Stoichiometry has been a long running theme in my research, but I have often been dissatisfied in how to graphically display information about carbon (C), nitrogen (N), and phosphorus (P) ratios. This is often done with a triplot: a three-panel graph of scatterplots of the contents of C vs. N, C vs. P, and N vs. P in insects, lakes, etc. I have certainly used this format, but it forces the viewer to integrate multiple lines of information – a task that may be particularly difficult if differences between subjects are small and spread over three different figures.
Through the LiWe project, I have collected lots of information about the C, N, and P contents in the lakes and streams of Cuatro Ciénegas. Now, I wanted a new way to look at the CNP ratios across the system (but not using the dreaded triplot). Below is my (current) solution. I began with the notion that if I scaled all CNP ratios to P = 1, then I could focus on two numbers of the ratio, C:P and N:P. For example, the Redfield Ratio is 106:16:1. If a lake’s CNP ratio is the same as Redfield, it’s two descriptor numbers would be 106 (for C:P) and 16 (for N:P). Scaling the CNP ratios for all the lakes and streams in my dataset in this way would give my two values for each lake or stream – a scenario that allows the graphing of CNP ratios on a single XY scatterplot. Yay!
But, the graph needed a few modifications: (1) Scaling the axes. Nutrient concentrations in the lakes and streams of Cuatro Ciénegas often have wide ranges – these ranges become even larger when ratios are calculated. Therefore, I used a log transformation on the scale of both axes to better visualize the data. (2) Relevant reference lines. Large divergences from the Redfield Ratio may suggest organisms living in that system are more likely to be N- or P-limited. Therefore, I added vertical and horizontal lines to the plot that represent the Redfield Ratio (CP = 106, NP = 16, respectively). (3) Information on nutrient concentrations. One advantage of the triplot is that it integrates information on both nutrient concentrations (“quantity”) and nutrient ratios (“quality”) in the same plot. I wanted to get at this information in my stoichiometry plot, as well, so with the aid of ggplot (yes, the graph was made in R), I was able to scale the size of each point to the concentration of P (in terms of total dissolved P).
Now, when I read a paper that says “The CNP stoichiometry of the subjects changed,” I will probably visualize this graph – did the subjects move from the upper right quadrant of the graph to the bottom left quadrant (suggesting a move from a phosphorus poor to phosphorus rich environment)? Did the subjects shift from a near-Redfield stoichiometry to a non-Redfield stoichiometry? This visualization packs in a lot of information; I’d love to hear ways in which you think it is useful or ways in which it could be improved to be even more useful! And, while this graphing technique is new to me, please let me know if you’ve seen it elsewhere.
One final note: You perhaps have noticed that I did not include the CN ratios. CN ratios could definitely be used on the x-axis instead of the CP ratios. But, perhaps Jim is right in saying that the CN ratios are usually pretty boring... you’ll have to look at your data and see. ;)