Combining multivariate statistics and analysis of variance to redesign a water quality monitoring network

Abstract

The objective of this paper was to demonstrate how multivariate statistics combined with the analysis of variance could support decision-making during the process of redesigning a water quality monitoring network with highly heterogeneous datasets in terms of time and space. Principal Component Analysis (PCA) and Hierarchical Cluster Analysis (HCA) were selected to optimise the selection of water quality parameters to be monitored as well as the number and location of monitoring stations. Sampling frequency was specifically investigated through the analysis of variance. The data used were obtained between 2007 and 2010 at the Long-term Environmental Research Monitoring and Testing System (OPE) located in the north-eastern part of France in relation with a geological disposal of radioactive waste project. PCA results showed that no substantial reduction among the parameters was possible as strong correlation only exists between electrical conductivity, calcium or bicarbonates. HCA results were geospatially represented for each field campaign and compared to one another in terms of similarities and differences allowing us to group the monitoring stations into 12 categories. This approach enabled us to take into account not only the spatial variability of water quality but also its temporal variability. Finally, the analysis of variances showed that three very different behaviours occurred: parameters with high temporal variability and low spatial variability (e.g. suspended matter), parameters with high spatial variability and average temporal variability (e.g. calcium) and finally parameters with both high temporal and spatial variability (e.g. nitrate).

A Proposal for Redesigning the Water Quality Network of the Tunjuelo River in Bogotá, Colombia through a Spatio-Temporal Analysis

“…It is common to observe a relationship between the first [45,47,48] or second [49] principal components and organic matter variables (COD and BOD) and an inverse relationship with DO in PCA analyses of basins contaminated by urban processes. However, as Gigues et al [50] and Pejman et al [51] demonstrate, different behaviors are found in rural basins. Fats and oils (OG), dissolved oxygen (DO), and pH exhibited low variability throughout the river and did not significantly contribute to either of the first two principal components.…”

Determining Multivariate Analysis Sampling Frequency for Monitoring Contamination Caused by Trout Farms

“…Likewise, Mohamed et al (2015) used these methods to optimize a water quality monitoring network in the basin of the Klang River and were very satisfied with the obtained results [30]. Guigues et al (2013) integrated ANOVA and multivariate statistical methods to redesign a water quality monitoring network [31]. They concluded that this technique was very applicable in understanding the complex nature of water quality issues and determining the priorities to improve water quality.…”

Temporal variability of the optimal monitoring setup assessed using information theory

“…Regarding the water management, whose influence during the period of investigation appeared to be rather low (except for the pumping stations at gauge Q025), a clear change (e.g., the activation of a further pumping station) would cause the optimal network to be reevaluated [cf. Guigues et al ., ]. However, to define the point in time when the change becomes severe would be hard to predict.…”