Dynamic Interactions between Forest Structure and Fire Behavior in Boreal Ecosystems

This paper reviews and synthesizes literature on fi re as a disturbance factor in boreal forests. Spatial and temporal variation in the biophysical environment, specifi cally, vegetative structure, terrain, and weather lead to variations in fi re behavior. Changes in slope, aspect, elevation, and soil affect site energy and water budgets and the potential plant community. These terrain features also have a major infl uence on fi re-caused disturbance through their role in determining moisture conditions and fl ammability of fuels on hourly, seasonal, and successional time-scales. On fi ne time scales (minutes to hours), changes in weather, specifi cally wind and relative humidity, signifi cantly affect a fi re’s intensity and aboveground effects. Normal seasonal changes in dryness and periodic drought infl uence fi re intensity and severity principally by affecting the depth of burn and belowground effects. On decades-long time scales changes in vegetative structure affect the mass of fuel available for burning and therefore the potential energy that can be released during a fi re. The severity of fi re varies in time and space depending not only on the biophysical environment, but also on the location on the fi re’s perimeter (head vs. fl ank vs. rear). Spatial and temporal variation in severity within a fi re can have long-lasting impacts on the structure and species composition of post-fi re communities and the potential for future disturbances. Characteristic temperature histories of ground, surface, and crown fi res are used to illustrate variations in fi re severity. A soil-heating model is used to illustrate the impact of varying depth of burn on the depth at which various fi re effects occur in the soil profi le. A conceptual model is presented for the effects of fi re severity on fi re-plant regeneration interactions. The conceptual model can be used by restoration ecologists to evaluate the differential effects of controlled or prescribed fi res and wildfi res and to plan and implement fi re treatments to conserve biodiversity.


Introduction
Disturbances (White and Pickett 1985), and fi res in particular, are important processes for maintaining community and landscape biodiversity in boreal ecosystems (Wein and MacLean 1983a, Johnson 1992, Shugart et al. 1992, Goldammer and Furyaev 1996).Excessive and often haphazard use of fi re historically occurred and continues to occur in some areas (Malluik 1995, Parviainen 1996, Pyne et al. 1996, Östlund et al. 1997, Hörnberg et al. 1999, Pitkänen and Huttunen 1999).In many areas, natural resource managers have been largely successful at removing fi re from the ecosystem, leading to major changes in species composition and stand structure (Gromtsev 1996, Hardy and Arno 1996, Parviainen 1996, Arno et al. 1997, Smith and Arno 1999).Both scenarios, excessive fi re and fi re suppression, constitute signifi cant departures from the disturbance regimes that have prevailed through most of the Holocene (Tolonnen 1983, Clark andRichard 1996).Departures from long-standing disturbance patterns are widely believed to be major factors in the loss of biodiversity in many parts of the world (Gill and Bradstock 1995) and particularly in boreal forests (Granström 1996, Parviainen 1996).As a result, there is now considerable interest in reintroducing fi re in many boreal ecosystems.
An understanding of the role and use of fi re in restoring and maintaining biodiversity requires studies that integrate phytosociology, landscape ecology, ecophysiology, and wildland fi re science (Johnson 1992, Shugart et al. 1992, Ryan 1998, 2000a).There is considerable literature on shortterm changes in post-fi re community ecology.In contrast, relatively few studies have quantifi ed long-term community (Arno et al. 1997, Covington and Moore 1994, Östlund et al. 1997, Hörnberg et al. 1999) or landscape dynamics (Arno et al. 1993, Syrjänen et al. 1994, Gromtsev 1996, Minnich and Chou 1997, Turner et al. 1997, 1999, Hessburg et al. 1999, Kasischke et al. 2000a).Likewise, there have been few attempts to address the long-term consequences of fi re on biogeochemical processes at the individual, community, or landscape levels (c.f.Shugart et al. 1992, 2000, Keane et al. 1997, 1998, 1999).Morphological characteristics, plant architecture, and life history attributes interact with heat trans-fer mechanisms during fi res resulting in variable plant responses (Noble and Slatyer 1980, 1981, Gill 1981, Keeley 1981, Rowe 1983, Peterson and Ryan 1986, Trabaud 1987, Johnson 1992, Fischer et al. 1996).The physiological ecology of individuals and species affect their ability to respond favorably in the post-fi re environment.How individuals and species resist fi re injury, as well as their physiological response to injury, affect a number of cascading secondary disturbance and successional processes.For example, variations in fi re behavior lead to variations in tree injury, which in turn, affect such ecosystem processes as which species of insects attack fi redamaged trees (Geiszler et al. 1984, Ryan and Amman 1994, 1996), fungal fl ora (Littke et al. 1986, Penttilä and Kotiranta 1996, Richter et al. 2000), stand decomposition (Lowell et al. 1992, Richter et al. 2000), snag retention, and insectivorous avian dynamics (Hutto 1995, Saab andDudley 1998).Improved understanding of the role and use of fi re in the restoration and maintenance of biodiversity requires greater recognition of fi re's variability in space and time and how that affects pattern and process.
While climate has a dominant infl uence on the overall productivity and character of vegetation (Woodward 1987, Kasischke andStocks 2000), disturbances have a major impact on community structure and function during stand development (Grime 1979, White andPickett 1985).Landscapes are composed of communities with varying disturbance histories.A disturbance is a discrete event that has a signifi cant affect on community composition, structure, or function (White and Pickett 1985).Over time, aggregate multiple disturbances lead to dynamic communities and shifting landscape mosaics (Fig. 1).Weather, insects, diseases, exotic species, and fi re dynamically interact to affect the composition, structure, and function of the stand.The range and distribution of disturbance are just as important as the average and extreme events.Each type of disturbance has its own characteristic infl uence on community structure.For example, severe windstorms tend to thin the forest from above causing the greatest damage to the dominant overstory trees, which often are the earlier successional species (Foster 1988).In contrast, a surface fi re thins the forest from below causing the greatest

Ryan
Dynamic Interactions between Forest Structure and Fire Behavior in Boreal Ecosystems damage to smaller and younger trees, which tend to be the later successional species (Peterson and Ryan 1986, Ryan and Reinhardt 1988, Johnson 1992, Agee 1993).
Fire is a major disturbance in boreal ecosystems (Rowe 1983, Johnson 1992, Bonan and Shugart 1989, Shugart et al. 1992, Goldammer and Furyaev 1996, Kasischke and Stocks 2000).Fire science, the study of fi re behavior and heat transfer mechanisms, provides a basis for understanding fi re effects on individuals, landscapes and ecosystems (Albini 1976, Alexander 1982, Johnson 1992, Johnson and Miyanishi 2001).The purpose of this paper is to review fi re as a disturbance process in boreal ecosystems, describe how spatial and temporal changes in the biophysical environment affect fi re behavior and severity, and describe some relationships between fi re severity, soil heating, plant survival, and regeneration.
2 Fire as a Disturbance Process White and Pickett (1985) describe disturbance as a discrete event having attributes of kind or type, frequency, extent, seasonality, magnitude, and synergy.Fire is itself a dynamic process that varies in time and space, and each of these attributes has varying effects on the dynamics of boreal forests.

Fire Frequency
Fire frequency describes the number of fi res in a given period of time.It is defi ned by such attributes as the mean fi re return interval (the average number of years between successive fi res in a given area over a given time period) and fi re cycle (the time required to burn an area equal to the study area) (McPherson et al. 1990, Agee 1993, Brown 2000).Fire frequency is an important factor affecting species interactions because of age-to-reproductive-maturity considerations (Grime 1979, Noble and Slatyer 1980, 1981, Rowe 1983).Variations in fi re frequency can also lead to variation in regeneration strategy within a species, e.g., proportion of cone serotiny (Muir 1993, Gauthier et al. 1996).Very frequent fi res typically lead to dominance of species with short life-cycles, e.g., annual grasses and herbs ( Keeley 1981).Theoretically, frequent fi res favor small patch size (Clark 1990, Malamud et al. 1998).On average the mass of burnable fuel, and therefore fi re intensity, are inversely related to fi re frequency (Olson 1981).An exception occurs when successive fi res are much closer than the average fi re return interval.For example, fi re intensity and severity may be greater than normal when a stand replacement fi re occurs in a mature coniferous forest and is followed by a subsequent fi re in dense reproduction and heavy deadfall (Wellner 1970, Heinselman 1981, Hungerford et al. 1991, Gray and Franklin 1997).Variation in fi re return intervals favors multiple species with varied life histories (Gill 1981, Keeley 1981, Noble and Slatyer 1981, Rowe 1983).
Several authors have reviewed fi re frequency data for boreal forests (cf. Zackrisson 1977, Heinselman 1981, Wein and MacLean 1983b, Bergeron and Brisson 1990, Payette 1992, Wein 1993, Turner and Romme 1994, DeLong 1998, Brown 2000, Duchesne and Hawkes 2000).Fire frequencies vary from a few decades to several centuries depending on location.Fire frequency varies by climatic zone as seasonal patterns of precipitation (Johnson 1992, Payette 1992, Pojar 1996, Murphy et al. 2000, Flannigan and Wotton 2001) and lightning (Sannikov andGoldammer 1996, Latham andWilliams 2001) vary.For example, on average, the frequency of fi res decreases moving from the interior of Alaska and the Rocky Mountain Cordillera in the Northwestern part of the North American boreal forest toward the more humid boreal forest in southeastern Canada (Heinselman 1981, Johnson 1992, Pojar 1996, DeLong 1998, Murphy et al. 2000).Similar continental-scale patterns occur in the Eurasian boreal forest where fi re frequency increases on average east of the Ural Mountains (Bonan and Shugart 1989).However, the minimum requirements for fi res to occur are available dry fuel and a source of ignition.Because the energy and moisture constraints on combustion are spatially heterogeneous (Kunkel 2001), fi re frequency is highly variable and short fi re return interval vegetation types can be found on dry sites in more humid regions (Zackrisson 1977, Heinselman 1981, 1996, Rowe 1983, Bergeron and Brisson 1990, Engstrom and Mann 1991, Syrjänen et al. 1994).Within a climatic zone fi re frequency varies by aspect because of its affect on the radiation balance (Kunkel 2001), by hillslope moisture gradients (Syrjänen et al. 1994, Samran et al. 1995, Kushla and Ripple 1997, Larsen 1997), and by proximity to human-caused ignitions (Syrjänen et al. 1994, Parviainen 1996, Swetnam 1996, Östlund et al. 1997, Hörnberg et al. 1999, Pitkänen and Huttunen 1999).Likewise, local drainage patterns (Larsen 1997) and landform (Zackrisson 1977, Hessberg et al. 1999, DeLong 1998) affect fi re frequency.At the landscape scale fi re movement may be viewed as somewhat analogous to the fl ow of a buoyant fl uid.Natural barriers such as avalanche paths (Malanson andButler 1984, Keane et al. 1997), lakes (Bergeron 1991, Gauthier et al. 1996), rivers, and barren ground reduce the likelihood of fi re spreading into an area (Heinselman 1981, 1996, Turner and Romme 1994).Thus two sites that are otherwise similar can have different fi re frequencies.Barriers that are effective in mild burning conditions (e.g., low fuel mass, high moisture content, and low wind speed) are ineffective under more severe conditions (e.g., high wind speed and spotting) (Heinselman 1981, Turner and Romme 1994, Schimmel and Granström 1997, Agee et al. 2000, Finney 2001).Thus, much of the boreal forest exhibits a pattern of periodic small fi res with infrequent large fi res that are associated with high wind and drought (Van Wagner 1983, Fryer and Johnson 1988, Larsen 1989, Bergeron and Brisson 1990, Johnson et al. 1990, Bergeron 1991, Johnson and Larsen 1991, Bessie and Johnson 1995, Korovin 1996, Murphy et al. 2000, Flannigan and Wotton 2001, Hess et al. 2001, Senkowsky 2001) and episodic climate variations (Clark 1990, Sirois and Payette 1991, Payette 1992, Campbell and Flannigan 2000).
The complexity of factors controlling the spatial and temporal distribution of dry, burnable fuel and ignition sources affect fi re return intervals within and between vegetation types.Modal and extremes in the range of fi re return have important implications on the species composition of stands (White and Pickett 1985).Attempts to restore or conserve biodiversity by reintroducing fi re need to consider the natural range in variability of fi re frequency for a given area (Hunter 1993, DeLong 1998, Lertzman et al. 1998).

Ryan
Dynamic Interactions between Forest Structure and Fire Behavior in Boreal Ecosystems

Fire Size
The area within a fi re's perimeter is often used to describe the extent of a fi re, but the actual area burned, patch size, and burn mosaic should also be considered in bioconservation studies (Eberhardt and Woodward 1987, Turner and Romme 1994, Turner et al. 1997).Burn mosaic needs to be assessed in three dimensions: the pattern of area burned, i.e., green versus black within the overall fi re perimeter, and the above-and belowground impacts within the burned area, i.e., the size-magnitude interaction within the disturbance.This interaction is important because the scale of the burn mosaic relative to species niche requirements and mobility can have major impacts on early fl oral and faunal dynamics (Noble and Slatyer 1977, 1980, 1981, Gill 1981, Heinselman 1981, 1996, Noble 1981, Rowe 1983, Eberhardt and Woodward 1987, Sirois and Payette 1991, Fischer et al. 1996, Smith and Fischer 1997, Turner et al. 1997, Kasischke et al. 2000a, Miller 2000, Smith 2000).Heterogeneity in vegetation structure and microenvironment leads to heterogeneity in fi re behavior and effects that can increase the heterogeneity of post-fi re vegetation (Heinselman 1981, Rowe 1983, Mushinsky and Gibson 1991, Turner et al. 1994).Within the boreal forest, many sub regions have relatively few species adapted to a site (Nikolov and Helmisaari 1992), and the size of the burned area may be large relative to species mobility (Payette 1992).Thus, the increase in heterogeneity may not be readily apparent in the species present but may be manifest in the rates of numerous biogeochemical processes (Shugart et al. 1992, Kasischke andStocks 2000).Heterogeneity occurs at all spatial scales within fi res (Turner et al. 1994(Turner et al. , 1997)).Large patches of crown fi re-killed forests often contain a mosaic of varying depths of burn into the soil due to spatial differences in the depth of the duff (mor) (i.e., the combined fermentation and humus soil horizons) and moisture content (Ryan and Noste 1985, Fryer and Johnson 1988, Zoltai et al. 1998, Kasischke et al. 2000a, 2000b, Miyanishi 2001).Lightly burned and unburned patches provide refugia for fi re sensitive fl ora and fauna.Likewise, deep burning ground fi res often occur with little surface fl aming, resulting in large areas of deeply heated soil with minimal direct heat effects on aboveground tissues (Rowe 1983, Wein 1983, Ryan and Noste 1985, Ryan and Frandsen 1991).Because of subtle changes in the biophysical environment, fi res often burn in combinations of intensity over short distances (Turner et al. 1994, Kafka et al. 2001).For example, creeping surface fi res commonly transition into passive crown fi res when they encounter low branches (Van Wagner 1977, 1983).These branches provide the vertical fuel continuity necessary to carry the fi re into the crown.The torching of individual trees and small clumps of trees results in localized patches of high upward heat pulses.This type of fi re is common in boreal forests burning under low winds and low relative humidity (Heinselman 1981, 1996, Rowe 1983).With increasing wind, these fi res can transition into intense crown fi res (Van Wagner 1977, 1993, Fryer and Johnson 1988, Scott and Reinhardt 2001) that result in heterogeneous fi re treatments of varying size within the larger fi re perimeter.
Homogeneous environments lead to larger, more uniform fi res.When fuels, weather, and terrain are relatively uniform within a region, a large portion of the area will be receptive to ignition and burnout at one time (DeLong 1998).Once ignited, fi res will burn until a signifi cant change occurs in either the weather or fuels (Johnson 1992, Bessie andJohnson 1995).Patch size is then determined by how far the fi re can spread before encountering a signifi cant change in burning conditions.Dry continental air masses, strong persistence patterns such as blocking high-pressure ridges, and the strong wind events associated with passing of dry cold fronts create conditions suitable for rapid fi re growth and extended severe fi re weather (Johnson 1992, Johnson and Wowchuk 1993, Flannigan and Wotton 2001).Thus, most of the area burned in the boreal Alaska, Canada, and Siberia is burned by a relatively small number of large fi res that burn for several days in relatively dry years (Johnson et al. 1990, Johnson and Larsen 1991, Johnson 1992, Bessie and Johnson 1995, Korovin 1996, Valendik 1996, Flannigan and Wotton 2001, Hess et al. 2001).

Seasonality
Fires may burn during the pre-growing dormant period, the active growing season, or the postgrowing dormant season with variable results that may be related to differences in the fi re's energy release characteristics, plant susceptibility to injury, or physiological response to injury.
Seasonality is important because of direct changes in fuel moisture that affect fl ammability (Albini 1976, Van Wagner 1983, Andrews 1986, Stocks et al. 1989, Johnson 1992).Crown fi re potential increases with decreasing foliar moisture (Van Wagner 1977, 1993, Alexander 1998, Scott and Reinhardt 2001).Foliar moisture content is lowest in the spring prior to bud-break when soils are still at or near freezing (Chrosciewicz 1986).Early in the fi re season low relative humidity and high wind can combine to yield high crown fi re potential while the forest fl oor is still too wet to sustain combustion (Artsybachev 1983, Stocks et al. 1989).Even though it is somewhat more diffi cult to initiate and sustain a crown fi re when foliar moisture content is higher in the mid and late summer, high intensity crown fi res still occur when relative humidity is low, wind speed is high, and fi ne fuels are abundant (Stocks et al. 1989, Johnson 1992).In contrast to spring fi res, these later-season fi res often are accompanied by deep depth-of-burn ground fi res due to the lower duff moisture content (Kasischke et al. 2000a, 2000b, Miyanishi 2001).The depth of meristematic tissues varies by species (Flinn and Wein 1977, Schimmel and Grandström 1996, Fischer et al. 1996, Smith and Fischer 1997, Miller 2000).Thus, species susceptibility varies with the depth of heat penetration during burning, which can be expected to increase with seasonal dryness of the duff leading to a species-dependent seasonal effect on survival of meristematic tissues.Given its infl uence on fi re severity, the most signifi cant seasonal effect in boreal forests is related to duff moisture content (Heinselman 1981, Johnson 1992).
Independent of the duff moisture effect on combustion, species-dependent variation should be expected in responses due to seasonal changes in phenology (Wright and Bailey 1982, Flinn and Pringle 1983, Peterson and Ryan 1986, Agee 1993, Miller 2000).These differences may be due either to morphological factors affecting heat transfer to meristematic tissues (Peterson and Ryan 1986) or seasonal variations in plant water relations and the availability of stored carbohydrates (Flynn and Pringle 1983, Ryan 1990, 1998, 2000a, Rigolot et al. 1994, Ducrey et al. 1996).For example, in controlled laboratory studies Flinn and Pringle (1983) found that rhizomes of eight boreal species varied in their heat tolerance.All were most sensitive to injury during the summer, and all but one species exhibited better growth following spring heating as compared to fall heating.Several species from lower latitudes have been shown to respond most favorably, e.g., by fl ower production, following fi res that occur in the 'natural' fi re season as opposed to those contrived for cultural purposes (c.f.Fischer et al. 1996, Miller 2000).Thus, seasonally-dependent species responses likely exist in boreal areas dominated by early vs. mid vs. late season fi re occurrence.
Other seasonal effects likely occur.For example, large boreal forest fi res are commonly associated with drought (Johnson 1992, Bessie and Johnson 1995, Korovin 1996, Murphy et al. 2000, Flannigan and Wotton 2001, Hess et al. 2001, Senkowsky 2001).Drought affects tree physiology in a variety of ways, including reduced carbon allocation to stem growth and host resistance to bark beetles (Larsen and MacDonald 1995, Ryan 1998, 2000a).There is little direct evidence of seasonal effects on host resistance to insect attack because of the lack of research.However, one would anticipate that fi re injury prior to bark beetle fl ight would subject a tree to an increased likelihood of attack as opposed to fi re injury that occurred after beetle fl ight.There are likely a number of seasonally-dependant fi re effects other than the obvious ones related to fi re behavior, but these remain to be elucidated

Fire Magnitude
In fi re ecology there is no universally accepted defi nition for the magnitude of fi re.White and Pickett (1985) give examples of intensity and severity as two measures of magnitude.Accordingly, intensity pertains to the '…physical force of the event per area per time period (e.g., heat released per time period for a fi re…)' and severity pertains to the 'impact on the organism, community or ecosystem (e.g., basal area removed).' Fire managers have long recognized that the amount of available fuel, weather conditions, and terrain steepness have a dominant effect on a fi re's energy release characteristics and fi re suppression capabilities (cf.Rothermel 1972, 1991, Albini 1976, Stocks et al. 1989, Pyne et al. 1996, Grishin 1997).This triplet, fuels, weather, and terrain, is referred to in the North American fi re science community as the fi re environment concept.Of more interest in bioconservation and restoration studies is the understanding that the energy released by fi re has the potential to do ecological work, i.e., to change a host of ecosystem state variables.Thus, quantifi cation of the energetics of fi res is desirable in ecological studies (Johnson 1992, Johnson andMiyanishi 2001).However, fi re behavior is highly variable in non-uniform fuels, instrumentation is costly, and it is often impractical to sample fi re behavior except on small experimental plots, making it diffi cult to quantify the magnitude of fi re treatments in ecological studies and restoration projects.
The fi re environment concept can be extended from its suppression-derived simplicity to a more ecological construct (Fig. 2a).The extension of the fi re environment concept to ecological studies requires that fuels be considered in the broader context of the structure of biomass on the site.Structure includes the quantity, distribution, and horizontal and vertical arrangement of live and dead trees, understory vegetation, woody debris, litter, and humus (Artsybashev 1983, Brown and Bevins 1986, Johnson 1992).Structure defi nes the total amount of biomass that can be burned, and therefore the total energy that can be released in a fi re.The size distribution of the structural components defi nes the rate at which energy will be released during favorable burning conditions.The rates at which fuels wet, dry (Nelson 2001), and burn (Anderson 1969) are functions of particle surface-area.These rates can be approximated from diameter for most fuels above the duff layer.
As a fi re burns across the landscape, it encounters different communities with different disturbance histories that result in varying stand structures and fl ammability.For example, stands with a high open crown and low understory fuels have poor vertical fuel continuity.They have an increased likelihood of burning due to increased sunlight and wind at the surface (Albini 1976, Stocks et al. 1989, Kunkel 2001) but have a low crown fi re potential (Van Wagner 1977, 1993, Artsybashev 1983, Grishin 1997, Scott 1998, Scott and Reinhardt 2001).In contrast, stands with a dense understory of shrubs or immature trees have high vertical fuel continuity.If they have a patchy overstory, i.e., poor horizontal fuel continuity in the canopy layer, they are less likely to burn because of the typically moister microenvironment but readily support passive crowning (torching) and spotting under low relative humidity.Stands with high vertical and horizontal fuel continuity have the highest crown fi re potential (Van Wagner 1977, 1993, Alexander 1998, Finney 1998, 1999, Scott 1998, Scott and Reinhardt 2001).The availability of these fuels varies not only in space, but also in time with changes in weather, principally relative humidity, temperature, and drought (Johnson 1992, Bessie andJohnson 1995).
Weather, specifi cally relative humidity, wind, and drought, defi ne the fraction of the total fuel that is available to be consumed in a given fi re.The short-term weather history is the primary determinant of the fl ammability of the moss and lichen layers, loose litter, foliage, and fi ne twigs (Albini 1976, Stocks et al. 1989).Long-term weather determines the moisture content and combustibility of deeper organic layers and surface logs (Stocks et al. 1989).Wind is perhaps the single most important cause of spatial and temporal variation within boreal forests.Fires frequently pulsate between intense surface fi res and crown fi res with only modest changes in wind speed (Van Wagner 1977, 1993, Finney 1998, Scott 1998, Scott and Reinhardt 2001).The result is a mosaic of small crown fi re patches instead of the large expanses that occur in sustained winddriven fi res.
In ecological studies terrain needs to be considered in the broader context of how it affects not only the site water and energy budgets that dominate site productivity, but also fuel moisture (Rothermel 1972, Albini 1976, Andrews 1986, Miyanishi 2001) and wind fl ow patterns (Pyne et al. 1996).Thus, terrain includes slope angle, aspect, elevation, hill-slope drainage, and land-form.An additional aspect of terrain is that terrain features can present barriers to fi re spread.Barriers have two general effects.Either the area's fi re frequency is reduced due to its reliance on local ignitions as discussed in fi re frequency above, or at scales smaller than the typical fi re's size, the fi re fl anks or backs into the area with a reduced intensity and severity (Catchpole et al. 1982, 1992, Agee et al. 2000, Finney 2001).
Independent of the biophysical environment in which the fi re is burning, major differences in fi re behavior are associated with the location on the fi re's perimeter, that is whether an area is burned by a heading fi re, fl anking fi re, or backing fi re (Catchpole et al. 1982(Catchpole et al. , 1992) (Fig. 2b).The heading portion of the fi re burns with the wind or upslope.The backing fi re burns into the wind or down slope.The fl anking fi re burns perpendicular to the wind's axis.The direction of fi re spread is a function of the slope and wind vectors, with the latter dominating except at low wind speeds weather, and vegetative structure and with b) whether or not the area experiences a head fi re, fl ank fi re, or backing fi re.c) As the fi re behavior changes so do the effects.(Rothermel 1972, Albini 1976, Finney 1998).The greater the wind speed or slope, the greater the difference between the intensity of the heading fi re and backing fi re.Commonly, fi reline intensity in a backing fi re is on the order of 0.1 to 0.2 times that of a heading fi re in a given biophysical environment, while fl anking fi res are about 0.4 to 0.6 times the headfi re intensity (Catchpole et al. 1992).Variations in the fi re environment and location on the fi re perimeter lead to signifi cant variations in the fi re behavior and effects (Fig. 2c).For example, it is common to see fi res spread across a slope running with the wind when the vegetation structure is not suffi cient and continuous enough for the fi re to carry up the slope.The ignition pattern that is used in a restoration burn can also be expected to affect the pattern of fi re behavior.
Fireline intensity (kW/m) is the product of the fuel value, i.e., the fuel's heat content (kJ/kg), the mass of fuel consumed (kg/m 2 ), and the rate of spread (m/s) (Byram 1959).It is proportional to the fl ame length in a spreading fi re and is a useful measure of the potential to cause damage to aboveground structures (Van Wagner 1973, Alexander 1982, Ryan and Noste 1985).Rothermel (1972) defi ned a somewhat different measure of fi re intensity, heat per unit area, which is commonly used in fi re behavior prediction in the United States (Albini 1976, Andrews 1986, Scott 1998, Scott and Reinhardt 2001).Likewise, the Canadian forest fi re danger rating system calculates the intensity of surface fi res and crown fi res (Stocks et al. 1989).One problem with using current fi re behavior prediction systems in ecological studies is that they do not predict all of the combustion, and therefore all of the energy released, over the duration of the fi re (c.f.Johnson and Miyanishi 2001).
Fires burn throughout a continuum of energy release rates (Table 1) (Artsybashev 1983, Rowe 1983, Van Wagner 1983, Rothermel 1991).Ground fi res burn in compact fermentation and humus layers and in organic muck and peat soils (Fig. 3a).Ground fi re spread is predominantly by smoldering combustion.Such fi res typically burn for hours to weeks, exhibit forward rates of spread in the range of decimeters to meters per day, and exhibit temperatures in excess of 300 °C for several hours (Frandsen and Ryan 1986, Ryan and Frandsen 1991, Hartford and Frandsen 1992, Agee 1993) (e.g., Fig. 3b).The conditions necessary for ground fi res are organic soil depth greater than about 4 to 6 centimeters and extended drying  a) Rates of spread, fl ame length and fi reline intensity vary widely in transitional fi res.In subalpine and boreal fuels it is common for surface fi res to creep slowly until they encounter conifer branches near the ground, then individual trees or clumps of trees torch sending embers ahead of the main fi re.These embers start new fi res, which creep until they encounter trees, which then torch.In contrast, as surface fi res become more intense, torching commonly occurs prior to onset of active crowning.b) Flame lengths are highly variable in crown fi res.They commonly range from 0.5 to 2 times canopy height.Fire managers commonly report much higher fl ames but these are diffi cult to verify or model.Such extreme fi res are unlikely to result in additional fi re effects within a stand but are commonly associated with large patches of continuous severe burning.(Reinhardt et al. 1997, Miyanishi 2001).Surface fi res spread by fl aming combustion in loose litter, woody debris, and understory herbaceous and shrubby plants typically less than two meters tall.Under marginal burning conditions surface fi res creep along the ground at rates of decimeters per hour with fl ames less than fi ve decimeters (Table 1).As fuel, weather, and terrain conditions become more favorable for burning, surface fi res become progressively more active with spread rates ranging from tens of meters to kilometers per day.The duration of surface fi res is on the order of one to a few minutes (Vasander andLindholm 1985, Frandsen andRyan 1986, Hart- ford and Frandsen 1992) except where extended residual burning occurs beneath logs or in concentrations of heavy woody debris (e.g., Fig. 4).
Here fl aming combustion may last a few hours resulting in substantial soil heating (Hartford and Frandsen 1992).If canopy fuels are plentiful and suffi ciently dry, surface fi res begin to transition into crown fi res (Van Wagner 1977, Scott andReinhardt 2001).Crown fi res burn in the foliage, twigs, and epiphytes of the forest or shrub canopy above the surface fuels (Fig. 5a).Such fi res exhibit the maximum energy release rate but are typically of short duration, 30 to 80 seconds (Fig. 5b).Fires burn in varying combinations of ground, surface, and crown depending on the local conditions at the specifi c time a fi re passes a point.Ground fi res burn independently from surface and crown fi res and often occur some hours after passage of the fl aming front (Artsybashev 1983, Rowe 1983, Van Wagner 1983, Hungerford et al. 1995).Changes in surface and ground fi re behavior occur in response to subtle changes in the microenvironment, stand structure, and weather leading to a mosaic of fi re treatments at multiple scales in the ground, surface and, canopy strata.

Fire Severity
Within the fi re effects literature, there is increasing acceptance of the use of the term fi re severity to describe the ecological impacts of fi res (Rowe 1983, Ryan and Noste 1985, Moreno and Oechel 1989, Turner et al. 1994, Schimmel and Granström 1996, Smith and Fischer 1997, DeBano et al. 1998, Feller 1998, Neary et al. 1999, Kafka et al. 2001).Given the many ecological components and their peculiar metrics, it is not possible to come up with a single system to quantify fi re severity (DeBano et al. 1998).Consistent criteria for classifying fi re severity have yet to emerge.Numerous authors have used measures of the depth of burn into the organic soil horizons or visual observation of the degree of charring and consumption of plant materials to defi ne fi re severity (cf.Rowe 1983, Schimmel and Granström 1996, DeBano et al. 1998, Feller 1998, Pérez and Moreno 1998).However, focusing severity only on ground-based processes ignores the aboveground dimension of severity implied in the defi nition of White and Pickett (1985).This is especially important because soil heating is commonly shallow even when surface fi res are intense (Wright and Bailey 1982, Frandsen and Ryan 1986, Hartford and Frandsen 1992) (Fig. 4).Ryan and Noste (1985) summarized literature on depth of burn and charring of plant materials and developed descriptive characteristics.A condensed and revised (Moreno and Oechel 1989, Pérez and Moreno 1998, DeBano et al. 1998, Feller 1998) description of their characteristics is provided for clarifi cation of subsequent discussion of fi re effects.
Unburned: Plant parts are green and unaltered, there is no direct effect from heat.Scorched: Fire did not burn the area but radiated or convected heat caused visible damage.Mosses and leaves are brown or yellow but species characteristics are still identifi able.Soil heating is  In areas with deep organic soils deep depth-ofburn occurs when ground fi res consume the rootmat or burn beneath the root-mat.Trees often topple in the direction from which the smoldering fi re front approached.
The moderate depth of burn class is a broad class.Some investigators have chosen to divide the class into two classes (c.f.Feller 1998).In practice I have found it diffi cult to do so on the basis of post-hoc examination of the mineral soil alone but rely on the preponderance of the evidence, which includes reconstructing the prefi re vegetative structure.The depth-of-burn characteristics are appropriate for quadrat-level descriptions.At higher spatial scales logic needs to be developed for defi ning fi re severity on the basis of the distribution of depth of burn classes (c.f.Ryan andNoste 1985, DeBano et al. 1998).Ryan and Noste (1985) combined fi re intensity classes with depth of burn (char) classes to develop a two-dimensional matrix approach to defi ning fi re severity.The basis for these characteristics is that fi re-intensity classes qualify the relative energy release rate for a fi re, whereas depth-of-burn classes qualify the relative duration of burning.Their concept focuses on the ecological work performed by fi re both above ground and belowground.The matrix provides an approach to classifying the level of fi re treatment or severity for ecological studies at the scale of the individual and the community.The approach has been used to interpret differences in plant survival and regeneration (Willard et al. 1995, Smith andFischer 1997) and to fi eld-validate satellite-based maps of burned areas (White et al. 1996).

Synergy
Synergism occurs when multiple disturbance factors interact (White and Pickett 1985).There are numerous examples of synergy related to fi recaused disturbances.For example, fi re injury in conifers leads to increased incidence of insect attack on both short (Ryan andAmman 1994, 1996) and long (Geiszler et al. 1984, Gara et al. 1985) time scales.Insect-caused mortality leads to increased fuel and fi re potential (Lotan et al. 1985).High winds periodically blow down large forest tracts.The increased solar drying, surface wind, and the heavy fuel that result increase the likelihood and potential severity of subsequent wildfi re (Stocks 1975, Ryan 2000b, Finney 2001).Perhaps the best example of synergy is the fi re/fl ood cycle common to mountainous areas of the world.The potential for erosion increases with depth of burn, surface fi re intensity, and slope steepness (Wright and Bailey 1982, DeBano et al. 1998, Gresswell 1999, Pannkuk et al. 2000).

Fire Severity and Fire Effects
Fire severity changes as vegetation structure changes (Olson 1981, Van Wagner 1983, Turner and Romme 1994, Schimmel and Granström 1997, Kafka et al. 2001).At least at a coarse scale, predicted or observed fi re severity can be integrated with plant vital attributes (Noble and Slatyer 1977, 1980, 1981, Noble 1981, Rowe 1983) to postulate which individuals are likely to survive and which species are likely to increase vs. decrease following a fi re.It is possible to predict survivorship of trees based on knowledge of fi re behavior, tree morphology, and species' life history attributes (c.f.Peterson and Ryan 1986, Ryan and Reinhardt 1988, Ryan 1990, 1998, Johnson 1992, Agee 1993, Fischer et al. 1996, Reinhardt et al. 1997, Dickinson and Johnson 2001).In general, the probability of crown injury decreases with increasing plant height and height of the live-crown-base, at roughly the two-thirds power of the fi reline intensity (Van Wagner 1973, Ryan 1998, Dickinson and Johnson 2001).Resistance to stem injury increases with the square of the bark thickness, which increases approxi-mately linearly with tree diameter (c.f.Sofronov and Volokitina 1977, Peterson and Ryan 1986, Ryan and Reinhardt 1988, Johnson 1992, Agee 1993).The diameter-to-bark-thickness ratio is a species-specifi c parameter commonly available in the forest mensuration literature.Thickbarked trees frequently survive active surface fi res, whereas thin-barked trees only occasionally survive creeping fi res.Survival of thin-barked trees usually only occurs when the fi re is patchy and does not circumnavigate the entire stem or where only fi ne, light, fl ashy surface fuels (e.g., lichens and grasses) are quickly consumed.Thick-barked and deep-rooted trees commonly survive ground fi res several centimeters deep (Ryan and Frandsen 1991) but not deep peat fi res (Artsybashev 1983).Shallow-rooted species rarely survive ground fi res (Wein 1983, Johnson 1992, Smith and Fischer 1997).
In contrast to Eurasian boreal forests that contain a number of thick-barked fi re resistant trees (Sofronov andVolokitina 1977, Nikolov andHelmisaari 1992), only red pine (Pinus resinosa) is fi re resistant in the North American boreal (Heinselman 1981, Engstrom andMann 1991).Red pine is a minor component of the North American boreal forest (Heinselman 1981).In the North American boreal crown fi res are more common than in the Eurasian boreal (Shvidenko and Nilsson 2000), apparently because of different stand structure.Due to thin bark, active surface fi res kill most trees in North American boreal fi res regardless of depth-of-burn (Johnson 1992).In contrast, mortality in similar fi res in pine and larch forests in the Russian boreal is approximately 20 percent in light depth-of-burn fi res (Shvidenko and Nilsson 2000).Although surface fi res and crown fi res both cause high mortality in thin-barked species, their effects on other ecosystem processes can be expected to be quite different.For example, the surface microenvironment, shaded vs. exposed, in surface fi res vs. crown fi res can be expected to affect post-fi re species dynamics (Rowe 1983).Needles are killed by heat rising above the fi re (Van Wagner 1973, Dickinson andJohnson 2001), thereby retaining their nutrients.Thus, litterfall of scorched needles vs. no litterfall in crown fi re areas can be expected to affect nutrient cycling.Rainfall simulator experiments have shown litterfall also reduces erosion on sites where duff was completely consumed (Pannkuk et al. 2000).Thus, a lethal stand replacement crown fi re, i.e., one that kills the dominant overstory (Brown 2000) represents a more severe fi re treatment than a lethal stand replacement surface fi re.
Many non-tree species in boreal forests rely primarily on vegetative reproduction from belowground tissues to survive fi res (Flinn and Wein 1977, Flinn and Pringle 1983, Granström and Schimmel 1993, Schimmel and Granström 1996).Noble andSlatyer (1977, 1980) described plant adaptive traits affecting fi re survival.Rowe (1983) summarized these traits.Survival of species that rely on vegetative reproduction, the V and W species in Rowe (1983) (Table 2) depends on the depth of burn relative to the depth of the meristematic tissue (Flinn and Wein 1977, Wein 1983, Schimmel and Granström 1996).Organic soils are excellent insulators (Uggla 1974, Vasander and Lindholm 1985, Hungerford et al. 1991, Hartford and Frandsen 1992).Unburned residual organic soil effectively protects deeper levels from significant temperature rise.Survival of species with V and W reproductive strategies is relatively independent of surface fi re intensity, as aboveground organs are nearly uniformly killed by all fi re inten-sities (Granström and Schimmel 1993, Schimmel and Granström 1996, Feller 1998).
Reproduction from seeds, either from on-site seed banks (S and C species) or from off-site dissemination (D species) (Table 2) is also differentially affected by fi re severity (Granström andSchimmel 1993, Schimmel andGranström 1996).Similar to V and W species the deeper the depth of burn, the greater the destruction to the soil seed bank (S species).A number of S species store seeds for decades to centuries awaiting the next fi re (Heinselman 1981, 1996, Keeley 1981).Low depth of burn may or may not favor soil seed bank species (S species) depending on the depth of the stored seed (Granström 1982, Rydgren andHestmark 1997).The fate of S species is relatively unaffected by surface fi re intensity except when such seeds are exposed on the surface.In contrast, canopy-stored (C species) seed bank survival is reduced by severe heating in intense surface and crown fi res (Despain et al. 1996).Even species storing seeds in serotinous cones experience reduced survival in crown fi res (Ellis et al. 1994, Despain et al. 1996).However, C species are relatively insensitive to the depth of burn because the lethal temperature height above ground fi res is insignifi cant.Fire severity affects Mode of regeneration and reproductionfi rst vital process (Noble and Slatyer 1980) Vegetative-based V species -able to resprout if burned in the juvenile stage W species -able to resist fi re in the adult stage and to continue extension growth after it (although fi re kills juveniles) Disseminule-based D species -with highly dispersed propagules S species -storing long-lived propagules in the soil C species -storing propagules in the canopy Communal relationshipssecond vital process (Noble and Slatyer 1980) T species -tolerants that can establish immediately after a fi re and can persist indefi nitely thereafter without further perturbations R species -tolerants that cannot establish immediately after fi re but must wait until some requirement has been met (e.g., for shade) I species -intolerants that can only establish immediately after a fi re.
Rapid growth pioneers, they tend to die out without recurrent disturbances regeneration of D species by determining the quality of the seedbed and the amount of early post-fi re competition.The mode of regeneration and reproduction (i.e., the fi rst vital process) and the communal relationship (i.e., the second vital process) (Noble and Slatyer 1977, 1980, Noble 1981), can be evaluated in the context of the fi re severity matrix (Fig. 6) and Table 2. Three examples can be used to illustrate differential responses.First, fi res with low depth of burn favor vegetative regeneration by V and W species.However, the surface fi re intensity will tend to favor one mode over another.A V species that establishes immediately after a disturbance (VT species) can be expected to respond more favorably than a species that requires shade (e.g., VR species) following low depth of burn crown fi res.Second, deep depth of burn fi res favor  Variability in fi re severity affects both the amount and depth of seed and belowground organs that survive fi re, as well as the communal relationships.These lead to spatial variability in the initial post-fi re fl oristics.Secondary successional processes increasingly dominate population dynamics with time since fi re.However, a mechanistic understanding of fi re severity can be coupled with knowledge of species attributes to further understand the role of fi re as a disturbance process.Clearly, more integrated fi eld studies are needed to better understand how complex phytosociological, ecophysiological, and fi re science relationships affect biodiversity.
Models exist for predicting the energy output of fi res in relatively homogeneous stands (cf.Albini 1976, Andrews 1986, Stocks et al. 1989, Van Wagner 1998) and for multiple stands across landscapes (Finney 1998).However, fi re behavior models predict rates of spread (m/s), fi reline intensity (kW/m), heat per unit area (kW/m 2 ), and fl ame length (m) (Table 1), whereas fi re effects on fl ora, fauna, and soils are typically described in terms of temperature histories (c.f.Wright and Bailey 1982, Hungerford et al. 1991, DeBano et al. 1998).The energy output of a fi re cannot be directly related to the temperature history of an entity within a fi re.The temperature reached by an entity in a fi re depends not only on the fi re's behavior but also the heat transfer mechanism and the thermal properties of the heat transfer medium.The only fi re effect that has been clearly demon-strated to relate to the existing fi re behavior model outputs is crown scorch which is defi ned as the height of the 60 °C lethal isotherm (Van Wagner 1973).Empirical measurements of the temperature history associated with fi res of varying behavior provide insights into the type of effects to be expected (Fig. 6).Active surface fi res with no secondary burning (i.e., no organic soil or coarse woody fuel consumption) are of suffi cient temperature (310 °C) and duration to kill foliage, thinbarked stems, and exposed seeds near the surface but not those in the forest canopy or deeper than 2 centimeters in the mineral soil (Fig. 4, Fig. 6 lower left temperature history).Crown fi res burning over wet organic soil horizons produce substantially higher surface temperature (1060 °C) but the depth of lethal heat penetration (60 °C) is only 4 centimeters (Fig. 5, Fig. 6 upper left temperature history).In contrast, a smoldering ground fi re in the absence of an active surface fi re causes few effects above the ground level but temperatures in excess of 300 °C persist for several hours at depths greater than 4 centimeters (Fig. 3, Fig. 6 lower right temperature history).The combined effect of a crown fi re followed by a ground fi re provides the maximum potential to kill aboveground and belowground tissues (Fig. 6 upper right temperature history).These representative temperature histories can be related to the fi re severity matrix developed by Ryan and Noste (1985).This illustrates the combined effect of the heat pulse up, represented by the fl ame length class, and the heat pulse down, represented by depth of burn class.As the fl ame length increases, the fi re's ability to cause ecological change to aboveground ecosystem components increases.Likewise, as the depth of burn increases, the fi re's ability to cause ecological change to belowground ecosystem components increases.
Active growing tissues die at lower temperatures or exposure times than dormant tissues (Wright andBailey 1982, Peterson andRyan 1986).Seeds are more resistant to temperature than active growing tissues (Wright andBailey 1982, DeBano et al. 1998).Killing of plant tissues is, however, only one of a number of fi re effects related to high temperatures (Hungerford et al. 1995, DeBano et al.1998).Campbell et al. (1994Campbell et al. ( , 1995) ) developed and Albini et al. (1996) tested a model for predicting temperature histories in the soil given fi res of different energy release rates and durations.The model predicts the effects of varying soil texture and moisture content on temperature history.The model can be used to predict the maximum temperature for varying depths of burn (Fig. 7).An increase in the depth of burn increases the depth at which any temperaturerelated effect will occur.This further illustrates that the greater the depth of burn, the more severe the disturbance.Models are available to predict the depth of burn in many North American vegetation types (Reinhardt et al. 1997, Johnson 1992).These empirical models provide a fi rst approximation for predicting the effects of fi re on belowground ecosystem components.By cou-pling these predictions with fi re behavior models (Andrews 1986, Stocks et al. 1989, Finney 1998, 1999, Scott and Reinhardt 2001), ecologists engaged in restoration burns can predict the severity of disturbance in advance and can compare the expected disturbance to historical reference conditions.It is important, however, to account for spatial variation in model input parameters.

Conclusions
Fires are a dominant force in shaping boreal landscapes.The severity of these fi res varies in time  Campbell et al.1994Campbell et al. , 1995)).Conditions are for coarse dry soil, which provides the best conduction (i.e., a worst-case scenario).

Ryan
Dynamic Interactions between Forest Structure and Fire Behavior in Boreal Ecosystems and space depending on the vegetation structure, terrain, short-and long-term weather, and location on the fi re's perimeter (head vs. fl ank vs. rear).
Fire behavior and effects models can be used to understand this variation.Individual plants and species vary in their resistance to fi re injury in predictable ways.A combined understanding of plant attributes and fi re severity concepts can be used to predict or evaluate the effects of wildfi res on biodiversity and for defi ning restoration goals.
In order to understand the effects of fi re-caused disturbances on bioconservation and restoration more emphasis is needed on fi eld studies that integrate complex phytosociological, ecophysiological, and fi re science relationships.

Fig. 1 .
Fig. 1.Landscapes are aggregates of stands with multiple disturbance histories that affect the potential for future disturbances (from Arno et al. 1993).Photo by Steve F. Arno.

Fig. 2 .
Fig. 2.Fire behavior varies in time and space with a) changes in the terrain, weather, and vegetative structure and with b) whether or not the area experiences a head fi re, fl ank fi re, or backing fi re.c) As the fi re behavior changes so do the effects.

Fig. 3 .
Fig. 3. a) Example of smoldering ground fi re in deep duff (average 17 cm) (from Ryan and Frandsen 1991).b) Such fi res typically produce temperatures in excess of 300 °C for several hours.Duff depth = 6.5 cm, moisture content = 18.3% (from Hartford and Frandsen 1992).Photo by Kevin C. Ryan.

Fig. 4 .
Fig. 4. Example of an active fl aming surface fi re in grassland.Such fi res typically produce surface temperatures in excess of 300 °C but only for 1 to 2 minutes.

Fig. 5 .
Fig. 5. a) Example of a crown fi re in jack pine (Pinus banksiana) in the Northwest Territories, Canada.b) Such fi res typically produce temperatures in excess of 1000 °C for about 1 minute.Photo by Kevin C. Ryan.

Fig. 6 .
Fig. 6.Representative temperature histories (top) for fi res of varying severity: A-crownfi re/low depth of burn (DOB), B-crownfi re/moderate DOB, C-active surface fi re/low DOB, D-creeping surface fi re/ moderate DOB.(See text and Table1for fi re intensity and DOB descriptions).Relationship between fi re severity and the mode of regeneration following fi re.(See Text and Table2for description of plant regeneration attributes).Changes in site variables, including terrain and vegetative structure, and weather variables lead to fi res of differing peak temperature and duration.Arrows indicate increasing site and weather potential.Both site and weather conditions must be met to affect fi re severity.

Fig. 7 .
Fig. 7. Temperature ranges associated with various fi re effects (top) (from Hungerford et al. 1991) compared to the depth of heat penetration into mineral soil (bottom) for a crown fi re over exposed mineral soil (observed in jack pine Pinus banksiana in the Canadian Northwest Territories) or for ground fi re burning in 5-, 15-, and 25-cm of duff (predicted viaCampbell et al.1994Campbell et al. ,  1995)).Conditions are for coarse dry soil, which provides the best conduction (i.e., a worst-case scenario).

Table 1 .
Representative ranges for fi re behavior characteristics for ground, surface, and crown fi res.
RyanDynamic Interactions between Forest Structure and Fire Behavior in Boreal Ecosystems negligible.Light: In forests the surface litter, mosses, and herbaceous plants are charred to consumed but the underlying forest duff or organic soil is unaltered.