Assessment of urban landslide groundwater characteristics and origin using artificial tracers, hydro-chemical and stable isotope approaches

In the framework of landslides, the hydrogeological features play an essential role in slope stability, governing water movement and thus resulting in modification of the effective stress in the soil. In this framework, the hydrogeological conceptualization of landslide areas and the identification of groundwater origin are key points to developing risk mitigation measures. In fact, groundwater recharge cannot always be attributed to local precipitation alone. Mixing processes between water derived from local infiltration and deep water upflow along tectonic lineaments or anthropogenic water can affect the groundwater balance on a local scale. This study aims to define the potential groundwater origin of one of the highest risk urban landslides in central Italy and to define a hydrogeological conceptual model by exploiting its existing drainage system network. This research is based on a multiple-techniques approach based on hydrological water balance, artificial tracer tests during recharge period, seasonal monitoring of the water stable-isotope content, hydro-chemical survey during low-flow periods, and analysis of the piezometric level fluctuation. All these analyses are coupled with a detailed reconstruction of the geology of the area depicted from boreholes and drill holes. Two groundwater bodies have been evidenced from the study. The shallower one is located in the landslide unstable zone and is hydraulically connected to a deeper groundwater body hosted in the underlying bedrock. Results highlighted that the local rainfall regime could not fully explain the hydro-chemical facies. Local water contributions to the landslide area coming from leakage of the urban sewerage system have been evidenced, excluding deep groundwater upflow from the fault system.


Introduction
In the context of landslides, several studies focused the attention on the relation between hydrogeology and landslide dynamics (Iverson and Major 1987;Baum and Reid 1992;Van Asch et al. 1999;Coe et al. 2003;Cappa et al. 2004), highlighting the importance of that discipline for mitigating the associated risk. A peculiar attention has been paid to the groundwater recharge, which may be not related to precipitation and/or snow melt alone (Bogaard et al. 2007). In fact, a considerably amount of water may derived from the deep-water circulation and upflow into landslide deposits (Colombetti and Nicolodi 1998;Bertolini and Gorgoni 2001;Ciancabilla et al. 2004;Baraldi 2008;Ronchetti et al. 2009Ronchetti et al. , 2020, or even by the presence of multiple aquifers of different depths (Binet et al. 2009). Other studies investigated the consequences of sewerage breaking due to the landslide activities, which can produce a re-activation of the landslide when the wastewater volume is not negligible (Preuth et al. 2010), or can cause alteration of groundwater quality and possible impacts on the environment and its ecosystem (Geertsema et al. 2009). To this aim, a variety of methods based on hydrochemistry have been introduced in the literature to investigate the groundwater origin in landslide areas, such as water isotopes (Guglielmi et al. 2002;Lin and Tsai 2012;Bogaard et al. 2007;Sajinkumar et al. 2017), artificial tracer tests (Bonnard 1988;Binet et al. 2007b), trace elements determination (Cervi et al. 2012) and geochemical analyses (Cappa et al. 2014;Binet et al. 2007a). Above others, the importance of water contribution different from the local precipitation has been enhanced using oxygen-18 and tritium (Mikos et al. 2004) and tracer tests (Ronchetti et al. 2009;Vallet et al. 2015), while the groundwater flow paths have been defined coupling hydro-chemical data and 18-oxygen/deuterium (Guglielmi et al. 2000(Guglielmi et al. , 2002Mikoš et al. 2004;Peng et al. 2012;Vallet et al. 2015).
In this study, five complementary approaches including hydrologic water balance, artificial (i.e., fluorescein and NaCl) and natural tracers (oxygen-18), geochemistry and leaching analysis are joined to define a potential origin of groundwater feeding the landslide and to draw a detailed groundwater conceptual model.
The groundwater hydro-chemical data were compared with leaching test results carried out on the lithologies outcropping in the landslide area, and the comparison between chemical groundwater content with local aqueduct water has been used to investigate the likely occurrence of sewerage system leaks. Finally, the landslide conceptual model has been developed combining all the hydrogeological data in some geological cross sections.

The Montelupone landslide
In Marche Region (central Italy), gravitational phenomena mainly affect the inhabited centers located on the top of hills composed by Plio-Pleistocene geological formations, which are constituted by alternances of pelitic and marly lithologies characterized by sparse arenaceous lenses and intercalations of rare arenaceous-conglomeratic bodies. The hills of the Marche Region display slopes between 20% and 70% making most of the municipalities in the territory classified as 'unstable inhabited centers' (Principi et al. 2007). In this framework, the Montelupone village has been affected by a large-scale landslide since 1600, with a significant increase in landslide movement, recorded from the '70 s, which caused several damages to the buildings and infrastructures. The Montelupone village has been classified by Guzzetti et al. (1994) as the most damaged village in Marche Region, due to landslide phenomena. The landslide topographic elevation ranges between 220 and 285 m a.s.l, affecting about 36% of the historical centre of Montelupone, both inside and outside the historical city walls. The landslide volume is approximately 2 × 10 6 m 3 and is characterized by a prevalent translational movement (Morgoni 2012), in which the sliding surface detaches the colluvial-eluvial deposits from the underlying bedrock (ISPA 1979). Landslide risk mitigation measurements have been adopted since 1980, such as the construction of two drainage tunnels (DT), with radial drains located 20 m up to 50 m below the historical centre and, respectively, located on the Northern-east and Southern-east flanks of the hill.
According to the scientific literature available to the authors, only the technical reports produced by ISPA (1979) and Morgoni (2012) reported preliminary considerations about the origin of groundwater intercepted by the drainage systems, assuming a possible groundwater amount coming from deep groundwater bodies through the fault system, in addition to water infiltrating in the landslide area by meteoric contribution.

Geological and hydrogeological setting
From a geological point of view, the study area belongs to the External Umbria Marche Stratigraphic Domain (Cantalamessa et al. 1986). The Montelupone village is located at the top of a hill, 285 m above sea level (a.s.l.), and its hillslope is characterized by the presence of the Offida member of the Argille Azzurre Fm. (Lower Pleistocene), covered by the colluvial-eluvial Quaternary continental deposits. The Offida member is, respectively, subdivided into two lithofacies (Fig. 1a). The arenaceous-pelitic lithofacies (AAF1) is characterized by alternating sandstones and marly clays (total thickness of about 10 m), with sandstones predominating; in this lithofacies both sandstones and clays are lenticular and in general the sandstone layers/lenses are responsible for the horizontal permeability, while the marly clay layers/lenses lead to a low vertical permeability. The pelitic-arenaceous lithofacies (AAF2) is composed by marly clays with minor alternances of sandstones, with a maximum thickness of about 200 m; the permeability is low due to the wide diffusion of marly clays, although a local horizontal permeability is favored by the presence of sandstone lenses. The bedding is generally flat-parallel with ENE and NNE immersion between 10 ÷ 12°.
Above the bedrock (Argille Azzurre Fm.), the colluvial-eluvial Quaternary continental cover (Musone Synthem) outcrops extensively, with an average thickness of 20-30 m and reaching a maximum of 70 m. The Musone Synthem is subdivided into two different facies, from the oldest to the younger: (a) the sandy and clayey loam with predominating clays (ECD), classified as scarcely permeable and thus acting as aquitard; (b) the silty clay and clayey silt landslide deposits (LD), classified by ISPA (1979) as impervious-scarcely permeable, with 10 -11 < k < 10 -10 m/s, thus acting as aquitard; in the medium-low portion of this unit, the grain size becomes silt-sandy or sandy-loamy, increasing the permeability and incoherence of the sediment. The heteropic variation within the Musone Synthem is particularly apparent in the zone between the historical center and the eastern landslide area at about 220 m a.s.l. LD displays convoluted sedimentary structures, syn-sedimentary slumping or anomalous strata patterns as minor-folds.
From ISPA (1979) two overlapping groundwater bodies were identified. The main one is located at the contact between ECD Fm. and the bedrock of AAF1 Fm. (from now named deeper groundwater body), while the shallower one has limited extension and is characterized by several perched aquifers located in the sandstone lenses of LD, separated each other by the marly clay layers. Figure 1b shows the piezometric map of the deeper groundwater body, from which a prevalent direction of the groundwater flow paths towards E-NE, and a minor SE directed path in the eastern side of the historical center can be observed. In the highest portion of the Montelupone hill, below the inhabited center, the water table varies between 20 and 30 m below ground level (b.g.l.).

Materials and methods
A multi-techniques hydrogeological investigation, based on hydrogeochemical, and hydrological analyses supported by artificial tracer tests and isotopic surveys has been set up to investigate the groundwater origin and to quantify the groundwater amount feeding the landslide body. The discharge of the drainage system was monitored over time, the piezometric level was measured, groundwater was sampled to determine the isotopic content (oxygen-18) and major ions concentration. Finally, soil and rock samples were taken for leaching analyses. Respectively, 3 standpipe piezometers and 11 drains of the landslide's drainage system have been selected for specific monitoring. Mixing phenomena and endmembers were assessed to characterize the groundwater chemistry and oxygen-18 value. Leaching experiments contributed to verify whether the groundwater chemical composition can be of geogenic origin or if anthropogenic pollution on groundwater exists. The complementary approach based on stable water isotopes, artificial tracer tests and groundwater balance has been used to: (i) identify the groundwater flow paths and velocity; (ii) assess the source of water drained by the tunnels; (iii) evaluate potential mixing of water from different sources. To support and develop the hydrogeological conceptual model of the landslide, a detailed geological survey has been integrated with stratigraphic data of fourteen geognostic surveys (twelve boreholes and two dynamic penetrometric tests).

Hydrogeological monitoring system
An integrated hydrogeological monitoring system has been set up to observe the relation between meteoric recharge processes and groundwater discharge trends during the study period in the landslide area. Rainfall and temperature data have been collected from the Marche Region Hydrometereological monitoring system (https:// www. regio ne. marche. it/ Regio ne-Utile/ Prote zione-Civile/ Conso le-Servi zi-Prote zione-Civile/ SIRMIP-online) using the Recanati rainfall station (located 3 km far from the landslide) and Villa Potenza thermometric station (located 10 km far from the landslide). Starting from 8 years. record (2014)(2015)(2016)(2017)(2018)(2019)(2020)(2021), the mean monthly temperature and monthly rainfall were used to calculate the mean monthly water surplus using the WaterbalANce WebApp (Mammoliti et al. 2021). The recharge area extension was iteratively selected between 50,000 m 2 (as assumed by ISPA 1979) and 80,000 m 2 including the area located further to the west of Montelupone village, in which the gently NE dipping sandy layers of the outcropping geological formations may convey the groundwater flow towards the drainage tunnel (DT). In the largest recharge area selected for calculation, the impervious portion occupied by the historic center was included, with an overall area of 67,000 m 2 . Based on rainfall and temperature data, the discharge calculated by the water balance was compared to that measured by the drainage system at the outlet of the DT.
The hydrogeological data (available from previous investigations) consist of discontinuous time series from different monitoring phases. Some data came from the exploration phase aimed at designing interventions to mitigate landslide risk; others come from a new monitoring phase using in situ devices. In particular, two piezometers named P1 and P8 (having daily water head fluctuation data available from January 2014 to May 2015), located in the Montelupone historical center, have been used to preliminary assess the presence of hydrogeological heterogeneities within the landslide body. One standpipe piezometer (named P6) has been equipped with a hydrometric pressure transducer (CTD diver Eijkelkamp, accuracy ± 0.5 cmH 2 O and resolution 0.2 cmH 2 O) compensated by atmospheric pressure, from October 2018 to March 2019, for periodically monitoring groundwater level fluctuation and water conductivity, with an acquisition time interval of 10 min. The groundwater level trends were then compared to the rainfall regime of the area to assess the groundwater response to the precipitation regime.

Tracer tests
Two artificial tracer tests have been performed to identify the groundwater flow path and evaluate groundwater velocity, giving thus information on the presence of different groundwater bodies, the hydraulic contacts between them, and preferential groundwater drainage patterns from the injection point to the monitored ones. The artificial tracers were chosen considering the interaction between tracers and the geological matrix (Tazioli and Palpacelli 2013) using sodium chloride (NaCl) and fluorescent dye (Na-Fluorescein, C 20 H 10 Na 2 O 5 ). The first test (TEST1) started on 19/10/2018 at 11:30, injecting 38 kg of NaCl into the piezometer P4 and 82 kg of NaCl into piezometer P2 at 12:40. The second test (TEST2) started on 20/12/2018 at 15:30 with the injection of 1.8 kg of Na-Fluorescein into the piezometer P3. During the TEST1, 5 drains (D3, D106, D254, D273, D196), 2 piezometers (P6, P5), and the drainage tunnel outlet (DT) were instrumented by water electric conductivity sensor (CTD diver Eijkelkamp, accuracy ± 1% μS/cm and resolution 0.1% μS/cm) to observe the tracer arrivals. As concerns TEST2, 7 drains (D119, D122, D138, D152, D186, D196, D273), 3 piezometers (P3, P6, P8) were instrumented by active carbon traps used to fix tracer and observe qualquantitative arrivals of fluorescein in the monitoring points.
Every 15 days, the substitution of activated charcoals was performed, and a groundwater sample was collected at the same monitoring point. Once collected, the fluorescein (if present) is extracted by the carbon-active traps using a potassium hydroxide solution in methanol ). The solutions obtained by the extraction and the punctual groundwater sample were analyzed by an RF-6000 laboratory spectrofluorometer produced by Shimadzu Corporation (Kyoto, Japan) after calibration.

Hydrochemical and isotopic investigations
Groundwater was sampled from drains of the DT, during two campaigns, respectively, on 2015 (sampling campaign A), and 2017-2018 (sampling campaign B). Samples were collected using 500 mL polyethylene sampling bottles, stored at 4 °C and analyzed at Università Politecnica delle Marche Geochemical laboratory using an ion chromatography system (ICS-1000, Dionex, Waltham, MA, USA). One of the sample aliquots was filtered upon sampling through 0.45 um membrane filters and then acidified with 1% of 1:1 diluted HCl. HCO 3 − concentration was determined by acid titration with 0.01 N HCl using methyl orange as indicator. Analyses were conducted to establish major ion concentrations, such as K + , Na + , Ca 2+ , Mg 2+ , NO 3 − , F − , Cl − , SO 4 2− and HCO 3 − . The total relative uncertainty is less than 5% for all compounds (IRSA-CNR 2003). The groundwater geochemical results were then compared to the major ions contents of the aqueduct water provided by the local water management company. During the sampling campaign B an isotopic investigation has been performed to determine the origin of the groundwater captured by the drainage systems. The monitoring phase consisted in a periodic sampling of 13 drains of the DT and a monthly sampling of rainfall water in Montelupone village at 285 m a.s.l. using a rainfall sampler connected to a buried totalizers constituted by 20 L volume tank.
All the water samples for isotope analysis were collected into 50 mL high-density polyethylene bottles and they were tightly sealed to avoid evaporation. Isotopic analyses were performed at the Stable Isotopes Laboratory of National Research Council (CNR-Pisa, Italy) through the analysis of gaseous CO 2 , previously equilibrated with water at 25 °C (Epstein and Mayeda 1953) by mass spectrometry FINNI-GAN MAT252 (analytical precision less than 0.10‰) for the δ 18 O, while the δ 2 H value was measured by means of the Europa Scientific GEO 20-20, by reducing the water to elemental hydrogen (Coleman et al. 1982) using magnesium instead of zinc (analytical error 1.5‰). The "permil" notation is used to represent the isotope results, indicating the water sample ratio with respect to the international standard V-SMOW (Vienna-Standard Mean Oceanic Water) as reported by Rozanski et al. (1993).
Eventually, the mean annual oxygen-18 rainfall value obtained for Montelupone has been integrated with the ones of Monte Dago rainfall station (170 m a.s.l.), derived from the Global Network of Isotopes in Precipitation (GNIP) (IAEA International Atomic Agency 2019), and Portonovo station (15 m a.s.l.) obtained from Mussi et al. (2017). The oxygen-18 vs. elevation relationship has been used to identify the groundwater recharge area mean elevation for each isotopic monitoring point.

Leaching analysis
Three samples from the geological formations outcropping in the area, respectively, AAF2, AAF1 and LD were collected and prepared in the laboratory for batch leaching tests, in accord with the USGS field leach test procedures (USGS 2005). The samples have been collected outside the urbanized area to avoid any further anthropogenic contamination and make a comparison with those measured in groundwater samples. Fifty grams of dried material for each sample was sieved less than 2 mm and leached into 0.5 L of deionized water (liquid-to-solid ratio 100, according to Someya et al. 2021), continuously shaken by a mechanical agitator (with a constant rotation speed of 12 rpm). Each water sample was collected after 24 h of interaction to determine the major ions content (ASTM 2017).

Hydrometeorological features and hydrological balance of the landslide
The rainfall regime of the area is characterized by a seasonal variability with the highest values recorded during the late winter-early spring between February and March (80-100 mm per month). The total annual rainfall is about 744 mm, while the average annual temperature is about 16.6 °C. The highest temperature is recorded between July and August, while the lowest is recorded in the winter months (December-February). The preliminary hydrological balance identified the monthly water surplus occurring only between January and April, until the year 2018 (Fig. 2). Then, starting from May 2018 the water balance calculation evidenced no water surplus until the year 2021. The annual water surplus, if positive, varies between 18 mm (year 2017) and 290 mm (year 2014) with an average annual value of about 104 mm, which corresponds approximately to 15% of the total annual rainfall. Table 1 shows the basic statistics of the meteorological time-series considered in this research (i.e., R: monthly rainfall, T: mean monthly temperature, and the computed WS: monthly water surplus).
Starting from the annual water surplus, the average discharge calculated from the water balance for the period 2014-2018 is 0.39 L/s. This value is determined given an infiltration area of 80,000 m 2 ; the area of 50,000 m 2 investigated by (ISPA 1979), gives an average discharge of 0.25 L/s. Eventually, excluding the impervious area (about 67,000 m 2 ) from the overall area (80,000 m 2 ), a discharge rate of only 0.06 L/s is obtained. For the same time interval, the mean discharge measured at the outlet of the DT is 0.5 L/s. On the other hand, considering the period in which DT discharge rate data are almost continuous (2019-2021), the average measured discharge is 0.15 L/s, while the calculated one is 0 L/s (water surplus = 0). Based on that, an excess of groundwater flowing in the DT is evidenced by this analysis. Following, results of water level monitoring of piezometers P8 and P1 (years 2014-2015) together with P6 (years 2018-2021) are shown in Fig. 3. The water head above the compensated pressure transducers of P8 and P1 piezometers (Fig. 3a), shows a double hydrodynamic behaviour, suggesting the presence of hydrogeological heterogeneities inside the landslide body, in accordance with the occurrence of lithological alternances in the area. The piezometer P1 is characterized by impulsive increases and decreases of the piezometric level indeed, strictly related to the rainfall events in which water infiltrates in the shallower groundwater body, hosted by the silty-clay and clayey-silt deposits of the landslide (LD). On the contrary, the piezometer P8, located in the sandstone predominating geological formation (AAF1), shows a smoother trend, with seasonal variability expressed by the maximum values recorded in the late spring months, and the minimum values recorded in the winter period. From the data set, only precipitation events > 50 mm can generate a sharp increase in the piezometric level, a phenomenon that may be attributed nor to the kinematic pressure transfer generated by the precipitation event (Bailly-Comte et al. 2010), nor to water transfer from the upper to the deeper groundwater body. The same behaviour is still confirmed during the 2018-2019 monitoring period. In fact, the water head level variation of P6 (Fig. 3b), located within the landslide body (LD) shows a similarity with P1. In light of these observations, the hypothesis of the occurrence of two groundwater bodies within a single aquifer, located at different depths, is, therefore, confirmed. The water head fluctuations recorded in the deepest groundwater body by P8 piezometer highlights the hydraulic connection between the groundwater bodies.

Tracer test
Results of tracer tests with an indication of the injection and monitoring points are summarized in Fig. 4a, and 4b. In Table 2, details of velocities (m/s) for TEST 1 and the spectrofluorometric analysis results performed on the active carbon traps for TEST 2 are shown. With regard to TEST 1, two impulsive peaks of salinity increase have been recorded in D3 (Fig. 5a). The first arrival on D3 was recorded after  18 days from the tracer injection, thus resulting in a maximum velocity of about 4.86 × 10 -5 m/s. This evidence is in accordance with the general groundwater flow direction and the presence of a high hydraulic gradient between the tracer injection points and D3 (Fig. 1b).
For what concerns the monitoring points located in the ENE portion (P5 and D196), the maximum tracer velocity is about 2.54 × 10 -5 -3.01 × 10 -5 m/s. The increase of water electrical conductivity due to the tracer arrival in the drains is confirmed by the trend of the water electrical conductivity measured at the outlet point of DT. The tracer is not recorded in D354 (Fig. 4a), and this fact can be attributable to its shallower position in respect to the injection point. D354 involves, in fact, the uppermost portion of LD, approaching the DT outlet.
TEST1 stresses a hydraulic connection between the shallow groundwater body hosted by the landslide deposits LD geological Fm., and the underlying groundwater body hosted by the bedrock (AAF1 geological Fm.). Moreover, the different flow velocities underline the heterogeneity of the lithologies characterizing the area. As far as TEST 2 is concerned, although only qualitative, it confirms what was observed in TEST1: (i) the drainage system plays an important role in directing groundwater northwards ii) the groundwater body feeding the landslide is hydraulically in contact with the one located at the boundary between the deposits and the bedrock in the western portion of the landslide, where the tracer injection points are located (Fig. 4a, b).

Hydrochemistry
The classification of water sampled from drains is reported in the Piper-trilinear plot. According to the Piper diagram (Fig. 6), the water samples are mainly located within the central part of the plot, indicating a mixed type of water with an overall amount of Cl − ranging from 50% to 80%. More in detail, two groups are identifiable: (i) the former with SO 4 2− -Cl − and Ca 2+ -Mg 2+ ranging from 70% to 80%    Table 3. After 24 h of leaching, samples ML2 and ML3 (representing the leaching water coming from the bedrock AAF1 and AAF2 geological Fms.) show slight enrichment in Cl − content, while ML1 (representing water coming from leaching of the landslide deposits LD) shows higher content in K + with respect to ML2 and ML3. The results of leaching analyses have been compared with the chemical analyses of water provided by the local water management company (ASTEA 2021).
Coupling the geochemical data for groundwater samples and the results of leaching tests, an enrichment in SO 4 2− and Cl − of the Group 1 of drains is evidenced, which is even higher both than water coming from the ASTEA aqueduct, both than that derived from leaching analyses. This preliminary result leads to the hypothesis that some samples of water drained by the system may be affected by anthropogenic pollution. The high concentration of Cl − can be attributable to the road de-icing operation during winter periods, whereas the high concentration of NO 3 − can be due to urban sewerage system leaks.

Isotopes
The average groundwater stable isotope data of drains (divided into the two groups identified through the hydrochemical analysis) are reported in Fig. 7 Table 4. With regard to the drains, two main groups are identified: (i) those characterized by a mean recharge elevation ranging from 175 to 208 m a.s.l., which is rather lower with respect to the mean topographic elevation of the Montelupone landslide area (D13, D31, D186, D234, D237, D354); (ii) those displaying a mean recharge area elevation comparable to the mean topographic elevation of Montelupone landslide area (about 225 m a.s.l). The isotopic results and the subdivision in two groups strength the previous hypothesis of possible interaction between groundwater and water derived by sewage system leaking. In fact, the local aqueduct wells system is located at about 140 m a.s.l. and it provides sanitary and potable water to the Montelupone village.

Discussion
To develop the discussion about the obtained results, three geological cross sections (Fig. 8) have been created by integrating the geological knowledge with the data achieved by the combined hydrogeological investigations;   they suggest a detailed hydrogeological conceptual model of the landslide body. Geological cross sections have been obtained by the observation of fourteen direct drill cores results and geological field survey, which allow for the identification of hydrostratigraphic constrains for groundwater conceptual model and permit to make assumptions about groundwater origin and dynamics within the landslide body. The hypothesis made by previous studies (ISPA 1979;Morgoni 2012) about the presence of two different groundwater bodies within the landslide is validated by the groundwater level fluctuation, recorded by the piezometers, located  With regard to the tracer tests, it is important to remark that injection of NaCl during TEST1 is made in P2 and P4 piezometers, in which the fenestrated section is located in the shallow groundwater body (Fig. 8b), whereas the tracer's arrivals is recorded in the drains of the drainage tunnel system, involving the deeper groundwater body. The tracer arrivals in the drains highlight a hydraulic connection between the shallow and the deeper groundwater body, not visible by the observation of the lithology setting alone, which depicts, in fact, small perched sandy lens or sandy layers (the blue dot line in Fig. 8a, c) occurring in the colluvial eluvial deposits (LD and ECD). From a sedimentological point of view, LD and ECD are Synthems, defined as complex sedimentary prisms bounded above and below by regional discontinuity surfaces, deposited in different depositional systems and having a complex stratigraphic architecture. In accord with Irace et al. (2010), the internal stratal pattern of this complex stratigraphic architecture strictly controls the groundwater flow directions and the secondary permeability at basin scale. The analysis of boreholes data, therefore, highlights high heterogeneity of the hydrogeological characteristics of landslide materials. At different depths, in fact, syn-sedimentary slumping and anomalous strata patterns (such as minor-folds near tectonic discontinuities) act as zones of preferential flow, increasing the percolation towards the deeper groundwater body. Therefore, at basin scale, the colluvial-eluvial Quaternary continental cover (LD and ECD Fms.) is considered as aquitard, wherever it is not affected by the peculiar stratigraphic architecture. All this evidence explains the tracer migration and, therefore, the percolation of water from the shallow groundwater body through the deeper groundwater body with a prevalent direction towards the north. The latter is located at the transition between the colluvial-eluvial deposits (ECD) and the bedrock of AAF1 in the western portion of the Montelupone hill, while moving eastwards the saturated zone involves even the deeper portion of the LD Fm. (the dashed blue line in Fig. 8b, c). In this way, the shallow groundwater body is recharged by meteoric precipitation (Fig. 8b, c), as suggested by the piezometric level fluctuation observed in P1 and P6 (Fig. 3a, b), in which a fast response to rainfall events is recorded as piezometric level increase. The deeper groundwater body is recharged by two different contributions: the main is the percolation through the sandy horizons of the bedrock, occurring during the months in which the water surplus is positive (February-April), as revealed by P8, not influenced by single rainfall events, except for daily rainfall higher than 50 mm. The secondary mechanism of recharge is given by slow infiltration process occurring on the western hillslope area through the sandy and clayey loam (ECD geological formation) and through the alternating sandstones and marly clays with predominating sandstones (AAF1 geological formation), driven by a general bedding flat-parallel with ENE and NNE immersion between 10 ÷ 12° (Fig. 8b, c). As concerns the groundwater drained by the drainage tunnel, a decrease in discharge is observed between 2014 and 2021. The water balance model stresses a groundwater discharge value slightly lower than the one calculated from the mean discharge collected by the drainage tunnel for the same period. Although the scarcity of meteoric groundwater recharge is evidenced by the water budget between 2018 and 2021, a surplus of discharge drained by the drainage tunnel is continuously recorded over the years. The discharge calculated considering the area of urban fabric-which is almost totally impervious and consequently it leads to surface runoff instead of infiltration-is always lower than the measured discharged, even if values are quite similar. Moreover, the geochemical analyses and the isotopic oxygen-18 values lead to the identification of two main groundwater groups: (a) the former coming from local meteoric infiltration, for which the oxygen-18 values display a mean recharge area elevation comparable with the Montelupone hill; (b) the latter deriving from mixing with the local aqueduct water (characterized by lower isotopic elevation). The second group shows also high concentration of Cl − and NO 3 − . As reported by Howard and Haynes (1993) and Marques et al. (2019), high Cl − concentration can be attributed to road de-icing salt dumping operations, frequently occurred during the winter periods. Those de-icing chemicals are leached from rainfall and conveyed to the streets sewage system which experienced several breakups during the years (ISPA 1979) and subsequent leakages into the groundwater landslide body. The high concentration of NO 3 − can be connected to urban sewerage system leaks, as supported by previous bacteriological analyses that revealed the presence of a fair number of colonies of Colibacilli and other typical fecal bacteria (ISPA 1979). Similar evidence was observed in other hillslope villages and cities characterized by the presence of landslides, in which the sewage system is often affected by contaminant dispersion in groundwater (Tarcan and Koca 2001;Plumlee et al. 2012;Vivalda et al. 2017). The partition in two different groups observed from the geochemical and the isotopic analyses highlights the following mechanism: where the aqueduct water, showing higher oxygen-18 isotopic values than the local groundwater (as its recharging altitude is lower), interacts with groundwater (e.g., after a leakage in the sewerage system), it yields an enrichment in the isotopic composition together with an increase in Cl − and SO 4 2− content. In this case the isotopic content can be considered as a natural tracer of pollution, alternative to the chemical elements. Wastewater coming from roads runoff and the sewage water of the houses (isotopically enriched) flows in fact into the same sewage system that, due to the breaks caused by the landslide movement, is leached into the groundwater body. Moreover, based on the geochemical composition of the groundwater sampled (Supplement 1), the previous hypothesis made by ISPA (1979) and Morgoni (2012) of shallow groundwater mixing with deep fluid upwelling operated by the normal fault system can be excluded. In fact, as reported by several studies (Nanni and Vivalda 2005;Capaccioni et al. 2001;Frondini 2008;Acero et al. 2015), the presence of a deep regional aquifer interacting with the evaporites and connected to the local aquifers through the fault system of the area, should result in saline fluids strongly enriched in Na-Cl component.

Conclusions
This research represents an all-around hydrogeological investigation of an important urban landslide in central Italy using tracers, isotopes, hydrochemistry, and groundwater in situ monitoring. The outcomes of this study showed that the traditional hydrogeological monitoring techniques alone, such as the hydrological balance, the piezometric levels trend analysis and the hydro-chemical analysis can give only partial information on the groundwater system, misleading the hydrogeological model in complex geological frameworks. To this aim, the use of techniques, such as tracer tests and water isotopic characterization, coupled with the hydro-chemical analysis, can further complete the hydrogeological scenario, and avoid misinterpretations of the conceptual model. In detail, the hydrogeological model proposed by ISPA (1979) and Morgoni (2012), in which two connected groundwater bodies are identified was validated by the groundwater level fluctuation recorded by the piezometers located at different depths and by the artificial tracer test results. At the same time, the hypothesis of deepwater inflow (Morgoni 2012;ISPA 1979) has been rejected based on the hydro-chemical characteristics of the groundwater sampled. The leakage of the sewerage system has been evidenced by multiple observations, including the isotopic signature and the hydro-chemical composition of the sampled water and the leaching tests. The artificial tracer tests clearly support the connection between the shallow aquifer (in which the contaminant dispersion occur) and the deep aquifer, which is drained by the landslide DT. With regard to the groundwater surplus derived by the water balance calculation, is important to note that during scarce groundwater recharge conditions, such as from 2018 (Fronzi et al. 2022), the amount of water coming from the leakage system might be not negligible and it is even larger than water coming from the rainfall recharge. There is no scientific evidence that this source of groundwater could be correlated with instability events or that this water contribution should be accounted for the design of mitigation works. However, this study remarks the importance of investigating the groundwater origin to identify the sources of water feeding the landslides, which in some cases could enhance the instability phenomena and/or alter the groundwater chemical quality.