Introduction

Landslides are a common hazardous phenomenon occurring in mountain chains. Bedrock lithology, together with rainfall infiltration and groundwater paths, is considered a main parameter in landslide initiation and kinematics; different lithotypes exhibit different geotechnical properties. Weak rocks can promote the development of failures, collapses and landslides. Accordingly, comprehensive knowledge of the aerial and underground distributions of weak rocks (evaporites, marls, and clays) is required to identify the main areas subjected to deformation.

Evaporite formations are typically considered weak rocks (Yılmaz 2001). Gypsum is characterized by high solubility and low mechanical strength and is commonly associated with other soluble minerals, such as anhydrite, calcite, and dolomite. These characteristics, together with a high degree of tectonization, can lead to collapse and the development of large depressions, fissures or caves (Chiesi and Forti 2009; Gutiérrez 2010; Gutiérrez and Cooper 2013) that can play a double role in the stability of rocks. On the one hand, they drain the groundwater hosted in a slope and produce an increasing slope stability factor; on the other hand, they lead to the development of collapses that can provoke slope deformation or produce damage to the infrastructure in the affected area.

Evaporite units underlie approximately 25% of the continental rocks of the earth (Klimchouk et al. 1996; Gutiérrez et al. 2013) and are often the bedrock of overlying formations (Denchik et al. 2019). Several evaporite outcrops are scattered across Italy from the Alps to Sicily with gypsum units cropping out from or lying beneath unconsolidated cover (Forti and Sauro 1996; Klimchouk et al. 1996; Chiesi et al. 2010; Calligaris et al. 2017).

In the northern Apennines, outcrops of Triassic evaporites are very common and in particular are widely distributed in the USRV (Fig. 1). Triassic evaporites outcropping locally in the USRV consist mainly of metre- to decametre-scale gypsum, anhydrite, black dolomite and minor halite (Lugli 2001). Layers of dolomitic clasts within a carbonatic matrix are locally detected. In northern Italy, many cases of deformation promoted by evaporite collapses that involve anthropic housing and infrastructure have been reported (Buchignani et al. 2008; Intrieri et al. 2015; La Rosa et al. 2018). For this reason, the underground distribution of gypsum units and evaporites in general is important since the presence of fissures or caves in the bedrock can influence the stability of the overlapping formations with the development or reactivation of landslides (Gutiérrez and Cooper 2013). Gypsum units outcropping along the USRV are commonly covered by slope debris or landslide deposits. An accurate identification of gypsum outcrops and their extents is needed to understand the destabilizing role that gypsum may play in slope debris or landslides and to identify areas that are prone to collapse.

Fig. 1
figure 1

a Geographic location of the study site; b distribution of gypsum in the USRV and location of study site. MC, Montecagno Landslide, 1 impermeable rock masses, clayey, marls etc.; 2 medium-low permability rock masses, flysch; 3 medium-high permeability rock masses, sandstones, limestone; 4 high-permeability karstified rock masses; 5 river/torrent; c geological sketch map of the Montecagno study site; d cross-section of the Montecagno slope

Full size image

In addition, to get a better understanding of the stability/instability conditions of slope debris/landslides, a clear identification of the recharge area, the groundwater origin and the residence time is required. In recent decades, new and indirect methods (based on water chemistry and isotopes) have been used to better understand groundwater processes in unstable slopes (de Montety et al. 2007; Cervi et al. 2012; Vallet et al. 2015; Marc et al. 2017; Deiana et al. 2018) and to demonstrate the existence of lithologies and structures that are not visible with surface-based geological investigations (Bogaard et al. 2007).

This paper focuses on the landslide affecting the urbanized Montecagno slope (northern Apennines of Italy). The landslide at the sliding surface depth exhibits a cyclic and seasonal deformation pattern that is correlated with seasonal groundwater variations inside the landslide body (Deiana et al. 2017). Moreover gypsum formations outcrop along the Montecagno slope, but their exact distribution is not well known. Stratigraphic information obtained during borehole drilling is available only for some points (P1–P12) (Fig. 1).

To fill in the lack of information on the bedrock lithology in the other portions of the landslide, and clarify relations between groundwater circulation and bedrock lithology in the Montecagno landslide new investigations based on a coupled hydrologic and geochemical approach were performed.

Specifically, the study aims to clarify performing the following aspects:

  • define the origin, flow paths and transit time of groundwater circulating inside the landslide body using groundwater level monitoring, δ18O-δ2H and 3H isotope analyses and FLOWPC modelling (Malozweski and Zuber 2002);

  • investigate evidence of gypsum deposits inside the landslide body using water chemistry (main ions) and isotopes (87Sr/86Sr, δ11B) and highlight the main reactivating factor acting on montecagno landslide mechanism.

Study area

Hydrogeological and hydrochemical setting

The study site is located in the USRV in the northern Apennines of Italy.

The northern Apennines are a fold and thrust mountain belt made up by sedimentary rocks, many of which are turbidite sequences (flysch rock masses) and clayey chaotic deposits (clayshales, marls) (Fig. 1b). The complex structural features of outcropping formations, the high degree of tectonization and the wide presence of faults prevent the development of a regional deep flow path developed in this portion of the mountain chain. Only in some limited portion in the east part of the Apennines chain, where thick permeable rock masses with lateral continuity outcrops, a deep regional flow is discovered (Gargini et al. 2008; Cervi et al. 2014). In the research area, groundwater flow is limited at the slope scale, where permeable rock masses or slope deposits outcrops and shallow unconfined aquifers develop (Segadelli et al. 2017; Deiana et al. 2017; Tazioli et al. 2019). Only at the bottom of the Secchia Valley where Evaporite Triassic rocks outcrops with a certain lateral continuity a deep flow path is developed (Colombetti and Fazzini 1987; Chiesi et al. 2010) (Fig. 1b).

Hydrochemical features of groundwater hosted in this portion of the Apennines mainly depend by water–rock interaction with outcropping formations (Cervi et al. 2012). Specifically groundwater discharging by evaporitic formations are characterized by Ca-SO4 hydrofacies with Sr and B contents lower than 15 mg L−1 and 2 mg L−1, respectively (Toscani et al. 2001; Duchi et al. 2005). Groundwater interacted with flysch formations or clay-rich formation mainly belong to Ca-HCO3 hydrofacies with a low ion contents and Sr and B values normally lower than 1 mg L−1 and 0.05 mg L−1 (Cervi et al. 2012).

Study site

The studied landslide is located at elevations ranging between 1050 and 800 m a.s.l. (Fig. 1). It consists of an earth/debris slide (Cruden and Varnes 1996) and covers an area of 0.2 km2, while the mean thickness of the sliding mass is 28 m. The total estimated volume of the landslide is 3.6 Mm3, and the mean deformation velocity is 2.4 mm month−1.

The activity of the landslide is related with the seasonality of the groundwater level variation which is strictly related with the rainfall pattern of the area. During the summer and dry periods when the groundwater depth in the piezometer is normally higher than 5.2 m below the ground surface (b.g.s.) the landslide velocity is near to zero. In the humid seasons, when in the same piezometer the groundwater depth is around or lower than 4 m b.g.s., the landslide starts to accelerate and the velocity reacts or exceeds 3 mm month−1 (Corsini et al. 2013; Deiana 2017). Along the slope, four sedimentary geological formations outcrop (Fig. 1). The geological formations consist of evaporitic and clay-rich rock masses that exhibit a marked structural complexity and a large lithological heterogeneity. Along the slope, the following geological formations outcrop (from the bottom to the top): (1) a gypsum formation (GSB-Upper Triassic); (2) a shale with limestone formation (AVC-Lower Cretaceous); (3) a flysch formation (CAO-Upper Cretaceous); and (4) a marls formation (MMA-Oligocene).

The GSB formation is composed mainly of gypsum alternating with layers of anhydrite and highly fractured black dolomite. Locally, layers of dolomitic clasts within a carbonate matrix with an Oligocene–Miocene tectonic origin are present. The AVC formation consists of clayey shale alternating with carbonatic, pelitic and marly layers. Locally, layers of calcareous clasts are encompassed within a clayey matrix outcrop. The MMA formation is composed mainly of marls and pelitic marls with local layers of sandstone and clayey shale, especially at the bottom of the formation. The CAO flysch formation is composed mainly of a calcareous-marly component, but locally, arenaceous-pelitic layers can alternate with layers made by centimetric clasts in a clay matrix.

All the geological formations are characterized by a high degree of tectonization with several faults.

The slope is bordered downward by a water body, the Freddana stream that is oriented in a NE–SW direction in the study area. The discharge pattern of the Freddana depends mainly on the distribution of precipitation throughout the year.

Clay-rich formations and deposits, namely, the AVC and MMA formations, are considered aquicludes. Flysch and evaporite formations, namely, the CAO and GSB formations, are considered to be aquifers or aquitards due to their medium–low permeability (Gargini et al. 2008), which is the result of a medium–high degree of fracturing.

Along the Montecagno slope, a contact spring (Fetter 1980), namely, MS1, is detected at the bottom of the GSB slab that outcrops in the northeast portion of the study area. This spring, which drains a catchment area composed of the GSB slab tectonically overlapping the AVC formation, shows low discharge values depending on the distribution of precipitation throughout the year.

Boreholes drilled into the landslide area have been equipped with one standpipe piezometer (MP3) and five inclinometers (MP4, MP5, MP8, MP10, and MP16) (Table 1; Fig. 1).

Table 1 List of the selected surface and groundwater monitoring points
Full size table

Two well screens were performed along the slope and transverse to the landslide to reduce the groundwater level inside the slope deposits. The wells are linked underground, and the endpoint of the drain system is represented by MS9, which collects groundwater coming from the entire landslide body (Fig. 1).

Close to MS9 at the same elevation, a rainfall collector PLMC was installed to collect rainfall (Fig. 1).

The mean annual rainfall during the monitoring period (2013–2016) was 1730 mm (ARPAE-RER). The observed rainfall distribution is reported in Fig. 2. Frequent snow events occurred from November to March, and each single event accumulated up to 100 cm of snow on the ground surface. The mean annual air temperature during the monitoring period was 10.5 °C; the coldest month was February, and the warmest month was July (Fig. 2).

Fig. 2
figure 2

Monthly pattern for total rainfall, effective rainfall and temperature in the period September 2013-November 2016 recorded in the Ligonchio weather station

Full size image

According to Thornthwaite and Mather (1957), the average annual effective rainfall is 1240 mm.

Methods

To identify the geology of the bedrock and the hydrogeological features of the landslide, a multidisciplinary study was performed on the Montecagno landslide and the surrounding area.

To consider the groundwater variation inside the landslide, the groundwater level and discharge were monitored using a piezometer, a broken inclinometer and a drainage system between 2014 and 2016.

The δ18O–δ2H signatures were characterized every 4 months between 2014 and 2018 and coupled with the 3H content and FLOWPC modelling to assess the groundwater origin and flow paths to obtain a semi-quantitative evaluation of the groundwater transit time inside the landslide. The chemistry, 87Sr/86Sr ratios and δ11B values were determined to investigate the water–rock interaction processes of the groundwater along the flow paths.

In situ activities

Groundwater level monitoring

One borehole in the landslide was equipped with a standpipe piezometer (Table 1); this piezometer, MP3, is located in the central part of the landslide at an elevation of 960 m a.s.l. and is slotted from 2 to 28 m below the ground surface (Fig. 1). In MP3, a CTD-Diver (Eijkelkamp) was used to continuously acquire groundwater level data with a frequency of 1 h from January 2015 to April 2016; non-continuous groundwater level monitoring was performed monthly (every 3–4 months) from January 2014 to April 2016.

The other five boreholes were equipped with inclinometers, but in a few cases, the inclinometers were destroyed by landslide deformation. After a preliminary investigation, the inclinometers were cut and highly deformed, and they worked almost similarly to piezometers. For this reason, at these points (MP4, MP5, MP8, MP10, and MP16), non-continuous groundwater level monitoring was performed monthly (every 6 months) from January 2014 to June 2015 (Fig. 1).

During October 2011 and July 2015, two well screens were created along the landslide at elevations of 932 and 928 m a.s.l. to reduce the groundwater level along the slope.

The wells (diameter: 1 m, depth: 15 m) were linked underground and collected groundwater from the entire landslide. The endpoint of this drain system is represented by MS9, located at 920 m a.s.l. (Fig. 1, Table 1). In MS9, discharge was monitored continuously from January 2015 to May 2016 using an OTR water level probe installed in 2015; the frequency of the readings was every 1 h. Moreover, non-continuous measurements of discharge using a graduated container were performed monthly (every 4 months) from January 2014 to January 2015 and daily (every 20 days) from February 2015 to June 2016.

Surface and groundwater sampling

To collect groundwater and surface water samples for chemical and isotopic analyses, 2–30 sampling campaigns were performed between January 2014 and March 2018 at the Montecagno study site. The sampling locations are presented in Fig. 1. Sampling locations were selected to obtain representative samples of surface water and groundwater both inside (MP3, MP4, MP5, MP8, MS9, MP10, and MP16) and outside (MS1, MF2, MF6, MF7, and MF12) the landslide; the sampling sites were associated with different geological formations outcropping in the study area (Table 1). For comparison, sampling sites located outside the landslide were also included in this study. The rainfall collector, which was located at 920 m a.s.l. (PLMC; Fig. 1), collected water for chemical and isotopic analyses from January 2016 to January 2018.

Moreover, during the monitoring period, four springs (S1, S2, S3 and S4) located approximately 0–20 km away from the study area were selected to monitor the stable isotopic composition of the water (δ18O-δ2H). These springs were selected because the features of their hydrogeological catchments were consistent with the isotopic characteristics determined by Vespasiano et al. (2015). Specifically, they have small and well-defined recharge areas as well as small differences between the infiltration/recharge elevations and the spring elevations. For this reason, these selected springs contributed to obtaining a valid 18O-elevation relationship and more representative background isotopic values in sub-surface and surface water in the area, consistent with recent studies (Lauber and Goldscheider 2014; Vallet et al. 2015; Marc et al. 2017; Deiana et al. 2018).

Samples from springs, ditches and drain systems were collected directly from discharge points, whereas groundwater was collected from piezometers and broken inclinometers using a bailer or a low-flow pump (0.1 L s−1), which occurred after removing three aliquots from the water volume contained in the piezometer.

Water samples were collected at a mean frequency of 4–6 months using 500 mL polyethylene (PE) bottles with double caps. The samples were filtered through a 0.45-μm cellulose membrane filter, and the aliquots used for cation analysis were acidified with Suprapur Merck 65% HNO3. For all water samples, temperature (T), electrical conductivity (EC) and pH were measured in the field using a Crison MM40+ multimeter equipped with a Ross glass electrode to measure pH. In addition, at two points (MP3 and MS9), continuous EC and water T monitoring were performed from February 2015 to November 2016 using a CTD-Diver datalogger. The CTD-Diver allowed the measurement of the water temperature with a range of − 20 to 80 °C, an accuracy of ± 0.1 °C, and a resolution of 0.01 °C. The EC was measured with a range of 0–120 mS/cm, an accuracy of ± 1% and a resolution of ± 0.1%.

Water samples for stable isotope analysis (δ18O, δ2H) were collected at a mean frequency of 3 months using 50 mL PE bottles with double caps. Water samples for 3H analysis (collected in 2 L PE bottles with double caps) were collected during four sampling campaigns in January, March, May and November 2015. Water samples for 87Sr/86Sr and δ11B analyses were collected during a single campaign in November 2015 using 500 mL PE bottles with double caps. Samples collected for isotopic analysis were stored at 4 °C to avoid evaporation after collection.

Analytical methods

Chemical analyses on the water samples collected at the selected monitoring points in the Montecagno study site are reported in the Supplementary material (Table SM1).

Chemical analyses

Chemical analyses were performed at the Department of Chemical and Geological Sciences at the University of Modena and Reggio Emilia and the Chrono-Environment Laboratory at the University of Franche-Comté.

An inductively coupled plasma spectrometer (ICP-AES Perkin Elmer Optima 4200) was used for the analysis of Ca, Mg, Na, K, Sr, and Btot. Anion (SO4, Cl, NO3) concentrations were assessed using high-pressure ion chromatography (Dionex DX 100). Total alkalinity (expressed in HCO3) was assessed using the Gran titration method (Gran 1952). The total relative uncertainty was less than 5%. Silica was analysed with a Spectroquant spectrophotometer (Pharo 300, Merck) using a silica test kit (Merck). A total of 29 chemical analyses were completed during this study.

Isotopic analyses

The results of the isotopic investigations conducted from 2014 to 2018 are shown in Table S1 (Supplementary material).

The isotopic compositions of the water molecules were measured in water collected from all sampling locations at a mean frequency of 3 months; notably, the 18O/16O ratio was determined in 106 samples using an IRMS spectrometer, and the 2H/1H ratio was determined in 88 samples using a Los Gatos Research liquid water isotope analyser LGR-LWIA-24d. Isotopic analyses were performed at the IGG-CNR, Pisa. The results are reported as δ-‰, which reflects deviations from the standard isotopic value (V-SMOW). The analytical precision is ± 0.10‰ and ± 1.0‰ for δ18O and δ2H, respectively.

The 3H content was measured using a PerkinElmer Quantulus liquid scintillation counter based on the electrolytic enrichment method. The results are reported as tritium units (T.U.). The analytical precision is better than 0.8 T.U.

The 87Sr/86Sr ratio was measured in 5 samples in November 2015 using a Finnigan MAT 262 multi-collector mass spectrometer, which occurred after performing the ion-exchange separation of Sr from the matrix. The measured 87Sr/86Sr ratios were normalized to 86Sr/88Sr = 0.1194. The reproducibility of the measured 87Sr/86Sr ratios was tested using replicate analyses of the NIST SRM 987 (SrCO3) standard; during the analytical period, this standard yielded an average value of 0.710237 ± 0.000020 (2SD, n = 6).

The δ11B value was determined using an MC-ICP-MS Neptune Plus with a combination of internal standardization and bracketing standards for instrumental mass bias correction; this instrument can achieve precision for isotope ratio measurements in the range of 0.001–0.002%.

Data integration and analysis

Lumped parameters modelling (FLOWPC)

FLOWPC is a software program based on lumped parameter models applicable to the interpretation of environmental tracers in groundwater systems (Maloszewski and Zuber 2002; Viville et al. 2006; Sànchez-Murillo et al. 2015). In this study, the δ18O isotopes of rainwater and groundwater collected from the landslide are processed to obtain information about the turnover time (Maloszewski and Zuber 2002), and the groundwater flow inside the landslide is evaluated using a combined method made by a piston flow model (PFM) and an exponential model (EM). In the PFM, the flow lines are assumed to have the same transit time, and hydrodynamic dispersion and diffusion are negligible (Maloszewski and Zuber 2002). In the EM, the flow lines are assumed to have an exponential distribution of transit times, and no exchange of tracers between the flow lines happens (Maloszewski and Zuber 2002). This method allows us to obtain a semi-quantitative evaluation of the groundwater transit time inside the landslide.

Results

Groundwater level, EC and T

The results of the groundwater level monitoring in MP3 are reported in Fig. 3 and Table 2.

Fig. 3
figure 3

a Groundwater level compared with daily rainfall in MP3; b discharge compared with daily rainfall in MS9

Full size image
Table 2 List of manual groundwater level monitoring (expressed in m a.s.l.) performed using piezometer and broken inclinometers
Full size table

During the monitoring period, the groundwater level in MP3 ranged between a minimum value of 953.9 m a.s.l. recorded at the end of the summer period (September 2015) and a maximum value of 955.9 m a.s.l. recorded during winter (February 2016). The temperature ranged between 10.2 °C and 11.6 °C; the EC varied between 600 μS cm−1 and 1400 μS cm−1.

In MS9, the discharge ranged between 0.7 L s−1 recorded during winter (March 2016) and 6.0E−4 l s−1 recorded during summer (June 2015). T ranged between 10.5 and 11.5 °C; the EC varied between 700 and 1400 μS cm−1.

The groundwater level measured in the broken inclinometers (Table 2) ranged between a minimum value of 934.7 m a.s.l. (MP4) and a maximum value of 961.8 m a.s.l. (MP16).

Water chemistry

The results of the water chemistry analyses are reported in Fig. 4 (see also Supplementary material).

Fig. 4
figure 4

a Durov diagram of hydrochemical analyses; b stiff diagrams of representative samples; c spatial representation of sampled points

Full size image

In the landslide and surrounding areas, four different groundwater hydrotypes are identified.

Group 1 corresponds to Ca-SO4-rich water and includes MS1, MF6, MF7 and MF12. Samples belonging to this group show high contents of Ca (up to 610.4 mg L−1) and SO4 (up to 1493 mg L−1). Low contents are detected for Na (mean value 13.2 mg L−1), K (mean value 6.8 mg L−1) and Cl (mean value 14.5 mg L−1). Variable contents are detected for HCO3 (ranging from 132 to 320 mg L−1) and Mg (ranging from 2.3 to 66.0 mg L−1). The EC varies between 996 and 2570 μS cm−1. The mean pH value is 8.0, and the mean T value is 10 °C.

Group 2 corresponds to Ca-HCO3-rich water and includes MF2, MP3, MP4, and MP10. Samples belonging to this group predominantly comprise Ca (up to 235.6 mg L−1) and HCO3 (up to 336 mg L−1). Cations Mg, Na, and K show mean values of 12.0, 16.5, and 8.2 mg L−1, respectively, and non-negligible amounts of SO4 and Cl are detected in MP3 (up to 200 and 35.9 mg L−1, respectively). The EC varies between 536 and 985 μS cm−1. The mean pH value is 7.6, and the mean T value is 12.7 °C.

Group 3 corresponds to Ca-HCO3-SO4-rich water and includes MS9. More specifically, water belonging to this group exhibits high concentrations of both HCO3 (mean value of 323 mg L−1) and SO4 (mean value of 228 mg L−1). Ca and Na are the most abundant cations (mean values of 129.4 and 55.1 mg L−1, respectively). Subordinate concentrations are detected for Mg and Cl (mean values of 22.6 and 29.0 mg L−1, respectively). Variable concentrations are detected for K (ranging from 7.9 to 40.1 mg L−1). The EC varies between 1034 and 1255 μS cm−1. The mean pH value is 7, and the mean T value is 12.4 °C.

Group 4 is represented by rainfall with main components of Na (1.7 mg L−1), K (0.7 mg L−1) and Cl (3.3 mg L−1) corresponding to Na–K–Cl-rich water.

The sample CAO reported in Fig. 4 is an external spring collected in a carbonate flysch (Deiana et al. 2018) and is assumed to be an end-member of the water–rock interaction with the carbonate flysch.

Water isotopes (δ18O-δ2H, 3H, 87Sr/86Sr, δ11B)

The results of the isotopic analyses are reported in Table SM1 (Supplementary material) and Fig. 5.

Fig. 5
figure 5

δ18O-δ2H diagram. All water samples and rain samples are reported as well meteoric line valid for Western Mediterranean area (WMWL; Celle-Jeanton et al. 2000), North Italy (NIMWL; Longinelli et al. 2003) and Global line (GMWL; Craig 1961). Dashed black line corresponds to the calculated Local Line

Full size image

The isotopic values of water from the Montecagno sampling points are reported in the δ18O-δ2H plot. In general, samples fall between the northern Italy meteoric water line (NILM) and the global meteoric water line (GMWL) and identify a best-fit line with the equation δ2H = 7.96* δ18O + 9.45 (r2: 0.98).

The δ18O and δ2H isotopic compositions of the spring and surficial water range from − 10.25‰ (MS1) to − 8.85‰ (MF7) and from − 70.1‰ (MS1) to − 60.6‰ (MF7), respectively. The standard deviations of the isotopic values are higher than 0.29‰ (δ18O) and 2.30‰ (δ2H); the coefficient of variation (CV, Koch and Link 1971) is 4.1% (δ18O) and 5.2% (δ2H).

During the monitored period, springs S1, S2, S3, and S4 showed δ18O values ranging from − 10.25 to − 8.47‰ and δ2H values ranging from − 69.1 to − 54.5‰. At each site, the standard deviation of isotopic values measured on differing dates ranges from 0.06 to 0.3‰ (δ18O) and from 0.6 to 2.3‰ (δ2H). The CV is 1.9% (δ18O) and 2.3% (δ2H). The mean values of the isotopic content and the standard deviations of S1, S2, S3, and S4 are reported in the Supplementary material in Table SM2. In groundwater collected from piezometers, drains and broken inclinometers, the δ18O values range from − 10.05‰ (MP3) to − 8.13‰ (MP4), while δ2H values range from − 69.4‰ (MP3) to − 55.6‰ (MP4). In groundwater, the standard deviations of the δ18O and δ2H values are lower than 0.27‰ (δ18O) and 2.03‰ (δ2H) and the CV is 0.9% (δ18O) and 2.2% (δ2H).

Rainfall collected by PLMC shows δ18O values ranging from − 10.91 to -3.63‰ and δ2H values ranging from − 77.6 to − 18.7‰. The standard deviation is 2.2 (δ18O) and 18.2 (δ2H); the CV is 29.8% (δ18O) and 37.1% (δ2H).

To identify the altitude of infiltration of precipitation in the study area, a δ18O-elevation relationship was determined using springs MS1, S1, S2, S3, and S4 and following the approach proposed by Mussi et al. (1998) and Vespasiano et al. 2015. These selected springs, located 0–20 km around the study site, are used as natural pluviometer (Doveri and Mussi 2014; Cervi et al. 2016; Vespasiano et al. 2015; Deiana et al. 2018; Tazioli et al. 2019) as they are characterized by small, geologically defined catchment areas, and small differences between the elevations of maximum recharge and discharge points.

The relationship obtained for the Montecagno area is characterized by an isotopic gradient of − 0.44‰/100 m (r2: 0.98; not shown), which allows an estimation of the infiltration altitude of the groundwater inside the landslide. Infiltration altitudes consistent with the main elevation of the area were obtained for MP3 (1174 m a.s.l.), MS9 (1084 m a.s.l.), MP5 (1084 m a.s.l.) and MP8 (1086 m a.sl.). Inconsistent values were estimated for MP4 (852 m a.s.l.) and MP10 (1215 m a.s.l.). However, these points are associated with broken inclinometers in which groundwater circulation does not occur easily, such as in a piezometer. The broken inclinometers could have affected the isotopic compositions. Accordingly, these samples will not be considered in further discussions.

The isotopic gradient of − 0.44‰/100 m is higher than the regional gradient of − 0.22‰/100 m obtained for northern Italy (Longinelli and Selmo 2003) and is probably due to local effects at the slope scale, as recently described in other case studies in the northern Apennines (Deiana et al. 2017, 2018).

The results obtained using FLOWPC modelling and the combined PFM-EM method on δ18O values were selected using the sigma criterion-based approach (Malozweski and Zuber 2002). In this approach, the goodness of fit of a model is defined by a variable, namely, sigma, and the goodness of fit increases with a decrease in the sigma value. For the modelling performed in the Montecagno study site, the lowest sigma value obtained was 0.52, which is consistent with a turnover time (Tt) of 90–100 days.

The 3H content was assessed in surficial water and groundwater by four sampling campaigns during 2015. During the first campaign, which was performed in January, a 3H content of 3.9 ± 0.6 TU was measured in MS1. In March, values of 4.3 ± 1.0 TU and 4.3 ± 0.6 TU were measured in MP3 and MS9, respectively. In May, the following 3H contents were measured: 4.1 ± 0.7 TU (MS1) and 3.7 ± 0.7 TU (MP3). In November, the following 3H values were measured: 5.1 ± 0.7 TU (MP3), 4.0 ± 0.6 TU (MS9), 3.2 ± 0.6 TU (MF6) and 4.2 ± 0.7 TU (MF12).

The 87Sr/86Sr ratio was assessed in November 2015. The 87Sr/86Sr value of spring MS1 was 0.70797. In ditches MF6 and MF12, 87Sr/86Sr values of 0.708083 and 0.708024, respectively, were measured. Groundwater collected inside the slope debris showed 87Sr/86Sr values of 0.708151 (MP3) and 0.708400 (MS9).

The δ11B contents, which were assessed in November 2015, are as follows: 16.84 ± 0.1 (MS1), 7.24 ± 0.5 (MP3) and 21.75 ± 0.4 (MS9).

Discussion

Groundwater origin, flow paths and transit time

Changes in the groundwater level and pore pressure are crucial for the stability of a landslide. Groundwater hosted inside a landslide usually has a meteoric origin and responds to rainfall patterns. In the northern Apennines, however, recent studies have highlighted the presence inside the landslide of groundwater characterized by a deep origin (Cervi et al. 2012), by a long transit time and by deep flow paths in which the pressure transfer prevails over the mass transfer (Deiana et al. 2018).

In the Montecagno study site, the results show that a relevant portion of the landslide is saturated. The groundwater level inside the landslide deepens westward. In general, variations in the groundwater level over time suggest that rainfall patterns influence the groundwater level. However, in the left portion of the landslide, groundwater seems to be more static (0.20 m of variation in the groundwater level) than in the central and right portions, in which variations up to 3.8 m are measured.

The δ18O-δ2H results support a meteoric origin of the groundwater in all the sampled sites. Seasonal variations are evident in almost all samples with the more depleted values measured at the beginning of spring (March 2014) and the heaviest values measured in autumn (November 2015). In the northern Apennines, autumn 2015 was particularly dry (300 mm of cumulative rain), and the main precipitation events happened after the end of summer. This could explain the heaviest isotopic values measured in this period.

The more depleted values in the area are measured in the upper portion of the slope and in the head zone of the landslide, specifically in MS1 (− 10.25‰) and MF2 (− 10.17‰) while the heavier values are measured in MP4 (− 8.13‰ and − 8.37‰) in a lateral portion of the landslide. The high variability of the isotopic composition measured in general the head zone of the landslide (MF2, δ18O CV: 5.6%) is probably correlated with the presence of shallow circuits directly recharged by precipitation (PLMC, δ18O CV: 29.8%). The presence of scarps and fracture promotes infiltration in this portion of the landslide. Lower variability is measured in the central and lower portions of the landslide (MP3, MS9). In this portions the stability of isotopic values (δ18O CV of 0.9%) highlights that groundwater is hosted in deeper and longer circuits in which the integration of the precipitation isotopic signal by the aquifer is promoted.

As supported by δ18O-elevation relationship, groundwater shows a local origin. The infiltration altitudes obtained using the δ18O signatures for MS9, MP5 and MP8 are consistent with the elevation of the area and of the main scarp and, for the lower portion of the landslide, exclude water sources other than local meteoric recharge. In particular, in MP3, the infiltration altitude (1174 m a.s.l.) almost coincides with the highest peak of the area, confirming a local recharge that involves the entire upper portion of the slope (Fig. 10).

Moreover groundwater hosted inside the landslide has a recent origin. The 3H content measured in groundwater from both shallow and deeper hydrological circuits (3.8 TU and 4.3 TU, respectively) is close to the 3H content recorded in springs from the northern Apennines (4.2 TU) sampled in the same period (Deiana et al. 2016, 2017). The groundwater turnover time, estimated by the PFM-EM model, is 177 days; this suggests that in the Montecagno landslide meteoric water takes approximately 6 months to infiltrate and circulate after being finally discharged from the landslide body.

Distribution of the evaporite bedrock under the landslide and mixing processes

Along the Montecagno slope, gypsum, flysch and clay formations outcrop. Specifically, flysch and clay formations outcrop continuously along the slope; gypsum formations outcrop in two different and separated parts: the first one in the upper portion of the slope and the second one in the lower portion of the slope. During the borehole execution, gypsum blocks were found below the lower portion of the landslide (corresponding to MS9 and MP10). However, the landslide affects the slope from 1050 m a.s.l. to the stream at 800 m a.s.l. Because of the landslide deposits, the nature of the bedrock and the lateral continuity of the gypsum blocks below the lower portion of the Montecagno landslide remains unclear. The presence of gypsum below and inside the landslide can influence the stability of the landslide body (Elorza and Santolalla 1998; Gutierrez 2010; Gutierrez et al. 2013); indeed, by means of fissures and fractures, gypsum can promote the drainage of groundwater hosted inside the landslide body and increase the stability of the slope. Moreover, the ongoing dissolution affecting gypsum can promote the enlargement of fractures and caves and increase the possibility of collapses. Additionally, the stability of the slope is influenced by the dimensions of the gypsum blocks; decametric blocks can have a greater influence and can affect a larger portion of the slope than blocks with other dimensions.

The aspects of the subsurface bedrock structure and the extent of the geological formation composing the bedrock of the Montecagno landslide were investigated and clarified using chemical analyses (Bogaard et al. 2007) and 87Sr/86Sr–δ11B analyses. The chemical analyses seem to suggest the presence inside the landslide of water–rock interaction processes involving gypsum formation. The predominant role of gypsum dissolution process is highlighted especially in the lateral portion of the landslide (MF6, MF7 and MF12 are positioned under the equiline, Fig. 6). In the central and lower portions of the landslide (represented by MP3 and MS9, respectively), the position of these points in Fig. 6 (very near to the equiline) suggests that groundwater could interact with both carbonate and gypsum materials. Specifically in the central portion of the landslide (MP3, above the gypsum equilibrium line) water–rock interaction with carbonate material seems to be the main process affecting groundwater. The occurrence of water–rock interaction with gypsum is likely the main processes affecting groundwater in the lower portion of the landslide (Fig. 7). Moreover in this part, chemical features of groundwater are probably affected secondly by mixing processes between a Ca-SO4 end-member (MS1), a Ca-HCO3 end-member (MF2-MP3) and a Na–K–Cl end-member (PLM). The Ca-HCO3 groundwater that characterizes the upper portion of the landslide (MP3), moving into hydrological circuits toward the lower portion becomes enriched in SO4 content. The increase in SO4 content is promoted by ditches that laterally bordered the landslide.

Fig. 6
figure 6

Plot of HCO3 vs SO4 content in sampled groundwater

Full size image
Fig. 7
figure 7

Plot of Ca vs SO4 content in sampled groundwater. The equiline indicates the gypsum equilibrium line

Full size image

A further contribution is related with the preferential infiltration of rainfall, that is promoted by an upper secondary landslide scarp and contribute to the increasing of Na and K (Fig. 4a, b) in the lower portion of the landslide.

Data provided by 87Sr/86Sr seem to support the presence of water–rock interactions with gypsum in the lower portion of the slope (Fig. 8). In Fig. 8, end-members for the Triassic gypsum (Dinelli et al. 1999) and flysch (i.e., CAO) formations (Deiana et al. 2018) are reported. Groundwater from MS1 has 87Sr/86Sr ratios that match those of its evaporitic host rocks; the 87Sr/86Sr signature for MS1 is very similar to that reported for Triassic gypsum both in northern Apennines (Dinelli et al. 1999) and in Alps (Spötl and Pak 1996). The 87Sr/86Sr value measured in the lateral portion of the landslide (MF12), slightly higher than that of MS1, indicates a dominant interaction with gypsum and a subordinate influence of flysch formation (CAO). In the central portion of the landslide (MP3) the 87Sr/86Sr signature matches the CAO 87Sr/86Sr value (Fig. 8), highlighting a clear acquisition from preferential interaction with the flysch formation (CAO). In the lower portion of the landslide (MS9) due to the differences in 87Sr/86Sr signature (Fig. 8) the interaction of groundwater with pure gypsum seems to be excluded; here, the high 87Sr/86Sr content is probably due to the dominantly interaction of groundwater with flysch (CAO), which outcrops extensively along the slope, and subordinately with the gypsum that is found in the landslide.

Fig. 8
figure 8

1/Sr vs 87Sr/86Sr plot; gypsum field is by Dinelli et al. (1999). Note: CAO data are by Deiana et al. (2018)

Full size image

This hypothesis was also supported by the δ11B results. The δ11B value measured in MS9 falls in the field identified for limestone (Hemming and Hanson 1992), confirming that the interaction with the carbonatic component of the flysch formation is the main water–rock interaction process affecting groundwater in the lower portion of the landslide (Figs. 9, 10). Moreover, the similarity of the MS9 δ11B value to the gypsum field (Lemarchand and Gaillardet 2006; Mao et al. 2019) confirms the subordinate interaction with gypsum that likely outcrops in the landslide body as small lenses or blocks. In the central portion of the landslide, the δ11B value measured in MP3 suggests that water–rock interaction processes involve only the flysch formation, and allow to exclude the presence of gypsum in this portion of the landslide body and the bedrock (Figs. 9, 10). The lower δ11B value found in this point is likely due to the presence in the central portion of the landslide of a silicatic component that generally imparts to groundwater a low isotopic signature (Barth 2000; Pennisi et al. 2000; Lemarchand and Gaillardet 2006; Millot and Négrel 2007).

Fig. 9
figure 9

87Sr/86Sr vs 11B/10B plot. Evaporites field is by Négrel et al. (2009), Elderfield (1986); carbonate field is by Négrel et al. (2009), Millot et al. (2011); rainwater by Négrel et al. (2007), (Xiao et al. 2013). Seawater is by Millot et al. (2011)

Full size image
Fig. 10
figure 10

Conceptual model of the Montecagno slope and landslide showing the schematic groundwater flow path. Blue squares represent piezometer; green square represents drain system; red squares represent inclinometers; blue circle and arrow represents MS1 spring that is not in line with the cross section; continuous light blue arrows represent the local rainfall infiltration along the slope; continuous blue arrows represent the flow path occurring in the landslide. The 3H and δ11B contents are reported near the corresponding sampling points as well tritium content of rainfall. The 87Sr/86Sr content detected in landslide is also reported

Full size image

As supported by isotopic analyses, gypsum blocks found inside the landslide, due to their small dimension seem to play a negligible role in the landslide stability. Accordingly rainfall infiltration and groundwater paths result to be the main parameter in the kinematics of the Montecagno landslide.

Conclusions

In the Montecagno landslide, a comprehensive study of chemical and multi-isotope analyses (δ18O-δ2H, 3H, 87Sr/86Sr, δ11B), groundwater level monitoring and hydrological lump modelling (FLOWPC) were performed to define the groundwater origin, flow paths, turnover time and bedrock lithology. In the study area, Triassic gypsum outcrops are quite widespread and composes the bedrock of the landslide deposits; the gypsum outcrops have serious consequences on the stability of the landslide deposits since gypsum formations are affected by dissolution phenomena that can produce caves and collapses.

The results from the δ18O-δ2H and 3H analyses and the δ18O-elevation relationship suggest that the groundwater hosted inside the landslide has a recharge controlled by local rainwater. Groundwater inside the landslide is hosted in shallow flow paths. FLOWPC models suggest that in these flow paths, groundwater renewal occurs every 3–4 months.

Chemical analyses support a water–rock interaction with gypsum inside the slope; 87Sr/86Sr and δ11B allow to confirm this hypothesis and obtain important information about the bedrock structures. Groundwater inside the central portion of the landslide is characterized by interaction with the flysch formation. In the lower portion of the landslide, the groundwater interacts primarily with flysch and subordinately with gypsum that is present in this portion of the slope as small lenses or blocks.

The small dimensions of the gypsum blocks allow us to hypothesize that the investigated portion of the Montecagno landslide is characterized by a low possibility of the occurrence of collapse phenomena and that groundwater infiltration play a key role in the landslide stability.

This research demonstrates that the methodology used, based on isotopic investigations, chemistry, modelling and groundwater level monitoring, can contribute to improving knowledge about the lithology of structures that are not visible with surface-based geological investigations and about the hydrogeological aspects of these rocks and the other adjacent formations.