Tidal and environmental modulation of seismic noise at Mefite d’Ansanto (Italy) non-volcanic CO2 emission field

The Mefite d’Ansanto (Southern Apennines, Italy) is one of the largest non-volcanic CO₂ emission areas worldwide, representing a unique natural laboratory for investigating the interactions between geofluid circulation and external forcing mechanisms. In this study, we analyzed continuous seismic noise recorded by three stations deployed around the vent area, with the aim of exploring the influence of external drivers on noise parameters. We applied a combination of multiple linear regression (MLR) and Singular Spectrum Analysis (SSA) to seismic noise attributes, specifically RMS amplitude and polarization (Rlen), in two frequency bands associated with the activity of the emission vents. Our results indicate that environmental variables (mainly wind, solar radiation, air temperature, and soil moisture) and tidal strain contribute to signal modulations (20%–30% of the variance) on semidiurnal to monthly time scales. We propose that meteorological and tidal processes, by inducing variations in the physical properties of the shallow crust (permeability, pore pressure, stress distribution), can affect both near-surface seismic noise sources as well as the dynamics of the deeper degassing system. These findings highlight the potential of seismic noise as monitoring tool for tracking external forcing mechanisms acting on geofluid systems.

1 Introduction

CO₂ emissions driven by crustal processes as well as deeply sourced mantle gas, are influenced by a combination of internal geological processes and external environmental forces. Among these, tidal, hydrological, and atmospheric modulations have garnered increasing scientific attention due to their potential to regulate the migration and release of CO₂ along fault systems and in volcanic fumarolic fields. The role of external forcing in modulating CO₂ degassing in volcanic processes has been extensively investigated by many researchers. Soil gas emissions, including those of CO₂, are significantly influenced by environmental parameters such as barometric pressure, air temperature and humidity, rainfall, wind speed, wind direction, and soil water content (Rogie et al., 2001; Lewicki et al., 2007; Carapezza et al., 2003; Lewicki and Hilley, 2014; Boudoire et al., 2017). Semidiurnal, diurnal and seasonal cycles in CO₂ degassing have been observed worldwide in volcanic areas (see e.g., Viveiros and Silva, 2024). Viveiros et al. (2008), Viveiros et al. (2014) examined the impact of environmental/meteorological factors on soil gas emissions at volcanic sites (Fogos and Furnas) in the Azores. Similarly as Morita et al. (2019) at Aso volcano (Japan) for soil CO₂ flux, they confirmed the aforementioned observations at both short and seasonal time-scales. Rinaldi et al. (2012) demonstrated that at Furnas volcano atmospheric pressure variations were shown to enhance surface degassing, underscoring the role of exogenous factors in shaping gas emission patterns.

In addition to the atmospheric and hydrological factors, tidal forces induced by gravitational interactions between the Earth, Moon and Sun, can affect subsurface stress fields, altering the permeability of fracture networks and modulating gas flux (López et al., 2006; Viveiros et al., 2014). In a nutshell, the exogenous sources can control not only near-surface processes but also deeper sourced phenomena, through complex coupling mechanisms (Rinaldi et al., 2012; Lewicki and Hilley, 2014).

Cyclic patterns in CO₂ degassing have been observed not only in volcanic areas, but also in tectonic contexts, where deep crustal or mantle gases migrate through fault systems. As an example, diurnal variations related to the Earth’s tides characterize the soil CO₂ flux and concentration along seismically active faults in the West Bohemia region (Weinlich et al., 2006; Faber et al., 2009). The authors advocate that the main causes of these cycles are likely the periodic compression and extension of faults induced by tidal stresses, which contribute to altering their permeability and regulating gas migration. Tidal modulation of gas emissions has also been documented out of the seabed off SW Taiwan, where CO₂ and methane fluxes, as well as associated tremors, are strongly correlated with ocean tides, especially diurnal and semidiurnal constituents (Hsu et al., 2013). The authors suggest that the tidal response is related to the variation of the pore-fluid pressure due to the gas charging from a deeper source to the shallow sediments. Yang and Zhang (2021) revealed a significant correlation between crustal tilt magnitude and CO₂ leakage along Bayanhot (Inner Mongolia) active fault zone. They proposed that fault architecture and permeability changes, driven by tectonic or tidal stresses, enhance fracture opening and facilitate gas release. A recent study by Zuddas and Lopes (2023) highlights the relationship between lunar declination and CO₂ degassing from the natural aquifers in the central Apennines, suggesting a lunar periodic forcing of the aquifer chemical composition resulting in the tidal control of gas emissions.

These studies indicate that tectonic CO₂ degassing provides valuable insights into the permeability variations and stress state of active faults, as well as the influence of external forcing mechanisms. Nevertheless, the role of external forcing on non-volcanic CO₂ emissions still remains underexplored, particularly in tectonically active regions where gas migration is closely tied to fault dynamics. The present study focuses on the Mefite d’Ansanto, a prominent CO₂ degassing site in the southern Apennines of Italy (Figure 1). Known for its exceptional flux of mantle-derived CO₂, Mefite provides a unique natural laboratory to investigate the interplay between tectonic processes and external modulation of gas emissions. The Mefite d’Ansanto, situated between the Sannio and Irpinia seismogenic regions, emits approximately 2000 tons of CO₂ daily, ranking it among the largest non-volcanic degassing sites globally (Chiodini et al., 2010; Giustini and Brilli, 2020). This area is characterized by a high density of active faults, some of which are associated with significant seismic events, such as the Ms 6.9 Irpinia earthquake in 1980 (Improta et al., 2003). The CO₂ emissions originate from the lithospheric mantle, with gas migrating through deep-seated faults and fractures, and trapped in a shallower pocket (at a depth of 1,100–1,600 m), before being released at the surface over an area of about 4.000 m² (Chiodini et al., 2010; Improta et al., 2014). The primary emission site, known as Mefite Lake (or Gray Lake, Figure 1), consists of a bubbling mud pool surrounded by vents that release a lethal gas flow. The intersection of local minor transfer faults with the regional NW–SE normal fault system is likely to cause an increased crustal weakness that favors the upward migration of gases (Pischiutta et al., 2013). The Mefite d’Ansanto area hosts two main groundwater circuits: a deep one, reaching depths of up to about 100 m below the surface, and a shallow one, only a few meters deep, which includes very shallow, lens-shaped aquifers (Ortolani et al., 1981; Martinelli et al., 2021). Further recent studies have shed light on the seismic characteristics of the Mefite area. Among these, Morabito et al. (2023) conducted a passive seismic experiment, revealing stable wavefield properties over time and identifying distinct frequency bands: high-frequency signals (>5 Hz) were found to be closely linked to the gas emissions, while lower frequencies (<5 Hz) exhibited contributions from endogenous and exogenous (e.g., anthropogenic and meteorological) sources. By correlating seismic noise characteristics with environmental variables, these authors show that external forcing, such as atmospheric pressure and temperature fluctuations, significantly influences gas emission dynamics. The seismic wavefield associated with the Mefite gas emissions was further investigated by La Rocca et al. (2023), who attributed frequencies below 3 Hz mainly to a stationary source, whereas higher frequencies (up to 20 Hz) to a non-stationary one, causing significant variations in the signal amplitude. In both bands, surface waves generated by very shallow sources are dominant. These findings are complemented by the work of Panebianco et al. (2024), who applied machine learning techniques to seismic signals, advancing the automated detection and classification of seismic tremors associated with non-volcanic CO₂ degassing.

Figure 1

Figure 1. Mefite d’Ansanto site. Location of AME seismic array (squares and triangles, which indicate the stations used in the present study). In the top right inset the position of the Italian study area is marked by a red square. The red arrow indicates the position of Mefite Lake, where the main emission vents insist. Base map: Google Earth, Image©2025 Airbus.

Morabito et al. (2024) developed a comprehensive model for the hydrothermal tremor source at Mefite d’Ansanto. Their study characterized the seismic wavefield associated with the degassing vents in the intermediate frequency band (1–15 Hz), which was found to be composed of a stationary background component (2–5 Hz), linked to the shallower bubbling activity (few meter depth below the surface) of the hydrothermal system, and an intermittent one of higher energy (10–15 Hz), likely generated by the passage of overpressured gas within the upper 15–25 m. The model proposed by Morabito et al. (2024) suggests that the tremor source is linked to the dynamic interaction between ascending CO₂-rich fluids and the surrounding rock matrix, modulated by the shallow groundwater circulation, which in turn is strongly influenced by meteoric waters. Thus, this model provides valuable insights into subsurface processes involving the coupling between gas flux dynamics and environmental conditions.

Building on this growing body of knowledge, the present study aims to explore the periodicities in seismic signals recorded at Mefite and their significance in terms of potential modulation by tidal forces, hydrological and meteorological factors. Using multiple linear regression (MLR) and Singular Spectrum Analysis (SSA) to decompose signals and identify cycles, we investigate the mechanisms underlying the interplay between external forcing and gas emission processes.

2 Materials and methods

2.1 Data

The dataset analyzed in this study comprises continuous seismic recordings acquired at the Mefite d’Ansanto site, as described in Morabito et al. (2023). These data were collected by a temporary seismic array (AME) consisting of 7 three-component stations equipped with short-period 1 Hz Lennartz 3D-Lite sensors (Figure 1). The sampling rate for all stations was set at 100 Hz. Seismometers were buried in the ground, at a depth of about 50 cm. The deployment spanned approximately 4 months (10 June-27 September 2021), capturing a diverse range of seismic signals under varying environmental and atmospheric conditions. All instruments were synchronized via GPS time signals to maintain precise timing across the array.

For the present analysis, we selected the stations AME4, AME2 and AME0, which have the longest time records and a small number of signal gaps due to sensor breakdowns, ensuring reliability for robust statistical analysis. The three stations are characterized by increasing distance from the emission vents, with AME4 being the closest one. The seismic signal (Supplementary Figures S1–S3) associated with the CO₂ degassing was extracted from the raw data by filtering the waveforms in the 2–5Hz and 10–15 Hz frequency bands, corresponding to the background and intermittent wavefield generated by hydrothermal activity, according to Morabito et al. (2024). Then, we computed the hourly RMS amplitude (averaged over the three directions of motion) and the spread of the azimuthal distribution of the polarization vector, expressed in terms of the hourly-averaged resultant vector length (Rlen). This statistical parameter is used to discriminate directional from non-directional motion, and it has been shown to be sensitive to the presence of fluids in the medium, which induce depolarization and a corresponding Rlen decrease (Pischiutta et al., 2022; Petrosino and De Siena, 2021). The resulting seismic time series are shown in Figure 2 (for more details see Morabito et al., 2024 and references therein).

Figure 2

Figure 2. Time series of RMS and Rlen at the three stations AME4 (red line), AME2 (blue line), and AME0 (green line) in the frequency bands 2–5 Hz (B1) and 10–15 Hz (B2).

The dataset also includes auxiliary environmental data, such as atmospheric pressure, air temperature, precipitation, air humidity, wind speed and solar radiation (insolation) recorded at an hourly sampling rate by a weather station located in Flumeri, at about 12 km from the Mefite site (Figure 3). Daily soil moisture data were available for a further station located in Pietradefusi, at a distance of about 23 km (Data accessible at www.agricoltura.regione.campania.it/meteo/agrometeo.htm; last access 25 June 2024). The soil moisture time series was resampled at hourly intervals to ensure consistency in temporal resolution (Figure 3).

Figure 3

Figure 3. From the top to the bottom, time series of air temperature, rainfall, air humidity, wind speed, atmospheric pressure, and insolation recorded at Flumeri station, and soil moisture recorded at Pietradefusi station.

In addition to these local parameters, we also considered a global parameter, corresponding to the Earth’s rotation velocity, known as the length-of-day (LOD), and whose short-term variations record the tidal action on Earth (Lambeck, 2005; Le Mouël et al., 2019). We used the daily measurements of LOD from the EOP14C04 dataset covering the 10 June to 27 September 2021 interval, same time span as the temporary seismic survey, which is provided by the International Earth Rotation Service (IERS, Paris, France, https://www.iers.org/IERS/EN/DataProducts/EarthOrientationData/eop.html).

These various auxiliary data sets were utilized to explore potential correlations between gas emissions and external forcing mechanisms, both meteorological and tidal.

2.2 Methods

Pearson’s correlation coefficients (r) between seismological and environmental parameters were computed to investigate their potential relationships (Chatfield, 2013). A t-test at a significance level of 0.05 was used for testing that each pair of variables is uncorrelated against the alternative hypothesis of a nonzero correlation. It is important to note that correlation does not imply causation, but represents a first step for exploring such a relation.

Stepwise multiple linear regression (MLR) analysis (Draper and Smith, 1998) was performed to quantify the combined effect of meteorological variables on the seismic signal. The procedure follows a series of steps: starting from an initial model, the independent variables (predictors) are first tested using the t-test at a significance level of 0.05. Then, each significant variable is added to the model if it increases the adjusted R² (which is a measure of how much of the variation in the dependent variable is explained by the model) by more than 1%. The process stops when no additional variables improve the model significantly. In multiple regression analysis it should be kept in mind that some of the independent variables may exhibit multicollinearity if they are moderately or highly correlated (Draper and Smith, 1998). This can negatively affect the regression results. The variance inflation factor (VIF) is usually estimated to check for the presence of multicollinearity in a set of multiple regression variables, as it is a measure of how much the variance of an independent variable is influenced by its correlation with the others. A VIF equal to 1 indicates that variables are not correlated. A VIF between 1 and 5 suggests low/moderate multicollinearity which is usually not severe. A VIF greater than 10 corresponds to strong dependence among the independent variables and in this case, corrective actions to reduce multicollinearity, such as the application of Principal Component Analysis (PCA) or regularization, are needed.

We complemented these statistical investigations by considering the Singular Spectrum Analysis (SSA) to assess the presence of periodicities and their modulation over time in the seismic signals recorded at the Mefite site. This non-parametric technique is widely used in time series analysis to decompose complex signals into additive components, such as trends, periodic oscillations, and noise (Vautard et al., 1992; Golyandina and Zhigljavsky, 2013; Courtillot et al., 2022; Petrosino and Dumont, 2022; Lopes et al., 2024). SSA appears therefore as particularly interesting for detecting weak periodic signals in noisy datasets, making it an ideal tool for exploring the link between seismic signals and gas emissions. The details of SSA method can be found in Golyandina and Zhigljavsky (2013), Gibert et al. (2024) and Lopes et al. (2024). It can be resumed in four main steps: (1) the Embedding, with the building of the trajectory matrix from lagged parts of the seismic time series; (2) the Singular Value Decomposition (SVD), which consists in the projection of the trajectory matrix on an ad hoc orthogonal basis whose eigenvalues correspond to the relative importance or contribution of the different components extracted, while the eigenvectors represent the principal directions of variability in the data and dual spaces; (3) the Grouping or reconstruction; step consisting in regrouping the eigen-triplets (eigenvalues and eigenvectors) showing a homogeneous behavior, in our case it was mainly based on the eigenvalue spectrum and modulation, and (4) the Diagonal average allowing to obtain the final time series of each component.

The robustness of the SSA decomposition was assessed by varying the window length L and evaluating the stability of the extracted components. We performed the iterative SSA, which consists in subtracting each component of interest to the original time series and applying the SSA to the residuals as in Dumont et al. (2025). With time series length of 2,635 elements (N), we initiated our analysis with L ∼ N/2, such as 1,300; by exploring the first 15–20 eigenvectors. The residual time series were then similarly analyzed by increasing progressively the window length L. The identification of pseudo-periods with tidal constituents was realized by taking into account the associated uncertainties, estimated from the half-width at half maximum amplitude.

The results of the statistical MLR analysis provide insights into the extent to which meteorological conditions influence the seismic signal, complementing the SSA for identifying periodic components related to tidal forcing.

3 Results

3.1 Descriptive statistics

The analyzed meteorological data exhibit significant variability across different parameters (Table 1). Air temperature has a mean value of 23 °C, ranging from 11 °C to 37 °C, while relative humidity shows a notable fluctuation, with values spanning from 34% to 98%. Precipitation was irregular over the analyzed period, with a peak in mid-July and more frequent rainfall events from mid-August onward. Solar radiation displays the typical pattern associated with diurnal variations, with values ranging from 0.9 to nearly 1000 W/m² and peaks during the central hours of the day. Wind average speed is 2.5 m/s, reaching a maximum of 9.1 m/s. Atmospheric pressure remains relatively stable, with no substantial variations. Soil moisture exhibits smaller fluctuations, with a mean value of 25.8 %VWC and a range between 22.5% and 33.1% VWC. Regarding the seismic time series, the RMS and Rlen values differ across frequency bands and stations. At AME4, the station closest to the degassing area, the mean RMS in the 10–15 Hz band is significantly higher than in the 2–5 Hz band (0.7 vs. 0.09). Conversely, at AME0 and AME2, the ratio between the two bands is less pronounced. Regarding Rlen, AME4 exhibits higher values in the 2–5 Hz band than in the 10–15 Hz band, whereas AME0 and AME2 show less distinct behavior.

Table 1

Table 1. Descriptive statistics of the seismic data acquired at Mefite d’Ansanto and environmental parameters.

3.2 Correlation analysis

The Pearson correlation analysis (Tables 2, 3) reveals some relationships between seismic parameters and environmental variables. Wind speed is positively correlated with RMS in both frequency bands across all stations (r ranging from 0.46 to 0.54, p < 0.001). Solar radiation and air temperature show a positive correlation with RMS (particularly in the 2–5 Hz band), with stronger effects at AME0 and AME2. Air humidity is negatively correlated with RMS and positively correlated with Rlen in the 2–5 Hz band. Soil moisture exhibits different effects depending on the seismic parameter: it is negatively correlated with Rlen in the 2–5 Hz band at all stations, whereas in the 10–15 Hz band, correlations are either positive or negligible depending on the station. Atmospheric pressure is correlated with some variables, although with lower coefficients compared to other environmental parameters.

Table 2

Table 2. Summary of the correlation analysis: correlation coefficients, r, between environmental and seismological variables (RMS and Rlen) and the p-values, for data acquired at AME4, AME2 and AME0 in the 2–5 Hz frequency band. p-values > 0.05 (no significant correlation) are marked in red.

Table 3

Table 3. Summary of the correlation analysis: correlation coefficients, r, between environmental and seismological variables (RMS and Rlen) and the p-values, for data acquired at AME4, AME2 and AME0 in the 10–15 Hz frequency band. p-values > 0.05 (no significant correlation) are marked in red.

3.3 Multiple linear regression analysis

The results from stepwise multiple linear regression (Supplementary Tables S1–S12) indicate that among the independent variables, wind speed is the primary predictor of RMS variation in the 2–5 Hz band for all stations, with a contribution to the adjusted R² between 0.216 and 0.317. Solar radiation emerges as a secondary predictor for AME0 and AME2. In the 10–15 Hz band, the predictive capacity of the regression model is generally lower (adjusted R² between 0.029 and 0.061), with atmospheric pressure and soil moisture emerging as influencing factors, albeit with limited contributions.

For Rlen in the 2–5 Hz band, environmental parameters explain part of the variability (adjusted R² between 0.204 and 0.283), with temperature, wind speed, and soil moisture being the main determinants. Conversely, the 10–15 Hz band exhibits a weaker dependence on environmental factors, with soil moisture identified as the primary predictor (adjusted R² between 0.057 and 0.148).

Overall, the results of the MLR analysis (Supplementary Tables S1–S12) indicate that in the 2–5 Hz frequency band, the meteorological variables explain approximately 20% of the variation of the seismological parameters at AME4 and about 30% at AME2 and AME0 (except for Rlen at the latter, where the explained variance is around 20%). In contrast, in the 10–15 Hz frequency band, the effects of environmental factors are markedly weaker accounting for less than 10% of the variance in seismic data at all the stations, except for Rlen at AME0, where they contribute approximately 15%.

The VIF computed for each independent variable included in the regression models remains below 1.5, suggesting a low degree of multicollinearity. Thus, the current set of predictors does not show problematic interdependencies, ensuring the robustness of the regression estimates.

3.4 Singular Spectrum Analysis

The SSA results (Tables 4–6) reveal the presence of distinct periodic components, identified after Ray and Erofeeva (2014), in the seismic noise RMS and Rlen time series across the three stations, for both the 2–5 Hz and 10–15 Hz frequency bands. Several of these components match well-known long- and short-term tidal periodicities.

Table 4

Table 4. Periodicities detected via SSA, percentage of reconstruction of the original time series (R%) and number of extracted components (Nc) for RMS and Rlen in the 2–5 Hz band, and LOD. Values are in days (mean ± std).

Table 5

Table 5. Periodicities detected via SSA, percentage of reconstruction of the original time series (R%) and number of extracted components (Nc) for RMS and Rlen in the 10–15 Hz band, and LOD.

Table 6

Table 6. Periodicities detected via SSA, percentage of reconstruction of the original time series (R%) and number of extracted components (Nc) in the environmental data.

In the 2–5 Hz band (Table 4), both RMS and Rlen at all stations exhibit statistically significant components near the Msm (31.8 days) and Mm (27.55 days) periods. In the initial portion of the waveform, signal amplitudes are comparable across stations, but subsequently AME2 and AME0 exceed AME4, with AME2 showing the major increase for Rlen (Figures 4A,B). Periodicities in the 14–15 days range, compatible with the Msf (14.77 days) and Mf (13.66 days) constituents, are present in RMS at AME4 and in Rlen at both AME2 and AME4, but are absent from AME0 (Figures 4C,D). This fortnightly signal is generally weak, with AME4 appearing as the most responsive. A 9–10 days component, potentially associated with Mst and Mt, is detected in the Rlen across all stations. Most time series (except Rlen at AME0) also display ∼7.3 days periodicities, which may correspond to Mt subharmonics or reflect the influence of anthropogenic weekly cycles. Mq and Msp occur only in two cases (RMS at AME4 and Rlen at AME0). Diurnal tidal constituents (P1/S1/K1) are consistently detected; in RMS, the daily component reaches the highest (and most stable) amplitude at AME0, followed by AME2, whereas in Rlen both AME2 and AME0 dominate over AME4, which shows the weakest response (Figure 5). Semidiurnal tidal periodicities (S2, M2, K2) are sporadically found.

Figure 4

Figure 4. Tidal components extracted in the 2–5 Hz band: (A) 27–30 days in the RMS, (B) 27–30 days in Rlen, (C) 13–14 days in the RMS, and (D) 13–14 days in Rlen, for AME0 (orange curves), AME2 (purple curves) and AME4 (green curves) stations. The 27-day and 13-day components extracted in the LOD (gray dotted line) are shown for comparison. The respective normalized FFT spectra are also shown on the right.

Figure 5

Figure 5. The 1-day component extracted in the 2–5 Hz band for the RMS (upper panel) and Rlen (lower panel) data for AME0 (orange curves), AME2 (purple curves) and AME4 (green curves) stations.

In the 10–15 Hz band (Table 5), tidal signatures are generally less pronounced compared to the 2–5 Hz band, as confirmed by lower SSA reconstruction percentages of the time series. Nevertheless, components close to long-period tidal constituents remain detectable, especially in Rlen. For the monthly component, AME2 shows a clear predominance in amplitude over both AME4 and AME0, even more marked than in the 2–5 Hz band (Figures 6A,B). The fortnightly signal is poorly extracted in RMS at AME4 (opposite to the 2–5 Hz case), whereas AME2 displays the highest amplitude in both RMS and Rlen. AME0 remains lower, particularly in Rlen (Figures 6C, D). Diurnal constituents P1/S1/K1 appear only at AME0 and AME2, in both RMS and Rlen, and are absent in AME4. Semidiurnal tidal periodicities (S2, M2, K2) are not detected.

Figure 6

Figure 6. Tidal components extracted in the 10–15 Hz band: (A) 27–30 days in the RMS, (B) 27–30 days in Rlen, (C) 13–14 days in the RMS, and (D) 13–14 days in Rlen, for AME0 (orange curves), AME2 (purple curves) and AME4 (green curves) stations. The 27-day and 13-day components extracted in the LOD (gray dotted line) are shown for comparison. The respective normalized FFT spectra are also shown on the right.

We evaluate that in general, the tidal components do not exceed 30%–40% of the original seismic signal. The variance explained by tidal constituents appears systematically larger in the lower frequency band than at higher frequencies for both RMS and Rlen analysis.

SSA applied to the meteorological parameters (Table 6) indicates the presence of monthly, fortnightly and weekly periodic components, which, in the case of pressure and soil moisture, account for less than 20% of the variance. In addition, air temperature, air humidity, wind speed and insolation exhibit daily and half-day periodicities, likely reflecting the diurnal cycle of insolation. The detected harmonic components reconstruct a significant portion of the original time series (50%–90%), consistent with their strong and regular cyclicity.

Finally, the LOD time series decomposition (Tables 4, 5) exhibits periodic components aligned with the Mm, Mf, Mt, and Msp typical frequencies.

4 Discussion

The results presented in the previous sections highlight the presence of multiple forcing mechanisms acting on the seismic noise recorded at the Mefite d’Ansanto site. MLR and singular spectral decomposition point out specific pattern and modulations in the seismic amplitude (RMS) and polarization (Rlen), likely originated by the interplay of tidal action on Earth and meteorological factors.

The MLR analysis reveals that meteorological variables, particularly air temperature, wind speed, and solar radiation, contribute to the variability of seismic observables, with the main effects observed in the 2–5 Hz frequency band. The statistical models explain up to 30% of the variance in certain stations. Conversely, the 10–15 Hz band exhibits weaker correlations, suggesting a reduced sensitivity to environmental drivers at higher frequencies. This difference likely reflects that the hydrothermal tremor in the 2–5 Hz band originates from a shallower and low-energy source, whereas the 10–15 Hz band is characterized by more energetic, intermittent signals generated slightly deeper within the system (Morabito et al., 2024). Consequently, the low frequency noise, compared with the high frequency one, is more strongly influenced by meteorological factors, which primarily act at the surface.

Among the variables tested by MLR, wind emerges as the primary controlling environmental factor for both RMS and Rlen across all stations. Insolation and air temperature also emerge as predictors, but with a least impact. Intriguingly, Tables 2, 3 show that wind speed, insolation, and air temperature all display positive correlations with RMS and negative correlations with Rlen, with the effect being strongest in the 2–5 Hz band.

Similar inverse relationships between these variables and CO₂ flux have been documented in other environments (Lewicki et al., 2007; Viveiros et al., 2008; Rinaldi et al., 2012). These results have been interpreted as resulting from the infiltration of strong winds in the vadose zone, diluting upward gas transport and thereby suppressing surface degassing. For example, Lewicki et al. (2007) observed a strong inverse correlation between wind speed and CO₂ emission at Mammoth Mountain, attributing it to wind-driven airflow through shallow, permeable soils, which alters pressure gradients within the unsaturated zone and partially impedes fluid/gas migration toward the surface. Similarly, Viveiros et al. (2008) proposed that wind-induced infiltration can delay gas exhalation by mixing air into the shallow subsurface, an effect enhanced by highly porous and fractured soils. At Mefite itself, Dioguardi et al. (2025) reported an anticorrelation between wind speed and CO₂ concentration in vent degassing. In parallel, numerical modelling by Rinaldi et al. (2012) indicates that air temperature affects fluid mobility, as higher values reduce CO₂ density and increase viscosity, thereby decreasing the flux.

If we consider the lithostatic pressure located below a 1 × 1 × 1 m cube of weathered soil (density: 800–2000 kg.m^-3) or the same surface but at 10 m depth, we obtain pressure estimates of 78.48–196.20 hPa and 784.8–1962.0 hPa for 1 and 10 m depth, respectively. Considering that the atmospheric pressure variations measured in the area (Figure 3) are of the order of 20 hPa, the modulation of the wind is likely confined to the near-surface, e. g., the first meter. In line with the afore-mentioned studies and our first-order calculations, we hypothesize that in the Mefite system, variations in pressure gradients within the near-surface unsaturated zone, favored by stronger winds or by reduced CO₂ flux during warm periods, could alter pore pressure, local stress distribution and some medium properties such as porosity and permeability. Thus, these changes may promote temporary CO₂ accumulation within the subsurface soil, which could increase seismic noise amplitude (RMS) and reduce azimuthal coherence (Rlen). Laboratory and modelling studies have shown that CO₂–rock interactions can modify porosity and permeability, ultimately affecting bulk density and seismic wave velocity (Kumar et al., 2008). These effects have been confirmed in controlled CO₂ injection experiments, which recorded large amplitude anomalies in the seismic noise near the injection site (Ivanova et al., 2012; Li et al., 2022). Taken together, these lines of evidence support the plausibility that near-surface CO₂-rich fluids redistribution, induced by both wind and thermal conditions, can influence the seismic noise characteristics at Mefite, modulating both amplitude and polarization.

Alternatively, wind might generate noise of purely meteorological origin, such as surface ground vibrations or turbulence-induced pressure fluctuations (see e.g., Lott et al., 2017), or tremor-like signals resulting from interactions with nearby trees (Johnson et al., 2019). However, this meteorological noise would increase RMS and diminish Rlen by adding incoherent energy into the seismic record, independent of any CO₂-driven mechanism. We do not exclude such influence, although the seismometers were buried in areas with sparse vegetation (mainly small bushes and reeds), where the topography does not promote such turbulent behavior. Nevertheless, both interpretations remain plausible, and further investigation, preferably involving direct CO₂ flux measurements, will be essential to discriminate between fluid-mediated versus purely atmospheric influences on seismic noise.

Finally, in both frequency bands, soil moisture emerges as a predictor for Rlen, albeit with a modest contribution, underscoring its usefulness as a sensitive proxy for fluid-driven processes. This result is consistent with previous studies that demonstrated the strong diagnostic potential of polarization parameters in detecting and tracking fluid circulation (Petrosino and De Siena, 2021; Pischiutta et al., 2022; Morabito et al., 2024).

Complementary insights are provided by the spectral decomposition provided by SSA. In the 2–5 Hz band, both RMS and Rlen exhibit statistically significant components matching the periods of well-known tidal constituents (Ray and Erofeeva, 2014), including the monthly, fortnightly, weekly, diurnal and semidiurnal cycles of luni-solar origin. The SSA reconstructions in this band account for a moderate percentage of the variance (13%–37%) of the original time series. The 10–15 Hz band, by contrast, is characterized by lower variance (<10% in most cases, Table 4), indicating a reduced contribution of periodic components in this range. This parallels the observation of a weaker meteorological forcing in the same band. Nevertheless, certain tidal signatures, including those associated with long- and short-period constituents, are still detectable. These findings may reflect a more intermittent source mechanism acting at higher frequencies but still influenced by external forcing, consistent with the interpretation proposed by Morabito et al. (2024), which links the 10–15 Hz band to episodic or impulsive degassing events, in contrast to the persistent background noise observed in the 2–5 Hz band. It is likely that in the case of intermittent seismic signal, the periodicity induced by the external forcing may be masked.

The identification of weekly, 24 h and 12 h periodicities in RMS and Rlen indicates a complex interaction of several forcing mechanisms modulating the seismic noise. In fact, such periodicities may arise not only from tidal forcing but also from environmental influences (as also found in Morabito et al., 2023; Morabito et al., 2024), since the corresponding meteorological series exhibit similar 24 h and 12 h cycles (Tables 4–6), as well as from possible anthropogenic activity. In contrast, fortnightly and monthly periodicities match the characteristic periods of tidal constituents, leaving little ambiguity about their origin.

Further support for a tidal influence is provided by the spectral content of the LOD time series, used as a proxy for global tidal forcing. The match between LOD tidal components (Mm, Mf, Mst, Msp) and those extracted from the seismic records indicates a plausible coupling between large-scale tidal strain and local seismic noise (Dumont et al., 2020; Dumont et al., 2021; Petrosino and Dumont, 2022). While this coupling is not deterministic, its consistency over time suggests that tidal strain may act as a modulating factor. Earth tides contribute to the redistribution of fluid mass within and at the surface of the crust, inducing changes in the stress field and pore-fluid pressure (Dumont et al., 2023, and references therein). Consequently, they can modulate the release of CO₂, independently of its origin, from greater depths to subsurface. In fact, periodic cycles of compressive and tensile stress induced by Earth tides along faults and fracture systems can enhance or inhibit permeability, temporarily opening or closing migration pathways and thus regulating the upward flow of CO₂-rich fluids from depth (Weinlich et al., 2006; Faber et al., 2009; Yang and Zhang, 2021). These observations support a model in which tidal stresses not only act on the Mefite shallow aquifer, whose mixing with CO₂ modulates seismic tremor generation, but can also extend their effects to deeper rock volumes, affecting both the physical state of the medium and the fault/fracture networks, as well as the dynamics of gas migration. Such a mechanism could play a role in controlling the source dynamics of the Mefite CO₂ field, potentially modulating degassing and modifying the properties of the seismic noise. In volcanic environments, both long- and short-period tidal components have been detected in the RMS amplitude and polarization parameters of volcanic and hydrothermal tremor (De Lauro et al., 2013; Dumont et al., 2021; Petrosino and Dumont, 2022). The present observations extend this framework to a tectonic CO₂ degassing setting, providing, to our knowledge, the first indication of tidal modulation in seismic noise parameters (RMS and polarization) associated with CO₂ release of non-volcanic origin.

A slight spatial variability emerges in the response of Mefite’s system. The station closest to the main degassing vents (AME4) systematically displays weaker environmental and tidal signatures in terms of the amplitude of the extracted components, whereas the more distant stations (AME2 and AME0) exhibit stronger responses (Figures 4–6). For the monthly, fortnightly, and daily periodicities, AME2 often shows a dominant behavior, with the highest amplitudes in both RMS and Rlen time series. Overall, considering all the extracted periodic components, in the 2–5 Hz band, the variance of the RMS and Rlen reconstructed time series at AME2 and AME0 exceeds that at AME4 by about 10% (Table 4; Figure 7). This also holds for the meteorological parameters identified as predictor variables in the MLR (Figure 7). Even in the 10–15 Hz range, where environmental and tidal contributions are reduced, AME4 still displays a moderate variance (compared with the other stations) at least for Rlen (Table 5; Figure 7). Such differences may be linked to site effects, consistently with Morabito et al. (2024), who observed a slight amplification of the H/V spectral ratio in the 2–5 Hz band for AME2 and AME0, and in the 10–15 Hz band for AME2. Conversely, AME4 shows an almost flat H/V curve (no amplification) in the 2–5 Hz range. The influence of site effects on the tidal response of geophysical observables is in line with previous findings from seismic noise and ground tilt analyses (De Lauro et al., 2018; Petrosino et al., 2020; Petrosino and Dumont, 2022).

Figure 7

Figure 7. Explained variance percentage of the RMS and Rlen time series from MLR and SSA, in the 2–5 Hz (left panel) and 10–15 Hz (right panel).

The combined results from MLR and SSA indicate that the seismic signal at the Mefite site is influenced by both environmental and tidal drivers. Tidal forcing acts globally on our planet, from the atmosphere to deeper crustal volume, as shown by the tidal variations recorded by the Earth’s rotation axis, affecting shallow aquifers, fluids and crustal rocks up to greater depths. This mechanism alters stress and permeability conditions, and can regulate CO₂ migration from deep reservoirs, along faults and fractures, to the surface. In addition, the variation of the meteorological factors, such as wind and air temperature, induce permeability, porosity and pore pressure changes within the near-subsurface vadose zone. The combined effect of tidal and environmental forcing results in the modulation of the seismic noise originating from shallow sources located in the first tens of meters (Morabito et al., 2024). Overall, these results support a conceptual model (Figure 8) in which meteorological processes modulate the near-surface noise source, whereas tidal strain acts as a global modulating factor, at large-spatial scale, that can influence both the shallow and deeper dynamics, consistently with previous interpretations of the Mefite hydrothermal system (Chiodini et al., 2010; Improta et al., 2014; Morabito et al., 2024).

Figure 8

Figure 8. Conceptual model (not in scale) illustrating environmental and tidal effects on the Mefite d’Ansanto system. The left panel shows an unperturbed state: the seismic noise generated by the source is recorded with a given amplitude and polarization (upper left) by a hypothetical seismic station (green triangle). The right panel illustrates the action of insolation (yellow dotted arrow), wind (light blue dotted arrow) and tides (deep blue arrows), which induce variations in medium properties, reduce outgassing and, consequently, lead to amplitude variation and depolarization of the seismic noise.

Although the variability of the seismic noise at Mefite d’Ansanto appears partly driven by meteorological and tidal factors, alternative endogenous processes may also contribute. Nonlinear interactions between the solid rock matrix and fluids can produce sustained tremor (Julian, 1994), whose amplitude and frequency may vary with changes in fluid pressure, composition, or in the geometry of conduits and fractures. For instance, at Ischia Island, the modulation of the sustained tremor has been linked to energy variations triggered by pressure build-up in the hydrothermal system (Falanga et al., 2021). Similarly, the variable patterns of the seismic signals associated with methane degassing at Salse di Nirano mud volcanoes have been interpreted as the result of a stick-slip mechanism, that regulates gas outflow through the interaction between exsolved gas bubbles, mud plugs, and vent walls (Carfagna et al., 2024). These observations suggest that a complex interplay between endogenous and exogenous factors should be considered to fully describe the characteristics of the seismic noise.

5 Conclusion

This study shows that seismic noise at the Mefite d’Ansanto site reflects the influence of both environmental and tidal drivers, contributing to modulations over a range of time scales. The combined application of MLR and SSA reveals a consistent framework in which meteorological factors mostly affect shallow, near-surface seismic noise sources, whereas tidal strain exerts a larger-scale control, extending its influence on both shallow and deeper processes within the system. The polarization features of the seismic noise, in particular, appear to offer a sensitive window into these interactions, highlighting their potential use in monitoring strategies aimed at identifying external forcing on fluid-driven seismic sources.

To our knowledge, this work provides the first observation of tidal modulation in amplitude and polarization of the seismic noise sourced within a tectonic CO₂ degassing setting. The obtained results and their preliminary interpretation form a solid basis for future numerical modelling of this complex phenomenology.

Future developments could focus on exploring lagged relationships between environmental/tidal factors and seismic time series, to investigate potential delayed system responses and better constrain the coupling between exogenous forcing and fluid-flow-related seismic noise. Additionally, joint field measurements (e.g., CO₂ flux, meteorological data, seismic noise and ground deformation) could further refine the understanding of the mechanisms governing such interactions.

Data availability statement

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Author contributions

SP: Data curation, Formal Analysis, Investigation, Writing – review and editing, Supervision, Writing – original draft, Conceptualization. SD: Investigation, Writing – review and editing, Formal Analysis, Data curation. SM: Data curation, Writing – review and editing. PC: Data curation, Writing – review and editing, Project administration.

Funding

The author(s) declared that financial support was received for this work and/or its publication. Data used for the current study are acquired in the framework of the 2020–2024 FURTHER project (INGV) and are available from the authors on reasonable request. This work was also supported by the Portuguese Fundação para a Ciência e Tecnologia (FCT) I.P./MCTES through national funds (PIDDAC): UID/50019/2025 and LA/P/0068/2020 (https://doi.org/10.54499/LA/P/0068/2020).

Acknowledgements

We thank the Rocca San Felice community for their kind availability and hospitality.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/feart.2025.1708254/full#supplementary-material

References

Boudoire, G., Di Muro, A., Liuzzo, M., Ferrazzini, V., Peltier, A., Gurrieri, S., et al. (2017). New perspectives on volcano monitoring in a tropical environment: continuous measurements of soil CO2 flux at Piton de la Fournaise (La Réunion Island, France). Geophys. Res. Lett. 44 (16), 8244–8253. doi:10.1002/2017GL074237

CrossRef Full Text | Google Scholar

Carapezza, M. L., Badalamenti, B., Cavarra, L., and Scalzo, A. (2003). Gas hazard assessment in a densely inhabited area of Colli Albani Volcano (Cava dei Selci, Roma). J. Volcanol. Geotherm. Res. 123 (1-2), 81–94. doi:10.1016/S0377-0273(03)00029-5

CrossRef Full Text | Google Scholar

Carfagna, N., Brindisi, A., Paolucci, E., and Albarello, D. (2024). Seismic monitoring of gas emissions at mud volcanoes: the case of nirano (northern Italy). J. Volcanol. Geotherm. Res. 446, 107993. doi:10.1016/j.jvolgeores.2023.107993

CrossRef Full Text | Google Scholar

Chatfield, C. (2013). The analysis of time series: theory and practice. Springer.

Google Scholar

Chiodini, G., Granieri, D., Avino, R., Caliro, S., Costa, A., Minopoli, C., et al. (2010). Non-volcanic CO2 Earth degassing: case of mefite d'Ansanto (southern apennines), Italy. Geophys. Res. Lett. 37 (11). doi:10.1029/2010GL042858

CrossRef Full Text | Google Scholar

Courtillot, V., Le Mouël, J. L., Lopes, F., and Gibert, D. (2022). On the nature and origin of atmospheric annual and semi-annual oscillations. Atmosphere 13 (11), 1907. doi:10.3390/atmos13111907

CrossRef Full Text | Google Scholar

De Lauro, E., Petrosino, S., Ricco, C., Aquino, I., and Falanga, M. (2018). Medium and long period ground oscillatory pattern inferred by borehole tiltmetric data: new perspectives for the campi flegrei caldera crustal dynamics. Earth Planet. Sci. Lett. 504, 21–29. doi:10.1016/j.epsl.2018.09.039

CrossRef Full Text | Google Scholar

De Lauro, E., De Martino, S., Falanga, M., and Petrosino, S. (2013). Synchronization between tides and sustained oscillations of the hydrothermal system of Campi Flegrei (Italy). Geochem. Geophys. Geosyst. 14 (8), 2628–2637. doi:10.1002/ggge.20149

CrossRef Full Text | Google Scholar

Dioguardi, F., Chiodini, G., and Costa, A. (2025). Probabilistic hazard analysis of the gas emission of mefite d'Ansanto, southern Italy. Nat. Hazards Earth Syst. Sci. 25 (2), 657–674. doi:10.5194/nhess-25-657-2025

CrossRef Full Text | Google Scholar

Doodson, A. T. (1921). The harmonic development of the tide-generating potential. Proc. R. Soc. Lond. Ser. A 100, 305–329.

Google Scholar

Draper, N. R., and Smith, H. (1998). Applied regression analysis, 326. John Wiley & Sons.

Google Scholar

Dumont, S., Le Mouël, J. L., Courtillot, V., Lopes, F., Sigmundsson, F., Coppola, D., et al. (2020). The dynamics of a long-lasting effusive eruption modulated by Earth tides. Earth Planet. Sci. Lett. 536, 116145. doi:10.1016/j.epsl.2020.116145

CrossRef Full Text | Google Scholar

Dumont, S., Silveira, G., Custodio, S., Lopes, F., Le Mouel, J. L., Gouhier, M., et al. (2021). Response of fogo volcano (cape verde) to lunisolar gravitational forces during the 2014–2015 eruption. Phys. Earth Planet. Interiors 312, 106659. doi:10.1016/j.pepi.2021.106659

CrossRef Full Text | Google Scholar

Dumont, S., Custódio, S., Petrosino, S., Thomas, A. M., and Sottili, G. (2023). Tides, earthquakes, and volcanic eruptions. A Journey Through Tides, 333–364. doi:10.1016/b978-0-323-90851-1.00008-x

CrossRef Full Text | Google Scholar

Dumont, S., de Bremond d’Ars, J., Boulé, J.-B., Courtillot, V., Gèze, M., Gibert, D., et al. (2025). On a planetary forcing of global seismicity. Front. Earth Sci. 13, 1587650. doi:10.3389/feart.2025.1587650

CrossRef Full Text | Google Scholar

Faber, E., Horálek, J., Boušková, A., Teschner, M., Koch, U., and Poggenburg, J. (2009). Continuous gas monitoring in the west Bohemian earthquake area, Czech Republic: first results. Studia Geophys. Geod. 53, 315–328. doi:10.1007/s11200-009-0020-z

CrossRef Full Text | Google Scholar

Falanga, M., Cusano, P., De Lauro, E., and Petrosino, S. (2021). Picking up the hydrothermal whisper at ischia island in the Covid-19 lockdown quiet. Sci. Rep. 11 (1), 8871. doi:10.1038/s41598-021-88266-9

PubMed Abstract | CrossRef Full Text | Google Scholar

Gibert, D., Lopes, F., Courtillot, V., and Boulé, J. B. (2024). Information theory, signal analysis and inverse problem. arXiv Prepr., arXiv:2408.16361. doi:10.48550/arXiv.2408.16361

CrossRef Full Text | Google Scholar

Giustini, F., and Brilli, M. (2020). Mefite d’Ansanto, southern apennines (italy): the natural CO 2 seep which emits the largest quantity of non-volcanic CO 2 on Earth. Int. J. Earth Sci. 109, 1705–1706. doi:10.1007/s00531-020-01857-1

CrossRef Full Text | Google Scholar

Golyandina, N., and Zhigljavsky, A. (2013). Singular spectrum analysis for time series. Berlin, Germany: Springer.

Google Scholar

Hsu, S. K., Wang, S. Y., Liao, Y. C., Yang, T. F., Jan, S., Lin, J. Y., et al. (2013). Tide-modulated gas emissions and tremors off SW Taiwan. Earth Planet. Sci. Lett. 369, 98–107. doi:10.1016/j.epsl.2013.03.013

CrossRef Full Text | Google Scholar

Improta, L., Bonagura, M., Capuano, P., and Iannaccone, G. (2003). An integrated geophysical investigation of the upper crust in the epicentral area of the 1980, MS = 69, irpinia earthquake (Southern Italy). Tectonophysics 361, 139–169. doi:10.1016/S0040-1951(02)00588-7

CrossRef Full Text | Google Scholar

Improta, L., De Gori, P., and Chiarabba, C. (2014). New insights into crustal structure, Cenozoic magmatism, CO2 degassing, and seismogenesis in the southern apennines and irpinia region from local earthquake tomography. J. Geophys. Res. Solid Earth 119 (11), 8283–8311. doi:10.1002/2013jb010890

CrossRef Full Text | Google Scholar

Ivanova, A., Kashubin, A., Juhojuntti, N., Kummerow, J., Henninges, J., Juhlin, C., et al. (2012). Monitoring and volumetric estimation of injected CO2 using 4D seismic, petrophysical data, core measurements and well logging: a case study at ketzin, Germany. Geophys. Prospect. 60 (5), 957–973. doi:10.1111/j.1365-2478.2012.01045.x

CrossRef Full Text | Google Scholar

Johnson, C. W., Meng, H., Vernon, F., and Ben-Zion, Y. (2019). Characteristics of ground motion generated by wind interaction with trees, structures, and other surface obstacles. J. Geophys. Res. Solid Earth 124 (8), 8519–8539. doi:10.1029/2018jb017151

CrossRef Full Text | Google Scholar

Julian, B. R. (1994). Volcanic tremor: nonlinear excitation by fluid flow. J. Geophys. Res. Solid Earth 99 (B6), 11859–11877. doi:10.1029/93jb03129

CrossRef Full Text | Google Scholar

Kumar, A., Datta-Gupta, A., Shekhar, R., and Gibson, R. L. (2008). Modeling time lapse seismic monitoring of CO2 sequestration in hydrocarbon reservoirs including compositional and geochemical effects. Petroleum Science Technology 26 (7-8), 887–911. doi:10.1080/10916460701825505

CrossRef Full Text | Google Scholar

Lambeck, K. (2005). The Earth’s variable rotation: geophysical causes and consequences. Cambridge University Press.

CrossRef Full Text | Google Scholar

La Rocca, M., Galluzzo, D., Nardone, L., Gaudiosi, G., and Di Luccio, F. (2023). Hydrothermal seismic tremor in a wide frequency band: the nonvolcanic CO 2 degassing site of mefite d’Ansanto, Italy. Bull. Seismol. Soc. Am. 113 (3), 1102–1114. doi:10.1785/0120220243

CrossRef Full Text | Google Scholar

Le Mouël, J. L., Lopes, F., Courtillot, V., and Gibert, D. (2019). On forcings of length of day changes: from 9-day to 18.6-year oscillations. Phys. Earth Planet. Interiors 292, 1–11. doi:10.1016/j.pepi.2019.04.006

CrossRef Full Text | Google Scholar

Lewicki, J. L., and Hilley, G. E. (2014). Multi-scale observations of the variability of magmatic CO2 emissions, Mammoth Mountain, CA, USA. J. Volcanology Geothermal Research 284, 1–15. doi:10.1016/j.jvolgeores.2014.07.011

CrossRef Full Text | Google Scholar

Lewicki, J. L., Hilley, G. E., Tosha, T., Aoyagi, R., Yamamoto, K., and Benson, S. M. (2007). Dynamic coupling of volcanic CO2 flow and wind at the horseshoe Lake tree kill, Mammoth Mountain, California. Geophys. Res. Lett. 34 (3). doi:10.1029/2006GL028848

CrossRef Full Text | Google Scholar

Li, T., Gu, Y. J., Lawton, D. C., Gilbert, H., Macquet, M., Savard, G., et al. (2022). Monitoring CO2 injection at the cami field research station using microseismic noise sources. J. Geophys. Res. Solid Earth 127 (12), e2022JB024719. doi:10.1029/2022jb024719

CrossRef Full Text | Google Scholar

Lopes, F., Gibert, D., Courtillot, V., Mouël, J. L. L., and Boulé, J. B. (2024). On the optimization of singular spectrum analyses: a pragmatic approach. arXiv Preprint arXiv:2412.17793.

Google Scholar

López, D. L., Bundschuh, J., Soto, G. J., Fernández, J. F., and Alvarado, G. E. (2006). Chemical evolution of thermal springs at arenal volcano, Costa Rica: effect of volcanic activity, precipitation, seismic activity, and Earth tides. J. Volcanology Geothermal Research 157 (1-3), 166–181. doi:10.1016/j.jvolgeores.2006.03.049

CrossRef Full Text | Google Scholar

Lott, F. F., Ritter, J. R., Al-Qaryouti, M., and Corsmeier, U. (2017). On the analysis of wind-induced noise in seismological recordings: approaches to present wind-induced noise as a function of wind speed and wind direction. Pure Appl. Geophys. 174 (3), 1453–1470. doi:10.1007/s00024-017-1477-2

CrossRef Full Text | Google Scholar

Martinelli, G., Ciolini, R., Facca, G., Fazio, F., Gherardi, F., Heinicke, J., et al. (2021). Tectonic-related geochemical and hydrological anomalies in Italy during the last fifty years. Minerals 11 (2), 107. doi:10.3390/min11020107

CrossRef Full Text | Google Scholar

Morabito, S., Cusano, P., Galluzzo, D., Gaudiosi, G., Nardone, L., Del Gaudio, P., et al. (2023). One-year seismic survey of the tectonic CO2-rich site of mefite d’Ansanto (southern Italy): preliminary insights in the seismic noise wavefield. Sensors 23 (3), 1630. doi:10.3390/s23031630

PubMed Abstract | CrossRef Full Text | Google Scholar

Morabito, S., Cusano, P., Nardone, L., and Petrosino, S. (2024). A model for the hydrothermal tremor source of the mefite d’ansanto (italy) CO2 non-volcanic emissions in the intermediate frequency band (1–15 Hz). Sci. Rep. 14 (1), 19480. doi:10.1038/s41598-024-70022-4

PubMed Abstract | CrossRef Full Text | Google Scholar

Morita, M., Mori, T., Yokoo, A., Ohkura, T., and Morita, Y. (2019). Continuous monitoring of soil CO₂ flux at Aso volcano, Japan: the influence of environmental parameters on diffuse degassing. Earth, Planets and Space 71 (1), 1. doi:10.1186/s40623-018-0980-8

CrossRef Full Text | Google Scholar

Ortolani, F., De Gennaro, M., Ferreri, M., Ghiara, M. R., Stanzione, D., and Zenone, F. (1981). Prospettive geotermiche dell'Irpinia centrale (Appennino meridionale); studio geologico-strutturale e geochimico. Boll. Della Soc. Geol. Ital. 100 (1), 139–159.

Google Scholar

Panebianco, S., Satriano, C., Vivone, G., Picozzi, M., Strollo, A., and Stabile, T. A. (2024). Automated detection and machine learning-based classification of seismic tremors associated with a non-volcanic gas emission (Mefite d’Ansanto, Southern Italy). Geochem. Geophys. Geosystems 25 (2), e2023GC011286. doi:10.1029/2023GC011286

CrossRef Full Text | Google Scholar

Petrosino, S., and De Siena, L. (2021). Fluid migrations and volcanic earthquakes from depolarized ambient noise. Nat. Commun. 12 (1), 6656. doi:10.1038/s41467-021-26954-w

PubMed Abstract | CrossRef Full Text | Google Scholar

Petrosino, S., and Dumont, S. (2022). Tidal modulation of hydrothermal tremor: examples from ischia and campi flegrei volcanoes, Italy. Front. Earth Sci. 9, 775269. doi:10.3389/feart.2021.775269

CrossRef Full Text | Google Scholar

Petrosino, S., Ricco, C., De Lauro, E., Aquino, I., and Falanga, M. (2020). Time evolution of medium and long-period ground tilting at campi flegrei caldera. Adv. Geosciences 52, 9–17. doi:10.5194/adgeo-52-9-2020

CrossRef Full Text | Google Scholar

Pischiutta, M., Anselmi, M., Cianfarra, P., Rovelli, A., and Salvini, F. (2013). Directional site effects in a non-volcanic gas emission area (mefite d’Ansanto, southern Italy): evidence of a local transfer fault transversal to large NW–SE extensional faults? Phys. Chem. Earth, Parts A/B/C 63, 116–123. doi:10.1016/j.pce.2013.03.008

CrossRef Full Text | Google Scholar

Pischiutta, M., Petrosino, S., and Nappi, R. (2022). Directional amplification and ground motion polarization in casamicciola area (ischia volcanic island) after the 21 August 2017 Md 4.0 earthquake. Front. Earth Sci. 10, 999222. doi:10.3389/feart.2022.999222

CrossRef Full Text | Google Scholar

Ray, R. D., and Erofeeva, S. Y. (2014). Long-period tidal variations in the length of day. J. Geophys. Res. Solid Earth 119, 1498–1509. doi:10.1002/2013JB010830

CrossRef Full Text | Google Scholar

Rinaldi, A. P., Vandemeulebrouck, J., Todesco, M., and Viveiros, F. (2012). Effects of atmospheric conditions on surface diffuse degassing. J. Geophys. Res. Solid Earth 117 (B11). doi:10.1029/2012jb009490

CrossRef Full Text | Google Scholar

Rogie, J. D., Kerrick, D. M., Sorey, M. L., Chiodini, G., and Galloway, D. L. (2001). Dynamics of carbon dioxide emission at Mammoth Mountain, California. Earth Planet. Sci. Lett. 188 (3-4), 535–541. doi:10.1016/S0012-821X(01)00344-2

CrossRef Full Text | Google Scholar

Vautard, R., Yiou, P., and Ghil, M. (1992). Singular-spectrum analysis: a toolkit for short, noisy chaotic signals. Phys. D. Nonlinear Phenom. 58 (1-4), 95–126. doi:10.1016/0167-2789(92)90103-t

CrossRef Full Text | Google Scholar

Viveiros, F., and Silva, C. (2024). Hazardous volcanic CO2 diffuse degassing areas–a systematic review on environmental impacts, health and mitigation strategies. Iscience 27 (10), 110990. doi:10.1016/j.isci.2024.110990

PubMed Abstract | CrossRef Full Text | Google Scholar

Viveiros, F., Ferreira, T., Vieira, J. C., Silva, C., and Gaspar, J. L. (2008). Environmental influences on soil CO2 degassing at furnas and fogo volcanoes (são miguel island, azores archipelago). J. Volcanol. Geotherm. Res. 177 (4), 883–893. doi:10.1016/j.jvolgeores.2008.07.005

CrossRef Full Text | Google Scholar

Viveiros, F., Vandemeulebrouck, J., Rinaldi, A. P., Ferreira, T., Silva, C., and Cruz, J. V. (2014). Periodic behavior of soil CO2 emissions in diffuse degassing areas of the azores archipelago: application to seismovolcanic monitoring. J. Geophys. Res. Solid Earth 119 (10), 7578–7597. doi:10.1002/2014JB011118

CrossRef Full Text | Google Scholar

Weinlich, F. H., Faber, E., Boušková, A., Horálek, J., Teschner, M., and Poggenburg, J. (2006). Seismically induced variations in mariánské lázně fault gas composition in the NW Bohemian swarm quake region, Czech republic—A continuous gas monitoring. Tectonophysics 421 (1-2), 89–110. doi:10.1016/j.tecto.2006.04.012

CrossRef Full Text | Google Scholar

Yang, D., and Zhang, L. (2021). Carbon dioxide leakages through fault zones: potential implications for the long-term integrity of geological storage sites. Aerosol Air Qual. Res. 21 (12), 210220. doi:10.4209/aaqr.210220

CrossRef Full Text | Google Scholar

Zuddas, P., and Lopes, F. (2023). The effect of lunar declination on CO2 degassing from central Italian apennines. Innovation Geoscience 2 (2), 100073. doi:10.59717/j.xinn-geo.2024.100073

CrossRef Full Text | Google Scholar