Volume 56, Issue 9 e2020WR027848
Research Article
Free Access

Managed Versus Natural Recharge of Pre-Alpine Phreatic Aquifers

Pietro Teatini

Corresponding Author

Pietro Teatini

Department of Civil, Environmental and Architectural Engineering, University of Padova, Padova, Italy

Correspondence to:

P. Teatini,

[email protected]

Search for more papers by this author
Grazia Martelli

Grazia Martelli

Polytechnic Department of Engineering and Architecture, University of Udine, Udine, Italy

Search for more papers by this author
Andrea Comerlati

Andrea Comerlati

Department of Civil, Environmental and Architectural Engineering, University of Padova, Padova, Italy

C+P Engineers, San Pietro in Cariano (VR), Italy

Search for more papers by this author
Giovanni Paiero

Giovanni Paiero

Polytechnic Department of Engineering and Architecture, University of Udine, Udine, Italy

Search for more papers by this author
Claudia Zoccarato

Claudia Zoccarato

Department of Civil, Environmental and Architectural Engineering, University of Padova, Padova, Italy

Search for more papers by this author
First published: 29 August 2020
Citations: 2

Abstract

Managed aquifer recharge (MAR) is becoming a common practice worldwide. MAR is carried out in different environments from coastlands to highlands, in megalopolis, farmlands and pristine areas, and in arid and humid regions. Pre-Alpine aquifers represent an optimal target when MAR is aimed at storing large amounts of high-quality waters. In fact, pre-Alpine aquifers are generally characterized by high permeability and a thick unsaturated zone, with the catchments crossed by watercourses rich of high-quality water. Here, we focus the attention to a representative pre-Alpine aquifer system located in the Friuli region, northeastern Italy. A 1-year long MAR test was carried out through a ~700 m2 infiltration basin recharged by water diverted from a nearby channel. The site was characterized from the hydrogeological viewpoint, and the MAR test was monitored through time-lapse hydrogeophysics, water level and piezometric records, and physicochemical water characterization. The data set was used to calibrate a local groundwater flow model, showing that MAR recharged the 50 m deep phreatic aquifer with 1,000 m3/day. Hydrogeologic data made available by previous studies were processed to develop a groundwater model of the regional aquifer that allowed for estimating the natural groundwater recharge of the phreatic system and, subsequently, evaluating the MAR effects in the context of the natural balance. If a single MAR site, like the tested one, plays a certain effect at a local scale only, the MAR implementation on several gravel pits and large-diameter wells scattered in the region could store several million cubic meters of water per year, significantly raising the water table and improving the groundwater quality.

Key Points

  • TL-ERT, piezometric measurements, and chemical analyses integrated by an FE groundwater flow model capture MAR effects in pre-Alpine aquifers
  • MAR is effective in pre-Alpine aquifers due to their high permeability, thick unsaturated zone, and availability of surplus surface waters
  • MAR implementation represents a promising method to offset the decrease of water resources due to climate changes in the pre-Alpine region

1 Introduction

Managed aquifer recharge (MAR) is a promising methodology that is increasingly used worldwide for different purposes, such as to reduce vulnerability to climate changes and hydrological variability (Scanlon et al., 2016), increase groundwater availability (e.g., Brothers & Katzer, 1990), mitigate land subsidence (e.g., Hoffmann et al., 2001; Zhang et al., 2015), control seawater intrusion (e.g., Abarca et al., 2006; Herndon & Markus, 2014), improve water quality through infiltration (e.g., Behroozmand et al., 2017; Schwarz et al., 2016), and reduce evaporation loss (Dillon & Arshad, 2016). According to Dillon et al. (2019), implementation of MAR has increased at a rate of 5% per year since the 1960s.

Clogging (e.g., Martin, 2013; Pavelic et al., 2007), geochemistry of aquifer recharge (e.g., Grischek & Paufler, 2017), fate of pathogens, organics, and pesticides (e.g., Weiss et al., 2005), and groundwater quality changes are the main issues that must be investigated and managed during MAR planning and operation, especially when MAR is aimed at aquifer storage and recovery for potable uses. These aspects are generally regulated by specific legislation and requirements ruled at national level (Dillon et al., 2019). Rodríguez-Escales et al. (2018) developed a methodological approach and a specific tool (MAR-RISKAPP) to evaluate the risk of failure of MAR facilities, taking into account both technical and nontechnical factors.

MAR has been effectively and largely implemented in coastal areas (e.g., Zuurbier et al., 2014), arid regions (e.g., Schwarz et al., 2016), and in the surroundings of large cities (e.g., Zhang et al., 2015). In this contribution, we focus on the role of MAR in the context of pre-Alpine regions. Despite the general large availability of high-quality surface and subsurface water, the Alpine and pre-Alpine regions are experiencing large environmental transformations connected to climate changes with more intense warming compared with the global average trend and to higher frequency of extreme hydrological events, such as droughts and intensive rainfall (ARPA FVG, 2018; Kiese et al., 2018). Brunetti et al. (2006) show that during the 20th century, the south-eastern part of the Alpine and pre-Alpine region has been subject to a significant drying, mainly caused by pronounced negative trends in autumn. Climate changes superpose to a widespread increase of human pressure on the pre-Alpine environment, yielding a general reduction in groundwater resources (Haas & Birk, 2019). The analyses of the piezometric levels have shown a decreasing trend (Calligaris et al., 2016), still sustainable today but possibly no longer manageable in the next decades considering the temperature increase and precipitation decrease predicted by climate models by the end of the century (Gobiet et al., 2014).

In view of these trends, MAR may represent an increasingly important strategy for effective water management in pre-Alpine environments. Pre-Alpine aquifers are usually characterized by a large depth (tens of meters) to the water table and highly permeable gravel soils (Martelli & Granati, 2007; Wild et al., 2018) and are crossed by watercourses rich in good-quality waters, with water surplus at least from autumn to spring. These conditions are largely favorable to MAR.

In this study, we focus on the upper plain of the Friuli Venezia Giulia (FVG) region, northeastern Italy (Figure 1). The area is representative of the pre-Alpine region. The plain is crossed by several rivers collecting water in their alpine catchments and conveying it to the northern Adriatic Sea. Fifty kilometers separate the Alps front from the coastline. Especially between autumn and spring, when the river discharges are large and irrigation is unnecessary, a portion of the waters flowed into the Adriatic Sea could be diverted and stored into the pre-Alpine aquifer system. We tested the effectiveness of using the waters from the Tagliamento River to recharge the phreatic aquifer at Mereto di Tomba in the upper plain, 10 to15 km from the Prealps. MAR was carried out using an infiltration pond excavated in the early 2000s (Teatini et al., 2015).

Details are in the caption following the image
Map of the study area showing the main geomorphological and hydrological features and the average elevation (m above msl) of the water table in the UP over the period between 1967 and 2008. The Mereto di Tomba site where MAR was carried out and the boundary of the local and regional groundwater flow models are highlighted.

Our main aim is to investigate the potential role of MAR to support the natural groundwater flow regime in pre-Alpine aquifers. This is accomplished by a three-phase two-scale analysis. An almost 1-year long test was carried out in 2014, with a first preliminary phase to set up the civil infrastructure, followed by a second 2-day recharge phase aimed at characterizing the shallowest soil below the pond, and a third longer one when MAR was fully developed. The last two phases were extensively monitored using time-lapse geophysics, piezometric records, electrical conductivity measurements in boreholes, and chemical analyses on water samples. The data set was used to calibrate a local 3D variably saturated groundwater flow model of the 1-km MAR scale. A 10-km regional scale groundwater flow model is then developed to estimate the major components of the natural flow regime. Finally, the outcome of the MAR-scale model in terms of infiltration rate is introduced as a new water source to evaluate its effects on the depth to the water table and the springs located at the passage between the undifferentiated phreatic and the downstream multiaquifer systems. A few scenarios are investigated.

The paper is organized as follows. The hydrogeological setting of the pre-Alpine upper FVG plain and, more specifically, the Mereto di Tomba area is first reviewed. Then, the modeling setup at the local and regional scales is introduced. The results of the monitoring approach and their use to quantify accurately the recharge rate by means of a 3D numerical model of the MAR experiment are shown. After the calibration of the regional groundwater flow model, few scenarios are presented to investigate the MAR role in the context of the natural groundwater balance. Finally, a discussion section and the conclusions analyze the results and draw the main findings of this work.

2 Hydrogeological and Climatic Setting of the Pre-Alpine Upper Friuli Plain

2.1 Regional Setting

The Friuli plain is a relatively small alluvial area extending 2,900 km2 in the northeasternmost tip of Italy. Urban centers, farmlands, and industrial zones are scattered throughout the plain and take advantage of a highly productive aquifer system formed of a thick unconfined aquifer and a multilayer aquifer system hydraulically connected. The plain (Figure 1) is bounded to the north by the Moraine Amphitheater and the Julian and Carnic Prealps and by the Karst to the east and the Adriatic Sea to the south. The Livenza River represents its western geographical border. The Isonzo, Natisone, Torre, Cormor, Tagliamento, Cellina, and Meduna are the major Alpine rivers flowing through the area, and their catchment basins consist mainly of carbonate rocks.

From a geological point of view, the Friuli plain represents the 6,000 m thick Miocene to Holocene sedimentary infilling of the subsiding foreland basin related to the opposing thrust belt chains of the Southern Alps and the Northern Apennines (Castellarin et al., 1992; Massari et al., 1986).

The Quaternary geological evolution of the plain was characterized by sedimentary processes constrained by intense tectonic-structural activity (Zanferrari et al., 2008) and glacial-eustatic variations related to the late Quaternary climatic events (Florineth & Schluchter, 2000; Orombelli & Ravazzi, 1996). This has produced depositional mega-fans of regional extension (Fontana et al., 2008) characterized by a downstream decrease in grain size sediment. Within 10 to 25 km from the mountain front, the mega-fans are cone-shaped with topographic gradients ranging between 7‰ and 3‰ and consisting mainly of gravel deposits, hundreds of meters thick, irregularly cemented in conglomerate beds and intercalated by sandy and rarely clayey levels (Figure 2a). This sector of the plain, known as Upper Plain (UP), has a dry surface because of the high permeability of its sediments and is interested by an important fluvial leakage feeding the unconfined aquifer.

Details are in the caption following the image
Geological setting of the UP: (a) digital terrain model (DTM) of area included in the regional model; (b) geological sections available at the regional scale (modified after Martelli Granati, 2010); (c) DTM of the area included in the MAR model; (d) lithological sequence in wellbore PZ6; and (e) NS seismic section at the Mereto di Tomba site (modified after Teatini et al., 2015).

The southern limit of the UP corresponds to the transition from the gravels to the silty-clayey sediments alternated with sand units forming the multilayered aquifer system of the Lower Plain (LP). This lithological transition forces the groundwater table to rise up to the land surface, supplying a number of springs that lie along an “alignment” crossing the study area from east to west (Feruglio, 1925; Vecchia et al., 1968). South of the springs, the general slope of the plain gradually decreases to less than 1‰. The thickness of the Quaternary body increases westward, ranging from 50 to 900 m (Nicolich et al., 2004) with average gradients of 2.5‰ and 15‰ to the right and left side of the Tagliamento River, respectively.

The elevation of the water table in the UP (Figure 1) was mapped using a geostatistical analysis (Kitanidis, 1997) of long-term piezometric records (1967–2008) through 122 wells scattered in the area. The aquifer recharge from Tagliamento River is evident from the shape of the contour lines representing the water table elevation. The groundwater flows along an NE–SW direction to the west of the watercourse. To the east, a piezometric divide due to the uplift of the pre-Quaternary deposits south of Udine gives rise to two main groundwater flow directions showing a fan-shaped arrangement: an NW–SE flow in the westernmost sector and an NE–SW direction in the eastern one, with an SE deflection near the spring area probably connected with tectonic activity.

The groundwater is characterized by a temperature less than 14°C and a total dissolved solid less than 500 mg/l, with a predominantly Ca-HCO3 facies as expected from limestone. A Ca-SO4 hydrochemical facies characterizes the Tagliamento River because of chalk dissolution within the Bellerophon formation that outcrops in the mountain basin. The groundwater sulfate decreases from the river eastward and westward (Martelli & Granati, 2010). Like other pre-Alpine catchments (e.g., Stoewer et al., 2015), groundwater in the UP is largely polluted by nitrates mainly due to agricultural practices (35 ÷ 85%) and, secondarily, to breeding (5 ÷ 45%) and civil and industrial releases (3 ÷ 25%) (ISPRA, 2015). In large areas of the UP aquifer (Figure 3a), nitrate concentrations exceed 50 mg/l—the limiting value for the potable use according to the Italian legislation (D.Lgs.152/2006).

Details are in the caption following the image
(a) Nitrate distribution in the phreatic aquifer of the UP as of 2000. (b) Annual average precipitation and depth to the water table in the central UP between 1976 and 2014. The linear trend of the latter is provided.

Despite of the modest population in the FVG region, groundwater withdrawal was estimated at about 1,860 mm3/year, 70% of which from the multilayer aquifer in LP where the pumping exceeds the natural recharge by 6.6 m3/s (Martelli & Granati, 2007; Zini et al., 2011). Piezometric levels in the unconfined and multilayer aquifer systems clearly highlight the sign of this long-term water deficit. The annual piezometric data collected in 57 piezometers in the UP showed an average increase of the depth to the water table ranging between 3.2 m (0.08 m/year) to the north and 2.4 m (0.06 m/year) near the spring line over the period from 1976 and 2014 (Figure 3b). This increase caused the spring line to shift seaward by about 1 km (Martelli & Granati, 2007). A significant pressure decrease was also recorded in several shallow (within 110-m depth) wells in the LP.

The decline of aquifer storage is caused by both the rise of anthropogenic pressure, confirmed by an increase of water wells, and the ongoing climate change. The analysis of the average annual precipitation carried out for the same period on 61 rain stations shows a slightly increasing trend, totaling about 9 mm/year, mainly due to the wetter years between 2008 and 2014 (Figure 3b). However, the annual trend is uneven, with a few percent increase in the western part of the plain and an average decrease, estimated at 15 ÷ 20% over the considered period, in the eastern sector. More significant is the trend of rainy days, which decreased significantly in the spring/summer seasons (ARPA FVG, 2018).

The analysis of the annual temperatures collected in 29 meteorological stations over the same period (1976–2014) shows an average increase of 1.6°C (0.041°C/year), both in the mountain and in the plain sectors, causing a drastic decrease of Alpine glacier extent. These changes were responsible for a decrease in effective infiltration, an increase in evapotranspiration, estimated in about 10%, and thus the reduced availability of both surface water and groundwater.

2.2 The Mereto di Tomba Setting

The Mereto di Tomba area (46° 3′46.53″N; 13° 4′15.74″E) is located in the central part of the UP, to the east of the Tagliamento River (Figure 1). With a surface elevation of about 105 m above mean sea level (amsl), the site was selected to evaluate the impacts of MAR in this pre-Alpine aquifer in terms of both groundwater storage capacity and improvement of groundwater quality through enhanced mixing and natural attenuation.

The lithostratigraphic data obtained through 11 wells drilled within a 10 km2 zone revealed the presence of a top 15 m thick gravel unit, below which two relatively low-permeability clayey gravel and cemented gravel layers, 10 to 15 m thick, are encountered. The gravel system overlies a fractured conglomerate about 30 m thick (Figure 2b). Moving southward, a compact conglomerate, probably representing the bottom of the aquifer, was found at a depth of 97.5 m from the land surface. A seismic acquisition pointed out that this sequence develops in the whole area with significant continuity (Figure 2b).

The long-term (1976–2014) behavior of the regional phreatic aquifer at Mereto di Tomba shows an average water table deepening of about 0.11 m/year, that is, approximately 4.0 m over this 36-year period. To investigate in detail the local groundwater behavior, weekly measures of water level were performed, between November 2012 and October 2014 at nine monitoring wells located around the recharge site. The piezometric level was characterized by an NE–SW flow direction due to the prevailing recharge from the Moraine Amphitheater, with a gradient ranging between 2 × 10−4 and 5 × 10−4. A fluctuation of the depth to the water level between 55 to 33 m below the land surface was observed, together with the development of a temporarily perched aquifer above the clay gravel layer at about 14-m depth below the land surface.

The physicochemical groundwater characterization of groundwater was carried out through the nine monitoring wells over the same 2-year period using a multiparameter probe YSI556. Samples were collected and analyzed at the Water Geochemistry and Hydrogeology Laboratory of the University of Udine by means of a Metrohm “850 Professional IC” ion chromatograph. The water properties of the San Vito channel used for MAR were also quantified. Groundwater is characterized by temperature of about 13.5°C, electrical conductivity of 700 μS/cm, and total dissolved solid of 400 mg/l, with a Ca-HCO3 facies. The sulfate and nitrate contents amount to about 35 and 60 mg/l, respectively. The waters of the San Vito channel have bicarbonate-calcium facies, with an average sulfate content equal to 70 mg/l, higher than that of groundwater, and a very low nitrate concentration averaging 6 mg/l.

3 MAR Tests at Mereto di Tomba

We took advantage of an infiltration basin excavated in the early 2000s at Mereto di Tomba (Figure 4). The possibility of MAR implementation was allowed by the regional environmental agency only 10 years later. An accurate survey of the basin geometry was made by Lidar (Figure 4e): the pond is 5.5 m deep, 45 × 7 m2 wide at the bottom, with an areal extent of approximately 700 m2 at the top, and its volume amounts to approximately 1,000 m3 at a 102.15-m amsl elevation, which corresponds to a water level of about 2.5 m above the pond bottom.

Details are in the caption following the image
(a–d) Photos of the MAR test at Mereto di Tomba. (a) The basin filled by water in July 2014; (b) TL-ERT acquisition system in 20 March 2014; (c) the infiltration basin in mid-June 2014 in empty condition because of the management activities on the San Vito channel; (d) detail of the basin bottom with the pressure transducer used to monitor the water level and highlighting a 2–5 mm thick layer of fine deposits sedimented on the bottom; and (e) DTM of the infiltration basin obtained by Lidar.

Three infiltration tests were carried out over the period between December 2013 and October 2014 using the water provided by the nearby San Vito channel. A first preliminary test aimed at verifying the civil infrastructures (channel diversion, pipes, screening system, and sedimentation basin), and the basin response started in December 2013 and lasted 39 days. Then, two fully monitored tests were carried out. The first test lasted 2 days (19 and 20 March 2014), whereas the second one lasted 237 days from 21 March to 14 October 2014.

Various parameters were recorded during the tests: the water discharged into the basin (Figure 5a), the basin water level (Figure 5b), and piezometric levels at four 70 m deep boreholes (PZ6, PZ8, PZ11, and PZ12 located at 50, 400, 400, and 10 m far from the basin, respectively) and two 15.5 m deep boreholes (A1 and A2) drilled a few meters upstream of the basin edge. A detailed location of the boreholes and the recoded piezometric levels are provided in the following. Water conductivity and basic chemistry analyses were conducted on samples collected weekly from PZ12. Moreover, a time-lapse electrical resistivity tomography (TL-ERT, Wenner configuration) was carried out over the 2-day long infiltration experiment (Figure 4b). A rough estimate of the total volume of water diverted to the pond from 19 March to 14 October 2014, as obtained by integrating the curve provided in Figure 5a, amounts to 290,000 m3.

Details are in the caption following the image
(a) Water volumetric discharge into the infiltration basin as measured at the channel diversion and (b) water level versus time in the basin. A zoom of the water level fluctuations during the first 2 days (19 and 20 March 2014) when the TL-ERT survey was carried out is provided in (c). The starting time (Day 0) corresponds to 19 March 2014.

Notice in Figure 5 that MAR was stopped between 12 and 25 June 2014, due to restoration work on the San Vito channel supplying the water to the infiltration basin. Unfortunately, the water diversion was not promptly shut down, causing the sedimentation of a few millimeter thick layer of fine deposits on the pond bottom (Figures 4c and 4d) and affecting the infiltration efficiency as presented in the sequel. The large fluctuations of the discharge rate over the first 3 months of MAR were caused by a significant transport of materials (leaves, brushwood, etc.) causing a frequent obstruction of the screening system. The system was improved during the mid-June period of MAR shutdown.

4 Quantifying Managed and Natural Aquifer Recharge

Natural and managed recharge of an aquifer generally develops at different scales, on the order of some tens of kilometers for a pre-Alpine basin and some hundred meters, respectively. The quantification of the two components of the recharge requires the development of regional- and local-scale numerical groundwater flow models. The calibration of the former is carried out by means of regional piezometric networks and other regional hydrologic variables such as precipitation, evapotranspiration, leakage from surface water bodies, and spring outflows (e.g., Faunt et al., 2010). Conversely, investigating MAR effectiveness requires the development of specific tests and the use of models to reproduce the hydrologic observations collected locally around the recharge plant (Ringleb et al., 2016). The modeling framework described in the following is used in this work for both the local and the regional-scale investigations.

4.1 Mathematical and Numerical Model

The parabolic partial differential equation describing water flow in variably saturated porous media is obtained by combining Darcy's law with the continuity equation. For a 3D system, the so-called Richards equation (Philip, 1969) reads:
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0001(1)
where σ(Sw) = SwSs+ϕ∂Sw/∂ψ, Sw(ψ) being the water saturation, Ss the specific storage coefficient, ϕ the porosity, ψ the pressure head, Δ = (/∂x, /∂y, /∂z) is the gradient operator, and Ks is the hydraulic conductivity tensor, represented in our case by a diagonal matrix with elements Kx, Ky, and Kz, the saturated conductivity coefficients along the x, y, and z coordinate directions (assumed to coincide with the principal directions of anisotropy). Kr(Sw) is the relative hydraulic conductivity function, ηz = (0,0,1)T, z is the vertical coordinate upward, and qs represents distributed source or sink terms (volumetric flow rate per unit volume).

The Richards equation module that is part of the more general code CATHY (Camporese et al., 2010) was used in this study. Equation 1 is solved by tetrahedral-based finite elements with linear basis functions and a weighted finite difference approach for time stepping. The code handles temporally and spatially variable boundary conditions, including seepage faces and atmospheric inputs, and heterogeneous material properties and hydraulic characteristics. Equation 1 is nonlinear due to the pressure head dependencies in the storage and conductivity terms and is solved in the code using a Picard or Newton iterative technique (Paniconi & Putti, 1994). Efficient heuristic time stepping control is used to increase the robustness of the simulator to ensure practical convergence of the iterative schemes in all conditions.

Soil characteristics are specified using the van Genuchten (1980) model, which provides water saturation and relative hydraulic conductivity as functions of the pressure head (Paniconi & Putti, 1994):
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0002(2)
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0003(3)
where Swr is the residual water saturation, urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0004, ψs is the capillary or air entry pressure head value, n is a constant, and m = 1 − 1/n for n approximately in the range 1.25 < n < 6.

4.2 Model Setup

4.2.1 MAR Model

A 3D tetrahedral finite element (FE) mesh of the phreatic aquifer system around the infiltration pond was constructed considering the digital terrain model of the farmland (Figure 2c) and the pond (Figure 5e) and the distribution of the stratigraphic units obtained by integrating wellbore (Figure 2d) and seismic information (Figure 2e) (Teatini et al., 2015). The model spanned the upper ~70 m of the sedimentary sequence, which was divided in four main lithologic units: a ~15 m thick gravel layer, below which clayey gravel, cemented gravel, and fractured conglomerate about 9, 20, and 27 m thick, respectively, are located (Figure 2d).

Figure 6a shows an axonometric view of the 3D mesh, with a zoom on the infiltration pond. The domain extended 500 and 800 m along the west–east and south–north direction, respectively, and was bounded above by the land surface and below by a 35-m amsl horizontal plane. The grid consisted of 551,978 nodes and 3,272,223 tetrahedra with characteristic element sizes ranging between 0.5 m at the infiltration pond to 3 m along the model boundaries. The mesh was generated using TetGen (Si, 2015).

Details are in the caption following the image
3D view of FE grids developed to simulate (a) the MAR at Mereto di Tomba and (b) the regional groundwater flow. The lithostratigraphic units are displayed.
The initial and boundary conditions were prescribed as follows. The water table in undisturbed condition, that is, prior to the beginning of the MAR tests, was set at 52 m amsl in correspondence of borehole PZ6 and then areally extended using the average regional north to south gradient equal to 0.2% as provided by the piezometric measurements described above. This means that the water table in correspondence of the infiltration pond was located at the interface between the cemented gravel and the conglomerates units. Then, the initial pressure head ψ0 was assigned to the generic node of coordinates (x,y,z) according with the relation:
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0005(4)
where zW is the water table elevation in correspondence to the (x,y) coordinates and z the node elevation (i.e., hydrostatic pressure distribution). Equation 4 was used also as boundary condition along the outer bounds, which were set at a distance sufficiently far from the pond to assume negligible the infiltration influence. The bottom was assumed impermeable. A null recharge was also assumed on the nodes at the model surface except for those located in the pond below the water level. Here, a Dirichlet boundary condition was prescribed:
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0006(5)
with urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0007 the prescribed head, zL the water level in the pond changing over time, and zb the elevation on the bond bottom at node of coordinate (x,y). Notice that the number of Dirichlet nodes had to be varied over time to account for the switch to no-flow boundary conditions when zL falls below zb.

4.2.2 Regional Model

The regional model comprised the UP between the Tagliamento and the Cormor rivers to the west and east, respectively. Along the NS direction, the domain extended between the Moraine northward and 5 km to the south of the spring line, within the LP (Figure 1). The model area amounted approximately to 24 × 35 km. To the north of the spring line, the lithostratigraphic sequence recognized along a few NS sections (Figure 2b) was approximated in two macro-layers, the upper one representing the “gravel system” with the “clay-conglomerate system” at the bottom. The thickness of the two units averaged 50 and 125 m, respectively, and varied in the ranges 10 ÷ 70 m (gravel) and 90 ÷ 150 m (conglomerate). The multiaquifer system in the southernmost portion of the domain was approximated by a homogeneous unit, whose hydraulic permeability was calibrated in order to match the spring line location. On top, the model domain was confined by the ground surface (Figure 2a) and the lower limit of the conglomerate formation represents the bottom boundary.

Figure 6b shows a 3D view of the FE, in which the lithostratigraphic sequence described above can be recognized. Like the development of the MAR mesh, the 3D tetrahedral FE grid was generated using TetGen (Si, 2015) and consisted of 138,958 nodes and 732,718 tetrahedra. The mesh represented accurately the ground topography and the lithostratigraphic sequence, with a characteristic element size that ranged from 20 to 500 m in the horizontal direction and 2 to 10 m along the vertical one. A finer mesh was used to discretize the gravel unit where the water table mainly fluctuated, in correspondence of the recharge zones, and along the spring line.

The following boundary conditions were prescribed. The Tagliamento River significantly recharged the phreatic aquifer system, as outlined by the curvature of the piezometric isolines shown in Figure 1. By means of an integrated investigation (geostatistical elaborations of piezometric, rainfall, and climate records together with chemical and isotopic analyses) of a large data set spanning the decades from the 1970s to the 2010s, the average leakage from the Tagliamento riverbed was estimated in 54 m3/s (Martelli & Granati, 2010; Zini et al., 2011). This value corresponds to 2,332,800 m3/day, assuming an equal subdivision between the two river sides. The largest recharge takes place close to the Alpine foothills, when the river enters in the UP. Due to the lack of more quantitative information, the leakage was modeled as linearly distributed along the ~25 km long portion of the river crossing the UP, with a null value in correspondence of the spring line. A Neumann boundary condition of 2.16 to 0.0 m3/(s·km) was thus imposed, with the largest value at the Prealps and Moraines foothills, decreasing southward. Available measurements suggested to use a no-flow condition along the Comor River (east boundary). The model bottom was assumed impermeable too. On the northern boundary along the Moraines bound and the southern one within the LP, Dirichlet conditions were prescribed. A hydrostatic pressure was used along the two bounds, with the pressure head computed as:
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0008(6)
where zr is the water table elevation in correspondence to the generic (x,y) coordinates along the Moraines margin and the ground elevation along the southern boundary, respectively, and z is the elevation of the generic node with coordinates (x,y). The water table elevation along the Moraines was obtained by the interpolation of the available piezometric map (Figure 1). Because of the relatively short distance between the southern bound of the model and the spring line and a 1 × 10−3 ÷ 2 × 10−3 average hydraulic gradient in the upper multiaquifer system (Martelli & Granati, 2010), it was reasonably assumed that any significant overpressure developed within the simulated northernmost strip of the LP, with the piezometric head remaining at the land surface. Finally, a uniformly distributed infiltration equal to 730 mm/year was prescribed on the land surface. This value was obtained by considering a yearly precipitation equal to 1,550 mm (Figure 3b), an evapotranspiration estimated in 700 mm (Martelli & Granati, 2007), and a potential infiltration coefficient equal to 0.8 since most of the area is farmland. Therefore, the daily recharge amounted to about 890,000 m3/day since the areal extent of the UP within the model boundary amounted to 445 km2 (Martelli & Granati, 2010).

Because only steady-state simulations were performed at the regional scale, initial conditions lose importance.

4.3 Model Calibration

4.3.1 MAR Model

In situ investigations using sieve analyses, borehole infiltration, and pumping tests provided a preliminary characterization of the hydraulic parameters of the main hydrogeologic units, that is, gravel and conglomerate, of the Mereto di Tomba site (Teatini et al., 2015): Ks = 10−5 ÷ 10−4 m/s, ϕ = 0.20 ÷ 0.30, and Ss ~10−3 m−1. Concerning the van Genuchten parameters, the values were computed using the Rosetta software (Schaap et al., 2001) on a few grain size distribution curves representative for the Mereto di Tomba sandy-gravel soil (Teatini et al., 2015): Swr = 0.1 ÷ 0.15, ψs = 0.7 ÷ 1.8 m, and n = 2.5 ÷ 3.0.

These values were used as initial estimate in the calibration procedure of the MAR model carried out in this work. The reliability of the system characterization, specifically the saturated hydraulic conductivity and the air entry pressure, was improved by taking advantage of the data set collected during the MAR tests. A preliminary sensitivity analysis showed that the variations of Ss, Swr, and n within reasonable ranges negligibly affected the model response. Consequently, the values defined in Teatini et al. (2015) were used, thus avoiding overparameterization during model calibration.

The calibrated model parameters are reported in Table 1. Calibration was carried out as follows. The properties of the shallower gravel unit constituting the bottom of the infiltration basin were characterized using the evolution versus time and depth of the electrical resistivity ρ provided by the TL-ERT survey over the initial 2-day infiltration test. Figure 7 compares the ρ acquisitions and the model outcome in terms of Sw for a few times representative of the pond filling and the following emptying due to the recharge shutdown and the natural infiltration thought the pond bottom. The maximum water level in the pond was constrained to remain below 0.7–0.8 m to allow the correct functioning of the electrical equipment (Figures 4b and 5c) and the infiltrated water remained located within the top gravel layer. The model outcome matches the geoelectrical results satisfactorily. Notice the different behavior versus time of the infiltration along the basin that depended on the irregular elevation of its bottom (Figure 4e). The acquisition of a detailed pond geometry and its accurate representation through the FE discretization played a major role in reproducing the TL-ERT data set.

Table 1. MAR Groundwater Flow Model: Calibrated Soil Parameters
Material Ks (m/s) Ss (m−1) ϕ ψs (m) Swr n
Gravel 1.8 × 10−5 10−3 0.3 1.9 0.1 3.0
Clayey gravel 1.1 × 10−6 10−3 0.2 6.3 0.1 3.0
Cemented gravel 1.1 × 10−5 10−3 0.3 2.4 0.1 3.0
Conglomerate 1.1 × 10−3 10−3 0.3 0.2 0.1 3.0
Details are in the caption following the image
Comparison between the TL-ERT survey and calibrated MAR model outcomes on a vertical section crossing the infiltration basin along its main axis. The ρ (top inset in each sub-panel) and Sw (bottom inset of each sub-panel) sections are provided for a few significant times: (a) 0 hr (10:00 a.m., 19 March, before the beginning of the MAR experiment); (b) 1.5 hr (11:30 a.m., 19 March); (c) 4.5 hr (14:30 p.m., 19 March); (d) 7 hr (17:00 p.m., 19 March); (e) 13 hr (00:00 a.m., 29 March); and (f) 25 hr (11:00 a.m., 20 March). The inset shows the selected times on the water level versus time behavior.

Depth to the water table recorded at the various piezometers played a main role in characterizing the deeper geologic units. The 15.5 m deep A1 and A2 piezometers highlighted the formation of an unexpected perched aquifer (Figure 8a). The clayey gravel layer was semi-pervious and precluded a direct connection between the pond and the regional water table aquifer located at approximately 50 to 55 m below the land surface. The hydraulic properties of the clayey gravel were calibrated in order to match the A1 and A2, located a few meters to the north of the basin (see map in Figure 8). The comparison between the model results and the piezometric records is shown in Figure 8a. The water table below the pond rose from 2 to 3 m, with a dynamics that strictly followed the fluctuation of the water level in the infiltration pond (Figure 5b). A much smaller dynamics developed in the deep aquifer. Figure 8b shows that the water table variation was limited to a few tens of centimeters. This required that the fractured conglomerate was characterized by a high hydraulic permeability, about 2 orders of magnitude larger than that of the gravel. Concerning ψs, the value ψs = 0.2 was selected in agreement with values published in the literature for fractured rocks (e.g., Finsterle, 2000; Trautz & Wang, 2002). Any specific information was available to calibrate the cemented gravel unit because the water table remained always within the conglomerate during the MAR test. Therefore, a value of Ks was selected only based on the grain size distribution.

Details are in the caption following the image
Observed versus simulated pressure head changes: (a) 15.5 m deep piezometer A1 and (b) 70 m deep PZ6, PZ8, PZ11, and PZ12 piezometers. The manually recorded values obtained by means of a phreatimeter in the A2 borehole are provided in (a). The records in (b) represent the differences between the head values in the wells surrounding the infiltration basin and those recorded in the well 0156 of the regional network located 1 km to the SE of the MAR site. The well locations are provided in the map. The starting time (Day 0) corresponds to 19 March 2014.
It is interesting to highlight that some restoration works were needed along the San Vito channel at the beginning of June 2014. This caused a temporary shutdown of the basin recharge (Figure 5a). Unfortunately, the closure of the diversion between the channel and the pond was not sufficiently prompt causing the deposition of a few millimeters of fine sediments on the pond bottom (Figures 4c and 4d). The MAR experiment resumed after 12 days. The improvement of channel diversion and the screening system during the stop allowed to increase by ~25% the derived water volume (from approximately 1,300 to 1,700 m3/day, Figure 5a). The water level in the pond increased correspondingly, from 2 to 2.7 m. However, piezometers A1 and A2 revealed a thickness of the perched aquifer that, returned to the pre-shutdown value soon after the MAR restart, gently decreased over time with a behavior significantly different from the phase before June (Figure 8a). An in situ inspection on the basin bottom after the MAR end revealed the movement of the fine sediments into the top part of the gravel causing a partial clogging. To consider the clogging effect in the numerical simulations, the following empirical relationship was developed of the hydraulic conductivity of the gravel:
urn:x-wiley:00431397:media:wrcr24843:wrcr24843-math-0009(8)
with Kz(t0) the hydraulic conductivity before clogging, ΔKz its changing rate over the time interval (tf − t0), and tf and t0 the final time of the test and the starting time when clogging started. In the present case, Kz(t0) = 1.8 × 10−5 m/s (Table 1), tf =239 days, and t0=100 days, which corresponded to the resume of MAR. The value ΔKz = 1.8 × 10−5 m/s was calibrated to reproduce the records. Figure 8a shows that the model represented satisfactorily the evolution of the perched aquifer if clogging was accounted for. The sudden shutdown and recovery of the recharge provided interesting information of the system dynamics. For example, the comparison between the pressure head variation in the local perched and regional aquifers over the period 12 to 25 June allowed to quantify in about 10 days the time interval required by the perturbations on the land surface to reach the aquifer.

The model allowed to accurately compute the amount of water that recharged the aquifer through the MAR test. Figure 9 shows the behavior versus time of the computed infiltration rate and the cumulative infiltrated volume. The infiltration rate remained quite stable at 1,000 m3/day, with short-term fluctuations corresponding to the main water level fluctuation in the pond. The total volume of infiltrated water at the end of the MAR test amounted to 210,000 m3, which differs reasonably from the total volume of water diverted to the pond (290,000 m3) considering the uncertainty characterizing this latter value and some evaporation.

Details are in the caption following the image
Infiltration rate and cumulative infiltrated volume as computed by the calibrated MAR model. The starting time (Day 0) corresponds to 19 March 2014.

Figures 10 and 11 provide a general view of the MAR outcome obtained with the calibrated model. The numerical results in terms of water saturation at a few significant times (after 30, 90, 104, 204, and 239 days from the MAR beginning) are provided in Figure 10 along two vertical sections crossing the infiltration pond in the south–north and west–east directions. The formation of the perched aquifer (Sw ≅ 1 in the upper 15-m depth) is apparent, with the water saturation increasing from the residual value of 0.1 to about 0.5 between the water table in the conglomerate and the bottom of the perched aquifer. The streamlines of a few particles released at the pond bottom are shown in Figure 11. Over the 239 days of the MAR experiments, the maximum distance traveled by the particles amounted to ~115 m, with a prevailing southward component related to the natural groundwater regime.

Details are in the caption following the image
MAR model outcomes: Sw along an N–S and a W–E vertical section through the infiltration pond at the few significant times: (a) 0 day (19 March, before the beginning of the MAR experiment); (b) 30 days (18 April); (c) 90 days (17 June); (d) 104 days (1 July); (e) 204 days (9 October); and (f) 239 days (14 October). The inset shows the selected times on the water level versus time behavior.
Details are in the caption following the image
Computed streamlines over the 239 days spanned by the simulation for a few particles released at the bottom of the infiltration pond. The colors are representative of the distance.

4.3.2 Regional Model

Modeling of the regional phreatic system in the UP where the MAR site is located was mainly aimed at developing a rough water balance of the subsurface system. A steady-state simulation was carried out to reproduce the average long-term features of piezometric trend.

The calibration was focused on the hydraulic conductivity of the model units, with the final values summarized in Table 2. Specifically, the Ks of the gravel and conglomerate layers derived by the MAR model were updated to match the piezometric map obtained by interpolating the depth to the water table recorded at the borehole networks over the past decades. For the multiaquifer system, the hydraulic conductivity of the units was calibrated by matching the location of the spring line, that is, by verifying the location of the mesh nodes on the land surface where Sw = 1. A comparison between the records and the model outcome is provided in Figure 12.

Table 2. Regional Groundwater Flow Model: Calibrated Soil Parameters
Material Ks (m/s) Ss (m−1) ϕ ψs (m) Swr n
Gravel system 1 × 10−4 10−3 0.3 1.0 0.1 3.0
Clay/conglomerate system 2 × 10−5 10−3 0.2 1.0 0.1 3.0
Multiaquifer system 1.2 × 10−4 10−3 0.3 1.0 0.1 3.0
Details are in the caption following the image
Regional groundwater flow model of the UP: comparison between the (a) measured and (b) simulated elevation of the water table (m amsl). The maps represent the average pattern over the period 1967–2008.

According to the imposed boundary conditions and the parameter calibration, the water balance of the system resulted as follows: the recharge from the Tagliamento River, the Moraines, and the effective rainfall amounted to 2,332,800, 717,000, and 890,000 m3/day, respectively, with a discharge from the southernmost boundary of the UP equal to 3,940,000 m3/day. These values are in agreement with regional estimates carried out in the past by means of geostatistical elaborations of records from regional networks (Martelli & Granati, 2007, 2010): the effective meteoric infiltration accounted for approximately 25–30% of the whole recharging contribution; the total discharge from the UP between the Tagliamento and the Isonzo rivers (Figure 1) amounted to 106 m3/s. Since the study area represents approximately 40% of the total region addressed in Martelli and Granati (2007, 2010), this volume can be quantified in ~3,600,000 m3/day.

5 MAR Versus Natural Aquifer Recharge

The final aim of this study was to investigate the MAR as a best management practice for water resources in pre-Alpine basins. Managing an MAR site as the Mereto di Tomba infiltration basin, where recharge is performed only through gravity, the pond is already available, and the surficial water used to recharge the aquifer is characterized by a quality generally better than the groundwater quality, still requires the effort of technicians to survey and maintain the engineering system, with the possibility of a prompt intervention in the case of malfunctioning.

A first evaluation about the effectiveness of the MAR at Mereto di Tomba was carried out by estimating the water budget in portion of the UP around the MAR site for the year 2014, which was a wet year. The 2014 precipitation at the Udine, where the rain gauge closest to the MAR site is located (Figure 1), was equal to 2,281 mm, with an annual evapotranspiration quantified in 772 mm (Turc, 1954). Considering a potential infiltration coefficient of 80%, the aquifer recharge due to rainfall was estimated in ~1,200 mm. The areal extent of the semi-pervious clayey gravel layer detected below the infiltration basin and responsible for the development of the perched aquifer was considered as reference for the budget analysis. By processing available lithostratigraphic and geophysical data, its extent was approximated 1,600,000 m2. The volume of rainfall water recharging this area amounted to 1,920,000 m3, a value that must be compared with the 210,000 m3 infiltrated through the Mereto di Tomba site during the test. MAR increased the annual infiltration by ~11%. Assuming that a daily infiltration rate equal to 1,000 m3/day lasted over the entire year, the MAR at Mereto di Tomba would increase the aquifer recharge by ~19%, which is a quite significant quantity.

A significant impact was also recorded in terms of groundwater quality (Figure 13). At PZ12, which is located downstream of the basin along the groundwater flow direction (Figure 8), the electrical conductivity decreased from 738 to 580 μS/cm during the preliminary test in January 2014, and then to 490 μS/cm, a value which is only slightly higher than that of water of the San Vito channel, during the MAR phase from April to October 2014. Similarly, nitrates decrease from 62 to 12 mg/l in January 2014, continued to decrease from April 2014, and took the same value of the channel water (4 mg/l) from August until the end of the MAR test. The sulfate content increased from 35 to 69 (January 2014) and to 80 mg/l during the last part of MAR. The waters sampled form A1 piezometer, which intercepted the perched aquifer fed by the managed recharge, showed values of the three indicators in accordance with those of the San Vito channel. No significant changes were recorded at PZ8 and PZ11 wells, a few hundred meters downstream of the MAR site. The recorded reduction of nitrate concentration, although observed at the local scale only, assumes a certain importance as nitrate contamination in groundwater is a widespread problem also in high-plain pre-Alpine aquifers (e.g., Wild et al., 2018).

A more general evaluation of the MAR effects on the regional aquifer was carried out by integrating the outcomes obtained by the local and the regional models presented above. By means of the local model, it has been quantified in 1,000 m3/day the water volume that can be infiltrated in normal condition through an MAR site in the UP. This rate has been used as additional water inflow for the regional groundwater flow model, with the effectiveness of the MAR evaluated in terms of rise of the regional water table and northward shift of the spring line. The simulations were carried out in steady state (scenario M0), with the result provided in Figure 14a. As expected from the previous outcomes, the MAR site impacted the regional aquifer locally, with a few centimeter rise of the regional water table near the MAR site itself and no effect on the spring line location.

Details are in the caption following the image
Time behavior of the electrical conductivity (EC) and nitrate and sulfate concentration in the San Vito channel and the monitoring wells around the Mereto di Tomba pond during the MAR tests.

However, Mereto di Tomba is only one of the various ponds and gravel pits distributed in the UP. The potentiality of a cheap and widespread MAR practice in this pre-Alpine system is high due to the large number of pits and shallow large-diameter wells, the thick high-permeable unsaturated aquifer, and the availability (mainly from autumn to spring) of a large amount of high-quality surface water that is conveyed by the rivers and “lost” into the nearby Adriatic Sea. The average annual precipitation of the Carnic and Julian Prealps (Figure 1) ranges between 2,000 and 3,000 mm.

To address the influence of a possible extensive MAR practice on the regional groundwater flow, two different scenarios (M1 and M2) were investigated. In scenario M1, a second infiltration basin located in Carpaneto, about 10 km to the SE of Mereto di Tomba, was supposed to put in operation. In scenario M2, we simulated the simultaneous use of the two ponds and a number of 10 to 20 m deep, 2-m diameter wells drilled in Mereto di Tomba (two wells) and Carpeneto (10 wells) but never put in operation because of legal issues (Civita, 2005). The cumulative water volume infiltrated by MAR in scenario M2 amounted to 2,000 and 10,000 m3/day in Mereto di Tomba and Carpeneto, respectively. Figures 14b and 14c summarize the model outcome obtained with the two scenarios. The rise of the water table remained in the order of a few centimeters in M1 and increased to a maximum of 0.4 m in correspondence of the MAR site and to 0.1 m on a ~20 km2 area around Carpeneto in M2. Moreover, with scenario M2, an important upstream shift of the spring line, approximately equal to 30 m, was computed along the 5-km easternmost portion of the study area.

Details are in the caption following the image
Rise of the water table computed by the regional model of the UP in steady-state conditions for scenario (a) M0, (b) M1, and (c) M2.

6 Discussion

Important advancements have been achieved in MAR over the last decades (Dillon et al., 2019). With the need to face advancing climatic changes and growing human need of freshwater, MAR represents an increasingly important strategy for water management allowing to enhance stressed groundwater systems, protect natural subsurface water resources, and improve the quality of surface waters and treated wastewaters for their use and reuse (Dillon & Arshad, 2016).

Although (i) different types of MAR schemes are widely distributed and applied on various scales and for various purposes in the European countries (Sprenger et al., 2017), and (ii) the effectiveness of MAR in European mountain regions is well known since centuries ago (Martos-Rosillo et al., 2019), MAR is poorly implemented in Alpine and pre-Alpine regions. Available databases (Sprenger et al., 2017; Stefan & Ansems, 2018) show that very few are the locations (along the Rhine River close to Basel and the Arve River in Geneva, Switzerland; de los Cobos, 2002) where MAR is ongoing to replenish aquifers using surplus surface waters. This evidently contrasts, for example, with the ongoing practice in the Netherlands where managed recharge of coastal aquifers is widely implemented (Stuyfzand & Doomen, 2004). Despite an increase in yearly gross precipitation induced by climate change (van der Hurk et al., 2014), MAR has been largely investigated and implemented with the prospects of longer periods of drought and an increase in extreme rainfall events, which will require better management of aboveground water reservoirs for retention of intense rainfall. MAR provides the means to lower the levels of these reservoirs by early infiltration once potential extreme rainfall events are predicted and to store water without having to discharge (and lose) it to the sea.

In the pre-Alpine region of Italy, where climate changes have started to affect water availability since the last decade (Calligaris et al., 2016), only a few pilot initiatives have been developed over the last years, generally associated to research projects funded by the European Union (e.g., FOKS, Bertoldo et al., 2013; AQUOR, Mezzalira et al., 2014). None of these tests have been advanced to a permanent MAR plant. Conversely, the tests ended with the project closure. A main barrier to the development of aquifer recharge in Italy has been the lack of a specific legislation regulating MAR management since 2016. The EU Water Framework Directive (EU, 2000) has recognized MAR as a water management tool which may be used for supporting the achievement of good groundwater status, but it required member states to develop their own policies in relation to MAR application. In Italy, a regulation on licensing and permitting MAR plants was promulgated only in June 2016 (Ministero dell'Ambiente, 2016) with the Regional Environmental Agencies never allowing full applications before that date.

The MAR test at Mereto di Tomba followed the same process. Funded through the EU WARBO project (http://www.warbo-life.eu/, accessed April 2020), it finished in 2015 upon completion of the research. The application of MAR-RISKAPP tool (Rodríguez-Escales et al., 2018) to the study site shows that, among the main risks of MAR failure, the probability of nontechnical factors prevails over the technical ones (0.87 vs. 0.26 for the MAR design and construction phase and 0.73 vs. 0.58 for the MAR operational phase). MAR governance is the most probable cause for failure of the Mereto di Tomba facility.

However, the study developed in the framework of this test has allowed to accomplish two major advancements. First, it has been clearly outlined how direct and indirect monitoring surveys can be fruitful integrated with numerical modeling to follow the recharge behavior and quantify accurately the amount of water that can be stored in the subsurface through an MAR site. In agreement with previous results in different research fields (e.g., de Franco et al., 2009; Meyerhoff et al., 2014), the capability of time-lapse hydrogeophysics to provide a clear picture of the water movement at the local scale of the infiltration pond must be outlined. Second, MAR appears as a promising approach to counteract the decrease of water resources due to climate changes and human pressure in pre-Alpine regions allowing storing of approximately cubic millimeter per year in the subsurface. At least three issues support this consideration:
  1. aquifers bounding the Alps foothills are generally characterized by coarse textures and thick unsaturated zone;
  2. surplus surface waters, chemically compatible with groundwater (Teatini et al., 2015) and even qualitatively better than groundwater (e.g., with a lower nitrate concentration), are largely available; and
  3. MAR is economically inexpensive. Several gravel pits and ponds usable as MAR basins are scattered in these high plains. Moreover, the plains are crossed by a dense network of natural creeks and artificial channels used in summer for irrigation. Therefore, only minor economic investments are needed to put in operation MAR systems.

7 Conclusions

An experimental and modeling study was carried out at Mereto di Tomba, northeastern Italy, to investigate the effectiveness of a 1-year long MAR test. The site is representative of the typical hydrogeologic setting of the pre-Alpine region, with coarse permeable deposits and a depth to the water table of several meters. MAR was carried out through a 5 m deep infiltration pond, with the water supplied by the Tagliamento River through an irrigation channel.

A large monitoring program was carried out to follow quantitatively and qualitatively the water fate, with TL-ERT, piezometric level measurements in shallow and deep boreholes, and chemical analyses on water samples. The collected data set, together with previously available detailed information on the hydrostratigraphic units, was used to set up and calibrate a 3D FE variably saturated groundwater flow model of the MAR test. The main outcome of the local-scale model is the quantification in ~2 m3/day of water storable into the aquifer system per square meter of pond surface.

Successively, this amount was considered in the context of the natural groundwater balance of the regional aquifer system. Taking advantage of hydrologic analyses and geological investigations carried out in previous studies, a regional 3D FE groundwater model of the aquifer system analyses where the Mereto di Tomba MAR is located was developed. The model allowed to quantify the main inflow and outflow contributions to the undifferentiated phreatic aquifer, with a satisfactory match of the multiyear averaged piezometric level and location of the spring line representing the limit between the undifferentiated and the multiaquifer systems.

The infiltration rate computed by the local model was added to the regional one, and the effect was evaluated in terms of water table rise and upstream shift of the springs. The analysis showed that MAR at Mereto di Tomba can affect the aquifer condition only locally. However, a modeling scenario carried out considering the possible MAR implementation in other gravel pits already excavated into the alluvial fan sediments of the upper Friuli plain showed that this strategy is effective to counterbalance the decrease of water resources due to climate changes already ongoing in the pre-Alpine region.

Concluding, because of highly permeable soils, large depth to the water table, the availability of high-quality surface water, and a dense network of natural creeks and/or artificial channels for its distribution, MAR represents a useful way to safely replenish aquifers in pre-Alpine areas. Presently, mid- to long-term governance and management of MAR facilities appear as the main potential limitations for a more widespread application, at least in countries like Italy, where the environment, rivers, and artificial watercourses are managed by different state, regional, and local administrative/technical bodies with superposing and sometimes contrasting authority.

Acknowledgments

The study was funded by the EU LIFE+ Project “WATER RE-BORN—Artificial Recharge: Innovative Technologies for the Sustainable Management of Water Resources.” The authors gratefully acknowledge Daniel Nieto, Alessando Affatato, Luca Baradello, and Flavio Accaino and the National Institute of Oceanography and Experimental Geophysics (Trieste, Italy), for the management of the MAR tests and the geophysical characterization of the site; the personnel of the Consorzio di Bonifica Ledra-Tagliamento, particularly Massimo Canali and Stefano Bongiovanni, for the assistance in the field; EUREKOS Srl (Portogruaro, Italy) for TL-ERT acquisitions during the MAR tests; the well drilling company Botti Elio S.a.s. (Adria, Italy); Giorgio Mattassi and Davide Brandolin, ARPA Friuli Venezia Giulia (Palmanova, Italy), for their fruitful support to the project; and Cristina Granati for the chemical analyses of the water samples.

    Data Availability Statement

    The collected data and modeling inputs described in this manuscript can be found at HydroShare (https://doi.org/10.4211/hs.024d498535ff42e9ad88432a219e6134).