The use of multicomponent statistical analysis in hydrogeological environmental research

Abstract

The present article examines the possibilities of investigating NO₃⁻ spread in aquifers by applying multicomponent statistical methods (factor, cluster and discriminant analysis) on hydrogeological, hydrochemical, and environmental parameters.

A 4-R-Mode factor model determined from the analysis showed its useful role in investigating hydrogeological parameters affecting NO₃⁻ concentration, such as its dilution by upcoming groundwater of the recharge areas. The relationship between NO₃⁻ concentration and agricultural activities can be determined sufficiently by the first factor which relies on NO₃⁻ and SO₄²⁻ of the same origin—that of agricultural fertilizers. The other three factors of R-Mode analysis are not connected directly to the NO₃⁻ problem. They do however, by extracting the role of the unsaturated zone, show an interesting relationship between organic matter content, thickness and saturated hydraulic conductivity.

The application of Hirerarchical Cluster Analysis, based on all possible combinations of classification method, showed two main groups of samples. The first group comprises samples from the edges and the second from the central part of the study area. By the application of Discriminant Analysis it was shown that NO₃⁻ and SO₄²⁻ ions are the most significant variables in the discriminant function. Therefore, the first group is considered to comprise all samples from areas not influenced by fertilizers lying on the edges of contaminating activities such as crop cultivation, while the second comprises all the other samples.

Introduction

Multicomponent statistical analysis is a mathematical technique applied frequently to many problems in geological investigation, such as geochemistry [1], petrography [2], geological engineering [3], and environmental geology [4], [5].

To compare the different chemical parameters of the water analyses and determine the relationships between them, factor analyses were applied to a large number of water quality data samples. This statistical method has several advantages over classical graphical approaches as it takes into consideration more data such as neutral chemical species (e.g. SiO2) and non-chemical data (e.g. temperature) [6], as well as other parameters, such as the distance of a pollutant point from an aquifer. R-Mode factor analysis enables relations among variables, while cluster analysis enables relations among samples to be interpreted in terms of simpler relations that provide an insight into the underlying structure of the data-set [7]. The Discriminant analysis tries to explain why samples should be distinguished in groups according to their characteristics.

In shallow groundwater, increased nitrate concentrations are due to the extensive application of agricultural fertilizers, which are the most common human source of NO₃⁻ in groundwater systems [8]. The extent of nitrate leaching from agricultural land is influenced strongly by factors inherent in nature such as soil type and climatic conditions [9]. In groundwater and pore water that is strongly oxidizing, NO³⁻ (the most stable form of dissolved nitrogen) is transported with the groundwater and experiences no chemical transformation and little, or no, retardation [10].

The aim of this paper is to present a simple method based on multicomponent statistical analysis, which could be useful in examining the origin and transport of nitrate ions through the non-saturated and saturated zones of an aquifer. In other words using statistical analysis, we can draw easier conclusions about the factors affecting nitrate transport. Among the most important multicomponent statistical analyses used in this study, Factor analysis (R-mode), Cluster analysis, and Discriminant analysis are presented and proposed as the most efficient.