Introduction

The volcanic area of Campi Flegrei (Italy) is a subcircular caldera with a diameter of ~12 km1,2. The last eruption occurred in 1538 CE3 and formed the volcanic edifice of Monte Nuovo4. Today, the Campi Flegrei area is densely inhabited, with about 500,000 people living in this area5. The main city is Pozzuoli, which is located in the central part of the caldera (Fig. 1). Ground deformations are typical of this caldera and, in the past, they have resulted in phases of subsidence alternating with uplift phases6,7,8,9. This phenomenon is called bradyseism and has been studied since the 19th century by the scientists of that time10,11. In the first half of the 19th century, relative measurements of sea level compared to ground level12 demonstrated that the Campi Flegrei caldera was in subsidence. Since then, the first uplift phase began in the 1950s13. Then, two other bradyseism crises occurred in the 1969–1972 period14 and during the 1982–1984 time interval15,16,17,18,19, accompanied by significant uplift (1.7 m and 1.8 m, respectively) and seismicity, which caused significant damage to buildings in the area of maximum uplift20. These bradyseism crises were followed by temporary subsidence21,22. Finally, the last phase of bradyseism began around 2005 and is still ongoing. The current phase is characterized by increasing seismicity (Fig. 1c)23,24,25,26,27 and fumarolic tremor amplitude, which is indicative of hydrothermal activity28,29,30,31,32. Moreover, significant ground deformations (Fig. 1d)19,33,34,35,36,37,38,39,40,41, which produced about 1.2 m uplift in the central sector of the caldera, and increasing fluid emission (Fig. 1e)42,43,44,45,46,47,48,49,50,51 are typical of the current unrest.

Fig. 1: Parameters of the long-term unrest ongoing in the Campi Flegrei caldera.
figure 1

a Shaded relief map of the Campi Flegrei caldera with the edges of the caldera (red dashed line). The Digital Terrain Model (DTM) was obtained from Lidar data of the city of Naples (publicly available at https://sit.cittametropolitana.na.it/downloads.php), using software ESRI ArcGIS 10.4, adopting the method described in ref. 90. The indication of the horizontal displacement of the GNSS stations in the period 2016–2024 (black arrows) and locations of the earthquakes that occurred between 2005 and 2024 (red dots). The blue arrow indicates the horizontal displacement of the ACAE GNSS station, which deviates from the radial trend of the other stations and is located in the area characterized by a geodetic anomaly55. The RITE GNSS station, which is located in the center of the city of Pozzuoli, is indicated with a black square. Ellipses A and B mark the zones with the greatest concentration of epicentres, corresponding to the main seismogenic volumes. b East-West section showing the projection of the hypocentres of the earthquakes recorded between 2005 and 2024. c Temporal evolution of the cumulative earthquake count. d Uplift measured from the vertical component of the GNSS RITE station. e Soil CO2 flux at FLXOV5 station. The FLXOV5 geochemical station is located in the Solfatara crater and is marked by a blue circle on the map. The CSOB seismic station, positioned on the southern edge of the Solfatara, is marked by a blue triangle. The seismograms shown in this article are recorded by this station. Source data are provided as Source Data files.

During the bradyseism crises of the second half of the 20th century and in the current one the deformation pattern has always shown a bell-shaped radial pattern with the maximum uplift in the central area of the caldera (e.g., refs. 52,53,54,55), where the RITE GNSS station is presently located (Fig. 1a). In the 1982–84 crisis, deformation measurements were mainly based on topographic leveling, which provided vertical deformation measurements. For that period the deformation pattern was modeled as a Mogi source15,16 or as a horizontal crack17,18. The deformation pattern of the current unrest, which is measured through the GNSS network, managed by the Osservatorio Vesuviano of the Istituto Nazionale di Geofisica e Vulcanologia (INGV-OV) and through DInSAR measurements, has often been modeled as a horizontal crack (e.g., refs. 36,37,41,55), or a horizontal thermo-poro-elastic zone56.

As a consequence of the current unrest, in December 2012 the Italian Civil Protection decided to move to the yellow alert phase, which is the second of four alert levels (green = basic; yellow = attention; orange = pre-alarm; red = alarm). From 2012 onwards, the unrest continued and accelerated (Fig. 1c, d, e) in terms of increase in the earthquake occurrence rate, earthquake magnitude24,31, ground deformation rate24,25, and variations in geochemical parameters51. In particular, Bevilacqua et al.25 have shown that the increase in the cumulative number of earthquakes (in total or above a given Md threshold) versus the vertical uplift of the caldera is well-fitted with two exponential functions and that the inflection point between these functions falls in the period between 4/2020 and 9/2022. This transition between the two exponential functions, the second having a larger exponent, marks a clear acceleration of the unrest and, thus, of the seismicity in the caldera. Furthermore, analyzing in detail the deformation pattern, Giudicepietro et al.55 discovered a geodetic anomaly that clearly manifested in 2021, with a maximum uplift deficit, described in detail below, of ~9 cm (in October 2023) in the Mt. Olibano area (East of Pozzuoli) compared to the surrounding zones.

The aforementioned recent studies indicate that since 2021 the unrest of the Campi Flegrei caldera has intensified. For this reason, in this work, we analyze the seismicity that occurred in 2021–2024 with the aim of investigating the relationships between the characteristics of the earthquakes, the seismogenic processes, the evolution of the deformation field, and the geochemical variations (CO2 flux) of the Campi Flegrei caldera. In particular, we focus our attention on several seismic sequences characterized by very short inter-event times, of even a few seconds. Within this category, we identify a specific type of sequence, which we call burst-like swarms, following the classic literature of volcanic seismology (i.e. refs. 57,58). This type of sequence can provide clues on the seismogenic processes in relation to the evolution of the current accelerating unrest, which is controlled by intense ground deformation and escalating hydrothermal activity.

Results

In Campi Flegrei the INGV-OV seismic network has recorded ~18,500 earthquakes (max Md = 4.4; average Md = 0.26, minimum Md = −1.6) from March 2005 to May 2024. Not all of these earthquakes are located because some of them are too small to obtain a reliable location (those located are around 10,200). The locations show a distribution that approximately describes an ellipse in the central area of the caldera, however, the majority of earthquakes (about 90%) are located at Pozzuoli and east of Pozzuoli, in an area that includes the Solfatara-Pisciarelli hydrothermal system and the lava domes of Mt. Olibano (see the zone marked as A in Fig. 1a). A minor group of earthquakes (about 2% of the total) is located in the Gulf of Pozzuoli (zone B in Fig. 1a). The cumulative earthquake count highlights the progressive increase in seismicity rate over time (Fig. 1c). We focused on the seismicity in the period from 2021 to 2024 when the current unrest showed an intensification in terms of seismicity and uplift rates. Specifically, we looked at the types of seismic events that show peculiar characteristics.

The typical seismicity of Campi Flegrei is characterized by VT earthquakes. This type of earthquakes features a clear and generally impulsive onset and a broad frequency content. They are associated with brittle failure mechanisms, similar to tectonic earthquakes. The earthquakes in the Campi Flegrei caldera generally occur in swarms (Figs. S1 and S2 of the supplementary material show the seismograms of two typical seismic swarms of the Campi Flegrei). Details on the typical VT swarm seismicity of Campi Flegrei can be found in Giudicepietro et al.30, where the authors study two swarms in relation to the increase in uplift rate that preceded them and in gas emission that accompanied and followed them, and in Tramelli et al.26, where the six main swarms recorded at Campi Flegrei in the period 2000–2020 are analysed. As an example, predating the 2021–2024 period, which is the temporal focus of this study, we consider here the swarm of 6 December 2019 (Fig. S1 in Supplementary Material), with about 30 earthquakes (Md max = 3.1). At the time, it was the most significant swarm since 1984, in terms of maximum magnitude. In this case, the average inter-event time was about 200 s, with a standard deviation of about 370 s, indicating an irregular occurrence of the events within the typical Campi Flegrei swarms. Similarly, the seismic events recorded in the Campi Flegrei caldera from 2021 to 2024 are generally of VT type. Within this VT swarm seismicity, we recognized sequences characterized by a very short time interval between two consecutive events, which we call inter-time, sometimes even less than a few seconds. Specifically, we identified sequences that we call burst-like swarms, following the definition of Hill et al.57. Galleries showcasing examples of these categories of seismic sequences are provided in Figs. S3 (burst-like swarms) and S4 (generic very short inter-time sequences) of Supplementary Material.

The very short inter-time sequences are generally characterized by events of short duration, ranging from some seconds to several tens of seconds, with impulsive onset. In some cases, the inter-event times between these earthquakes appear to be relatively regular. For instance, the sequences shown in Fig. 2 have inter-times of 7.2 s (±4.1 s) for the 21 January 2022 sequence (25 events, max Md = 0.8) (Fig. 2a), 29.6 s (±8.4 s) for the 20 April 2022 sequence (5 events, max Md = 0.7) (Fig. 2b), and 15.5 s (±4.5 s) for the 28 December 2022 sequence (6 events, max Md = 2.7) (Fig. 2c).

Fig. 2: Seismic recordings in the Campi Flegrei caldera.
figure 2

ac Seismograms and spectrograms of examples of generic very short inter-event time sequences. d Seismogram and spectrogram of the earthquake of 27 April 2024, 03:44:56 UTC (Md = 3.9) located in the Gulf of Pozzuoli (group B in Fig. 1a) reported as an example of a Campi Flegrei earthquake not included in a very short inter-event time sequence. eh Seismograms and spectrograms of examples of burst-like swarms. These sequences also include earthquakes with relatively high magnitudes such as the one with Md 4.2 occurred on 27 September 2023 at 01:35 (UTC) (g) and the largest Campi Flegrei earthquake, with Md = 4.4, recorded on 20 May 2024 at 18:10 (UTC) (h). All examples are recorded by the multiparametric seismic station CSOB, component HNZ (vertical component of the accelerometer with a sampling rate of 200 samples per second).

The burst-like swarms (Fig. 2e–h), which we identified among the very short inter-event time sequences, are rapid-fire sequences of VT earthquakes with widely overlapping coda57. The overlap of events is chaotic, does not occur in a regular manner, and shows very short inter-event times that are often not easily recognizable. For this reason, these sequences show a spasmodic appearance, where the typical VT waveforms with impulsive onset and gradual amplitude decay are no longer recognizable, but instead, they become chaotic signals with highly irregular waveforms (Fig. S3 of Supplementary Material). These swarms are typically accompanied by a continuous background signal with the same frequency content as the VT earthquakes (Fig. 2e, f, g, h). Examples of this signal are shown in Fig. 2e, which depicts the continuous background signal recorded during the burst-like swarm on 22 July 2022 (Md max = 0.5), with remarkable amplitude in the second half of the recording, and Fig. 2f, which shows a burst-like swarm recorded on 19 April 2023 (Md max = 1.2), beginning with a continuous signal of increasing amplitude. Burst-like swarms can also be accompanied by larger events, as in the case of the earthquake with the greatest duration magnitude (Md = 4.4) recorded in the Campi Flegrei on 20 May 2024 at 18:10 (UTC), which belongs to this type of sequence (Fig. 2h). The spectrogram helps to distinguish individual events within the burst. For comparison, Fig. 2d shows the seismogram and spectrogram of an earthquake not associated with a very short inter-event time sequence. The selected event is the largest one (Md = 3.9) of the group of events located offshore in the Gulf of Pozzuoli (zone B in Fig. 1a).

To further investigate these seismic sequences and highlight the features that distinguish burst-like swarms within very short inter-time sequences, we selected two examples as representative cases: the 12 October 2023 generic very short inter-time sequence (Md max = 1.7) and the 14 April 2024 burst-like swarm (Md max = 3.7).

Figure 3 shows the comparison between these two sequences. The burst-like swarm includes two earthquakes with Md > 3.0 and exhibits the continuous background signal, which is clearly recognizable in the spectrogram (Fig. 3d).

Fig. 3: Comparison between the very short inter-time sequence of 12 October 2023 and the burst-like swarm of 14 April 2024.
figure 3

a 30-min seismogram of the two sequences. b Location of the events of the generic very short inter-time sequence (blue circles) and the burst-like swarm (semi-transparent red circles). In background a contour plot produced with matplotlib (Python library) using data from SRTM GL1 DEM at 1 arcsec. c Seismogram and spectrogram of the 5-min window of the generic very short inter-time sequence highlighted with a yellow rectangle in (a). d Seismogram and spectrogram of the 5-min window of the burst-like swarm highlighted with a yellow rectangle in (a). Both examples are recorded by the CSOB station, component HNZ.

In the 5-minute signal windows of the two sequences reported in Fig. 3c, d, we counted 38 events in the generic very short inter-time sequence (Fig. 3c), with an average inter-event time of ~8 s. For burst-like sequences, it is not possible to reliably estimate the inter-event times, as the events are spasmodically and systematically overlapped. Therefore, we can only count the distinguishable events within the sequence, identified as peaks of spectral amplitude in the spectrogram, obtaining 36 events that emerge from the continuous signal within a 5-min window. Actually, the continuous background signal characteristic of burst-like sequences makes the concept of inter-event time inapplicable to this type of sequence, except for the major events that are macroscopically distinguishable in the spectrogram. This is the defining characteristic of burst-like sequences: they show a chaotic and irregular overlap of VT earthquakes, with bursts of events so dense that it is generally not possible to discern the inter-event times.

We also compared the locations of the events within these two sequences (Fig. 3b). For each sequence, we selected the locations of the events within the half-hour signal shown in Fig. 3a. For the sequence of 12 October 2023, we obtained 10 event locations with depths between 700 and 900 m and Md ranging from −0.5 to 1.7. For the burst-like swarm sequence of 14 April 2024, we retrieved 37 event locations with depths between 730 and 2870 m and Md values between −0.3 and 3.7. Both sequences are located within the Solfatara-Mt. Olibano area (zone A in Fig. 1a). This comparison shows that the two type of cases share the same source region. However, the events of the generic very short inter-time sequence (blue circles in Fig. 3b) are concentrated at shallow depths beneath Mt. Olibano area, which is characterized at surface by lava domes, whereas the event locations of the burst-like sequence (red circles in Fig. 3b) extend to greater depths and are more clearly distributed beneath the Solfatara crater, with a slight northward shift beneath the Solfatara- Pisciarelli area, which is characterized by the hydrothermal system.

Finally, we selected all the sequences with very short earthquake inter-times that we recognized in the period between January 2021 and May 2024 and compared them with the clearest examples of burst-like swarms, such as those reported in Figs. 2 and 3. In particular, the selected burst-like swarms include those shown in Fig. 2e–h, the April 14, 2024 sequence reported in Fig. 3a, and a sequence recorded on June 2, 2023 (total number of events = 70; depth ranging between 450 and 2870 m and Md between −0.8 and 4.4). We show the distribution of event epicenters belonging to the clearest examples of sequences classified as burst-like swarms (red circles), together with the generic very short earthquake inter-time sequences (light blue circles) in Fig. 4a. In addition, Fig. 4a outlines the area of the geodetic anomaly discovered in Giudicepietro et al.55 (solid line) and the Solfatara-Pisciarelli hydrothermal area (dashed line), based on the Solfatara Diffuse Degassing Structure (Solfatara DDS,59). The two groups of events appear to share the same source region. Specifically, the seismogenic volume is elongated, shallower within the Mt. Olibano geodetic anomaly area, and extends deeper with a slight dip towards the north, coinciding with the Solfatara-Pisciarelli hydrothermal system. Notably, this analysis highlights a greater concentration of burst-like swarms within the Solfatara-Pisciarelli hydrothermal area. Moreover, the distribution of the locations of the generic very short earthquake inter-time sequences also reveals a very shallow seismicity around the area of the geodetic anomaly (Fig. 4a, b) that became evident, and gradually more pronounced, in the period 2021–2024.

Fig. 4: Relationship between the sequences with very short inter-event times, the hydrothermal system of Solfatara-Pisciarelli, and the geodetic anomaly of Mt. Olibano.
figure 4

a Map of the locations of the events of the sequences with very short earthquake inter-time. Red circles: events of burst-like swarms; light blue circles: events of generic very short earthquake inter-time sequences. The dashed line includes the Solfatara-Pisciarelli hydrothermal area. The solid line contains the area of the Mt. Olibano geodetic anomaly. The earthquake locations are provided as Source Data files (Data 1, 2). In background a contour plot produced with matplotlib (Python library) using data from SRTM GL1 DEM at 1 arcsec. b Red scale: residuals from the inversion of DInSAR Sentinel 1 data estimated in Giudicepietro et al.55, highlighting the geodetic anomaly. The residuals are negative and are concentrated in the Mt. Olibano area with a minimum of ~−10 cm relevant to the period January 2021–October 2023. The image of the Solfatara diffuse degassing structure, modified after Cardellini et al.90, is shown on the map. c Residuals between observed and expected vertical displacements of the ACAE GNSS station, in the area of the geodetic anomaly. d Gray circles indicate CO2 flux in the target area of the Solfatara diffuse degassing structure, measured monthly by the INGV-OV staff. The red dots indicate the uplift deficit (expressed in cm), which is the opposite of the residual shown in (c). Source data are provided as Source Data files.

In Giudicepietro et al.55, analyzing data up to October 2023, the residual between the observed and estimated uplift in the Mt. Olibano geodetic anomaly area—calculated using time series data from two pixels of the Sentinel-1 dataset, one representing the anomalous area and the other representing the maximum uplift area—was found to be ~ − 9 cm. This residual, calculated for a single pixel, was also consistent with the residuals obtained from the inversion of the DInSAR measurements to determine the deformation source (see below), which were concentrated around the Mt. Olibano area, with a minimum of about −10 cm. These residuals, represented in Fig. 4b, illustrate the spatial extent of the geodetic anomaly identified in Giudicepietro et al.55. Using a similar method to the one applied to the Sentinel-1 selected pixels in Giudicepietro et al.55, in this study, we calculated the uplift residual for the ACAE GNSS station in the period January 2004 –May 2024 (Fig. 4c), from the comparison with the temporal evolution of the uplift recorded at RITE GNSS station, used as the reference station (see “Methods”). This analysis revealed that by May 2024, the actual uplift residual at ACAE had reached ~−11 cm, with an increase of about 2 cm compared to October 2023 (over ~6 months). It also revealed that the relative subsidence of the Mt. Olibano area began to emerge before 2021, albeit at a slow rate (Fig. 4c).

In the Solfatara-Pisciarelli hydrothermal area, where the majority of the burst-like swarms are located, the diffuse degassing from the Solfatara crater increased significantly in the period 2021–2024, when the sequences with very short inter-event times became evident, reaching a flux of 1557 tons of CO2 per day in 2024 (Fig. 4d), comparable to that of an open-conduit volcano59.

To frame this type of sequences in the general context of the ongoing unrest in the Campi Flegrei, we considered the entire seismicity of the caldera and compared it with the seismicity of zone A (Fig. 1), where these sequences are located, and with the geodetic measurements and geochemical data.

To show the distribution of earthquakes in the caldera, in relation to ground deformation, we selected earthquakes with Md > 1.0 and plotted them on the uplift map obtained with Sentinel 1 data (Fig. 5). We processed the Sentinel 1 data from 2015 to May 2024 according to the approach proposed by Casu and Manconi60 (see the “Methods” section) to obtain the map of the vertical displacement of all the coherent pixels (maximum uplift ≈103 cm). We reported on the map the indication of the deformation sources obtained in Giudicepietro et al.55. The primary source (blue star in Fig. 5) was found to be compatible with a horizontal circular crack whose location parameters are: latitude = 40.817814; longitude = 14.126922; depth = 3826, while a secondary source is found in the area of the geodetic anomaly of Mt. Olibano (magenta star in Fig. 5) (details on the inversion procedure are reported in Giudicepietro et al.55). The spatial distribution of seismicity in the Campi Flegrei caldera reveals that earthquakes occur predominantly along an elliptical ring surrounding the primary ground deformation source (indicated by the blue star in Fig. 5), as well as within the ring itself. This suggests a relationship between the deformation field and the entire seismicity of the caldera. The seismic sequences of interest fall within a sector of this elliptical distribution, referred to in this work as zone A. For this zone, previous studies report normal fault focal mechanisms compatible with an extensional stress regime in this sector of the caldera (e.g.,28,30,61,62). In particular, Giudicepietro et al.30, analyzing the focal mechanisms of the main events from two swarms that occurred in zone A on 7 October 2015 and 6 December 2019, found normal fault mechanisms. Tramelli et al.26 studied the focal mechanisms of six main swarms recorded at Campi Flegrei during the period 2000–2020 and consistently found normal fault mechanisms in zone A. Similarly, Scotto Di Uccio et al.61 once again found only normal fault mechanisms in zone A, analyzing seismicity from 2014 to 2024. Finally, the focal mechanisms reported in the INGV-OV surveillance bulletins (https://www.ov.ingv.it/index.php/monitoraggio-e-infrastrutture/bollettini-tutti/bollett-mensili-cf, last accessed 10 October 2024), which are preliminary, also consistently show normal fault mechanisms in zone A.

Fig. 5: Distribution of earthquakes and ground deformation in the Campi Flegrei caldera.
figure 5

Map with the locations of earthquakes with Md >1.0 (orange circles), which are distributed on an elliptical ring. The size of the circles is proportional to the magnitude, which ranges between 1.0 and 4.4. The vertical displacement of the ground from 2015 to May 2024 obtained from DInSAR data is shown in blue scale on the map. The blue star indicates the location of the primary source of the deformation (depth 3826 ± 45 m). The magenta star indicates the location of the secondary source of the deformation (depth 528 ± 78 m). The parameters of the deformation sources are from Giudicepietro et al.55. The red ellipse indicates zone A, the gray ellipse indicates zone B, and the two blue dashed ellipses delimit the ring where part of the seismicity is distributed. In background a contour plot produced with matplotlib (Python library) using data from SRTM GL1 DEM at 1 arcsec.

In zone B (Fig. 5), however, earthquakes with focal mechanisms consistent with reverse faults have been documented in the INGV-OV surveillance bulletins; however, some of these are less constrained than those in zone A, which is better covered by the seismic network.

The distribution of the entire seismicity of the caldera on an elliptical ring around the location of the center of the primary deformation source and the concentration of earthquake locations in zone A, where the secondary deformation source is located, suggest the overlap of multiple seismogenic processes. In particular, the seismogenic processes in zone A can be linked to the increased activity of the Solfatara hydrothermal system, where, as already mentioned, the CO2 diffuse degassing shows a progressive increase over time and is currently comparable with the average flux of CO2 in the plume of active volcanoes with continuous degassing51, and to the geodetic anomaly of Mt. Olibano, that introduces a further complication in the deformation field of the Campi Flegrei caldera.

To describe the evolution of the geodetic anomaly up to the last months, taking advantage of the high temporal sampling of the GNSS network, we exploited the vertical displacement deficit (uplift deficit) of the ACAE station. We used the RITE GNSS station, which accurately represents the uplift of the central area of the caldera, as a reference to derive the correlation coefficient used to estimate the expected uplift at the ACAE station over the period 2004–2024 (well before the onset of the acceleration in 2021). We obtained the residuals by subtracting the expected uplift at ACAE from the observed uplift data (Fig. 4c). Finally, we defined the opposite of these residuals as the uplift deficit (see “Methods” for details).

This approach enabled us to track the temporal evolution of the uplift deficit at the ACAE station, located within the geodetic anomaly. Consequently, we were able to correlate the uplift deficit in the anomaly area with the cumulative earthquake count, dominated by events occurring in zone A. We found a very high correlation between the uplift deficit and the cumulative number of earthquakes (r = 0.998) (Fig. 6a, b). We also compared the vertical displacement at the RITE station with the cumulative earthquake count, obtaining a slightly lower correlation between the two parameters (r = 0.988) and a non-linear relationship (Fig. 6c), which is consistent with the exponential trend analyzed in Bevilacqua et al.25. Lastly, we correlated the vertical displacement at the ACAE station with the cumulative earthquake count, finding a correlation coefficient r = 0.979 (Fig. 6d). This analysis highlighted a strong correlation between the temporal trend of seismicity, in terms of event frequency, and the time evolution of the geodetic anomaly (Fig. 6a).

Fig. 6: Temporal evolution of seismicity and ground deformation.
figure 6

a Comparison between the uplift deficit of the GNSS ACAE station (cyan line), representative of the geodetic anomaly of Mt. Olibano, and the cumulative earthquake count in the period January 2021–April 2024. The black line indicates the cumulative count of earthquakes recorded in the caldera. The correlation coefficient between earthquakes and ACAE uplift deficit is r = 0.998. b cumulative earthquake count in the period January 2021–April 2024 versus uplift deficit of the GNSS ACAE station. c Comparison between the uplift of the GNSS RITE station (orange line), representative of the general deformation process (see deformation source Fig. 5), and the cumulative earthquake count in the period January 2021–April 2024. The correlation coefficients between all earthquakes and RITE up is r = 0.988. d Comparison between the uplift of the GNSS ACAE station (green line) and the cumulative earthquake count in the period January 2021–April 2024. The correlation coefficients between all earthquakes and ACAE up is r = 0.979. The time series are on a daily basis and are normalized.

Discussion

In this article, we focus on seismic sequences with very short inter-event times, which can be significant indicators of hydrothermal and geodetic processes. These sequences occur in the context of general unrest and overall seismic activity. They are located within a sector of an elliptical distribution of earthquakes, concentrated roughly along the perimeter of the bell-shaped, axisymmetric deformation, at the flexure between the highly deformed central part of the caldera and the outer region experiencing less intense deformation (reasonably due to the concentration of stress in this flexure zone). This seismicity can be interpreted as the brittle response of the crustal rocks above the primary deformation source. The interpretation of this deformation source located at a depth of about 3800 m in the center of the caldera remains non-univocal. It could represent an accumulation zone of magmatic fluids63, the expansion of a porous medium under increased pressure56, or even a magma intrusion64.

The specific segment of the elliptical seismic distribution affected by the seismic sequences we are studying is zone A (Figs. 1 and 5), where the Solfatara-Pisciarelli hydrothermal system is located, and where a geodetic anomaly has been observed since 2021, linked to a secondary deformation source. This sector of the elliptical ring is the most seismically active zone of the caldera and earthquakes occurring in this area dominate the seismic catalog of Campi Flegrei. The very short inter-event time sequences originate within an elongated source volume, which slightly inclines northward. The shallowest part of this volume is beneath Mt. Olibano (the geodetic anomaly area) and deepens towards the Solfatara-Pisciarelli area, adjacent to Mt. Olibano. This pattern is characterized by a generally continuous distribution of hypocenters, with the events belonging to the burst-like swarms more concentrated in the Solfatara-Pisciarelli hydrothermal area. The example sequences shown in Fig. 3 well illustrate this hypocenter distribution well (see also Fig. 4).

The focal mechanisms reported in the previously cited literature (which include some events belonging to burst-like swarms) and in the INGV-OV surveillance bulletins indicate that while in zone A they are always indicative of normal fault activity, in zone B they often appear to be related to reverse fault activity.

This pattern of focal mechanisms may indicate an irregular distribution of stresses and possible deformation asymmetries, masked by the sea in the Gulf of Pozzuoli area. This condition has sometimes been observed in calderas, even in potentially pre-eruptive states, such as at the Sierra Negra volcano, where it has been transiently observed several times over the past decades, associated with caldera uplift, as recently reported in Bell et al.65.

Seismic sequences similar to those with short earthquake inter-times, mainly concentrated in the seismogenic zone A, which are the main focus of this study, have been described in the literature. Burst-like swarms have been observed in other volcanoes such as Mammoth Mountain (United States), a hydrothermally active area that occurs in a lava dome complex along the rim of the Long Valley caldera57. For them, Hill et al.57 hypothesized a process driven by a transient increase in local fluid pressure, based on the similarity of these sequences with signals associated with pressure transients generated by shutdown operations on production wells in Japanese geothermal fields66. Nishi et al.67 recorded sequences of this type on White Island and observed that they were located in the same seismogenic volume as volcano-tectonic earthquakes. Specifically, the authors found that the shallow seismicity occurs within the main volume of the volcano-hydrothermal system that underlies the Main Crater floor. This is similar to what has been observed in the Campi Flegrei caldera. The mechanism that caused them was attributed to brittle failure due to rapid fluid pressure fluctuations. Also, Lin et al.68 recorded burst-like swarms on the Tatun Volcano Group (Taiwan) and interpreted them as the response of an extended network of cracks to the pressure variations due to fluid injection. Moreover, a close association between these signals and hydrothermal activity is also suggested by McCausland et al.69, who consider them possible precursors of phreatic explosions when recorded in association with low-frequency and hybrid events. Finally, Yukutake et al.70 analyze similar signals associated with the 2015 phreatic eruption of the Hakone volcano, in Japan. This leads us to interpret the burst-like swarms of the Campi Flegrei as the brittle response to the increase in hydrothermal system fluid pressure. This interpretation is compatible with the appearance of the continuous background signal, which may be generated by a mechanism involving near-continuous brittle failure. It is also consistent with the model proposed by Fournier71 for the movement of fluid from a shallow magmatic zone to a hydrothermal environment. The increase in the hydrothermal system fluid pressure in Campi Flegrei is documented in several studies (e.g.,54). Furthermore, we have evidence that this variation in hydrothermal fluid pressure occurs in extensional stress regime. Indeed, the burst-like swarms share the same source region as other earthquakes occurring in zone A, which have focal mechanisms consistent with a local extensional stress regime26,30,61. Additionally, the burst-like swarm locations are close to the Solfatara-Pisciarelli hydrothermal zone and Mt. Olibano area, where a geodetic anomaly is located, which is further evidence of the local extensional stress regime72,73,74. In fact, as explained in Giudicepietro et al.55, the extensional stress regime in the upper crustal layers is induced by the inflation of the deformation source in the central part of the caldera72,73,75. This extensional stress regime interacts with the hydrothermal system76,77, which has shown significant increases in pressure and temperature in recent years51,78. The potential effect of an intensifying extensional stress regime on an increasingly pressurized hydrothermal system could be an increase in rock permeability, further promoting gas emissions and rapid circulation of hot fluids in the upper crustal layers. Conceptually, these conditions may predispose the system to phreatic activity79,80,81, and the occurrence of burst-like sequences could be an indicator of this possibility in the Solfatara-Pisciarelli area. However, this is a speculative interpretation, and further research will be needed to confirm this hypothesis.

The extensional stress regime also interacts with the heterogeneity of the Mt. Olibano area, which is characterized by strong lithological contrasts between lava bodies and pyroclastic rocks, and by faults that generate the observed seismicity61. This interaction contributes to the non-uniform behavior of the Mt. Olibano area, which responds inelastically under the action of the overall inflation of the Campi Flegrei caldera, leading to the localized relative subsidence documented in Giudicepietro et al.55. Notably, this relative subsidence of the Mt. Olibano area is not an aseismic process, as the uplift deficit in this zone shows a strong correlation with the cumulative earthquake count (Fig. 6), which is largely dominated by the seismicity of zone A.

The seismicity associated with the uplift deficit of Mt. Olibano also includes very short inter-event time sequences. This type of sequences, located near Mt. Olibano, appears to be more directly connected to the tectonic evolution of this area, which is more affected by the geodetic anomaly effects. This interpretation agrees with the recent study on the spatial distribution of the b value27. The authors find a b-value close to 1 for the earthquakes occurring below Mt. Olibano in contrast to the higher values estimated for the volume below Solfatara and Pisciarelli. This value can be related to the mainly tectonic origin of the earthquakes occurring in this area. The b-value is, indeed, close to 1 for tectonic areas and greater for volcanic ones, as found by the same authors for the closer areas of Sofatara and Pisciarelli. This suggests differing physical characteristics of the medium in these two adjacent areas of the caldera.

In general, in the Campi Flegrei current unrest, the spatial distribution of earthquakes and the characteristics of the seismic sequences suggest the strong interconnection of the geodetic, hydrothermal, and seismic phenomena, that are controlled by the primary deformation source located at the center of the caldera. When the uplift of the center of the caldera (e.g., in correspondence of the RITE GNSS station) accelerates, the uplift deficit of the geodetic anomaly (e.g., in correspondence with the ACAE GNSS station) and the cumulative earthquake count also accelerate, as also highlighted in Giudicepietro et al.55 and here confirmed with data updated to May 2024. However, previous studies24,25 have shown that seismicity, as represented by the cumulative earthquake count (total or above a given magnitude), depends exponentially, with increasing exponent in time, on GNSS data, such as the vertical component of the RITE station, which shows the greatest uplift among the GNSS stations. This difference in the temporal evolution of the two parameters is also observed in the period 2021–2024 (Fig. 6c, d). On the other hand, in this study we computed the daily time series of the uplift deficit in the Mt Olibano area, identified as a geodetic anomaly55, exploiting the high acquisition rate of data from the INGV-OV GNSS network. It is noteworthy that this uplift deficit, due to the different uplift rate of the Mt. Olibano zone compared to the surrounding areas, evolves similarly to seismicity, and shows a linear relationship with the cumulative earthquake count (Fig. 6a, b). This linear relationship suggests that the uplift deficit is a good indicator of the inelastic component of the behavior of this sector of the caldera in response to the overall deformation process. Therefore, we interpret this finding as the effect of the lithological and mechanical discontinuity of the crustal rock, which in this area begins to deviate from elastic behavior when subjected to deformation. This condition, combined with the aforementioned potential for rapid ascent of hot fluids and the occurrence of burst-like swarms in the surrounding area, suggests that this region can be susceptible to phreatic activity.

Methods

The seismic network of the Campi Flegrei is developed and managed by the INGV-OV. It includes 27 stations distributed within the rim of the caldera. Moreover, four stations were installed in the Gulf of Pozzuoli82. Most of the stations are equipped with three-component broadband seismometers, with a sampling rate of 100 samples per second, and with accelerometers, whose signals are acquired at 200 samples per second. The data is transmitted in real-time to the INGV-OV acquisition center. The figures showing seismograms and spectrograms use data from the CSOB station, which is in an optimal position to record the seismic sequences that are the object of this work (Fig. S5 in Supplementary Material). In 2020, the location sensibility in space of the Campi Flegrei seismic network was characterized by a magnitude threshold ranging between 0 and 0.523. In the Solfatara-Pisciarelli area, where a very local seismicity has become increasingly evident since 2010, the seismic network is denser and allows us to locate earthquakes with magnitude (Md) < 0, when the seismic noise is particularly low. To describe the general trend of the current long-term unrest, we used the locations of the earthquakes from 2005 to 2024, and the seismic catalogue of the Campi Flegrei from 2004 to 2024 (Fig. 1a, b, c). The locations and the seismic catalogue are extracted from the INGV-OV seismological database (https://terremoti.ov.ingv.it/gossip/flegrei/index.html, last accessed 18 October 2024). The catalogue used in this article is also provided as a Source Data file (Data 3). The velocity model used by INGV-OV is reported in Calò and Tramelli83.

Since the very short inter-event time seismic sequences share the same source region as the other A-zone earthquakes, shown in Fig. 1, we also created a specific A-zone earthquake catalog. This catalog includes only earthquakes with first arrival recorded by the seismic stations located in the A-zone (see the seismic station map in Fig. S5 and the caption of Fig. S6 of Supplementary Materials). In selecting the A-zone earthquakes, we considered both localized and non-localized events, including those recorded by only one or two stations, which are typically very small. Thus, we generated plots similar to those in Fig. 6, both with the cumulative count of all earthquakes recorded in the entire caldera and with the cumulative count of the A-zone earthquake catalog. The two counts show negligible differences in their temporal evolution (see Fig. S6 of Supplementary Materials), therefore, we no longer took into account the A-zone earthquake catalog, and in Fig. 6, we only showed the cumulative count of all earthquakes in the caldera (which is dominated by the seismicity of zone A).

The selection of the burst-like swarms and generic very short inter-event time sequences was conducted through visual analysis of the seismograms. The spectrogram analysis and time series correlations were performed using ObsPy84, a system for seismic data analysis, along with other Python-based packages. Additionally, since the ground deformation taking place at Campi Flegrei is significant, and is closely correlated with the seismicity, we also used data from the INGV-OV GNSS network and DInSAR Sentinel-1 data.

Data from the continuous GNSS network of Campi Flegrei were processed using the Bernese GNSS software on a daily basis with the IGS final and reprocessing products to obtain homogeneous results. To remove the regional tectonic background from the volcanic deformation pattern, the time series and velocity fields were transformed into a local reference frame including six stations from the INGV RING network, located outside the Neapolitan volcanic area. No correction was applied to the vertical component, as the tectonic contribution was considered negligible. Furthermore, seasonal signals were removed (for more details on the data processing of the Campi Flegrei GNSS network see ref. 85. To obtain the time series of the vertical residuals of the ACAE GNSS station (Fig. 4), we applied the geodetic anomaly identification method from Giudicepietro et al.55, based on the DInSAR measurements, to the GNSS data. Considering that the RITE station is the closest to the area of maximum uplift of the radially symmetrical bell-shaped deformation pattern of Campi Flegrei, and the ACAE station is located in Mt Olibano, we used these two stations to derive the uplift deficit in the area characterized by the geodetic anomaly. Specifically, we: (1) assumed RITE as a reference for the uplift trend in the central area of the caldera; (2) calculated a proportionality coefficient (α) between the RITE and ACAE uplift time series using data from 2015–2018, when the ACAE area was not yet affected by the geodetic anomaly; (3) estimated the expected ACAE uplift time series by multiplying the RITE uplift time series by α; (4) subtracted the expected ACAE uplift time series from the actual ACAE uplift time series, obtaining the residual series from 2004 to 2024 as shown in Fig. 4c. Then, we calculated the opposite of the residuals to obtain the time series of the uplift deficit shown in Fig. 6a. The RITE up and ACAE up time series are provided as a Source Data file (Data 4).

The DInSAR measurements were generated by applying the Parallel Small BAseline Subset (P-SBAS) approach86,87. In particular, we separately processed 430 ascending (Track 44) and 429 descending (Track 22) orbits data acquired by Sentinel-1 constellation from 24 March 2015 to 24 May 2024 to obtain the Line of Sight (LOS) displacement time series for each coherent pixel. Then, the vertical displacement component was calculated by combining the LOS P-SBAS time series using the method described in Casu and Manconi60. The mean velocity of the vertical component computed between 3 January 2021 and 24 May 2024 period represented as DInSAR fringes (every color cycle corresponds to a velocity variation of 4 cm/year) is shown in Fig. 7a. It clearly highlights that the uplift deficit in the Mt. Olibano area, starting from 2021, makes the fringes to deviate from a circumferential pattern, which instead is preserved in the whole caldera. The vertical component of the Sentinel-1 data, in the March 2015–May 2024 time interval, is also shown in blue scale in Fig. 5.

Fig. 7: Pattern of vertical deformation in the Campi Flegrei caldera during 2021–2024.
figure 7

a Mean vertical velocity computed for the 3 January 2021–24 May 2024 (every color cycle represents a velocity variation of 4 cm/year). Inset shows a closed look at the velocity patterns in the Solfatara—Mt. Olibano area. The map was produced with IDL8.8 (NV5 Geospatial Software). The background (in grayscale) was obtained by calculating a shaded relief starting from SRTM GL1 DEM at 1 arcsec. b Difference between the actual and the expected displacement time series for the vertical component for a pixel, identified by the white square in (a), representative of the deformation behavior in the Mt. Olibano area. Dashed red lines correspond to 2σ (standard deviation) error bar computed before 2021.

Finally, we used geochemical data from the INGV-OV geochemical monitoring network. In particular, we showed data from a single geochemical station for automatic soil CO2 flux measurement (Fig. 1e), and data of the diffuse CO2 emission of the entire Solfatara degassing structure (Fig. 4d). The former dataset consists of continuous recordings from the FLXOV5 station, which is located in the Solfatara crater. The latter dataset consists of the CO2 emission (in t/d) from an area of ~90,000 m2 (“target area” in Fig. 8) located inside the Solfatara crater. The CO2 emission is monthly computed through the measurement of 63 fixed points with the accumulation chamber method88. This monitoring activity started in 2004 and is still ongoing. The data show a marked increase in the CO2 emission in the last years (Fig. 4d). Further details of the technique measurement and data treatment are reported in Chiodini et al.51 and Cardellini et al.89. Source data are provided as Source Data files (Data 5 and 6).

Fig. 8: Solfatara diffuse degassing structure (DDS) and locations of the target area and of the 63 points that are monthly measured.
figure 8

Source data are from Chiodini et al.51. The map shows the probability that the CO2 flux is >50 g m−2 d−1, a suitable threshold for a pure biogenic CO2 flux. Above this threshold, the CO2 emission is at least partially fed by the hydrothermal-magmatic source. The map synthesizes 30 measurement campaigns, each of 400–500 measurement points performed in the 1998–2016 period (see Cardellini et al.89 for further details). The measuring points inside the target area are measured monthly to monitor the CO2 emission in the time (Fig. 4d; see Chiodini et al.51 for further details).