Time to do some work

In the course fundamentals, we covered the history and theory behind digital soil mapping. Now, with a solid foundation in R, it’s time to apply this knowledge—i.e., perform DSM with R.
The following pages explore key preparatory steps (not all strictly “mapping”) that streamline a DSM workflow:

What are you in for?

These pages present common methods for preparing and exploring soil data.

Soil point databases are inherently heterogeneous because sampling protocols vary by site. However, most records share a label and spatial coordinates for each sample.

Depth‐wise, observations also differ: some studies sample per horizon, others at fixed intervals, some only the topsoil, and others down to bedrock. Attributes measured (e.g., carbon, texture, pH) may appear in some profiles but not others.

As you work with these data, you’ll see that several preprocessing steps are needed to meet your analysis goals.

When preparing data for a DSM project (sensu Minasny & McBratney 2010), ask:

  1. What is the target attribute or class?
  2. What is the support of observations?
    • Are they point samples or aggregated over areas?
    • Over what depth interval (e.g., 0–10 cm, 0–1 m, full profile)?

You may map a single depth interval (e.g., 0–1 m) or model depth variation alongside lateral variation (3-D mapping). Planning these objectives up front guides your choice of depth functions, covariate extraction, and modelling approach.

Recent DSM research combines depth functions with spatial models to produce near-3D soil maps. The next page dives into one such method: a mass-preserving soil depth function. Then you’ll link your processed soil observations to environmental covariates and prepare those covariates for spatial modelling.

References

Minasny, B., & McBratney, A. B. (2010). Digital Soil Mapping: Bridging Research, Environmental Application, and Operation. In J. L. Boettinger et al. (Eds.), Digital Soil Mapping (pp. 429–425). Dordrecht: Springer.