Blog‎ > ‎

Challenges to Forest Landscape Modeling Under Climate Change: What Are Our Objectives?

posted Jan 16, 2013, 5:52 PM by   [ updated Dec 27, 2013, 3:22 PM ]
Any modeling endeavor – be it projecting climate change effects on forests or hurricane forecasting – must begin with both explicit and implicit objectives.  How well we define our objectives will determine our success.

When modeling forest response to climate change, explicit objectives typically include descriptive narratives (How will community composition change if the climate changes by 5°C?) and reflect the information needs of forest managers (How should forest managers alter their harvesting practices if climate changes?).  Such objectives are results or ‘product’ driven.  Explicit objectives should more frequently include testable hypotheses.  Although becoming more common, testable hypotheses remain the exception.  Hypotheses could address expected shifts in community composition (e.g., the emergence of novel communities), the relative importance of different drivers (e.g., comparing the effects of climate to disturbance), or the relative importance of different processes (e.g., mortality vs. establishment, colonization vs. competition).  

Implicit objectives are those that go unsaid (and maybe should be).  Implicit objectives may include demonstrating the applicability of a model at a given location, utilization of existing or available data, and utilization of existing models and technology.  Implicit objectives may also include adherence to a spatial, temporal, or taxonomic resolution that has been used successfully by the researcher, group, or organization.  Publishing and achieving tenure are not uncommon implicit objectives.

In general, a higher ratio of explicit to implicit objectives will result in a more efficient enterprise.
Our objectives will necessarily guide us towards a compromise among the accuracy, realism, and generality that can be achieved by any particular model (Levins 1966).  
Within forest modeling, realism is often equated with fine-scale mechanistic detail.  Such structural realism may be superficially desirable, but it often requires a high parameterization cost (‘realism requires more reality’ as my advisor David Mladenoff always said).  Extensive parameterization limits spatial extent and would necessitate the exclusion of critical processes that operate at larger spatial scales, thereby self-limiting ‘realism’. Hence, determining the appropriate level of process complexity and the degree of structural "realism" in a forest landscape model is a non-trivial task and therefore should be an explicit objective.

Accuracy is often an implicit objective, but we need to balance what’s possible against what’s meaningful and interesting. Statistical models may be highly accurate in their predictions, yet they do not readily allow for insights regarding the processes that led to the observed behavior.  For example, we could accurately model how forests have changed in the past using statistics without advancing our understanding. However, when we attempt to project future dynamics of landscapes, statistical models fail because they were not built upon the underlying processes. Simulation modeling of these processes is, however, often much more uncertain as compared to statistical models. Hence, the accuracy of process-oriented models for assessing climate change effects will often be lower than that of statistical models.  Another compromise is between local and global accuracy, which tends to separate scientists from managers.  Management typically seeks high local accuracy, whereas landscape ecologists tend towards a more global view and seek patterns that emerge across many locations. We should acknowledge our definitions of accuracy (become more explicit).

Finally, generality within forest models has historically been regarded as a compromise – we often implicitly assumed that more general is better.  But, again, if we explicitly examine this objective (generality), we may find it unnecessary.  Ultimately, the questions must drive the choice of scale and subsequently model structure. Do we seek generality for generality’s sake?  Unless there are explicit objectives or hypotheses that require generality (a cross-landscape comparison, for example), local accuracy with limited generality is fine.  On the other hand, if an objective is to save money by recycling a more general model, that’s fine, too, but be explicit about it.

The more explicitly we frame our objectives when modeling climate change effects, the more likely we are to arrive at the appropriate levels of realism, accuracy, and generality.

- Robert Scheller

Levins, R. "The Strategy of Model Building in Population Biology", American Scientist, 54:421-431, 1966