Hong S. He

SPATIALLY EXPLICIT AND STOCHASTIC SIMULATION OF FOREST- LANDSCAPE FIRE DISTURBANCE AND SUCCESSION

Understanding disturbance and recovery of forest landscapes is a challenge because of complex interactions over a range of temporal and spatial scales. Landscape simulation models offer an approach to studying such systems at broad scales. Fire can be simulated spatially using mechanistic or stochastic approaches. We describe the fire module in a spatially explicit, stochastic model of forest landscape dynamics (LANDIS) that in- corporates fire, windthrow, and harvest disturbance with species-level succession. A sto- chastic approach is suited to forest landscape models that are designed to simulate patterns over large spatial and time domains and are not used deterministically to predict individual events. We used the model to examine how disturbance regimes and species dynamics interact across a large (500 000 ha), heterogeneous landscape in northern Wisconsin, USA, with six land types having different species environments, and fire disturbance return intervals varying from 200 to 1000 yr. The model shows that there are feedbacks over time between species, disturbance, and environment, resulting in the re-emergence of patterns that char- acterized the landscape before extensive alteration. Landscape equilibrium of species com- position and age-class structure develops at three scales from the initial, disturbed landscape. Over 100-150 yr, fine-grained successional processes cause gradual disaggregation of the initial pattern of relatively homogenous composition and age classes. Species such as eastern hemlock ( Tsuga canadensis), largely removed from the landscape by past human activities, only slowly re-invade. Next, patterns on the various land types diverge, driven by different disturbance regimes and dominant species. Finally, aging of the landscape causes the prob- abilities of larger and more severe fires to increase, and a coarse-grained pattern develops from the disturbance patches. Influence of adjacent land types is shown as fires spread across land type boundaries, although modified in spread and severity. As found by others, altered landscapes are likely to retain their modified pattern for centuries, suggesting that nonequilibrium conditions between tree species and climate will persist under predicted rates of climate change. The results suggest that this modeling approach can be useful in examining species- level, broad-scale responses of heterogeneous landscapes to changes in landscape distur- bance, such as modified management or land-use scenarios, or effects of global change.

An aggregation index (AI) to quantify spatial patterns of landscapes.

There is often need to measure aggregation levels of spatial patterns within a single map class in landscape ecological studies. The contagion index (CI), shape index (SI), and probability of adjacency of the same class (Qi), all have certain limits when measuring aggregation of spatial patterns. We have developed an aggregation index (AI) that is class specific and independent of landscape composition. AI 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). AI is similar to SI and Qi, but it calculates aggregation more precisely than the latter two. We have evaluated the performance of AI under varied levels of (1) aggregation, (2) number of patches, (3) spatial resolutions, and (4) real species distribution maps at various spatial scales. AI was able to produce reasonable results under all these circumstances. Since it is class specific, it is more precise than CI, which measures overall landscape aggregation. Thus, AI provides a quantitative basis to correlate the spatial pattern of a class with a specific process. Since AI is a ratio variable, map units do not affect the calculation. It can be compared between classes from the same or different landscapes, or even the same classes from the same landscape under different resolutions.

Linking an ecosystem model and a landscape model to study forest species response to climate warming

No single model can address forest change from single tree to regional scales. We discuss a framework linking an ecosystem process model {LINKAGES) with a spatial landscape model (LANDIS) to examine forest species responses to climate warming for a large, heterogeneous landscape in northern Wisconsin, USA. Individual species response at the ecosystem scale was simulated with LINKAGES, which integrates soil, climate and species data, stratified by ecoregions. Individual species biomass results, simulated by LINKAGES at year 10, were quantified using an empirical equation as species establishment coefficients (0.0-l.0). These coefficients were used to parameterize LANDIS, thus integrating ecosystem dynamics with large-scale landscape processes such as seed dispersal and fire disturbance. Species response to climate warming at the landscape scale was simulated with LANDIS. LANDIS was parameterized with information derived from a species level, forest classification map, and inventory data. This incorporates spatially-explicit seed source distributions. A standard LANDIS run with natural fire disturbance regime and current climate was conducted for 400 years. To simulate the effects of climate change, the differences in species establishment coefficients from current and warmer climates were linearly interpolated over the first 100 years assuming climate warming will occur gradually over the next century. The model was then run for another 300 years to examine the consequences after warming. Across the landscape, the decline of boreal species and the increases of temperate species were observed in the simulation. The responses of northern temperate hardwood species vary among ecoregions depending on soil nutrient and water regimes. Simulation results indicate that boreal species disappear from the landscape in 200-300 years and approximately same amount of time for a southern species to become common. Warming can accelerate the re-colonization process for current species such as found for eastern hemlock, where moisture does not become limiting. However, the re-colonization is strongly affected by available seed source explicitly described on the landscape. These phenomena cannot be simulated with most gap models, which assume a random seed rain.

The Effects of Seed Dispersal on the Simulation of Long-Term Forest Landscape Change

ABSTRACT The study of forest landscape change requires an understanding of the complex interactions of both spatial and temporal factors. Traditionally, forest gap models have been used to simulate change on small and independent plots. While gap models are useful in examining forest ecological dynamics across temporal scales, large, spatial processes, such as seed dispersal, cannot be realistically simulated across large landscapes. To simulate seed dispersal, spatially explicit landscape models that track individual species distribution are needed. We used such a model, LANDIS, to illustrate the implications of seed dispersal for simulating forest landscape change. On an artificial open landscape with a uniform environment, circular-shaped tree species establishment patterns resulted from the simulations, with areas near seed sources more densely covered than areas further from seed sources. Because LANDIS simulates at 10-y time steps, this pattern reflects an integration of various possible dispersal shapes and establishment that are caused by the annual variations in climate and other environmental variables. On real landscapes, these patterns driven only by species dispersal radii are obscured by other factors, such as species competition, disturbance, and landscape structure. To further demonstrate the effects of seed dispersal, we chose a fairly disturbed and fragmented forest landscape (approximately 500,000 ha) in northern Wisconsin. We compared the simulation results of a map with tree species (seed source locations) realistically parameterized (the real scenario) against a randomly parameterized species map (the random scenario). Differences in the initial seed source distribution lead to different simulation results of species abundance with species abundance starting at identical levels under the two scenarios. This is particularly true for the first half of the model run (0–250 y). Under the random scenario, infrequently occurring and shade tolerant species tend to be overestimated, while midabundant and midshade tolerant species tend to be underestimated. The over- and underestimation of species abundance diminish when examining long-term (500 y) landscape dynamics, because stochastic factors, such as fire, tend to make the landscapes under both scenarios converge. However, differences in spatial patterns, and especially species age-cohort distributions, can persist under the two scenarios for several hundred years.

An object-oriented forest landscape model and its representation of tree species

LANDIS is a forest landscape model that simulates the interaction of large landscape processes and forest successional dynamics at tree species level. We discuss how object-oriented design (OOD) approaches such as modularity, abstraction and encapsulation are integrated into the design of LANDIS. We show that using OOD approaches, model decisions (olden as model assumptions) can be made at three levels parallel to our understanding of ecological processes. These decisions can be updated with relative efficiency because OOD components are less interdependent than those designed with traditional approaches. To further examine object design, we examined how forest species objects, AGELIST (tree age-classes), SPECIE (single species) and SPECIES (species list), are designed, linked and functioned. We also discuss in detail the data structure of AGELIST and show that different data structures can significantly affect model performance and model application scopes. Following the discussion of forest species objects, we apply the model to a real forest landscape in northern Wisconsin. We demonstrate the models capability of tracking species age cohorts in a spatially explicit manner at each time step. The use of these models at large spatial and temporal scales reveals important information that is essential for the management of forested ecosystem.

Simulating the effects of different fire regimes on plant functional groups in Southern California

A spatially explicit landscape model of disturbance and vegetation succession, LANDIS, was used to examine the effect of fire regime on landscape patterns of functional group dominance in the shrublands and forests of the southern California foothills and mountains. Three model treatments, frequent (35 year), moderate (70 year), and infrequent (1050 year) fire cycles, were applied to the landscape for 500 year. The model was calibrated and tested using a dataset representing an initial random distribution of six plant functional groups on an even-aged landscape. Calibration of the three fire regime treatments resulted in simulation of fire cycles within 7% of these intended values when fire cycles were averaged across ten replicated model runs per treatment. Within individual 500-year model runs, the error in the simulated fire cycle (average area burned per decade) reached 11% for the moderate and frequent fire cycle treatments and 53% for infrequent. The infrequent fire regime resulted in an old landscape dominated by the three most shade tolerant and long-lived functional groups, while shorter-lived and less shade tolerant seeders and resprouters disappeared from the landscape. The moderate fire regime, similar to what is considered the current fire regime in the southern California foothills, resulted in a younger landscape where the facultative resprouter persisted along with the long-lived shade tolerant functional groups, but the obligate seeder with low fire tolerance disappeared, despite its moderate shade tolerance. The frequent fire regime resulted in the persistence of all functional groups on the landscape with more even cover, but the same rank order as under the moderate regime. The model, originally developed for northern temperate forests, appears to be useful for simulating the disturbance regime in this fire-prone Mediterranean-type ecosystem.

The adequacy of different landscape metrics for various landscape patterns

The behavior of several landscape pattern metrics were tested against various pattern scenarios generated by neutral landscape models, including number of classes, scale-map extent, scale-resolution, class proportion, aggregation level-RULE, and aggregation level-SimMap. The results demonstrate that most of the metrics are sensitive to certain pattern scenarios, yet are not sensitive to others; therefore, none of them is appropriate for all aspects of a landscape pattern. Despite these limitations, some of these metrics are recommended for future use, which include total number of patches, average patch size, total edge density, double-logged fractal, contagion, and aggregation index. Special attention should be paid to the relationships between metric values and ecological processes rather than the numbers themselves. es.

Study of landscape change under forest harvesting and climate warming-induced fire disturbance

We examined tree species responses under forest harvesting and an increased fire disturbance scenario due to climate warming in northern Wisconsin where northern hardwood and boreal forests are currently predominant. Individual species response at the ecosystem scale was simulated with a gap model, which integrates soil, climate and species data, stratified by ecoregions. Such responses were quantified as species establishment coefficients. These coefficients were used to parameterize a spatially explicit landscape model, LANDIS. Species response to climate warming at the landscape scale was simulated with LANDIS, which integrates ecosystem dynamics with spatial processes including seed dispersal, fire disturbance, and forest harvesting. Under a 5 degrees C annual temperature increase predicted by global climate models (GCM), our simulation results suggest that significant change in species composition and abundance could occur in the two ecoregions in the study area. In the glacial lake plain (lakeshore) ecoregion under warming conditions, boreal and northern hardwood species such as red oak, sugar maple, white pine, balsam fir, paper birch, yellow birch, and aspen decline gradually during and after climate warming. Southern species such as white ash, hickory, bur oak, black oak, and white oak, which are present in minor amounts before the warming, increase in abundance on the landscape. The transition of the northern hardwood and boreal forest to one dominated by southern species occurs around year 200. In the sand barrens ecoregion under warming conditions, red pine initially benefits from the decline of other northern hardwood species, and its abundance quickly increases. However, red pine and jack pine as well as new southern species are unable to reproduce, and the ecoregion could transform into a region with only grass and shrub species around 250 years under warming climate. Increased fire frequency can accelerate the decline of shadetolerant species such as balsam fir and sugar maple and accelerate the northward migration of southern species. Forest harvesting accelerated the decline of northern hardwood and boreal tree species. This is especially obvious on the barrens ecoregion, where the intensive cutting regime contributed to the decline of red pine and jack pine already under stressed environments. Forest managers may instead consider a conservative cutting plan or protective management scenarios with limited forest harvesting. This could prolong the transformation of the barrens into prairie from one-half to one tree life cycle.

INTEGRATION OF GIS DATA AND CLASSIFIED SATELLITE IMAGERY FOR REGIONAL FOREST ASSESSMENT

New methods are needed to derive detailed spatial environmental data for large areas, with the increasing interest in landscape ecology and ecosystem management at large scales. We describe a method that integrates several data sources for assessing forest composition across large, heterogeneous landscapes. Multitemporal Landsat Thematic Mapper (TM) satellite data can yield forest classifications with spatially detailed information down to the dominant canopy species level in temperate deciduous and mixed forests. We stratified a large region (106 ha) by ecoregions (103–104 ha). Within each ecoregion, plot-level, field inventory data were aggregated to provide information on secondary and sub-canopy tree species occurrence, and tree age class distributions. We derived a probabilistic algorithm to assign information from a point coverage (forest inventory sampling points) and a polygon coverage (ecoregion boundaries) to a raster map (satellite land cover classification). The method was applied to a region in northern Wisconsin, USA. The satellite map captures the occurrence and the patch structure of canopy dominants. The inventory data provide important secondary information on age class and associated species not available with current canopy remote sensing. In this way we derived new maps of tree species distribution and stand age reflecting differences at the ecoregion scale. These maps can be used in assessing forest patterns across regional landscapes, and as input data in models to examine forest landscape change over time. As an example, we discuss the distribution of eastern white pine (Pinus strobus) as an associated species and its potential for restoration in our study region. Our method partially fills a current information gap at the landscape scale. However, its applicability is also limited to this scale.

A simulation study of landscape scale forest succession in northeastern China

Abstract Changbai Natural Reserve in northeastern China provides an excellent opportunity to explore how temperate and boreal forest ecosystems in northeastern China will evolve and recover over large spatial and temporal scales. Such studies are increasingly needed to design scientifically sound forest management and restoration plans in this region. Long-term (300 years) successional trajectories of the dominant tree species are simulated on the heterogeneous, undisturbed area (within the reserve) using a spatially explicit landscape model. We also examine the spatial and temporal constrains of landscape recovery on the human disturbed areas (surrounding the reserve). Simulation results suggest that an equilibrium in landscape structure and composition is approached on the large landtypes dominated by shade tolerant species, but not on landtypes altered by humans. Such equilibrium can be observed in spruce-fir, mountain birch, and larch forests and not in aspen-birch forests. Our results suggest that direct and indirect human impact may produce long-term alterations to forest landscape patch structure that persist for decades to centuries. For example, even with complete natural succession over 300 years, Korean pine only recovers on 1/3 of the areas in the landtypes it can dominate. We estimate a full recovery would take another 200–300 years without human disturbance. Our results also indicate that landscape-scale recovery is often limited by the available seed sources and this is particularly true for Korean pines in this region. Comparison of simulation results for the entire study area with land types (two scales) reveals the greatest variations at the land type scale. This discrepancy indicates that the ‘space-for-time’ substitutions can be flawed as species composition and age class at a given site and time may represent only the specific successional history of that site. This is particularly true for human disturbed forest landscapes where higher variations are observed.