A recent project, which is still in the very early stages, is looking for changes in patterns in the surface ocean due to el Nino. El Nino is a decadal mode of variability in the atmospheric forcings on the Pacific Ocean leading to anomalously warmer ocean temperatures throughout the eastern Pacific. This water, like a warm blanket, alters the internal dynamics of the water (i.e. hydrodynamics) and the ecosystem (i.e. ecodynamics).
This project is looking for what some of these ecosystem changes are and how they may (or may not) impact the sinking of carbon within the water column. The removal of organic carbon from the surface ocean, a process termed export, has important ramifications on atmospheric carbon dioxide levels as well as climate change more generally.
While mechanistic theories abound, my work will be looking at an extensive dataset collected over the past 10 years off the coast of California. First we will dissect the overall patterns in the area–how does our particular ecosystem work overall? Then we can try to parse apart how these relationships change during the advent of el Nino and what that may mean for climate, especially if el Nino is expected to be more prevalent in future years.
Preliminary Ideas
Here are two correlation matrices displayed as color-coded plots (blue indicates positive correlation while magenta is negative correlation). My initial idea is to use these correlation matrices as virtual fingerprints for the two states of the ecosystem: non-el Nino and el Nino. On the left we have a picture of the sort of relationships (and the strength of the relationships) during cool, non-el Nino years while on the right we have the same plot with the warm, el Nino like years.
Between the two, we see patterns of similarity as well as unique features just as we would expect. The next step in this assessment would be the inclusion of uncertainty within the fingerprint since many of the data points used for the correlation have measurement uncertainty associated with them (e.g. Net Primary Production is typically +/- 10 percent).
I know how I would include the uncertainty, but for now I’ll hold off from describing it until I have a working example.