Evaluating the Role of Visualization

Each time step of the SORTIE model involves interaction among several thousand trees and seedlings. In a very real sense, the model system is comprised of literally millions of interactions in each time step. We have demonstrated that large-scale behavior of the model is mediated by these myriad local interactions and is sensitive to both changes in the environment and the specific nature of each species' strategies. In this work, our understanding of "why" was largely dependent on our ability to visualize these interactions. As an example, we present a pair of visualizations from this work.

A strong analogy can be made with exploratory data analysis (EDA). The EDA method is predicated on the idea that the human eye and brain are incredibly good at pattern recognition (Tukey 1977; Tufte 1983; Cleveland 1985). Thus, graphical representations of complex data are a fundamental part of any statistical analysis. This idea is a powerful one, with implications for any work with complex models, in ecology as well as other disciplines.

Light
Dispersal

The interaction between the canopy trees and light availability at the forest floor.

From these two panels you can quickly ascertain the patchy distribution of beech, hemlock, and birch; that the forest canopy remains closed even under frequent disturbance; the differences in shade cast by the three species; the ubiquitous nature of yellow birch seedlings; and in direct contrast, the dispersal limitation of beech seedlings. None of these insights are obvious from the statistical summaries of the forest like average tree density. If the system is complex enough to require individual, spatially explicit interactions, then the model should be evaluated only after visualizing these same interactions.