Search

Forest landscape models (FLM) are widely used for simulating forest ecosystems. As FLMs have become more mechanistic, more input parameters are required, which increases model parameter uncertainty. To better understand the increased mechanistic detail provided by LANDIS-II/PnET-Succession, we studied the effects of parameter uncertainty on model outputs based on three different approaches. Global sensitivity analyses summarized the influence of each parameter, a local sensitivity analysis determined the magnitude of and degree of nonlinearity of variation in model outputs alongside variation in individual parameters, and a regression tree analysis identified hierarchical relationships among and interaction effects between parameters. Foliar nitrogen, maintenance respiration, and atmospheric carbon dioxide concentration were the most influential parameters in the global analysis. Knowing where parameter influence is concentrated will help model users interpret results from LANDIS-II/PnET-Succession to address ecological questions and should guide priorities for data acquisition.

Validation of the long-term biomass predictions of forest landscape models (FLMs) has always been a challenging task. Using the space-for-time substitution method, forest biomass curves over stand age were generated from a forest survey dataset (FSD) in the Lesser Khingan Mountains area (LKM), Northeastern China and compared with long-term biomass predictions of LANDIS-II model. The results showed that mean forest age and mean biomass of the LKM in 2000 were 51.6 years and 84.2 Mg ha−1, respectively. Significant linear correlations were found between FSD derived biomass and simulated biomass in the aggradation phase for the entire LKM and most subregions. However, a considerable difference in the mean maximum biomass (53.45 Mg ha−1) existed between from FSD and simulation during the post-aggradation phase. The space-for-time substitution method has potential in validating time series biomass predictions of FLMs in aggradation phase when only limited forest inventory data is available.

Megafires are large wildfires that occur under extreme weather conditions and produce mixed burn severities across diverse environmental gradients. Assessing megafire effects requires data covering large spatiotemporal extents, which are difficult to collect from field inventories. Remote sensing provides an alternative but is limited in revealing post-fire recovery trajectories and the underlying processes that drive the recovery. We developed a novel framework to spatially reconstruct the post-fire time-series of forest conditions after the 1987 Black Dragon fire of China by integrating a forest landscape model (LANDIS) with remote sensing and inventory data. We derived pre-fire (1985) forest composition and the megafire perimeter and severity using remote sensing and inventory data. We simulated the megafire and the post-megafire forest recovery from 1985 to 2015 using the LANDIS model. We demonstrated that the framework was effective in reconstructing the post-fire stand dynamics and that it is applicable to other types of disturbances.

Hurricanes in the southeast United States are infrequent disturbances that affect large areas and have a large effect on forest succession. In order to understand and quantify this effect, we added a new module to the LANDIS-II landscape change model. Focusing on the southeast coast of the United States, we simulated stochastic hurricanes for 50 years. For each simulated storm, the new model extension generates the maximum sustained wind speed over the region and uses the resulting parameter surface to compute maximum sustained wind speed for each cohort cell in a raster grid. Mortality is estimated for each species and age cohort in each cell based on the maximum sustained wind speed, altering forest succession. Results indicate that hurricanes reduce average aboveground biomass by > 20% over 50 years on a landscape in Fort Bragg, North Carolina (USA) compared to a scenario without hurricanes and increased uncertainty of projected succession.

Forest landscape models are effective tools for exploring the effects of long-term and large-scale landscape processes such as seed dispersal, fire, and timber harvest. These models have been widely used for about a decade, and although significant advances in theory and technology have been incorporated into their development, evaluating the veracity of simulated results from forest landscape models remains challenging. In this study, we evaluated simulated forest succession and the effects of simulated fire and harvest by a spatially explicit forest landscape model (LANDIS PRO), initialized using forest inventory data (second and third tier data from years 2000 and 2010). Our results suggested that the initialized forest landscape constructed from the year 2000 forest inventory data adequately represented the forest composition and structure from that year. The simulated density and basal area from year 2010 adequately represented the forest inventory data from that year at landscape scales. Our results indicated that the simulated fire and harvest effects were comparable to the field data (measured density and basal area). Results in this study quantified the near-term reliability and confidence of the model as well as prediction uncertainties.

Evaluation of forest landscape model (FLM) predictions is indispensable to establish the credibility of predictions. We present a framework that evaluates short- and long-term FLM predictions at site and landscape scales. Site-scale evaluation is conducted through comparing raster cell-level predictions with inventory plot data whereas landscape-scale evaluation is conducted through comparing predictions stratified by extraneous drivers with aggregated values in inventory plots. Long-term predictions are evaluated using empirical data and knowledge. We demonstrate the applicability of the framework using LANDIS PRO FLM. We showed how inventory data were used to initialize the landscape and calibrate model parameters. Evaluation of the short-term LANDIS PRO predictions based on multiple metrics showed good overall performance at site and landscape scales. The predicted long-term stand development patterns were consistent with the established theories of stand dynamics. The predicted long-term forest composition and successional trajectories conformed well to empirical old-growth studies in the region. We present a framework for evaluating the short- and long-term forest landscape model predictions at site and landscape scales.Site-scale evaluation is conducted through comparing cell-level predictions with inventory plot data.Landscape-scale evaluation is conducted through comparing predictions stratified by extraneous drivers with aggregated values in inventory plots.We successfully evaluated the LANDIS PRO forest landscape model predictions using empirical data and knowledge and showed reasonable performances at both scales.

LANDIS is a cell-based spatially explicit forest landscape model designed to explore successional dynamics under natural and anthropogenic disturbances. Species age cohort (10-year cohort of a given tree species) information is required for each cell in LANDIS. However, providing such information for a landscape comprising millions of cells is challenging. In this study, a stand-based assignation (SBA) approach was developed to stochastically assign species age cohorts to each cell based on forest inventory data. As a probability-based approach, SBA will introduce errors in LANDIS input. In order to assess the effect of errors produced by SBA on LANDIS results, 20 Monte Carlo simulations were produced. For each species simulated in LANDIS, the recurrence frequency (RF) of the majority species age cohort (MSAC, the most frequently occurring species age cohort) from 20 Monte Carlo simulations were used to quantify the uncertainty in species age cohorts for individual cell. Average recurrence frequency (ARF) of the MSAC was used to quantify the overall uncertainty in species age cohorts at the cell level. For each species, the coefficient of variation (CV) for the percent area and an aggregation index for the 20 Monte Carlo simulations was used to quantify the uncertainty at the landscape level. Results showed that at the cell level, uncertainty was relatively low at the beginning of the simulation (ARF was larger than 10). Seed dispersal, seedling establishment, mortality, and fire disturbance caused uncertainty to increase with simulation year. The uncertainty finally reached an equilibrium state, where input errors in original species age cohorts had little effect on the simulation outcomes. At the landscape level, species percent area and their spatial patterns were not substantially affected by the uncertainties in species age structure at the cell level. Since the typical use of LANDIS is to predict the long-term landscape pattern change, SBA can be used to parameterize species age cohorts for individual cells.

Over the last 20 years, we have seen a rapid development in the field of forest landscape modeling, fueled by both technological and theoretical advances. Two fundamental challenges have persisted since the inception of FLMs: (1) balancing realistic simulation of ecological processes at broad spatial and temporal scales with computing capacity, and (2) validating modeled results using independent, spatially explicit time series data. The paper discusses the current status and future directions regarding these two challenges.

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 model's 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.