Origin and assessment of deep groundwater inﬂow in the Ca’ Lita landslide using hydrochemistry and in situ monitoring

. Changes in soil water content, groundwater ﬂow and a rise in pore water pressure are well-known causal or triggering factors for hillslope instability. Rainfall and snowmelt are generally assumed as the main sources of groundwater recharge. This assumption neglects the role of deep water inﬂow in highly tectonized areas, a factor that can inﬂuence long-term pore-pressure regimes and play a role on local slope instability. This paper aims to assess the origin of groundwater in the Ca’ Lita landslide (northern Italian Apennines) and to qualify and quantify the aliquot attributable to deep water inﬂow. The research is essentially based on in situ monitoring and hydrochemical analyses. It involved 5 yr of continuous monitoring of groundwater levels, electrical conductivity and temperature and with groundwater sampling followed by determination of major ions (Na + , K + , Mg 2 + , Ca 2 + , Cl − , HCO − 3 , SO 2 − 4 ) , tracers (such as B tot and Sr 2 + ) , and isotopes ( δ 18 O, δ 2 H and 3 H). Leaching experiments on soil samples, hydro-chemical modelling and water recharge estimation were also carried out. Results show that the groundwater balance in the Ca’ Lita landslide must take into account an inﬂow of deep and highly mineralised Na-SO 4 water (more than 9500 µS cm − 1 ) with non-negligible amounts of Cl − (up to 800 mg l − 1 ) . The chemical and isotopic ﬁngerprint of this water points to oil-ﬁeld water hosted at large depths in the Apennine chain and that uprises through a regional fault line crossing the land-slide area. It recharges the aquifer hosted in the bedrock underlying the sliding surface (at a rate of about 49 000– 85 700 m 3 yr − 1 ) and it also partly recharges the landslide body. In both the aquifers, the hydrochemical imprint of deep water mixed with rainfall and snowmelt water was observed. This indicates a probable inﬂuence of deep water inﬂow on the mobility of the Ca’ Lita landslide, a ﬁnding that could be applicable to other large landslides occurring in highly tec-tonized areas in the northern Apennines or in other mountain chains. The paper demonstrates that hydrochemistry should, therefore, be considered as a valuable investigation method to deﬁne hydrogeological limits and the groundwater sources in hillslope and to assess groundwater ﬂow patterns in deep-seated landslides.


Introduction
Instability of hillslopes is generally triggered by hydrological and hydrogeological factors governing infiltration, increase of pore water pressure and resulting decreases in effective stress in the soil (Wieczorek, 1996;van Asch et al., 1999).In deep-seated landslides, effective groundwater infiltration on a slope scale over long periods of time can increase hydrostatic levels and determine groundwater flow, thus, playing an important role in the reactivation of slope movements (Hutchinson, 1970;Iverson and Major, 1987;van Asch et al., 1999).
However, the activation/reactivation of deep-seated landslides is a complex issue and groundwater recharge is not always due to precipitation alone.For instance, deep water upflow along regional tectonic structures can affect groundwater balance on a slope scale (Tóth, 2009).The presence of deep water into landslide deposits was reported in several cases in the northern Apennines (Colombetti and Nicolodi, Published by Copernicus Publications on behalf of the European Geosciences Union. 1998; Bertolini and Gorgoni, 2001;Ciancabilla et al., 2004;Baraldi, 2008;Ronchetti et al., 2009) as well as in other mountain chains (Bonzanigo et al., 2001;de Montety et al., 2007).
Deep water can easily be detected by hydrochemical surveying (Bogaard et al., 2007) due to its very distinct chemical imprint depending on depth, temperature and pressure conditions, the mineral composition of hosting rocks or deposits, time of interaction between water and aquifer, and the mixing of different water types (Freeze and Cherry, 1979).However, differentiation of water types in hillslope hydrology is not often conducted (Guglielmi et al., 2000;Cappa et al., 2004).This paper deals with the analysis of groundwater in the Ca' Lita landslide, a large rotational rock slideearth flow affecting highly tectonized flysch rock masses in the northern Italian Apennines.The landslide occurs along a major regional fault line, through which deep water inflow was proven to occur (Ronchetti et al., 2009).The aim is to assess the origin of groundwater below and inside the landslide, and to qualify and quantify the aliquot of groundwater from deep water inflow.The research makes combined use of groundwater monitoring and chemical/isotopic analyses.The results allow an assessment of the contribution of deep water to the hydrological processes and development of instability involved in the landslide.

Hydrogeological and hydrochemical settings of the northern Apennines
The northern Apennines (NA) are a fold-and-thrust mountain belt generated by the closure of the Ligure-Piemontese Ocean basin and the subsequent collision of the Adria and European continental plates (Boccaletti et al., 1971;Kligfield, 1979;Bettelli and Vannucchi, 2003;Molli, 2008).As a consequence of the polyphasic evolution of the accretionary wedge, the northern Apennines are made up of several tectono-stratigraphic units of marine sedimentary rocks, many of which are turbidite sequences (flysch rock masses) and clayey chaotic deposits (clayshales) (Fig. 1).Therefore, the hydrogeological setting of the northern Apennines is largely dominated by low permeability formations, which in many cases are to be considered as aquicludes unless they are affected by regional tectonic features inducing significant secondary permeability (Gargini et al., 2008).Limestone or gypsum formations acting as aquifers, that are quite common in the Alps, are indeed quite sporadic in the northern Apennines and they are prevalently exposed in the higher portion or at the front of the chain, far away from the case study where, in turn, permeable hydrogeological units are constituted by fractured and faulted turbidites-like formations.More specifically, the Tuscan Units (TU), the Ligurian Units (LU) and the Sestola Vidiciatico Unit (SV) are mainly composed of thick and highly tectonized turbidite sequences (flysch rock masses) and clayey chaotic deposits (clayshales).Within these units, Triassic evaporites (TUG gypsum) are also found.The Epiligurian Units (EL) are mainly composed of turbidites (flysch rock masses).The Messinian and Post Messinian units, are made up of evaporites (MME gypsum) and marine clayey rocks (QM) outcropping at the front of the mountain chain.The main hydrogeological characteristics of these units and the typical hydrochemical imprint of hosted groundwater are described below.

Tuscan Units (TU)
Groundwater circuits developing within TU flysch rock masses are mostly shallow and widely distributed.The unconfined aquifers feed a large number of low-yield springs which discharge where the groundwater  Tóth, 1999) has recently been identified by Gargini et al. (2008): groundwater slowly (from 10 −12 to 10 −14 m s −1 ) flows from the upper NA chain toward the Po Plain.Supplied by rainfall and snowmelt water infiltrating near the watershed divide, the BRS is driven by regional gradients, but in some cases tectonic lineaments or topography concentrate discharge in isolated springs or directly into stream beds.Systematic changes in the anion facies have been reported: fresh infiltrated waters start from HCO 3 (point no. 5 in Fig. 1a, b) and pass through SO 4 (point no.7 in Fig. 1a, b) finally reaching Cl (point no. 4 in Fig. 1a, b), while TDS (Total Dissolved Solids) increases up to 9-10 g l −1 .pCO 2 constantly increases up to 2.0 kPa.Water stable isotopes (δ 18 O and δ 2 H) range within precipitation values, shifted towards negative values, which characterise the higher recharge areas (δ 18 O up to −10.5 ‰ to −11 ‰ while δ 2 H can reach −75 ‰: Gargini et al., 2008).In the longest and deepest flow paths (main anion: Cl − ), accumulation of trace ions, such as B tot and Sr 2+ and transported hydrocarbons and/or immiscible fluids (oil, gas) were noted (point no. 4 in Fig. 1a, b: Minissale et al., 2000;Capozzi and Picotti, 2002;Heinicke et al., 2010).In several cases Sr 2+ and B tot contents are strongly related; B tot , which can be mobilised from the organic matter buried during sedimentary processes, is normally associated with oilfield water if its value is in the order of 100 mg l −1 (White, 1965) while Sr 2+ ranges from 100 to 500 mg l −1 (Conti et al., 2000;Capozzi and Picotti, 2010).

Post Messinian Units (QM)
Along the boundary with the Po plain, several mud volcanoes (points 3a, b in Fig. 1a, b) affect the impermeable QM.These phenomena, which are widely observed along compressive margins around the world (Martinelli and Judd, 2004;Martinelli and Dadomo, 2005), consist of Na-Cl waters (up to 90 g l −1 ) flushing out particles of clay and entrapped connate water of the old Pliocene sea (Bonini, 2007).They originate from the buried QM-TU contact (located at a depth of more than 2000 m), whereas oilfield fluids start to rise up towards the surface following inverse faults.Stable isotopes ( 18 O and 2 H) are sometimes highly positive (δ 18 O up to +5 ‰ and δ 2 H equal to 0 ‰) while pCO 2 varies widely (0.03-0.9 kPa).B tot and Sr 2+ can be as high as 300 mg l −1 (Boschetti et al., 2010).

Ca' Lita landslide settings
The Ca' Lita landslide is located between 640 and 230 m a.s.l. in the NE mountainside of the northern Apennines (coordinate: Lat.44.458657 • , Long.10.626566 • , Aspect 135 • North).Over the last 40 yr, the average annual rainfall in the area was 810 mm yr −1 and the mean air temperature was 11.7 • C (ARPA, 2010).Precipitation is distributed over 70 to 100 days yr −1 with two peaks: a main one in autumn (October-November) and a secondary one in spring (April).Fazlagic et al. (2004) report an average annual snowfall of about 42 cm from 1830 to 1998.The Ca' Lita landslide affects a hillslope composed of formations belonging to the Ligurian Units (LU).One formation consists of poorly cemented sandstone flysch (MOH), another is composed of clayshales (MVR) (Fig. 2a, b).In the slope, the stratigraphic relation between MOH and MVR is complicated by the presence of a high angle regional fault (Papani et al., 2002) that is linked to a regional faults system (Val Rossenna fault as reported by Bettelli et al., 2002;Fig. 1a).The landslide can F. Cervi et al.: Origin and assessment of deep groundwater inflow be classified as a reactivated complex landslide (WP/ WLI, 1993;Cruden and Varnes, 1996), associating rotational rock slides in the crown and head zones (in the MOH formation) with earth flows in the lower main body (in the MVR formation).The total length is 2.7 km with a mean slope angle of 15 • and a total estimated volume of 40 Mm 3 .The area involved by the retrogressive rotational slide movements is 0.5 km 2 and it extends from 640 to 475 m a.s.l.over a length of 525 m (Fig. 2a, b).It was reactivated several times in the last century (Borgatti et al., 2006;Corsini et al., 2009).Recently, it resumed activity in 2002 and underwent paroxysmal phases in winter 2003 and spring 2004 due to 200 mm of cumulated rainfall in the two preceding months combined with the rapid melting of about 80 cm of snow.After the 2004 event, an extensive geological investigation and monitoring campaign was set up in order to design mitigation structures (Corsini et al., 2006).Boreholes and inclinometers showed that the active rock slide deposit had a maximum thickness of 44-48 m while the active earth flow deposit had a mean thickness of about 10 m (Borgatti et al., 2006).Since summer 2006, no further deep seated rock slide movements have been recorded and only shallow slides have occurred close to the main scarps.Later, inclinometers and 5 standpipe piezometers were installed and equipped with pressure transducers.Electrical conductivity probes were also installed inside two of these.Piezometers were used in the research as groundwater monitoring and sampling points.Other monitoring and sampling points were drainage wells and subhorizontal drains constructed during mitigation work at different locations on the slope (Fig. 2a, b).
Deep drainage wells in the rock slide (Borgatti et al., 2008) drain groundwater from the landslide body and convey it to well A (WA in Fig. 2a, b), from where it is removed by pumping.At the base of the frontal scarp of the rock slide zone, a shield of shallow drainage wells drain water from the superficial debris and convey it to drainage point B (DrB in Fig. 2a,  b).In the same area, a set of sub-horizontal drains of length from 50-100 m drain water from the bedrock below the sliding surface and convey it to drainage point A (DrA in Fig. 2a,  b).
The in situ characterisation performed by Ronchetti et al. (2009) highlighted high heterogeneity of the hydrogeological characteristics of flysch bedrock and of landslide materials, which are affected at different depths by discontinuities that act as zones of preferential flow.In situ slug tests and pumping tests were used to estimate the hydraulic conductivity of the different units.The flysch bedrock has a hydraulic conductivity in the range 10 −6 /10 −7 m s −1 , the main landslide body (made of disarranged flysch rock masses) 10 −5 /10 −6 m s −1 and the clayshales less than 10 −8 m s −1 .According to Civita (2005), the infiltration coefficient (I c ) ranges from 0.4 to 0.7 for flysch and the main landslide body and can be considered negligible for the clayshales.

Methods
With reference to the case study, an interdisciplinary investigation was performed.It consisted of several steps involving hydrological (groundwater level and discharge monitoring, estimation of the groundwater balance for several hydrologic years), hydrochemical (groundwater conductivity monitoring, groundwater isotopic and chemical analyses, leaching experiments, hydrochemical modelling), and mineralogical (soil composition) surveys.More specifically, the groundwater chemistry and stable isotope contents were characterised in order to assess end-members and mixing phenomena.Based on existing local databases, the reference isotopic values (δ 18 O and 3 H) for rainfall recharge water were defined.An independent validation, to check if the peculiar water chemistry could be due only to the mineralogy of the site, consisted in leaching experiments with soil samples belonging to the two outcropping geological formations.In addition, equilibrium-based model PHREEQC (Parkhurst and Appelo, 2004) was defined in order to simulate the long-term chemical interaction between rainfall water and host rocks.The aliquot of deep water was estimated by coupling radioactive isotopes content with groundwater balance.

Groundwater level, conductivity and discharge monitoring
The 5 standpipe piezometers are slotted at different depths and monitor groundwater in the bedrock at the crown and along the side of the rock slide (P1, P2), inside the rock slide body (P3), and across the deepest sliding surface (P4, P5) (Fig. 2a, b).The characteristics of the boreholes are reported in Table 1.
They have been equipped with pressure transducers since 2005 in order to monitor Groundwater Levels (GW L ) with an acquisition frequency of 1 h.Groundwater Electrical Conductivity (GW EC ) was measured from June 2009 to March 2010, with an acquisition frequency of 6 h in P3 and P4.In the other piezometers, GW EC and Groundwater Temperatures (GW T ) were measured periodically from 2006 to 2009 and from 2010 to date.
The discharge of groundwater drained by the wells from inside the rock slide body was monitored from 2007 to 2010 by means of a graduated weir located at the outlet of the WA pumping system.The discharge of groundwater drained by sub-horizontal drains from the bedrock underneath the main rock sliding surface was monitored from 2006 to 2009 by means of a graduated weir located at the outlet of DrA.In WA and DrA, the GW EC and the GW T were measured seasonally from 2006 to August 2010.The physical-chemical parameters were checked using a Crison MM40+ multimeter equipped with a Ross glass electrode for pH.

Groundwater chemical analyses
Groundwater sampling was conducted in 2006 (sampling campaign A), 2007 and 2009 (sampling campaigns B and D; red square in Fig. 3b).All the 5 standpipe piezometers and both DrA and DrB were sampled.All samples were collected using bailers except samples collected during summer 2009 from DrA and P1, which were collected using a low-flow pump (0.1 l s −1 ) after removing the standing water in the piezometer.Water for laboratory study was filtered through 0.45 µm cellulose membranes, and the aliquot for cation analysis was acidified with 65 % HNO 3 Suprapur Merck.Analyses were conducted to establish major ion concentrations (K + , Na + , Ca 2+ , Mg 2+ , Cl − , SO 2− 4 and HCO − 3 ) as well as the B tot and Sr 2+ contents.Cation contents were assessed by atomic absorption spectrometry (Spectr AA-640, Varian).Anions were determined using an HPLC (high performance liquid chromatography, Dionex DX-120).Total alkalinity was assessed by Gran titration (Gran, 1952).Data are reported in mg l −1 (Table 3).The total relative uncertainty is less than 5 % for all compounds.

Groundwater isotopic analyses
Stable oxygen and hydrogen isotope analyses (δ 18 O, δ 2 H) were carried out by mass spectrometry on water samples collected from all the piezometers and DrA.Five sampling campaigns scattered over 5 yr were conducted starting from 2006 (sampling campaigns A, B, C, D, E, F; yellow diamond in Fig. 3b).The samples represent the summer (A, B, D, F), winter (C), and spring (E) periods.The results are reported as differences between the sample and the standard (Vienna Standard Mean Oceanic Water: V-SMOW).This deviation is presented in the standard δ-notation as per mil (‰) where δ = [(R S /R SMOW )−1]1000; R S represents either the 18 O/ 16 O or the 2 H/ 1 H ratio of the sample, and R SMOW is 18 O/ 16 O or the 2 H/ 1 H ratio of the SMOW.The precisions are in the order of ±0.05 ‰ (δ 18 O) and ±0.7 ‰ (δ 2 H).Deuterium excess values were calculated according to Dansgaard (1964) Tritium analysis ( 3 H) of groundwater was carried out on samples collected in August 2010 from DrA (sampling campaign F; green diamond in Fig. 3b).Water samples from DrA were selected as representative of water that can discharge, by natural or artificial drainage, from the deepest and longest circulation path developed through the slope.Analysis was performed using the electrolytic enrichment and liquid scintillation counting method (Thatcher et al., 1977).Results were expressed as Tritium Units (T.U. in which one T.U.equals one tritium atom per 10 18 hydrogen atoms).Analytical errors are in the order of ±0.2 T.U.With reference to rainfalls in northern Apennines, Longinelli and Selmo (2003) reported a mean annual δ 18 O value in the period 1993-2002 around −7 ‰ to −8 ‰ along the Po Plain (about 30 m a.s.l.) increasing up to −10 ‰ near the highest relief (about 2100 m a.s.l.) with a mean altitude effect of about −0.2 ‰ δ 18 O/100 m.By considering the years 2006-2007-2008, Cremaschi (2008) ) found an average isotopic gradient of about −0.1 ‰ δ 18 O/100 m for the same area.This is in agreement with Longinelli et al. (2006), which reported more depleted values of δ 18 O in precipitations over the Po Plain areas during years 2002, 2003 and 2004 which were characterised by high temperatures.Considering these altitude gradients, a mean annual δ 18 O value for precipitations in Ca' Lita should not be more negative than −8.5/−9 ‰.Anyway, Iacumin et al. (2009) reported the δ 18 O values for shallow groundwater (sampled in wells and springs) and rivers in northern Apennines.In rivers, monthly δ 18 O observations from 2004 to 2006 varied from −9.8 ‰ (winter) to −6.8 ‰ (summer).In shallow groundwater, monthly δ 18 O observations varied between maximum of −8.0 ‰ to minimum of −8.7 ‰.The authors concluded that springs and wells provided smoothed records of δ 18 O compared to the river and the isotopic values are shifted towards more negative values.This is due to a homogenisation effect made by the flysch aquifers on the isotopic signal of rainfalls, together with the inability of summer rainfalls to reach the groundwater.Similarly, Minissale and Vaselli (2011) reported for central Italy that springs can be used to provide a mean annual isotopic value of rainfall recharge water.Based on Iacumin et al. (2009), a mean annual δ 18 O value for the Ca' Lita landslide could be even more negative than those reported taking into account the mean isotopic gradients.Given these uncertainties, and considering that some of the samples collected in Ca' Lita might actually represent rainfall water just after a shallow infiltration, the reference δ 18 O of rainfall water was estimated graphically in a δ 18 O-δ 2 H plot (as suggested by Gibson et al., 2005) as the intersection point between the line defined by the groundwater samples in Ca' Lita and the meteoric line for north Italy proposed by Longinelli and Selmo (2003).Other springs with similar elevation recharge area of Ca' Lita were sampled and analysed to provide further comparison and the results confirmed the reference δ 18 O value of −9.90 ‰.
With respect to reference 3 H values for rainfall water, the ARPA Emilia-Romagna 3 H data (ARPA, 2009) was considered as representative of rainfall recharge water.This database includes three-monthly 3 H values for water samples collected in the hydrologic year 2008-2009 just a few km north of the Ca' Lita landslide.The 3 H dataset from ARPA can be considered an isotopic marker of rainfall.It refers to water sampled in a very shallow well (depth of 20 m) in an unconfined gravel aquifer which is directly recharged by Secchia river water.This 3 H record, obtained using the same analytical method used in Ca' Lita, shows that T.U. is on average of 9.8 T.U., with small variability throughout the year.This value is comparable to the monthly weighted average of 8.4 T.U.measured by Tazioli (2011) in the 2008-2009 period in a rain-gauge located in the eastern-sector of the Apennines.The slight discrepancy between the 3 H rainfall-recharge values measured by ARPA (2009) and Tazioli (2011) can be explained as a result of the different atmospheric/precipitation regimes between coastal and continental stations (Ferronsky and Polyakpov, 2012) which results in  Martinelli (2005).

Leaching analyses
Four soil samples from the two different geological formations (MVR and MOH) were collected from in-situ outcrops (about 7 km south of Ca' Lita, Fig. 1a: MVRb and MOHb) and regolith (i.e., weathered material still in its original location at Ca' Lita, Fig. 1a: MVRa and MOHa).According to the USGS field leach test procedures (USGS, 2005), 50 g of dried material for each sample was sieved less than 2 mm and leached into 1 l of deionized water, continuously shaken by a magnetic agitator (with a constant rotation speed of 200 rpm).The experiments extended over a maximum of 10 h while an electrical conductivity probe was used to monitor water mineralisation on a fixed time-table (30 s, 1 min, 2 min, 5 min, 10 min, 30 min, 1 h, 10 h).Each water sample was collected after 1 h of interaction to determine the major ion contents.

Mineralogical analyses and PHREEQC modelling
The mineralogy was investigated using 2 soil samples collected at Ca' Lita (MOHa, MVRa).The analysis of the < 2 µm fraction was conducted using X-ray diffractometry on oriented paste.Starting from these two mineralogical assemblages, the PHREEQC model (Parkhurst and Appelo, 2004) simulates the interaction (precipitation and dissolution of phases) with an aqueous solution (specifically: rain water with a mean temperature equal to the mean annual air temperature of the site, i.e., 11.7 • C and pH = 5.8) until equilibrium is reached.

Isotopic groundwater balance
The mean annual groundwater recharge from rainfall was estimated using rainfall and temperature data over the last 40 yr derived from available official ARPA Emilia Romagna rainfall databases (ARPA, 2010) for the nearest weather station (Baiso), located 4.5 km northwest of Ca' Lita with the same elevation and aspect.The Thornthwaite and Mather (1957) formula was used to assess the potential mean annual evapotranspiration.Total groundwater recharge volume was calculated by multiplying the water surplus for the slope recharge area.The mean annual deep water inflow was estimated by   solving a simplified annual water-balance equation and by processing the3 H data as described by Mazor (1997).In particular, the aliquot of deep water stored inside the aquifer (groundwater aliquot a) was estimated by considering a 3 H of deep water ( 3 H dw ) equal to 0 T.U., starting from Eq. ( 1): where 3 H r is the tritium content of rainfall recharge water (groundwater aliquot b) and 3 H f is the tritium content of groundwater discharged (naturally or artificially) from the flysch aquifer sampled in DrA (groundwater aliquot c).Subsequently, the groundwater aliquots a and b were used to estimate the theoretical δ 18 O of deep water recharge (δ 18 O dw ) on the basis of Eq. ( 2): where δ 18 O r and δ 18 O f are the oxygen isotope values for rainfall recharge water and flysch groundwater, respectively.In terms of annual volumes, aliquot V dw corresponds to the volume of annual deep water inflow, while the aliquot V r corresponds to volume of annual rainfall recharge and aliquot V f corresponds to the annual volume of water discharged (naturally or artificially) from the flysch aquifer.In these terms, Eq. ( 3) can be written as: Considering a steady-state budget (V IN = V OUT , so that V r + V dw = V f ), Eq. ( 3) can be rewritten as: Finally, the volume of annual deep water inflow (V dw ) can, therefore, be calculated as: where I r is the relative δ 18 O isotopic ratio.
The calculation is affected by uncertainties related to the estimated annual rainfall recharge volume V r (in the order of 20-25 % according to Fetter, 2001), and by uncertainties related to measured or calculated isotopic values (the latter assessed using the common error propagation methods reported in Taylor, 1997).

Groundwater monitoring
The main groundwater recharge into the slope takes place in the autumn and spring when the effective rainfall, i.e., water surplus (Thornthwaite and Mather, 1957), is available.Similarly to the rainfall pattern, GW L show two peaks: one in November-December and another in April (GW L patterns are reported in Fig. 3a).These peaks are normally followed by groundwater regression periods; the main one between May and October.In the crown area (P1), the groundwater response delay after rainfall events is in the order of a few hours up to a couple of days, while the GW L can increase by more than 5 m.These GW L peaks then discharge completely within 5-10 days.The maximum seasonal amplitude is about 7 m.The flysch rock mass in the crown is intensively fractured (P2) and behaves differently.GW L variations for each event are modulated or even absent (as the consequence of high hydraulic conductivity).However, during an intensive rainfall-recharge period, an increase of more than 10 m was detected.In the head zone, above the main sliding surface (P3), GW L are affected by the high-frequency pumping in WA; as a result, they are stabilised at about 29 m from the ground surface.In the deepest piezometers, which monitor deep multi-compartmented horizons (P4, P5: since the records are overlapping, only the second is represented in Fig. 3a), GW L are 20 m higher than the GW L observed inside the main landslide body (P3).Only the most intense precipitations (such as 150 mm on 10 June 2008) can influence the GW L record (total head increased by 1 m after 1 month) while seasonal peaks are modulated (2-4 m) and shifted by some months compared to the rainfall records.
Despite GW EC continuous monitoring only are available for the head zone, different trends can be observed (Fig. 4).Above the sliding surface (P3), a continuous decrease in GW EC is evident.Starting from 5450 µS cm −1 , it fell to a final value of around 5100 µS cm −1 with a minimum of 4570 µS cm −1 on 22 October 2009.This drop was the result of automatic pumping switching-on in WA and the consequent downflow of less mineralised water coming from the shallow part of the landslide body.A new equilibrium was reached only after 3 months (5100 µS cm −1 ).In summer 2011, due to continuous pumping, GW EC reached a new minimum value of 1247 µS cm −1 .For the same time window, GW EC measured in P4 increased from 5500 to 5800 µS cm −1 with a smoothed peak of more than 6000 µS cm −1 in November 2010.The latter was preceded by 300 mm of effective rainfall over 3 months, which provided slight GW L increases even in the deepest piezometers (P2).The discharge monitoring network started to collect data in 2006-2007 and was stopped in 2009 (WA) and so only the hydrologic year 2007-2008 is available for further comparison, when a total of 30 000 m 3 was drained from WA and 6000 m 3 from DrA.

Groundwater chemical analyses
Groundwater salinity ranged between a minimum value of about 1.5 g l −1 measured in the upper part of the slope (P1), to 6 g l −1 measured in the lower part of the slope (DrA, sampling campaign F).According to Drever (1997), the latter can be classified as salt (brackish) water.These waters are enriched in Na + and SO 2− 4 with substantial amounts of HCO − 1.4 kPa (DrA), while pH fluctuated around 7 (slightly acid in the upper part of the slope: 6.6-6.9 observed in P1).
The most mineralised water collected at Ca' Lita during sampling campaign A (DrA) is reported in a Piper diagram together with other GW of NA (Fig. 5a).Whereas common shallow GW falls on the left side of the diamond, the Ca' Lita sample plots in the right corner (Na-SO 4 GW-type).This hydrofacies characterises all water sampled at Ca' Lita (Fig. 5b), with P4 shifted downwards owing to the higher HCO − 3 content.Regarding the main ion contents (reported in Table 3), P3, P5 and DrB had the lower values for both ions (Na + of about 500-600 mg l −1 and SO 2− 4 800-1000 mg l −1 ) while the other samples exceeded 800 mg l −1 and 1000 mg l −1 , respectively.Marked discrepancies were detected in the Cl − contents.The head zone (P3, P4, P5) is characterised by lower levels of Cl − (less than 90 mg l −1 ).The piezometers inserted inside the flysch rock masses (P1, P2) reached 300-400 mg l −1 with a maximum of 790 mg l −1 in DrA.
A comparison between sampling campaigns A, B and D was only possible for P1 and DrA.In 2009 (sampling campaign D) the Cl − contents were halved, while Na + and SO 2− 4 slightly increased.The most pronounced increase involved DrA: in this case, Na + increased to about 1600 mg l −1 while SO 2− 4 tripled (more than 3000 mg l −1 ).This fact is responsible of the upward displacement of the last DrA sample in the Piper diagram (Fig. 5b).Trace ion levels are available only for the last sampling campaign: B tot concentrations reached 8.2 mg l −1 (DrA) and 6.4 mg l −1 (P1); Sr 2+ varied from 6.8 (DrA) to 3.3 (P1) mg l −1 .

Groundwater isotopic analyses
The isotopic values δ 18 O and δ 2 H from groundwater samples collected in Ca' Lita are summarised in Table 3 and are plotted in Fig. 6 together with the meteoric water line proposed by Longinelli and Selmo (2003).Several samples   , in summer 2006).It was also noticed that as soon as the deep drainage mitigation system made of wells began to drain the landslide body, δ 18 O in DrA became more negative (−6.93 ‰ in summer 2010).In the δ 18 O-δ 2 H plot (Fig. 6), samples from P2, P1 and DrA are aligned along a δ 2 H line which has a lower inclination that of the meteoric line.The closest point to the meteoric line is P2, while the most δ 2 H depleted points refer to samples from DrA (from −31.04 ‰ to −53.54 ‰).Samples from other piezometers (P3, P4, P5) plot close to the local meteoric line.Following the approach of Gibson et al. (2005), the intersection between the alignment of samples from Ca' Lita and the meteoric water line can be considered as representative of δ 18 Oδ 2 H rainfall water content.This gives δ 18 O of −9.90 ‰, which is coincident to the lowest value measured in springs located in the Secchia river valley in the vicinity of Ca' Lita area (see springs labeled as Group 5 in Table 3, which are located in Fig. 1a).The lowest value of δ 18 O from Ca' Lita samples falls on the meteoric water line and was obtained with one of the samples collected in P2, which indicates rapid infiltration in the vicinity of the sampling point.In the d-δ 18 O plot (Fig. 7), Ca' Lita samples are distributed along a line linking progressively more negative deuterium excess values.This alignment further evidences that a substantial differentiation exists between isotopic values of groundwater and of rainfall recharge water.Finally, the 3 H content measured in groundwater sampled from DrA during summer 2010 was 3.5 T.U., a value that is significantly lower than the 3 H of 9.8 T.U.considered as reference for rainfall recharge water.(Longinelli and Selmo, 2003).

Leaching analyses
The results of the leaching experiments are summarised in Table 4.The time required to reach the final concentration was generally short for all the samples: at least 70 % of the total chemical content was achieved within 10 min, and 90 % within 1 h.Clearly, the 1 h samples are representative of the lixiviation interactions between soil and water.Once leached, all the MOHa and MOHb provided Ca-HCO 3 water with similar mineralisation (close to 200 µS cm −1 after 1 h).Otherwise, appreciable differences were detected between MVR samples.In particular, an increase of mineralisation (up to 930 µS cm −1 while the furthest sample MVRb rearound 180 µS cm −1 ) was noticed in MVRa.Starting from the Ca-HCO 3 water of the former, the samples leached with MVRa soil collected along the main fault were enriched with Na + and SO 2− 4 up to 160 and 175 mg l −1 , respectively.These results are only partially in agreement with the chemical characteristics of the water hosted in the deepest aquifer of Ca' Lita (DrA).In this case, unlike the water collected from DrA, only a few mg l −1 of Cl − were detected during the leaching experiment.
In the PHREEQC modelling, two different scenarios were tested according to the state of the hydrogeological system:  in the first, it was considered as closed to the surface (pCO 2 reached values of 2.5 kPa and 3 kPa, respectively), while in the second a fixed partial pressure of CO 2 was guaranteed (pCO 2 equal to 0.03 kPa).In both cases, some ions showed discrepancies even with the results obtained from the leaching experiment.Mg 2+ was not present in the final solution (<0.1 mg l −1 ) while Ca 2+ was assessed as around 2-4 mg l −1 .Na + content was increased through albite dissolution, reaching 100 mg l −1 (closed system) and some thousands in the second case.Regarding anion contents, only HCO − 3 (more than 100 mg l −1 ) could be produced by these assemblages of minerals while no SO 2− 4 was detected.

Origin of groundwater and deep water inflow
Results suggest the existence of a complex hydrogeological system comprising overlapping aquifers characterised by the presence of Na-SO 4 GW.In the landslide body (head zone), the aquifer hosted above the main sliding surface (P3) exhibits an unconfined behaviour (Fig. 2b).Its response to rainfall events is fast.The same patterns can be noticed in GW EC : values range between 1000 to 5000 µS cm −1 .Below the sliding surface (P4 and P5), in the thick deformation band made up of disarranged flysch, the same aquifer becomes semi-confined (or multi-compartmented). GW L have no direct relationships with rainfall as only modulated (changes less than 5 m) seasonal variations are observed.P4 had higher GW EC values and ion contents, together with less depleted values of δ 18 O (between −8.51 ‰ and −8.73 ‰ compared to −9.22 ‰ observed for P5).
However, in both cases the chemical imprints are quite similar and characterised by low Cl − contents (less than 88 mg l −1 ).All the water samples taken lie on the local meteoric isotopic line, and the most depleted values of δ 18 O (close to −9.22 ‰) can be assumed to be strongly linked to infiltration processes.The aquifer hosted in the undisturbed flysch (P1 and DrA) and in the side crown (P2) behaves differently.Depending on the denser and more persistent fracturing pattern characterising the crown zone, GW L variations in P1 are higher than those detected in P2.However, both piezometers are characterised by water depleted in δ 2 H together with high contents of Cl − (300-400 mg l −1 ).The latter reaches about 800 mg l −1 at the bottom of the aquifer (DrA).Discharged water is the most mineralised (GW EC up to 9500 µS cm −1 ).
A sharp decline in chemical parameters occurred during November 2009: GW EC in the landslide body repositioned at about 1500 µS cm −1 (P3).Due to the continuous pumping in WA, the unconfined aquifer hosted in the head zone was progressively renewed by the inflow of more fresh and less mineralised water from the upper landslide body.This presence was confirmed by Ronchetti et al. (2009) after conductivity loggings in P1: the authors detected a rapid increase of GW EC with depth (stabilised at about 4000 µS cm −1 at 20 m below GW L ).The undisturbed flysch aquifer suffered by the hydrological changes of the head zone: in sampling points P1 and DrA, Cl − contents were halved.
A change in hydrochemical characteristics following continuous pumping in WA was also observed in the δ 18 O record.In DrA, δ 18 O decreased from −1.61 ‰ in July 2006, to −5.33 ‰ in July 2007 up to −6.93 ‰ in August 2010.This can be interpreted as the result of an increased contribution of lighter rain water infiltrating into bedrock from the upper landslide body, thus, producing a mixture of rainfall recharge water and deep groundwater.As previously reported, a δ 18 O = −9.90‰, can be considered as representative rainfall recharge isotopic content (point of intersection between the alignment of samples from Ca' Lita and the meteoric water line).On the other hand, an indicative value for the mean annual δ 18 O of deep groundwater of about +3 ‰ can be estimated by solving Eq. ( 2) described in Sect.3.7.An estimated value of δ 18 O of about +3 ‰ for deep water inflow is consistent with δ 18 O values measured in mud volcanoes located 15 km to the north (points no.3a, b in Fig. 1a) (Martinelli et al., 1989) as well as in other springs flowing out of TU flysch formations (i.e., Well SalsMag1 in Boschetti et al., 2010) or from LU-TU formations in the middle-frontal part of the Apennine chain (i.e., Regnano site point 3c in Fig. 1a; Capozzi and Picotti, 2010).
The hypothesis that the isotopic differentiation between groundwater sampled in Ca' Lita and rainfall recharge water is likely to be due to the inflow of a certain aliquot of deep water, is corroborated by the consideration that the influence of evaporative processes on δ 2 H depletion and enhancement of δ 18 O values can most likely be excluded on the basis of a number of reasons.First, the slope of the regression line linking groundwater samples in the δ 18 O-δ 2 H plot is 3.6 (see Fig. 6).This is lower than those reported by Celico (1986) and Kendall and Caldwell (1998) in similar environmental contexts (3.9 to 6.8 with humidity of 0 % to 95 %, respectively).Furthermore, a distribution of Ca' Lita samples along a line linking progressively more negative deuterium excess values is evident in the d-δ 18 O plot (Fig. 7).Evaporation only occurs if the groundwater table is close to the surface (Mazor, 1997;Kendall and McDonnel, 1998) whereas the GW L depth in Ca' Lita ranges from −5 to −30 m and groundwater temperature GW T is generally stable at around 13.5 • C throughout the seasons, showing no influence of atmospheric temperature variations.It should be considered that the most depleted δ 2 H and the higher δ 18 O values refer to DrA, which collects water at depths greater than 50 m, making evaporation very improbable.In addition to that, DrA conductivity and δ 18 O values for three different sampling campaigns (see B-D-F in Table 3) are available.As reported by Gonfiantini et al. (1974) and Simpson et al. (1987) evaporation process should lead to positive correlation between salinity and δ 18 O in the residual water.On the contrary, this is not occurring in DrA, where samples showing the most enriched values of δ 18 O are characterised by lower conductivities.For these reasons, isotopic values in DrA can be considered as reference values for groundwater in the flysch aquifer underlying the landslide.As a consequence, the 3 H content of 3.5 T.U.measured in groundwater sampled from DrA during summer 2010, can be used as the reference value of 3 H for the flysch aquifer.This value, it should be recalled, is significantly lower than the 3 H value of 9.8 T.U. which has been considered as reference for rainfall recharge water.
Other evidence corroborates the hypothesis of a mixing taking place between deep water inflow and rainfall-recharge water inside the bedrock underneath the landslide.The content of B tot and Sr 2+ in DrA is three orders of magnitude larger than the values normally found in shallow springs of NA (a few µg l −1 as reported by Duchi et al., 2005;Toscani et al., 2001).Considering the data reported in Sect. 2 and Mather and Porteous (2001), the long-term interaction between water and host rocks (such as flysch and/or evaporites) cannot in itself justify the measured B tot values.Mud volcanoes and oilfield waters are consistent with the B tot contents detected at Ca' Lita, while Sr 2+ is much higher than   Taylor (1997).
Rainfall recharge volume (V r ) is calculated with the Thornthwaite and Mather (1957) formula by using infiltration coefficient (I c ) ranging from 0.4 to 0.7.The uncertainty related to V r is considered equal to ±20-25 % by Fetter (2001), mostly due to the quality of the original rainfall and temperature datasets.As the quality level of the adopted ARPA dataset could not be assessed specifically in this study, an indicative value of ±20 % is given.expected.It should be noted that, unlike Sr 2+ , B behaves in a more conservative way and can hardly precipitate from the solution (Grew and Anovitz, 1996;Walker, 1975).Even admitting the presence of adsorption processes such as those detected in marine environments by Couch (1971), B would have been incorporated in clays' tetrahedral sheet structures at a very low rates (Keren and Mezuman, 1981) and it cannot be re-dissolved at the Ca' Lita site temperatures.Therefore, B tot contents linked to the deep water inflows could even be larger than those measured.

Data
If mud volcanoes and oil field waters are plotted in a δ 18 O-Cl − graph (Conti et al., 2000), they represent one of the endmembers of a hypothetical mixing line connecting them with shallow GW.GW from Ca' Lita falls on this line (Fig. 8).At the same time, the δ 18 O-B tot graph (Fig. 9) makes it possible to separate these two components.Ca' Lita water is completely left-shifted from the mud volcano points, which are polluted by an extra source of B contained in the connate waters of QM.This suggests a purely foredeep origin (oilfield waters) for the Ca' Lita GW.Leaching experiments on soil samples strengthen this conclusion.In particular, water which interacted with MVRa samples collected at Ca' Lita became more enriched in Na + and SO 2− 4 than MVRb leading to the supposition that these clayey outcrops were impregnated with deep water.The interactions with MOH samples (MOHa and MOHb) did not produce any changes in water chemistry and ion contents.A simple chemical equilibrium analysis was conducted coupling the mineralogical compositions of the soil samples using PHREEQC software.The investigation underlined that, even at equilibrium, the observed levels of Ca 2+ and SO 2− 4 cannot be reached.Moreover, there are no local mineralogical phases that could provide Cl − .(Conti et al., 2000), 3b and 3c (Capozzi and Picotti, 2010) -mud volcanoes; 5 -common shallow groundwater; 6 (Cervi, 2003), 7 (Venturelli et al., 2003) -springs; 8 -hydrocarbon seep.Black line: evaporation line from seawater.

Estimated aliquot and volume of deep water inflow in the landslide
The aliquot and the annual volume of deep water inflow at the landslide slope was estimated by using the isotopic groundwater balance approach described in Sect.3.7.The mean annual rainfall recharge was estimated using the Thornthwaite and Mather (1957) formula using a 40 yr long records of rainfall and temperature.A annual water surplus of 200 mm yr −1 was calculated, which is concentrated between December and April (Table 2).Table 2 compares GW L variations and the total and effective rainfall recharge calculated for each hydrological year.A direct relation can be observed.The 2006-2007 hydrological year was characterised by the lowest recharge value and consequent minimum GW L in each piezometer.Conversely, 2009-2010 was very wet and the maximum GW L was recorded at all points.Considering the annual water surplus and recharge area, the mean annual volume of rainfall recharge in the slope (V r ) was estimated to be in the range of 27 000 to 48 000 m 3 , depending on the infiltration coefficients used (I c = 0.4 to 0.7, which is the range of values flysch rocks in the northern Apennines reported by Civita, 2005).The uncertainties related to rainfall recharge volume (V r ) estimates are considered as large as 20-25 % by Fetter (2001), mostly due to the quality of the original rainfall and temperature datasets.As the quality level of the adopted ARPA rainfall dataset (ARPA, 2010) could not be assessed specifically in this study, an indicative value of 20 % can be assumed.
Equation (1) allows the aliquot of rainfall recharge water (aliquot b) to be estimated in 36 % (0.36 ± 0.02) and the aliquot of deep water inflow (aliquot a) to be estimated in 64 % (0.64 ± 0.02).By considering a steady state (Eqs.3 to 5) in which the total annual recharge of groundwater (given by rainfall recharge water and deep water inflow) equals annual discharge from the slope (assumed to have the isotopic characteristics of water hosted in the flysch, and sampled in DrA) (Fig. 10), the annual deep water inflow aliquot in terms of volume (V dw ) can be estimated as 49 000 to 85 700 m 3 .In other words, the deep water annual inflow V dw can be estimated to be about 1.80 ± 0.16 times larger than the rainfall and it represents a constant influx (in the order of 1.5 to 2.7 l s −1 ) that can be considered an influential factor for the long-term groundwater level fluctuations.

Conclusions
This research showed that in Ca' Lita landslide a significant aliquot of deep water inflow has to be considered in conjunction with rainfall recharge in order to interpret the chemical and isotopic composition in the landslide and in the surrounding slopes.Moreover, it demonstrates that the hydrogeological basin of the landslide is significantly different from the hydrological one, as it includes deep sectors of the slope.This conclusion is supported by evidences coming from multiple observations, including the anomalous chemical and isotopic composition of sampled water, the results from mineralogical analyses, the hydrochemical modelling using PHREEQC and the leaching tests.The chemical and isotopic fingerprint of deep water in the landslide slope points to oilfield water hosted at large depths in the northern Apennine chain.Several authors have documented the presence of this type of deep water in many locations in the Apennines and have generally ascribed their uprising motion as related to the presence of tectonic lineaments.The Ca' Lita landslide is characterised by the presence of a regional fault line that crosses the landslide area and which can be considered as the preferential flow path for deep water uprise.Deep water is then diffused in the slope via a network of discontinuities affecting both bedrock and landslide body.
Therefore, this study highlights how deep fluids can condition geomorphic processes such as favouring slope instability.Actually, the results obtained in Ca' Lita indicate that the amount of groundwater recharge due to deep water inflow might be larger than the amount due to rainfall recharge.Thus, correlations of instability events to rainfall only, would be erroneous and misleading.Furthermore, this deep water contribution has to be accounted for the design of mitigation works, such as for instance deep drainage systems.Monitoring data from the weirs at the drainage system outlets registered around 6000 m 3 yr −1 of discharge from DrA, and around 30 000 m 3 yr −1 from WA, a value that is within the range of the estimated deep water annual inflow if uncertainties in groundwater balance are considered.
In conclusion, the paper demonstrates the applicability of hydrological and hydrochemical analyses for quantifying the relative contribution of different sources of groundwater in slope-scale water budgets.This type of approach can be very useful in the study of large deep seated landslides, in which heterogeneity of constituting materials is high, and in which the hydrogeological basin can differ significantly from the hydrological one.A good understanding of the hydrological limits is essential when calculating representative hydrological budgets and if long-term stability modelling has to be carried out by interpreting the processes governing groundwater levels and their fluctuations in time and by evaluating the effectiveness of deep drainage mitigation systems.Hydrochemistry should, therefore, be considered as a valuable investigation method to assess the groundwater flow patterns in deep-seated landslide.

Fig. 4 .
Fig. 4. Continuous groundwater electrical conductivity (GW EC ) monitoring for piezometers P3 and P4 and groundwater level (GW L ) monitoring in WA and P5.
from the landslide body (P4) or the lateral unstable flank (P2) were characterised by values of δ 18 O which varied significantly during the different sampling campaigns, ranging from −8.51 ‰ to −9.88 ‰.In one piezometer inside the landslide body (P5), δ 18 O did not change significantly over the seasons (−9.13 ‰; −9.22 ‰).In P1 and DrA, δ 18 O values were significantly higher than rainfall water (up to −1.61 ‰

Fig. 10 .
Fig. 10.Reservoir-concept model of the slope.Water volumes (V r : annual rainfall volume; V dw : annual deep water volume inflow; V f : annual volume of water hosted inside the flysch; V OUT : annual water outflow volume) and their corresponding isotopic signals.

Table 1 .
Characteristics of piezometers, wells and drainage systems in Ca' Lita landslide.Location in the landslide area, material, depth of the filters, depth of the main slide surface, type of monitoring are also reported.

Definition of reference isotopic values for rainfall recharge water
In recent years, no station of the Global Network of Isotopes in Precipitation (GNIP) of the International Atomic Energy Agency (IAEA) located in the proximity of the study site measured δ 18 O and 3 H values(WISER, 2012).In particular, the San Pellegrino Alpe station (1520 m a.s.l., about 30 km south-west of Ca' Lita) has monthly records of δ 18 O from 1993 to 2002 and of 3 H from 1994 to 1995, while the Bologna station (35 m a.s.l. about 50 km north-east of Ca' Lita.) has only a monthly record of δ 18 O from 1996 to 2000.The isotopic values in these records are too distant in time to be used as reference values for the 2006-2010 analysis period.

Table 2 .
Piezometer monitoring results and total-effective rainfall.Data are distinguished by hydrologic year.

Table 3 .
Groundwater chemical (sampling campaign A, B, D) and isotopic (sampling campaign A, B, C, D, E, F) analyses of the Ca' Lita samples.Other samples are reported for further comparison: mud volcanoes (point: 3b), common shallow groundwater (point: 5) and hydrocarbon seep (point: 8).

Table 4 .
Leaching results for the 4 soil samples collected.Chemical analyses for the 1 h samples are also reported in mg l −1 .Electrical conductivity -25 • C (µS cm −1 )

Table 5 .
Relevant data and results of isotopic groundwater balance at the landslide slope scale.Precision of measured values is that of the analytical determination methods.Uncertainties in calculated values are assessed using error propagation methods presented by Table5summarises relevant data and results.The rainfall recharge area corresponds to about 340 000 m 2 , which is approximately the extent of the slope including stable flysch at the crown, at the sides and the landslide area, thus, excluding only the areas in which impermeable clayshales outcrop.