Why Do Models Require Data?

Last week, I wrote about some of the basics of ecological modeling. Modelers use mathematical equations to describe the relationships between organisms and the environment, and the equations become more complicated or more abundant the more interactions we try to represent in our models. Most of these equations don't write themselves, so how do we create the formulas that describe these relationships?

Food web modeling in Ecopath with Ecosim, which is what I work on, is based on two master equations. These two equations describe the production of each model group--often called functional group--and the consumption of each group. These equations ensure energy balance in a group and in the ecosystem. To populate the values in these two equations, modelers need data, often from gut content analyses, field surveys and observations. The gut content analyses data provide information on the diet of each functional group; for every functional group in a model, the modeler needs to provide data on how much of every other functional group is consumed, which can get to be quite cumbersome depending on the model size. Additionally, modelers can find other information for these master equations published in scientific literature.

Modelers add information regarding the environmental conditions to their model to help build out their models. The Ecosim component refers to building and running model simulations to evaluate how the interactions shape ecosystem dynamics and how these relations might change stock (the amount of a given functional group) in the future. For these simulations and runs, modelers need data to help calibrate their models. Model calibration is the process of using data to fine-tune the model parameters and make the model more accurate. By using data from field research, modelers can evaluate the predictive success of the model. Then, modelers can create additional simulations, knowing that the calibration process provided an accurate/successful run, and evaluate the future condition of the system they are studying. 

In my postdoc project, I am currently working on putting together information about how the organisms respond to environmental conditions. We want to learn how changes in these environmental conditions will change the health of this system, so this important building block step will inform the success of organisms under specific conditions in the model runs. This portion of my work, the species response curve portion, will be the focus of next week's blog, so stay tuned to learn about where we get this information, why sometimes trapezoids are better than hills, and why sometimes you enter 1000 data points for one species and only two for another.