Anthony's Science Blog

How big should I make my soil cores?

March 10, 2008 – 4:29 pm

While perhaps not a profound question, “How big should I make my soil cores?” is an important question that I think needs some attention.

From a conceptual standpoint, we know that any study of the environment is scale-dependent, meaning that the scale you choose for your investigation can affect the results and conclusions of your study. Or perhaps more accurately, we can say that results and conclusions need to be qualified by the scales at which they are relevant.

All sampling-based studies have a “universe of interest” (aka “extent” or study area), and a “window size” (aka “grain size”). There’s a good summary in this paper (JSTOR, requires subscription). The idea of window size is exemplified by the size of a soil core. People, including me, tend to pick this soil core size based on convenience. Most people collect a core that’s manageable to dig up. Another common technique is to use composites of several soil cores to reduce heterogeneity.

Either of these might be fine, but how does our choice of the soil core size affect our landscape-level estimates of soil and ecosystem functions? How much heterogeneity is reduced when particular composites are made? Perhaps I just don’t know where the papers on this topic are located, but I don’t think there are good quantitative investigations of this issue. I’d like to see some solid theory here since I believe that it will continue to remain both impossible and unethical to measure all of the soil in a study site. Though perhaps we can cut some kind of deal with the mountaintop removal miners.

I think the way to figure this out is to get out there and measure soil variables using different window sizes but within the same landscape. It will also be important to quantify how difficult each soil window size sample is to implement. For example, washing-machine-sized soil cores may be difficult to manage, but perhaps there is a way to do it using homogenization and subsampling of the mixture. With a good sampling regime, you could then compare estimates of variable means, variances, variogram parameters and all sorts of other stuff.

On one level, this kind of research is really boring. It’s unsatisfying because it is research on how to do research and thus is not immediately applicable to anything anyone cares about. But when you’re out there trying to figure out how to quantify global change and ecosystem processes, this kind of thing sure would be useful. Also, trying to figure out something that is really fundamental to how the Earth is structured could yield other interesting and unexpected results.

Like I said above, I could be missing the great work already done on this topic and if I discover such work, I’ll update and post it here (though I predict there will still be important missing pieces even so). Share if you know some literature on the matter or want to refine my thinking on the topic.

Update (3/25/2008): I was thinking this morning that in this post I was confusing two different concepts when I was talking about window size. Window size can still be the size of the soil cores, but another important factor is how close together the samples are, as in the average distance between them or the grain size of a grid. In a remote sensing image, these things are the same, which is why I think they get confused, but they are two different and important sampling parameters that need to be taken into account for soil samples. The distance between samples is talked about a lot but the size of the samples is not. There is a brief section on it in the book Model Based Geostatistics by Diggle and Ribeiro in a section called “regularisation” but I don’t know any other sources of the top of my head that discuss it. Again, we probably need more info on the effects this has on our numbers overall. Whew! Just had to get that off my chest.