Introduction

In recent years, extreme rainfall events and variations in seasonal cumulated rainfall distribution occurred, increasing also the arise of slope instabilities, mostly in the more susceptible areas, such as the hilly area of Oltrepò Pavese (northern Italy). Heavy rainfall events represent in fact one of the most important triggering factors in shallow landslides occurrence, and a better understanding of the trigger processes is necessary, also for early warning systems development and improvement.

Shallow landslides triggered by heavy rainfalls are slope instabilities, developed in the most superficial eluvial layers, involving the first 2 m from the ground level. They are instantaneous processes characterized by a high velocity against small volumes of involved soil (101-105m3). However, shallow landslides may represent a huge hazard for human activities, by means of significant damages to cultivations and infrastructures and, sometimes, causing the loss of human lives (Lacasse et al. 2010). Increasing population and economic activities in very prone areas made these phenomena much more relevant and impacting on socio-economic activities (Froude and Petley 2018).

In the last years many strategies of EWS (early warning system) and risk mitigation were implemented to reduce the impact of slope instability at local and regional scale (Segoni et al. 2014).To improve EWS reliability it becomes necessary to understand the predisposing factors of these phenomena, in terms of soil features, geomorphological and hydrological conditions (Guzzetti et al. 2008; Segoni et al. 2014).For this purpose, spatial and temporal components are fundamental parameters, both to determine the most prone area and to define the frequency of slope instability occurrence, according to predisposing and triggering factors (Bordoni et al. 2019).

A topic key to lead to a reliable soil characterization, also in terms of ground water circulation, is the SWC (soil water content), since it influences the main soil pedogenesis processes, such as physical, chemical, and biological ones. Those processes are strictly related to shallow landslides predisposing factors, such as rainfall infiltration, water runoff, and soil erosion (Merz et al. 2006; Norbiato et al. 2009; Western et al. 2004). SWC depends on the soil properties, such as porosity, heterogeneities, hydrogeological conditions, and meteorological parameters (precipitation, air temperature and humidity and solar incidence) (Teuling and Troch 2005).SWC is generally monitored through soil probes, such as TDR (time-domain reflectometers), FDR (frequency-domain reflectometer) and dielectric sensors, that provide specific-site data. Since the local predisposing factors are involved in probability estimation of slope failures, it is necessary to shift the hydrological characterization from a specific-site scale to a slope scale.

Electrical Resistivity Tomography (ERT)appears particularly suitable for understanding the variations, both in the spatial and temporal domain, of the most superficial layers, in terms of soil water content within the landslide bodies, being one of the most applied geophysical techniques to characterize landslide properties, such as sliding body thickness, lateral extension, depth, and temporal and spatial changes in water content (Crawford and Bryson 2018; Crawford et al.2015; Gance et al. 2016; Uhlemann et al. 2017; Yilmaz 2007). ERT is a non-invasive geophysical technique, widely employed since the 1990s, useful to understand the subsurface conditions, in terms of water content, soil total bulk porosity and soil mineralogy, trough the determination of electrical resistivity spatial (2D-3D) and temporal (Time-Lapse) distribution, (Oldenburg and Pruess 1993; Loke and Barker 1996; Kim et al. 2009). ERT also allows to evaluate the percentages of sand and clay, in soils or rocks with a certain clay content and therefore to estimate the bulk total porosity and permeability. In non-clayey soils or rocks it allows to estimate the total porosity (bulk total porosity derived from formation factor) and therefore to hypothesize the permeability (Archie 1942; Shevnin et al. 2006, 2007; Torrese et al. 2022; Torrese 2023; Vogelgesang et al. 2020; Waxman and Smits 1968; Wildenschild et al. 2000). On the other hand, electrical resistivity interpretation is still quite complex, due to the different involved factors, such as porosity, pore water resistivity, water content: the deeper is the knowledge of these factors, the better is the accuracy of the ERT measurements interpretation (Lapenna and Perrone 2022). This technique is based on the estimation of the electrical resistivity of the subsoil and consist in the application of direct current into the ground by means of two current electrodes, displaced along a profile, and the measure of the resulting voltage via two potential electrodes (Torrese and Pilla 2021).

Since the electrical resistivity is influenced by parameters such as lithology, petrophysical and hydrological conditions, ground geological settings can be assumed by means of the knowledge about the conductivity distribution. The interpretation of electrical resistivity measurements is also useful to characterize vertical soil profiles, in terms of water content and water deficit, strictly related to different horizons porosity (Brunet et al.2010).

ERT results are inverse resistivity tomographic images, 2D, 3D or 4D (time-Lapse) (Karaoulis et al. 2011), that model the field subsurface measured resistivity, giving out the spatial pattern of features in the subsurface and addressing an interpretation of the hydrogeological conditions, bedrock type, soil type and thickness. Therefore, ERT survey becomes a useful instrument to characterize landslides predisposing factors, since they are strictly related to electrical resistance of rock and soils (Reynolds 2011). ERT technique is also particularly useful in the identification of preferential ways of fluids infiltration(Ivanov et al. 2020) and may represents a significant contribution in early warning systems for rainfall induced landslides in terms of water saturation effects in the estimation of the rainfall thresholds (Bordoni et al.2019). Hence, ERT measurements have been integrated and compared with meteorological and hydrological data, for example with TDR measurements or other geophysical data (Vanella et al. 2021).

In the last 20 years several applications were conducted on ERT on the field, proving the reliability of this technique in landslide characterization and monitoring. Beff et al. (2013) have conducted an ERT field experiment, focusing on 3D soil water content distribution at the plot scale determination and soil water content dynamics, by investigating on the rainfall and root water uptake effects. They combined ERT and TDR (Time Domain Reflectometer) measurements, finding a very good correlation between ERT and TDR acquisitions (R2 = 0.98). They carried out several ERT permanent and non-permanent profiles, to monitor SWC and water deficit, accessing the hydrogeological parameters of “Cevennes” landslide, located in the SE of France. The results from ERT performed on the landslide shaley soils confirmed the feasibility of such measures for a more systematic exploration of SWC and water deficit in superficial soils (Brunet et al. 2010). A multi-scale approach was conducted by Lebourg et al. (2010) on the “Vence” landslide (South-eastern France), focused on the identification of spatial and temporal distribution on rainfall derived inflowing water in the sliding mass, by means of resistivity variations. Basing on the temporal and spatial variations of resistivity, a water infiltration and drainage monitoring was performed with high-resolution Time-Lapse ERT in order to understand the hydrogeology of a complex landslide and to detect the sliding surfaces, in Lixian County, Southwestern China (Ma et al. 2021). Chengpeng et al. (2016) characterized the internal structure and the relative groundwater circulation of a translational landslide in Kualiangzi, Southwest China, by means of the interpretation of ERT profiles combined with drill cores and inclinometer data. Ivanov et al. (2020) approached the characterization of shallow landslides triggered by rainfall extreme events in a mountain landscape, performing hydrogeological and Time-Lapse ERT measurements. The time series of the ERT images were used for mapping the waterfront time-dependent changes, being a useful tool for implementing procedures for the validation of the rainfall thresholds. An automatic ERT monitoring system to monitor the saturation degree, combined with field data of volumetric water content was performed by Wicki and Hauck (2022) in a landslide-prone hillslope in the Napf region (Switzerland). Zieher et al. (2017) performed a sequence of 2D ERT images during an irrigation period, through a spray irrigation in an Alpine zone, prone to shallow landslides. A grid of TDR sensors was installed at different depths to monitor SWC and to retrieve the pore pressure. The results have shown the negative resistivity changes associated with the vertical progressing water, although the model was not able to investigate the small-scale variations in pore pressure that takes effect in shallow landslides triggering.

Other geophysical techniques for the estimation of water content and total porosity are high resolution active seismic techniques (surface: seismic tomography, MASW-Multi-channel Analysis of Surface Waves, refraction seismic, reflection seismic) (borehole: cross-hole, down-hole, seismic tomography), electromagnetic techniques (surface: GPR-Ground Penetrating Radar, TDEM-Time Domain Electro-Magnetic, FDEM-Frequency Domain Electro-Magnetic) (borehole: cross-hole, VRP-Vertical Radar Profiling). Unlike electrical techniques, in which water content and total porosity estimation is based on electrical resistivity measurement, seismic and GPR techniques estimate water content and total porosity from the measurement of seismic and electromagnetic (dielectric constant) waves propagation velocity, respectively. In general, for this purpose, electrical techniques provide higher accuracy and resolution than seismic techniques and less or comparable accuracy than GPRones. The disadvantage of electromagnetic techniques lies above all in the low depth of investigation in clayey or water-saturated soils or rocks (high electrical conductivity).

Given the previous bibliography, this work aims to characterize the hydrological conditions of two test sites with different geological setting, both prone to shallow slope instabilities, to improve the knowledge of the predisposing factors in shallow landslides triggering, by means of non-invasive techniques and directly involving the slope stability analysis, considering the impact of geological hazards, such as landslides, on engineering applications. The estimations of soil water content and average bulk permeability provided by this analysis are in fact important parameters in the stability modeling, necessary for engineering geology feasibility studies. Indeed, soil conditions before a precipitation event, in terms of water content, water infiltration and circulation may influence rainfall-induced shallow landslides triggering (Bordoni et Al. 2019).

For this work, an innovative integrated approach was applied, combining different ERT surveys (2D-3D-4D) at different scales (specific site scale and slope scale) with geomorphological and geotechnical investigations and soil water content monitoring. Twore presentative test sites prone to shallow landslide were chosen in Oltrepò Pavese area, Montuè and Costa Cavalieri, representing both the typical geological and geomorphological contexts in northern Apennines. The test site of Montuè (MON) is located in the NE sector of Oltrepò Pavese, while Costa Cavalieri test site (CCV) is located in the central sector of Oltrepò Pavese. In both the test sites, several stratigraphic logs derived from manual drills and trenches were performed, to calibrate ERT surveys in order to set up a hydrogeological model at site and slope scale as well as the soil samples collection during the field surveys for laboratory analyses (geotechnical parameters, infiltration tests).

Given that ERT is strongly influenced by soil lithological characteristics and it provides for a qualitative water content estimation, a calibration with soil water content probes was necessary. In both the test sites, low and high detailed 3D arrays were performed to compare ERT results with TDR monitoring data, at the same time of the ERT acquisition, linking direct soil measurements to ERT models. 3D models were successively integrated with2D longitudinal and transversal ERT arrays, providing a deep survey along test sites slopes, actually extending the punctual measurements derived from hydrological monitoring stations to a spatial characterization. Finally, multitemporal 4D arrays were performed, with injection of water between the first and the second acquisitions, in order to simulate a very intense rainfall event for bulk average permeability estimations.

Data and methods

As summarized in the flowchart of Fig. 1, this analysis combined different input data, starting from the field surveys retrieved by micro-boreholes, trenches and UAV (unmanned aerial vehicle), in terms of soil logs and soil horizons interpretation and DTMs (digital terrain models), useful for reconstructing lithological 3D models and relative derivatives, such as lithological sections. Within the trenches soil probes were also installed at different depths (Fig. 5), to monitor the soil water content and the pore water pressure for each soil horizon above the bedrock.

Fig. 1
figure 1

Flowchart showing the methods of the research

Full size image

Geophysical surveys were performed using different techniques and different detailed arrays in order to enlarge the scale of analysis, from the punctual site scale of the monitoring stations to a slope scale as well as to investigate more in depth. From all data input were achieved different preliminary results suitable for ERT calibrations and interpretations, such as soil logs, lithological sections, time series of hydrometeorological data. Different scaled ERT surveys and 4D time-lapse acquisitions filled the gap to perform slopes characterizations, from which it was possible to quantify the average areal water infiltration rate, to build hydrogeological conceptual models and to reveal the presence of deep landslides sliding surfaces.

The test sites

This work was carried out in two test sites, Montuè (MON) and Costa Cavalieri (CCV), that are representative areas of the Oltrepò Pavese hilly region, located in the northern termination of the Italian Apennines (Fig. 2). These locations were selected for the following reasons: presence of past shallow failures closes to the monitoring station: typical geomorphological and geological features of the test sites and of the areas prone to failures.

Fig. 2
figure 2

Geographical, geological and geomorphological settings maps of the test sites. Monitoring stations position is indicated by red stars

Full size image

According to Koppen’s classification, the climatic regime of test sites is temperate/mesothermal, with a mean yearly temperature of 11–12 °C and an average yearly rainfall amount between 669 mm for CCV and 684 mm for MON, according to rainfall data collected from rain-gauge stations of ARPA Lombardy monitoring network in the period 2004–2019.

The Oltrepò’ Pavese hilly area presents a complex geological setting, due to the different emplacement phases, during the chain growth in middle Eocene-late Miocene. The complex set of tectonic units in the area is formed by the thrust of the Internal Ligurian Units on the External ones and the sin-tectonic turbidites deposits and submarine landslides, that compose the Epiligurian Units (Toscani et al. 2006; Bosino et al. 2019; Tibaldi et al. 2023). This area was progressively interested by post-Messinian deposits, superposed on these Units, concurrently with the covering by Quaternary alluvial deposits due to rivers and creeks action, presenting NW/SE, NE/SW and N/S oriented faults and lineaments, according to the direction of units imposed during the chain formation (Panini et al. 2002).

As regards the land use, more than 40% of the Oltrepò Pavese is covered by vineyards, with a significant presence (> 20%) of woodlands and shrublands developed in hill-sloped, where agricultural activities were abandoned.

MON test site is characterized by a succession composed by poorly cemented sandstones and conglomerates, with a marl and evaporite deposits bedrock. Superficial horizons derived from weathered bedrock and they are essentially composed by low plastic clayey or clayey–sandy silts, with a thickness ranging from few tens of centimeters and 2 m. From the morphological point of view MON site shows steep hillslopes, with an average slope angle between 15 and 35°, covered with shrubs and woods of black robust. The arenaceous–conglomeratic bedrock is characterized by a medium–low hydraulic conductivity, due to its low primary porosity and the limited number of fractures, not enough to develop a high secondary hydraulic conductivity. The deep-water circulation is then confined in less cemented or more fractured levels, at different depths in the bedrock. Those levels correspond to of poorly cemented gravels, sands or conglomerates horizons with a limited lateral extension and thickness around 0.2 and 1.0 m. The presence of water in the bedrock can be identified only by considering the more permeable levels as isolated bodies. These bodies, indeed, do not seem to be a continuous more permeable level, that can develop a deep aquifer. Shallow landslides are triggered mainly when the shallowest horizons are saturated by water infiltration, forming a perched water table (Bordoni et Al. 2015).

CCV test site is characterized by marly and calcareous flysches, which are alternated with sandstones and marls. The bedrock is composed of alternated sandstones, marls and claystone with a peculiar block-in-matrix, predisposing both the soil lithology and thickness. This area is also interested in the presence of deep slow-moving landslides deposits that influence the eluvial soils thicknesses, ranging from 1 to 5 m. Slopes have medium steepness, ranging from 8° to 20°. In this site, shallow soils are essentially high plastic silty clays. The groundwater circulation is not attributable to a proper aquifer but is strictly related to the geological contact between a calcareous plate and the marls. Shallow landslides are in fact triggered when the water coming from the calcareous plate form a perched water table, in correspondence with water infiltration from heavy rainfall events.

Moreover, both the study areas are very prone to heavy rainfall-induced shallow landslides (Meisina 2004; Bordoni et al. 2019). Slopes are affected by shallow landslides, that may evolve into flows, with a sliding surface at about 1 m in depth and ratio between length and width of 1.0–7.1 (length between 10 and 500 m, width between 10 and 70 m) (Bordoni et al. 2015). In general, most of the shallow landslides are classified as complex phenomena, according to Cruden and Varnes (1996) classification. In fact, the soil horizons above sliding surfaces are characterized by a lower density (unit weight, γ, of 16.7–19.0 kN/m3), a higher permeability (saturated hydraulic conductivity, K s, of 10 -5 /10 -6 m/s) and they are less strong (especially due to an effective cohesion, c’, around 0 kPa) than the layers underlying the sliding surfaces (γ of 18.6–20.3 kN/m3, K s of 10 -7 m/s; c’ up to 29 kPa) (Bordoni et al. 2021a, b).

Stratigraphic logs and geological sections

In both MON and CCV test sites, two trenches and several micro-boreholes with an average depth of 2 m and 6 cm of diameter were carried out, in order to assess the soil thickness and the lithological characterization (Fig. 3).

Fig. 3
figure 3

Maps of ERT surveys performed on the two test sites. In A is shown MON test while in B CCV site. The trenches correspond to the green stars, at the same location of monitoring stations

Full size image

In MON test site a rectangular (2 m × 1 m) trench, with a depth of 1.85 m was performed (Fig. 3A). To improve the site characterization 5 micro-boreholes (Fig. 3A) with a depth ranging from 0.6 to 1.85 m, depending on the bedrock depth, were performed: S2 (0.6 m), S3, S4, S5 (1.2 m) and S6 (1 m) (Fig. 5). In CCV test site a rectangular (2.5 m × 1 m) trench with a depth of 2.0 m and 1 micro-borehole until 2 m were performed (Figs. 3B and 5).

Soil thickness and lithostratigraphic logs were interpolated in Rockworks software (RockWare Inc. Golden, CO 80401 USA) to perform 3D geological models derived from the interpretation and interpolation of each soil horizon. From the test sites geological/lithological models two lithological sections were retrieved, in order to interpret and calibrate ERT resistivity values in respect of the lithological variations.

From the trenches 5 samples (one for each horizon) were collected both in MON and in CCV test site. Samples were analyzed in geotechnical laboratory in order to retrieve the main geotechnical parameters, such as plasticity and grain size distribution, furthered in the results section and summarized in Table 1.

Table 1 Geotechnical parameter (plasticity parameters and percentage of grain size distribution) of test sites soils samples, collected at different depths
Full size table

Monitoring stations

In both the test sites, which were assumed representative of the surrounding area, hydrological monitoring stations were installed (Bordoni et al. 2021a, b), recording data every 10 min. Montuè monitoring station, installed in March 2012, measures meteorological (rainfall, air temperature, air humidity, atmospheric pressure, wind speed and direction, and net solar radiation) and hydrological (water content, pore water pressure) parameters. Six TDR probes, installed at different depths, are used to measure water content, while three Jet-fill tensiometers and three HD (heat dissipation) sensors, installed in pairs at three different depths, measure pore water pressure (Table 2) (Fusco et al 2022). Costa Cavalieri monitoring station was installed in December 2015 and collects only hydrological data. Five GS3 probes installed at different depths allow to monitor soil water content, while three MPS-6 dielectric sensors and three T4e-UMS tensiometers, installed in pairs at three different depths, allow to measure pore water pressure (Table 2). Meteorological data (rainfall amount, air temperature, air humidity, wind speed, and direction) are retrieved from the meteorological station of ARPA Lombardia at Fortunago, 2 km away from the test site.

Table 2 Soil monitoring probes technical characteristics with relative depths, installed in the test sites of MON and CCV
Full size table

Geophysical survey

2D ERT data collection

In both test sites different ERT experimental layouts were performed: 2D, large and small 3D and 4D (time-lapse) ERT surveys were carried out (Fig. 3). According to Maerker et al. (2023), Szalai et al. (2009) and Torrese (2022) the quadrupole sequence is a suitable array geometry, providing more accurate inverse models and more reliable ERT imaging, through merging and jointly inverting datasets from different arrays (i.e., Wenner and dipole–dipole).

Two ERT 2D profiles were acquired at each test site. The 2D surveys were undertaken orthogonally each other and centered to the to the soil monitoring stations, MON-AA’ and MON-BB’ (Fig. 3A) profiles are longitudinal and CCV-AA’ and CCV-BB’ are transversal to the slope (Fig. 3B). Each profile is 70.5 m long and was obtained using 48 electrodes 1.5 m apart. Each profile was collected using 306 Wenner-Schlumberger array quadrupoles which ensure high vertical resolution and signal amplitude and 328 dipole–dipole array quadrupoles which provide enhanced lateral resolution. A fully automatic multi-electrode resistivity meter SYSCAL Jr. Switch-48 by IRIS Instruments was used for data collection.

3D ERT data collection

In MON and CCV test sites both small sized (shallow and accurate) and large sized (deeper) 3D ERT surveys were performed, to calibrate ERT models with the monitoring stations soil probes and to extend the ERT investigation to a slope scale.

3D-S ERT are high-accuracy, small-scale surveys (Fig. 3A), acquired centered on the monitoring stations, involving the use of a snake grid comprised of 12 × 4 electrodes spaced 1 m apart both along the X and Y axes.

3D-L ERT are deeper, large-scale surveys (Fig. 3B), acquired centered on the monitoring stations, involving the use of a snake grid comprised of 12 × 4 electrodes spaced 10 m apart both along the X and Y axes. A sequence comprising 270 quadrupoles, using all feasible X, Y and diagonal dipole–dipole array configurations was used.

Time-lapse surveys 4D-S were finally performed, involving the same experimental setup of 3D-S ERTs and were undertaken before and after water injection (Fig. 3). Time-lapse resistivity monitoring was used to assess soil drainage conditions during the water infiltration following a simulated precipitation. For this purpose, multitemporal high detailed 3D acquisitions were performed around the monitoring stations, effectively producing a 4D acquisition. Those acquisitions consisted in two sequential surveys (pre-irrigation and post-irrigation) using the same arrays, reconstructing in fact time-lapse surveys.

In both test sites, immediately after the first 3D-S acquisitions, very intense precipitations of about 40 mmwere simulated for 1.5 h, before the post-irrigation surveys, to drive water infiltration up to the deepest horizons (about 1.2 m from the ground).

According to the Alpert et al.’s (2002) classification the irrigations simulated heavy-torrential rainfalls, typical of dry period convective storm in Oltrepò Pavese (Bordoni et al. 2019). Moreover, a landslide event occurred on 5–6 March 2016 close to MON monitoring station was triggered by a cumulated rainfall of 42.6 mm in 24 h (Bordoni et al. 2021a, b). More in detail, in both test sites, water injection was performed manually over the array’s areas, through a continuous and regular flow of 280 l of water distributed a very small buffer (0.05 m of thickness), along a grid composed by 7 rows, with a length of 10.5 m, and 9 columns with a length of 7.5 m, for a total area of irrigation of about 7.25m2. In fact, second acquisitions (post-irrigation) recorded during water infiltration, to give up the temporal variable related to the drainage velocity of both test sites.

For MON_3D_S model (Fig. 3A), the center of the survey is positioned on the sensors, 2.2 m upstream from the station with a solar panel and batteries with azimuth 263°. For CCV_3D_S model (Fig. 3B), the center of the survey was positioned on the sensors, 2.2 m in the SW direction with azimuth 265° from the station with solar panel and batteries.

Data inversion was performed using ERTLab Solver (Release 1.3.1, by GeostudiAstiers.r.l.—Multi-Phase Technologies) based on tetrahedral FEM (finite element modelling).

ERT data inversion

To obtain a true resistivity model of the subsurface an inversion procedure was needed, summarized in the schematic diagram of Fig. 4 (Loke and Barker 1996). Changes in electrical resistivity correlate with variation in solid material, water saturation, fluid conductivity and porosity, which may be used to map stratigraphic units, geological structure, fractures, cavities, groundwater (Burger et al. 2008; Meglich et al. 2013; Torrese and Pilla 2021). The depth of investigation depended on the distance between the current electrodes. The arrangement of current and potential electrodes during the measurement was dependent on the chosen electrode array. Also, the resistivity signature of different hydrogeological bodies depends on the size of the target in relation to its depth and on the contrast between the resistivity of the target and that of the host rock. The amplitude of resistivity anomalies is an inverse function of the distance between the measurement points and the target. The deeper the target, the lower the reliability for the method in identifying it. The depth of investigation, the vertical and horizontal resolutions of electrical resistivity surveys were linked to i) the electrode spacing, ii) the configuration array, iii) the quadrupole sequence, iv) the signal-to-noise ratio (S/N) ratio, v) the contrast between the resistivity of the target and the surrounding rock and/or background resistivity (Torrese and Pilla 2021).

Fig. 4
figure 4

ERT inversion schematic diagram used in 2D-3D and 4D tests, to retrieve 2D resistivity sections and 3D resistivity models

Full size image

Resistance is calculated by Ohm’s Law which relates voltage (V), current flow intensity (I) and resistance (R) as:

$$R=V/I$$

Resistivity (ρ) is calculated by multiplying resistance (R) by a geometrical factor (Κ) which is related to the arrangement of current and potential electrodes during the measurement:

$$\rho =R\mathrm{\;x\;}K=V/I\mathrm{\;x\;}K$$

Tetrahedral discretization was used in both forward and inverse modeling. The foreground region was discretized using a 0.75 m, 5 m and 0.5 m cell size, for 2D, large 3D and small 3D surveys, respectively, i.e., half of the electrode spacing, to give the model higher accuracy. The background region was discretized using an increasing element size towards the outside of the domain, according to the sequence: 1 × , 1 × , 2 × , 4 × and 8 × the foreground element size.

The forward modeling (Fig. 4) was performed using mixed boundary conditions (Dirichlet–Neumann) and a tolerance (stop criterion) of 1.0E-7 for a SSORCG (Symmetric Successive Over-Relaxation Conjugate Gradient) iterative solver. Data inversion was based on a least-squares smoothness constrained approach. Noise was appropriately managed using a data-weighting algorithm that allows the variance matrix after each data point iteration that was poorly fitted by the model to be adaptively changed. The inverse modeling was performed using a maximum number of internal inverse PGC (Preconditioned Conjugate Gradient) iterations of 5 and a tolerance (stop criterion) for inverse PCG iterations of 0.001. The amount of roughness from one iteration to the next was controlled to assess maximum layering: a low value of reweight constant (0.1) was set with the objective of generating maximum heterogeneity.

The inverse resistivity models were obtained by merging and jointly inverting datasets from different arrays which can deliver better detectability and imaging and, hence, provide more accurate inverse models and more reliable ERT imaging (Szalai et al. 2009; Torrese 2020; Torrese and Pilla 2021; Torrese et al. 2021, 2022). Inversion involved the application of homogeneous starting models that set at each node the average measured apparent resistivity value. The final inverse resistivity models were chosen based on the minimum data residual or misfit error.

Results

Lithological sections and ERT calibration

In both the test sites, the eluvial horizon was characterized through log surveys derived from manual drilling (micro-boreholes) and trenches. The first results from the micro-boreholes confirmed that in MON test site soil thickness ranges from few tens of centimeters at slope top to about 1.6 m downslope of the monitoring station, while in CCV test site the soil thickness ranges from few tens of centimeters at slope top to about 1.7 m downslope.

From the samples collected within the trenches in both the test sites, a characterization of the main geotechnical parameters, such as the soil plasticity and the grain size distribution, was performed.

According to USCS classification, soil horizons in MON test site are essentially low-plastic soils (CL), locally alternated by heterogeneities of medium-plastic or high-plastic soils (CH) soils, while for CCV soil horizons are mainly high plastic soils (CH). The grain size curves confirmed the predominance of silty components with high content of clay and low percentages of sand and gravel in MON test, while in CCV test site prevail clayey components with high content of silt. Complete results of samples analyses from MON and CCV test site are summarized in Table 1.

In MON test site the shallowest soil horizon consists in dark-brown clayey sandy silt, reach of organic matter and carbonate concretions and plant remains (0.0–0.2 m depth), overlapped to a discontinuous calcichorizon,

enriched in harder carbonate concretions (0.2–0.9 m depth), up to the weathered bedrock composed by weakly-cemented conglomerates (Fig. 5). CCV soil profiles enlightened instead a first horizon (0.0–0.2 m depth) characterized by clayey-silty soil, with carbonates pebbles above a deeper level with increasing plasticity (0.2–0.7 m depth). Increasing in depth (about 1.0–1.5 m depth) soil becomes more clayey and shows the evidence of water table variations (Fe oxides) going towards the marly fractured bedrock (> 2 m depth) (Fig. 5).

Fig. 5
figure 5

Representative stratigraphic setting of soil covers. Soil logs were performed on MON-CCV test sites trenches, used for ERT calibration.Lithological sections derived from geological models of MON and CCV test sites, with micro-boreholes logs.Soil horizons were indicated following pedological (USDA 2014), lithological and engineering soil classification (USCS). Schemes of monitoring stations with soil probes depths

Full size image

The analysis of micro-boreholes performed in the test sites allowed instead to build the soil profiles of the test sites. From the interpolation of the soil profiles lithological models of the two test sites were reconstructed, to carry out the lithological sections along the ERT profiles.

MON slope is characterized by an increasing thickness of the eluvial coverage from the ridge (S2 – 0.70 m) towards the slope (1.20 m—S1 and S4) up to the lower part of the slope, where the soil coverage ranges from 1.10 m (S5) and 1.85 (S7).

For the MON_2D calibration 5micro-boreholes drilling and 1 trench were considered, were correlated respectively for the MON 2D_1 survey the S3-S4-S5-S6 surveys, and for the MON 2D_2 survey the S2-S6 (Fig. 5). At CCV test site, the slope soil profile was performed based on a stratigraphic log (Op 67/15), analyzing the lithology of the first 2 cm of the colluvial coverage. The soil profile log Op 67/15 (Fig. 5) was carried out close to the soil monitoring station and was used for the ERT calibration.

According to USDA 2014 pedological classification the two representative test sites soils logs, derived from trenches, develop with the sequence A-B-C, as shown in Fig. 5.

2D ERT models

For both the test sites, 2D ERT surveys were performed during the wet period (March). The most superficial (0.2 m) SWC measured by the monitoring stations during the ERT acquisitions were respectively 0.268m3/m3in MON and0.171m3/m3in CCV, corresponding to a saturation degree respectively of 81% for MON and 35% for CCV.

In MON test site 2D array has shown high values of resistivity in the superficial horizons, becoming lower increasing the depth, according to the presence of silty and clayey soils in the superficial layers. While, in CCV test site 2D array measured a decreasing of resistivity from the ground level to the deeper layers.

More in details, in MON_A-A’ 2D section (Fig. 6A) a discontinuous layer with low resistivity (9–50 Ω m) characterizes the first decimeters of depth and can be interpreted as the clayey-sandy silt coverage, which constitutes the most superficial part of the eluvial layer. Below this coverage there is a layer where resistivity increases (50–75 Ω m), reaching an average depth of ≈3 m. These resistivity values ​​define a layer rich in carbonate concretions and multi-centimeter pebbles. Within this continuous level there are high resistivity anomalies (> 75 Ω m) which may represent areas of accumulation of material from surface landslides. The deeper part of the section shows low resistivity values (9–50 Ω m) and can be interpreted as the conglomerate bedrock, that presents local low resistivity anomalies (lower than 9–50 Ω m) that identify areas with higher permeability (fractured-weathered bedrock). This section also enlightens the presence of deep structures, similar to roto-translational landslide main scarps, in the middle of the slope. In particular, horizons above this scarp are interested by lower values of resistivity, potentially related to the presence of deep groundwater layers.

Fig. 6
figure 6

Resulting models from 2D acquisitions. In figure A is shown the model related to MON test site. A and B are the models derived from the MON 2D ERT profiles (A along the maximum slope line and B along the slope direction) C and D are the models derived from the CCV 2D ERT profiles (C along the maximum slope line and D along the slope direction)

Full size image

MON_B-B’ 2D (Fig. 6B) section revealed widespread heterogeneities of resistivity below the surface. A discontinuous layer with low resistivity (9–50 Ω m) characterizes indeed the first meter of depth and can be interpreted as a clayey-sandy silt, which constitutes the most superficial part of the eluvial layer. Below, there is a layer with high resistivity (75–291 Ω/m) with an average thickness of 2 m, probably due to lithological variations, according to the presence of buried ancient landslides along this slope. These resistivity values ​​define a layer rich in carbonate concretions and multi-centimeter pebbles. Within this continuous level there are high resistivity anomalies (> 100 Ω/m) which represent areas of accumulation of relict landslide material, probably related to past landslides events. The deeper horizon presents a continuous low resistivity layer (9–50 Ω/m) with an average thickness of 4–5 m, decreasing from SSE to NNE, overlying the deepest horizon with higher resistivity, related to the compactness of the substrate.

In CCV_A-A’ 2D (Fig. 6C) section a discontinuous layer with low resistivity (3.7–6 Ω/m) characterizes the first decimeters of depth and can be interpreted as clayey silt which constitutes the most superficial part of the eluvial layer. On the top of the slope, to N-NE this superficial layer shows high resistivity values, probably due to drier conditions of the topsoil. Below, there is an area with higher resistivity (9–15 Ω m) which reaches an average depth of ≈3 m and may represent a very fractured calcareous marly bedrock, alternated with clayey lenses. Below the monitoring station, an anomaly of high resistivity values was detected and can be due to the remolded material during the station installation. The lower area presents low resistivity values (6–7 Ω m) which can be interpreted as the marly substrate, inside which low resistivity anomalies (< 5 Ω m) were identified at a depth of about 7 m, representing a higher clay content area. CCV_B-B’ 2D (Fig. 6D) section revealed a continuous low resistivity layer (3.7–6 Ω/m), that characterizes the first meter of depth and can be interpreted as a clayey, silty, plastic horizon, rich in carbonate concretions. Below this there is an area with higher resistivity (9–15 Ω/m) which reaches an average depth of 3–4 m: these resistivity values define a layer where there are very fractured calcareous marls alternating with clayey lenses. The lower layer has low resistivity values (6–7 Ω/m) and can be interpreted as the marly substrate, with low resistivity local anomalies (< 5 Ω/m) at a depth of about 7 m, which represents an area (laterally extended for about 40 m) with a higher clay content.

3D ERT models

For both the test sites, 3D-L ERT surveys were performed at the end of the wet period (May). The most superficial (0.2 m) SWC measured by the monitoring stations during the ERT acquisitions were respectively 0.312 m3/m3 in MON and 0.337m3/m3 in CCV, corresponding to a saturation degree respectively of 63% for MON and 69% for CCV. 3D-L ERT models were performed to obtain more detailed surveys of the test sites, close to the monitoring stations, allowing the correlation with the soil moisture values measured at the same time of the ERT acquisitions.

MON_3D-L planimetric section (Fig. 7B) shows the surface conditions in terms of resistivity. The lower part of the slope (E-NE) surface presents low resistivity values (5–40 Ω/m) while the medium and the highest part (to W-SW) of the slope present higher resistivity values (75–366 Ω/m). This suggests a change in drainage that becomes smaller, from the upper to the lower part of the slope. MON_3D-L model revealed an area with heterogeneous resistivity values (Fig. 7A). The lower portion of the slope is characterized on the surface by relatively low resistivity values (5–50 Ω/m) which develop in depth along the entire slope to reach the surface in the opposite side of the slope (WSW). Anomalies with higher resistivity (75–366 Ω/m) are present in the upper portion of the slope, developing from the surface to a depth of 10 m and, for a smaller volume, at the base of the slope at a depth of 5 m with an average thickness of 7 m. This resistivity range identifies areas of the slope that have a high drainage and that have been affected by surface landslides and the consequent dislocation of the eluvial material.

Fig. 7
figure 7

Resulting models from 3D-L acquisitions. A–C represent the 3D bulk of both test sites, while B –D represent the planar imaging of the topographic surface of both test sites

Full size image

Through the CCV_3D-L planimetric section (Fig. 7D) it is instead possible to observe the surface variation of the soil resistivity. The surface is homogeneously occupied by low resistivity values (2.1–7 Ω/m) that can be correlated to the clayey silty horizon that characterizes the eluvial layer. On the two opposite sides there are two belts with higher resistivity (7–16.7 Ω/m), attributable to surface drainage areas. The conditions observed on the surface recur overall also in depth. The CCV_3D-L model (Fig. 7C) highlights volumes of soils with lower resistivity (2.1–5 Ω/m) in the southern sector at the foot of the slope under investigation, with an average thickness of 10 m. Geological bodies with higher resistivity (7–9 Ω/m) occupied the eastern area from the monitoring station to the S-SW sector of the slope up to the maximum depth of investigation. The presence in depth of bodies characterized by a high anomaly (7–16.7 Ω/m) is also revealed in the planimetric section, as higher resistivity belts.

Time-lapse (4D) ERT models

4D ERT surveys were performed at the end of summer, in the driest period (early September), 6 days (MON) and 9 days (CCV) after a convective precipitation event (4–5 mm), both from 11:00 AM to 14:00 PM, with water infilling at 11:30 AM. In both the test sites two acquisitions were performed sequentially, with an identical array characteristic (3D-S—high resolution survey). First acquisitions were performed under undisturbed conditions (natural soil water content), followed by second acquisitions achieved under disturbed conditions (after water infilling), implementing in fact the Time-Lapse surveys. From all the acquisitions were built 2 models for each test site: 3D-S undisturbed (Fig. 8A and B) and 3D-s disturbed (Fig. 8C and D). Before starting the ERT sessions the soil water content at 0.2 m from the ground was 0.156m3/m3 for MON and 0.225 m3/m3 for CCV, corresponding to a saturation degree respectively of 47% for MON and 46% for CCV.

Fig. 8
figure 8

Resulting models from 4D (time-lapse) acquisitions. A–B show the results of the Undisturbed acquisition, before water infilling. B–C show the results of the Disturbed acquisition, during and before water infilling

Full size image

In MON_3D_S model (Fig. 8C), after irrigation, the resistivity values decreased on the surface (positive values) along the lateral bands of the investigated area due to the induced perturbation and increased in depth (negative values) due to the deep drainage of rainwater. The central strip of the area (where the station is located) has not undergone any changes.

In CCV_3D_S model (Fig. 8D), after irrigation, the resistivity values decreased in depth (negative values) due to the rapid infiltration of irrigated water due to the plowing of the soil and increased on the surface (positive values) due to evaporation due to solar radiation. The variations in resistivity were not very high due to exposure to solar radiation and, secondarily, to the limited general drainage of the lithotypes. The models also enlighten the water infiltration as resistivity values variations, going from higher to lower resistivity values, respectively for undisturbed and disturbed conditions. Those trends were also compared with the monitoring stations TDR of both test sites, which measured water content variations after the water infilling, at different times, depending on the permeability of the different soil horizons (Table 3).

Table 3 Soil water content measured by TDR onboard the monitoring stations during the disturbed 3D Time-Lapse acquisitions for both test sites, after water infilling
Full size table

Subtraction models, derived from the difference between undisturbed models and disturbed models of each test site, allowed to evaluate the infilled water perturbation in depth (Fig. 9A and B). In both test sites negative resistivity values (-10 to -0.01 Ω/m) correspond to water perturbation (Fig. 9C and D), while no resistivity values variations (0 Ω/m) correspond to water perturbation limit (Fig. 9E and F).

Fig. 9
figure 9

Resulting models from 4D (time-lapse) acquisitions. Those results derived from the differences between the undisturbed and the disturbed acquisitions. A–B show the results of the full range of resistivity values, representing the complete volume of soil. C–D show the negative resistivity values, showing the soil volumes reached by infilled water (perturbation). E–F show no resistivity variations, representing the limit of water perturbation in depth

Full size image

Water infiltration in both test sites was thus given by the measured resistivity variations during the Time-Lapse, in terms of negative resistivity changes associated with the vertical progressing water. For both test sites, the average depths of water perturbation were retrieved by means of the difference between the topographic surfaces from MON and CCV DTMs and the 0 Ω/m subtraction models, that correspond to the limit of infilled water infiltration in depth.

In particular, for MON test site the model shown an average depth of water perturbation of about 0.7 m, locally up to 1 m from ground level, achieved in 127 min of the completed water infilling and corresponding to a bulk hydraulic conductivity of 1.4410−4 m/s (Fig. 9E). This estimated value is higher than laboratory (through an evaporation test) and field (through an amoozemeter) measured values, which were in the order of 10–6 and 10–5/10–6 m/s, respectively (Bordoni et al. 2015, 2021b). MON test site is interested by the presence of deep soil shear fractures, related to the presence of ancient sliding surface. These fractures have an irregular distribution pattern, due to the strong soil heterogeneities of the test site and are preferential paths for fast water deep infiltration. Simultaneously, TDR onboard the MON monitoring station confirmed the water drainage from the surface as an increasing of soil water content (SWC): < 40 min to reach 0.4 m and about 40 min to reach 0.6 m. TDR monitoring in MON enlighten a slow water infiltration in the surface, while the velocity increases with depth, due to a faster deep drainage.

For CCV test site instead the model shown an average depth of water perturbation of about 0.9 m, locally up to 1.1 m from ground level, achieved in 122 min of the completed water infilling, corresponding to a bulk hydraulic conductivity of 1.6410−4 m/s (Fig. 9F).As in MON test-site, the estimated value is higher than laboratory (through an evaporation test) and field (through an amoozemeter) measured values, which were in the order of 10–6 and 10–5 m/s, respectively (Bordoni et al.2021b). CCV test site is interested by the presence of deep soil cracks, related to the soil amount of clay minerals in dry conditions. The fractures network presents a regular distribution pattern, with spacing from 20 to 40 cm (Bordoni et Al. 2021a), while the fractures opening may reach 5-10 cm with a depth up to 1 m (Fig. 10), becoming a preferential path for fast water deep infiltration. Differently from MON test site, in CCV test site the model shown instead no significant changes in resistivity values in surface, related to a very fast water infiltration from the surface to the ground, while more in depth, resistivity values become negative, according with a slower infiltration velocity. TDR onboard the CCV monitoring station confirmed the fast water drainage from the surface to 0.2 m (instantaneous) and to 0.4 m(< 20 min) while, from0.4 m water infiltration strongly decreases in velocity, TDR reaching 0.6 m after < 2:00 h.

Fig. 10
figure 10

Soil desiccation cracks forming after prolonged dry periods in CCV test site

Full size image

Conceptual models from ERT interpretation

Conceptual models are the resulting combination of both resistivity values and lithological classification (Fig. 11).

Fig. 11
figure 11

Conceptual models from the combination of resistivity values and soil logs. A–B–C–D represent the 2D sections of both test sites. E–F represent the slope scale 3D conceptual models of both test sites, derived from 3D_L

Full size image

2D models performed in MON test site (Fig. 11A and B), reveal a superficial layer with a general low permeability, characterized by dense clayey soils, intercalated by some anomalies of higher permeability, characterized by clayey-silty soils with pebbles. The permeability increases definitively at a depth of about 3-4 m from ground level, probably for the presence of fractured conglomeratic bedrock.

2D models performed in CCV test site (Fig. 11C and D), reveal instead a superficial layer of about 4-5 m from the ground level where the permeability is high, and the drainage to the deepest horizon is faster. Below this horizon, on the top of the slope the permeability increases due to a fractured marly-arenaceous bedrock, while in the bottom part of the slope the permeability decrease, for the presence of a chaotic-clayey bedrock.

3D-L models were built from ERT surveys both at MON test site (Fig. 11E) and CCV test site (Fig. 11F) trough interpolation of 2D conceptual models.

Discussion

All ERT surveys produced data useful to characterize the test sites at different scales, from small to slope scale, allowing also a calibration with lithological logs and hydrological data from soil monitoring stations. Indeed, the inversion of the experimental dataset has on average shown a good correlation to the available stratigraphic logs, as well as a fair detection of the different lithological levels of the surface blanket. However, there were some issues related to the investigation resolution, not always enough to resolve particularly thin layers. This implies that the greater the investigation depth the less the ability to resolve a slight lithological change, not associated with a real variation of resistivity (and chargeability) distinguishable and therefore analyzable. In addition, the rapid changes in resistivity values produce a buffer zone that does not detect a concrete change in lithology: this effect is due to the data inversion and has therefore been considered for the purposes of geological interpretation.

Another consideration regards the decreasing resolution at the lateral edges in the 2D inverse tomographic sections. 2D ERT arrays were useful to understand the subsurface conditions at slope scale, reaching a depth of ≈15 m from the ground level and allowing the understanding of the groundwater circulation, in both the test sites slopes.

Given that the hydrogeological conditions are an essential parameter for slope instability characterization, with particular reference to shallow landslides, ERT analysis performed in MON and CCV test sites have proved to be a useful tool for hydrogeological conditions characterization. All the surveys performed in this work demonstrated that the geological differences may affect in different ways water circulation and drainage overall, both at local and slope conditions.

Moreover, stratigraphic logs shown a good correlation with ERT investigations, validating the results of the geophysical investigations. In general, stratigraphic logs have found good correlation with the bodies identified in the 3D_S tomographic models, that have a higher resolution.

In MON test site 2D resistivity model, high resistivity horizons were identified in the first 4–5 m of soil, finding a good correlation with the lithology of the first 5–6 m horizons, mainly composed by hard dense clayey-sandy silts, with lower permeability. Deeper horizons, below 5–6 m from the ground, have shown lower values of resistivity, according to the lithology of this horizon, with a higher amount of sand, which is more permeable. In addition, geoelectric investigations confirmed the depth and the trend along the slope of the sliding surfaces of shallow landslides and since it was possible the detection of possible incipient fractures in the sliding surfaces. On the other hand, in CCV test site superficial layer shows high resistivity values, probably due to drier conditions of the topsoil. Below, there is an area with higher resistivity (9–15 Ω m) which reaches an average depth of ≈3 m and may represent a very fractured calcareous marly bedrock, alternated with clayey lenses.

The interpretation of the inverse models, made it possible through calibration with stratigraphic logs, associates the lithological variations with the different resistivity ranges, allowing to interpret and set up conceptual models of the first subsoil horizons. The combination of geophysical data with observation of stratigraphic logs allowed to develop 2D and 3D conceptual models, highlighting the lithological and hydraulic features that characterize the subsoil of the investigated test sites.

The time-lapse high-resolution 4D-ERT results derived from the differences between the imaging of undisturbed/perturbed conditions comparison, after simulating a heavy-torrential precipitation of about 40 mm, above the cumulated thresholds for shallow landslides triggering in the study area (Bordoni et al. 2021a, b). The ERT monitoring provided information about the way in which the slope responds to the infiltration of rains and the consequent variations of the soil moisture. 4D confirmed thus the differences between the two test sites, in terms of water drainage from ground level to deeper horizons.

The velocity of infiltration in MON test site increased from the surface to the deepest horizons, probably related to the presence of deep shear cracks due to ancient shallow landslides sliding surfaces, which can modify deeper water circulation, increasing in fact the velocity of drainage. This feature finds a good correlation with the lithotypes of the first 5–6 m horizons, mainly composed by hard dense clayey-sandy silts. However, the strong variations in resistivity after irrigation were probably influenced by lithological heterogeneities, due to the presence of ancient landslides. Differently, in CCV test site infiltration velocity was very fast from the surface to about 0.4 m, to become slower in depth, as also confirmed by TDR onboard the CCV monitoring station. A more compact soil horizon below 0.4 m may reduce the permeability in depth. In surface, resistivity variations were not so strong, probably for the fast drainage of the most superficial horizon, interested by deep soil cracks (up to 1 m), typical of the dry period. This feature finds a good correlation both with the heterogeneities of lithotypes in the first meters (weathered marls, silty clays, fractured marls) and the mechanical conditions of the first decimeters of soil, which was strongly plowed by agriculture activities. Moreover, the high velocity of the superficial drainage might be influenced by the strong solar radiation exposure of the field.

In MON test site bulk water infiltration could be affected also by the local lithological heterogeneities due to the presence of ancient shallow landslides, as well as the presence of tree roots, while in CCV test site water infiltration is mainly affected by the presence of deep fractures and cracks that increase the bulk permeability. The CCV colluvial cover is indeed characterized by a very high and high swelling-shrinking potential (Meisina 2004). The estimated bulk hydraulic conductivity thus turns out to be higher than the laboratory and field point measures, for both test sites. 4D surveys, indeed, were performed in a soil bulk, over a large portion in field, involving the effect of fractures and cracks present during dry months. The effect of these cracks becomes thus more evident as an increase in bulk hydraulic conductivity estimated through geophysical datasets, compared to the limited portion of soil matrix analyzed through sample or field tests.

Bulk hydraulic conductivity allowed in fact to cover a more representative setting of each studied hillslope in respect of punctual measures, such as laboratory analyses (from soil samples) or infiltration tests (amoozemeter). Bulk hydraulic conductivity and surface soil drainage after heavy rainfall precipitations are significant parameters in shallow landslides triggering, as well as the soil moisture, that can determine changing in physical properties, such as shear strength. This suggests, as possible implementation of this work, the development of an integrated early warning system at slope scale which includes an automated ERT system (A-ERT). This would provide a continuous resistivity monitoring, allowing an automated and continuous long-term measurement, at slope scale: it is worth pointing out that: i) electrode spacing directly influences the depth of investigation, as well as the vertical and spatial resolution, allowing to reach the sliding surface depth; ii) the signal to noise ratio (S/N) affects the depth of investigation; iii) more accurate inverse models and more reliable ERT imaging are provided by merging and jointly inverting datasets from different arrays (i.e. Wenner and dipole–dipole).

Conclusions

This experimental work confirmed the reliability of geophysical surveys, by means of electrical tomography (ERT)for hydrogeological characterization, from the high detailed to the slope scale.

ERT surveys combinations were performed in the experimental test sites of Oltrepò Pavese area (northern Italy) and focused on slopes susceptible to shallow landslides investigation. For this purpose, inverse resistance models were generated starting from experimental datasets, collected during the geophysical campaigns carried out at the experimental sites of MON and CCV.

The geophysical surveys involved linear geometry (2D) and snake type (3D) geoelectric layouts with Wenner-Schlumberger and Dipole–Dipole electrode arrangement. An ERT time-lapse survey was also conducted for both experimental sites, investigating the change in the resistivity of the subsoil in a defined portion (10.5 × 7.5 m) of a slope subject to intense precipitation (artificially simulated by manual irrigation).

The results produced by the inversion of the experimental dataset generated two-dimensional sections and 3D models, proving to be very useful for improving knowledge of the investigated phenomenon by characterizing the lithology and hydrogeological characteristics of the subsoil of the studied area. To complete the analysis, ERT resistivity models were calibrated on stratigraphic logs to discretize and characterize the subsoil with higher accuracy as well as ERT resistivity values were compared to soil probes water content measurements, to compare resistivity models to real soil water content. This study highlighted, in addition to the nature of the material composing the colluvial cover and the underlying substrates, also the sliding surfaces (cover-substrate limit) and the superficial and deep drainage systems present along the slope.

Finally, 4D-S time-lapse surveys provided information on the hydraulic dynamic characteristics of the first subsoil, subject to a simulated precipitation by artificial irrigation. The interpretation of ERT time-lapse models allowed to understand the water drainage and the hydrological rheology of the different soil horizons, in the proximity of both the soil monitoring stations of test sites. In MON test site were observed two different velocities of drainage from the surface to the depth: deepest layers seem to drain much more quickly than superficial ones. In CCV test site, drainage was faster in the surface, becoming slower below 4–5 m from ground level, probably related to the field managements within the first 40 cm from the ground, that strongly affect the permeability, combined to the presence of deepest soil fractures (cracks). These surveys allow thus to estimate the bulk hydraulic conductivity of shallow soils, also involving the effects of cracks and fractures which cannot be caught well through small samples or point field measures, providing a more realistic infiltration model of the natural system.

Under the light of the results obtained in this work, is possible to assert that the proposed ERT method finds a good applicability in the characterization of the areas affected by rainfall-induced shallow landslides, allowing a better knowledge of the hydrogeological precursors of landslides (infiltration of water in the first soil horizons). The application of geophysical methods combined with geological and geotechnical analyses is definitely an excellent tool for understanding and improving the prediction of surface landslides, becoming much more significant in light of climate changes of last decades. Indeed, in the last 15 years, climate changes affected the rainfall distribution during the hydrological year, as observed by the hydrometeorological monitoring stations installed in both test sites: prolonged periods of water deficit alternated to heavy rainfall events was observed. After such long dry periods clayey soils are affected by remarkable cracks opening, in some cases with 10–15 cm opening with a depth higher than 1 m, which is the typical depth of sliding surfaces in the study area. This research highlighted the importance of mud cracks as preferential ways of water infiltration towards the sliding surfaces, in fact increasing slope failures occurrence probability.

Since this research highlighted the suitability of the proposed ERT-based experimental integrated approach to define shallow landslides predisposing and triggering factors, some possible future implementations of this work should be slope stability assessment and integrated early warning systems development. Slope stability assessment could be performed, for example, using data from this approach (e.g., bulk infiltration, slope scale water content), not only for the research test sites, but also for other sites with different geological-geomorphological contexts. Possible future development of an integrated early warning system at slope scale would include an automated ERT system (A-ERT), providing a continuous resistivity monitoring and allowing an automated and continuous long-term measurement, at slope scale. For this purpose, the setup of empirical and physically based rainfall thresholds remains essential, combined to meteorological-soil monitoring, in terms of cumulative rainfalls and soil water content (soil probes and A-ERT monitoring), in order to define the slope-specific critical hydrogeological conditions. While our test sites are representative of northern Italian Apennines, areas with different climatic, hydro-geological and geomorphological conditions could be good experimental sites to improve the knowledge of hydro-geological dynamics in such prone to shallow landsliding slopes.

This paper thus opens interesting prospects for the assessment of soil water content and water deficit, through application of ERT measurements. The proposed approach supports future research also related to climate changes, to develop landscape management strategies (e.g., agricultural managements and land use), to understand and reduce the water deficit impact on environment and to prevent slopes degradation.