Ionomics and compositional data analysis

Constancio Asis, Senior Research Agronomist, Darwin

Ionomics is the study of the ionome—the mineral component of plant and soil ecosystems. Ionomics can be used to determine changes in mineral composition in response to physiological stimuli.

Since ionomic data has units in percentages, parts per million or milligram per kilogram, it is considered compositional data. Compositional data is the quantitative description of the parts making up a whole and can be further expressed as proportions, concentrations and fluxes of properties. It is the type of data most commonly collected in ecology and agronomy. However, statistical analysis of compositional data is potentially biased, often leading to conflicting interpretations.

I attended a training course on 4-8 July at the University of Girona in Spain, to learn how to analyse compositional data using the free software ‘CoDaPack’. The training was conducted by the Research Group in Compositional Data of the University of Girona and consisted of lectures, laboratory sessions and case study presentations. There were 11 participants from Australia, Canada, China, Colombia, Denmark, Italy and Spain.

The training provided me with the theoretical and practical aspects of statistical analysis for compositional data. This is a valuable skill for analysing nutrient balances of mango trees. DPIR will be working with leading local mango growers to investigate whether this approach will improve nutrition management in local mango orchards. It may also have applications in other horticultural crops in the Northern Territory.

For more information, material and ideas about compositional data analysis, please visit the CoDaWeb website, maintained by the Research Group in Compositional Data, at http://www.compositionaldata.com/

Researchers whose interest goes from real case studies to the mathematical foundations of compositional data are welcomed to join the website forum. The CoDaPack software can be downloaded from http://ima.udg.edu/codapack/

Last updated: 19 December 2016