Bestelmeyer et al. 2011

Post date: Jul 26, 2012 7:14:04 PM

You can find a great review paper about bringing spatial phenomena into state-and-transition models here (opens a pdf in a new window).

The authors allude to some of the problems that I encountered when trying to make a spatially-explicit state-and-transition model. (This class of model is a type of Markov Chain. Maybe someday I'll be so snooty that I only refer to them as Markov Chains. Today is not that day.) Some of these problems include:

• Spatially aggregated averages may never occur in a heterogeneous landscape

• At different spatial grains, different processes dominate. For example, soils may dominate at an intermediate spatial grain, while climate dominates at a coarse spatial grain

• Contagion is important in real ecological systems. Think of seed dispersal, or wildfire spread, for example. These can be impossible to include in a classically non-spatial STM

Regarding the issue of contagion: I agree that this is hugely important. That's why when I build my model, I designed to work in a GIS environment, so that it could work in conjunction with a mechanistic model of fire spread and effects. As for seed dispersal, pathogen spread, etc., well, I've skipped that for now. But it should be incorporated... at some point. The authors refer to such additions to STM as "elaborations". I like that term.

The authors also suggest that since previous land use strongly affects transitions and stable states, alternative STM submodels may be necessary within a landscape. I definitely saw this problem in my tree growth study in the Willamette Valley. I think their suggestion might be an elegant solution. On the other hand, it could lead to tremendously complex, and possibly uninterpretable, models.

I found this paper easy to read, and it provided good justifications for why my modeling work is important.