Water-table and discharge changes associated with the 2016–2017 seismic sequence in central Italy: hydrogeological data and a conceptual model for fractured carbonate aquifers

The seismic sequence recorded in central Italy in 2016–2017 included nine main events (Table 1) with moment magnitude (Mw) ≥5.0 (four of which were Mw ≥ 5.5) occurring on four separate days (August 24th 2016, October 26th and 30th 2016 and January 18th 2017), as described in detail in Chiaraluce et al. (2017) and ISIDe Working Group (2016). The main events caused several observed changes in groundwater dynamics, including spring discharge variation, water-table anomalies and river discharge alteration in different basins located up to 100 km from the epicentral zone. The fractured and locally fissured carbonate nature of the aquifers outcropping in the earthquake area favors a quick co-seismic response in terms of pore pressure propagation; however, the observed sustained changes, which developed during several days after the main shocks, affected groundwater dynamics for several months after the seismic events.

Hydrogeological changes caused by earthquakes have been historically reported. Instrumentally measured responses, however, have become available only in the last few decades. These responses include changes in water level (Leggette and Taylor 1935; Cooper et al. 1965; Roeloffs 1998; Brodsky et al. 2003; Roeloffs et al. 2003; Lachassagne et al. 2011; Shi et al. 2015), temperature (Mogi et al. 1989), chemical composition (Claesson et al. 2004; Skelton et al. 2014), stream flow (Manga et al. 2003; Montgomery and Manga 2003; Manga and Rowland 2009; Muir-Wood and King 1993; Rojstaczer et al. 1995), and spring attributes (Wang and Manga 2015). Understanding the origin of these hydrological and hydrogeochemical phenomena may have significant impacts on the comprehension of the occurrence of liquefaction (Cox et al. 2012), water supply and quality (Gorokhovich and Fleeger 2007), underground storage (Wang et al. 2013) and pore-pressure triggered seismicity (Brodsky et al. 2003).

The effects of earthquakes on groundwater are commonly divided into “transient oscillations” (Cooper et al. 1965) and “sustained offset”, which include abrupt rises or falls and sustained gradual rise lasting for several days after the shock (Roeloffs 1998; Yan et al. 2014). The most frequent consequences of earthquakes are spring and river discharge increase and water-table rise, which are generally attributed to four general classes of possible explanations: (1) co-seismic static strain increases pore pressure that may contribute to change permeability (e.g. Wakita 1975; Jonsson et al. 2003); (2) earthquake-related dynamic strains may increase permeability, permitting a more rapid flow, which in fractured aquifers can be enhanced by fracture cleaning, eventually increasing discharge (e.g. Briggs 1991; Rojstaczer and Wolf 1992; Rojstaczer et al. 1995; Sato et al. 2000; Wang et al. 2004a; Curry et al. 1994; Amoruso et al. 2011); (3) breaching of hydraulic barriers or seals (e.g. Sibson 1994; Brodsky et al. 2003; Wang et al. 2004a); (4) the excess of water discharged after the earthquake lies in the shallowest subsurface where water is liberated by the consolidation or even liquefaction of near-surface unconsolidated materials (e.g. Manga 2001; Manga et al. 2003; Montgomery et al. 2003).

Looking at the relationships between tectonic framework, hydrogeological setting and earthquakes from a wider point of view, recent research activities have highlighted the role of fluids at crustal scale during the seismic cycle. Doglioni et al. (2014), for instance, suggest that fluid flow rates differ during the different periods of the seismic cycle (inter-seismic, pre-seismic, co-seismic and post-seismic periods), also in connection with the tectonic style. In particular, they hypothesize that in extensional tectonic settings like central Italy, the wedge of crust above the brittle ductile transition remains “suspended” while a dilated area forms during the inter-seismic period. This area would trap deep fluids, which when the wedge of crust above the brittle ductile transition starts to drop in the pre-seismic period, would be squeezed above due to the progressive fracture closing. Consequently, in the co-seismic period, aquifers can host changes in hydrochemistry (Barberio et al. 2017) and in water levels, independently from changes due to the previously listed local mechanisms, which can affect the hydrodynamics of the struck aquifers during and after the seismic sequence. Such a comprehensive tectonic model allows looking at changes induced in groundwater after earthquakes in a general framework of crustal deformation, suggesting the role of deep inputs in triggering the aforementioned well-known processes (as pore pressure changes, permeability increase, liquefaction/consolidation, etc.) acting at the aquifer scale.

The effects of past earthquakes on groundwater in central Italy have been described by previous papers. Esposito et al. (2001) describe the effects of four earthquakes in southern Apennines including the 1980 Irpinia earthquake, which generated important hydrogeological changes as far as 200 km from the epicenter, including a significant increase of Caposele spring flow. Amoruso et al. (2011) describe the hydrogeological changes in a fractured aquifer after the L’Aquila 2009 earthquake, inferring that those changes were probably connected with the increase of bulk hydraulic conductivity at the aquifer scale, mainly due to fracture cleaning, raising hydraulic heads in the discharge zones, and correspondingly lowering them in the recharge areas (Adinolfi Falcone et al. 2012; Galassi et al. 2014).

The aim of this paper is to present an overview of the effects of the 2016–2017 seismic sequence on the dynamics of groundwater flow in central Apennines, analyzing the extent of the impacted area, possible relations between tectonic environments, geological-hydrogeological setting, and groundwater changes, and providing preliminary considerations on the possible causes of the observed phenomena.

The central Apennines (Italy) is a Meso-Cenozoic ENE-dipping thrust-and-fold belt mainly developed during upper Miocene-Quaternary, composed by a pre-orogenic Triassic-Miocene sedimentary succession overlain by Miocene and Pliocene synorogenic sediments, resulting in a highly variable facies and thickness distribution. A Meso-Cenozoic carbonate platform domain extends in the SE part of the study area (Latium Abruzzi Apennine, Fig. 1), consisting of a 5,000-m-thick sequence of limestone and subordinate dolomite of Upper Triassic to upper Miocene age (Brandano and Loche 2014 and references therein). In the western side of the area (Umbro Marchean Apennine), a Lower Jurassic carbonate shelf unit is overlain by stratified pelagic sediments (middle Lias-lower Miocene), with an overall thickness of 2,500–3,000 m (Marchegiani et al. 1999). The Apennine orogenesis overthrusts the Umbria-Marche succession onto the Latium–Abruzzi platform along the main regional thrust fault system named Olevano-Antrodoco line (Pierantoni et al. 2005 and references therein). From the upper Miocene to lower Pliocene, thrust migration towards the east was coupled with the progressive development of fore-deeps in front of the migrating fold-and-thrust belt (Cipollari and Cosentino 1995).

Since the upper Miocene–lower Pliocene, extensional faulting connected with the opening of the back-arc Tyrrhenian Basin has been dissecting the compressive structures (Boncio and Lavecchia 2000 and references therein), leading to the development of intermontane basins filled with thick continental sequences of Quaternary alluvial, detrital and lacustrine deposits (Cavinato and De Celles 1999). Some normal faults show evidence of Holocene activity, suggesting that they may be responsible for the seismic activity occurring in this sector of the Apennines (Cello et al. 1998), mainly confined within the upper part of the crust (<16 km; Lavecchia et al. 1994; Boschi et al. 1995).

In the study area, the 2016–2017 seismic sequence includes some of the largest instrumental earthquakes of the last 40 years (Norcia 1979 Mw = 5.9, Irpinia 1980 Mw = 6.9, Gubbio 1984 Mw = 5.2, Colfiorito 1997 Mw = 5.9, L’Aquila 2009 Mw = 6.3; Pantosti and Valensise 1990; Boncio and Lavecchia 2000; Deschamps et al. 2000; Chiarabba et al. 2009). The 2016–2017 sequence and its main shocks (Table 1) were generated by the Gorzano Mt.-Vettore Mt.-Bove Mt. faults (Galadini and Galli 2003, LMF and MVF in Fig. 2). The seismic crisis started with the August 24th 2016 event (Mw 6.0) and a further significant event on October 26th. The Vettore Mt. fault experienced a rupture with tectonic segments ~10 km long and a surface-displacement of ~30 cm (Smeraglia et al. 2017). The October 30th 2016 event (Mw 6.5) was generated by the rupture of the central zone of the fault by a normal movement. The focal mechanism, identical to the previous earthquakes, was a strike-angle of N155°, a WSW dip slip and a dip angle about of 50° in depth (RCMT 2016). During the October 30th 2016 event, the entire Vettore Mt.-Bove Mt. fault system gave origin to important surface faulting occurrences, reusing the pre-existing fault plane and redisplacing the fault segments previously broken.

The acquisition of the interferometric satellite data ALOS-2 from JAXA (Japan Aerospace Exploration Agency) and further interferometric analyses (INGV Central Italy Earthquake Team 2016) provided an estimation of the co-seismic subsidence along the NW–SE component reaching a maximum of ~80 cm (LOS: Satellite Observation Line). The horizontal co-seismic maximum movements consist of ~40 cm towards NE, ~30 cm towards SW as well as a maximum vertical movement of about 20–40 cm, when considering also the October 30th event (INGV Working Group GPS 2016; INGV Central Italy Earthquake Team 2016).

In the study area, the fractured carbonate ridges host the main aquifers, feeding several perennial springs (Nanni and Vivalda 2005; Martarelli et al. 2008; Mastrorillo et al. 2009; Mastrorillo and Petitta 2014) with steady regimen, located mostly at the external boundaries of the aquifers (Fig. 1). Groundwater flows in fissured to locally karstified carbonates. The Miocene-Pliocene synorogenetic silicoclastic sediments surrounding the carbonate aquifers, as well as the Plio-Quaternary deposits, filling the intermontane plains and the river valleys, act like aquitards (Petitta et al. 2011).

Widespread karst development, including endorheic basins, ensures high infiltration rates, from 500 to 700 mm/year, in the Umbria Marchean aquifers and up to 900 mm/year in the Latium-Abruzzi aquifers, collectively feeding a total discharge of about 300 m³/s (Boni et al. 1986, 2010). Fractures and karst conduits allow for fast vertical flow in the vadose zone, while the large thickness of the saturated zone facilitates a steady flow towards the basal springs that show outstandingly high and steady discharge (Petitta 2009; Amoruso et al. 2014; Fiorillo et al. 2015).

Co-seismic changes were examined in several observation sites located within the area affected by the earthquakes. Within the framework of a continuous monitoring, the collected data refer to piezometric heads in wells and piezometers, spring discharges and river hydrometric levels or discharges.

Figures 3, 4, 5 and 6 show the time plots of the available data from monitoring sites (location in Fig. 2), summarized in Tables 2, 3 and 4, which refer respectively to water levels in piezometers (W1–W5), spring discharge (S1–S12) and river discharge or levels (R1–R6). Red bars in Figs. 3, 4, 5 and 6 indicate the four main Mw > 5.5 events. For the sake of simplicity, the August 24th 2016, October 26th 2016, October 30th 2016 and January 18th 2017 earthquakes will be named hereinafter 1st, 2nd, 3rd and 4th events respectively. See Fig. 2 for the locations of epicenters and the main active fault systems. Both the 1st and the 2nd events were clearly perceived in the S1, S2 and S3 northern sites (Fig. 3a), with abrupt step-like variations. The 3rd event, the strongest of the sequence, only slightly affects the discharge at S1 and S3.

Further south, Forca Canapine spring and Pescara spring (S6 and S7, Fig. 3b) show a clear increase of discharge after the 1st event, more step-like for S6 but gradual and sustained for S7. These two springs are located at high elevation (1,350 and 850 m a.s.l.) and they both suffered for a sharp decrease after the 3rd event, which completely dry up at S6. Aso River nearby (R2, Fig. 3b), monitored at the spring outlet, is also clearly influenced by the 1st event with a step-like increase and limitedly affected by the following events.

Different effects were registered in the area closest to the 2nd and 3rd events, in the San Chiodo spring area (Fig. 4a) and in the Norcia area (Fig. 4b). In the San Chiodo area (Fig. 4a) a water supply system is operating by a tunnel drainage; periodical operational changes, opening and closing of different drainage tunnels, produce clear variations in the water levels (black vertical dashed lines in Fig. 4a). The responses of two of the 14 available piezometers (W1 and W2), considered representative of the entire monitoring network, and the discharge of the Upper Nera River downstream at Castelsantangelo (R1), are shown. After the 1st event, the system reacted with a sharp step-like increase in the downgradient part of the aquifer (W1, 779 m a.s.l.) and a clear step-like decrease in the upper part of the aquifer (W2, 823 m a.s.l.). The same happened on the 2nd event, while on the 3rd event the water level in W1 increased gradually by nearly 7 m, while in W2 it firstly decreased, then steadily increased for several days up to 6 m in height. After the 3rd event, the discharge of the Upper Nera River (R1) doubled; this quick increase reached a steady state in December 2016. In the Norcia area (Fig. 4b) the Torbidone spring (S5), reactivated after the 3rd shock, shows a gradual increase in discharge up to 1.5 m³/s in about 3 months after the shock. The Sordo River (R3), receiving the Torbidone discharge, also reacted with a clear gradual and sustained increase lasting several months, due also to a different direct groundwater inflow. In addition, the water table of the porous clastic aquifer of the Norcia Plain has shown a hydraulic head increase, reaching +15 m at the contact with the carbonate aquifer, with respect to the water table recorded in 2010–2011.

The western-located S4, S8 and W3 (Fig. 4c) registered abrupt positive step-like increases for the 1st, 2nd and 3rd events, with some differences among them. The Nera River at Torre Orsina (Fig. 4d) receives the entire inflow of the aforementioned flow systems, and others not described here due to the absence of recordings; however, its discharge is clearly influenced by the 1st, 2nd and 3rd events, as shown by the abrupt step-like increases coincident with the first three red lines. Although no significant precipitation is recorded, the river discharge did not decrease for the 3 months following the 3rd event. Overall, the Nera River suffered for a total discharge increase of about 9 m³/s considering the first three main events, corresponding to about 30% surplus of its natural baseflow.

The 4th event, which is the southernmost, was not perceived at all so far to the north, but locally affected the southern monitoring points (Fig. 5), located very close to its epicenter (Fig. 2). All four springs (Fig. 5a) show significant abrupt step-like increase of the discharge after the 1st and the 3rd event, while the 2nd event does not at all modify the hydrographs. The 4th event, in spite of being so close, only slightly influences the discharge of the sole S9 station.

Similar evidence has been recorded at a horizontal borehole (W4, Fig. 6) located in the underground laboratories of the INFN in the Gran Sasso massif (Petitta and Tallini 2002; Amoruso et al. 2013). The time plot of the pressure head variation shows sudden increases in the hydraulic pressure (MPa) with a sharp rise of about 2 m recorded on the 1st and on the 3rd events, while no evidence was recorded after the 2nd and 4th events (Fig. 6).

Further south, changes in the water table were clearly recorded in the monitoring well at Bussi sul Tirino (W5, Barberio et al. 2016; Fig. 5b). The magnitude of the water level variation is about 20 cm for the first three events, with a gradual sustained type variation, while an abrupt step-like increase up to 90 cm is observed for the closest 4th event. The river monitoring sites R5 and R6 do not show any significant variation of the hydrometric level in the 1st, 2nd and 3rd events (Fig. 5b). However, after the 4th earthquake, both the Aterno River (R6) and, most clearly, the Tirino River (R5) hydrometric level responses show a sharp and sudden increase of the hydrometric levels, which drop rather quickly to nearly the prior level than the earthquake discharge.

The hydrogeological changes caused by the 2016–2017 seismic sequence are of remarkable intensity specially if compared to the relatively limited magnitude of the events; similar or larger hydrological responses are very rare (Mohr et al. 2015). The estimated amount of extra discharge drained by springs and rivers since August 24th 2016 for the following 6 months exceeds 0.1 km³. This amount has been obtained looking at the discharge of the entire Nera Basin (R4, Nera at Torre Orsina) before and after the seismic sequence: the additional discharge was about 1.5 m³/s between the 1st and 3rd event (about 8 × 10⁶ m³), and about of 9 m³/s after the 3rd event, which until the end of February 2017 correspond to more than 0.095 km³. This estimation does not consider other changes observed in other basins, which released a minor amount of discharge.

Other documented examples of earthquake induced groundwater release include (ordered by decreasing earthquake magnitude): the Maule Mw 8.8 earthquake in Chile (1.1 km³, Mohr et al. 2016), the Chi-Chi Mw 7.5 earthquake in Taiwan (0.7 km³, Wang et al. 2004b), and the Hebgen Lake (Mw 7.3, 0.5 km³), Borah Peak (Mw 6.9, 0.3 km³; Muir-Wood and King 1993,; USGS 2017), and Loma Prieta earthquakes in the USA (Mw 6.9, 0.01 km³; Rojstaczer et al. 1995); however, these are earthquakes larger than the central Italy events. At a comparable magnitude, the Mw 6.0 South Napa earthquake (USA) produced extra water of about 0.001 km³ (Wang and Manga 2015).

The extent of the area affected by hydrogeological changes approaches 10,000 km². The recorded hydrological variations fall in the known fields of abrupt and sustained water level changes in groundwater due to earthquakes (Fig. 7, modified from Wang and Chia 2008). Nevertheless, peculiar responses in the study area have been observed in sites very close to the epicenters (<10 km), where abrupt changes have been frequently followed by a sustained increase with time, especially after the 3rd event (October 30th), having the highest magnitude. This behaviour could be due to the fractured nature of the aquifer, where pressure changes can easily and quickly propagate along the effective porosity network, reaching the boundaries of the aquifers, causing change in hydrodynamic independently from seismic-induced stresses (Amoruso et al. 2011). Furthermore, the occurrence of several subsequent events repeatedly impacted the aquifers, as happened on October 30th 2016 for the third time in 2 months. Similar events have been not reported in the literature so far.

In addition, the characteristics of the fractured aquifers impacted by the shocks may also have influenced the entity of the response. Differently from the Latium-Abruzzi carbonate aquifers, where a basal aquifer is usually governing groundwater flow, in the Umbria-Marchean aquifers, a network of interconnected faults plays a key role in determining the dynamic groundwater divide location, seepage velocity and extent of the recharge area. It follows that groundwater flow can be easily influenced by seismic events; furthermore, during the 2016–2017 sequence, the study area suffered from many earthquakes whose epicenters were differently located along the central Apennine chain; therefore, numerous aquifers and springs were hit by repeated events, or by a single group of seismic events only.

Generally, a decreasing intensity of the recorded effects was observed moving from the earthquake epicenters towards the distal areas. Figure 8 shows the spring discharge/water level variation, as a function of the distance between the monitored points and the epicenters of each seismic event. The result is a cloud of values from about 10—200% in the near-field (less than 10 km from epicenters), clearly decreasing with distance. Up to +50% in discharge and +2 m water level increase have been recorded between 20 and 30 km away, while minor changes have been observed for sites located between 60 and 100 km from the epicenters. Locally, discharge decreases have been encountered at less than 10-km distance from the epicentral area, but only for the October 30th event, which was the more energetic one.

The mechanism causing both sustained negative and positive hydraulic changes is frequently related to static stress modifications evidenced by comparison between pre-seismic and post-seismic conditions (Jonsson et al. 2003; Montgomery and Manga 2003; Parvin et al. 2014; Mohr et al. 2016). The difference between transient dynamic oscillations (Cooper et al. 1965) and the offset-type water-level changes affecting, in a more permanent way, groundwater flow, is also well described in Yan et al. (2014). Whereas the mechanism for explaining the first type of effects is substantially accepted as being transient oscillations due to crustal dynamic poro-elastic deformation in an aquifer during the passage of seismic waves (Rexin et al. 1962; Kitagawa et al. 2006; Yan et al. 2014), the cause of permanent offsets, corresponding to sustained changes, is still debated.

In this study case, the distribution of changes observed after each main event, and their positive or negative effects, are synthetized in Fig. 9. The impact on most of the springs is the increase of discharge with varying magnitude. Most of the recordings in wells and in rivers clearly indicated a sudden and sharp increase simultaneous with the earthquakes, generally followed by a steady increase lasting for a few days after the shock, and a subsequent smooth decrease. Few points show a decrease of water level or discharge: the disappearance of one tapped spring (S6) and a significant discharge decrease of the spring S7, both located at high elevation (Fig. 3b). Similar decreases or drying-up have been observed in other minor not-monitored springs located in the same recharge area. Another monitored site experiencing post-earthquake decrease in the water table is in Upper Nera Valley, where only the highest-elevation monitoring well (W2, Fig. 5a) suffered from a sharp lowering, quickly balanced in the following days.

The decrease of discharge or water levels and the disappearance of springs have been also been associated in literature to proximity to the active faults. With respect to the water-level changes observed in the footwall area of the fault activated by the Chi-Chi earthquake, Chia et al. (2001) report that water-level rise was the predominant effect in most of the area, whereas water-level fall prevailed in a narrow zone adjacent to the fault trace. Amoruso et al. (2011) report that after the L’Aquila Mw 6.3 earthquake, the two highest springs, located on the trace of the activated fault, suddenly dried up after the main shock. The opposite phenomenon, the reactivation of dry springs or stream, as observed in this case for the Torbidone spring (S5), is also well known. After the 2014 Mw 6.0 South Napa earthquake, many streams and springs, which were dry or nearly dry, started to flow after the earthquake (Wang and Manga 2015). The possible explanation given by the mentioned authors is the enhanced permeability in the recharge areas.

Generally, after the co-seismic peak, discharge and water levels remained on higher values with respect to pre-seismic conditions. A similar mechanism was observed after the 1980 Irpinia earthquake at Caposele spring (Esposito et al. 2001) and, more recently, in the Abruzzi region after the L’Aquila 2009 earthquake (Adinolfi Falcone et al. 2012). This last case has been explained by a double effect: (1) pore pressure propagation due to dynamic stresses caused by the seismic waves, which determined the sudden peak, and (2) an increase of bulk hydraulic conductivity of the fractured aquifer due to fracture cleaning triggered by the pore pressure propagation, which induces mobilization by shaking fine particles that block fracture throats (Amoruso et al. 2011).

Recorded hydrological changes for the 2016–2017 earthquakes may be preliminarily attributed to the pore pressure propagation in the aquifers, which would be followed by a sustained discharge increase attributable to fracture cleaning, mobilizing fine particles from fractures (Brodsky et al. 2003; Wang and Manga 2010; Adinolfi Falcone et al. 2012), as reflected by turbidity increase, clearly recorded in several monitoring points (e.g. R1, S5, S11, S12). In this case, the superimposition of the post-seismic changes on a recession phase makes the post-earthquake evolution clearer. The dynamic stress due to pore pressure propagation has been clearly observed at the high-frequency-monitoring site W4 (Fig. 10), where the short-term (30 s) changes in hydraulic head are the response to the seismic wave of August 24th, whose effects on groundwater levels ended after a few minutes.

In this study, the sustained response of several monitored sites to the subsequent seismic events, considering the fractured nature of the struck aquifers, support the hypothesis of fracture cleaning and the consequent increase of the bulk hydraulic conductivity. In many cases, discharge increase has reached its peak a few days after one of the main events, followed by a decreasing trend due to the recession phase. Accordingly, in several cases, mainly located far from epicenters, the next earthquakes did not cause permanent changes, but only a temporary discharge increase, favoring the model of fracture cleaning. The location of negative effects at high-elevation sites confirms this hypothesis, as water-table decrease is expected in recharge zones of struck aquifers, while increase in discharge is common in low-elevation zones.

A more complex response has been recorded in the core area, i.e. between the epicenters of August 24th and October 30th events. In the Upper Nera River Valley (R1, W1 and W2), in the Norcia Plain (S5, R3) and partially in the eastern side of the Sibillini Mts. (S7, R2), the effects of both events is still evident in the following months. The discharge of springs and water-table elevations remained very high with respect to pre-earthquake conditions, as testified by the Nera main gauging station (R4), not simply justifiable by seismic stresses. The new conditions of groundwater flow are testified by the observed changes in the recession curve of San Chiodo spring (R1, W1, Fig. 5a): previous values of α = 0.003 calculated for the 2011–2015 period, strongly decrease to α = 0.001 after the earthquakes, testifying to a continuous additional contribution from the aquifer with time.

Consequently, other possible additional factors may have influenced the post-seismic response of the area of Sibillini Mts. The impressive increase of spring and river discharge observed in the Upper Nera River valley and Norcia plain in the mid-term responses, may be correlated with the subsidence induced by the toe of Vettore Mt. faulting, which might have created in the core of the Sibillini Mts. aquifer an additional “squeezing effect”. Other possible mechanisms, not further investigated in this paper, could be related to shaking of the aquifer, tilting, settlement and uplifting of the seismogenetic fault-bounded structures and the consequent dislocation of permeability thresholds induced by faulting. Further, a possible decrease in aquifer storativity due to fracture-width reduction could be envisaged, which would have directly triggered the additional volume of groundwater released in the months following the shocks, modifying groundwater dynamic divides and groundwater flow directions. This hypothesis, still not verified, should be carefully evaluated in the future, for prevision purposes.

The response of the monitored fractured aquifers to the earthquakes poses questions in terms of future management of groundwater resources of central Apennines. Furthermore, the discharge increase recorded along the Nera River basin has raised flood risks for the urban areas struck by the earthquake, as in Castelsantangelo sul Nera and Norcia. As several of the monitored springs are tapped for drinking purposes, it is necessary to carefully evaluate the consequences and the mid-to-long-term evolution of the spring discharge.

The abrupt and sustained variations of spring discharge and groundwater levels, observed in carbonate fractured aquifers in central Italy by a wide selection of water points during the 2016–2017 central Apennine seismic sequence, cannot be attributed to natural hydrological drivers and have therefore to be related to the earthquakes. The main findings obtained by analyzing data from more than 20 monitoring sites are the following:

Based on the suggested conceptual models, two different evolutions of the groundwater flow could possibly be faced in the near future: