The Response Curve Rollercoaster

Last week, I focused on why models require data and wrote about some of the steps in model calibration (which we are about to start in the lab). This week, however, I want to talk more about what I have been working on in the lab, and some of the spatial components of a model. While food web modelers often use Ecopath with Ecosim (EwE) to describe how broad changes in the environment affect the interactions between organisms, we can also include a spatial component since interactions are likely limited by spatial features, such as habitat type, water flow and currents, and the presence or absence of physical features in the water (sandbars, etc.). When we model, we define spatial cells in the model where interactions between functional groups take place. When we use Ecospace--the spatial modeling portion of EwE--we can account for spatial variations in environmental conditions or species distributions/interactions, that likely provide a more informative outcome. A tradeoff, of course, is the amount of computing power required and the amount of information needed to build the model in the first place.

My current work provides some of the specifics of how organisms behave or respond in different geographic environments.  These 'response curves' describe the suitability of specific environmental conditions for all the organisms in the model, with one response curve per environmental factor per functional group. I am using field data from multiple states bordering the Gulf of Mexico to generate these response curves. And while some response curves are quite easy to populate in the model (especially if the response curve has been published by researchers modeling the Gulf of Mexico), many response curves require a combination of coding and mathematics. 

Let's take, for instance, a simple scenario using humans as an example. Humans have tolerances for certain environmental conditions, and we as a species thrive under specific oxygen levels, temperatures, air quality, etc. If we were to consider just temperature and plot the abundance of humans on Earth and the average temperature where they live, we could generate an understanding of the temperature tolerance of humans. Very few humans live at the extreme temperatures on Earth and most live somewhere around 57 F (according to NOAA). From these data, we can generate a tolerance or response curve, which describes the likelihood of encountering a human based on the temperature.

In my work, I am generating these response curves for all the functional groups in the model, and right now I am focused on three variables: water depth, water temperature, and salinity. From the field data, I am plotting the abundance of individuals collected under certain environmental conditions and creating response curves, like the one shown here. While there's nothing mathematical about this image, it provides an estimated shape of the response curve. I can then use coding (and some math) to describe the shape of this response curve, and enter the information about the shape and some key points on the curve into the Ecospace model. Then, as we run the model and environmental conditions change, these response curves will help inform which areas remain suitable for species, which areas become suitable, and which areas become unsuitable where the species should not exist. Here for instance, we don't seem to often find this species at depths > 20 meters, so if we have a scenario where water depth increases, we might not find this species in areas in which it used to thrive.

I hope you learned a little about how we can use biological data to generate mathematical relationships to inform our models. Stay tuned for next week's blog!

---