Predicting Changes in Forest Composition and Dynamics – Landscape-scale Process Models

Synthesis

Synthesis: 

Preparers

Eric Gustafson, Institute for Applied Ecosystem Studies, Northern Research Station; Robert Keane, Missoula Fire Sciences Laboratory, Rocky Mountain Research Station

Issues

Natural resource managers need predictions of how climate change will affect landscape-level tree community composition and ecosystem function.  Ecologically based process models are critical tools for exploring vegetation dynamics under climate change. 

There are four general categories of process models distinguished primarily by the scale and detail at which they operate: forest gap models, ecosystem models, forest landscape models and dynamic global vegetation models (DGVMs). 

Forest gap models were developed to simulate the effects of ecophysiological drivers, such as water, nutrient, and light availability, on the rates of establishment, growth, and mortality among competing species of trees within a relatively homogenous forest stand (1, 2).  Early gap models were not spatially explicit, but some later gap models simulated spatial interactions among trees at fine scales (3: SORTIE, 4: FM) and the effects of specific disturbances (5: FORSKA; 6: FORCLIM).  Most gap models simulate process for a plot or stand so they cannot incorporate effects of exogenous disturbance regimes and spatial ecological processes such as dispersal, hydrology, and topography.

Ecosystem process models are similar to forest gap models in that they simulate the effects of biogeochemical process (e.g., fluxes of energy and mass) on ecological dynamics (e.g., forest growth rates, carbon accumulation, decomposition), but they emphasize biogeochemical and water dynamics for coarsely defined vegetation types or life forms, rather than individual trees or species (7).  Ecosystem process models have been applied at both fine and broad-scales, typically using land-cover datasets from remotely-sensed imagery, with each pixel representing a site.  However, similar to gap models spatially-explicit interactions among landscape-scale processes are rarely simulated in ecosystem models (8).

Forest Landscape Models (FLMs), the main subject of this page, are used to simulate forest generative processes (establishment, development, aging) and degenerative processes (disturbance, senescence) over appropriate spatial (landscape) and temporal (centuries) scales.  They explicitly model spatial processes such as seed dispersal and disturbance spread, and account for interactions in both time and space.  FLMs include stochastic or ‘random’ elements that cannot easily be predicted, such as sudden disturbances.  They are used to study how abiotic environmental factors (including changing climate), disturbances and management activities interact to affect forest dynamics such as tree species and age composition, spatial pattern and other ecosystem attributes.

FLMs are primarily intended to simulate forest disturbance and successional processes, as well as their interactions, across broad spatial and temporal scales.  FLMs also generally provide spatially continuous projections of disturbance and vegetation dynamics that are valuable for determining key drivers of landscape-level forest composition or structure or disturbance behavior.  Within this framework, the diverse FLM family of models can be further classified based on ability to simulate multiple processes or operate at fine temporal resolutions (9), or whether community change is static or dynamic, with the former determined by a priori successional stages and the latter by the life history attributes, behavior (e.g., seed dispersal), and physiological requirements of individual species (8).  Some FLMs directly or indirectly incorporate the influence of biogeochemical process on forest productivity and can be coupled with gap or ecosystem process models to derive inputs representing climate effects on species establishment probabilities or productivity (e.g., 10 using PnET-II).  Unlike DGVMs (discussed below), FLMs do not incorporate feedback loops with climate as simulated by General Circulation Models (GCMs), and applications are currently not feasible at continental to global scales. 

DGVMs are similar to terrestrial biogeochemical models, but simulate competition among vegetation types (e.g., biomes) and are directly coupled to GCMs, allowing feedbacks to climate at regional to global scales (11).  DGVMs have the potential to simulate climate change effects on tree establishment and mortality via mechanistic plant responses to biogeochemical and hydrological dynamics (e.g., 12: SEIB–DGVM), but, because DGVMs operate at very broad scales, many processes are highly simplified and some are omitted.  For example, DGVMs are unable to simulate important species- and plant-level ecological responses to disturbances.

Forest Landscape Models

FLMS operate at a scale that is highly relevant to strategic forest management planning and decision-making.  They are useful to describe the landscape and long-term context for more tactical planning tools such as FVS, by integrating global drivers (e.g. climate) and landscape disturbance processes with potential management actions.

FLMs are designed using either a mechanistic or a phenomenological approach, or more often, a combination.

Mechanistic models (sometimes called process-based) take a reductionist approach by explicitly representing the mechanisms that lead from cause to effect.  For example, tree growth, is simulated using the ecophysiological processes of photosynthesis and respiration.

Phenomenological models (sometimes called empirical or statistical) take a more holistic approach by using information about how a system has typically behaved in the past to predict how causes produce effects (phenomena).  That is, the outcome of the process is predicted using surrogates for the mechanism.  Tree growth in this case might be predicted as a change in diameter increment calculated from data about how trees have grown under various climate, shade, and nutrient conditions. The use of the term “empirical” for this type of model is somewhat confusing because mechanistic models often estimate mechanistic functions using empirical data.

Given a rapidly changing world, it may be increasingly difficult to rely on analysis of past conditions to predict the future because the relationships used by the models may no longer hold.  Key questions include:

  • Will successional pathways (sequence, timing and likelihood) remain the same as the climate changes?  Will the species that comprise successional communities change? 
  • How will rapidly changing factors such as CO2-fertilization, increased temperatures, reduced rainfall, physiological changes in water use efficiency of trees and changes in cloudiness and solar radiation interact to affect tree species growth rates and species competition? 
  • How will disturbance regimes (rotation intervals, event size, severity and frequency) change?  How will the complex interactions among disturbances (fire, pests, grazing, harvests) and with climate, vegetation, fuels, and topography dictate future landscape conditions?

New landscape behaviors and conditions may not be predictable if we only look at how systems behaved in the past.  For these reasons, forest landscape modelers are gradually moving toward a more mechanistic approach to simulate landscape dynamics under changing climate conditions.  Mechanistic modeling comes at a cost, though.  Mechanistic models are difficult to build and run because quantitative physical relationships may not be available or parameterized for many ecological processes; model uncertainty and instability usually increases with model complexity; and operation of the model usually requires more computing resources, higher levels of expertise, and greater amounts of input data.

Modeling Tree Community Changes

There are approximately 50-150 FLMs that have been published in the peer-reviewed literature and each one was designed for specific ecosystems or simulation objectives (for reviews see 13; 14; 15; 8; 9; 16).  We present two models that the authors have used in their studies.

LANDIS-II is a grid-cell FLM that simulates forest development processes of establishment, growth, competition and the forest degenerative processes of senescence and disturbances such as fire, wind, insect outbreaks and timber harvesting (17).  The model tracks living and dead biomass within cohorts of species on each cell.  LANDIS-II encapsulates distinct ecological or physical processes that act on the biomass of cohorts within cells on the landscape, interacting with one another to produce overarching forest dynamics across the landscape.

FireBGCv2 is a mechanistic, individual-tree succession model containing stochastic properties implemented on grid cells at landscape scales (18).  Tree growth, organic matter decomposition, litterfall, and many other ecological processes are simulated using detailed physical biogeochemical relationships.  Tree establishment and mortality are modeled probabilistically using empirically estimated probabilities.  FireBGCv2 also includes a spatial simulation of fire, where fire spreads across the landscape based on slope and wind information from a fire start ignited based on fuel moisture and loading.  Fire effects are simulated by computing fire intensity, then estimating fuel consumption, tree mortality, and soil heating using algorithms in the embedded FOFEM model (19).  Daily weather inputs drive primary canopy processes (e.g., transpiration, photosynthesis and respiration) and are also used to ignite and spread fire across the landscape.  Carbon allocation to the stem of a tree is used to calculate diameter and height growth.  Material from trees (fallen needles, leaves, and branches) are added to the fuelbed and eventually decompose based on available water, nitrogen, and light.  Cone crops are stochastically generated and subsequent seed dissemination is spatially modeled across the landscape using empirical dispersal relationships representing wind and bird dispersal.

The following examples offer a glimpse into the ability of landscape models to project how climate change can interact with many ecosystem processes and properties to profoundly impact forested ecosystems.

LANDIS-II: Gustafson et al. (20) simulated scenarios of climate change and harvest of virgin forests in Siberia (Russia).  They found that the direct effects of climate change on forest composition and landscape pattern was relatively modest, but that indirect effects, specifically the arrival of a cold-limited defoliating insect, dramatically changed composition and pattern.  This in turn affected the fire regime.  A strength of the modeling approach was that the interactions among the novel conditions (climate, insects, harvesting) and existing disturbances (fire, wind) emerged from the simulations because each process independently operated on the ecosystem vegetation in a mechanistic way.

FireBGCv2: Keane et al. (18) found that under warming climates, the pattern and composition of landscapes in the US northern Rockies will result from a complex net of interactions between disturbance, vegetation, fuel, genetics, and spatial scale.  The set of disturbances and ecological processes included in the model dictate subsequent landscape behavior.  Wildland fire, for example, will create significantly different landscapes under warming climates depending on whether they are suppressed or allowed to burn.  Mountain pine beetle and the exotic white pine blister rust will also influence future landscapes as they interact with fire, fuels, and vegetation (21)

Management Options

Landscape process models are complex, but managers can work with researchers to achieve robust and defensible results.  A collaborative, iterative approach involving researchers, management decision makers and local resource experts is a promising model for using landscape models to inform management decisions (see 22).

FLMs are simplifications of reality, so results may be somewhat inaccurate but highly precise.  Model results should not be interpreted as specific predictions, but as expected behavior.  Because of this it is most useful to compare scenarios or management alternatives because the magnitude of differences between scenarios is more reliable than the specific values of predictions.  FLMs are great tools to conduct “what-if” landscape experiments.  Like ecological systems, FLM’s have a random component.  Therefore replicated model runs give a sense of the range and variation that can occur for the modeled ecosystem dynamics.  The value of FLM results greatly depends on the quality of the input data; accurate model parameters and initial conditions will reduce the uncertainty in simulation results.

Although FLMs can be a valuable tool for guiding management decisions, certain precautions are helpful in interpreting and using model results.

  • FLMs are best viewed as general learning or insight tools rather than specific predictive tools.  Although they can make increasingly robust predictions, uncertainty is probably higher than managers would like.  Scenarios and “what-if” experiments should be designed to maximize insight and learning rather than to simply try to predict an uncertain future.
  • Extrapolating results simulated for one landscape to others is not recommended.  Simulation results may be landscape-specific because each landscape has a unique set of biophysical characteristics.  While general trends may be the same, the specifics of modeled outcomes may be significantly different.

Despite the limitations of FLMs, they are extremely powerful tools to integrate a great number of complex and interacting factors, account for spatial processes, make projections at landscape spatial and temporal scales and produce quantitative and comparable projections of many ecosystem characteristics of interest to managers.  They have tremendous potential to provide critical information that can improve strategic management decisions in the face of multiple global changes.

Publication date: 
Tue, 04/15/2014
References: 
  1. Botkin, D.B.; Janak, J.F.; Wallis, J.R. 1972. Ecological consequences of a computer model of forest growth. Journal of Ecology. 60:849-872.
  2. Shugart, H.H. 1984. A theory of forest dynamics: the ecological implications of forest succession models. Springer-Verlag, New York.
  3. Pacala, S.W.; Canham, C.D.; Silander, J.A.J. 1993. Forest models defined by field measurements: I. The design of a northeastern forest simulator. Canadian Journal of Forest Research. 23:1980-1988.
  4. Miller, C.; Urban, D.L. 2000. Modeling the effects of fire management alternatives on Sierra Nevada mixed-conifer forests. Ecological Applications. 10:85-94.
  5. 5. Prentice, C.I.; Sykes, M.T.; Cramer, W. 1993. A simulation model for the transient effects of climate change on forest landscapes. Ecolological Modeling. 65:51-70.
  6. Bugmann, H.; Cramer, W. 1998. Improving the behaviour of forest gap models along drought gradients. Forest Ecology and Management. 103:247-263.
  7. Cushman, S.A.; McKenzie, D.; Peterson, D.L.; Littell, J.; McKelvey, K.S. 2007. Research agenda for integrated landscape modeling. RMRS-GTR-194. USDA Forest Service, Rocky Mountain Research Station, Fort Collins, CO.
  8. Scheller, R.M.; Mladenoff, D.J. 2007. An ecological classification of forest landscape simulation models: tools and strategies for understanding broad-scale forested ecosystems.  Landscape Ecology. 22:491-505.
  9. He, H.S.; Keane, R.E.; Iverson, L.R.  2008.  Forest landscape models, definition, characterization, and classification. [pdf]  Forest Ecology and Management. 254:484-498.
  10. Xu, C.; Gertner, G.Z.; Scheller, R.M. 2009. Uncertainty in the response of a forest landscape to global climatic change.  Global Change Biology. 15:116–13.
  11. Medlyn, B.E.; Duursma, R.A.; Zeppel, M.J.B. 2011. Forest productivity under climate change: a checklist for evaluating model studies. Wiley Interdisciplinary Reviews: Climate Change. 2:332-355.
  12. Sato, H.; Itoh, A.; Kohyama, T. 2007. SEIB–DGVM: A new Dynamic Global Vegetation Model using a spatially explicit individual-based approach. Ecological Modelling. 200:279-307.
  13. Gardner, R.H.; Romme, W.H.; Turner, M.G. 1999.  Predicting forest fire effects at landscape scales. In: Mladenoff, D.J.; Baker, W.L. (eds.). Spatial modeling of forest landscape change: approaches and applications. Cambridge University Press, Cambridge, United Kingdom, pp 163-185.
  14. Keane, R.E.; Cary, G.; Davies, I.D.; Flannigan, M.D.; Gardner, R.H.; Ra, L.F.; Lenihan, J.M.; Chao, L.H.  2004.  A classification of landscape fire succession models: spatially explicit models of fire and vegetation dynamics. Ecological Modelling. 256:3-27
  15. Perry, G.L.W.; Enright, N.J. 2006. Spatial modelling of vegetation change in dynamic landscapes: a review of methods and applications. Progress in Physical Geography. 30:47-72.
  16. Baker, W.L. 1989. A review of models of landscape change. Landscape Ecology. 2:111-133
  17. Scheller, R.M.; Mladenoff, D.J. 2004.  A forest growth and biomass module for a landscape simulation model, LANDIS: design, validation, and application.  Ecological Modelling. 180:211-229
  18. Keane, R.E.; Loehman, R.A.; Holsinger, L.M. 2011. The FireBGCv2 landscape fire and succession model: a research simulation platform for exploring fire and vegetation dynamics.  USDA Forest Service Gen. Tech. Rep. RMRS-255.  Rocky Mountain Research Station, Fort Collins, Colorado, USA.
  19. Reinhardt, E.; Keane, R.E. 1998. FOFEM - a First Order Fire Effects Model. Fire Management Notes. 58(2), 25-28.
  20. Gustafson, E.J.; Shvidenko, A.Z.; Sturtevant, B.R.; Scheller, R.M. 2010. Predicting global change effects on forest biomass and composition in south-central Siberia. Ecological Applications. 20:700-715.
  21. Loehman, R.A.; Clark, J.A.; Keane, R.E. 2011. Modeling Effects of Climate Change and Fire Management on Western White Pine (Pinus monticola) in the Northern Rocky Mountains, USA. Forests. 2(4):832-860
  22. Gustafson, E.J.; Sturtevant, B.R.; Fall, A. 2006. A collaborative, iterative approach to transfer modeling technology to land managers.  In Perera, A.H.; Buse, L.; Crow, T.R. (eds.). Forest landscape ecology: Transferring knowledge to practice.  Cambridge Press, London, UK.  pp 43-64.
How to cite: 

Gustafson, E.; Keene, R. (April, 2014). Predicting Changes in Forest Composition and Dynamics – Landscape-scale Process Models. U.S. Department of Agriculture, Forest Service, Climate Change Resource Center. www.fs.usda.gov/ccrc/topics/process-models

Reading

Recommended Reading: 

Canham, C.D.; Cole, J.J.; Lauenroth, W.K. 2004. Models in Ecosystem Science. Princeton University Press, Princeton, New Jersey, USA.

Gustafson, E.J. 2013. When the past can’t be used to predict the future: using mechanistic models to predict landscape ecological dynamics in a changing world.  Landscape Ecology. 28(8):1429-1437.

Gustafson, E.J.; Sturtevant, B.R.; Fall, A.  2006.  A collaborative, iterative approach to transfer modeling technology to land managers.  In Perera, A.H.; Buse, L.; Crow, T.R. (eds).  Forest landscape ecology: Transferring knowledge to practice.  Cambridge Press, London, UK.  p. 43-64.

Green, D.G.; Sadedin, S. 2005.  Interactions matter—complexity in landscapes and ecosystems. Ecological Complexity 2:117-130

LeBauer, D.S.; Wang, D.; Richter, K.T.; Davidson, C.C.; Dietze, M.C. 2013. Facilitating feedbacks between field measurements and ecosystem models. Ecological Monographs. 83:133-154

Scheller, R.M.; Mladenoff, D.J. 2007. An ecological classification of forest landscape simulation models: tools and strategies for understanding broad-scale forested ecosystems.  Landscape Ecology. 22:491-505.

Seidl, R.; Fernandes, P.M.; Fonseca, T.F. et al. (2011)  Modelling natural disturbances in forest ecosystems: a review. Ecological Modelling 222:903-924

Shifley, S.R.; Thompson, F.R.; Dijak, W.D.; Fan, Z. (2008)  Forecasting landscape-scale, cumulative effects of forest management on vegetation and wildlife habitat: A case study of issues, limitations, and opportunities.  Forest Ecology and Management. 254:474-483

Links