Skip to main content

A large-scale forest landscape model incorporating multi-scale processes and utilizing forest inventory data


Wen Wang

Hong He

Martin Spetich

Stephen Shifley

Frank Thompson

David Larsen

Jacob Fraser

Jian Yang

+3 more
Publication Type:
Journal Article
Year of Publication:
Secondary Title:


Two challenges confronting forest landscape models (FLMs) are how to simulate fine, stand-scale processes while making large-scale (i.e., >107 ha) simulation possible, and how to take advantage of extensive forest inventory data such as U.S. Forest Inventory and Analysis (FIA) data to initialize and constrain model parameters. We present the LANDIS PRO model that addresses these needs. LANDIS PRO adds density and size mechanisms of resource competition. This is achieved through incorporating number of trees and DBH by species age cohort within each raster cell. Forest change is determined by the interactions of species-, stand-, and landscape-scale processes. Species-scale processes include tree growth, establishment, and mortality. Stand-scale processes include density and size-related resource competition that regulates self-thinning and seedling establishment. Landscape-scale processes include seed dispersal, as well as natural and anthropogenic disturbances. LANDIS PRO is designed to be straightforwardly comparable with forest inventory data, and thus the extensive FIA data can be directly utilized to initialize and constrain model parameters before predicting future forest change. We initialized a large landscape (107 ha) from historical FIA data (1978) and the predicted forest structure and composition following 30 years of simulation were statistically calibrated against a prior time-series of sequential FIA data (1978 to 2008). The results showed that the initialized conditions realistically represented the historical forest composition and structure at 1978, and the constrained model parameters predicted reasonable outcomes at both landscape and land type scales. The subsequent evaluation of model predictions showed that the predicted forest composition and structure were comparable with old-growth oak forests; predicted forest successional trajectories were consistent with the expected successional patterns in oak-dominated forests in the study region; and the predicted stand development patterns were in agreement with the established theories of forest stand development. This study demonstrated a framework for forest landscape modeling including model initialization, calibration, and evaluation of predictions.