Data-Driven models for groundwater level forecasting and improvement of waterresource management: example of the Foot-hill aquifer system in the Brenta riverplain (Veneto, Italy)

 In most regions worldwide, the groundwater usage is increasingly common due to the progressive decreasing
of the effective availability of surface water for both quantity regime and quality issues, as a consequence of
global population increasing, global climate change and growing water pollution. Therefore, groundwater is
the most important and safest source for water supply being less affected by pollution and climate changes.
For example, in European Union countries, groundwater provides nearly 70% of the piped water supply and
80% of the drinking water. However, the overexploitation of groundwater may sometimes exceed recharge
over long periods and over extensive areas and the subsequent decline in water table level may affect natural
groundwater discharge and quality, which in turn may have harmful impacts on groundwater dependent
streams, wetlands and ecosystems. For these reasons a correct management of the groundwater resources
is of paramount importance. In this scenario, groundwater modelling, both conceptual and numerical, is
particularly crucial for the sustainable and efficient management of groundwater resources, even more in
the context of expected climate change.
The middle-high Brenta river plain (NW of Veneto, Italy) is characterized by the existence of an important
unconfined to semi-confined foothill aquifer system, that is made of a very thick single-layer of gravellypebbly alluvial deposits, in the northernmost part, whereas in the southern part the aquifer becomes a
multilayer composed of gravelly deposits and levels of silt and clay. The existence of an important aquifer
system is tied to the high annual rainfall amount, about 1200 - 1500 mm/year, however the main
groundwater recharge component of the aquifer is related to the water dispersion from the Brenta river.
Groundwater hosted in the aquifer represents an important resource for drinkable supply, industrial and
agricultural usages. However, over the last decades, the exploitation of groundwater resource and the
meteo-climate regime caused a decrement of the piezometric level alerting the local authorities. Thanks to
the existence of a consistent and continuous monitoring network made up of several meteoric, hydrometric
and piezometric stations, very long time-series of data are available. The availability of a long time-series
allowed to develop Data-Driven models, specifically Multiple Linear Regression Models, of the Brenta river
hydrometric level (using rainfall, snowfall and atmospheric temperature as independent variables) and of the
piezometric level of groundwater in the middle-high plain (using atmospheric temperature, the local rainfall
and the model of Brenta river as independent variables). The regression models were used to make
predictions on the development of the hydrometric level of Brenta river and consequently of the
groundwater level under extreme weather and climate conditions as those of the last years, thus providing
useful information for steering the best water management practices in a zone where strategic groundwater
exploitation systems are located.