Comparing the physico-chemistry dynamics of running waters (North-East of France) based on sequence clustering

Abstract

This article presents a clustering-based process for analyzing temporal sequences of physico-chemical measurements from French river monitoring networks. The method was originally developed for processing satellite images and adapted for river data, that are characterized by sampling variability and a great number of parameters, monitored in different ways. Dissimilarity between time sequences is computed on the basis of their chronology, not dates. The method handles sequences of various lengths. A subset of about 300 sequences from monitored sites in north-eastern France, over the 2007–2012 period, was selected and processed with the adapted method. One macro-parameter and one group of pesticides were analyzed. Results are presented and commented, through various visualization tools. The method allowed to highlight groups of sites with similar temporal dynamics and pollution level, which may be related to geographical and agricultural characteristics over the period considered.