Designing a forest landscape model that simulates density-based stand dynamics at regional scales

Background/Question/Methods

Niche- and process-based models are the primary tools for projecting climate impacts on forests and guiding mitigation efforts.  Both approaches lack the capability to include forest landscape processes (FLPs), such as forest management, disturbance, and invasive species.  However, these processes can be stronger drivers of forest change than the direct effects of climate.  Without incorporating FLPs, predictions of vegetation dynamics may significantly underestimate the magnitude of change at regional scales. Forest landscape models (FLMs) are explicitly designed to simulate FLPs in a spatially interactive manner. To date, they have contributed relatively little to forecasting regional vegetation dynamics because they have been unable to simulate sufficiently large landscapes that are comparable to the size of niche or process model. We present an innovative approach implemented in LANDIS PRO, a variant from LANDIS FLM family. LANDIS PRO simulates forest succession and dynamics by incorporating species-, stand-, and landscape-scale processes. Within each raster cell, the model records number of trees by species age cohort, and size (e.g., DBH) for each age cohort is derived from empirical age-DBH relationships. The model derives key stand parameters such as basal area, stocking, importance value, biomass and carbon from tree density and size information for each species. 

Results/Conclusions

We parameterized LANDIS PRO for a large Central Hardwood Region that contains a continuous forested area of over 100 million hectares, including portions of Arkansas, Missouri, Illinois, Indiana, Kentucky, and Ohio. LANDIS PRO simulations were conducted at 90 m resolution with 17000 (col) x 7000 (row) pixels, over 80 land types (landforms within each ecological subsections), and 16 major tree species. We showed that LANDIS PRO could be directly initialized from nearly a million FIA plots. We also showed that the model could be calibrated against FIA data to constrain the initial landscape and model parameters so that the starting landscape approximated the reality closely.  Finally, we showed how model predictions could be evaluated using stand management diagrams (Gingrich stocking charts and Reineke density diagrams), which were based on the basic theories of density-dependent behavior of populations including competition, site occupancy, and self-thinning. These diagrams are considered as the best available approach to evaluate the long-term predicted stand development trajectories. Since LANDIS PRO simulates interactions of fine-scale forest succession with FLPs at regional scales, it is possible to evaluate the effects of succession and FLPs on forest change predictions at regional scales, which are made by niche and process models.