Content uploaded by Marco Rotiroti
Author content
All content in this area was uploaded by Marco Rotiroti on Jun 22, 2017
Content may be subject to copyright.
3rd National Meeting on Hydrogeology
Cagliari, 14-16 June 2017
39
HYDROCHEMICAL CHARACTERIZATION OF GROUNDWATER
AND SURFACE WATER SUPPORTED BY MULTIVARIATE
STATISTICAL ANALYSIS: A CASE STUDY IN THE PO PLAIN (N
ITALY)
Marco ROTIROTI 1, Tullia BONOMI 1, Letizia FUMAGALLI 1, Chiara ZANOTTI 1,
Sara TAVIANI 1, Gennaro A. STEFANIA 1, Martina PATELLI 1, Veronica NAVA 1,
Valentina SOLER 1, Elisa SACCHI 2, Barbara LEONI 1
1 Department of Earth and Environmental Sciences, University of Milano-Bicocca, Piazza della Scienza 1-
20126 Milano, Italy, marco.rotiroti@unimib.it
2 Department of Earth and Environmental Sciences, University of Pavia, Via Ferrata 1-27100 Pavia, Italy
Multivariate statistical analysis is a useful method for supporting the interpretation of
experimental data, particularly in the case of large datasets. In the present study, cluster analysis
(CA) and factor analysis (FA) are used to support the hydrochemical characterization of
groundwater and surface water in an area located in the Po Plain (N Italy), highly impacted by
human activities related to agriculture.
The study area is located in the Oglio River basin, between the outflow from Lake Iseo and the
confluence into Mella River, and covers ~1900 km2. The northern part of the study area (higher
plain) hosts a mono-layer aquifer mainly composed of sands and gravels, whereas the southern
part of the area (lower plain) hosts a multi-layer aquifer constituted by a vertical alternation of
sands with silty clays; the transition between higher and lower plain is marked by the so called
“spring belt”.
During a field survey performed in fall 2015, 58 groundwater, 20 river (Oglio River and its
main tributaries), 1 Lake Iseo and 7 spring samples were collected for chemical analysis.
Physico-chemical parameters, major ions, trace elements and water isotope were measured. The
CA was performed on total 86 samples and 18 variables; data were autoscaled. The Ward
hierarchical method, based on squared euclidean distance, was used. The FA was done using
82 samples (4 outliers were excluded) using the Kaiser criterion to select significant factors.
Results of multivariate statistical analysis were combined with the geomorphological and
hydrogeological knowledge of the study area in order to give a hydrogeological explanation of
each data cluster.
Results led to the identification of 5 main clusters: (1) higher plain groundwater and springs,
characterized by an oxidized hydrofacies with higher NO3, (2) lower plain groundwater,
characterized by a reduced hydrofacies with higher As, Fe and Mn, (3) Oglio River, (4) Oglio
River tributaries and (5) outliers. Within the cluster of higher plain groundwater, three
subgroups can be identified: (a) samples with the highest NO3 and a more enriched isotopic
signature attributable to recharge by local precipitation, (b) samples located around the spring
belt and characterized by intermediate NO3 concentrations (average ~50 mg/L) and (c) samples
located around the Oglio River and characterized by lower NO3. Also within the cluster of
lower plain groundwater, three subgroups can be identified: (a) samples with more reduced
IAH Italian Chapter
University of Cagliari – Department of Chemical and Geological Sciences
3rd National Meeting on Hydrogeology
Cagliari, 14-16 June 2017
40
states, (b) samples with earlier reduced states, likely due to some interactions with surface
waters and (c) samples with the highest As concentrations.
The Oglio River cluster can be subdivided into 3 subgroups: (a) the river stretch with losing
behavior, (b) the river stretch with draining behavior and (c) groundwater and spring directly
fed by Oglio River water.
In conclusion, this work confirms how multivariate statistical analysis can sustain the
interpretation of large hydrological datasets in order to support a hydrochemical
characterization. The latter will bear the development of the hydrogeological conceptual model
of the area, also oriented to groundwater/surface water interactions, that, in turn, will support
the numerical flow modeling of the system.
Acknowledgements
This work was supported by Fondazione Cariplo, grant 2014-1282.
IAH Italian Chapter
University of Cagliari – Department of Chemical and Geological Sciences
