A1082
Title: Estimating graphical models varying on a spatial network for water quality assessment
Authors: Rosaria Ignaccolo - University of Turin (Italy) [presenting]
Nicola Pronello - University of Turin (Italy)
Alex Cucco - Imperial College London (United Kingdom)
Vito Frontuto - University of Turin (Italy)
Natalia Golini - University of Turin (Italy)
Luigi Ippoliti - University G.d'Annunzio Chieti-Pescara (Italy)
Abstract: The purpose is to illustrate how graphical models that vary across a spatial network provide an effective way to accurately depict the conditional dependence relationships among water pollutants and other features observed over a fluvial network. Once all graphs - one for each site - have been estimated, their analysis enables tracking changes in the ecological and chemical status of water bodies along the fluvial network. By considering distances among graphs, it is possible to cluster spatial sites and identify potentially critical zones. Additionally, graph analytics can be used to detect spatially varying clusters of variables throughout the fluvial network. The motivating case study focuses on assessing the quality of water bodies in the Italian region of Piedmont.
