 .27273 1/ Data obtained from an unpublished study of wildland fuel conditions in the Northeastern United States (Roussopoulos 1973, 92: Cit.). 47 In Figure 6, the form of the cumulative distribution function is shown in comparison with measured values for the first entry in Table 3, while Figure 7 (a-d) plots the predicted cumulative percentage points against the corresponding measured values for all entires in Table 3. The 45° lines in Figure 7 indicate the loci of perfect agreement between predicted and actual values. Points falling below these lines represent over-estimates of the proportional volume of slash material below the indicated height. Agreement between predicted and actual values appears quite good for the smaller diameter classes-especially at the lower datum heights where most of the slash is found. Agreement deteriorates somewhat for the larger size classes, though, principally above the .91 meter datum height. The increased scatter with size class is not surprising, due to poorer representation of larger diameter fuels in the samples, while the apparent increased scatter in the upper right portion of each graph is due more to the scaling of axes than to absolute prediction errors. Furthermore, the four most serious underestimates in Figure 7d are attributable to the Shingleton 3 sample area where trees were felled but crowns were left intact and no products were removed. All other sample areas were commercial harvests. The Shingleton 3 area was the only case showing a significant difference (at the .05 level of confidence) between predicted and actual vertical distributions by the Kolmogorov-Smirnov nonparametric test for goodness-of-fit. When the Shingleton 3 data are deleted, Figure 7d shows reasonable agreement. Despite the noted discrepancies between the predicted and actual vertical distributions, the fact that the plotted points are scattered roughly evenly about the "perfect agreement" line, and the fact that the larger diameter particles-~where poorer agreement was found--are of 48 Measured Values .8" ‘EEEL____ Predicted Distribution (YI - 0.16111) .4 1 Cumulative Proportion of Slash Below Indicated Height 1 0 0.5 1.0 1.5 Height Above Litter Surface (meters) Figure 6. Relationship between predicted and measured vertical slash distribution for size class I in the Roscommon 1 sample area. 49 a. Size Class I b. Size Class II 999 ‘ .99 q Actual Proportion of Slash Below Datum Height Actual Proportion of Slash Below Datum Height 0 T I I 0 t 1 I .9 .99 .999 .9 .99 .999 Predicted PrOportion of Slash Below Predicted Proportion of Slash Below Datum Height Datum Height U U n 2 .3 c. Size Class III .3 d. Size Class IV 3 _999 . g .999-l s u o a 8 39‘ o' 4- 8 '99‘ 5. 9° 5- g -+ 1O 2 “a ”‘5 0 .. g ,9 4 x 5 .9‘ v4 -v-1 u u u u o o m a 3 O 04 N a a H .— g 0 I f I g 0 I T U 9 ,9 .99 .999 g .9 .99 .999 ‘ Predicted Proportion of Slash Below a Predicted Proportion of Slash Below Datum Height Datum Height Figure 7. Predicted versus actual cumulative percentage points for vertical slash distributions on nine Michigan clearcuts. 1/ Cumulative percentage points are shown by particle size class for datum heights of .30 m (a), .61 m 0), .91 m (+). 1.22 m (x), 1.52 m (0),- and 1,83 m (A). 1/ Data obtained from an unpublished study of wildland fuel conditions in the Northeastern United States (Roussopoulos 1973, 22, cit.) 50 minor importance in a fire behavior context, suggest that Equation 4.9 yields suitable vertical distribution estimates in untreated logging slash. In non-slash fuels like those encountered in the BWCA study area, this algorithm is yet untested. Nevertheless, since the distribution of particle inclination angles in one-year-old and older slash has been found to be similar to that of naturally fallen branch material (Brown and RouSSOpoulos 1974), it seems likely that the general form of the vertical distribution function would also be similar. Minor Vegetation Because herbaceous and small woody plants (including shrub and tree seedlings less than 30.5cm tall) were actually clipped and weighed in the field, only three supplementary data items were required: plant moisture content and associated dry weight conversion factors, the fractional distribution of biomass between foliar and woody parts, and the vertical distribution of these materials. These will be discussed in turn. Dry_Weight Conversion Using the ranked set sampling procedure of McIntyre (1952), the arithmetic mean loading for all clipped plots provides an efficient, unbiased estimate of the fresh biomass of these materials per unit area. Since actual determination of plant moisture content was not feasible in this study, an indirect approach was taken to convert the fresh weight estimates to oven-dry values in grams per square meter. Two concurrent sampling efforts were established during the period June 24 to August 26, 1976--while the field fuel inventory was underway. 51 One of these efforts involved the periodic subsampling of clipped vegetative material during the field inventory process. Subsamples were weighed in the field, bagged, and transported via fire detection aircraft to the Kawishiwi Field Laboratory near Ely, Minnesota within five days. At the laboratory they were oven-dried for 24 hours at 105°C and reweighed. The dry weights were divided by the field fresh weights to obtain the dry weight conversion factor (CF) for the sample. The second sampling effort was conducted at the Kawishiwi Field Laboratory where plant moisture contents could be monitored more continuously.l/ Plants representing 21 species or related species groups found commonly in the BWCA study area were collected at intervals of one to several days (excluding non-work days and days with rain) throughout the sample period. Forty-two subsamples were taken between 1400 and 1600 CDT each sample day (3 subsamples each from 14 species), with individual species or species groups being represented in proportion to their occurrence. The number of sample days for any given species or species group ranged from 3 for starflower to 22 for large-leaved aster (Table 4). All above-ground plant parts were included in each sample. Moisture contents and dry weight correction factors were determined gravi- metrically as described previously and each subsample triplet was averaged for the day. At the end of the sample period, the time series of daily correction factors were examined and compared on the basis of magnitude and seasonal Loomis, Robert M., Peter J. Roussopoulos, and Richard W. Blank. 1978. Summer dry weight conversion factors and associated moisture contents of some herbaceous and other plants in northeastern Minnesota (Man- uscript in preparation for publication). 552 Table 4. Seasonal average dry weight conversion factors (C.F.), standard errors (5.8.), number of sample collection days, and species composite membership for some grasses, forbs, and small woody plants collected near the Kawishiwi Field Laboratory. 1/ Species Plant Groupg n C.P 8.8. Composite Labrador Tea (Ledum groenZandicum) 11 .422 .012 l Blueberryszcccinium spp.) 18 .402 .009 1 Club Moss3 (Lycopodiwn spp.) 18 .397 .011 1 Grasses“ 10 .378 .018 1 minus:5 (Rubus spp.) 4 .330 .024 2 Spreading Dogbane (Apocynum andnosaemifblium) 6 .323 .024 2 Twinflower (Linnaea boreaZia) 3 .303 .035 2 Bracken Fern (Pteridium aquilinum) 13 .291 .009 2 Wild Sarsaparilla (Aralic nudicaulis) 22 .285 .005 2 Strawberry (Fragria spp.) 10 .283 .014 2 Bunchberry (Cbrnus Canadensis) 18 .277 .005 2 Bush Honeysuckle (DierviZZa Zonicera} 5 .264 .024 2 Pearly Everlasting (Anaphalis margaritacea) 4 .250 .032 2 False Solomon's Seal (Smilacina rccemoaa) 3 .250 .017 2 Other Ferns 5 .240 .018 2 Wood Horsetail (Ecuisetum syZvaticum) 4 .228 .005 3 Large-Leaved Aster (Aster macrophyllus) 22 .212 .007 3 Starflower (Trientalis borealis) 3 .213 .003 3 Dwarf Solomon's Seal (Maianthemum canadénse) 17 .205 .006 3 Sweet Coltsfoot (Parasites spp.) 4 .158 .019 3 Bluebead-Lily (Clintonic borealis) 15 .090 .003 4 1/ After Loomis 22.‘21. (1978)(Qp, Cit.) 2Vacciniwn mym'tilloides and V. angustifolium are predominant. 3Lycopodium clavatum and obscurum are predominant. l‘Csmear: spp. and Oryzopais asperifblia are predominant. Rubus strigous and R. pubescent: are predominant. sFragria vesca and F. virginiana are predominant. 7Athyrium spp., Onoclea sensibilis and Osmunda cinnamcmea are predominant. 53 trend. Seasonal CF averages ranged from .090 for Clinton's lily to .422 for labrador tea (Table 4). Four species composites were identi- fied by graphic comparison. The membership of each composite is also shown in Table 4. For each sample day the samples representing each species composite were averaged, obtaining four time-series of composite correction factors (Figure 8), covering most of the field inventory period. Correction factor estimates for non-sample days were derived by linear interpolation between sample days, as indicated in Figure 8. Four observations concerning Figure 8 are worthy of note. First and most obvious, is the wide range of CF values, corresponding to moisture contents from 122 to 1150 percent, and consistent ranking of the four composite groups. Both the most and least succulent species apparently retain these distinctions throughout the season. Second, the general parallelism exhibited by these four signatures indicates that, though the moisture content levels vary considerably among the four groups, their relative moisture responses to environmental stimuli are quite similar. Third, a net upward trend in CF values is apparent for all composites, reflecting the impact of the 1976 summer drought throughout the upper midwest. In fact, termination of sampling on August 26 was due in part to the conscription of inventory personnel to fight several large forest fires attributable to the lower moisture conditions. Finally, the absolute range of fluctuation in the four series is noticeably higher at higher CF levels. In relative terms, though, they are similar. The range of CF values as a percentage of range mid-points is nearly constant for all species composites (37, 42, 40, and 40 percent respectively for composites 1, 2, 3, and 4). These observations suggest that 1) it is necessary to account for seasonal DRY WEIGHT CORRECTION FACTOR 0.4.. 03.. 02.. 01_ 00. m. m JUNE Figure 8. Composite 1 Composite 2 Composite 3 Composite 4 DATE (1976) 25 2O 25 25 10 15 1o 15 20 JULY AUGUST Time series of measured dry weight correction factors for composite groups Of grasses, forbs and small woody plants collected near the Kawishiwi Field Laboratory during 1976. 55 plant moisture content trends in converting fresh biomass estimates to dry-weight equivalents, and 2) these seasonal trends can be represented using a moisture content analog if calibrated prOperly for differences in general CF levels among species and locations. In this study, the CF measurements at the Kawishawi Laboratory provide a suitable analog for study area values. Calibration was achieved by linear regression Of correction factors for samples extracted from the study area against the corresponding values from Figure 8. The resulting regression equation is: CFi = 0.00817 + 0.86059 (KCFi) where: . . .th . . CFi = dry weight correction factor for the i speCies C0mp081te in the BWCA study area . . . .th KCFi = corresponding dry weight correction factor for the 1 species composite Obtained at the Kawishawi Field Laboratory. Additional regression statistics are: n=269, sy.x=0.0869, F=264.5, and r2=0.50. Although the coefficient Of determination (r2) is not overly impressive, the F statistic indicates an extremely high level of signi- ficance for the regression. The regression estimates of CFi were computed for each fuel inven- tory sample day and used to convert measured field fresh weights Of minor vegetation to oven-dry fuel loadings. Loadings were expressed in grams per square meter by species for each inventory sample. Component Biomass For small woody plants, a rough estimate Of the distribution of plant biomass among woody and foliar components was needed to define a representative surface area-to-volume ratio for the fuelbed. These 56 estimates were Obtained by clipping a few stems (less than 30.5cm tall) of each prominent species, separating the foliar and woody components, and determining the fresh weight of each. This was done in late summer following the period of major plant growth. The percentage of total fresh weight attributable to foliage was computed for each Species and rounded to the nearest 10%. Resulting values averaged 30% for labrador tea; 40% for blueberry, sweet fern (Comptonia peregrina), Rubus spp., Ribes spp., and bog rosemary (Andromeda glauoophylla); and 50% for all other shrubs less than 30.5cm tall. For small tree seedlings, results were 40% for balsam fir and red pine, 50% for black spruce and northern white-cedar, and 30% for all other species. These proportions were assumed to be representative Of the dry weight distribution as well. All woody parts were less than 0.64cm in diameter. Vertical Distribution Due to the definition of minor vegetation in this study, most of these plants were confined to the lowest height stratum (0-30.5cm), making it unnecessary to establish vertical biomass distributions. A few Of the herbaceous and low shrub Species, though, would frequently exceed 30.5cm in height. For these Species, field inventory personnel were asked to subjectively estimate the overall percentage distribution Of above-ground biomass among the 0-30.5, 30.5-61.0, and 61.0-9l.5cm height strata. The consensus Of these estimates (Table 5) provided a means of representing the vertical loading distribution for these species. 57' Table 5. Consensus estimates Of percentage biomass distributions among three height strata for taller minor vegetation species. Height Stratum (m) 0-.30 .30-.61 .61-.91 Species PERCENT Blueberry (Vdccinium spp.) 90 10 0 Wild Sarsaparilla (AruZia nudicaulis) 30 60 lo Spreading Dogbane (Apocynum androsaemifblium) 90 10 0 Bracken Fern (Pieridium aquiZinum) 20 6O 20 Sweet Fern (Comptonia peregrine) 40 4O 20 Wild Raspberry (Rubus spp.) 60 40 0 Wild Pea (Lathyrus spp.) 70 30 0 Current (Ribes spp.) 60 40 0 Labrador Tea {Ledum groenlandicum) 70 30 0 Bog Rosemary (Andromeda glauccphylla) ' 80 20 - 0 58 Organic Mantle TO convert litter and duff layer depths to oven-dry loadings (gm/m2), dry-weight bulk densities are required. Loomis (1977) reports forest floor weights, depths, and bulk densities for nine jack pine and six aspen stands near the Kawishawi Field Laboratory. Measured bulk densities and standard errors in the jack pine stands were 18.2 i 1.67, 48.2 i 1.85, and 66.6 i 3.14 kg/m3 for the L, F, and H layers respective- ly. In the aspen stands, corresponding values were slightly lower: 15.1 1 1.03, 42.0 1 2.27, and 54.1 i 2.40 kg/m3. The jack pine values were applied to all inventories in conifer types, while aspen values were applied in all hardwood types. Loading estimates were expressed in grams per square meter. In keeping with the National Fire Danger Rating System (Deeming et_al, 1977) the upper 0.64cm of litter was classified as 1 hr. timelag fuel, the next 1.9cm was 10 hr., and the remainder 100 hr. timelag. Mosses and Lichens Bulk densities were also needed for ground cover plants to derive loadings from estimates Of percent cover and height. These values were Obtained by a procedure similar to that Of Loomis (1977) for forest floor materials. Four 0.2 hectare sample plots were established near the Kawishawi Field Laboratory-~two in mature jack pine, one in a dense stand Of spruce-fir, and one in a rock outcrop community within a jack pine stand. The plots showed no evidence of recent disturbance and were representative of much of the area in the BWCA. In each sample plot, an eight point by eight point systematic sam- pling grid was established. At each point, a 12.7cm diameter circular 59 sample of moss or lichen was extracted and bagged. Samples were located as close to the sample point as possible, but such that the entire area within the 12.7cm diameter circle was covered by a continuous mat of mosses or lichens. If no suitable ground cover was found within a 3 meter radius of the grid point, no sample was taken. Sample material was extracted by cutting around a 12.7cm metal cylinder at the ground with a knife. The thickness Of the extracted ground cover mat was measured to the nearest 0.25am at the time Of sampling. Samples were bagged and removed to the Kawishawi Field Laboratory where soil particles were separated from vegetative material using a water bath, and dry weights were Obtained to the nearest 0.1 gram by oven-drying at 105°C for 24 hours. Sample volumes were computed from the field depth measurements and resulting dry-weight bulk densities were expressed in kg/m3. In all, 227 samples were obtained on the four plots. Samples were post—stratified into two groups—-mosses and lichens. The moss group included 100 samples, comprised mainly Of Plurosium schreberi with minor representation from Dicranum spp. The mean and standard error of the bulk density measurements for this group were 14.954 1 0.420 kg/m3. For the remaining 127 lichen samples (Cladonia spp.), both the mean and standard error were roughly twice that for the mosses--26.873 i 0.839 kg/m3. These bulk densities were used to compute ground cover loadings from inventory Observations by: Lgc = 0.1 C D 0b 4.10 60 where: Lgc = ground cover fuel loading in gm/m2 C = field estimate of percent cover D = field estimate of ground cover mat depth in cm pb = appropriate bulk density estimate (kg/m3) Shrubs and Tree Seedlings Estimation Of the amount and distribution of biomass for shrubs and small trees from field measurements of stem diameter required the development Of regression equations. Although a few studies have produced relevant estimators for total and component above-ground plant biomass (Ohmann 85:21: 1976, Crow 1977, Telfer 1969), they provide little insight into the size distribution of dry matter as desired for fire modeling. Brown (1976) reported similar equations for 25 shrubs of the northern Rocky Mountains. In addition to the biomass equations, though, he included a table of stemwood weight percentages attributable to specific fuel size classes. A sampling effort similar to that of Brown (1976) was undertaken at the Kawishawi Field Laboratory to pro- vide the needed functions.l/ Methods Sample stems Of 17 species Of shrubs and tree reproduction (<2.5cm dbh) were collected during July and August Of 1976. At least 20 stems Roussopoulos, Peter J. and Robert M. Loomis. 1978. Biomass and dimensional properties of shrubs and small trees in northeastern Minnesota. (Manuscript in preparation for publication. North Central Forest Experiment Station, E. Lansing, Mich.). 61 per species were cut at ground level and were taken to the Kawishawi Field Laboratory for processing. For each sample stem, the following measurements were recorded: stem diameter at ground level and at 15cm above ground level to the nearest 0.25cm (measurement Of diameter at 15cm above ground avoids the region of high stem taper normally found at ground level), plant height, and length (depth) of crown to the nearest 15cm. Each plant was divided into components Of foliage and woody parts. Dead and live woody parts were also separated. All woody parts were further separated into size classes by diameters. Each component group was weighed to the nearest 0.1 gram and its moisture content determined by subsampling and ovendrying for 24 hours at 105°C. All fresh weights were converted to ovendry in this manner. Analysis Total plant weight, foliage weight, total wood weight (live and dead), and live woody weight all in grams dry weight (Y), were regressed with the 15cm stem diameter (X) in cm using the allometric model: Y = a Kb 4.11 Regression statistics were evaluated in linearized form using natural logarithm transformations of X and Y. The "a" coefficient was adjusted for bias inherent in this procedure (Baskerville 1972). For each stem, the dry weights Of all woody material less than 2.5cm inclusive, were divided by the overall weight for total wood and for live wood only. These ratios (Y) represent the proportional contribution Of size classes 0 tO 0.6cm and 0 to 2.5cm to the weight Of the live and the total woody components. They were regressed against the stem diameter at 15cm (X) using the hyperbolic model: Y = X/(a + bX) 4.12 62 The regressions were accomplished using the following linearized form: X/Y = a +bX 4.13 To help evaluate the vertical distribution of these fuels, linear regression equations were also developed for plant height and crown length on the 15cm diameter. Results In all, 460 stems were collected and processed representing 14 deciduous species of trees and shrubs and three coniferous trees (Table 6). For all Species, the range Of sampled stem diameters was 0.3 to 5.1 cm at 15 cm above ground. All species were represented over the bulk of this interval except Diervilla lonicera, Lonicera canadensis, and Rosa acicularis. These small shrubs rarely attain stem diameters outside the range sampled. Total above ground biomass ranged from 1 to 2,714 grams dry weight per stem for all species. Component Weights Regression statistics were calculated for dry weights of all above- ground components, foliage, total wood, and live wood (Table 6). Exami- nation Of the coefficients of determination (r2) shows reasonably good fits for all species except Diervilla lonicera and Lonicera canadensis. These low r2 values may be partially due to the narrow range (0.2cm for Diervilla) of sampled stem diameters compared to the measurement pre- cision (: 0.12cm). These equations agree quite well with those of Ohmann e£_al, (1976) except for Corylus cornuta, where their estimates Show somewhat lower weights--especially at the larger stem diameters. Because their samples were also collected in the vicinity Of the Kawishawi Field Laboratory, 63 .eweseu no seasons ha now 88.. 2:8 58 3a 88.. 2.: s: .88.. 35 3.2a 88.. a»: on. oooo .3929: .828 .238. 123...- saouo 268. Honey no Islam ca veeeoumxe one canvas: .oaneu ozu eouu eu:o«0wuuoou scaeueuoou one A use a use .ueooum o>ons enouoaausoo ma censuses enouoadusou cu wouoadwo sous ecu e« x .eleuu cm unwed) ucocomsou ecu ed » ewes: axe I a show end no and essaeneumoz \m am vm. ~mn.u «no.0n mo vo. vow-N oom.om on ow. mvv.a www.mn om em. mom.a nnv.mo a.mun.o mm nwweusopwuoo shock mom mm. mwv.n www.mm hoe mm. nwv.n oom.mn HA no. Hoe-w mmo.m 0mm mm. mmm.m van.vv o.num.o va eceuwuoss season mm mm. HNh.N mov.Nv we mo. w~>.~ wan.nv mm vm. Qua-n om~.~a maa om. vom.N mNo.mm o.mlm.o mm .mmu awash o no. v-.n NmN.OO m «o. v-.m ova.mo m on. ~m~.~ mmm.- 0 mm. nmm.~ ov~.nm m.alm.o mm neuowsuwbo seem nvu om. nm~.~ mm~.vm «ma so. men.” Ono-mm «v be. vmo.~ mmm.ma mma Om. ham-m avo.mo m.num.o ma .amu museum Av hm. mmo.~ oom.vm He no. 5mm.m vom.mm ma hm. «mo.m m~m.o~ am we. ham.N chm.®v m.MIm.o 5N .mmo nSHSQOE on no. nvm.~ oom.ha on we. wow-N o~o.m~ we no. evo.~ mow-Om we no. nmm.~ ems.mo m.num.o ma .mmw moons v mo. o~o.~ hao.mm v #0. «va.H www.mm N am. mew-H mam.m m cm. mmb.a oom.mm o.HIm.o mm nwmcouscofi QHOUNCOQ a On. ave-H on~.o n mm. moo.~ mo~.~H A ma. mam. «wo.m v me. ham-H HH~.¢H m.onm.o Hm casewsOH ewuwsuown vv om. mom.~ men.mv he om. mwm.~ vma.om m Hm. oHO.~ mHH.~H we mm. omv.~ mam.mo m.~nn.o on absence newmuoo «MA mm. amm.N mmo.vm 0nd 0a. Hmm.m www.mm ma mm. mmo.~ Hma.ha vma om. hmv.~ v-.vh w.mcm.o hm unooah wacuou he no. ohm.~ 0mm.ao me mm. mhm.~ 0mm.~o ha om. mum.a nan-va me no. o>~.~ mam.os O.nun.d mm showwhameq squash owe Hm. mnv.n mmm.mm OOH vm. wee-N emm.oo Hm mm. wmm.H mhv.oH end no. amn.~ Vnm.an H.vnm.o hw .dmu Mowzoce~os< owH om. vmv.~ hho.mv va om. mom.N moh.mv ma om. mum.a mmn.v~ cod mm. omm.N omN.mo H.VIm.o mN .mmw nocm< nNH mm. hav.N «mm.~m «ma mm. nov.~ men.vm ON mm. omo.H mOm.pH avg mm. own-N NmH.Mh m.vnm.o mw EsuoUan Room com «o. Omv.~ moo.mv van mm. mom.n nvm.mv mu Hm. ovm.H an.m~ mew vm. mom.“ hom.oo H.vnm.o on assess Roo< mv hm.o vmm.N omn.av om no.0 vov.N vom.~v mm v0.0 HHO.N mum.mm om om.o omN.n mah.~h n.nlm.o mm eosflnflln mean! xxm mm A a saw Nu n e th we n a xxw -Nu a e mwouosdwo swam oouuoaaoo too: e>wu 8003 add ummaHom Hence no cocoa msoum newuoom ..nnu 50m.~vv noon» Hanan pad abouts no mu£Ow031>uo unocomsoo mcwunawumo new uCOAmeuooo scanmouoow one owwn oamadm .0 wanna a. 64 and because they also used stem diameter measured at 15cm as the independent variable, close agreement is not surprising. Brown (1976) and Telfer (1969), on the other hand, used stem diameter at ground level. To facilitate comparision with the results Of these studies, the relation between the 15cm stem diameter and basal stem diameter was examined. Scatter diagrams suggested that ground diameter could be predicted from the 15cm diameter using simple linear regressions. The resulting coefficients were remarkably similar for all species (Table 7). Telfer's (1969) predictions for woody plants in eastern Canada, after diameter adjustment, were also in close agreement. Brown's (1976) equations, on the other hand, yielded lower weight estimates for most species, perhaps partially due to the different environmental conditions of the northern Rocky Mountains. Both Brown and Telfer predicted greater weights for Lonicera spp. at larger diameters (Brown's weights were lower than Telfer's). Brown had the broadest diameter range for Lonicera (0.3 to 1.7cm); Telfer's was similar to this study (0.1 to 0.7cm). Size Distribution Examination of scatter diagrams revealed that the proprotional contributions of the 0 to 0.6cm (I) and 0 to 2.5cm (I + II) size classes to total woody weight are discontinuous functions of stem diameter. They equal 1.0 at low stem diameters and fall quickly away from this value above some "critical stem diameter" near the upper size class limit. TO ensure realistic size class predictions on both sides of this discontinuity, two measures were necessary. First, for each size class the smallest sample stem was found that contained woody material in the next larger size class. Naturally, its stem diameter at 15cm was 65 mm mmmm. ammo. memo.a mmma. mm swam. meow. mm ommm. ommo. wwwmusoowooo whose om Ohma. mmhm. mnma.a memo. em Nmme. Nmmm. om Chev. maom.a MEMOwHOEm monsom mm mvma. oamm. Omha.a momo. 5H voov. boob. ha mmmm. nvhm.a swamm mm mmoa. Naem. mavo.a mmmo. mm moma. doom. mm hood. momo.a meMHsOHom mmom mm «mom. baom. whoo.H HmHH. ma mmom. meme. ma omhm. mmHN.H .mmm unease hm mmha. meow. bamo.H vmmH. mm meow. omam. mm mamm. mHmN.H .mmm mswomom mm mmma. HHhm. HomH.H maho. mm mmma. omom. mm mesa. thm. .mmm mmowm mm mmaa. womb. omwm. mowo. mm ohma. mwvn. mm movm. vaN.H mwmsmomcoo ohmowcon Hm ONHH. mamm. mamm.- mooa. Hm omma. thm. an mmma. momm.a MROOHGON mwmwbwmwn mm eama. whom. mmmm. omma. mm mwom. oamm. om Noam. eamm.a essence mommhou hm moma. vahm. mmmo.a memo. hm some. comm. mm mmbo. mmba.a mmoosh mosuoo mm woma. memo. mmvo.H mama. Hm moms. hmwm. Hm eomm. omnm.a MHmeHQOQ mwoumm hm moma. mama. nmoa.a Neao. ha vHHm. Hmom. ha boom. wham.a .mmm.amw£osmme< mm mama. Nmmm. mmmo.a mood. o mmom. Hmmm. o mmmm. mmNH.H .mmm macaw mm mmvm. mmvm. mmvo.a mvwa. mm mmov. moon. mm ammo. OOHN.H E3u00wmm Room om. mmom. memo. man.H mooo. mm bump. mmmm. mm Hmmm. Honm.a Saunas Room mm mmmm.o mamm.o moma.a emoo.o mm mhmm.o mmvo.o mm momm.o «005.0 mmfimmwmn mmwnm s x.>m -Nu Q m S x.%m Q s x.mm n moaoomm AEOV Houmsmflo Hommm Amuouoev sumsoa SBOHU Amuouwev unmfiom .H0>OH ocsomm 0>Obm so ma um Asov Houseman ewum momu0> nouoeoao swam demon Row Asp + m u we mGOammoummn mucosa one .nuosoa cacao one nomads Rom Axb n we samfluo smoounp msOflmmmumwm .R magma 66 always close to the upper diameter limit--about 0.5cm for the 0 to 0.6cm class and 2.1cm for the 0 to 2.5cm class. These diameter values varied little among species. Stems with diameters below these values were deleted from the respective size class regressions. This eliminated samples from the "flat" section of the curve where the proportional size class contribution is 1.0 and allowed separate mathematical representation of the "flat" and "falling" curve sections. Second, the critical stem diameter—-the point separating the two sections--was defined from the coefficients of each hyperbolic regression as a/(l-b). The regression equation applies only to stem diameters above this value, which results in the following expression for the fractional contribution of each size class (Y) in terms Of stem diameter (X): ( a . < X < l' M ' l O, for 0 Yifj—BS- ( flat section) Y = 4 4.14 -——3£——- for-——Ji—- < X ("fallin " section) . (a+bX)' (1-b) 9 Regression coefficients were calculated for use with equation 4.14, both for all wood and for live wood only (Table 8). For the <0.6cm size class regressions, Diervilla lonicera was the only species that had no samples with stem diameters more than 0.5cm. Data for all other species were subjected to regressions. Diervilla, Corylus cornuta, Lonicera canadensis, Rosa acicularis, and Sorbus americana were exempted from the 0.0 to 2.5cm size class regression analysis because each had less than five sampled stems that were 2.1cm or more in diameter. Good fits were Obtained for most of the remaining species with this model (Table 8). Actual weight estimates for each size class are Obtained by multi- plying the appropriate fractional weight contribution estimate (equation 67 .unaao one. sou-oases rune 0» cousnauuu- sensor an coauuauo or» as s o:- uobeu ocsouo e>one lo mu possess! .lu. Robe-eqo lobe nu u cupcakes usevcoaevcu sues) .xa + e n r\x ea novel scanneuoOL 0:9 I \u nu umun.~ mne. hmm>.v ~mmv.¢u nu he'd.“ ove. OOOn.v OOOn.cu Om ooh~.v For. ammo-o conn.01 en veto-n moo. cone-m 65mm.vn nanoutoumuuo execs c o o c o c c a c c cw ~o-.n mno. moov.m~ ov~o.ou on homo-n «no. cmm~.ma cano.o~: enouwtuse nausea & vmv~.~ «he. mo~w.n omma.cl m nmv~.« owe. mmhm.n v0m0.01 MN vva.v Nah. hmuh.h wau.vl nu mvmv.v ooh. Hocv.h oo-.vl .maa kw~um c o c a o . c c c o c an nONN. duo. nocn.n unea.au nu newn. dam. neon.n Hhm~.~u swue~s0auo one: m hmmo. ham. who~.m vcwh.vl a memo. v00. HHOH.M aooh.v| vn n?na.n vow. nnnm.n hmnv.ns ow v~m¢.~ «ah. mmw~.m nvmo.ul .mme qazshm o mnmm. ado. mmmn.n o-v.ou o newm. one. vo~5.n comm.on hm sno~.n o~m. mean-o doom.v| on vwo~.n Ava. Homn.o -n¢.¢l .mae homomoa m onwn. one. vo«w.~ mmo~.v1 o m~mn. ~ne. hmuh.u mooo.vu m~ ovow. nmo. mbOm.~ area. I m~ come. «we. when.“ mean. a .Qan sebum . c c c c c c c c c m~ vaH. Ono. nomo.u name. n ma vaH. Ono. mono.“ name. I nwncouucsu quotatoa a e e a a c c e a e e a e o e e e e e e Ghouwco~ 1-~>hwwe e c a e a e e e c 0 mm ovoo. v00. Omo~.m Onom.nu nn NoNQ. 0am. whao.v domO.NI fluschoo mafimkpu h «men. one. snow-H ~omv. I s mmcu. who. smog-A mmOe. : cm Nmo~.~ won. ~v~e.o v~o~.nu vm mec~.~ «we. Oqoo.m oooo.~| snooze eschew OH mnun. «ha. vm0h.n hoov.0I OH «can. fire. vavm.n hmoo.wl nu mnnH.N mow. moom.m Ov-.h| nu omHs.H mam. «mob.h ammo.ml cu0h-mnoh u~=uwm o mmom. moo. nem~.v o~mo.nu o cave. «do. m~mo.v omev.o: ha envv.~ moo. ~ow~.h waav.vv nu so-.~ mam. envm.o OOvO.vn .oao stagecomwfis w novm. ova. mohm.N v~O¢.nI w meow. eve. 0v~m.~ momm.nl mm ~woo.~ uwm. OnNh.b ~N00.m| mm comm-N vmm. Ov®@.@ a~¢n.vl .QGG mszaa w some. ovo. wmv~.v nova-ml o damn. Ono. A~m~.v «wee-mu nw mmoo.u «do. nubm.o nchn.m| mm nea~.~ mum. mnoo.h vows-v: sensuous use: 0 venn.~ boo. moan.n m~n~.o| 0 --.~ mom. momm.n ovmo.ou an Oahm.o one. vwbm.- vvnv.>| an Hmeo.m was. o-n.OH O~m~.OI sauceh use: a hemm.o m~m.o homo-n ommn.vv o ocmn.o ova-O mwhm.n hnO~.v| m~ v~ns.0 -m.o memo.“ ammo-HI m~ o-¢.o «No.0 amon.« uv~m.ou eoEon~ed nowa< : saw NM 2 e : wa -Ny A e a mum w» b o : saw nu n a mouse Noooa osaa an caea>ao muted anon: ass a: cooa>so uuuumxxeoo: o>wmxxmlmuoa>se sauce Recon sac-um eooa>ae moauoom so w.” v mused >600: v>aq so m.~ v mused >600: ~H< EU 0.0 v apnea hoes) o>aa EU 0.0 v mason hooo: mu: .momuu Hanan one unsunw mUOmmsswz onwnuuoc mo mwaoomm ha Mom Anson HO o>wav sewuwocoo not mmmHO ouwm an mucocomaoo zooos mo msofluonfluusoo usofio3 HchfluOMHw msfiumsfiumm Rom mowumfiumum :memoumom .m wanna 68 4.14, Table 8), times the corresponding predicted wood weight (equation 4.11, Table 6). Weights Of the 0.6 to 2.5cm and > 2.5cm Size classes, as well as dead wood weights are found by subtraction. Vertical Distribution Knowledge Of the total heights and crown lengths of understory vegetation is useful in representing vertical distributions and packing ratios for these fuels. Equations were developed to predict these dimensions using stem diameter at 15cm as the predictor variable in a forced zero-intercept regression model (Table 6). Plant heights for the conifers were roughly half those of deciduous plants with the same stem diameter. The crown length ratios for coniferous samples (crown length/total height) were characteristically l l/2 times those of deciduous species. Estimated biomass for each shrub tallied in the field fuel inven- tory was distributed vertically according to the following assumptions: 1. The crowns are cylindrical in shape with uniform bulk density throughout. 2. The proportional contribution Of each size class is uniform throughout the crown volume. 3. Below the crown base (plant height minus crown length), stem taper is represented by a paraboloid of rotation with altitude equal to estimated total plant height and diameter at a height Of 15cm equal to the field measurement. The vertical distribution of Shrub layer loadings for the bulk fuelbed was obtained by summing the vertical distributions for individual tallied stems. 69 Crown Fuels As for the Shrub layer fuels, the analysis of crown fuel inventory data required auxiliary estimators for total crown weight, distribution of weight among components and size classes, and vertical fuel distribu- tion. Methods patterned after those Of Sando and Wick (1972) were employed. Weight Of Living and Dead Crown Components Crown weight regressions were Obtained entirely from the plant biomass and forest fuel literature, exploiting published information from other regions and for exotic species where necessary. Table 9 displays published regressions used in this analysis for the weight of live and dead crown components.l/ Also included in Table 9 are reported coefficients Of determination (r2), literature sources, and species for which equations were originally developed. Dead crown weight regressions were not available for the pine species, northern white cedar, paper birch, and red maple. For the conifers, equations developed for taxonomically and structurally similar western species were adapted from Brown (1978). NO attempt was made to differentiate between living and dead components Of paper birch and red maple. These compromises, and the fact that the equations of Brown (1963), Young et_al, (1964), and Ribe (1973) were derived from data collected some distance from Minnesota, introduced substantial uncertainty con- cerning the canopy fuel models. However, their use was confined to evaluating potential crown fire behavior under extreme burning 1/ _' Tree crown materials are defined here as living or dead foliage, twigs, and branches less than 7.62cm in diameter. '70 .Ameoae x0e: one ocean sore must «0 mansanca sumocootm ooam_saso:o . \a moaooam uncensoo .meoaeooam mm. sano.ao memo.o+omee.ao ammo-on: ozone dance muoeno Estes; ton. .voose.mm tum 6:50» so. oaom.o=ma~o.~o emao.ouoo~v.o=asoa.~o mmmo.ou3 cacao Hence Estes; t.n< capstamnam o~zomm AvomH..MM amm ocJo> hm. Oeom.o=mfimo.~o nmao.olmo~v.oroso~.~o Hemo.ou3 cacao deuce atfiarsmaam soroum mobwo~éesaa ostnom Ammoavnodsomonmsom a masooq no. ovno.mo «moo.o+maoa.~o mnoo.ou3 czouo o>wq ombmoNSEmLo or~zmom AmnmavooHSOQOmnoom a masoog mm. voom.mo mmao.on3 :BOMO dmuoe outmONSSmao QSNSAQL gaseous sexes Amemsecaoum ma. o.mo mace-cu: cacao ammo oeNOQrofiwboo ofizg~ Anomavuoxo mm. oo.~=¢m.~o «moo.ouom.amav.mo mmao.ou3 :3ouo deuce owwoormtwuto owzga oeuota2.ooom Amemaexoaz a ocean om. mmmm.ao mom~.ouz cacao o>aa 6:0wLSE mek \W oocom om. ~m0m.da hovN.ou3 c30uo HouOB ozomkie 52mm GNOOwuzgz mzrwm Amhmuvc3oum om. whoo.mo mooo.on3 czouo flown assures agree Avoaso.mm..mm mazes mm. o~om.ozma~o.~o emao.o-~m~v.o=aeos.~o Revo.ou3 cacao sauce «septum use»; emotmtzoa carom .msmaeczonm am. osmm.~o mooo.ouz cacao sumo omGCNooa extra Anomavcsoum am. O mom.oo moo.mnz csouo o>aq enormous nsrwi omaaumruo . inseam: o>aoemm~.o desserts ozzsm Amhmavcsoum 1:. n3 csouo oooo soc.mmmn_. inseam: u>aoeim~o.ou o Node-o. orommxroa osrwm Amemflvxoaz a cocom Hm. o>~o.mo HmoH.ou3 csouo o>aq growoxrab mzzua masseuse amuse Amsaeexoaz 4 ocean no. so4o me-.ou3 czouo o>aa SmEdmNUQ mmuQV \N oocom om. mvm.~0 mvmd.ouz :30uo HOOOE UdenNUQ mmwbfi mowuomm uoohnsm ouuoom H coduofiwm scamnoumum acocomsou moaowmm um Rousseau Beam 9 .mx cw usage: neocooeoo n 3. .mucocomeoo csouo noon one o>a~ ecu moouSOm one macauoooe copmmoumvu atoms: hue .Amuouofi :« anode: menu a = .50 :a anode: ammoun .m canoe 71 conditions-—when even relatively large errors in fuel inputs are of small consequence. Use Of relationships derived elsewhere to represent BWCA species with similar form and growth habit appears reasonable for this application. Because variable radius plot sampling was employed for crown fuels, the crown weight estimates for each tallied tree were expanded to areal loadings by: L = 140.45 W/D2 4.15 where: L = crown component loading (gm/m2) W = crown component weight for individual tree (kg) D = tree diameter at bremfizheight (cm) For standing dead trees, an occular estimate of the proportion of the crown remaining intact was used to calculate an appropriate dead fuel weight from the crown weight regressions. Size Class Distribution The combustion of standing tree crowns, either by torching or continuous crowning, notably involves little but the finest fuel elements--principally foliage (Van Wagnerl977). It is important, then, that these crown components be distinguished from the heavier, less flammable branches and limbs. Little information is available concerning the size distribution of crown materials for BWCA species. Data for balsam fir and black 1/ spruce were provided by Sando—-, with regressions for quaking aspen 1/ Unpublished secondary analysis of data from Sando and Wick (1972). 72 size class distributions adapted from Loomis and Roussopoulos (1978). For other species, reasonably representative estimators were sought from studies alien to the BWCA. Selected functions for live crown size class contributions are given in Table 10, while corresponding functions for dead materials are given in Table 11. These relationships were used to apportion the live and dead crown weight estimates among component size classes. Vertical Distribution The proximity Of crown fuels to the surface fuelbed and the bulk density of materials (especially foliage) within the canopy space are Of crucial importance to the development of torching and crowning fire behavior (Van Wagner 1977). A geometric approach similar to that described by Sando and Wick (1977) provided means for modeling these features. In this model, the vertical distribution of live fuel materials was represented using the live crown dimensions and shape designations assigned in the field. Crowns of individual tallied trees were considered uniform in bulk density Of all components throughout the crown volume. With this assumption, the model vertically distributes live component weight estimates for each tallied tree and sums for all trees in the sample. Dead crown components were distributed within and below the live crown according to the field records of the height to the point where unpruned dead branches predominate the hole, and the number Of dead branches tallied by size class. Size class weights for tallied dead branches, regardless Of species, were computed from equations reported by Loomis and Roussopoulos (1978). These weights were distributed uniformly from the base Of the conceptual cylinder containing unpruned '73 .lmsase sea: 6cm ocean soup some no nonsense >umocoomm omcmaanaacs omazuocuo . a ONOOmoruE msswm Amhmavcsoum pm. «3 HH+H+m EOOHAQ . Q ahoo.ouo A omo.a oNoomo205 ozewm Amhoch3oum am. AAm.on vnoo.OIv~m.ovu3 H+m owooweros warms Abeoaecsoum mm. o Omao.ono A 0mm.OI3 m areasuo msrwi omfizuozuo . cassette; carom Amnoavcsoum mm. u: HH+H+E vm.a~o . o a mam.o a NNHO.OI obstetrcm carom .mhmavcaoum 1:: AAHO.O+O nma.ouo mmm.ovu3 H+m 86230294 638.304 83388.5 mm. o 4.3.010 A @3613 m soot-some 8:84. . omwsuonuo . A obscures oarwm Amno~vc30um mm. «3 HH+H+h . EOOHAO . JAG mmoo.oumvo.H. dokooroo oarwm .mhmavcsoum oh. AAQ bmoo.o|hhh.ovu3 H+m assesses serum Amemavczonm as. . alo mace-a-mmv.ocnz a steamerso manna omwzuonuo . A moors.m moose .memaeczonm «a. n: HH+H+E aov.hAQ . 4A0 mmoo.OInmo.HV oesdoNoo moves \M modem 1:: A mam.ou3 H+m cmsoowob oemo< \M spasm 1:: a mnm.ons m sosomNob some: mmsoomw uoowSSM oousom -hm cofiumsom nobowoouo mmnHo osmm moaooom o>w~ aduOu I 4 .Amuv anode ouqm uo anode: uao a 3. .AAEV anode: menu a 2 ..EO. anode: uneven um Rousseau menu .mommoao ouwm >n armies >st :30uo o>fi~ new moousom use maOuoeaumm gouaumauoum .o~ manna n o .Amzv unwed: :3ouo '74 .Amhmav x0e: oco oncom scum sumo mo mamxaocs >umocooom nonmaansmca \A 0| 0| 0 I " + + otmemtmisa emzsum AeomAe As no 0:50» Ao quo.~-:qomm.mo Ameo o 3 HH H a mmpwoNrsmLo maosmom AmemHVmOASOQOmmsox a maSOOA In: AmmoAAOmucsouo o>AA AoOOBVh>.OIO mm.Au3 AhacovH . . I 8.3 Share a zoom Eating 3.38 28:32 8 2:48 88 013 a a. N mm. omw3uozuo . Q vmoo.o oi.o.m u: .hHCOVHHH e e \ I monsoessste esNaxom AmsermvoooOmmsom a naeooo om Boo mane mWNCiao mon o e o a o mm omasuocfl ovno.mo wmoo o u: AxAcoeHH a o . s . mmtwo~semts osasnoa imeermoAsooommsom a assoc: no 50mm oAmo acme-Ho mmmo o - . . . n3 AsAoovH outwoNseots ms~snon AmnmAemoASOQOmmsom a masooA mm Aves-Ac mon o - - . o .0": m omnwoNSEmso msNaaom 38.83853 38600 Amhmavmogomonnmsom a 3500A no NAOAJA mnoo omfi3uocuo . A cooomNA onset Amnerczoum mm. u: HH+H+a Eovéno . a A we; a mmoo.o: . e A omh.ou3 H+m assume; sense AmnmAeczoum mm. o mmoo o: . . a me a to Soto oarse sesame; oases AmnmAeczoum mm o mooo.o-0 A 5A6 o 3 a .N u at. A omazuozuo . A 3:858?» .08....- 632565 8. n: :31 EOOHAQ . AAQ Hooo.oammo.av 63.5.82 embed \w Ono—Adm 11: A mmm.on3 H+h 853.82 85.05 \I ousom 1:: A serious .m 38.85: emu-Am e e H moaoomm boonbsm OOROOm u COAumawm Hobowooum mmoao swam newcomm A.ucouv .Cq panda 75 mtwo~zsmsu msmsaom AmhoavaAsomOmmoom a maSOOA nun omowaouucsono o>AA Houoevnn.ono m0.au3 A>ASOVH $609.- owfiéoo 83:32 S. o ammo-on: m 39:0 0mvn.a n 4 ul all a o o " Estazh «mow Avomav As no snow on 00¢m.0:0~00.00 Ham 0 3 HH+H+E ofimostmso ozszom AmnmdvmvooaOmmsom a masooa In: AomsAHoHucsouo o>AA Amuoev O mm.Hu3 A>Acovu no.0: Exams; 3804 Annmavonam mm. 0 mv~0.0u3 0 Essex; some mA00.A mofloomw bomfimsm ouusom Nu coausowm noboaoouo mnvo Oman newcomwl A.uc000 oOH ”finch—- 76 oma3uozuo . Avmvo.0|o\vmm.mvu .05 .mm onDtAVrom wszw~ .meaAvcsoum 00. "3 .m 50~.05NQ . U v00.0 operate; mxzmm omw3u0£uo . U meteoroo maze; AmsmAvc30um he. I00.0m”0msom.mm . o ma.0 o U mas-N n3 A+m 500.0mmo . O 0m~.0 omfi3uozuo . U 3.82100 85.: Ame-03:30.5 3. 500.3 00.06006 . 0m2.010 O m2..~ u: a Eom.om~0 . O and-0 cromnxzoa 032mm oma3uozuo . U 8.533 $32. 185...- --- "z 3:: \A . Bow-mama . A0 vao.0umv~.avu ona3honuo . U UNSUQNUQ auwQV IOflzom Ho. n3 H+m \A so . 10 . 0 O . m AAA mmm.n mom oquuozuo . U mwwzosm woman A0>0A0c30um he. «3 m 500.5Nd . o 0 one. ouEUoN80 samba Q naho.01 moaoomm uooHAJw condom Nu scaoomwm ROAOAUOHm mmoao ouwm newcomm .AAEV unmaoz wow» u = .AEOV unmwoz suntan no Rousseau some n a ..mxv Anode: c30uo poop AmuOu u U .Ame anodes >uo n z. .mommeao omen >2 utmaos xuo csouo peso ROM m00u30m use muOutEwumo Hoowumwutum .AH oAnoB 77 wmw3uwzuo . U agcowNm unzgh Ambmdvcxoum me. u: H+m Eomémma . m o mmv. Q -0.0I . wmfi3uozuo . o mucumNm Ufizgu Amhmavczoum hm. u: m 556.8 . $30.05 02.3 mfigrmfimnno 3.35. mmfizhmnuo . U UrdeUE cmmma oucmm III "3 HH+m \m auo.mamo . An mmoo.o-omm.~. 6:323: UNIS”: 005m mo. 0 U hmooé .I.3 Tim \M a 28.0- mmazumfiuo . 0 3555.398 6mm”: 333.598 2.. u: m ..BoéNQ . o o 03.9.m 38.56% 9...ng mvodI mma3honuo ~ 0 UNOngzos 332»; AmanVCBOHm NH. HM“: H+m Eom.:No . E moo.OImmo.:. 3Noowuros mzzwk Amhmavc30um av. omv.o Q U mavm.a u m m:&okum mxzmm mmflzflmnuo . a HmaodImooévu smeamfirom mxzr~ Amnmavczoum me. u: aka Emmdeo . O 00.0 mmaommw nomflnsm mousom Nu ceaumnmw Mauoflcwum mmmflu swam mewowmm A.ucouv .AH manna '78 mm~3~o o . +HH+H mmmao o>fiqu . :u AHHH vnmm.m D U mmao. "3 HHH+HH+H mmfiwostmLu mesz~ .mnmavmoHsomommnom a mflsooq III . . + + mmm o m>flAvI a U v-o.o sum mmma AHHH HH H H omoq.~ omw3uo£uo .AHH+H mmaao 0>MAVI a U mmmo.o .vnmm.m . "3 HH+H mmfimowxsmau mszmok AmhoavaHSOQOmmsom a mflsooa mm a . I . mmn o o>flAVI a o mooo.o Um oHAo AHH+H A boom.~ Q J . I. . I~u miN: 5% . o o> I a u mmoo 0 I3 H mm~wsNxsp amfiwo~zsmku mszmsa AmhmavmoHQOQOmmsom a maaooq mm AH mama any hem.a - . A acaommm pawnnsz mJHJOm «u :oHuaswm nouuficvum mmmao muflm mvaamnm A.ucoo. .HH wand? 79 dead branches to the base of the live crown. Additional dead material (from Table 11) not accounted for by tallied branches (if any) was distributed uniformly throughout the live crown space. Resulting representations of vertical crown fuel distribution were used to evaluate conditions required for the initiation of torching and crowning fire behavior. Inventory Results In all, 210 subsample units were inventoried within the pilot study area. Figure 9 shows the locations of land survey sections in which sampling was conducted and Table 12 indicates the distribution of sub- samples among management zones and Grigal-Ohmann (1975) community types. Thirty subsamples were inventoried in one of the ten selected sections that was drawn twice in the sample. Logistical problems, though, pre- vented completion of the remaining ten subsamples within that section-- hence a total of 210 subsamples. The distribution of inventory subsamples between the two management zones closely reflects the actual pr0portion of the study area within each zone. Sample allocation between community types differs little between zones, with the aspen-birch type accounting for 86% and 76% of the Portal and Interior subsamples respectively. Overall, 80% of the inventory subsamples were classified within the aspen-brich community type. This striking predominance of broadleaf types is apparently atypical of the BWCA on the whole (Table l), and is likely attributable to the 19th century fire history of the study area (Heinselman 1973). For all 210 subsamples, there was no evidence of fire disturbance since the major fire of 1875. Only four subsamples showed evidence of 80 Figure 9. Locations of land survey sections randomly selected for fuel inventory. 81 $8: 52:8 v.5 13.6 \~ v v m m h v awn m a «Na om OHN .34 III III III III III III III III II ON ON 3.». Zno v N III A v III III M." III III on II ON 3s. zmo nu III III III III III III III III Om II ON 3o ZMO 0 III III v III III III III III II 2 on :m 23 .3 ~ III a III III III III III II cm 3. 3m 23 3 III 9 III III III A v III ON II ON 3m Zen ha III III III III III III 2 III III on II 3 5 23 n III III N III III III 3 III III ow II an :o :3 on III III III A a . N S III III 2 II 2 3... 23 2.. III III III III III H mm m A v on am so 25 ow gaumm mImn. .3000 3 oz unotImImnm on: cum szuamImImt u: 151:. zonamIchB. sunumlfm namImLTez mafia—i . :30... com \— vs xuwgsoo an . I I econ E m . I 62 umqnumoa :53 £093 >u«:=§a8 can uvcoN aces—swung. ocean 313 nagging» no :ouunfluuafiu 3.3 0:39.35 new vouoo~am «Goduno- >u>n=n v3 .«4 cans... 82 other noteworthy disturbance in the past 25 years. Two were harvested in 1971 within the Portal zone, while the other two, also in the Portal zone, were cut in 1956. Despite the area's relatively uniform upland forest cover and disturbance history, though, several stand and site measurements showed considerable variation among inventory subsamples. The ages of dominant trees ranged from 32 to 117 years (Figure 10a), with a mean value of 80. Also quite variable was tree basal area. It ranged from 3.4 to 34.4 mz/ha (Figure 10b) with a mean and standard deviation of 16.6 i 6.3. Site index averaged 14 meters at age 50 and ranged from 9 to 21. Floristic Composition The noted over-representation of broadleaf community types in the fuel inventory sample indicates that the chosen study area is not representative of the BWCA in general. Since hardwood types are notably less prone to severe fire behavior than conifers, it will be difficult to extrapolate results of the pilot program to the BWCA in general-- where coniferous types are more prevalent. Comparing the floristic composition of the fuel inventory sample with that of a more extensive survey of BWCA vegetation offers further detail on this anomaly. Ohmann and Ream (1971b) report measurements of basal area of trees, stem density of tall shrubs and tree seedlings, and percent cover of low shrubs, herbaceous plants, and ground cover species for 106 randomly selected natural stands throughout the BWCA. Similar values are reported by Grigal and Ohmann (1975) for 77 of the original 106 natural stands combined with 50 randomly located stands in areas that have been logged. The natural stands of Ohmann and Ream (1971b) should be directly compar- able to the fuel inventory subsamples of the Interior zone. Furthermore, .muoam >uoucw>qfl stw an mono Human wwuu A.n can .ucmcomEOU huoumuw>o unusfisoc we won A.m Mom madumoumfin :oflusnfluumww hufiafindnoum Amn\mav «mud Hammm “mummhv wm¢ vcmum om ow omH ooa om om ov om saIdmesqns AloquaAuI go 1u3319a Ios .n .oH museum IOH I ON Iom Iov .m satdmesqns KxoquaAuI go quaozad 84 since the prOportional representation of natural stands in the Grigal- Ohmann data set (61%) is nearly identical to that for Interior zone sub- samples in the study area (59%), the values reported by Grigal and Ohmann (1975) should be comparable with those for the aggregate fuel inventory subsamples. Table 13 summarizes these values by vegetative stratum for all four data sets, while Table 14 shows the percentage contribution of individual species (Ohmann (1973) calls this relative dominance or relative density). Note in Table 13 that tree basal area is considerably lower for the study area than for the extensive BWCA. Furthermore, it appears that tree seed- lings within the pilot study area have been largely displaced by tall shrubs. The low shrub and herbaceous strata cannot be compared directly, since percent cover was not measured in the fuel inventory. Comparison of relative dominance for these strata (Table 13), therefore, may be questionable. Ground cover percentages in the study area are only slightly lower than found elsewhere. The most striking observation in Table 14 concerns the relative contribution of aspen and paper birch to sample basal area. In the Interior zone of the study area these species account for 47%, and they account for 55% in the aggregate sample. Respective values for the BWCA in general are only 17 and 18%. Coniferous species make up much of the difference, reflecting the noted contrast in community type distributions. Except for the relative density of red maple and balsam fir tree seedlings in the general BWCA stands, though, and some minor rank reversals in species density for tall shrubs, there are few addi- tional differences among the data sets. Even in the herbaceous stratum, the small variations apparent here can be partially attributed to 85 Table 13. Summary data for vegetative surveys in the study area and the BWCA in total. Vegetative Pilot Study Area BWCA in Total Stratum Interior Zone All Subsgggles Natufal Stands l/ All Stands 2/ Tree Basal Area 17.8 16.8 30.8 30.9 (mz/ha) Seedling Density 4999 7275 27,308 46,169 (stems/ha) Tall Shrub Density 60,772 63,446 25,358 26,042 (stems/ha) Low Shrub: t cover 5.6 5.9 gm/In2 4.7 4.4 Herbaceous Plants: \ cover 59.2 31.7 Whiz 19.5 13.3 Ground Cover: t cover 15.8 14.6 23.4 22.6 Data from Ohmann 2/ . Data from Grigal and Ohmann 1975. and Rean 1971b. 86 h.on hocuo o.~n nozuo TN Steamer? 6.38636 n .m 6:28.69? 6.2.6636 n .2. 350 n .3 :50 a.mH .aaa Exrdnowh a.mu .mma EJKULowN a.ma .maa 6wrofi6~Q m.n~ .mmn 6mrofi6Nb uo>oO Gaucho s.mv mkubmzrom Esmoosst a.ov «Lebesgue EswwoLst v.o~ mLmbmxxoo sammozst mm.- «Loaonxoo EsmooLst ~.o~ ~28 o.mn nonuo o.~n H05c n.m 6u~6oton 6w=ourmNo n.6n nocuo m.m 61~66Lo¢ 6mrourw~b n.m 6w~6mnoo 6uroorwNC ~.o Earmesos pawnmtmum m.m earmNuscG sawwmtoom 5.0 amormfi6fl6o msrnob v.n 66:666K66 E:meuz6~6> v.- 6m~66mbsr 6mN6L< v.na 6m~6owhsr 61~6L< oucdam nsoooonuo: ad 63.666266 €565§~6t m6 36:36:66 6356.» 72 dds 5.666663 0.: dd» 5666663 mév nsNNSxQEIIEE L33 m.: §~me66t6cs L36... 0.: 33369865 L33. m.v~ 3353986: L33. 73 2.50 52 35o 3: 350 6.2 h.ofio v6 8895626 £265ng 92 3683626 6.26536. 6.3 dun .33 0.3 doc 93:: m6” €66.83 62.5.8.5 6.3 £66.33 62.55.? 7: 263:3 63.92me m.$ goon—=6» 63.5.3.5 .335 so.“ m.o~ EsmNnxwuosmru Esmrm666s o.o~ E3m~mxmuosmzd Esmrw666a ~.mn .aaa sawrwboos h.mm .daa sawrmoous w.a~ nonuo ~.o~ nonuo m.m~ nonuo ~.n~ nozuo m.m smomto carat ~.- chowno oxrtw ~.H~ ESu6owQ6 L661 0.0 Esuoommu L66< n.m~ Exoscmmm Loot a.mm Eswsomum L66” o.o~ 6Q6mL6 mxrww m.o~ 6amvgo merq unauzm Hank v.3 663266 3:989 0.: 66:53.0 3598. vém 6.3566 §~P~ou v.06 693.58 6336.89 6.6 350 N6 633266.866 6.3:: o.~ .350 m.m Essex; L666 o.o~ 6L6HmLma6m 6~summ v.2 98.3.82 68E 5.3 56:; .83. Ens safes a.m~ H650 m.o~ nozuo m.o~ 66566~6n mowaw o.m~ owNoarmvvooo ufisss nonaavoom soak 6.3 68638 monk? 0.?” 66.5366 66.3w, a.m. 663345.68 wisdom o.m~ 6656366 66.23. v.~o ESLQSL Loot o.ov EJLQSL Lzow m.e~ sumawtmd6u 6&306m n.0n 666m6~sswgu nstmbm m6 .350 0.3 .350 «.2 350 m4 :65? .866. ~.n £3.33; smut n.~ 5.5:; 566‘ «Km 636686666 6.63:3 «.6 66.533 6683 «.5 633666.866 6.32: v.2 350 ad 645956 354 m6 636656866 6.33% «4. 6mE636Q 3.2? m6 6.8833 66?? m6 66.56366 6636 28 Ion .m A. 0.0 6:66.36 2.er 0.: 6:66.56 6.er 0.2 656.8385 636.5 0.3 6.6.26: 6.6.6.5 «003. o.- 6:6mL6E 666wk «.ma 6:6»L6E 666mm n.n~ 6:6wk65 6wtwl v.m~ 6:6m6xr6o osrum 1: 68.3.6266 9.5mm odd 666.63.35.26 63:63 6.: gathered 336m mg: Emkwnga Ssumm n.6H owbwoussogu as~sdom ~.n~ 626m6xr6a warrw H.5m mmpwoNssenu ms~snom 0.0m owwwostonu uSNSQDR .358 2.333 330% .358 «>333. 3825 3.58 «>323. .38on .158 «>333. «Seam 3.3 .2220 65. 1.628 323 5.3. 65. 2:955. 3&5.QO 2< Eflo «can 3235 .653. 6033 65 H3322 50m :5 3.8% 15.33. <03m ouuucm 60nd xvsum uoaum ucoaummmMV chum Edumuum .Hauou :« <03m on» 6:6 nous >6aun ozu uOu isuauun >n >ancoo no 00:6:«506 «saunas o>uuaaou .vm OHAIP 87 disparate measures of species abundance used in the two sampling efforts. Apparently, the study area's major distinction is its comparatively low representation from coniferous tree species, possibly due to the loss of seed sources in the fire of 1875. The community type and relative species composition of the fuel inventory sample, along with the visual appearance of the landscape, leaves one with an impression of a broadly uniform, but floristically diverse forest community (often called the "Minnesota mix") occupying most of the study area's upland sites. The "patchwork" character of other BWCA landscapes, on the other hand, could present very different fire management problems when prescribed natural fire programs are initiated elsewhere. Fuelbed Properties In spite of the floristic homogeneity of upland communities in the study area, biomass and fuel loading estimates are quite variable among fuel inventory subsamples. Table 15 shows the mean loading of major fuel components in grams dry weight per square meter area, as well as the maximum value, minimum value, and standard deviation by community type (Grigal and Ohmann 1975). Similar values are given by fuel size class and condition (live or dead) for surface fuel materials in Table 16, and for crown fuels (above 1.8m in height) in Table 17. In Table 16, H and F layer fuels have been deleted. Total loading of organic material ranged nearly an order of magnitude from 3186 to 17,220 gm/m2 (Table 15). Of the ten fuel components listed, humus shows the greatest range of variation among subsamples--from 172 to 11,219 gm/mz, while downed deadwood had the 88 mum .ram mom moa mad NHN sum dmm o m 0 vv u had we ON em mm 0mm :a: cow mvh mac mam 5mm aom ammo «CNH 0mm x3: 5mm 0mm 0mm mmm mmm 5mm ore vmv mmm m Aw>fi4v naducm vow Neva Aden mxm owed mmm awm 30am 0 m vow nnaa mmv mom mum Mom mm nmv @mom ca: mmmq mm_v Hmmm ommm mem mmma ammo Nomm cmom xv: coczcuva m>v~ momm mood mwofl omoa nmaa mama mama omen m 13:30: ma v mv NH 0 v ma Ad 0 m NH H mm mg 0 v m x on :4: hm OH hmu av em mg on Nq om xv: umbcaum ma o no vm ha m mm mm on m mzomuqnuv: mo mnm vac mg Om om vm and o m a. hm mod Ha 0 mm o 0 HA ca: Non How mmaa wm on 09H com mmm Ha xv: uo>au on 0mm mmv em we m¢ no mom Ha m m::o»o OHH Om vmm an mm and and VHH o a mom mma am Nam mhm mam mHH mo com :«2 omo vmn mmh ovv mmm mum omoH o_v com xm: 9:. my: Sq Sm 3m 9% 03 can com m 3&3 a wax mom mmm whH mom can va own 0 m hmm 09v mmq was hem mam mma mma now ca: mmom om» c0©m mmaa OboH mova mmom omma new xv: I: one 82 m3 «B So mm: Sm So m ~33 m Hana mamm Nmmm mmm mung mood Nmma mam o m QHHN noam ovmm obnm mead mmaa «ma mama me :«2 mmao mowo mumna moan obvv Homm memh somv mmw xmz 33 3% 2mm $3 $3 3:” Sum 8mm mwm m 9&3 2 pages mummwm yum mEmum o~mu mnmw Havoc 2 oz mmot hummm mafia 0mm m3|uwmnmw< nah ammnmh souflmncmmc Loufimuuwm awhnmu9 audacaaoo Amnmdv :cm5503_akuu Hoax .mm>u auwcdfisoo vcd uzmcomsoo Hmzu >9 Am“ macaucw>oo cus©:mum can .momcuu ~Ax. mcoue ozavmoH Hana uo >hasssm .ma easy? 89 may“ vaoa maom mmmd 05mm Hmmm mmma mpVH o m mmvo namb ooood mmmo mmmm move wamm mavm mvmm :fi: moema mumaa ONNBH vmmm whama menoa mmoma mvmoa avmm xmz . one: ommm vmmva ommm mhmm Nmom mmz. vmow mvmm m mucmcomfiou Had hon m1 Cam mam mm mm mma man 0 m mmm mmm mmv Ava nmw mmm mm Hma nva :«z mooa can hmaa mmo Hmm vow mom mmoH nva xmz moo man now nmm oov hoe Hum mom hca m Aomwov mmwue ¢~ma mmm com has ovm com woo How 0 m Nada Nvma mvoH mqaa mvma mmwa mmm Omod mmoa :Hz mmov ovdm Nmnm maam ommm vavm camm momm mmoa xx: mmmm moma vmam mvmm aomm ovmm mafia ONON mmou m Am>fiav mwqu mN Hm mm am av ha mmH mm o m m A 0 mm AN 0 A N Z. :M: do mv voa Oma ova mm mwma nma an xv: ma om an up we ma am an an M Aumocv manusm umuma muummm uwm‘msnum oumw mama ustwt ? oz mmo: mumwm mcfim 6mm mmrufimnmm¢ ham ammumn nouwmscmm< zcufimuufim uMmImu¢IB xuacseaoo Amhmav :cmszOanmauo Hush A.u:oo. .mH m~nme 9O VN Nm ma mm ma 0v mm on o m ON hm Hm ov me $H 0N mm QNH :42 hh DNA mm 00H om moA mam mmd oma xv: Av wm hm mod v0 #6 mo mOa emu m H @H mm mm m h o NN Om o m cm mm ma had vh NOA h hm m¢ CH2 mad ma hHA mad mm mad wad VNH mm Xaz com mm mh dad mm Odd mm mm m$ m h @000 o o o o o o o o o m o o c o o O o o O CH: O O o o o o m C O xmz o O o o O O o o o m >H om mmm mmm ON ¢N Hm 00H em 0 m h 4H 0 o o O o o h CH2 OMH Amm mvh vv mo 00 mama mmm h x02 he and mum ed mm mm mm mv h m HHH no mad $n~ hm . om mo mod hhH O m mm NQA hom . me no oh ma om hmm :flz mmm new mom hma mmm VMN mam woo hmm xmz mNH MOM mam @NH Omd Oha mmm mmm hmm m HH vN Mh ow ma mm mm . mo 00 o m m mh N 90 ON NH Ca ov ONH CH2 #0 «mm Cam MHH mam mod th vmm DNA 352 mm Onm mm mm Nh ow OHM mam ONH m H Om Mh om v0 mm mo Nm no 0 m vv VOA ow Om vv hm vm om cod :fiz mom vow Hmm mam QMH mad 0mm mmm oo~ xmz mo ova 00A had hm voH NHH med OOH m m w>wq uwuwe wumnmm gum,memuo 93mm mumn 288 3 oz mac: a-mmm 95a cum mzLEumwc Sm 1min :ofimémmc nofimLE uEnméé: aha? >uwchEou Amhmav ccmszOIHmmauu QNam ceaufifi:oc .om>u huficsaaoo can .cofiuwccoo .mmmao wwwm >3 AS m.a 3040p. mmcfitmoa szw OQMwusm uo xumsssm .oa wandfi 91 'n MNM mama HOAH aoh mood New Omm omNH o hdmm acmm vmma NNha mmaa Ommm mmm heed Hmam CA: hvom nvmm mmmm mhmm hmmm hNHN Nmmo mmmm Hmam x02 VONN mvmm mmam hmvm aOhN mmma homm oooN Hmam m Hq< vmm 0mm Omma Nam hvoa mvm Mhh vvm o a mac mvm . mmm hOfi hON vow 0 00a acmm CH2 Nama :mNN Omen mmvm oNxN ama amHm whhm Acmm xmz vo- new“ mama HOHH mood ovh Nam OMHH HOmm m >H hhm 3mm vmm avm om oha ham ~QN o n ma «AN mNH oma 0m~ NMH A Hh mam 2%: mhv . mmha mmm ohh ham mvm hand How mmo xv: hmm Hhh Mom Ham vmm ham m~m 0mm moo m HHH hh Nh omu ho ~@ AHA vmm QmH o m mam ma Nm th mOm mom $m v0 owe :HZ com haw mac mmv hem hnv th hmm mmv xmz van oom Hmm mav hmm Non Nmm mvm mac m HH uwuua oumnmm ummwmamum woumm mnmn Havoc 3 02 mac: human ozwm cam manufimnmm< ugh ammnmh nouumnqmm< nouwmuuwm uqmlmuH moH mom OOH OOm VhH mmm mhm Hmm O m cm NO Oh man mhv 0mm em OOH vhm CH: mmv 0mm OOO HOHH mom vaH Nva mam vhm xv: own xnm MHv Ono mOh Ovo moo mwv qhm m HHH mmv mHH mmH mmm OOH mam hmm OHN O m Hmm van hHm VhH Hem mam hm OHM mmm CH: mmHH mom HHO Hmm MHm mam mONN mvoH mmm xv: mhh Hmv MHO hvm moo vmm oov mow mmm m HH hhm Hm ONH NON mmH mmH VVH hMH O m Hmm th mOm OOH mmm mmm hm mmm mom CH: OHOH 00m mvo mHo owm Hmo HNOH hmo mom xmx vmw mom ovv mmm Omv mvv mam com mom m H mom ONH mOm Omm OOH OOH OOH mmm o m nmv Ohm Ohm mmm hmm va we hmm th :H: MHmH mm: mma Nam mmh omh Ova hmm mhm so: vaH omv mmo mmo mvm mmm th vow mhm m m 0>Hq umuofi oumzwm uwm‘msdum ooamw,mumn havqu 3 oz mmo: mymmm mafia own m3-uHmme< uHm Hcm-mh :uuflm-cmm< :ouflnuuaz uHmumucumz 0Q>B >uHCdEEOU HmhmHv CCMESOnmeHuU mNHm :oHuHUCOU .vm>u qucanoo new ~coHuHmEoo .mmnHo oNHm an .8 O.H w>oan mmcHUmoH Hmsm czouo mo xumaazm .hH mHAue ‘93 NamH «em mmo cmoH ~mo ohm BHm «Ho o m mmMH nhmH mmoH mHHH ommH HmmH com HohH oomH :Hz mon whmw mnmm mnmm vomm Noam mmwm mmmm oomH xcz «hmm vamH comm mmom emhm ommm mmom ooqm oomH m HHa o o m m o o H o o m o o o o o o o o o :H: o NH VH HH H H HH H o xmz o m m m o o H o o m >H m mH HoH Hm oH aH HN we 0 m m m NH 0 oH NH o m mH =Hz HH Nm mHm om Hm mm moH mmH mH xm: o mH hh NH Hm mm mm mv mH m HHH on Hm ooH mm he av ch an o m mm mm hoH Hh noH mm 5H he mo :H: Haw moH «he «Hm Hmm NHN vom mom mm xv: qu Hh mmm hmH omH mmH «HH HmH no m HH uoums 0.2239qu 9535 ooumw‘m'mn no 3 oz mmoz human mcHa com mx-uHm-mm4 “Hm Hanna: cuuHmucmm¢ nuuHm1HHm “Hmum-<-H mean one ..*o HHH was» amumumsHo Hsou ousH mmHmEmonom mo msHmsoum HMOHHme:: unmoommnsm can scam mHonehm \w Izmusosomaoo oanu one umuHm on» no oson one umuHu on» no osmHm ecu sH muHsa msHHQEMonm .mmum mosum uoHHm on» :H muHcs >Houso>sH Hush mo :oHumswouo mucosomeoo HmmHosHum .NH 6.66.a bananas—IOU 45:08.3.— OZOOUG + 1+ + .u#+ fiw+ H + ++ + I +Hh + + I. + + 4 ++ #++ H.’ O a + + I + + ++++ # + ++ + N 1. 11¢ + o 0 * +£$+. ++tfl.v do .I + + a“ + + + 0 d + “ + ”7* # + 0 v # + +**++++ + O 1 ++++ .0. +... ‘+ .7 + + o 0 + + + + + O + ++ ++ O o + .¢+t.+ _6 o I + w + + I .I I + s l O o N + .I s IO 3 _ as o o N s s 0 I. ~ 0 a $ . O s s 0 o s WJOH L 124 accounts for 83% of the variation (Figure 12a), while the third princi- pal component accounts for an additional 4% (Figure 12b). Figure 12 also shows the results of the cluster analysis. Four fuel types were identified and designated types I, II, III, and IV. Groups that do not clearly separate in the plane of the first two principal components (Figure 12a) can be distinguished in the plane of the first and third (Figure 12b). Eighty-one percent of all sampling units were grouped as type I, 9% as type II, 8% as type III, and 2% as type IV. The number and plotted density of the type I samples reflects the visually apparent homogeneity of upland stands in the pilot study area, as well as the noted predominance of a single community type (aspen-birch) in the sample. The dendrogram resulting from this analysis also shows the relative sizes of the four groups (Figure 13). This is a diagrammatic represen- tation of the hierarchical relationships among inventory plots. At the bottom of the graph, each inventory plot is represented individually. As the clustering process begins, grouping of similar plots or clusters is represented by the merging of branches. The height at which the lines merge corresponds to the distance separating the groups in the principal component space (the distance function). The shortest distance function between any two plots was 0.0041, while all but eight individual plots had been combined below the 0.1 level. The final combination of clusters occurred at a distance function of 0.3722. Chaining, a common problem in cluster analysis that involves the systematic addition of small groups to larger groups in the clustering process,does not appear to have been a problem. Units of generally equal size were joined together at each level in the dendrogram. 125 .mwmhu Hwow msHmwo on somoso Hwosmumflov %UHHMHHEHmmHo mo Hm>mH on» one wuon owHHousw>sH esp moose mmHnmsoHuMHoH HmoHnonmuwHS on» mcHsonm Edumouosmo msHHoumoHU .MH musmHm g— ? FE __ EEgg:____=_F__________E______E__.______ £13; _é.EE===EEHFE==. >H HHH HH H "coHumsmHmwo oth Hash Iv- :OHuosom wocmumHo 126 As is apparent in Figure 13, the number of fuel types identified could range from 210 to 1, depending on the level of intercluster dis- tance chosen to define them. A distance function of 0.27 was chosen in this analysis (dashed line in Figure 13), primarily for operational convenience. More than five or six types would be cumbersome for fire management decision makers. Fuel types I through IV, described previously, were identified at this level of dissimilarity. Discriminant Analysis and Relationship to Cover Type Following clustering, the data were subjected to a discriminant analysis (Cooley and Lohnes 1962) to allow classification of uninventori- ed stands on the basis of readily observed site and stand characteris- tics. In this procedure, a set of discriminant functions was computed, based on the composition of the four fuel types. Undesignated stands can be assigned to the most appropriate fuel type according to the discriminant function scores. Forty-two site and stand variables were subjected to a stepwise discriminant analysis that, at each step, selects the variable that minimizes Wilks' lambda (Rao 1952). If the matrix W is defined as the within-groups cross-products matrix and T is the total sample cross- products matrix, the elements of W and T are computed as follows: 9 Nk _ _ w.. = 2 2 (x. - x. )(x. - x. ) 6.18 13 k=l n=1 ikn ik jkn jk N t..= Z (x -X)(x. -x) 6.19 1] n=1 3n where: g = the number of fuel types 127 Z 7? II the number of plots included in type k N total number of plots i and j run from 1 to p, the total number of variables Wilks' lambda (A) is defined as: W gA -+¥+ 6.20 This ratio of determinants decreases as the discriminating power of the included variables is improved. At each step in the process, then, the variable is selected that contributes most to the reduction of A. The coefficients of the discriminant functions produced by this algorithm are simply the normalized eigenvectors of the matrix W-1(T - W). Table 23 lists the 42 stand and site variables in order of their contribution to reducing A. Maximum tree height for upland black spruce, jack pine, and balsam fir, the most prominent conifers in the study area, were by far the most important variables. This makes intuitive sense, since torching is dependent on the height to conifer crown bases, and crowning depends on the density of conifer foliage in the canopy. Other noteworthy variables were site index, and basal area of all conifer species. Resulting discriminant functions, when all 42 variables were included, were able to correctly classify 91% of the measured inventory plots. A relationship exists, therefore, between fire behavior potential and general site properties. Although 91% success is an impressive figure, it must be remembered that 81% success could be achieved simply by assigning all inventory plots to fuel type I. Furthermore, use of the discriminant functions to classify uninventoried stands involves estimates of variables that require at least brief on-site examination. For only 10% 128 Table 23. Entry sequence for the 42 stand and site variables in a stepwise discriminant analysis of fuel types. Step Number Variable Entered Wilks' Lambda 1 Height of spruce' 0.83992 2 Height of jack pine 0.69891 3 Height of balsam fir 0.66101 4 Site index for dominant species 0.64360 5 Basal area of all conifer species 0.62547 6 Management history code 0.60942 7 Density of dead paper birch stems 0.59597 8 Soil texture code 0.58320 9 Basal area of jack pine 0.56857 10 Topographic slope 0.55673 11 Aspect code 0.54571 12 Density of dead aspen stems 0.53521 13 Stand age 0.52550 14 Time since last fire 0.51273 15 Density of dead maple stems 0.50317 16 Height of paper birch 0.49515 17 Height of maple 0.49001 18 Basal area of maple 0.48009 19 Basal area of miscellaneous species 0.47520 20 Basal area of paper birch 0.47020 21 Height above water 0.46657 22 Height of red pine 0.46232 23 Height of white'pine' 0.45684 24 Density of dead balsam fir stems 0.45361 25 Density of dead jack pine stems 0.45041 26 Basal area of red pine 0.44758 27 Density of dead red pine stems 0.44207 28 Height of northern white cedar 0.43889 29 Height of aspen 0.43583 30 Density of dead white pine stems 0.43348 31 Basal area of aspen 0.43131 32 Physiographic class 0.42925 33 Basal area of white pine 0.42784 34 Depth to bedrock 0.42647 35 Percent crown closure 0.42495 36 Basal area of northern white cedar 0.42367 37 Height of miscellaneous species 0.42262 38 Density of dead hardwood stems 0.42168 39 Density of dead northern white cedar 0.42080 40 Basal area of spruce 0.42018 41 Time since last disturbance 0.41978 42 Density of dead conifer stems 0.41957 129 improvement in efficiency, the additional effort appears questionable. Hopes to classify and map study area fuels on the basis of general site and stand characteristics were thus abandoned. As a second alternative, the relationship between fuel types and an existing covertype classification was examined (Table 24). Since the classification of sites according to the community type classifi- cation of Grigal and Ohmann (1975) also requires on-site description, it was deemed unsuitable for this purpose. Instead, the vegetation type map prepared for the study area by the University of Minnesota College of Forestry (Appendix)l/ was used. An overlay of inventory plot locations on the vegetative type map facilitated assignment of type designations to each plot. A key to the type designations is given in Table 24 and on the type map (Appendix). Note that for all types, at least 50% of the assigned subplots belonged to the cluster representing fuel type I. By maximum likeli- hood, then, all upland types should be represented by fuel type I. Apparently, in this area the distinctions among identified fuel types are due to variations in fuel prOperties that take place within forest types, and in general, do not vary significantly among forest types. This observation suggests that a single fuel type may be sufficient to represent upland fuels for the entire study area. Vegetative type map prepared for the North Central Forest Experi- ment Station, USDA Forest Service, by the University of Minnesota College of Forestry using 1:15840 Aerochrome infrared photography flown August 17 and September 16, 1976. 130 m.~ v.m v.Hm c.0c— weeds HH< Hoes oHoa noHHn-comu< - In :- c.cOH mmv.o .u¢\=oc\=to EDHoex «Hon e hHu-cozuam Esmtoz quEHu3em osHm sees: nu -- -- c.ccH moe.o :00\2oc ErHooz cha HHNIQDSHLm E:Htez nonfiHutsm ech oonz an 1.. II O . OCH mcv . O :o<\..co 5.2.0: menEHu3ev. 30..» H5159: EDH H04: hone—H uimw cc HQ 0» .213 - H.4H - 5.me oa~.m gee ---- ----- -------- esac:z sensHuism ease uoHes - - - c.ceH cmo.c .oe ---- ----- ezHea: uwneHuicm «cam sex - - - c.ocH cee.H =05 ------- - ------ - -------- achmx eHom nouHcoo ecaHacH u- In - c.5cH mom.H .om\:to noon nonEHusem ocHa xoeo ssHoez nonEHusmm uHunooaumm -- - - o.OoH oom.H zoo u-u unoI-III-nnolu-I EsHov: umasHazem uHu-sUSHQm :1 n: - o.ccH o~o.o .o<\=uo Moon 0H0; suan-ccom< EsHoo: oHom uHu-oosudm - H.HH - o.em eaH.v zoo ---- -- assoc: 0Hoa uHm-eosuam n: In is o.coH Hoh.m :oe\:tm ESHooz mHon uHuuoosunu E=Ho¢£ uvasHurmm osHa x0e: nu - In c.ocH omo.o xnv\xtm ESHooz vaHHosm scoot Emmott honEHusmm ecHQ xoeo -- -| -- c.ccH Ono.o zo<\:om EflHocz HOAEHuaem coantcont EoHow: umHEHusmm vcHa sumo -- - In o.coH Ono.o .o¢\zom noon usAEHusmm zuuHAIcoamc EsHooz uwneHukmw cch some I: II u- o.ccH cmo.o :U¢\:mm ESHooz vHom soanlcech Eoutoz uwnfiHuiom 6cm; sumo -- -| an c.ccH Onm.o .oa\:om Moon opom suanlcommc szHooz quEHuxsm wsHo xtmo - H.e - 5.66 5HH.m =om - --------------- -------- asHee: aeoEHuzem ecHs some I- O.Cm nu o.om omm.m :H\.tm - --------------- swans ocmHm: Loon uQHEHuzew ocHo xoeo - - - o.ooH ue6.o .cm ---- ------ ----- ------ none .uwseHusem ceHa some a- I- an c.0cH mov.o m\.om menouzo :00: soon 0H0“ och xomo - - - c.ccH emm.~ gov ---- ----- echcz ocHHecm recto nu an - o.ocH mcv.o .tm tt-I-In Inn-auuua uHu-cosumm oboe ueAEHusmm OCHn-zoan-cozms - - m.m 5.Ho ~mm.m suoxzea eszez oHca uHo-musuem ensue: nonsHuaam rouse-cosu< us In an o.c0H mov.o .UO\:o< noon mHom HHu-oushmm ESHOQZ HenEHu3ew couHa-soanc -- I: -- c.ocH Ono.c som\:o¢ ESHow: HonEHuBmm ocHa xueo EflHowz HenEHuxsm zuuHo-cc&m< I- as In o.ooH Ono.o .om\:o< .uooa McAEHuzem esHa xoeo EsHooz umnEHusmm zooms-coam< a.mH a.mH 5.5 m.He ceo.o gucxgee esHee: oHoe eoan-coeme ezHcoz noneHuanm :uan-cwem< :- c.mm -- o.mh oom.H .o<\:o< noon 0Hoa soan-commm ssHoc: usAEHusmm :UHHn-Cwans o.~ m.m m.HH v.wh mmh.mm :o< nun-ill nun-until III-nulunuIII-nl fisHoer henEHussm couHo-ccamt 1- I: In c.ccH mom.H :H\.o< IIIIIII Inuit-un- zmsun tcmHm: poo; umAEHusem soantcwdu< m.mH m.mH m.mH m.me H~5.m ..ua ---- -------- - c000 0Hca sonHo-ceam< II II 0.0m O.Cm Omm.0 K\IU< llllltl III! lllll LOHUSDO SUCK gag: WHCQ SDHMQI—hwflmd - a.mm - 5.66 mom.H e\.om ---- ----- douaoso xuom soon oHoe nouHs-cooma In O.m~ -- a.mh omm.~ :H\:o< nun-III -u-uou-u swans tssHaO Echcz oHom souHo-COQma In o.m C.Nm c.cc GmcfHH ..C< IIIIIII IIII IIIII .-Illlialilliltll EDHCCZ emom nuufifllcvcmé a.mm - - 5.oe mom.H ;H\.0< --------------- - awash cccho noon 0Hoe reuHs-ceam< - - - o.coH H~5.m :H - ----- - ------- - -- - resin occh: >H HHH HH H moo—m HHz coHumconoc huHecoc ouHm ueHcoam huHmceo oNHw mcHuedw INOIBC. Hos.» 0:95... =0HuH5 Huanwo was u,o.....-Lw-m. become.» an: aneuuuooc: NuOumuorao - .mmahu HQSH Use momhu uo>oo osofis muoHa wHLEmm uo :oHuanHuuch .vm oHomb 131 Further examination of the four fuel types adds support to this notion. Samples representing types II and IV were found to be entirely void of coniferous tree species. Since vertical fire development is extremely unlikely in pure hardwood stands of this area (Van Wagner 1977), and since these two types accounted for only 11 percent of the sampling units, they may be deleted from consideration without concern. Types I and III, however, include conifers at least as an understory component, enabling fire development into the crown space. The primary differences between these two types are the amount of dead and downed branchwood on the forest floor, amount of leaf litter, and abundance of herbaceous growth. Table 25 presents average fuel properties for types I and III. Mainly due to the high dead/live ratio, type III fuels burn with slightly greater intensity and offer greater opportunity for torching and spotting behavior. Little guidance can be given, though, in dis- tinguishing between these types on the ground. There is little use, then, in recognizing any difference between them. It is recommended that one or the other be chosen to represent burning conditions for the entire study area. The choice between fuel models I and III is perhaps best left a matter of administrative discretion. Type I is more representative of the area on the whole, but type III offers more conservative (more severe) estimates of fire behavior. Since torching and spotting is most commonly associated with jackpots or concentrations of fuel, rather than average conditions for broad areas, it may be argued that the more severe of these two models would be preferred for anticipating these phenomona. 132 Table 25. Summary of fuel properties for BWCA fuel types I and III. Fuel Type I III Fuel Loadings (gm/m2) Litter 449 651 Live + Dead Foliage 112 135 Live + Dead Class I 179 202 Live + Dead Class II 359 - 381 Live + Dead Class III 920 853 Surface Area-to-Volume Ratio (cm‘l) 84 83 Dead/Live Ratio 3.73 8.65 Height to Conifer Crown Base (m) 1.83 1.83 Surface Fuel Packing Ratio 0.0242 0.0255 APPLICATION AND VALIDATION OF RESULTS Even though efforts to classify and map fuels on the basis of fire behavior potential were not entirely successful, and a high level of uncertainty concerning fire behavior at a specific location still exists, the results of the fuel appraisal remain useful for anticipating general fire behavior for the area on the whole. In this section, tools are presented for identifying conditions that will likely produce torching and long—distance spotting, as well as predicting surface fire spread rates and intensities. Predicted versus actual fire behavior is com- pared for three project-scale fires that occurred in or near the study area in 1976, and historical weather records are examined to determine the frequency and seasonal distribution of spotting conditions. Suggestions are offered for applying the results of this study. Fire Behavior Prediction To provide simple means for identifying conditions when natural fires could escape the study area, fire behavior nomographs were developed for fuel types I and III. Methods of Albini (l976b) were employed to display predicted surface fire intensities and spread rates in a nomograph format. Threshold surface fire intensities required for vertical fire development (Van Wagner 1977) were incorpor- ated directly into the nomographs. Figure 14 displays the resulting fire behavior nomographs for types I and III in the form developed by Albini (l976b). These are graphical aids for computing surface fire intensities and spread rates by the Rothermel (1972) model under a wide variety of conditions. They 133 134 BWCA FUEL TYPE I LIVE NEEDLE MOISTURE (%) DEAD FUEL MOISTURE CONTENT (96) 8 9. s , O {a ’e S to PERBACEOUS .20 .- HERBACEOUS 2, J MOISTURE (%) z MOISTURE (W ” 5 a: ' 52! LU “b ,2 +2- EL'15 ~ 8 (k 9 r- 10 m i” as .5 $1o~ Q - 15 3 U. 0 4? Hi LU t2 5‘ ~20 5'3 cr 0 ‘< 8 0 . . . 25 0 I 2 3 FIRE INTENSITY (M BTU/FTg’MIN) UHCHV! wwc- "no wm‘ 0 2 A L..—d E E 4...’ Q —_.__—/ LU ‘ b D or I 1 1 z 40 to 20 E SLOPE We) 8 O 16 14 12 Figure 14a. Fire behavior nomograph for fuel type I. (95) DEAD FUEL MOISTURE CONTENT (”ECIIVE WIND SING MM _5 U1 |\) O 01 135 BWCA FUEL TYPE III LIVE NEEDLE MOISTURE (‘13) O 8 o no .— .— 8 O HERBACEOUS MOISTURE (%) —l d 01 O 01 M O DEAD FUEL MOISTURE CONTENT (7.) LL 0 E < c: FIRE INTENSITY (M NO .h c» WIND SPEED (MPH) on 16 14 12 10 ’Figure 14b. Fire behavior nomograph for fuel type III.- 136 also show threshold surface fire intensities required for vertical fire growth into the crowns (Van Wagner 1977). Using these graphs, one can evaluate burning conditions with regard to the likelihood of torching and subsequent spotting. For operational convenience, English units were used in these graphs.l/ Input variables include (1) dead fuel moisture content; (2) live fuel moisture content for surface (herbaceous) fuels; (3) wind speed; (4) topographic slope; and (5) foliar moisture content of coniferous trees. Periodic measurements or estimates of live fuel moistures are required, while dead fuel moisture content is computed as a weighted average of National Fire Danger Rating System (Deeming et_al,, 1972) l-hour, lO-hour, and lOO-hour timelag moistures. Appropriate weighting coefficients are given in Table 26. The noted differences in potential fire behavior for types I and III are readily seen by comparing the scaling of the two nomographs. Use of the nomographs is explained in detail by Albini (l976b). The only noteworthy departure from Albini's explanation occurs after the fire intensity and rate of spread have been graphically determined. These values serve as coordinates of a point in the upper right graph that corresponds to the surface fire line intensity (Byram 1959). Upper right locations (high fire intensity and high rate of spread) indicate high fire line intensities, while lower left locations indicate low ones. The hyperbolic curves corresponding to various needle moistures 1/ 1 foot/minute=0.3048 meters/minute l BTU/footZ/minute=0.27l3 calories/cmz/minute 1 mile per hour=0.447 meters/second 137 Table 26. weighting coefficients for dead fuel moisture calculation. Time-lag Class Fuelfnype I III l-hour (class I) 0.77 0.40 lO-hour (class II) 0.18 0.32 loo-hour (class III) 0.05 0.28 138 in the upper right graph are lines of equal fire line intensity. They identify torching threshold values from the Van Wagner (1977) model. If the point determined by the rate of spread and fire intensity is below and to the left of the apprOpriate torching threshold intensity curve (interpolate if necessary), the fire is expected to remain on the surface. On the other hand, if the point is above the appropriate threshold intensity curve, periodic flaming of tree crowns is predicted. Consider a fire reported in fuel type III on a day with the follow- ing conditions: 5% l-hour timelag moisture, 7% lO-hour timelag moisture, 10% lOO-hour timelag moisture, 150% herbaceous moisture, 8 mile per hour wind speed, 0% topographic slope, and 100% needle moisture content. The circled letters in Figure 14b refer to the steps listed below: A. Dead fuel moisture is computed using the coefficients in Table 26. (.40 x 5%) + (.32 x 7%) + (.28 x 10%) = 7.04% This value is found on the vertical axis of the upper right graph, and a horizontal line is extended from this point across both upper graphs. B. As described by Albini (l976b), a "turning line" is drawn on the upper left graph by extending a straight line through the graph origin and the intersection of line A with the curve representing 150% herbaceous moisture. C. The herbaceous moisture curves in the upper right graph ("3" curves) are interpolated for 150% herbaceous moisture at their intersection with line A. A vertical line is dropped from this point through the lower right graph. Estimated surface fire intensity (reaction intensity) can now be read on the horizontal 139 axis of the upper right graph. In this case, it is about 2700 BTU/ftZ/min (733 cal/cmZ/min). D. No slope correction is required in this example so the line corresponding to an 8 mph wind in the lower right graph is sought directly. At the intersection of this line with line C, a horizontal line is drawn to the diagonal turning line in the lower left graph. E. A vertical line is drawn from this intersection to turning line B in the upper left graph. F. From this intersection, a final horizontal line is drawn back through the upper right graph. Estimated forward rate of spread can now be read--about 14 feet per minute (4.27 m/min). G. The intersection of lines F and C (point G) corresponds to the surface fire line intensity (Byram 1959). If point G is below and to the left of the appropriate torching threshold intensity curve, the fire is expected to remain on the surface. On the other hand, if point G is above the appropriate thres— hold intensity curve, periodic flaming of tree crowns is pre- dicted. In this example, point G falls below the 100% needle moisture curve so a surface fire is expected. Depending on other circumstances, suppression could probably be deferred without jeOpardy. A relatively small change in wind speed can produce very different results though. For example, with all other values held constant, a lO-mile per hour wind would increase the spread rate to 20 feet per minute and place the intersection point above the threshold surface intensity, suggesting that torching is likely and spotting possible. 140 Solving the nomograph in reverse under these moisture conditions, we find that a 9 mile per hour wind speed is required to produce a fire- line intensity exactly at the threshold level. This, then, may be view- ed as a threshold wind speed for the given moisture and slope conditions. When winds are expected to exceed the threshold value, managers should be aware that torching is likely and, depending on moisture conditions, spotting is possible. Fires occurring under these conditions should be examined closely to determine whether spotting fire behavior could jeopardize the perimeter of the study area or other critical values within the study area. Validation To determine the reliability of the nomograph predictions, weather and fire behavior data were examined for three actual fires that occurred in or near the study area during August and September of 1976. Actual fire behavior, with emphasis on long-distance spotting, was compared with predictions obtained from the nomographs. The three fires chosen for this exercise were the Roy Lake Fire (August 21 to August 27, 1976), the Rice Lake Fire (August 30 to September 2, 1976), and the Fraser Lake Fire (September 7 to September 12, 1976). The Roy Lake Fire occurred just west of the end of the Gunflint Trail, about 12 km northeast of the study area. The Rice Lake Fire occurred near Forest Center Landing, a roughly equal distance to the southwest, and the Fraser Lake Fire burned within the northwestern part of the study area. Since these were all project scale fires, weather observations in the general vicinity of the fires were taken at roughly hourly intervals throughout each burning day. These observations provided 141 the dead fuel moisture and wind speed inputs required by the nomographs. Additional inputs were assumed constant and assigned the following values, as appropriate for local conditions: Herbaceous moisture content: 150% Needle moisture content: 100% Topographic slope: 0% Using the nomOgraph for fuel model III with the above assumptions, the periodic weather observations were converted to "yes or no" predictions of spotting potential. The prediction was "yes" if the following conditions were met: 1. Observed wind speed 2 10 mph (4.5 m/sec) 2. lO-hour timelag fuel moisture < 8% 3. Predicted torching of conifer crowns from the nomograph Actual spotting occurrence was determined through examination of maps and records documenting each fire's growth and activity, and through the author's recollection of on-site experience. A tabulation of the binary (yes or no) scores for both predicted and actual fire behavior, on a daily basis, is given in Table 27. If spotting was predicted or occurred at any time during a given day, the corresponding table entry is "yes". A "no" indicates no spotting (predicted or actual) at any time during the day. As a general rule, actual spotting occurred between the hours of 1400 and 1900 CDT. Most predicted spotting conditions fell within this period also. Thirteen of the 17 burning days examined showed agreement between predicted and actual spotting behavior. Of the four discrepancies, three involved underprediction. That is, spotting occurred on three days when no spotting was predicted. Two of these 142 Table 27. Comparison of predicted versus actual spotting occurrence on three project fires in or near the study area. Fire Date Predicted Actual (area) Spotting? Spotting? Roy Lake August 21 No Yes (1368 hectares) 22 Yes Yes 23 No Yes 24 Yes Yes 25 Yes Yes 26 Yes Yes 27 No No Rice Lake August 30 Yes Yes (435 hectares) 31 No No September 1 No Yes 2 No No Fraser Lake September 7 Yes Yes (415 hectares) 8 No No 9 No No 10 Yes Yes 11 Yes Yes 12 Yes No 143 days occurred on the Roy Lake Fire. Perhaps these differences can be attributed in part to the somewhat sheltered location of the anemometer (at the Seagull Guard Station) and the erratic wind patterns of the area. Wind speeds were found to be characteristically greater at the fire than at the guard station, roughly 8 kilometers to the southeast. The third case involving underprediction occurred on the third day of the Rice Lake Fire. On this day, actual spotting was moderate and the "nega— tive" prediction was marginal. The single case of overprediction took place on the day the Fraser Lake Fire was declared controlled. It is reasonable to expect that a free-burning fire would have produced spotting on this day. Although this limited comparison is by no means a rigorous test of the model, it does suggest that the approach may be operationally useful in anticipating the behavior of free-burning natural fires, and assessing their potential threat to the study area boundaries. It also suggests that fuel type III provides a more reasonable indication of spotting potential than fuel type I, which produced several addition- al underpredictions. Frequency_and Distribution of Spotting Conditions From an operational standpoint it is useful to know how often spotting conditions can be expected, and during what periods they are most likely. Perhaps even more important is information concerning the persistence of spotting conditions once they occur. Decisions regarding specific actions to be taken on naturally-caused fires may depend to a large extent on the expected persistence of spotting behavior. To provide some insight on this matter, an analysis was 144 performed on weather records from the Ely fire danger rating station. Records for the period 1970 to 1977 were obtained from the National Fire Weather Data Library (Furman and Brink 1975) and examined to determine the likelihood of spotting conditions, and of runs of consecutive spotting days by month. Spotting days were defined as days of record showing a wind speed at or above 10 mph (4.5 m/sec), lO-hour timelag fuel moisture at or below 8%, and torching predicted for BWCA fuel type III (from nomograph). The results of this analysis are shown in Table 28. The first column in Table 28 shows the proportion of all days of record for each month that met the spotting criteria. The month of May had the highest occurrence of spotting conditions (11%) with April and July having only slightly lower values. Only 2% of the September days of record met the criteria. In the conditional probability columns just to the right, the likelihood of consecutive spotting days is shown for various lengths of run. This may be interpreted as follows. If today (day 0) is a spotting day, it may be useful to know how likely it is that tomorrow and the next day will be spotting days also, that is, that two additional spotting days will follow today. To determine this, look in the column headed "2" and the row for the appropriate month to find the probability of oc- currence based on past records. In May, for example, there is an 11% chance that spotting will continue for two additional days. However, there is a 68% chance that there will be no spotting tomorrow--and in no case did spotting conditions prevail for more than two additional days in May. Table 28. 145 Probability that any day is a "spotting day"l/ by month, and conditional probability that any spotting day (day 0) is followed by exactly n additional consecutive spotting days. Probability that Number of Consecutive Spot- Month any day is a ting Days Following Day 0 (n) spotting day 0 l 2 3 4 Conditional Probability January 0 - - - - - February 0 - - - - - March 0 - - - - - April 0.10 0.89 0.11 0 0 0 May 0.11 0.68 0.21 0.11 0 0 June 0.06 0.93 0.07 0 0 0 July 0.10 0.61 0.23 0.12 0.04 0 August 0.04 0.89 0.11 0 0 0 September 0.02 1.00 0 0 0 0 October 0.03 0.50 0.17 0.17 0.16 0 November 0 - - - - - December 0 - - - - - .1_/ A spotting day is defined as one with wind at or above 10 mph, lO-hour timelag fuel moisture at or below 8%, and torching predicted for BWCA Fuel Type III (from nomograph). 146 July and October are the only months showing runs of spotting conditions for three days following the first spotting day. The con- ditional probabilities for October, though, are based on only 6 days meeting spotting criteria for the entire period of record. They should probably be somewhat lower than shown in the table, but the low fire occurrence in October indicates little cause for concern. From a climatological standpoint, July appears to be the month that is conducive to the development of large fires. Extra caution should be exercised during periods showing high spotting persistence, especially when coincident with a precipitation deficit. A word of caution is appropriate at this point. Since fire danger rating weather observations are taken at 1300 hrs. daily, and since spotting conditions are most likely between 1400 and 1900, it is likely that Table 28 underestimates both the frequency and persistence of spotting conditions. Nevertheless, it does indicate relative seasonal differences that may prove meaningful. SUMMARY AND CONCLUSIONS Mindful of the natural role of fire in the BWCA, the Superior National Forest has proposed a pilot study on the use of natural fire by prescription. The ecological need to restore fire to this wilder- ness is apparent, but several information needs must be satisfied before a wholesale fire management program may be implemented. Foremost among these needs is a rational and practicable means of deciding when to suppress fires and when to defer control action. This decision, of course, involves a variety of political, social, and economic, as well as physical and biological considerations. For the pilot study itself, the principal concern of fire managers is that naturally caused fires, if suppression action is deferred, could become uncontrollable, escape the 40,000 hectare pilot study area, and damage non-wilderness resources. Long distance spotting poses the most likely means for this occurrence. Spotting has been found to occur most frequently when: l) the lO-hour timelag fuel moisture content is at or below 8%, 2) the wind- speed is at least 4.5m/sec (10 mph) above the canopy, and 3) surface fire intensity is sufficient to cause vertical fire development into the canopy space (torching). In this study, an attempt has been made to inventory, appraise, and classify fuel conditions within the pilot study area to facilitate quantitative assessment of the daily potential for torching or crowning -- and ultimately spotting fire behavior. Results may be useful in establishing decision criteria concerning alter- native fire suppression or surveillance actions within the study area. 147 148 During the summer of 1976, a broad-scale inventory of forest fuels was conducted to provide data on the amount, character, and distribution of organic materials potentially available for combustion on upland sites. A two-stage sampling design was employed, and in all, 210 subsample units were inventoried. An assortment of established methods, including quadrat, tran- sect, and plotless sampling procedures was used to quantify the various living and dead fuel components for each subsample unit. In the interest of sampling efficiency, double-sampling methods were employed for all fuel components. Biomass regressions, bulk density estimates, and other required constants were evaluated from existing literature or through extractive sampling efforts near the Kawishiwi Field Laboratory. Both surface and aerial fuels were described, expressing fuel amounts in oven-dry grams per square meter by species, condition (live or dead), size class, and height above ground level in 30.5 centimeter increments. Lowland fuels were not inventoried in this study. Fires that burn in these fuels are generally low in intensity, and seldom produce long-distance spotting. Upland fuels, on the other hand, frequently support extreme fire behavior including crowning in conifer stands and spotting at distances over a kilometer. These burning conditions were observed in upland stands of the pilot study area during the Frazer Lake Fire of 1976. Upland fuel conditions within the study area range from barren rock outcrops or rock with a thin mantle of mosses and lichens to relatively fresh cutting slash. Total fuel loading ranged from 3.8 to 17.2kg/m2. 0f the 10 fuel component categories recognized, humus 149 showed the greatest range of variation, while downed deadwood had the next largest range. Most of the deadwood variation, though, is attributable to holes and branches greater than 3.6 cm in diameter. Except under extreme drought conditions, humus and large diameter deadwood have only minor influence on forest fire behavior. The more influential surface fuel components (L layer, ground cover, herbaceous plants, and small diameter deadwood), although highly variable in a relative sense, have a somewhat narrower range of absolute variation. Of the 210 randomly located inventory samples, 168 or 80% were within the aspen-birch community type, reflecting the overall repre- sentation of that type within the study area. This is not typical of the BWCA as a whole, where the aspen-birch type makes up less than 30% of the area. The other 20% of the samples were distributed as follows: Maple-aspen-birch-fir (0.5%), fir-birch (4.3%), jackpine- balsam fir (1.9%), aspen-birch-white pine (3.3%), red pine (2.4%), black spruce—feather moss (3.8%), cedar (1.9%), and jackpine-black spruce (1.9%). Among the community types sampled, the variability of fuel component loadings was for the most part minor. Variation appears to be greater within these types than it is among them. Yet, it is the fire behavior potential, not fuel loading, that concerns BWCA fire managers. Laboratory analyses of important fuel components were conducted to determine total and silica-free ash contents, low heats of combustion, particle densities, and surface area-to-volume ratios. 150 These data facilitated the use of existing mathematical fire models to assess fire behavior potential in sampled stands. The Rothermel (1972) fire model provided estimates of forward rate of spread and reaction intensity which were used to calculate fireline intensity (Byram 1959) and flame length for fires burning in surface fuels only. Three combinations of windspeed and dead fuel moisture content were used to represent a realistic range of burning conditions. Besides potential surface fire behavior, threshold surface fire intensities required for torching individual trees and thresh- old spread rates required for continuous crowning were predicted using the model proposed by Van Wagner (1977). An R-mode principal components analysis was performed on the set of fire behavior predictions for each inventory subsample unit to reduce the dimensionality of the data. Principal component scores were computed for each subsample and used in a Q-mode agglomerative clustering routine based on Euclidean distance in the principal component space. The sampling units comprising each cluster were taken to represent a fuel type with burning characteristics distinct from all other types. Following clustering, the data were subjected to a discriminant analysis to investigate relationships between the fuel type classification and readily observed stand and site characteristics. As a result of the cluster analysis, four fuel types were iden- tified and designated types I, II, III, and IV. Eighty-one percent of all sampling units were grouped as type I, 9% as type II, 8% as type III, and 2% as type IV. The high representation of type I in 151 the sample reflects the visually apparent homogeneity of upland stands in the study area, and is not surprising considering the representation of community types in the sample (Grigal and Ohmann 1975). Unfortunately, there was no clear relationship between the community type and fuel type classifications. The discriminant analysis produced a set of functions that were able to correctly classify 91% of the inventoried subsampling units on the basis of general site and stand characteristics alone. Maximum tree height for upland black spruce, jackpine, and balsam fir, site index and basal area of all conifer species were identified as the most important of 42 variables needed to assign stands to the appropriate fuel type. In light of the a priori distribution of samples among fuel types, though, use of the discriminant functions to classify uninventoried stands appears questionable. The effort required to quantify all 42 variables approaches that of the fuel inventory itself. In addition to the discriminant analysis, the relationship be- tween fuel types and an existing vegetation type classification was examined. By maximum likelihood, all upland types should be represented by fuel type I. Apparently, as was found for fuel loadings, the variation in fire behavior potential is greater within forest types than it is among them. The general conclusion drawn both from the analysis of fuel inventory data and from visual inspection of the study area is that in spite of the noted spatial variation in fuel loading and fire behavior potential, a single fuel model should be used to represent upland fuels within the study area. 152 Fuel types II and IV were inappropriate because they represent pure hardwood stands, and because an understory component of balsam fir is virtually omnipresent in the BWCA. Stands represented by types I and III, on the other hand, include conifers at least in the under- story. The primary differences between types I and III are the amount of dead and downed branchwood on the forest floor, amount of leaf litter, and abundance of herbaceous growth. Type III has heavier surface loadings of dead fuel and lighter herbaceous loadings than type I. Type I is more representative of the area on the whole, but type III offers more conservative (more severe) estimates of fire behavior. It may be argued that the more severe of these two models would be preferred for anticipating potential spotting conditions. For sample clusters representing fuel types I and III, methods of Albini (l976b) were used to construct nomographs that display predicted surface fire intensities and spread rates, as well as threshold conditions for vertical fire development into the crown space. Using the nomographs, one can evaluate burning conditions with regard to the likelihood of torching and subsequent spotting. Pre- dicted occurrence of long-distance spotting, using the fuel type III nomograph, compared favorably with actual conditions on three project fires in or near the study area during 1976. Thirteen of the 17 burn- ing days examined showed agreement between predicted and actual spotting behavior, suggesting that the nomographs may be operationally useful for anticipating spotting conditions and assessing the potential threat of fire spread beyond study area boundaries. Furthermore, since three of the four discrepancies involved under 153 prediction, and since fuel type III represents more severe fire behavior than fuel type I, the nomograph for fuel type III appears more suitable for predicting spotting conditions. A brief analysis of fire danger rating records from Ely, Minnesota, showed that weather conditions fostering long-distance spotting occur most commonly in April, May, and July. In addition, the expected persistence of these conditions, once they occur, is greatest in May, July, and October. Decision makers should consider historical spotting occurrence and persistence data as well as specific weather forecasts before initiating suppression action on fires that appear to threaten the study area boundary. Several general comments are in order concerning the scope and complexity of this study in a management decision context, as well as the applicability of study results to the total BWCA. The inventory, appraisal, and classification of fuel conditions in the pilot study area have involved detailed, tedious, and costly measurement and modeling procedures. Yet, the interpretation and "packaging" of study results for application has been guided by a desire to provide an operationally useful decision aid than can be employed at a management cost that is commensurate with resulting improvements in fire manage- ment decisions. The fact that identified fuel types could not be discriminated and mapped through simple, cost-effective means indicates that much of the inventory effort was superfluous. Equally useful results could have been produced with considerably less effort. Careful thought should be given to fuel appraisal needs when the pilot study is expanded to an operational prescribed natural fire 154 program for the whole BWCA. The apparent differences between the vegetation of the pilot study area and that found elsewhere preclude direct application of results of this study to other areas. Whether additional inventory efforts would be fruitful for these areas, though, is uncertain. The greater diversity of community types found outside the pilot area may be more conducive to identifying and discriminating communities with meaningful differences in fire behavior potential. Furthermore, the greater representation of conifer types may provide economic justification for additional fuel classification work. Perhaps by the time expansion is considered, additional knowledge of fuel information needs and uses in wilderness fire management will allow explicit consideration of operational costs and benefits in the design of fuel inventory and appraisal activities. This suggests a priority area for future research effort. In the meantime, results of this study should be applied strictly to the pilot study area. Even here, though, the tools developed in this study are merely first generation prototypes that have not been adequately field tested. Potential users are urged to carefully document their successes and failures so that they may be improved as experience with natural fire in the BWCA accumulates. Furthermore, these tools are by no means substitutes for practical experience or professional judgement. They can only supplement these important attributes. APPENDIX 155 VEGETATION/FUEL TYPE CLASSIFICATION Central Study Area-Boundary Waters Canoe Area Superior National Forest. Minnesota |977 LEGEND ViciVAIlnN/FUIL TYPE apud (ln Dart) (mm the A - Aspen-fllrch 3 , Awen-Mrch-Plne averalorv, Sprure-Fir unoeruory ur- c - :u-ovu Am and slur- (r09 - Lowland {uni'er Hy (w..-mun Mack Sun-u) sup-n Slli cuss inmates of Vegetation Etaaslncallan ”'5‘9’1"? -9‘ 2' -Predom'mantlv Upland unm- ole (SH-9“ DB“) h with stillered Jack Pine d-Saurlmber (9". Dan) b sawllm .- W 5-1"/Ad‘ - mm sewl‘lmber-hze Hand - - m-m a, . .. - l-l-7oz (medium) -o-u To: (goo-1) “<70: 01 {he my." cover and lspervflrrth who: u! the tram- (over. DYNER FUVURES ‘ - vuer are“ + - lnurlor pneu- cenler Ao”/A:"' Aspen-Blah sund with applemmtsly equal Hocking n' saulimber and pole llmber. m) 1 -L|ne (:rmlnnl pr-oxo .. _ _ -- _ tuner 36 leed stand cl Aspen B :h Vine sawfimber (ht-701 crmn wvu mm u- understory 0' $ ~0verlsy ”gm-"(m- marker Spruce-Hr. Overlay prepared r E can tum-l ram: Experiment smlou usnA-roru- Szrvlce, by m: Unlversllv or Nlnneiotl nullege or Forenry (usull Cooperulve or n- ll Research Agreemzn! U-SSI) using 1:15.840 1: al: Aerochrone ly.-"m, ”0,09,.th Howu by the rareu Suvlce on Aug-m l7 lnd September (6, ls7é. Alrphnle lMUDre-n-an. 'lzld chzcklng and map canon-Hon by Hm- 5. EM"; detail (mule- by Lucinda Nruska-Elleys; duklng by Murine Needham and mm a. Hagen and projeti mauvlslm by Dr. Merle P. Keyer. LITERATURE CITED LITERATURE CITED Abell, C. A. 1937. Rate of spread and resistance to control data for Region 7 fuel types and their application to determine strength and speed of attack needed. U.S.D.A. Forest Serv., Appalachian Forest Expt. Sta., Asheville, N.C., 7 pp., illus. Agee, J. K. 1974. Fire management in the national parks. West. Wildlands 1(3):27-33. Ahlgren, C. E. 1969. Eighteen years of weather in the Boundary Waters Canoe Area. Univ. Minn. Agric. Expt. Sta. Misc. Rep 88. Ahlgren, C. E. 1974. Effects of fires on temperate forests: North Central United States. In_"Fire and Ecosystems" (T. T. Kozlowski and C. E. Ahlgren, eds.), pp. 195—223. Academic Press, New York. Ahlgren, C. E. 1976. Regeneration of red pine and white pine following wildfire and logging in northeastern Minnesota. J. For. 74(3):l35-l40. Albini, Frank A. 1975. A computer algorithm for sorting field data on fuel depths. U.S.D.A. For. Serv. Gen. Tech. Rep. INT-23, 25 p. (Intermountain Forest and Range Experiment Station, Ogden, Utah 84401). Albini, Frank A. l976a. Computer-based models of wildland fire behavior: a users' manual. U.S.D.A. For. Serv. Intermountain For. and Range Exp. Sta. 68 p. Albini, Frank A. l976b. Estimating wildfire behavior and effects. U.S.D.A. For. Serv. Gen. Tech. Rep. INT-30, 92 p. 74 ref. Intermountain Forest and Range Experiment Station, Ogden, Utah, 84401. Aldrich, D. F. and R. W. Mutch. 1972. Wilderness fires allowed to burn more naturally. Fire Control Notes 33(1):3-6. Anderson, Hal E. 1969. Heat transfer and fire spread. U.S.D.A. For. Serv. Res. Pap. INT-69, 29 p., illus. Intermt. For. and Range Exp. Stn., Ogden, Utah. Anderson, Hal E. 1971. The consideration of forest fuels in fire planning. Planning for Fire Management Supervision Proc., Dec. 9—11, 1971, p. 132-149, Phoenix, A2. 156 157 Arneman, H. F. 1963. Soils of Minnesota. Univ. Minn. Agric. Ext. Serv. Ext. Bull. 278. 8 pp. Baker, D. G., D. A. Haines, and J. H. Strub, Jr. 1967. Climate of Minnesota. Part V. Precipitation facts, normals, and extremes. Univ. Minn., Agr. Exp. Sta. Tech. Bull. 254. Baker, D. G. and J. H. Strub, Jr. 1965. Climate of Minnesota. Part III. Temperature and its application. Univ. Minn. Agr. Exp. Sta. Tech. Bull. 248. Banks, W. G. and H. C. Prayer. 1966. Rate of forest fire spread and resistance to control in the fuel types of the Eastern Region. Fire Control Notes. 27(2):10-13. Baskerville, G. L. 1972. Use of logarithmic regressions in the estimation of plant biomass. Can. J. For. Res. 2:49-53. Bevins, C. D. 1976. Fire modeling for natural fuel situations in Glacier National Park. First Conf. on Scientific Res. in the National Parks, New Orelans, LA, Nov. 1976. Bitterlich, W. 1947. Die Winkel-Zahlmessung (Measurement of basal area per hectane by means of angle measurement). Allg. Forst. und. Holzwirtsch. Ztg. 58:94-96. Broido, A. and Maxine Nelson. 1964. Ash content: its effect on combusion of corn plants. Science l46(3644):652-653. Brown, James K. 1963. Crown weights in red pine plantations. U.S. Forest Serv. Res. Note LS-l9, 4 pp. Lake States Forest Expt. Sta., St. Paul, Minn. Brown, James K. 1968. Quantitative description of certain physical properties of ponderosa pine forest floor and cheatgrass fuels. Ph.D. Thesis, Univ. of Mich., Univ. Microfilms (Diss. Abstr. 28B:828), Ann Arbor, Mich., 152 pp. Brown, James K. 1970a. Vertical distribution of fuel in spruce-fir logging slash. U.S.D.A. Forest Serv., Intermountain Forest and Range Expt. Sta., Ogden, Utah, Res. Pap. INT-81, 9 pp. Brown, J. K. 1970b. Ratios of surface area to volume for common fine fuels. For. Sci. 16(1):101-105. Brown, J. K. 1971. A planer intersect method for sampling fuel volume and surface area. Forest Sci. 17:96—102. Brown, J. K. 1972. Field test of a rate-of—fire-spread model in slash fuels. U.S. For. Serv. Res. Pap. INT-116. 158 Brown, James K. 1974. Handbook for inventorying downed woody material. U.S.D.A. Forest Serv., Intermountain Forest and Range Exp. Sta., Ogden, Utah, General Technical Report INT-l6. Brown, J. K. 1976. Estimating shrub biomass from basal stem diameters. Can. J. For. Res. 6:153-158. Brown, James K. 1978. Weight and density of crowns of Rocky Mountain conifers. U.S.D.A. For. Serv. Res. Pap. INT-197, 56 p. Brown, James K. and Roussopoulos, Peter J. 1974. Eliminating biases in the planar intersect method for estimating volumes of small fuels. For. Sci. 20:350-356. Buehl, M. F. and W. A. Niering. 1957. Fir-spruce-birch forest in northern Minnesota. Ecology 38(4):602-610. Butters, F. K. and E. C. Abbe. 1953. A floristic study of Cook County, northeastern Minnesota. Rhodora 55:21-55. Butts, David P. 1976. Fire Management in Rocky Mountain National Park, Part I. Tall Timbers Fire Ecology Conf. 14:59-75. Byram, George M. 1959. Combustion of Forest Fuels. Chapt. 3 13_ Forest Fire Control and Use, Ed. Kenneth P. Davis, McGraw—Hill Book Co., Inc., New York. Carter, Mason C. and E. H. White. 1971. Dry weight and nutrient accumulation in young stands of cottonwood (Populus deltoides Bartr.) Ala. Agric. Exp. Sta. Circ. 190, 14 p. Cooley, W. W. and P. R. Lohnes. 1962. Multivariate procedures for the behavioral sciences. John Wiley and Sons, N.Y. 211 p. Cooper, W. S. 1913. Climax forest of Isle Royale, Lake Superior, and its development. Bot. Gazette 55:1-44, 115-140, 189-205. Cormack,R. 1971. A review of classification. J. R. Statist. Soc. A. 134(3):321-367. Crow, Thomas R. 1977. Biomass and production regressions for trees and woody shrubs common to the Enterprise Forest. In_The Enter- prise, Wisconsin, Radiation Forest--Radioecologica1 Studies. J. Zavitkovski, ed. p. 63-67. Tech. Inf. Cent., Energy Res. and Dev. Adm. TID-26113-P2. Daniels, 0. L. 1974. Test of a new land management concept: Fritz Creek 1973. West. Wildlands 1(3):23-26. Davis, Kenneth P. 1959. Forest Fire Control and Use. McGraw—Hill Book Company, Inc. New York, 584 pp. 159 Dean, J. L. 1971. Wetland forest communities of the eastern Boundary Waters Canoe Area. M.S. Thesis, University of Minnesota, 101 pp. Deeming, John E., James W. Lancaster, Michael A. Fosberg, R. William Furman, and Mark J. Schroeder. 1972. The National Fire-Danger Rating System. U.S.D.A. For. Serv. Res. Pap. RM-84, 165 p. Rocky Mt. For. and Range Exp. Stn., Fort Collins, Colo. Deeming, John E., Robert E. Burgan, and Jack D. Cohen. 1977. The National Fire Danger Rating System-—1978. U.S.D.A. For. Serv. Gen. Tech. Rep. INT-39, 63 p. Intermt. For. and Range Exp. Stn., Ogden, Utah 84401. Devet, D. D. 1976. DESCON: Utilizing benign wildfires to achieve land management objectives. Tall Timbers Fire Ecology Conf. 14:33-44. DeVries, P. G. 1973. A general theory on line intersect sampling, with application to logging residue inventory. Mededelingen Landbouw Hogeschool 73-11. Wageningen, Netherlands. 23 p. Dyer, Richard F. 1967. Fresh and dry weight, nutrient elements and pulping characteristics of northern white cedar Thuja occidentalis. Maine Agric. Exp. Sta. Tech. Bull. 27. 38 p. Fahnestock, George R. 1968. Fire hazard from precommercial thinning of ponderosa pine. U.S.D.A. Forest Serv., Pacific Northwest Forest and Range Expt. Sta. Portland, Oregon, Res. Pap. PNW-57, 16 pp. illus. Fahnestock, George R. 1970. Two keys for appraising forest fire fuels. U.S.D.A. Forest Serv., Pacific Northwest Forest and Range Expt. Sta., Portland, Oregon, Res. Pap. PNW—99, 26 pp. Fenneman, N. M. 1938. Physiography of the Eastern United States. McGraw-Hill Book Co., New York. Flaccus, E., and L. F. Ohmann. 1964. Old-growth northern hardwood forests in northeastern Minnesota. Ecology 45(3):448-459. Frandsen, W. H. 1971. Fire spread through porous fuels from the conservation of energy. Combustion and Flame 16:9-16. Frandsen, William H. 1978. A nonuniform fire behavior model. (Manuscript in preparation. U.S.D.A. Forest Service, Inter- mountain Forest and Range Exp. Sta.) Furman, R. William, and Glen E. Brink. 1975. The national fire weather data library: What it is and how to use it. U.S.D.A. For. Serv. Gen. Tech. Rep. RM-19, 8 p. Rocky Mt. For. and Range Exp. Stn., Fort Collins, Colo. 80521. 160 Gerloff, G. C., D. D. Moore, and J. T. Curtis. 1964. Mineral content of native plants of Wisconsin. Univ. Wis. Coll. Agric. Exp. Stn. Res. Rep. No. 14. pp 1-27. Gleason, H. A. 1926. The individualistic concept of plant association. Torrey Bot. Club Bu11., 53:7-26. Grigal, D. F., and H. F. Arneman. 1970. Quantitative relationships among vegetation and soil classifications from northeastern Minnesota. Can. J. Bot. 48:555-566. Grigal, David F., and John G. McColl. 1975. Litter fall after wild- fire in virgin forests of northeastern Minnesota. Can. J. For. Res. 5, 655-661. Grigal, D. F., and Lewis F. Ohmann. 1975. Classification, descrip- tion, and dynamics of upland plant communities within a Minnesota wilderness area. Ecological Monographs, 45(4):389-407. Gunzel, L. L. 1974. National policy change: Natural prescribed fire. Fire Manage. 35(3):6—8. Habeck, James R. 1973. A phytosociological analysis of forests, fuels and fire in the Moose Creek Drainage, Selway-Bitterroot Wilderness. University of Montana - U.S.D.A. Forest Serv. Publication No. RI-73-022. 114 pp. Habeck, James R. 1976. Forest, fuels, and fire in the Selway-Bitter- root Wilderness, Idaho. Tall Timbers Fire Ecology Conference and Intermountain Fire & Land Management Symposium. 14:305-353. Hahn, Gerald J., and Samuel S. Shapiro. 1967. Statistical models in engineering. John Wiley and Sons, Inc., New York. Haines, Donald A., Von J. Johnson, and William A. Main. 1975. Wild- fire atlas of the northeastern and north central States. U.S.D.A. For. Serv. Gen. Tech. Rep. NC-16, 25 p., illus. North Cent. For. Exp. Stn., St. Paul, Minn. Heinselman, M. L. 1963. Forest sites, bog processes, and peatland types in the Glacial Lake Agassiz region, Minnesota. Ecological Monographs 33, 327-374. Heinselman, M. L. 1965. Vegetation management in wilderness areas and primitive parks. J. For. 63(6):440-445. Heinselman, M. L. 1969. Diary of the canoe country's landscape. Naturalist 20, 2-13. Heinselman, M. L. 1970. Restoring fire to the ecosystems of the Boundary Waters Canoe Area, Minnesota, and to similar wilderness areas. Proc. 10th Ann. Tall Timbers Fire Ecol. Conf. pp. 211-214. 161 Heinselman, M. L. 1973. Fire in the virgin forests of the Boundary Waters Canoe Area, Minnesota. Quarternary Research 3:329-382. Hornby, L. G. 1936. Fire control planning in the Northern Rocky Mountain Region. U.S.D.A. Forest Serv., North. Rocky Mt. Forest and Range Expt. Sta. Prog. Rpt. l. Hough, W. A. 1969. Caloric value of some forest fuels of the south- ern United States. U.S.D.A. Forest Serv. Res. Note SE-120. S.E. Forest Expt. Sta., Asheville, N.C. 6 pp. Hough, W. A., and F. A. Albini. 1976. Predicting fire behavior in palmetto-gallberry fuel complexes. (Manuscript in preparation). Hovde, M. R. 1941. Climate of Minnesota. Pp. 925-934. Ig_Climate and man. Yearb. Agric. U.S. Govt. Print. Off., Washington, D.C. Howard, E. T. 1973. Heat of combustion of various southern pine materials. Wood Sci.5(3):194-197. Hughes, Malcolm K. 1971. Seasonal calorific values from a deciduous woodland in England. Ecology 52(5):923-926. Jamison, G. M. and J. J. Keetch. 1942. Rate of spread of fire and its resistance to control in the fuel types of eastern mountain forests. U.S.D.A. Forest Serv., Appalachian Forest Expt. Sta., Tech. Note 52. Kessell, S. R. 1975. The Glacier National Park basic resources and fire ecology model. Bull. Ecol. Soc. Amer. 56(2):49. Kessell, Stephen R. 1976. Wildland inventories and fire modeling by gradient analysis in Glacier National Park. Joint Proceedings of the Tall Timbers Fire Ecology Conference--Intermountain Fire and Land Management Symposium. 14:115-162. Kiil, A. D. 1965. Weight and size distribution of slash of white spruce and lodgepole pine. Forest chron. 41(4):432-437. Kiil, A. D. 1968. Weight of the fuel complex in 70-year-old lodge- pole pine stands of different densities. Can. Dept. Forest Publ. 1228, 9 pp. Kilgore, B. M., and G. S. Briggs. 1972. Restoring fire to high elevation forests in California. J. For. 70(5):266-27l. Kilgore, B. M. 1975. Restoring fire to national park wilderness. Kira, T., and T. Shidei. 1967. Primary production and turnover of organic matter in different forest ecosystems of the western Pacific. Jap. J. Ecol. 17:70-87. 162 Kourtz, Peter, Shirley Nozaki, and William O'Regan. 1977. Forest fires in the computer: a model to predict the perimeter location of a forest fire. Can. For. Serv. Inf. Rep. FF-X-65, 26 pp. Kozlowski, T. T., and C. E. Ahlgren (eds.). 1974. Fire and Ecosys- tems. Academic Press, New York. 542 pp. LaRoi, G. H. 1967. Ecological studies in the boreal spruce-fir forests of the North American taiga. I. Analysis of the vas- cular flora. Ecol. Monogr. 37:229-253. Leverett, F., and F. W. Sardeson. 1932. Quaternary geology of Minn- esota and parts of adjacent states. U.S. Geol. Surv. Prof. Pap. 154-A. Loomis, Robert M. 1977. Jack pine and aspen forest floors in north- eastern Minnesota. U.S.D.A. For. Serv., North Central For. Exp. Sta. Res. Note NC-222. Loomis, Robert M., and Peter J. Roussopoulos. 1968. Estimating aspen crown fuels in northeastern Minnesota. U.S. Dep. Agric. For. Serv. Res. Pap. NC-156, 6 p. North Cent. For. Exp. Stn., St. Paul, Minnesota. Loope, Lloyd C., and Robert P. Wood. 1976. Fire management in Grand Teton National Park. Tall Timbers Fire Ecology Conf. 14:87-98. Madgwick, H. A. 1970. Caloric values of Pinus virginiana as affected by time of sampling, tree age, and position in stand. Ecology 51:1094-1097. Maissurow, D. K. 1935. Fire as a necessary factor in the perpetuation of white pine. J. Forest. 33:373-378. Maycock, P. F. and J. T. Curtis. 1960. The phytosociology of boreal conifer-hardwood forests of the Great Lakes Region. Ecol. Monogr. 30:1-35. McIntosh, R. P. 1967. The continuum concept of vegetation. Bot. Rev. 33:130-187. McIntyre, G. A. 1952. A method of unbiased selective sampling, using ranked sets. Aust. J. Agric. Res. 3:385-90. McNab, William H., M. Boyd Edwards, Jr., and Walter A. Hough. 1976. Estimating weight of fuel components in slash pine/saw palmetto/ gallberry. (Manuscript in preparation). Miller, Roswell K. 1978. The Keetch-Byram Drought Index and three fires in upper Michigan. Proc. Am. Met. Soc. Fifth Nat. Conf. on Fire and Forest Meterology. March 14-16, 1978, Atlantic City, NJ. 163 Morrison, I. K. 1973. Distribution of elements in aerial components of several natural jack pine stands in northern Ontario. Can. J. For. Res. 3, 170-179. Muraro, S. J. 1971. The lodgepole pine fuel complex. Can. For. Serv., Pac. Forest Res. Cent. Inf. Rep. BC-x-53, 50 pp. Mutch, R., and C. W. Philpot. 1970. Relationship of silica content to flammability in grasses. Forest Sci. 16:64-65. Noble, Mark G., Linda K. DeBoer, Kenneth L. Johnson, Barbara A. Coffin, Lucia G. Fellows, and Neil A. Christensen. 1977. Quantitative relationships among some Pinus banksiana-Picea mariana forests subjected to wildfire and postlogging treatments. Can. J. For. Res. 7:368-377. Nordin, J. O. and D. F. Grigal. 1976. Vegetation, site, and fire relationships within the area of the Little Sioux fire, north- eastern Minnesota. Can. J. For. Res. 6:78-85. Ohmann, L. F., and Ream, R. R. 1971a. Wilderness ecology: A method of sampling and summarizing data for plant community classifica- tion. U.S.D.A. Forest Serv. Res. Pap. NC-49, 14 p. N. Cent. Forest Exp. Sta., St. Paul, Minn. Ohmann, L. F., and R. R. Ream. 1971b. Wilderness ecology: Virgin plant communities of the Boundary Waters Canoe Area. U.S.D.A. For. Serv. Res. Pap. NC-63, 55 p. Ohmann, L. F., C. T. Cushwa, R. E. Lake, J. R. Beer, and R. B. Brander. 1973. Wilderness ecology: The upland communities, woody browse production, and small mammals of two adjacent 33-year-old wildfire areas of northeastern Minnesota. U.S. Dep. Agric. For. Serv. Gen. Tech. Rep. NC-7. Ohmann, L. F., D. F. Grigal, and R. B. Brander. 1976. Biomass esti- mation for five shrubs from northeastern Minnesota. U.S. Dep. Agric. For. Serv., Res. Pap. NC-133, 11 p. North Cent. For. Exp. Stn., St. Paul, Minnesota. Orloci, L. 1967. An agglomerative method for classification of plant communities. J. Ecol. 55:193-205. Peek, James M., David L. Urich, and Richard J. Mackie. 1976. Moose habitat selection and relationships to forest management in north- eastern Minnesota. Wildlife Monographs No. 48, 65 p. Philpot, C. W. 1968. Mineral content and pyrolysis of selected plant materials. U.S.D.A. Forest Serv. Res. Note INT-84, 4 p. Philpot, C. W. 1970. Influence of mineral content on the pyrolysis of plant materials. For. Sci. 16:461-471. 164 Philpot, Charles W. 1977. Vegetative features as determinants of fire frequency and intensity. I£_Proceedings of the Symposium on the Environmental Consequences of Fire and Fuel Management in Mediterranean Ecosystems. U.S.D.A. Forest Service General Techni- cal Report WO-3. Poole, Robert W. 1974. An introduction to quantitative ecology. McGraw-Hill Book Company. 532 p. Radloff, David L., and David R. Betters. 1978. Multivariate analy- sis of physical site data for wildland classification. Rao, C. R. 1952. Advanced statistical methods in biometric research. New York: John Wiley and Sons. Reiners, W. A. 1972. Structure and energetics of three Minnesota forests. Ecol. Monograph 42:71-94. Reiners, W. A., and N. M. Reiners. 1970. Energy and nutrient dynamics in forest floors in three Minnesota forests. J. Ecol. 58:497-519. Reiners, William A., and Norma M. Reiners. 1972. Comparison of oxygen-bomb combustion with standard ignition techniques for determining total ash. Ecology 53(1):132—136. Ribe, John H. 1973. Puckerbrush weight tables. Life Sciences and Agr. Exp. Sta. Univ. of Maine, Orono. Misc.Rep. 152. Rothermel, Richard C. 1972. A mathematical model for predicting fire spread in wildland fuels, U.S.D.A. Forest Serv. Res. Pap. INT-115, 40 p., illus. Rothermel, R. C. and C. W. Philpot. 1973. Predicting changes in Chaparral flammability. J. For. 71(10):640-643. Roussopoulos, Peter J., and Von J. Johnson. 1973. Estimating slash fuel loading for several lake states tree species. U.S.D.A. For. Serv. Res. Pap. NC-88, 8 p. illus. North Cent. For. Exp. Stn., St. Paul, Minn. Roussopoulos, Peter J., and Von J. Johnson. 1975. Help in making fuel management decisions. North Cent. For. Exp. Stn., St. Paul, Minn. 16 p., illus. (U.S.D.A. For. Serv. Res. Pap. NC-112). Rowe, J. S. 1959. Forest regions of Canada. Can. Dep. N. Aff. Nat. Res. For. Br. Bull. 123. Sando, Rodney W., and Donald A. Haines. 1972. Fire weather and behavior of the Little Sioux fire. North Cent. For. Exp. Stn., St. Paul, Minn. 6 p., illus. (U.S.D.A. For. Serv. Res. Pap. NC-76). 165 Sando, Rodney W., and Charles H. Wick. 1972. A method of evaluating crown fuels in forest stands. North Cent. For. Exp. Stn., St. Paul, MN, U.S.D.A. For. Serv. Res. Pap. NC-84, 10 pp., illus. Seal, H. 1964. Multivariate statistical analysis for biologists. Wiley, New York. Sellers, Robert E., and Don G. Despain. 1976. Fire management in Yellowstone National Park. Tall Timbers Fire Ecology Conf. 14:99-114. Shafizadeh, Fred, Peter P. S. Chin, and William F. DeGroot. 1977. Effective heat content of green forest fuels. Forest Science 23(1):81-89. Sneeuwjagt, R. J. 1974. Evaluation of the grass fuel model of the National Fire Danger Rating System. M. S. Thesis. University of Washington, Seattle. Sneeuwjagt, R. J., and W. H. Frandsen. 1977. Behavior of experimental grass fires vs. predictions based on Rothermel's fire model. Can. J. For. Res. 7:357-367. Society of American Foresters. 1954. Forest cover types of North America (Exclusive of Mexico). Soc. Amer. Foresters, Washington, D.C., 67 pp. Soper, E. K. 1919. The peat deposits of Minnesota. Minn. Geol. Surv. Bull. 16. Susott, R. A., W. F. DeGroot, and F. Shafizadeh. 1975. Heat content of natural fuels. J. Fire and Flammability 6:311-325. Swain, Albert M. 1973. A history of fire and vegetation in North- eastern Minnesota as recorded in lake sediments. Quaternary Res. 3:383-396. Telfer, E. S. 1969. Weight-diameter relationships for 22 woody plant species. Can. J. of Bot. 47(12):1851-1855. Telitsyn, G. P. and A. P. Sosnovshchenko. 1969. (Calorific value of certain forest fuels in the Far East). Sbornik Trudov, 9:439-442. Turner, B. J. 1974. Applications of cluster analysis in natural resources research. Forest Sci. 20:343-349. U. S. Department of Agriculture. 1955. Wood handbook. Agric. Handb. 72, 528 p. U. S. Department of Commerce, Weather Bureau. 1956-1975. Minnesota monthly summaries. U. S. Govt. Print. Off., Washington, D.C. 166 Van Wagner, C. E. 1968. The line intersect method in forest fuel sampling. Forest Sci. 14:20-26. Van Wagner, C. E. 1972. Heat of combustion, heat yield, and fire behaviour. Petawawa Forest Experiment Station Inf. Rep. PS-X-35. Chalk River, Ontario. Van Wagner, C. E. 1977. Conditions for the start and spread of crown fire. Can. J. For. Res. 7:23-34. Walker, J. D., and B. J. Stocks. 1975. The fuel complex of mature and immature jack pine stands in Ontario. Can. For. Serv. Sault Ste. Marie, Ont. Rep. O-X-229. 19 p., illus. Whittaker, R. H. 1962. Classification of natural communities. Bot. Rev., 28:1—239. Whittaker, R. H. 1967. Gradient analysis of vegetation. Biol. Rev., 49:207-264. Whittaker, R. H., ed. 1973. Handbook of vegetation science, Part V, Ordination and classification of communities. Dr. W. Junk, the Hague, Netherlands, 737 p. Whittaker, R. H., and G. M. Woodwell. 1968. Dimension and production relations of trees and shrubs in the Brookhaven Forest, New York. J. Ecol. 56:1—25. Whittaker, R. H., and G. M. Woodwell. 1971. Measurement of net primary production of forests. ‘Ig: "Productivity of Forest Ecosystems," P. Duvigneaud (ed.), p. 159-175. Proc. Brussels Symp. 1969, UNESCO. Wright, H. E., Jr. 1969. Forest history of the BWCA since the last glacial period. Naturalist. 20, 14—17. Wright, H. E., Jr. 1971. Retreat of the Laurentide Ice Sheet from 14,000 to 9,000 years ago. Quaternary Research 1, 316-330. Wright, H. E., Jr. 1974. Landscape development, forest fires, and wilderness management. Science, 186 (4163):487-495. Wright, H. E., Jr. and Watts, W. A. 1969. Glacial and vegetational history of northeastern Minnesota. Minnesota Geological Survey, Special Publication Series. SP-ll, 59 pp. Young, H. E., L. Strand and R. A. Altenberger. 1964. Preliminary fresh and dry weight tables for seven tree species in Maine. Maine Agri. Exp. Sta. Tech. Bull. T-12.