Looking for similar patterns among monitoring stations. Venice Lagoon application

Abstract

The main purpose of this paper is to provide a classification, in terms of different trophic variables, of the sites of a water quality monitoring network located in Venice Lagoon. We apply a classification method based on functional data analysis (FDA), introduced by James and Sugar (2003), which allows to take into account sample information about the temporal dynamics of the variables of interest, as well as quantify the uncertainty in the partition. We found that classification based on FDA can provide useful insights to environmental scientists and public authorities with respect to the optimisation of monitoring networks and the identification of critical features deserving further investigation.

Time series clustering by a robust autoregressive metric with application to air pollution

“…medoid sites that are not virtual sites but real sites belonging to the set of considered sites. However, as noted by Pastres et al [33], among other this approach is not entirely satisfactory since the uncertainty in the partition is not quantified. "This issue could be relevant when deciding to redesign a monitoring network, since the location of 'uncertain' sites may be considered when establishing the boundaries between homogeneous areas" [33].…”

“…However, as noted by Pastres et al [33], among other this approach is not entirely satisfactory since the uncertainty in the partition is not quantified. "This issue could be relevant when deciding to redesign a monitoring network, since the location of 'uncertain' sites may be considered when establishing the boundaries between homogeneous areas" [33]. In order to overcome both limitations, Pastres et al [33] proposed to combine functional data analysis and probabilistic cluster analysis methods, which allow one to estimate the probability that a given object belongs to a given cluster.…”

“…However, various researchers note the importance of defying a suitable measure of uncertainty. As pointed out by Pastres et al [33] this issue could be relevant when deciding to redesign a monitoring network, since the location of 'uncertain' sites may be considered when establishing the boundaries between homogeneous areas. By adopting a fuzzy approach for clustering the air pollution time series monitored by a set of stations, our model is able to quantify the uncertainty -formalized in a fuzzy manner-in the assignment process of the monitoring stations to the clusters by means of the membership degrees.…”

“…"This issue could be relevant when deciding to redesign a monitoring network, since the location of 'uncertain' sites may be considered when establishing the boundaries between homogeneous areas" [33]. In order to overcome both limitations, Pastres et al [33] proposed to combine functional data analysis and probabilistic cluster analysis methods, which allow one to estimate the probability that a given object belongs to a given cluster. Clustering has been used to describe diurnal variation in gaseous and particle pollutants by Adame et al [63] and Flemming et al [64].…”

“…On the other hand, information collected by a redundant network, for example in the initial exploratory phase of a surveillance monitoring, could be extremely useful for partitioning a transition water body into homogeneous regions, for which different quality objectives may be established. With regard to the above issues, cluster analysis methods (or unsupervised classification) can play a very important role" [33]. In the literature, different clustering-based techniques have been proposed for analyzing air pollution.…”