Forest Ecology and Management 225 (2006) 82–93 www.elsevier.com/locate/foreco Simulating the effects of reforestation on a large catastrophic fire burned landscape in Northeastern China Xugao Wang a,c,*, Hong S. He b, Xiuzhen Li a, Yu Chang a, Yuanman Hu a, Chonggang Xu a,c, Rencang Bu a, Fuju Xie a,c a Institute of Applied Ecology, Chinese Academy of Science, P.O. Box 417, Shenyang 110016, China b School of Natural Resources, University of Missouri, Columbia c Graduate School of Chinese Academy of Science, Beijing 100039, China Received 20 May 2005; received in revised form 2 November 2005; accepted 16 December 2005 Abstract We use the LANDIS model to study the effects of planting intensity and spatial pattern of plantation on the abundance of three main species (larch (Larix gmelini), Mongolian Scotch pine (Pinus sylvestris var. Mongolica), and white birch (Betula platyphylla)) in the Tuqiang Forest Bureau on the northern slopes of Great Hing’an Mountains after a catastrophic fire in 1987. Four levels of planting intensity (covering 10%, 30%, 50%, and 70% of the severely burned area) and two spatial patterns of plantation (dispersed planting and aggregated planting) were compared in a 4
2 factorial design over a 300-year period. The results showed that increasing planting intensity positively influenced larch and Mongolian Scotch pine abundance, but negatively influenced white birch abundance. However, the increased degree of larch abundance with increasing planting intensity was significantly different between intensities. The difference in larch abundance between the 10% planting intensity scenario and the 30% planting intensity scenario was greater than that between the 50% planting intensity scenario and the 70% planting intensity scenario. However, the difference between 30% and 50% planting intensity scenarios was significantly low. Hence, given considerable labor input and economic costs, 30% planting intensity would be effective for forest recovery. In addition, dispersed planting showed more promising results on forest recovery than aggregated planting. However, the difference of larch abundance between dispersed planting and aggregated planting under intermediate planting intensity scenarios (30% and 50% planting intensity) was greater than that under a low planting intensity scenario and a high planting intensity scenario. Therefore, it is necessary to incorporate spatial pattern of plantation into planting practice, especially under an intermediate planting intensity scenario. These results have important implications for forest managers to design sound forest restoration projects for landscapes affected by large infrequent disturbances. In particular, the results suggest that the current planting strategy (50% planting intensity with aggregated planting) employed after the catastrophic fire in 1987 could not be optimum, and the dispersed planting strategy covering about 30% of the severely burned area would better stimulate forest recovery. # 2005 Elsevier B.V. All rights reserved. Keywords: Great Hing’an Mountains; LANDIS; Planting intensity; Spatial pattern of plantation; Forest recovery 1. Introduction The deciduous and coniferous forests of the Great Hing’an Mountains in northeastern China provide the most timber of any forested area in the country; simultaneously, this area encompasses rather unique ecological and environmental systems in the region (Zhou, 1991; Xu, 1998). Human activities, particularly timber harvesting, have substantially altered the spatial pattern and ecological functions of these * Corresponding author. Tel.: +86 24 83970350; fax: +86 24 83970351. E-mail addresses: wxg_7980@163.com, wangxugao@hotmail.com (X. Wang). 0378-1127/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.foreco.2005.12.029 systems. Decades of fire suppression have reduced fire size, prolonged the fire return interval (i.e., the number of years between two successive fire events for a specific area), and indirectly influenced forest composition and dynamics (Shu et al., 1996). The success of fire suppression, coupled with a warmer, drier climate due to global warming (Xu, 1998), have led to a fuel buildup and resulted in fires of greater intensity and extent than those that occurred previously in the region. Catastrophic fires can have disastrous effects on forest composition and structure, ecosystem processes, and landscape pattern (Romme, 1982; Turner et al., 1997, 1999). On May 6, 1987, a catastrophic fire occurred on the northern slopes of Great Hing’an Mountains, burning a total area of 1.3
106 ha. X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 This immense, high-intensity fire consumed vegetation cover and precipitated the exposition of mineral soils – as well as the subsequent erosion and runoff – during post-fire rain events (Xiao et al., 1988; Shu et al., 1996). Forest recovery in such vastly burned areas is challenging because the long-term landscape-level vegetation dynamic in a forest landscape is complicated by spatial and temporal interactions among multiple ecological and anthropogenic processes. In many cases, natural succession can eventually lead to post-fire recovery. This is especially true for cases in which there are sufficient residual forests remaining nearby to act as seed sources (Turner et al., 1999; Borchert et al., 2003). In cases such as the 1987 fire in the Great Hing’an Mountains, natural recovery is difficult because the severely burned area is extensive, the burn severities are high, and the seed sources are far removed (Xiao et al., 1988). In these situations, the process of vegetation recovery is slow, increasing the risk of soil erosion and environmental degradation. Thus, ecological restoration through human mediation is necessary. After the 1987 fire, forest management in this region shifted from timber harvesting to reforestation – particularly in the severely burned area – in order to accelerate forest restoration. Various approaches have been developed to restore forest vegetations for degraded systems where natural recovery is unlikely. Grass seeding provides quick, temporary vegetation ground cover; these are typically annuals or short-lived perennials that can hold soil (Beyers, 2004). However, such a treatment does not facilitate long-term ecosystem restoration. Long-term ecosystem restorations are accomplished either through the planting of a small number of early successional nursery trees or shrubs to create habitats for seed-dispersing birds (e.g., Lamb, 1998), or through high-density plantation of tree species once present in the disturbed areas (e.g., Smale et al., 2001; Nagashima et al., 2002). Although these approaches have shown promising results for development of forest structure, increase of species richness, and recovery of natural successional processes, they are often limited to relatively small areas (Lamb, 1998). In the Great Hing’an Mountains, it is not sufficient to restore only a few burned areas, and it is highly impractical to plant trees in the vast severely burned area due to the considerable labor and economic resources required. It is therefore necessary to assess the planting intensity (the proportion of the burned area for reforestation) and spatial patterns of plantation (the spatial allocation of reforestation in the field) for forest recovery. In the Great Hing’an Mountains, the planting strategy after the catastrophic fire in 1987 was to plant coniferous seedlings in an aggregated fashion over about 50% of the severely burned area where coniferous forests used to grow. Others such as dispersed planting, an alternative to the aggregated planting, have not been evaluated and compared. A dispersed planting would entail planting trees in a random fashion across the entire severely burned area. Descriptive studies and field experiments are often inadequate sources of input for managers developing and implementing reforestation plans. Comparing ecological restoration strategies on a large-scale landscape is often 83 beyond the limits of traditional or experimental studies. Models have therefore become an important tool for predicting the effects of alternative management options. Landscape models are particularly important in this study because other types of models, such as gap models and ecosystem process models, are often limited in spatial extents (He et al., 2002b; Mladenoff, 2004). While other anthropogenic disturbances – such as fire suppression and forest harvesting – have been studied using landscape models (Gustafson et al., 2000; Franklin et al., 2001; Sturatevant et al., 2004; Wimberly, 2004), the effects of reforestation on long-term forest dynamics has not been explicitly simulated. The purpose of this research is to study the effects of different reforestation scenarios on vegetation dynamics in the severely burned area of the Great Hing’an Mountains. We will examine the effects of planting intensities and spatial patterns of plantation (dispersed vs. aggregated) on forest landscape recovery. We will apply a factorial design of planting intensity and spatial pattern of plantation on the realistically parameterized forest composition maps of 1987 to identify the combinations of planting intensity and method that could best accelerate forest restoration. Understanding the probable postfire dynamics of the region under different reforestation scenarios will not only provide insights into landscape scale processes, but also will provide baseline information on forest landscape restoration in northeast China after catastrophic fire disturbances. 2. Study area The Tuqiang Forest Bureau (Fig. 1), encompassing approximately 4
105 ha on the northern slopes of Great Hing’an Mountains, is in the Mohe County of Heilongjiang province in northeast China (from 528150 5500 to 538330 4000 N, and 1228180 0500 to 1238290 0000 E). It borders Russia to the north (separated by Heilongjiang River), the Xilinji Forest Bureau to the west, the Amur Forest Bureau to the east, and the Inner Mongolia Autonomous Region to the south. The area has a cold, continental climate, with an average annual temperature Fig. 1. The location of Tuqiang Forest Bureau and burn severities after the catastrophic fire in 1987 (1, unburned area (MA 1); 2, severely burned area where conifers were dominant before the fire (MA 2); 3, severely burned area where conifers were not dominant before the fire (MA 3); 4, other burned area (MA 4)). 84 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 at 5 8C. Monthly mean temperature ranges from 47.2 8C in January to 31.4 8C in July. The average annual precipitation is 432 mm, with great inter-annual variations. Seventy-five percent of the rainfall occurs between June and August. Uplands and small hills characterize this region, though it possesses a relatively smooth topography. Slopes are generally less than 158; the maximum slope is less than 458. Hills undulate throughout this area, and the mountain ranges mostly run in north and south directions. Elevations range from 270 m to 1210 m; the mean elevation is 500 m. Brown coniferous forest soil is representative in the Bureau. Vegetation is dominated by larch (L. gmelini) forests. White birch (Betula platyphylla) is the major broad-leaved species in the region. In addition to larch and white birch, the tree species include Mongolian Scotch pine (P. sylvestris var. mongolica), spruce (Picea koraiensis), aspen-D (Populus davidiana), black birch (Betula. davurica), aspen-S (Populus suaveolens), and willow (Chosenia arbutifolia). On May 6, 1987, a catastrophic forest fire of 1.33
106 ha took place on the northern slopes of the Great Hing’an Mountains. The burned area covered 2.31
105 ha in the Bureau, and the severely burned area covered roughly 9
104 ha (Fig. 2). The conflagration caused incredible damage to Tuqiang Forest Bureau, and this has led to great difficulty in restoring the forest ecosystems. Fig. 2. Landtype maps of Tuqiang Forest Bureau. 3. Methods 3.1. Description of LANDIS LANDIS is a landscape disturbance and succession model that facilitates the study of the effects of natural and anthropogenic disturbances on forest landscapes, and has been described extensively elsewhere (Mladenoff and He, 1999; He and Mladenoff, 1999; He et al., 1999; Gustafson et al., 2000). Here we provide a general description of the model. In LANDIS, a landscape is organized as grid of cells, with vegetation information stored as attributes for each cell. Cell size can be varied from 10 m to 500 m depending on the research scale. At each cell, the model tracks a matrix containing a list of species by rows and the 10-year age cohorts by columns. The model does not track individual trees. This differs from most from stand simulation models than track individual trees (Grimm, 1999). Additionally, computational loads are greatly reduced, because actual species abundance, biomass, or density is not calculated. A species presence/ absence approach allows LANDIS to simulate large landscapes and avoids any false precision of predicting species abundance measures with inadequate input data or parameter information. LANDIS stratifies a heterogeneous landscape into land types, which are generated from GIS layers of climate, soil, or terrain attributes (slope, aspect, and landscape position). It is assumed that a single land type contains a somewhat uniform suite of ecological conditions, resulting in similar species establishment patterns and fire disturbances characteristic, including ignition frequency, mean fire return interval, and fuel decomposition rated (He and Mladenoff, 1999). These assumptions have been supported by many empirical and experimental studies (e.g., Brown and See, 1981; Kauffman et al., 1988). Furthermore, land types can be redefined by users to partition the landscape into strata that are most relevant for a particular application. Four spatial processes and numerous non-spatial processes are simulated by LANDIS. The spatial processes are fire, windthrow, harvesting, and seed dispersal. Fire is stochastic process based on the probability distributions of fire cycle and mean fire sizes for various land types (He and Mladenoff, 1999). LANDIS simulates fire levels of fire intensity from surface fire to crown fires. Fire intensity is determined by the amount of fuel on a site. Tree species are also grouped into fire fire-tolerance classes based on their fire-tolerance attributes and five age-based fire susceptibility classes from young to old, with young trees being more susceptible to damage than older trees. Thus, fire severity is the interaction of susceptibility based on species age classes, species fire tolerance and fire intensity. A new LANDIS fire module (Yang et al., 2004) employs hierarchical probability theory to allow even more explicit simulation of different fire regimes across landscapes. Windthrow is also stochastically simulated. In the LANDIS wind module the probability of windthrow mortality increases with tree age and size. Windthrow events interact with fire disturbance such that windthrow increases the potential fire intensity class at a site due to increased fuel load. 85 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 The LANDIS harvest module simulates forest harvesting activities based on management area and stand boundaries (Gustafson et al., 2000). These maps are predefined and are only used by the LANDIS harvest and fuel modules. Harvest activities are specified through rules relative to spatial, temporal, and species age-cohort information tracked in LANDIS. The spatial component determines where harvest activities occur and may be used to enforce stand boundary and adjacency constraints. The temporal component determines the timing (rotations) and manner (single versus multiple-entry treatments) of harvest activities. The species age-cohort component allows specification of the species and age cohorts removed by the harvest activities. For example, a clearcut removes all species and all ages, whereas a selection harvest typically removes only a few species and age cohorts. The ability to use a combination of spatial, temporal, species, and age information to specify harvest action independently allows a great variety of harvest prescriptions to be simulated (Gustafson et al., 2000). LANDIS simulates seed dispersal based upon species’ effective and maximum seeding distance (He and Mladenoff, 1999). Seed dispersal probability is modeled for each species using an exponential distribution that defines the effective and maximum seed dispersal distances. Non-spatial processes of succession and seedling establishment are simulated independently at each site. They also interact with spatial processes such as seed dispersal, harvesting, and disturbances. Succession is a competitive process driven by species life history parameters in LANDIS. It is comprised of a set of logical rules primarily using the combination of shade tolerance, seeding ability, longevity, vegetative reproduction capability, and the suitability of the land type (Mladenoff and He, 1999). These rules are used to simulate species birth, growth and death at 10-year intervals. For example, shade intolerant species cannot establish on a site where species with greater shade tolerance are present. On the other hand, the most shade tolerant species are unable to occupy an open site. Without disturbance, shade tolerant species will dominate the landscape given that other attributes (e.g., dispersal distances) are not highly limiting and the environmental conditions are otherwise suitable. Species establishment is regulated by a species-specific establishment coefficient (ranging from 0 to 1.00), which quantifies how different land types favor or inhibit the establishment of a particular species (Mladenoff and He, 1999). These coefficients, which are provided as input to LANDIS, are derived either from the simulation results of a gap model (e.g., He et al., 1999) or from estimates based on existing experimental or empirical studies (Shifley et al., 2000). Due to the stochastic nature of processes such as fire, windthrow, and seedling dispersal simulated in the model, LANDIS is not designed to predict the specific time or place that individual disturbance events will occur. Rather, it is a cause– response type of scenario model that simulates landscape patterns over time in response to the combined and interactive outcomes of succession and disturbance. It can provide managers with guidance about management practices that can mitigate current or anticipated problems on the forest landscape, and provide a better understanding of long-term, cumulative effects that may result from the combination of natural disturbances and management practices. LANDIS has been applied and tested with different species and environmental settings (Gustafson et al., 2000; Shifley et al., 2000; Franklin et al., 2001; He et al., 2002a,b; Mehta et al., 2004; Wimberly, 2004). In addition, the current version, LANDIS 4.0, added new capabilities that simulate explicit fuel dynamics, fuel-fire interactions, and biological disturbances (He et al., 2004). Parameterization of LANDIS 4.0 for the Tuqiang Forest Bureau involved several aspects: species’ vital attributes, a forest composition map that contains individual species presence/absence and age classes at each cell, a land type map, establishment probabilities for each landtype, fire disturbance regimes for each landtype, and forest management scenarios. 3.1.1. Species attributes and forest composition map A total of eight tree species were incorporated into LANDIS. Species’ vital attributes (Table 1) were estimated based on existing studies of the region (Ai et al., 1985; Duan, 1991; Hu et al., 1991; Xu, 1998; He et al., 2002a; Xu et al., 2004) as well as consultation with local experts. A forest composition map was derived from an extant forest stand map of 1987, a stand attribute database, and one scene of Landsat TM imagery taken in 1987. The forest stand map recorded boundaries of stands and compartments. (A compartment is a unit of forest inventory, generally containing 10–100 stands.) The stand attribute database provided information on the relative percentage of canopy species, the average age of dominant canopy species, timber production, and crown density. To reduce computational loads during model simulations, the forest composition map Table 1 Species’ life attributes for Tuqiang Forest Bureau in northeastern China Species LONG MTR ST FT ED MD VP MVP Larch (Larix gmelini) Mongolian Scotch pine (Pinus sylvestris var. mongolica) Spruce (Picea koraiensis) White birch (Betula platyphylla) Aspen-D (Populus davidiana) Black birch (Betula davurica) Aspen-S (Populus suaveolens) Willow (Chosenia arbutifolia) 300 250 300 150 180 150 150 200 20 40 30 15 20 15 25 30 3 1 4 1 1 1 1 2 3 1 2 3 3 4 4 2 150 100 10 200 1 200 1 1 300 200 150 4000 1 1000 1 1 0 0 0 0.8 1 0.8 0.9 0.9 0 0 0 40 40 40 40 30 LONG: longevity (years); MTR: age of maturity (years); ST: shade tolerance (1–5); FT: fire tolerance (1–5); ED: effective seeding distance (m); MD: maximum seeding distance (m); VP: vegetative reproduction probability; MVP: minimum age of vegetative reproduction (years). 1 represents unlimited seeding range. 86 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 Fig. 3. Spatial distribution of part landscape occupied by larch at simulation year 200. (I1, I2, I3, and I4 represent simulation scenarios where the planting intensity covers 10%, 30%, 50%, and 70% of the severely burned area, respectively. M1 and M2 represent dispersal planting and aggregated planting, respectively.) was processed at 90 m
90 m resolution, which yielded 1604 rows
873 columns. Each cell contains the presence/absence and age cohorts of all eight tree species. For each cell in a stand, we used a stand-based assignation (SBA) approach (Xu et al., 2004) to stochastically assign species age cohorts to each cell based on forest inventory data. Xu et al. (2004) used uncertainty analysis to evaluate the approach, and their results demonstrated that uncertainty was relatively low at the cell level during the beginning of the simulation. The uncertainty increased with simulation years, but the uncertainty finally reached an equilibrium state in which input errors in original species age cohorts had little effect on the simulation outcomes. At the landscape level, species abundance and spatial patterns were not substantially affected by the uncertainties in species age structure at the cell level. Since the typical application of LANDIS is to predict long-term landscape pattern changes, SBA can be used to parameterize species age cohorts for individual cells. 3.1.2. Landtype map LANDIS stratifies the heterogeneous landscape into relatively homogeneous units (landtypes or ecoregions). Within each landtype, similar environments for species establishment are assumed (Mladenoff and He, 1999). In the study, we derived eight landtypes, primarily based on terrain attributes, TM images taken in 1987, and the catastrophic fire of 1987 (Fig. 3). These landtypes include water, residential land, terrace, southern slope (SS), northern slope (NS), burned terrace (BT), burned southern slope (BSS), and burned northern slope (BNS). Because large burned areas came into existence after the 1987 fire, fire/fuel characteristics – such as fuel accumulation and the time elapsed since the previous fire – were different for unburned areas. Therefore, we differentiated BT, BSS, and BNS from unburned areas. All landtypes were interpreted from the previous forest inventory and the TM images taken in 1987. Non-active landtypes (water and residential land) account for 0.90% of the total area, while active landtypes (including terrace, SS, NS, BT, BSS, and BNS) account for 0.94%, 23.57%, 17.39%, 3.39%, 31.43%, and 22.38%, respectively. The species establishment coefficient is a critical feature of each landtype. It is an estimate of the probability that a species will successfully establish on a landtype, given the environmental conditions encapsulated by that landtype. The species establishment coefficients (Table 2) were derived from available literature as well as the existing LANDIS parameterizations on northeastern China (Li et al., 1987; Zhao et al., 1997; Liu et al., 1999; He et al., 2002a; Hu et al., 2004). Table 2 Attributes for each land type of Tuqiang Forest Bureau in China Land type MFRI TSLF EC1 EC2 EC3 EC4 EC5 EC6 EC7 EC8 SS NS T BSS BNS BT 0.25 0.2 0 0.25 0.2 0 0.01 0.01 0.001 0.01 0.01 0.001 0.1 0.05 0 0.1 0.05 0 300 350 1000 300 350 1000 150 180 500 10 10 10 0.3 0.25 0.002 0.3 0.25 0.002 0.03 0.05 0 0.03 0.05 0 0.3 0.25 0.004 0.3 0.25 0.004 0 0 0.01 0 0 0.01 0 0 0.05 0 0 0.05 MFRI: mean fire return intervals in years; TSLF: time since last fire disturbance; EC1, EC2, EC3, EC4, EC5, EC6, EC7, EC8 are the establishment coefficient for larch, Mongolian Scotch pine, spruce, white birch, aspen-D, black birch, aspen-S, and willow. SS: southern slope; NS: northern slope; T: terrace; BSS: burned southern slope; BNS: burned northern slope; BT: burned terrace. 87 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 3.1.3. Fire regimes The current fire regime for our simulations was parameterized based on a database of 10-year fire dating from 1990 to 2000 (Hu et al., 2004). Current fire return intervals were estimated by calculating the reciprocal of the annual proportion of forest land burned within each landtype in the Bureau. Only those fires greater than the resolution of the simulation recorded in the database (0.81 ha) were used to parameterize the fire regime (Table 2). Within the study area, the common planting strategy consisted of planting coniferous trees over about 50% of the total severely burned area (about 41300 ha). Two endemic coniferous species of larch and Mongolian Scotch pine (the ratio is about 7:3) were the main tree species planted. 3.2. Simulation scenarios We began with the realistically parameterized forest composition and landtype maps including species/age classes that represent the initial status of 1987. Four levels of planting intensity (10%, 30%, 50%, and 70% of the severely burned area) and two spatial patterns of plantation (dispersed planting and current/aggregated planting) were simulated in a 4
2 factorial design over a 300-year period (Table 3). Two coniferous tree species (larch and Mongolian Scotch pine) were planted in severely burned areas, and the proportion was 7:3. Five replicates of each scenario were simulated. The current planting intensity of planting coniferous species over about 50% of the severely burned area was simulated; additionally, three other planting intensities (including 10%, 30%, and 70% of the severely burned area) were selected to compare with the current planting intensity. The planting strategy, dispersed planting, consists of planting coniferous seedlings randomly throughout a severely burned area, whereas aggregated planting consists of planting coniferous seedlings in selected locations of severely burned areas where coniferous forests were dominant prior to a fire; both strategies were applied in our simulations. A ‘‘no reforestation’’ scenario was also simulated. All scenarios were simulated up to 300 years to examine the long-term effects of planting intensity and spatial pattern of plantation on forest succession. The study area was divided into four management areas (MA) indicating the planting options practiced there: (1) unburned area (MA 1), (2) severely burned area where conifers were dominant before the 1987 fire (MA 2), (3) severely burned area where conifers were not dominant before the fire (MA 3), and (4) other burned area (MA 4) (Fig. 1). The goal of the current planting strategy was to increase the proportion of coniferous trees in the Bureau. Dispersed planting was applied in MA 2 and MA 3, while current planting was restricted in MA 2. Within MA 2, planting activity was applied to 50% of the severely burned area in 10 years to mimic the current planting strategy designed to promote coniferous species. In MA 1 and MA 4, natural regeneration was not simulated because there were sufficient residual coniferous trees to provide seeds. Table 3 gives an overview of different combinations of planting intensities and spatial patterns of plantation. For each of the three species (larch, Mongolian Scotch pine, and white birch), the proportion of area (PA) was calculated for the mapped output of species for each 10-year step in the 300 simulation years to qualitatively evaluate temporal trends in the data. We divided the 300 simulation years into three stages: 10– 30 year as short-term stage, 40–100 year as mid-term stage, and >100 years as long-term stage. The average PA in each of the three stages was calculated for all three species. However, the average PA for the three species is dependent on each other at different simulation stages. Therefore, we analyzed the three response variables (average area of the three species) together in a multivariate analysis of variance (MANOVA) with planting intensity (n = 4) and spatial pattern of plantation (n = 2) in SPSS (SPSS 10.0) to test the global hypothesis that planting intensity and spatial pattern of plantation affect the area of the three species at different simulation stages. We used the Pillai’s Trace statistic to test our hypotheses because it is the least sensitive of the four multivariate tests provided by SPSS to the heterogeneity of variance assumption of MANOVA (Zar, 1999). In addition, to evaluate the effect of reforestation scenarios on the spatial pattern of species distributions, we calculated an aggregation index (AI) for the three species using the mapped output of each species throughout the 300-year simulation. The AI has a range between 0 and 1, with higher values indicating a higher level of aggregation in the measured spatial pattern (He et al., 2000). It assumes that a class with the highest level of aggregation (AI = 1) is comprised of pixels sharing the most possible edges. A class whose pixels share no edges (completely disaggregated) has the lowest level of aggregation (AI = 0). If ei,j represents total edges of class i adjacent to class j, for class i of area Ai, aggregation index measures ei,i, the total Table 3 The scenarios simulated by LANDIS Simulation scenarios Descriptions I1M1 I2M1 I3M1 I4M1 I1M2 I2M2 I3M2 I4M2 Planting Planting Planting Planting Planting Planting Planting Planting coniferous coniferous coniferous coniferous coniferous coniferous coniferous coniferous trees, trees, trees, trees, trees, trees, trees, trees, covering covering covering covering covering covering covering covering 10% 30% 50% 70% 10% 30% 50% 70% of of of of of of of of the the the the the the the the severely severely severely severely severely severely severely severely burned burned burned burned burned burned burned burned area area area area area area area area (70% (70% (70% (70% (70% (70% (70% (70% larch larch larch larch larch larch larch larch and and and and and and and and 30% 30% 30% 30% 30% 30% 30% 30% Mongolian Mongolian Mongolian Mongolian Mongolian Mongolian Mongolian Mongolian Scotch Scotch Scotch Scotch Scotch Scotch Scotch Scotch pine); pine); pine); pine); pine); pine); pine); pine); dispersed planting dispersed planting dispersed planting dispersed planting aggregated planting aggregated planting aggregated planting aggregated planting 88 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 edges shared by class i itself. The shared edges are counted only once in AI, and currently only four neighbors are considered. The level of aggregation of class i is calculated as: ei;i AIi ¼ max ei;i: Given a class is of area Ai, the maximum aggregation level is reached when A clumps into one patch that has the largest ei,i (e.g., it does not have to be a square). If n is the side of largest integer square smaller than Ai, and m = Ai  n2, then the largest number of shared edges for class i, max_ei,i, will take one of the three forms: max ei;i ¼ 2nðn  1Þ; when m ¼ 0; or max ei;i ¼ 2nðn  1Þ þ 2m  1; when m < n; max ei;i ¼ 2nðn  1Þ þ 2m  2; when m  n: or The two indices (PA and AI) were calculated in APACK (Mladenoff and DeZonia, 1997). This analysis was conducted for the entire study area. 4. Results MANOVA results indicate that both planting intensity and spatial pattern of plantation also significantly affect the PA of the three species (larch, Mongolian Scotch pine and white birch) across the study landscape at short-, mid-, and long-term stages (Tables 4–6), and their interaction of both treatments has also a significant influence. Planting intensity always has a significant influence on the successional dynamic of the three species for all simulation years. However, the effects include both positive and negative effects: greater planting intensity positively influences larch and Mongolian Scotch pine abundance, but negatively influences white birch abundance. Larch abundance is greater under high planting intensity scenarios than under low planting intensity scenarios. For example, under the 10% planting intensity scenario, it takes roughly 200 years for larch to reach 70% of the landscape, but it takes only 60 years to reach 70% of the landscape under the 70% planting intensity scenario. The PA of Mongolian Scotch pine responds to planting intensity in a similar way, with a greater abundance under high planting intensity scenarios than that under low planting intensity scenarios. In addition, the PA for larch increases with simulation years in all simulation scenarios, whereas the PA for Mongolian Scotch pine decreases with simulation years after roughly 10–20 years (Fig. 4). Mongolian Scotch pine has a relatively low abundance – less than 10% of the study area during most simulation years – because it is a fireintolerant and shade-intolerant species and is less competitive than either larch or white birch. Contrary to larch, white birch has a greater abundance under low planting intensity scenarios than that under high planting intensity scenarios. For example, under the 10% planting intensity scenario, the PA for white birch reaches 30% after only 10 years, and exceeds 40% after 40 years. However, under the 70% planting intensity scenario, it requires between 30 and 40 years attaining 30% abundance, and never reaches 40% in 300 simulation years. Individual ANOVA results also show that spatial pattern of plantation has a significant influence on the short-, mid-, and long-term PA for larch, and has a significant influence on the long-term PA for both Mongolian Scotch pine and white birch (Tables 4–6). In other words, spatial pattern of plantation has a significant influence on the abundance of larch during all simulation years, but has no short- or mid-term influence on Table 4 MANOVA and individual ANOVA results for the average area of the three species (larch, Mongolian Scotch pine and white birch) at short-term stage as a function of planting intensity and spatial pattern of plantation Effect df (h, e) MANOVA global test of hypotheses Intensity (I) Spatial pattern of plantation (P) I
P 9,96 3,30 9,96 Individual ANOVA tests of hypotheses Average area of larch at short-term stage–Model R2 = 0.96 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of Mongolian Scotch pine at short-term stage–Model R2 = 0.998 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of white birch at short-term stage–Model R2 = 0.988 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Pillai’s trace/type III SS F P 1.33 0.92 0.72 8.53 112.69 3.39 <0.001 <0.001 0.001 441.49 3.32 0.39 18.55 253.86 5.73 0.23 <0.001 0.023 0.88 6643.83 0.16 0.29 <0.001 0.69 0.834 73.49 0.0006 0.003 0.118 309.30 0.034 0.014 3.75 880.75 0.289 0.041 <0.001 0.594 0.989 89 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 Table 5 MANOVA and individual ANOVA results for the average area of the three species (larch, Mongolian Scotch pine and white birch) at mid-term stage as a function of planting intensity and spatial pattern of plantation Effect df (h, e) MANOVA global test of hypotheses Intensity (I) Spatial pattern of plantation (P) I
P 9,96 3,30 9,96 Individual ANOVA tests of hypotheses Average area of larch at mid-term stage–Model R2 = 0.994 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of Mongolian Scotch pine at mid-term stage–Model R2 = 0.996 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of white birch at mid-term stage–Model R2 = 0.975 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Pillai’s trace/type III SS F P 1.65 0.96 0.87 13.00 237.22 4.37 <0.001 <0.001 <0.001 521.9 52.16 6.12 3.66 1519.73 455.65 17.83 <0.001 <0.001 <0.001 2610.86 2.59 2.03 <0.001 0.118 0.13 423.24 0.32 0.50 <0.001 0.574 0.684 81.37 0.027 0.06 0.33 631.96 0.16 0.75 15.93 Table 6 MANOVA and individual ANOVA results for the average area of the three species (larch, Mongolian Scotch pine and white birch) at long-term stage as a function of planting intensity and spatial pattern of plantation Effect df (h, e) MANOVA global test of hypotheses Intensity (I) Spatial pattern of plantation (P) I
P 9,96 3,30 9,96 Individual ANOVA tests of hypotheses Average area of larch at long-term stage–Model R2 = 0.993 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of Mongolian Scotch pine at long-term stage–Model R2 = 0.989 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 Average area of white birch at long-term stage–Model R2 = 0.969 Intensity (I) 3 Spatial pattern of plantation (P) 1 I
P 3 Error (e) 32 either Mongolian Scotch pine or white birch. The interaction of treatment intensity and treatment method has no effect on the dynamics of Mongolian Scotch pine throughout the simulation years, and only affects the long-term stage of white birch abundance. The PA for larch under the dispersed planting scenario is higher than that under the aggregated planting scenario, very likely because dispersed planting increases the cover range of seed dispersal and benefits the regeneration of Pillai’s trace/type III SS 1.64 0.99 0.965 F P 12.8 796.98 5.06 <0.001 <0.001 <0.001 179.2 132.73 11.11 2.16 886.72 1970.36 54.97 <0.001 <0.001 <0.001 38.63 0.20 0.05 0.42 2610.86 15.30 1.15 <0.001 <0.001 0.346 287.37 16.88 3.38 9.88 310.41 54.69 3.65 <0.001 <0.001 0.023 larch. However, Mongolian Scotch pine does not respond to spatial pattern of plantation in a similar way. Spatial pattern of plantation has no influence on the abundance of Mongolian Scotch pine at the early simulation stage because the pine has a low proportion in the landscape and only relatively short dispersal distances. At the long-term simulation stage, spatial pattern of plantation shows an influence on Mongolian Scotch pine because such a long time frame captures its slow and 90 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 Fig. 4. Proportion of sites in the study landscape occupied by larch, Mongolian Scotch pine, and birch during each decade for all simulation scenarios. I1, I2, I3, and I4 represent simulation scenarios where the planting intensity covers 10%, 30%, 50%, and 70% of the severely burned area, respectively. cumulative dispersal behavior. However, the abundance of Mongolian Scotch pine is only slightly higher under the dispersed planting scenario than that under the aggregated planting scenario (Fig. 4). Due to its strong dispersal and colonization ability, white birch can invade open spaces created by fire. This results in no significant influence of the two spatial patterns of plantation on white birch abundance at the early simulation stage. However, in the later simulation stages, the abundance of white birch under the dispersed planting scenario is lower than that under the aggregated planting scenario due to strong competition with larch. The environment becomes more unfavorable for white birch in a dispersed planting as the canopy closure of larch forests increases. Compared with the no planting scenario, the abundance of larch in these planting scenarios was greater, which suggests that planting activity is effective for increasing the abundance of larch throughout these simulation years. Given its vital attributes and low abundance, the decreasing abundance of Mongolian Scotch pine with greater simulation years in all simulation scenarios is not surprising. The difference between these planting simulation scenarios and the no planting scenario – especially in the later simulation stage – is discernable, with the abundance of white birch under the no planting scenario higher than that in the planting simulation scenarios due to its diminished competition with larch. The simulated results indicate that planting intensity and spatial pattern of plantation not only significantly influence the abundance of the three species, but also influence the spatial pattern of these species (Fig. 5). However, the degree of aggregation of species distribution differs by different planting scenarios. The aggregation levels for larch under all planting scenarios are similar, probably because larch has relatively high proportions under all simulation scenarios; high proportions tend to have aggregated distributions in the study landscape. The aggregation level of larch is considerable high, with an AI larger than 0.7 at the start of the simulation, and stabilizing at 0.8 after only a few simulation decades. The distribution of Mongolian Scotch pine is more aggregated under the high planting intensity scenarios than that under the low planting intensity scenarios, whereas the aggregation level of white birch is negatively correlated with greater planting intensity. Also, the distribution of pine under the dispersed planting scenarios has a lower aggregation level than that under the aggregated planting scenarios. However, the difference in the level of aggregation between the two spatial patterns of plantation increases with the planting intensity. There is a sudden decrease in the AI for pine after about 200 simulation years; this is because the large number of pines planted at the beginning of the simulation began to die when they approached their 250-year longevity. The aggregation level of white birch responds to spatial pattern of plantation in a way similar to that for larch, with its AI lower under the dispersed planting scenarios than under the aggregated planting scenarios. X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 91 Fig. 5. Aggregation index in the study landscape occupied by larch, Mongolian Scotch pine, and birch during each decade for all simulation scenarios. I1, I2, I3, and I4 represent simulation scenarios where the planting intensity covers 10%, 30%, 50%, and 70% of the severely burned area, respectively. 5. Discussion Both planting intensity and spatial pattern of plantation play an important role in influencing tree species abundance. However, planting intensity exerts the dominant effect on the abundance of the three tree species examined. Greater planting intensities simulated in the study increase the abundance of larch and Mongolian Scotch pine, but decrease the abundance of white birch. Although larch abundance increases with increasing planting intensity, the degrees of the increase among the planting intensity scenarios are different. Species abundance has a greater, more sensitive response when planting intensity is low. For example, the difference in larch abundance under the 10% planting intensity scenario and the 30% planting intensity scenario is obviously higher than that under 50% planting intensity scenario and 70% planting intensity scenario (Fig. 4). Additionally, the aggregation levels for larch among the simulation scenarios are not significantly different. These results suggest that intermediate planting intensity, such as 30% or 50%, can achieve results for abundance and aggregation levels of larch comparable to those simulated under high intensity planting (e.g., 70%, 90%). For Mongolian Scotch pine, intermediate planting intensity scenarios should also be considered, since Mongolian Scotch pine cannot effectively compete with birch and larch. The relatively high abundance of Mongolian Scotch pine under a high planting intensity, which would involve both great labor and economic costs, will eventually decrease over the mid- and long-term periods. With respect to spatial patterns of plantation, larch abundance under a dispersed planting scenario is greater than that under an aggregated planting scenario. However, the interaction between spatial patterns of plantation and planting intensities can be intriguing. For example, the difference in larch abundance between dispersed planting and aggregated planting under intermediate planting intensity scenarios (30% and 50% planting intensity) is greater than that under both the low planting intensity scenario (10% planting intensity) and the high planting intensity scenario (70% planting intensity) (Fig. 4). In a low planting intensity scenario, the space created by the 1987 fire remained unplanted, which resulted in small differences between dispersed planting and aggregated planting. However, in a high planting intensity scenario, little space created by the 1987 fire remained unplanted, which also resulted in the small differences under dispersed planting and aggregated planting. These results indicate that it is necessary to incorporate information about spatial patterns of plantation into planting strategies under intermediate planting intensity scenarios. Therefore, under the current planting intensity (about 50%), dispersed planting would be better for forest recovery than aggregated planting. In Great Hing’an Mountains, recovery of the site and the reestablishment of mature ecological systems (coniferous forests) will probably require from 100 to 200 years to restore the life forms (e.g., Zhou, 1991; Xu, 1998). Our simulations show that coniferous seedlings planting have significantly longterm influence on forest recovery in the catastrophic fire burned landscape, which are consistent with some studies on 92 X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 reforestation (e.g., Miyawaki and Golley, 1993; Chen et al., 2003), for coniferous seedlings planting will eventually replicate the mature ecosystems of the site without going through the lengthy process of succession. Residual trees play an important role in forest restoration (e.g., Keeton and Franklin, 2005). However, only few residual trees remained in the severely burned area after the 1987 fire, which resulted in a lower abundance of coniferous trees under the no planting scenario than that under the planting scenarios. Therefore, large area clear-cutting should not be applied in forest management for this region. Coniferous trees planted in the severely burned area serve as residual trees, which further disperse seeds for forest recovery—that is, more residual trees are left with increasing planting intensity. In addition, the simulation results demonstrate that dispersed planting yields greater abundance of larch forests than aggregated planting, suggesting that some dispersed coniferous trees or seedlings should be left as seed sources for forest recovery when timber harvesting is applied in a forest landscape. Two key assumptions likely affect our conclusions. The first is that the species establishment coefficients do not vary at the scale of landtypes. Fine scale heterogeneity, such as soil moisture and temperature availability, are important variables to affect forest dynamics because they would affect plant establishment and growth, species diversity, nutrient cycling, rates of soil formation and other biotic activity (e.g., Xu, 1998). In reality, drought sites are often colonized by Mongolian Scotch pine rather than larch, for the pine has a lower shade tolerance than larch. Hence, the abundance of Mongolian Scotch pine may be underestimated in these scenarios. However, given its vital attributes discussed above, we believe that Mongolian Scotch pine could not have a high proportion of the study landscape. In addition, species establishment coefficients may vary with fire severity. In lowseverity burned areas, burning prepares seedbeds by removing duff and competing vegetation, and by altering the thermal regime of the soil, which favors germination. But in severely burned areas, organic layers of soil were almost completely consumed—an unfavorable condition for the regeneration of larch and Mongolian Scotch pine (Xiao et al., 1988; Xu, 1998). Using finer-scale land units to parameterize our simulations could alleviate these problems. However, no available data currently exists to estimate species establishments at a finer spatial scale. A second assumption is that once established, trees die only because of disturbance-caused mortality, approaching their longevity, and a small chance of random mortalities simulated in the model. In other words, other types of mortality such as stress related in the early stage of their life span are not considered in the LANDIS model. This assumption is appropriate, given that human interference is not considered. In this study, intensive planting is employed in the burned area, but the success rates of species establishment are also affected by the quality of the seedlings and aftermath care after planting (e.g., Xiao et al., 1988; Zhao, 1988; Tan, 1994). These factors are not addressed in LANDIS modeling. In addition, LANDIS did not simulate the effects of shrubs and herbs on forest recovery, which could have significant influence on coniferous seedlings. For example, some studies (Turner et al., 1997, 1999) have shown that shrub and herb recruitment following forest fires in burned areas where conifer seedlings were scarce may lengthen the time required for forest development, or even preclude them entirely, for shrub and herb may compete with trees for resources (e.g., light, nutrients). Previous studies in the Great Hing’an Mountains showed that shrubs and herbs would soon invade these open spaces created by the catastrophic fire in 1987. In some cases, if vegetation becomes too dense, light levels beneath shrubs and grasses may be insufficient for tree seedlings to maintain a net positive carbon balance, and thus inadequate radiation may cause low coniferous seedling survivorship. Therefore, coniferous seedlings could not regenerate in some places because of competition from the welldeveloped shrub and herb community (e.g., Zhou et al., 1989; Yang et al., 1998; Wang et al., 2003). If shrubs and herbs were not cleared before coniferous seedling plantation, some seedlings planted could not survival, which would cause lower abundance of coniferous tree species (larch and Mongolian Scotch pine). However, this may not be a significant factor: in our study areas, shrubs and herbs must be cleared prior to the coniferous seedling plantation (Shao, 1988; Wang et al., 2003). When coniferous seedlings grow older, they will out-compete shrubs and herbs. Thus, the simulated results could effectively reflect the long-term forest dynamics in the landscape based on this assumption. 6. Conclusion Landscape models are an efficient tool for evaluating the effects of different planting strategies on landscape-scale forest recovery. They provide insights into forest dynamics under natural disturbances and reforestation. In summary, we found that planting intensity can modify the abundance of the three species in the study (larch, Mongolian Scotch pine, and white birch) by encouraging larch and pine establishment that will eventually decrease white birch abundance. Although larch abundance increases with increasing planting intensity, the degree of increased abundance among these planting intensity scenarios is different. Given considerable labor inputs and economic costs, intermediate planting intensity (30% or 50%) would better facilitate forest recovery. In addition, spatial pattern of plantation also influences the three species, with higher larch and pine abundance and lower white birch abundance under dispersed planting than that under aggregated planting. This indicates that it is necessary to incorporate information about spatial patterns of plantation into planting strategies. These results have important implications for forest managers to design sound forest restoration projects for landscapes affected by large infrequent disturbances. Acknowledgements This paper is sponsored by the Chinese Academy of Sciences, National Natural Science Foundation of China (40331008, 30270225), and the Project of Chinese Academy of Sciences (No. KSCX2-SW-133) and Sino-Russian Joint Research Center on Natural Resources and Eco-Environmental Sciences. X. Wang et al. / Forest Ecology and Management 225 (2006) 82–93 References Ai, C.L., Wang, C.D., Ai, S.H., 1985. Study on the seed dispersal of Larix gmelinii. Forest. Sci. Technol. 4, 2–5 (in Chinese). Beyers, J.L., 2004. Postfire seeding for erosion control: effectiveness and impacts on native plant communities. Conserv. Biol. 18 (4), 947–956. Borchert, M., Johnson, M., Schreiner, D.S., Wall, S.B.V., 2003. Early postfire seed dispersal, seedling establishment and seedling mortality of Pinus coulterri (D. Don) in central coastal California, USA. Plant Ecol. 168, 207–220. Brown, J.K., See, T.E., 1981. Downed dead woody fuel and biomass in the northern Rocky Mountains. USDA Forest Service General Technical Report INT-117, pp. 1–47. Chen, X., Li, B., Lin, Z., 2003. The acceleration of succession for the restoration of the mixed-broadleaved Korean pine forests in Northeast China. Forest Ecol. Manage. 177, 503–514. Duan, X., 1991. Study on fire tolerance of Larix gmelinii. Forest Fire Prevent. 31 (4), 7–11 (in Chinese). Franklin, J., Syphard, A.D., Mladenoff, D.J., He, H.S., Simons, D.K., Martin, R.P., Deutschman, D., O_Leary, J.F., 2001. Simulating the effects of different fire regimes on plant functional groups in Southern California. Ecol. Model. 142, 261–283. Grimm, V., 1999. Ten years of individual-based modelling in ecology: what have we learned and what could we learn in the future? Ecol. Model. 115, 129–148. Gustafson, E.J., Shifley, S.R., Mladenoff, D.J., Nimerfro, K.K., He, H.S., 2000. Spatial simulation of forest succession and timber harvesting using LANDIS. Can. J. Forest Res. 30, 32–43. He, H.S., Mladenoff, D.J., Boeder, J., 1999. An object-oriented forest landscape model and its representation of tree species. Ecol. Model. 119, 1–19. He, H.S., DeZonia, B., Mladenoff, D.J., 2000. An aggregation index (AI) to quantify spatial patterns of landscapes. Landscape Ecol. 15 (7), 591–601. He, H.S., Hao, Z., Larsen, D.R., Dai, L., Hu, Y.M., Chang, Y., 2002a. A simulation study of landscape scale forest succession in China. Ecol. Model. 156, 153–166. He, H.S., Mladenoff, D.J., Gustafson, E.J., 2002b. Study of landscape change under forest harvesting and climate warming-induced fire disturbance. Forest Ecol. Manage. 155, 257–270. He, H.S., Li, W., Sturtevant, B.R., Yang, J., Shang, B.Z., Gustafson, E.J., Mladenoff, D.J., 2004. LANDIS 4.0 User’s Guide. School of Natural Resources, University of Missouri-Columbia, Columbia, pp. 1–91. He, H.S., Mladenoff, D.J., 1999. Spatially explicit and stochastic simulation of forest landscape fire disturbance and succession. Ecology 80, 81–99. Hu, H.Q., Yao, S.R., Shang, D.Y., 1991. The characteristics and effect of the forest fire in Northeast China. Forest Fire Prevent. 31 (4), 13–16 (in Chinese). Hu, Y., Xu, C., Chang, Y., 2004. Application of spatially explicit landscape model (LANDIS): a case study in Huzhong area, Mt. Great Hing’an. Acta Ecol. Sinica 24 (9), 1846–1856 (in Chinese). Kauffman, J.B., Uhl, C., Cummings, D.L., 1988. Fire in the Venezuelan Amazon. 1: Fuel biomass and tree chemistry in the evergreen rainforest of Venezuela. Oikos 53, 167–175. Keeton, W.S., Franklin, J.E., 2005. Do remnant old-growth trees accelerate rates of succession in mature Douglas-fir forests? Ecol. Monogr. 75 (1), 103–118. Lamb, D., 1998. Large-scale ecological restoration of degraded tropical forest lands: the potential role of timber plantations. Restor. Ecol. 6 (3), 271–279. Li, R.X., Zhao, H.X., Wang, F.Y., Zhang, D.L., 1987. The relationship between growth of natural Larix gmelinii and land type. J. Northeast Forest. Univ. 15, 55–59 (in Chinese). Liu, G.F., Yang, C.P., Liu, G.J., Yang, R.H., 1999. The configuration characters and germination percentage of various provenance seeds of Betula platyphylla. J. Northeast Forest. Univ. 27 (4), 1–4 (in Chinese). Mehta, S., Frelich, L.E., Jones, M.T., Manolis, J., 2004. Examining the effects of alternative management strategies on landscape-scale forest patterns in northeastern Minnesota using LANDIS. Ecol. Model. 180, 73–87. Mladenoff, D.J., 2004. LANDIS and forest landscape models. Ecol. Model. 180, 7–19. 93 Mladenoff, D.J., DeZonia, B., 1997. APACK 2.0 User’s Guide. Department of Forest Ecology and Management. University of Wisconsin-Madison, Madison, WI, USA. Miyawaki, A., Golley, F.B., 1993. Forest reconstruction as ecological engineering. Ecol. Eng. 2, 333–345. 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.), Advances in spatial Modeling of Forest Landscape Change: Approaches and Applications. Cambridge University Press, Cambridge, UK, pp. 1–13. Nagashima, K., Sands, R., Whyte, A.G.D., Bilek, E.M., Nakagoshi, N., 2002. Regional landscape change as a consequence of plantation forestry expansion: an example in the Nelson region, New Zealand. Forest Ecol. Manage. 163, 245–261. Romme, W.H., 1982. Fire and landscape diversity in subalpine forests of Yellowstone National Park. Ecol. Monogr. 52 (2), 199–221. Shao, C.H., 1988. Suggestions for the regeneration of burned areas in Tuqiang Forest Bureau. Chin. J. Ecol. 7 (sup.), 7–74. Shifley, S.R., Thompson III, F.R., Larsen, D.R., Dijak, W.D., 2000. Modeling forest landscape change in the Missouri Ozarks under alternative management practices. Comput. Electr. Agric. 27, 7–27. Shu, L.F., Zuo, J.C., Tian, W.T., 1996. Discussion on natural regeneration in burned areas of Great Hing’an Mountains. Manage. Forest. Resources 6, 43–45 (in Chinese). Smale, M.C., Whaley, P.T., Smale, P.N., 2001. Ecological restoration of native forest at Aratiatia, North Island, New Zealand. Restoration Ecol. 9, 28–37. Sturatevant, B.R., Zollner, P.A., Gustafson, E.J., Cleland, D.T., 2004. Human influence on the abundance and connectivity of high-risk fuels in mixed forests of northern Wisconsin, USA. Landscape Ecol. 19, 235–253. Tan, J., 1994. Reason and strategy on forest land reduction in Daxing’anling Mountains. Territory Natural Resour. Study 2, 15–19 (in Chinese). Turner, M.G., Romme, W.H., Gardner, R.H., Hargrove, W.W., 1997. Effects of fire size and pattern on early succession in Yellowstone National Park. Ecol. Monogr. 4 (67), 411–433. Turner, M.G., Romme, W.H., Gardner, R.H., 1999. Prefire heterogeneity, fire severity, and early postfire plant reestablishment in subalpine forests of Yellowstone National Park, Wyoming. Int. J. Wildland Fire 9 (1), 21–36. Wang, X., Li, X., Kong, F., 2003. Model of vegetation restoration under natural regeneration and human interference in the burned area of northern Great Hing’an Mountains. Chin. J. Ecol. 22 (5), 30–34 (in Chinese). Wimberly, M.C., 2004. Fire and forest landscapes in the Georgia Piedmont: an assessment of spatial modeling assumptions. Ecol. Model. 180, 41–56. Xiao, D., Tao, D., Xu, Z., 1988. Impacts of an extra-ordinarily disastrous fire on forest resources and environment. Chin. J. Ecol. 7 (sup.), 5–9 (in Chinese). Xu, H., 1998. Forest in Great Hing’an Mountains of China. Science Press, Beijing, pp. 1–231, (in Chinese). Xu, C., He, H.S., Hu, Y., Chang, Y., Larsen, D.R., Li, X., Bu, R., 2004. Assessing the effect of cell-level uncertainty on a forest landscape model simulation in Northeastern China. Ecol. Model. 180, 57–72. Yang, S., Liu, X., Cao, H., 1998. Study on vegetational change in burned areas of Great Hing’an Mountains. J. Northeast Forest. Univ. 26 (1), 19–23 (in Chinese). 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. Model. 180, 119–133. Zar, J.H., 1999. Biostatistical Analysis. Prentice-Hall Inc., Upper Saddle River, USA. Zhao, D., 1988. Vegetations and their restoration after the disastrous fire in the Daxing’anling Mountains. Chin. J. Ecol. 7 (sup.), 35–40 (in Chinese). Zhao, K.T., Li, Y.C., Zhang, G.L., 1997. The development and protection of Pinus pumila forest in Da Hingan Mountain. Territory Natural Resour. Study 2, 55–59 (in Chinese). Zhou, Y., 1991. Vegetation in Great Hing’an Mountains of China. Science Press, Beijing, pp. 1–216, (in Chinese). Zhou, Y., Wu, H., Chen, T., Zheng, Y., 1989. Accelerating the recovery of forest in the burned area in Great Hing’an Mountains according to the synecological characteristics. J. Northeast Forest. Univ. 17 (3), 1–10 (in Chinese).