Compositional Data Analysis

Volume 187 of the series Springer Proceedings in Mathematics & Statistics pp 31-43


Towards the Concept of Background/baseline Compositions: A Practicable Path?

  • A. BucciantiAffiliated withDepartment of Earth Sciences, University of Florence Email author 
  • , B. NisiAffiliated withCNR-IGG (Institute of Geosciences and Earth Resources)
  • , B. RacoAffiliated withCNR-IGG (Institute of Geosciences and Earth Resources)

* Final gross prices may vary according to local VAT.

Get Access


Water geochemistry is often investigated considering a large number of variables, including major, minor and trace elements. Some of these are usually well associated due to coherent geochemical behaviour, but the effect of anthropic factors tends to increase data variability, sometimes obscuring the natural laws governing their relationships. It may thus be difficult to identify geochemical features linked to natural phenomena, as well as to separate geogenic anomalies from the anthropogenic ones, or to define background or baseline concentrations for single chemical elements. This is particularly true at regional level, where numerous phenomena may interact and mix together, forming a complex pattern not easy to interpret. The identification of background or baseline values is particularly difficult due to the compositional nature of chemical variables, so that under the Compositional Data Analysis (CoDA) theory single background or baseline values lose their meaning. However, they are fundamental references for public institutions and government policies. In this contribution a new approach is proposed, aimed at investigating the regionalised structure of the geochemical data by considering the joint behaviour of several chemical elements. The approach is based on the robust CoDA theory, so that the proportionality features of abundance data are fully taken into account, enhancing their relative multivariate behaviour, as well as the influence of outliers. An application example is presented for the groundwater compositions in Tuscany Region, a surface of about 23,000 km\(^2\), where more than 6000 wells have been sampled and analysed. The mapping of robust Mahalanobis distance was able to indicate (1) in which part of the investigated area the pressure toward anomalous behaviour was higher, (2) where the compositions nearest to the barycentre were and (3) if spatial continuity was present in limited portions of the territory.


Water chemistry Compositional data analysis Baseline Fractals Dissipative structures