Geostatistical modelling for predicting soil organic carbon change – a tutorial

This tutorial explains how geostatistics can be used to predict the total Soil Organic Carbon (SOC) stock for an area, or to predict the change in total SOC stock for an area between two points in time. Geostatistical interpolation benefits from spatial correlation of the variable of interest, but it can also include information provided through correlated explanatory variables. This is usually done by incorporating a trend defined through a statistical regression model, such as a multiple linear regression or random forest machine learning model. An important strength of geostatistical modelling and prediction is that the uncertainty of the predictions is also quantified, by means of a prediction error variance or a prediction interval.

Furthermore, geostatistics has the advantage that, unlike sampling theory and design-based statistical inference, it does not require that the SOC stock observations used for modelling and prediction are a probability sample from the area of interest. It also works in cases where a convenience or purposive sample has been collected. A disadvantage is that geostatistics must assume a statistical model, meaning that its results are only valid under the model assumptions made, while sampling theory and design-based inference are completely model-free. Readers interested in sampling theory for estimating SOC stock and SOC stock change are referred to Sampling theory for estimating soil organic carbon change - a tutorial, also available in the ISRIC Resource Library.

While this tutorial makes use of geostatistics it is not a geostatistics tutorial. For this we refer to Geostatistics for soil mapping. Likewise, readers specifically interested in machine learning for digital soil mapping are referred to Machine Learning for Digital Soil Mapping.

The tutorial explains the underlying theory but the emphasis is on practical application using a concrete example and the R language for statistical computing. The tutorial provides all R scripts and datasets for the example area.

We developed this tutorial such that it should not be difficult to adapt it to other case studies, thus supporting the verification step of Monitoring, Reporting and Verification (MRV) projects.

We assume that users of the tutorial have basic knowledge of statistics, geo-information science and basic experience with R. Note that we also assume that R and required packages have been installed (version 4.4.3 or higher).

Materials

• Tutorial HTML
  
• Tutorial PDF
  
• Scripts and datasets

License: This tutorial is released under the GNU GPL v3.0 license. GNU GPL v3.0 is a strong copyleft license. This means that you may use the code and change/modify the code. If you distribute copies or modifications of the code, you are required to release these updates under the GPL v3 license.

Acknowledgements: This tutorial was developed within the MRV4SOC project, which received funding from the European Union’s Horizon Europe research programme under grant agreement n°101112754.

Disclaimer: This tutorial is provided without warranty. ISRIC is not obliged to provide updates or “bug fixes” of any kind. ISRIC will not provide user support for this tutorial.

Even though this tutorial is created with utmost care, ISRIC cannot be held liable for any damage caused by using this tutorial or any content therein in whatever form, whether or not caused by possible errors or faults nor for any consequences thereof.