LANDIS PRO is a raster-based forest landscape model (FLM) that evolved over 15 years of development and applications of the LANDIS model (Mladenoff et al. 1996, He and Mladenoff 1999a, Mladenoff 2004). Within each raster cell, the model records number of trees by species age cohort (Fig. 1), and size (e.g., DBH) of each age cohort, which is derived from empirical age-DBH relationships. LANDIS PRO incorporates species-, stand-, and landscape-scale processes (Fig. 1). Species- and stand-scale processes are simulated within each cell, and landscape-scale processes are simulated across the whole landscape, in addition these three scales interact with each other.
Figure 1. The conceptual design of LANDIS PRO.
Species-scale processes are regulated by stand-scale resource competition and FLPs. Stand-scale resources are occupied by species-scale seedling establishment, and released by disturbance and longevity caused mortality. Species-scale processes include tree growth, seedling establishment, stem resprouting, and mortality. These are simulated using species’ vital attributes (Mladenoff and He 1999) and empirical growth equations such as age-DBH relationships (Lowenstein et al. 2000).
Stand-scale processes include resources competition (e.g. light, water, and nutrients), which controls competition-caused mortality and seedling establishment. The competition intensity is quantified by the amount of growing space occupied (GSO) (Oliver and Larson 1996) within each cell. We implement competition-caused mortality using Yoda’s self-thinning (Yoda et al. 1963), and seedling establishment is determined by available growing space, species shade tolerance, and species establishment probability. Resource availability varies among different stand development stages due to the dynamics of establishment and mortality. The simulation of stand development patterns is governed by GSO and follows well documented stages of stand development.
Landscape-scale processes simulated in LANDIS PRO include seed dispersal (exotic species invasion), fire, wind, insect and disease spread, forest harvesting, fuel treatments and silvicultural treatments. These disturbance processes release growing space on one or more stands on the landscape, thereby, disturbance often resets the development stage of affected stands (Oliver and Larson 1996).
We stratified the landscape into relatively homogeneous units called landtypes to reflect environmental heterogeneity caused by factors such as topography, soil type, and land use (Fig. 1). Species establishment probabilities (SEP) and maximum growing space occupied (MGSO) are set for each landtype; SEP is the same parameter used in previous LANDIS models (Mladenoff and He 1999), LANDIS II (Scheller and Mladenoff 2004), and LANDCLIM (Schumacher et al. 2004). The SEP is a number from 0.0-1.0, reflecting how different environmental conditions favor a particular species in terms of its seedling establishment (He and Mladenoff 1999b). The second source of heterogeneity among landtypes is landscape disturbance and management.
LANDIS PRO simulates forest landscape change over large spatial (~ 108 ha) and temporal (~103 years) extents with flexible spatial (10-500 m pixel size) and temporal resolutions (1-10 years).
Key stand parameters such as species composition, basal area, density, stocking, and importance value can be derived for each species from the simulated density and size information. Biomass and carbon for individual species or for the cell can also be calculated using published empirical equations (Jenkins et al. 2003, Woodall et al. 2010). Taken together, these features of model input and output make model parameters and simulation results congruent to forest inventory data, thereby, current intensive forest inventory data can be directly applied to model initialization and to constrain model parameters.
In LANDIS PRO, heterogeneity of vegetation, disturbance, and management activities is modeled at multiple hierarchical levels from the landscape to the pixel. For vegetation heterogeneity, LANDIS stratifies the heterogeneous landscape into landtypes (also called ecoregions for broad-scale studies). Landtypes capture the highest level (coarse grain) of spatial heterogeneity caused by various environmental controls. Within a landtype, a somewhat uniform suite of ecological conditions that results in similar species establishment patterns is assumed, but the stochastic processes such as seed dispersal can result in intermediate level (fine grain, within land type) heterogeneity of a species distribution. Finally, succession, competition, and probabilistic establishment may result in heterogeneity of forest composition (density and size). Disturbance heterogeneity refers to various regimes a disturbance may cause on the simulated landscape.
LANDIS PRO is a stochastic model that uses random number generators to simulate the stochastic processes of seed dispersal, seedling establishment, disturbance, and management events. Therefore, LANDIS PRO predicts the statistical properties of landscape composition, age structure, and spatial pattern under particular disturbance and management regimes, but it does not accurately predict individual disturbance events or optimize management actions.
Compared to the previous versions of LANDIS models, number of trees is the only integer added to the original LANDIS model data structure, which requires one-fourth of the memory overhead compared to adding a biomass variable (Schumacher et al. 2004, Pennanen and Kuuluvainen 2002, Scheller et al. 2007). LANDIS PRO uses a sorted linked list to store number of trees occurring by species age cohort in sequential order (e.g. sorted by value or name) (Fig. 2), thereby, this data structure enhances memory efficiency and making large-scale simulation possible (Yang et al. 2011).
Figure 2. LANDIS PRO (sorted linked list) data structure for the representation species age-cohorts and number of individual trees in each cell.
LANDIS PRO 7.0 uses a component-based approach to conduct simulation, which breaks the monolithic program into multiple small, stand-alone, and functionally more specific components (He et al. 2002).
Figure 3. LANDIS PRO 7.0 uses a component-based approach to conduct simulation.
A component in a model is like a mini-model; it comes packaged as a binary code that is compiled, linked, and ready to perform certain tasks for the entire model. Components connect with each other at run time to form a complete model. With a component-based model, it is possible to replace some components while maintaining the integrity of the model. In LANDIS PRO 7.0, each component (module) simulates a particular process and collectively they simulate forest landscape change under natural and anthropogenic disturbances (Fig. 3).
LANDIS PRO is a 32- or 64-bit computer model that runs on Windows XP and Windows 7. 32-bit simulations are limited by Windows to accessing 2 GB of memory. The recommended memory amount for 64-bit simulations is 10 GB or more.