Forest Ecology and Management 242 (2007) 776–790 www.elsevier.com/locate/foreco Modeling the long-term effects of fire suppression on central hardwood forests in Missouri Ozarks, using LANDIS ZongBo Shang a,*, Hong S. He a, David E. Lytle b, Stephen R. Shifley c, Thomas R. Crow d a School of Natural Resources, University of Missouri-Columbia, 203 Anheuser-Busch Natural Resources Building, University of Missouri-Columbia, Columbia, MO 65211, USA b The Nature Conservancy, Ohio Chapter, 6375 Riverside Drive, Suite 50, Dublin, OH 43017, USA c USDA Forest Service, North Central Research Station, 202 Anheuser-Busch Natural Resources Building, University of Missouri-Columbia, Columbia, MO 65211, USA d USDA Forest Service, WFWAR, Stop Code 1113, 1400 Independence Avenue, SW, Washington, DC 20090, USA Received 15 March 2006; received in revised form 19 February 2007; accepted 19 February 2007 Abstract Fire suppression has been found to dramatically change fire regimes, lead to accumulation of fuels, and alter forest composition and species abundance in the Central Hardwood Forests in the Missouri Ozarks, United States. After a half century of fire suppression, fire hazards have increased to a high level and high intensity fires are more likely to occur. We used LANDIS, a spatially explicit landscape dynamics model, to simulate the long-term effects of fire suppression on forests in Missouri Ozarks. Specifically, we examined to what extent fire suppression would affect fuel loads and fire hazards, and how fire suppression would affect forest tree species abundance. Using a spatial modeling approach, we conducted 200-year simulations of two management scenarios: (1) a fire suppression scenario circa 1990s and (2) a historic fire regime scenario prior to fire suppression, with a mean fire-return interval of 14 years. Under the fire suppression scenario, the simulation showed that both fine and coarse fuels were at a medium-high level after a few more decades of fire suppression. Fire hazard also rapidly increased to a medium-high level within a few decades. After one century of fire suppression, simulated fire intensity increased to a dangerous level, with more than 3/4 of the fires at a medium-high intensity level. Fire suppression also led to distinct changes in species abundance; the pine and oak–pine forests which used to dominate the study area prior to fire suppression were replaced by mixed-oak forests. This study suggests that it may be desirable to re-introduce frequent fire. By greatly increasing the use of fire over current management levels, our simulation suggests less accumulation of dangerous fuels, reduced fire hazard, and decreased occurrence of high intensity fires. Results imply that frequent fire would greatly increase the abundance of fireresistant species (e.g., shortleaf pine) and decrease the abundance of more fire-sensitive species such as red oaks. Such a compositional shift should also decrease the recent phenomenon of widespread oak decline events. # 2007 Elsevier B.V. All rights reserved. Keywords: Fire suppression; Historic fire regime; Fire hazard; Oak; LANDIS model; Fire modeling 1. Introduction Understanding the effects of long-term fire suppression on fire regimes and forest ecosystems has become increasingly important when designing scientifically sound management plans (e.g., Johnson and Miyanishi, 1995; Backer et al., 2004; Dellasala et al., 2004; Dombeck et al., 2004). Fire suppression * Corresponding author at: Wyoming Geographic Information Science Center, University of Wyoming, Dept 4008, 1000 E. University Avenue, Laramie, WY 82071, USA. Tel.: +1 307 766 6207; fax: +1 307 766 2744. E-mail addresses: zshang1@uwyo.edu, shangzongbo@yahoo.com (Z. Shang). 0378-1127/$ – see front matter # 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.foreco.2007.02.026 efforts usually, but not exclusively, result in a departure from historic fire regimes. For example, the Government Accounting and Office (1999) reported that areas burned in large, intense fires have increased in National Forests of the western U.S. since the inception of fire suppression efforts. Baker (1992) found that both fire frequency and area burned decreased markedly with fire suppression efforts in the pine and spruce-fir forest within the Boundary Waters Canoe Area (BWCA) of Minnesota. Likewise, Romme and Knight (1981) and Agee and Huff (1987) reported an increase in fire intensity due to the accumulation of fuels associated with fire suppression in the Pacific Northwest forests. Many forest ecosystems show evidence of significant accumulations of surface and crown fuels since the inception Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 of effective fire suppression (e.g., Romme and Knight, 1981; Brown and Bevins, 1986; Agee and Huff, 1987; Clark, 1990; Baker, 1992; Johnson et al., 2001) in the 1940s and 1950s. Evidence from fire statistics on Forest Service lands indicates that after six or more decades of fire suppression, fires are more difficult to control (Conard et al., 2001) and more expensive to fight (Schuster et al., 1997). In the Missouri Ozarks forests, effective fire suppression over the last five to six decades has increased surface fuel loadings (Grabner et al., 1999) and there is concern that high fuel accumulations could foster large, intense wildland fires that were historically rare (Westin, 1992). Historic fire-return intervals (FRIs) ranged from approximately 18 years during the Native American depopulation period (1200–1760 A.D.) (Guyette and Larsen, 2000), to around 12 years during the Native American repopulation period (1760–1825 A.D.) (Guyette and Larsen, 2000; Guyette et al., 2002), to about 4 years during the Euro-American population settlement period (1825–1940 A.D.) (Pyne, 1982; Guyette et al., 2002). Fire suppression became increasingly effective in Missouri Ozarks after the 1940s, and fire records in Missouri (Westin, 1992) show that current wildfire-return intervals are 400 years or longer in most Missouri forests. Fire suppression, as well as other human activities, e.g., heavy logging during the late 19th and early 20th century, have dramatically changed forest components and species abundance in southeast Missouri. Historic reconstructions of tree species composition in the early 19th century (Batek et al., 1999) showed that shortleaf pine and white oak were the dominant tree species in the southeast Missouri before settlement by Euro-American. Several studies (Guyette and Dey, 1997; Batek et al., 1999) have reported a decline in the abundance of shortleaf pine (Pinus echinata Mill.) in the 20th century. Present shortleaf pine abundance in southeast Missouri is only 20–25% of that circa 1900. This decline was primarily caused by the heavy logging between 1890 and 1920 and has been exacerbated by fire suppression that began after 1940. In southeast Missouri, the abundance of white oaks in the 19th century was lower than current levels (Brookshire and Shifley, 1997; Shifley and Brookshire, 2000). White oak (Quercus alba L.)’s abundance is now ranked first among tree species in southeast Missouri. Other dominant tree species include black oak (Q. velutina Lam.), shortleaf pine (Pinus echinata Mill.) and sugar maple (A. saccharum Marsh.). Black oak (Q. velutina Lam.) and post oak (Q. stellata Wangenh.) abundance have been reported to be strongly and positively correlated with mean fire-return interval (Batek et al., 1999). Fire suppression favored the development of black oak and post oak forests (Batek et al., 1999). In southeast Missouri, the abundance of black oaks has significantly increased after the settlement by Euro-American (Batek et al., 1999; Abrams, 2003). In the Missouri Ozarks, effective fire suppression has been conducted for half a century, and will be continued in the future. The purpose of this study is to evaluate the potential changes in forest fuel loads and composition resulting from a continuation of this management practice. In addition, we contrast these effects with those that result from a re-introduction of the pre 777 Euro-American settlement fire regime. The comparison between these two scenarios (fire suppression and historic fire regime) could help us understand the potential problems caused by long-term effective fire suppression (e.g., build up fuels and fire hazards, danger of catastrophic fires and changes in species abundance). Simulation study on the historic fire regime scenario also helps us to evaluate the potential effects on forest ecosystems if we re-introduce fires into forests. More specifically, our goals were to model the long-term effects of two different fire management scenarios, continued fire suppression and a historic frequent fire regime, on tree species abundance and age structure, fuel loads, and fire hazard. We conducted this study using a modeling approach, as long-term empirical studies of fire effects over large landscapes are virtually impossible to conduct, and ecological models have proved to be useful tools for studying fire disturbance at landscape scales (e.g., Baker, 1992; He and Mladenoff, 1999a; Hargrove et al., 2000). We applied a spatially explicit landscape model of forest succession and fire disturbance, LANDIS (Mladenoff and He, 1999; He and Mladenoff, 1999a), and simulated two fire regimes: a historic fire regime (prior to fire suppression) and a fire suppression regime. For both regimes we examined and compared fire frequency, area burned, fire intensity, and fire size. We then compared predicted fuel loads and vegetation changes under these two fire regimes. 2. Methods 2.1. Study area We selected the southeastern Missouri Ozarks as our study area (Fig. 1A). The study area includes large, contiguous blocks of the Mark Twain National Forest that is surrounded by (and in some areas intermixed with) state and private forest ownerships. Information on species composition and ecological land types in the Mark Twain National Forest were readily available (Miller, 1981; Brookshire and Shifley, 1997; Shifley and Brookshire, 2000; Shifley et al., 2000a,b). In addition, sufficient historic data on fire regimes (e.g., Haines et al., 1972; Westin, 1992; Guyette and Larsen, 2000; Guyette et al., 2002) and studies on historic vegetation (e.g., Batek et al., 1999) exist in this area are helpful for validating our simulation results. Heavy logging between 1890 and 1920 (Shifley et al., 2000b) and fire suppression since 1940 (Guyette et al., 2002) have favored the development of mixed-oak forests in the region and have decreased the abundance of shortleaf pine (Pinus echinata Mill.) relative to the mid-19th century (Batek et al., 1999). Present-day forests are predominantly mixtures of white oak (Quercus alba L.), post oak (Q. stelatta Wangenh.), black oak (Q. velutina Lam.), scarlet oak (Q. coccinea Muenchh.), hickory (Carya spp.), and shortleaf pine. 2.2. LANDIS model LANDIS is a spatially explicit simulation model of forest landscape change in response to disturbance, succession and management (Mladenoff et al., 1996; Mladenoff and He, 1999; 778 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 Fig. 1. Study area in southeastern Missouri Ozarks and the land types. The study area covers 712 km2 Mark Twain National Forests, and 576 km2 non-federal lands. He and Mladenoff, 1999a). LANDIS is designed to simulate ecological dynamics and forest management at large extents (103 to 106 ha), over long periods of time (101 to 103 years). Within each 10-year iteration, the model can simulate: (a) species-level forest succession (i.e., vegetative reproduction, establishment, growth and competition among species, and mortality); (b) seed dispersal; (c) fire ignition, spread, intensity and severity; (d) windthrow occurrence, size, intensity and severity; (e) timber harvest size, type and frequency; and other silvicultural treatments. Each of these elements affects forest species composition and age structure on the landscape. We have recently developed a new fuel module for LANDIS 4.0 (He et al., 2004, 2005; Shang et al., 2004). LANDIS 4.0 (a) tracks fuel types and fuel loads; (b) incorporates a new approach to simulate fire effects on vegetation and fuels; (c) evaluates potential fire risks in forest landscapes; (d) simulates the long-term effects of fire management (e.g., prescribed burning or thinning) on fuel loads, fire frequency, tree species composition, and age structure; (e) provides a practical tool for resource and land managers to evaluate multiple fuel treatment scenarios and their impacts on forest ecosystems. Fuel loads as well as wildfire intensity are modeled as five categorical classes in the LANDIS fuel module: low (class 1); medium-low (class 2); medium (class 3); medium-high (class 4); high (class 5). Those categories are different from the fuel models used in the National Fire Danger Rating System (Deeming et al., 1977) or the standard National Forest Fire Laboratory (NFFL) fuel models (Anderson, 1982). In the LANDIS model, those five categories only capture the fuel loads from low to high, or the fire intensity from low intensity surface fire to high intensity surface fire. While explicitly modeling fuel biomass is possible, categorical fuel loads are sufficient for relative comparisons of responses to alternative disturbance or management scenarios, and such comparisons are the principle utility of landscape simulation models (Barrett et al., 2001). LANDIS does not independently predict the occurrence of individual disturbance or management events. Rather, it is a scenario model that compares long-term, large-scale, simulated effects of user-defined disturbance and management scenarios on real or hypothetical landscapes. LANDIS can operate with pixels (raster units) that range in size from 0.01 ha to more than 1 km2; in this study we used 30 m  30 m (0.09 ha) pixels. Additional information about LANDIS design and application is available from other sources (e.g., Mladenoff and He, 1999; He and Mladenoff, 1999a; Gustafson et al., 2000). The following paragraphs give a brief introduction to the modeling algorithms of the LANDIS model. 2.2.1. Succession and seed dispersal LANDIS is designed to simulate forest dynamics at large extents, by simulating individual species as 10-year age cohorts (Mladenoff et al., 1996). For each site, individual species are recorded as presence/absence of 10-year age cohorts. LANDIS (He and Mladenoff, 1999b) simulates forest succession dynamics for individual species (age class presence/absence) including seed dispersal and establishment, vegetative reproduction, growth, death (when reaching longevity), and mortality caused by disturbance. Seed dispersal is modeled as a function of species’ effective and maximum seeding distance. A negative exponential distribution is used to describe seeding probability in relation to distance from available seed sources. Seed establishment is determined based on shade tolerance of the seeding species relative to the species already occurring on the site. Vegetative reproduction is simulated stochastically based on the species’ sprouting probability. When species approach their longevity, mortality rate is modeled to increase. LANDIS deterministically removes the Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 779 species age cohort when the longevity of the species is reached. Mortality caused by disturbance (e.g., fire, windthrow, insect and disease), based on the disturbance intensity, species tolerance and age-specific susceptibility, is simulated by each disturbance module respectively. Details about LANDIS succession and seed dispersal has been published by He and Mladenoff (1999b). 2.2.2. Fuel Two types of dead fuels (fine fuels and coarse fuels) as well as live fuels were simulated in the model. Fine fuels, which correspond to 1- and 10-h lag fuels and litters (He et al., 2004), are leaves, twigs, ground litter, needles and fine woody debris which fall from trees annually (Deeming et al., 1977). Coarse fuels, also called coarse woody debris (CWD), include snags, logs, large pieces of wood (which result from the disintegration of larger snags and logs), branches, stems and coarse roots. Coarse fuels correspond to 100- and 1000-h fuels (Burgan, 1987). Live fuels consist of leaves, twigs, and stems of growing plants. Most live fuels are difficult to ignite and often do not burn readily by themselves. However, some conifer species provide vertical continuity between canopy layers and allow fire to move from surface fuels or understory vegetation to tree crowns. Detailed description of the LANDIS fuel module design and application can be found elsewhere (Shang and He, 2003; He et al., 2004; Shang et al., 2004). The following paragraphs gave a brief introduction about tracking of fuel loads in LANDIS model. Fine fuel loads are approximated in LANDIS by vegetation types (species composition) and stand age (He et al., 2004). In general, mature, old trees produce more needles, leaves, and dead twigs than small, young trees. A curve was defined for each species to approximate how fine fuel loadings varied with species age (Fig. 2A). Decomposition rates of fine fuels vary by sites. A study by Trofymow et al. (2002) on 18 upland forest sites across Canada suggests that about 80% of the original litter mass decomposed within their 6-year experiment. Mudrick et al. (1994) reported that leaf litter decomposes even faster in an Appalachian forest in West Virginia within the hardwood region. Chestnut oak leaf litter loses about 25% of its dry weight in the first year. Red maple losses about 65% dry weight in the first year. Generally, most of the leaf litter in a hardwood forest decomposes in a few years. Therefore, the LANDIS fuel module (He et al., 2004) assumes that most fine fuels decompose in less than 10 years, which is shorter than the LANDIS 10-year time step. At each time step, fine fuels are recalculated based on the live species/age cohorts. Fine fuel from different species may have different flammability due to differences in physical and chemical attributes. A fuel quality coefficient (0 < FQC  1) was assigned for each species to summarize such differences on a relative scale. Fine fuel loads for each species were weighted differently by FQC and aggregated together to calculate the total fine fuel loads for each site. He et al. (2004) addressed further details of how fine fuel loads are estimated in LANDIS. Unlike the fine fuels, coarse fuels are not derived using species-specific age cohorts. Instead, stand age (the oldest age Fig. 2. Modeling fine fuel and coarse fuel accumulation and decomposition in LANDIS: (A) fine fuel accumulation changed with species age; (B) coarse fuels accumulate over time in two different land types; (C) extra coarse fuels decompose after disturbance in two different land types. cohorts in the stand) in combination with disturbance history (e.g., time since last disturbance) are used to determine the coarse fuel loads (He et al., 2004). In the absence of disturbance, the accumulation process dominates until the amount of coarse fuel reaches a level where decomposition and accumulation are in balance (Fig. 2B). When disturbance happens (e.g., windthrow, insect defoliation, or harvest), a large amount of extra coarse fuels may be added into the coarse fuel pool. Those extra coarse fuels (caused by disturbance) then start to decompose in the next few decades. In general, coarse fuels decompose more rapidly on mesic land types than on xeric land types. This decomposition process is modeled based on the decomposition curve (Fig. 2C). The accumulation and decomposition processes together determine the coarse fuel loads. Details about the parameterization of accumulation and decomposition curves for eight ecological land types in the same study area were described in He et al. (2004). 2.2.3. Fire LANDIS 4.0 retains the statistical approach used in earlier versions of LANDIS to simulate fire behavior, whereas a more realistic fire spread model based on the fire growth simulation model FARSITE (Finney, 1998) has also been added (Yang 780 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 et al., 2004). The statistical fire simulation approach simulates fire ignition/occurrence, size, spreading, intensity and severity. Following Baker (1992) and Johnson and Van Wagner (1985), LANDIS model uses the mean fire-return interval and time since last fire to calculate the probability of fire occurrence, while fire size follows a log-normal distribution with small fires occurring more frequently than large fires (He and Mladenoff, 1999a). LANDIS uses a stochastic approach to simulate fire spreading over the landscape (He and Mladenoff, 1999a). The combination of the fine and coarse fuel loads in each site was used to determine fire intensity if the site was burned by wildfire (Table 1) (He et al., 2004). In this study, we refer to this potential fire intensity as the fire hazard. Species fire resilience is determined by fire intensity, species fire tolerance and agespecific fire-susceptibility (He and Mladenoff, 1999a). Details about the LANDIS fire module have been published by He and Mladenoff (1999a) and Yang et al. (2004). 2.3. Data preparation and model parameterization 2.3.1. Ecological land types The 20 ecological land types (ELTs) that Miller (1981) defined for the study area were combined to create a reduced set of seven principal ecological land types used for simulation in LANDIS (Shifley et al., 2000a,b). These include: (1) dry chert forest on south and west slopes; (2) dry-mesic forest on north and east slopes; (3) ridge tops and upland flats; (4) upland drainages; (5) mesic floodplain or low terraces; (6) side slope on limestone bedrock; (7) savannas/glades; (8) non-federal lands (Fig. 1B, Table 2). 2.3.2. Species composition, age cohorts and attributes LANDIS simulated the presence and absence of species by 10-year age cohorts for each 30 m  30 m pixel. We summarized information about species presence and age class for each scenario to show the impacts of different fire regimes. Based on the oldest tree cohort of each species for each pixel, we grouped results into five size classes: seedling (age 0–10 years), sapling (age 11–30 years), pole (age 31–60 years), sawlog (age 61–100 years), and old (age > 100 years). Depending upon patterns of succession or disturbance for a given scenario, each pixel could contain none, any, or all of the four species groups in one or in multiple size (or age) classes. Species composition in the study area was represented by four principal overstory species groups in the study area (Fig. 3, Table 3) (Shifley et al., 2000a): the white oak group, predominantly white oak (Quercus alba L.) and post oak (Q. Table 2 Land types in the study area, which includes 712 km2 Mark Twain National Forests, and 576 km2 intermixed non-federal lands Name Area (km2) Dry chert forest on south and west slopes Dry-mesic forest on north and east slopes Ridge tops and upland flats Upland drainages Mesic floodplain or low terraces Side slope on limestone bedrock Savanna/glade 211 182 262 35 13 8 3 30 26 37 5 2 1 0 Total national forests in the study area 712 100 Proportion (%) stellata Wangenh.); the red oak group, predominantly black oak (Q. velutina Lam.) and scarlet oak (Q. coccinea Muenchh.); the pine group, predominantly shortleaf pine (Pinus echinata Mill.); the maple group, predominantly red maple (Acer rubrum L.) and sugar maple (A. saccharum Marsh.). Life history characteristics for each species group (e.g., shade tolerance, fire tolerance, sprouting probability, seed dispersal distance, and longevity) were determined from published guides (Burns and Honkala, 1990). Forest inventory data for each stand in the study area were coupled with regional information on tree species composition from forest inventory and analysis databases (FIA) (Hansen et al., 1992) and the nearby Missouri Ozark Forest Ecosystem Project (MOFEP) study sites (Brookshire and Shifley, 1997; Shifley and Brookshire, 2000; Shifley and Kabrick, 2002) and used to establish a representative species composition for each forest stand of known location and stand age class (Table 3). Species locations within a given stand (Fig. 3) were randomly assigned (Shifley et al., 2000a). Data on vegetation, land types and stand boundaries were lacking for non-federal lands adjacent to and/or intermixed with national forests, so we parameterized a generic oak species (or pseudo-oak species) for the forest cover of the non-federal lands. This procedure allowed simulated wildfires to ignite in the non-federal lands and spread to the National Forest and vice versa rather than arbitrarily stopping wildfires at ownership boundaries. The species attributes of this pseudo-oak species were a composite of those of the dominant oak species in Missouri Ozarks. We parameterized a second generic species across the entire landscape to represent herbs, grasses, and understory shrubs. This species had a short longevity (20 years) and a low level of coarse fuel accumulation. The life span of this generic species is relatively short comparing to the simulation time step in the LANDIS model (10 years). Therefore, LANDIS Table 1 Fire hazard index defined by fine and coarse fuel loads (He et al., 2004) Coarse Coarse Coarse Coarse Coarse fuel fuel fuel fuel fuel class class class class class 1 2 3 4 5 Fine fuel class 1 Fine fuel class 2 Fine fuel class 3 Fine fuel class 4 Fine fuel class 5 1 1 2 2 3 1 1 2 3 3 1 1 2 3 4 1 2 3 4 5 2 2 3 4 5 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 781 Fig. 3. Initialized species distribution and age cohorts in the simulation study. could not precisely simulate those herbs, grasses and understory shrubs. However, those understory species are important as they allow surface fires to spread across non-forested areas such as grasslands. Therefore, we still simulated this generic species in our study so that surface fires can spread realistically across non-forested areas. Although our simulation runs included fires that moved to and from private lands, we restricted our analyses to National Forest lands where we had current, detailed information on initial forest conditions. 2.3.3. Fuel loads and fires A few studies describe historic or current fuel loadings in southeastern Missouri, and most studies were completed before the development of modern fuel sampling techniques and timelag classes (Scowcroft, 1965; Brown, 1970; Crosby and Loomis, 1974; Loomis, 1975). An inventory in the 1970s in southeastern Missouri (Brown, 1974) showed that total weight of forest floor fuels in 20-year-old black oak stands averaged 1.6 kg/m2, 0.5 kg/m2 of which were loose, fresh litter (mostly leaves; equivalent to fine fuel as we define it). Forty-year-old stands averaged 2 kg/m2 of forest floor fuel, including 0.7 kg/ m2 of litter. Another inventory in southeastern Missouri (Loomis, 1975) showed the annual changes in forest floor weights under an oak stand, with a minimum of about 1.3 kg/m2 present on September, before leaf-fall, and a maximum loading of 1.9 kg/m2 occurred in November after leaf-fall. A recent Table 3 Initialized species composition for the national forests within the study area Species group Species components Area (km2) Age cohorts (proportion) White oak group White oak (Quercus alba L.) Post oak (Q. stellata Wangenh.) 279 10–20 years: 19% 30–40 years: 31%; 60–70 years: 51% Red oak group Black oak (Q. velutina Lam.) Scarlet oak (Q. coccinea Muenchh.) 270 10–20 years: 19% 30–40 years: 31%; 60–70 years: 51% Pine group Shortleaf pine (Pinus echinata Mill.) 106 10–20 years: 19%; 30–40 years: 34% 60–70 years: 47% Maple group Red maple (Acer rubrum L.) Sugar maple (A. saccharum Marsh.) 46 10–20 years: 17% 30–40 years: 27%; 50–60 years: 56% Species proportions and age cohorts were estimated from forest inventory and analysis databases (FIA) (Hansen et al., 1992) and the nearby Missouri Ozark Forest Ecosystem Project (MOFEP) (Brookshire and Shifley, 1997; Shifley and Brookshire, 2000; Shifley and Kabrick, 2002). Species locations were randomly assigned within each stand based on the species proportions by ecological land type (Shifley et al., 2000a,b). 782 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 field study in southeastern Missouri (Kolaks, 2004; Kolaks et al., 2004) reported an average 1.7–2.2 kg/m2 of fuel, with 0.7 kg/m2 of litter, 0.1 kg/m2 of 1-h fuels, 0.1 kg/m2 of 10-h fuels, 0.2–0.4 kg/m2 of 100-h fuels, and 0.5–0.9 kg/m2 of 1000h (solid and rotten) fuels. We derived fuel loads classes in our simulation based on those inventory data (Tables 4 and 5). Wildfires in hardwood ecosystems often have lower intensity than in ecosystems with a substantial conifer component, because fuels are less flammable. In the Missouri Ozarks, typical surface fires burn the upper litter layer and small branches that lie on or near the ground. Crown fires rarely occur unless the stand structure, weather, and ladder fuels allow surface or ground fires to ignite tree crowns and spread to other crowns. Depending on the fire intensity, fuel condition and site condition, wildfires can cause different fuel consumptions. In general, wildfires, even medium-low and medium intensity fires, can readily burn most fine fuels. However, coarse fuels are often wet and difficult to burn. High intensity fires consume more coarse fuels than lower intensity fires but very high intensity surface fires can kill many trees (e.g., Regelbrugge and Smith, 1994) and add snags to the pool of dead coarse fuels. Details about fire intensity and fuel consumption can be found from Shang et al. (2004). Prescribed fires, depending on the fuel condition, weather, wind and slope, may have different effects on the surface fuel consumption and fire hazard reduction. A recent field study on the oak–hickory and oak–pine forests in southeastern Missouri (Kolaks, 2004; Kolaks et al., 2004) showed that prescribed burn-only treatment consumed nearly all the litter, 27–60% 1-h fuels, 16–45% 10-h fuels, 16–35% 100-h fuels, 12–37% solid 1000-h fuels, and 9–60% rotten 1000-h fuels. However, due to high variability, the consumption of 100- and 1000-h fuels were not statistically significant. To simplify the simulation study in Missouri Ozarks, LANDIS assumes prescribed fires burn most of the fine fuels on the ground, and reduce fine fuel loads to the low class. Prescribed fires have limited effects on the coarse fuels and only reduce coarse fuel loads by one category. For example, prescribed fires burn coarse fuels and reduce fuel loads from high level (class 5) to medium-high level (class 4). A simulation study aimed at fuel and fire management has been published by Shang et al. (2004). 2.4. Simulation scenarios and results verification We selected two scenarios for a 200-year simulation study. The first scenario approximated a historic fire regime (1680– 1940) before effective fire suppression. The second scenario approximated a fire suppression regime circa 1990s. We used Table 4 Definition of fine fuel loads by classes Litter, 1- and 10-h fuels (kg/m2) Class Class Class Class Class 1 2 3 4 5 (low) (medium-low) (medium) (medium-high) (high) 0.5 0.5–1 1–1.5 1.5–2 >2 Table 5 Definition of coarse fuel loads by classes 100- and 1000-h fuels (kg/m2) Class Class Class Class Class 1 2 3 4 5 (low) (medium-low) (medium) (medium-high) (high) 0.5 0.5–0.75 0.75–1 1–1.25 >1.25 dendrochronological studies spanning 1680–1940 in southeastern Missouri (Guyette and Dey, 1995; Guyette and Larsen, 2000; Guyette et al., 2002) to parameterize the historic fire regime scenario. We used fire statistics from 1970 to 1989 in Missouri (Westin, 1992) to parameterize the fire suppression scenario (Table 6). Historically, the species composition and age structure of Ozark forests shifted as disturbance regimes changed over time. For our simulations we used the same initial species composition and age cohorts (Fig. 3, Table 3) for both simulation scenarios so that we could readily compare simulated fire effects on fuel accumulation and species abundance and age structure for these two scenarios over time. The management policies have been varied substantially in recent centuries, e.g., heavy logging between 1890 and 1920 (Shifley et al., 2000b), fire suppression beginning in the 1940s, and prescribed burning after 1960s (Cleaves et al., 2000). However, this study did not include any timber harvest or prescribed burning in the simulation, so that any human effects other than effective fire suppression could be eliminated. This simplification process helps us to study only the contrasting effects of the two fire regimes. Of course, without a simulation scenario of realistic forest management activities (e.g., timber harvest and prescribed burning) this study could not be used as a prediction of future patterns. The LANDIS fire module includes numerous stochastic processes, including fire ignition and spread and the interaction of fire with vegetation. To ensure both scenarios are correctly simulated, we used historic fire data (Guyette et al., 2002) and fire suppression data (Westin, 1992) to calibrate the model iteratively until both scenarios matched the input data. After the 1940s, wildfires have been well documented in Missouri (Westin, 1992). We used recent (from 1970 to 1989) fire records (Westin, 1992) to verify our simulated mean fire-return interval (MFI), fire size and number of fires under the fire suppression. The comparison showed that our simulated fire suppression scenario corresponded well with the fire records (from 1970 to 1989). However, historic fire records before 1939 are very limited. Dendrochronological studies (from 1680 to 1940) (Guyette et al., 2002) recover the fire regime in the southeastern Missouri, which provide information that can be used to verify the simulated mean fire-return interval and number of fires under the historic fire regime scenario. As the fire suppression has just been initiated, but was not yet effective during 1939– 1945, we used fire records from that period to verify the simulated fire sizes under the historic fire regime. Verifications show the historic fire regime scenario matched the best available historic data (1939–1945) very well. 783 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 Table 6 Parameterization for two different scenarios: historic fire regime scenario and fire suppression scenario Simulation scenario Features Initialized mean fire-return interval (MFI) Initialized time since last fire Initialized mean fire size (ha) Historic fire regime scenarioa Forests with no wildfire suppression 10 years on savanna/glades 10 years on savanna/glades 36 14 years on other land types 0–30 years on other land types 20 years on savanna/glades 20 years on savanna/glades 480 years on other land types 70–80 years on other land types Fire suppression scenario b Forests with effective wildfire suppression 4.4 a The historic fire regime scenario approximated the period from 1680 to 1940, before effective wildfire suppression. During this period (1680–1940), the mean firereturn interval (MFI) typically varied between 2 and 19 years (Guyette and Dey, 1995; Pyne, 1982; Williams, 1989; Guyette et al., 2002). Therefore, the mean firereturn interval (MFI) was initialized at 10 years on savanna sites and 14 years on other land types. The time since last fire was set between 0 and 30 years for the landscape; the corresponding fire sizes were initialized based on the fire records (1939–1945) in Missouri (Westin, 1992). b The fire suppression scenario approximated a fire regime under fire suppression as practiced in the 1990s (Shifley et al., 2000a,b). Based on historic fire data between 1970 and 1989 (Westin, 1992), the MFI was initialized at 480 years for most land types, and 20 years on managed savanna/glades. Because wildfires burned frequently in the Euro-American settlement period (1821–1940) (FRI  4 years) (Guyette and Larsen, 2000), and infrequently in the fire suppression period (Westin, 1992), the time since last fire was initialized at 20 years for savannas and 70 or 80 years for other ecological land types. We initialized the corresponding initial fire sizes based on fire records (1970–1989) in Missouri (Westin, 1992). We used historic studies (Guyette and Dey, 1997; Batek et al., 1999) and forest inventory data for the Mark Twain National Forests and the Missouri Ozark Forest Ecosystem Project (MOFEP) study (Brookshire and Shifley, 1997; Shifley and Brookshire, 2000) to validate our simulation results on species composition and age structure under the two model scenarios. Comparisons indicate that our simulated species composition under the historic fire regime scenario is close to the historic reconstructions of historic tree species composition (circa 1818–1840) based on witness trees (Batek et al., 1999), which both indicated the dominance of pine (mostly shortleaf) and oak–pine forests. Our simulated species composition under the fire suppression scenario corresponded well with recent forest inventory data, which both showed the dominance of mixed-oak forests. 3. Results 3.1. Tree species composition and age structure Simulation results show distinct differences in species coverage between the two scenarios (Fig. 4). Percent coverage of white oaks was higher under the fire suppression scenario (Fig. 4B) than historic fire scenario (Fig. 4A). Trees in both the red oak (Fig. 4C and D) and maple groups (Fig. 4G and H) increased in the number of sites they occupied under the fire suppression scenario and decreased under the historic fire scenario. The pine group increased in percent cover under both scenarios, but more rapidly under the historic regime (Fig. 4E and F). Tree size class structure also differed between the two scenarios. For example, though percent coverage of old growth white oak trees did not differ distinctly between the two scenarios, more sawlog, pole and sapling size white oak forests existed under the fire suppression scenario than the historic fire scenario. All size classes of red oak and maple decreased in percent coverage under the historic fire regime scenario and increased under the fire suppression scenario. The cover of pine trees in all size classes increased rapidly under the historic fire regime scenario; all size classes of pine also increased, but much more gradually, under the fire suppression scenario. 3.2. Fuel loads and fire hazards Results showed differences in fine fuel loading between the two simulation scenarios (Fig. 5). Frequent fires under the historic fire regime scenario reduced the fine fuel loads to a low level along the 200-year simulation period. By the end of the 200-year simulation, about 10% of the study area had a low level of fine fuel accumulation, 60% had low or medium-low fine fuel loads, and another 30% had medium or medium-high fine fuel loads. Under the fire suppression scenario, fine fuels were at higher levels. At simulation year 200, on about 20% of landscape the fine fuel loads were low or medium-low, more than 30% of landscape had medium fine fuel loads, and more than 40% of landscape had medium-high fine fuel loads. The coarse fuel loads differed between these two scenarios (Fig. 6). In the historic fire regime, coarse fuel loads were generally at low levels over the 200-year simulation. At simulation year 200, about 20–30% of the landscape carrying a medium-low coarse fuel load, and another 70–80% of the landscape with a medium coarse fuel load. However, in the fire suppression scenario, there was a gradual accumulation of coarse fuel over the 200-year simulation. By the end of the 200year simulation, more than 40% of the landscape had mediumhigh coarse fuel loads, and about 30% of the landscape had high coarse fuel loads. Simulated coarse fuel loads under the fire suppression scenario were much higher than those under the historic fire regime scenario. Under the historic fire regime scenario, fire hazard was simulated at the medium-low level across most of the study area (Fig. 7). Only a few sites had a medium-high level fire hazard, and these sites were scattered far away from each other. Under the fire suppression scenario, within 50 years, a large number of sites had a medium level of fire hazard and a large number of sites had a medium-high level of fire hazard across the 784 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 Fig. 4. Simulated landscape coverage area of each species by age class under the two scenarios. Each 0.09 ha pixel could contain up to four species groups. (A) White oak group under historic fire regime scenario; (B) white oak group under fire suppression scenario; (C) red oak group under historic fire regime scenario; (D) red oak group under fire suppression scenario; (E) pine group under historic fire regime scenario; (F) pine group under fire suppression scenario; (G) maple group under historic fire regime scenario; (H) maple group under fire suppression scenario. landscape. In 100 years, about 60% of the landscape had a medium-high level fire hazard, and these sites with mediumhigh level fire hazard were contiguous. In 150 years, most of the landscape had a medium-high level fire hazard. In general, long-term fire suppression led to a medium-high level of fire hazard and increased the probability of high intensity fires. Under the simulated historic fire regime scenario (MFI = 14 years), about 98% of the study area was burned 10 times or more over the 200-year simulation period. In contrast, under the fire suppression scenario MFI was extended to over 480 years. Approximately 60% of the landscape never burned, 35% of the landscape burned only once, and only 5% burned two or more times within the 200-year simulation period. Fire sizes differed distinctly between these two scenarios (Fig. 8). As expected from our parameterization, the historic fire scenario simulated more and larger fires than the fire suppression scenario. Fires tended to be more frequent and burn larger area in the historic fire regime, while both fire number Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 785 Fig. 5. Comparison of fine fuel loads between the two scenarios. Fine fuel loads were classified into five categories: low, medium-low, medium, medium-high and high. Fig. 6. Comparison of coarse fuel loads between the two scenarios. Coarse fuel loads were classified into five categories: low, medium-low, medium, medium-high and high. 786 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 Fig. 7. Simulated fire hazard under the two scenarios. Fire hazard was estimated based on fine and coarse fuel loads. Fire hazard was classified into five categories: low, medium-low, medium, medium-high and high. and size of area burned declined in the fire suppression scenario. Simulation results differed between scenarios in the percentage of the area burned and fire intensity (Fig. 8). Under the historic fire regime scenario (Fig. 9A), about 70% of the study area was burned each decade, mostly by fires of mediumlow or medium intensity (classes 2 and 3). About 1% of forest were burned by medium-high intensity surface fires (class 4), and high intensity surface fires (class 5) rarely occurred in the 200-year simulation. Under the fire suppression scenario (Fig. 9B), about 2% of forests burned in each decade. Fires were less frequent, but more intense, with most fires at medium and medium-high intensity. Fig. 8. Comparison on number of fires by size class between the two scenarios. 4. Discussion 4.1. Modeling approach Physical fire-behavior models are one approach to simulate the real-time behavior of fires at the scale of the forest stand, e.g., BEHAVE (Andrews, 1986), FOFEM (Reinhardt et al., 1997), FIRE-BGC (Keane et al., 1995), FFE-FVS (Reinhardt and Crookston, 2003). Physical fire models have been widely applied to assess the effects of multiple fuel treatments on fuels, wildfire behaviors, and forest ecosystems (e.g., Van Wagtendonk, 1996; Brose and Wade, 2002). However, the use of highly detailed, fine-grained physical models for predicting fire effects at landscape scales for a long-term simulation is a ‘‘daunting task’’, because many variables (e.g., fuel moisture, packing and geometry of fuels, wind speed and direction, and humidity), are either unknown or un-measurable at the landscape scale; also, their long-term dynamics are difficult to predict (Gardner et al., 1999). Therefore, BEHAVE and FFE-FVS have often been applied at smaller spatial scales (e.g., a forest stand) and simulate fuel and fire behavior in a more mechanistic and deterministic way. Statistical fire models use the statistical description of fire regime (most often the historic fire regime) to estimate the probability density functions for characterizing and/or simulating fire regimes and forest dynamics, e.g., average fire interval, fire frequency, average age of the vegetation and the renewal rate of disturbed forests (Gardner et al., 1999). Studies have Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 787 step, LANDIS cannot precisely simulate ecological processes at shorter time scale. We hope to solve these problems in current and future model development, e.g., a new human effect module, a new climate module, and multiple time steps (1, 2, 5 or 10 years). 4.2. Fuel accumulation Fig. 9. Simulated area burned by fires and fire intensity under the two scenarios. Fire intensity was classified into five categories: low, medium-low, medium, medium-high and high. (A) Historic fire regime scenario; (B) fire suppression scenario. demonstrated that Weibull distribution (Johnson, 1979; Johnson and Van Wagner, 1985) is the most suitable form for representing the fire frequency, and the negative exponential distribution (Van Wagner, 1978) is suitable for the forest age component. Although the statistical approach developed by Van Wagner (1978) and Johnson (1979) lacks a spatial component, the method has been adapted and applied at landscape scales by modifications allowing the estimation of the effect of spatial patterns of disturbance, e.g., DISPATCH (Baker, 1992) and LANDIS (Mladenoff et al., 1996; Mladenoff and He, 1999; He and Mladenoff, 1999a). Statistical models have been proved to be useful in simulating wildfire events over long temporal scales and large landscape scale (Baker, 1992; He and Mladenoff, 1999a). However, successful application of statistical fire models in predicting fire behaviors is limited to a ‘‘test of hypothesis’’ that the statistical description of the past fire history is still valid in the future fire regime. Therefore, statistical models should not be expected to be precise predictions of the future states of the landscape (Baker, 1992). LANDIS does not independently predict the occurrence of future disturbance or management events. Rather, it is a scenario model that compares long-term, large-scale, simulated effects of user-defined disturbance and management scenarios on real or hypothetical landscapes. However, LANDIS also inherited the weakness of statistical fire models, e.g., LANDIS cannot be used to predict real-time fire behavior. LANDIS can only be used with the hypothesis that the past fire regime is still effective in future simulation. Because of its 10-year long time Currently many forest stands in the study area can be characterized as well-stocked and maturing, with most trees in sawlog and pole sizes, a medium level fine fuel load, and a low or medium-low medium level of coarse fuel accumulation. Our simulation results suggest that in about 30 years many of these stands will succeed into old forest stages, and under effective fire suppression fine fuel loads will rapidly increase to a medium or medium-high level due to more old growth and mature trees in the forest. Within the next 80 years, both fine fuel and coarse fuels will increase to a medium-high level, increasing the risk of high intensity fires. Widespread fuel management efforts (e.g., prescribed burning) that reduce fine fuels on the ground, reduce tree density, and maintain younger age classes, could become necessary in order to reduce fine fuels. For some stands, thinning, piling and burning the coarse woody debris may be required because prescribed burning can not significantly reduce the 100- and 1000-h fuels (Kolaks, 2004; Kolaks et al., 2004). Based on a survey conducted during 1985 and 1994 (Cleaves et al., 2000), about 5% of the forests in the national forests in Missouri Ozarks were scheduled for prescribed burning for the purpose of fuel reduction in this 10-year period. Our simulation results show more than 20% of the forests will have a mediumhigh level fine fuel load in the next 20 years, suggesting that the current extent of prescribed burning is not sufficient for controlling the accumulation of fuels. Our simulations, though, do not account for the fuel-reducing effects of the current prescribed burning practices, and it is likely that real fine and coarse fuel loads are lower than simulated fuel loads. Simple arithmetic, however, indicates even if our simulations included the effects of prescribed burning, at least 10% of the landscape with medium fine fuel loads escapes prescribed burning, and would reach medium-high fine fuel loads within the next 20 years. Control of these fuels would require a doubling of current prescribed burning activities. 4.3. Fire hazards Our simulation indicates that fire hazard would be significantly greater under long-term fire suppression than under the historic fire regime. The mean fire hazard was generally controlled at medium-low and medium level under the historic fire regime but reached medium and medium-high levels within a few decades under the fire suppression scenario. Given that the national forest in our study area is surrounded by (and in many areas intermixed with) private lands, an elevated fire hazard index suggests that intense, and potentially difficult to control fires could occur, and could pose significant risk to forests, property, and human lives in the wildland–urban 788 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 interface. Our simulation suggests that extensive management (e.g., prescribed burning and/or thinning) to reduce fire hazard may become necessary in the Missouri Ozarks if fire suppression continues into the 21st century. sizes and mean fire-return interval are simulated by the parameterization scheme: the total number of fires will not increase even if there are more fine fuel loads in the forests, nor can fire sizes become larger than the parameterization scheme allows, even if there is abundant fuel to burn on the landscape. 4.4. Fires 4.5. Species composition High intensity fires were anticipated to occur more frequently, and more areas would be burned by high intensity fires after a long-term of fire suppression in the Ozarks. However, our simulation results provided complex information (Fig. 9). By comparing the fire intensity and burned area between those two scenarios, we were very surprised to see that the total area that was burned by medium-high intensity fires were quite similar between those two different scenarios, with about 1–2% of the landscape burned by medium-high intensity fire in both scenarios. The fire suppression scenario was setup with a 480-year initialized mean fire-return interval on the whole landscape except savanna/glades, so the area burned could only be around 2% per decade. However, under the historic fire regime scenario a large area (about 60% of the landscape) was burned by medium-low or medium intensity fires, while less than 2% of the landscape was burned by medium-low or medium intensity fires. This suggests that the fire suppression scenario in our simulation effectively suppressed the low and medium intensity fires, but not the medium-high intensity fires. Even though the total burned area decreased to a very low level under the fire suppression scenario, the total area burned by medium-high intensity fires did not change extensively. Fire suppression did not lead to an increase in the area burned by high or medium-high intensity fires. Our simulation results are thus quite different from forests in the western United States National Forests, where catastrophic wildfires have become more frequent and burned larger areas after a long period of fire suppression (Government Accounting and Office, 1999). We suspect the difference in forests (e.g., species composition and tree density) and environment (e.g., humidity, precipitation, and soil) cause the differences in fuel accumulation and fire regime. For example, humidity in Ozarks forest is generally higher than western National Forests and therefore fuel tends to decompose faster in our study area. Also, living trees and surface fuels in mixed-oak forests are generally humid and inflammable, while trees and surface fuels in western National Forests may become dry and very flammable during the dry season and cause high risk of catastrophic fires. However, without verification by a long-term experimental study in the Ozarks, we could not exclude the possible reason that those differences were merely caused by the uncertainty in model parameterization. We expected that the accumulation of fuels caused by longterm of fire suppression would lead to larger, more intense fires. However, in our simulation, while fire intensity did increase as the number of years of fire suppression increased, the total number of fires did not increase, nor did the fire sizes or burned area (Fig. 9). This may result from the limitations of the LANDIS model. As a scenario model, LANDIS does not predict fire or other disturbance events. Rather, fire ignitions, The simulation study indicates that different fire regimes significantly affect forest composition. Under a historic fire regime characterized by frequent low-intensity fires and a short fire-return interval (14 years), pine and white oak forests dominate the study area. After a 200-year simulation, pines occurred on 70% of the area, while white oak group occurred on 60% of the study area. Red oak group only occurred on 10% of the study area. Pine and oak–pine forests were the major forest component under a historic fire regime. This result is close to the historic reconstructions of the tree species composition in the early 19th century (Batek et al., 1999), which showed pine forests (mostly shortleaf pines) and oak–pine forests covered more than 50% of the upper Current River watershed in the southeast Missouri, while red oaks (black and red oak) covered less then 10% of the area. Wuenscher and Valiunas (1967) reported a high abundance of white oak and a low abundance of black oak in the Mississippi River Hills Region in Missouri before Euro-American settlement: 32% of the forests were white oak, 11% were black oak, and 9% were sugar maple. Under the fire suppression scenario, white and red oak groups were dominant and pines were less abundant than under the historic fire regime. In general, the oak–pine forests of the historic fire regime were replaced by mixed oak forests. Due to heavy logging between 1890 and 1920 (Shifley et al., 2000b) and long-term of fire suppression, current coverage of the pine forest is much lower than historic levels in the mid-19th century (Batek et al., 1999). Therefore, more frequent fires in the historic fire regime significantly promoted the establishment of pine groups (Fig. 4E). Coverage of the pine group rapidly increased from under 20% to over 70% in the 200-year simulation, but remained lower than its historic coverage (ca. 1818–1840; Batek et al., 1999). In contrast, the fire suppression scenario showed much less pine establishment in mixed-oak forests. Because the current coverage of pine is quite low, it also increased under the fire suppression scenario, but to less than 40% coverage at the end of the 200-year simulation. Under the assumption that effective fire suppression will persist in the 21st century, red oak forests are anticipated to increase their abundance in the Missouri Ozarks. Red oak forests provide valuable wood products for commercial uses, and good crops of acorns that provide wildlife with foods. More red oak forests, however, may lead to more frequent oak decline events, which are more frequent and severe among red (Quercus rubra), scarlet (Q. coccinea), and black oak (Q. velutina) (Houston, 1973). Oak decline causes large areas of tree death and significant fuel accumulation. Environment stresses, such as drought, frost injury, or insect defoliation may weaken red oaks and increase the chance of insect invasion (Robert et al., 2002). Many red oak forests in southern Missouri Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 were already stressed to their limit before the late 1990s, when 2–3 years of severe drought hastened their decline by making them more vulnerable to disease-causing fungi, such as Armillaria, and wood-boring insects, such as the red oak borer. Reducing red oak abundance would reduce oak decline in the Missouri Ozarks forests. Greater use of prescribed burning, accompanied by commercial thinning of red oaks may significantly decrease the abundance of red oaks in the forest, and thus reduce the outbreak of oak decline events. Acknowledgments This work was funded by North Central Research Station of the U.S. Forest Service. We like to thank three anonymous reviewers for their helpful comments to this manuscript. The most recent version of LANDIS can be downloaded at http:// www.missouri.edu\landis. References Abrams, M.D., 2003. Where has all the white oak gone? BioScience 53, 927– 939. Agee, J., Huff, M., 1987. Fuel succession in a western hemlock/Douglas-fir forest. Can. J. For. Res. 17, 697–704. Anderson, H.E., 1982. Aids to determining fuel models for estimating fire behavior. USDA Forest Service, General Technical Report INT0122, Intermountain Forest and Range Experiment Station, Ogden, UT, p. 22. Andrews, P.L., 1986. BEHAVE: Fire Behavior Prediction and Fuel Modeling System BURN System. Part 1. Gen Tech. Rep. INT-194. U.S. Department of Agriculture, Forest Service, Intermountain Research Station, Ogden, UT, 130p. Backer, D.M., Jensen, S.E., McPherson, G.R., 2004. Impacts of fire suppression activities on natural communities. Conserv. Biol. 18, 937–946. Baker, W.L., 1992. Effects of settlement and fire suppression on landscape structure. Ecology 73, 1879–1887. Barrett, J.J., Shlisky, A.J., Barrett, R.H., Heald, R.C., Allen-Diaz, B.H., 2001. The effects of forest management on plant species diversity in a Sierran conifer forest. Forest Ecol. Manage. 146, 211–222. Batek, M.J., Rebertus, A.J., Schroeder, W.A., Haithcoat, T.L., Compas, E., Guyette, R.P., 1999. Reconstruction of early nineteenth-century vegetation and fire regimes in the Missouri Ozarks. J. Biogeogr. 26, 397–412. Brookshire, B.L., Shifley, S.R., 1997. In: Proceedings of the Missouri Ozark Forest Ecosystem Project Symposium: An Experimental Approach to Landscape Research. Gen. Tech. Rep. NC-193. U.S. Department of Agriculture, Forest Service, North Central Research Station, St. Paul, MN, pp. 1–25. Brose, P., Wade, D., 2002. Potential fire behavior in pine flatwood forests following three different fuel reduction techniques. Forest Ecol. Manage. 163, 71–84. Brown, J.K., 1970. Ratios of surface area to volume for common fine fuels. Forest Sci. 16 (1), 101–105. Brown, J.K., 1974. Some Forest Floor Fuelbed Characteristics of Black Oak Stands in Southeast Missouri. USDA Forest Service, North Central Forest Experiment Station (Research Note. NC-162) 4p. Brown, J.K., Bevins, C.D., 1986. Surface Fuel Loadings and Predicted Fire Behavior for Vegetation Types in the Northern Rocky Mountains. Research Note INT-358. U.S. Forest Service, Missoula, Montana. Burgan, R.E., 1987. Concepts and Interpreted Examples in Advanced Fuel Modeling. Gen. Tech. Re INT-238. US Department of Agriculture, Forest Service, Intermountain Research Station, Ogden, UT, 40 pp. Burns, R.M., Honkala, B.H., 1990. Silvics of North America: 1. Conifers; 2. Hardwoods Agriculture Handbook 654, vol. 2. U.S. Dept. of Agriculture, Forest Service, Washington, DC, 877p. 789 Clark, J.S., 1990. Fire and climate during the last 750 yr in northwestern Minnesota. Ecol. Monogr. 60, 135–159. Cleaves, D.A., Martinez, J., Haines, T.K., 2000. Influences on prescribed burning activity and costs in the National Forest System. USDA Forest Service, Southern Research Station, General Technical Report, GTRSRS-037. Conard, S.G., Hartzell, T., Hibruner, M.W., Zimmerman, G.T., 2001. Changing fuel management strategies—the challenge of meeting new information and analysis needs. Int. J. Wildland Fire 10, 267–275. Crosby, J.S., Loomis, R.M., 1974. Some Forest Floor Fuelbed Characteristics of Black Oak Stands in southern Missouri. USDA Forest Service, North Central Forest Experiment Station (RN-NC-162) 4p. Deeming, J.E., Burgan, R.E., Cohen, J.D., 1977. The national fire-danger rating system. U.S. For. Serv., Gen. Tech. Rep. INT-39. Dellasala, D.A., Williams, J.E., Williams, C.D., Franklin, J.E., 2004. Beyond smoke and mirrors: a synthesis of fire policy and science. Conserv. Biol. 18, 976–986. Dombeck, M.P., Williams, J.E., Wood, C.A., 2004. Wildfire policy and public lands: integrating scientific understanding with social concerns across landscapes. Conserv. Biol. 18, 883–889. Finney, M.A., 1998. FARSITE: Fire Area Simulator—Model Development and Evaluation. Res. Pap. RMRS-RP-4. U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station, Ogden, UT, 52p. Gardner, R.H., Romme, W.H., Turner, M.G., 1999. Predicting forest fire effects at landscape scales. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University Press, Cambridge, pp. 163–185. Government Accounting Office, 1999. Western National Forests: A Cohesive Strategy is Needed to Address Catastrophic Wildfires. United States General Accounting Office Report GAO/RCED-99-65, Washington, DC, 60pp. Grabner, K.W., Hartman, G., Dwyer, J., 1999. Characterizing fuel loading and structure using ecological landscapes (ELT’s) in the Missouri Ozarks. In: Proceedings of the Joint Fire Science Conference and Workshop. University of Idaho, The Grove Hotel, Boise, ID, pp. 278–282. Gustafson, E.J., Shifley, S.R., Mladenoff, D.J., Nimerfro, K.K., He, H.S., 2000. Spatial simulation of forest succession and harvesting using LANDIS. Can. J. Forest Res. 30, 32–43. Guyette, R.P., Larsen, D.R., 2000. A history of anthropogenic and natural disturbances in the area of the Missouri Ozark forest ecosystem project. In: Shifley, S.R., Brookshire, B.L. (Eds.), Missouri Ozark Forest Ecosystem Project Site History, Soils, Landforms, Woody, Herbaceous Vegetation, Down Wood, Inventory Methods for the Landscape Experiment. Gen. Tech. Rep. NC-208. U.S. Department of Agriculture, Forest Service, North Central Research Station, St. Paul, MN, pp. 19–40. Guyette, R.P., Dey, D.C., 1995. A Dendrochronological Fire History of Opeongo Lookout in Algonquin Park, Ontario. Forest Research Report No. 134. Ontario Forest Research Institute, Sault Ste. Marie, Ont., 4pp. Guyette, R.P., Dey D.C., 1997. Historic shortleaf pine (Pinus echinata Mill.) abundance and fire frequency in a mixed oak-pine forest (MOFEP, site 8). In: Brookshire, B.L., Shifley, S.R. (Eds.), Proceedings of the Missouri Ozark Forest Ecosystem Project Symposium: An Experimental Approach to Landscape Research. Gen. Tech. Rep. NC-193. St. Paul, MN: U.S. Dept. of Agriculture, Forest Service, North Central Forest Experiment Station, pp. 136–149. Guyette, R.P., Muzika, R.M., Dey, D.C., 2002. Dynamics of an Anthropogenic fire regime. Ecosystems 5, 472–486. Haines, D.A., Main, W.A., Crosby, J.S., 1972. Forest Fires in Missouri. Research Paper NC-87. U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station, St. Paul, MN, 19pp. Hansen, M.H., Frieswyk, T., Glover, J.F., Kelly, J.F., 1992. The Eastwide Forest Inventory Data Base: Users Manual Gen. Tech. Rep. NC-151. U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station, St. Paul, MN, 48pp. Hargrove, W.W., Gardner, R.H., Turner, M.G., Romme, W.H., 2000. Simulating fire patterns in heterogeneous landscapes. Ecol. Modell. 135, 243–263. He, H.S., Mladenoff, D.J., 1999a. Spatially explicit and stochastic simulation of forest landscape fire disturbance and succession. Ecology 80, 81–99. 790 Z.B. Shang et al. / Forest Ecology and Management 242 (2007) 776–790 He, H.S., Mladenoff, D.J., 1999b. The effects of seed dispersal on the simulation of long-term forest landscape change. Ecosystems 2, 308–319. He, H.S., Li, W., Sturtevant, B.R., Yang, J., Shang, Z.B., Gustafson, E.J., Mladenoff, D.J., 2005. LANDIS—a spatially explicit model of forest landscape disturbance, management, and succession. LANDIS 4.0 User’s Guide. USDA Forest Service North Central Research Station General Technical Report NC-263. He, S.H., Shang, Z.B., Thomas, R.C., Gustafson, E.J., Shifley, S.R., 2004. Simulating forest fuel and fire risk dynamics across landscapes—LANDIS fuel module design. Ecol. Modell. 180, 135–151. Houston, D.R., 1973. Diebacks and declines: diseases initiated by stress, including defoliation. Int. Shade Tree Conf. Proc. 49, 73–76. Johnson, E.A., 1979. Fire recurrence in the subarctic and its implications for vegetation composition. Can. J. Bot. 57, 1374–1379. Johnson, E.A., Miyanishi, K., 1995. The need for consideration of fire behavior and effects in prescribed burning. Restorat. Ecol. 3, 271–278. Johnson, E.A., Miyanishi, K., Bridge, S.R.J., 2001. Wildfire regime in the boreal forest and the idea of suppression and fuel buildup. Conserv. Biol. 15, 1554–1557. Johnson, E.A., Van Wagner, C.E., 1985. The theory and use of two fire history models. Can. J. Forest Res. 15, 214–220. Keane, R.E., Morgan, P., Running, S.W., 1995. FIRE-BGC—a mechanistic ecological process model for simulating fire succession on Northern Rocky Mountain coniferous forest landscapes. U.S. Department of Agriculture, Forest Service, Res. Rep. INT-RP-434. Kolaks, J.J., Cutter, B.E., Loewenstein, E.F., Grabner, K.W., Hartman, G.W., Kabrick, J.M., 2004. The effect of thinning and prescribed fire on fuel loading in the central hardwood region of Missouri. In: Yaussy, D.A., Hix, D.M., Long, R.P., Goebel, P.C. (Eds.), Proceedings of the 14th Central Hardwood Forest Conference, Gen. Tech. Rep. NE-316, Wooster, OH, March 16–19. U.S. Department of Agriculture, Forest Service, Northeastern Research Station, Newtown Square, PA, pp. 168–178. Kolaks, J., 2004. Fuel loading and fire behavior in the Missouri Ozarks of the Central Hardwood Region. Master’s Thesis, University of Missouri-Columbia, 115 pp. Loomis, R.M., 1975. Annual changes in forest floor weights under a southeast Missouri oak stand. USDA Forest Service, North Central Forest Experiment Station (RP-NC-104) 6p. Miller, M.R., 1981. Ecological Land Classification Terrestrial Subsystem, a Basic Inventory System for Planning and Management on the Mark Twain National Forest. U.S. Department of Agriculture, Forest Service, Eastern Region, 87pp. Mladenoff, D.J., He, H.S., 1999. Design and behavior of LANDIS, an objectoriented model of forest landscape disturbance and succession. In: Mladenoff, D.J., Baker, W.L. (Eds.), Spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University Press, New York, pp. 125–162. Mladenoff, D.J., Host, G.E., Boeder, J., Crow, T.R., 1996. LANDIS: a spatial model of forest landscape disturbance, succession, and management. In: Goodchild, M.F., Steyaert, L.T., Parks, B.O. (Eds.), GIS and Environmental Modeling: Progress and Research Issues. GIS World Books, Fort Collins, CO, USA, pp. 175–180. Mudrick, D.A., Hoosein, M., Hicks Jr., R.R., et al., 1994. Decomposition of leaf litter in an Appalachian forest: effects of leaf species, aspect, slope position and time. Forest Ecol. Manage. 68, 231–250. Pyne, S.J., 1982. Fire in America: A Cultural History of Wildland and Rural Fire. Princeton University Press, Princeton, NJ, USA. Regelbrugge, J.C., Smith, D.W., 1994. Postfire tree mortality in relation to wildfire severity in mixed-oak forests in the Blue Ridge of Virginia. N. J. Appl. Forest. 11, 90–97. Reinhardt, E.D., Keane, R.W., Brown, J.K., 1997. First Order Fire Effects Model: FOFEM 4.0. User’s Guide. Gen. Tech. Rep. EIIT-GTR-344. U.S. Department of Agriculture, Forest Service, Intermountain Research Station, Ogden, UT, 65p. Reinhardt, E.D., Crookston, N.L. (Eds.), 2003. The Fire and Fuels Extension to the Forest Vegetation Simulator. Gen. Tech. Rep. RMRS-GTR-116. Dept. of Agriculture, Forest Service, Rocky Mountain Research Station, Ogden, UT, p. 220. Robert, L., Bruce, M., Keith, M., 2002. Oak decline and the future of Missouri’s forests. Missouri Conservat. 63, 7. Romme, W.H., Knight, D.H., 1981. Fire frequency and subalpine forest succession along a topographic gradient in Wyoming. Ecology 62, 319–326. Schuster, E.G., Cleaves, D.A., Bell, E.F., 1997. Analysis of USDA Forest Service Fire-related Expenditure 1970–1995. Pacific Southwest Research Station Research Paper PSW-RP-230. USDA Forest Service, Albany, CA, 29pp. Scowcroft, P.G., 1965. The effects of fire on the hardwood forest of the Missouri Ozarks. Masters Thesis, University of Missouri, 126p. Shang, B.Z., He, S.H., 2003. LANDIS Fuel Module—For Tracking and Simulating Fuel Loading, Fire Intensity, Fire Risk and Fuel Management. LANDIS Fuel Module User’s Guide. University of Missouri-Columbia, 46pp. Shang, B.Z., He, H.S., Crow, T.R., Shifley, S.R., 2004. The effects of various fuel load reductions on potential fire risk—a spatial simulation study. Ecol. Model. 180, 89–102. Shifley, S.R., Brookshire, B.L. (Eds.), 2000. Missouri Ozark Forest Ecosystem Project: Site History, Soils, Landforms, Woody and Herbaceous Vegetation, Down Wood, and Inventory Methods for the Landscape Experiment. Gen. Tech. Rep. NC-208. U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station, St. Paul, MN, p. 314. Shifley, S.R., Kabrick, J.M. S, 2002. In: Proceedings of the Second Missouri Ozark Forest Ecosystem Symposium: Post Treatment Results of the Landscape Experiment. General Technical Report NC-227. St. Louis, MO, October 17–20, 2000. U.S. Department of Agriculture, Forest Service, North Central Forest Experiment Station, St. Paul, MN, 228p. Shifley, S.R., Thompson III, F.R., Larsen, D.R., Mladenoff, D.J., Gustafson, E.J., 2000a. Utilizing inventory information to calibrate a landscape simulation model. In: Proceedings of Integrated Tools for Natural Resources Inventories in the 21st Century, Gen. Tech. Rep. NC-212, 1998. U.S. Department of Agriculture, Forest Service, North Central Research Station, St. Paul, MN. Shifley, S.R., Thompson III, F.R., Larsen, D.R., Dijak, W.D., 2000b. Modeling forest landscape change in the Missouri Ozarks under alternative management practices. Comput. Electron. Agric. 27, 7–24. Trofymow, J.A., Moore, T.R., Titus, B., Prescott, C., Morrison, I., Siltanen, M., Smith, S., Fyles, J., Wein, R., Camire, C., Duschene, L., Kozak, L., Kranabetter, M., Visser, S., 2002. Rates of litter decomposition over 6 years in Canadian forests: influence of litter quality and climate. Can. J. Forest Res. 32, 789–804. Van Wagner, C.E., 1978. Age-class distribution and the forest fire cycle. Can. J. Forest Res. 8, 220–227. Van Wagtendonk, 1996. Use of a deterministic fore growth model to test fuel treatments. In: Sierra Nevada Ecosystem Project: Final Report to Congress, vol. II. Assessments and Scientific Basis for Management Options, Centers for Water and Wildland Resources, University of California, Davis, pp. 1155–1165. Westin, S., 1992. Wildfire in Missouri. Missouri Department of Conservation, Jefferson City, MO, 161pp. Wuenscher, J.E., Valiunas, A.J., 1967. Presettlement forest composition of the River Hills region of Missouri. Am. Midland Nat. 78, 487–495. Yang, J., He, H.S., Gustafson, E.J., 2004. A hierarchical statistical approach to simulate the temporal patterns of forest fire disturbance in LANDIS model. Ecol. Modell. 180, 119–133.