|
General Theory and Methods |
1 Dipartimento di Scienze della Terra, Università di Firenze, Via G. La Pira 4, I-50125, Firenze, Italy antonella.buccianti{at}unifi.it
2 Department Informàtica i Matemàtica Aplicada, Universitat de Girona, Campus Montilivi, P4, E-17071 Girona, Spain
The chemical composition of natural waters is derived from many different sources of solutes, including gases and aerosols from the atmosphere, weathering and erosion of rocks and soil, solution or precipitation reactions occurring below the land surface, and effects resulting from human activities. The chemical composition of the crustal rocks of the Earth, as well as the composition of the ocean and the atmosphere, are important in evaluating sources of solutes. Data used in the investigation of natural and non-natural contributions are obtained usually from chemical analysis of water samples, which may be statistically evaluated with the aim of summarizing the contained information. However, as these data are compositional and thus constrained to move in the simplex, application of usual statistical methodologies may lead to incorrect evaluations and/or interpretations. This paper focuses on how to draw information on natural processes by modelling univariate and multivariate frequency distributions using water data. The chemical composition of 977 samples collected in wells from Vulcano island (Italy) are used as a case study. The methodological approach can be transferred to the investigation of other geochemical or constrained data.
