Fluid injections in the subsurface are common operations in underground industrial activities such as oil and gas exploitation, geothermal energy development, and carbon capture and storage (CCS). In recent years, it became a focal point as new drilling technologies (e.g., hydraulic fracturing) enable the extraction of oil and gas in unconventional reservoirs and the development of CCS injection techniques became a key research topic in the context of the low-carbon energy transition. Fluid injections have drawn the attention also in the general public because of their main potential implications such as the induced seismicity phenomenon (Rubinstein and Mahani, 2015) and the environmental pollution (Burton et al., 2016, Pitchel et al., 2016). Considering the strong socioeconomic impact of fluid injection operations (National Research Council, 2013; Ellsworth, 2013; Grigoli et al., 2017) the current research in this field needs the integration of multidisciplinary studies, involving knowledge on geology, seismology, source physics, hydrogeology, fluid geochemistry, rocks geomechanics for a complete understanding of the phenomenon and to set-up the most effective and “best practice” protocols for the monitoring of areas where injection operation are performed. On this basis, this work applied a multidisciplinary approach integrating seismological methods, geochemical studies, and machine learning techniques. Two key-study areas characterized by high fluid-rock interaction and fluid-injection in the subsurface were analyzed: i) the High Agri Valley (hereinafter HAV), hosting the largest onshore oil field in West Europe, in which wastewater disposal operations have been carried out since 2006 at the Costa Molina 2 injection well and where both natural and induced seismicity clusters were recognized; ii) the Mefite d’Ansanto, the largest natural emission of CO2-rich gases with mantle-derived fluids (from non‐volcanic environment) ever measured on the Earth (Carcausi et al., 2013; Caracausi and Paternoster, 2015; Chiodini et al, 2010). Regarding the HAV study area, we reconstructed the preliminary catalogue of seismicity through accurate absolute locations in a 3D-velocity model (Serlenga and Stabile, 2019) of earthquakes detected from the local seismic INSIEME network managed by the CNR-IMAA. A total of 852 between local tectonic and induced earthquakes occurred in the HAV between September 2016 and March 2019. We tested the potential of the unsupervised machine-learning approach as an automated tool to make faster dataset exploratory analysis, founding the density-based approach (DBSCAN algorithm-Density-Based Spatial Clustering of Applications with Noise, Ester et al., 1996) particularly suitable for the fast identification of clusters in the catalogue resulting from both injection-induced events and tectonic local earthquake swarms. Moreover, we proposed a semi-automated workflow for earthquake detection and location with the aim to improve the current standard procedures, quite time-consuming and strictly related to human operators. The workflow, integrating manual, semi-automatic and automatic detection and location methods enabled us to characterize a low magnitude natural seismic sequence occurred in August 2020 in the southwestern area of the HAV (Castelsaraceno sequence) in a relatively short time with respect to the application of standard techniques, thus representing a starting point for the improvement of the efficiency of seismic monitoring techniques of both anthropogenic and natural seismicity in the HAV. Our multidisciplinary approach involved the geochemical study of the HAV groundwaters with the aim to: (1) determine the geochemical processes controlling the chemical composition; (2) define a geochemical conceptual model regarding fluid origin (deep vs shallow) and mixing processes by means isotopic data; (3) establish a geochemical baseline for the long-term environmental monitoring of the area. A total of 39 water samples were collected from springs and wells located at the main hydro-structures bordering the valley to determine chemical (major, minor and trace elements) and isotopic composition (e.g., dD, d18O, d13C-TDIC and noble gas). All investigated water samples have a meteoric origin, although some springs show long and deep flow than the other ones, and a bicarbonate alkaline-earth composition, thus suggesting the carbonate hydrolysis as the main water-rock interaction process. Our results demonstrated that HAV groundwater is chemically suitable for drinking use showing no criticalities for potentially toxic metals reported by the Italian and European legislation guidelines. Particular attention was given on thermal water of Tramutola well, built by Agip S.p.a. for oil & gas exploration, with the occurrence of bubbling gases. The geochemical study highlighted a substantial difference of these CH4-dominated thermal fluids with the rest of the dataset. Helium isotope (3He/4He) indicate a prevalent radiogenic component with a contribution of mantle-derived helium (~20%) and the average δ13C-CO2 value is of – 4.6 ‰ VPDB, consistent with a mantle origin. Methane isotope composition indicates a likely microbial isotopic signature (δ13C-CH4 =−63.1‰, −62.4‰, δD-CH4=−196‰, −212‰), probably due to biodegradation processes of thermogenic hydrocarbons. The methane output at the well, evaluated by means of anemometric measurement of the volume flow (m3/h) is of ~156 t/y, that represent about 1.5% of total national anthropogenic sources related to fossil fuel industry (Etiope et al., 2007). Our work highlighted that Tramutola well may represent a key natural laboratory to better understand the complex coupling effects between mechanical and fluid-dynamic processes in earthquake generation. Moreover, the integration of seismic and geochemical data in this work allowed us to identify the most suitable locations for the future installation of multiparametric stations for the long-term monitoring of the area and development of integrated research in the HAV. Regarding the Mefite d’Ansanto, we analyzed the background seismicity in the emission area recorded by a dense temporary seismic network deployed at the site between 30-10-2019 and 02-11-2019. First, we implemented and tested an automated detection algorithm based on non-parametric statistics of the recorded amplitudes at each station, collecting a total dataset of 8561 events. Then, both unsupervised (DBSCAN) and supervised (KNN-k-nearest neighbors classification, Fix & Hodges, 1951) machine learning techniques were applied, based on specific parameters (duration, RMS-amplitude and arrival slope) of the detected events. DBSCAN algorithm allowed to determine characteristic bivariate correlations among tremors parameters: a high linear correlation (r~0.6-0.7) between duration and RMS-amplitude and a lower one (r~0.5-0.6) between amplitude and arrival slope (first arrival parametrization). These relationships let us to define training samples for the KNN algorithm, which allowed to classify tremor signals at each station and to automatically discriminate between tremors and accidentally detected anthropogenic noise. Results allowed to extract new information on seismic tremor at Mefite d’Ansanto, previously poorly quantitively analyzed, and its discrimination, thus providing a starting workflow for monitoring the non-volcanic emission. Isotopic geochemistry (3He/4He, 4 He/20Ne, δ13CCO2) indicated a mixing of mantle (30%-40%) and crust-derived fluids. The source location of the emission related tremor would represent a step forward in its characterization, and for setting up more advanced automated detection and machine learning classification techniques to exploit the information provided by seismic tremor for an improved automatic monitoring of non-volcanic, CO2 -gas emissions.

Fluid injections in the subsurface: a multidisciplinary approach for better understanding their implications on induced seismicity and the environment / Panebianco, Serena. - (2022 Mar 10).

Fluid injections in the subsurface: a multidisciplinary approach for better understanding their implications on induced seismicity and the environment.

PANEBIANCO, SERENA
2022-03-10

Abstract

Fluid injections in the subsurface are common operations in underground industrial activities such as oil and gas exploitation, geothermal energy development, and carbon capture and storage (CCS). In recent years, it became a focal point as new drilling technologies (e.g., hydraulic fracturing) enable the extraction of oil and gas in unconventional reservoirs and the development of CCS injection techniques became a key research topic in the context of the low-carbon energy transition. Fluid injections have drawn the attention also in the general public because of their main potential implications such as the induced seismicity phenomenon (Rubinstein and Mahani, 2015) and the environmental pollution (Burton et al., 2016, Pitchel et al., 2016). Considering the strong socioeconomic impact of fluid injection operations (National Research Council, 2013; Ellsworth, 2013; Grigoli et al., 2017) the current research in this field needs the integration of multidisciplinary studies, involving knowledge on geology, seismology, source physics, hydrogeology, fluid geochemistry, rocks geomechanics for a complete understanding of the phenomenon and to set-up the most effective and “best practice” protocols for the monitoring of areas where injection operation are performed. On this basis, this work applied a multidisciplinary approach integrating seismological methods, geochemical studies, and machine learning techniques. Two key-study areas characterized by high fluid-rock interaction and fluid-injection in the subsurface were analyzed: i) the High Agri Valley (hereinafter HAV), hosting the largest onshore oil field in West Europe, in which wastewater disposal operations have been carried out since 2006 at the Costa Molina 2 injection well and where both natural and induced seismicity clusters were recognized; ii) the Mefite d’Ansanto, the largest natural emission of CO2-rich gases with mantle-derived fluids (from non‐volcanic environment) ever measured on the Earth (Carcausi et al., 2013; Caracausi and Paternoster, 2015; Chiodini et al, 2010). Regarding the HAV study area, we reconstructed the preliminary catalogue of seismicity through accurate absolute locations in a 3D-velocity model (Serlenga and Stabile, 2019) of earthquakes detected from the local seismic INSIEME network managed by the CNR-IMAA. A total of 852 between local tectonic and induced earthquakes occurred in the HAV between September 2016 and March 2019. We tested the potential of the unsupervised machine-learning approach as an automated tool to make faster dataset exploratory analysis, founding the density-based approach (DBSCAN algorithm-Density-Based Spatial Clustering of Applications with Noise, Ester et al., 1996) particularly suitable for the fast identification of clusters in the catalogue resulting from both injection-induced events and tectonic local earthquake swarms. Moreover, we proposed a semi-automated workflow for earthquake detection and location with the aim to improve the current standard procedures, quite time-consuming and strictly related to human operators. The workflow, integrating manual, semi-automatic and automatic detection and location methods enabled us to characterize a low magnitude natural seismic sequence occurred in August 2020 in the southwestern area of the HAV (Castelsaraceno sequence) in a relatively short time with respect to the application of standard techniques, thus representing a starting point for the improvement of the efficiency of seismic monitoring techniques of both anthropogenic and natural seismicity in the HAV. Our multidisciplinary approach involved the geochemical study of the HAV groundwaters with the aim to: (1) determine the geochemical processes controlling the chemical composition; (2) define a geochemical conceptual model regarding fluid origin (deep vs shallow) and mixing processes by means isotopic data; (3) establish a geochemical baseline for the long-term environmental monitoring of the area. A total of 39 water samples were collected from springs and wells located at the main hydro-structures bordering the valley to determine chemical (major, minor and trace elements) and isotopic composition (e.g., dD, d18O, d13C-TDIC and noble gas). All investigated water samples have a meteoric origin, although some springs show long and deep flow than the other ones, and a bicarbonate alkaline-earth composition, thus suggesting the carbonate hydrolysis as the main water-rock interaction process. Our results demonstrated that HAV groundwater is chemically suitable for drinking use showing no criticalities for potentially toxic metals reported by the Italian and European legislation guidelines. Particular attention was given on thermal water of Tramutola well, built by Agip S.p.a. for oil & gas exploration, with the occurrence of bubbling gases. The geochemical study highlighted a substantial difference of these CH4-dominated thermal fluids with the rest of the dataset. Helium isotope (3He/4He) indicate a prevalent radiogenic component with a contribution of mantle-derived helium (~20%) and the average δ13C-CO2 value is of – 4.6 ‰ VPDB, consistent with a mantle origin. Methane isotope composition indicates a likely microbial isotopic signature (δ13C-CH4 =−63.1‰, −62.4‰, δD-CH4=−196‰, −212‰), probably due to biodegradation processes of thermogenic hydrocarbons. The methane output at the well, evaluated by means of anemometric measurement of the volume flow (m3/h) is of ~156 t/y, that represent about 1.5% of total national anthropogenic sources related to fossil fuel industry (Etiope et al., 2007). Our work highlighted that Tramutola well may represent a key natural laboratory to better understand the complex coupling effects between mechanical and fluid-dynamic processes in earthquake generation. Moreover, the integration of seismic and geochemical data in this work allowed us to identify the most suitable locations for the future installation of multiparametric stations for the long-term monitoring of the area and development of integrated research in the HAV. Regarding the Mefite d’Ansanto, we analyzed the background seismicity in the emission area recorded by a dense temporary seismic network deployed at the site between 30-10-2019 and 02-11-2019. First, we implemented and tested an automated detection algorithm based on non-parametric statistics of the recorded amplitudes at each station, collecting a total dataset of 8561 events. Then, both unsupervised (DBSCAN) and supervised (KNN-k-nearest neighbors classification, Fix & Hodges, 1951) machine learning techniques were applied, based on specific parameters (duration, RMS-amplitude and arrival slope) of the detected events. DBSCAN algorithm allowed to determine characteristic bivariate correlations among tremors parameters: a high linear correlation (r~0.6-0.7) between duration and RMS-amplitude and a lower one (r~0.5-0.6) between amplitude and arrival slope (first arrival parametrization). These relationships let us to define training samples for the KNN algorithm, which allowed to classify tremor signals at each station and to automatically discriminate between tremors and accidentally detected anthropogenic noise. Results allowed to extract new information on seismic tremor at Mefite d’Ansanto, previously poorly quantitively analyzed, and its discrimination, thus providing a starting workflow for monitoring the non-volcanic emission. Isotopic geochemistry (3He/4He, 4 He/20Ne, δ13CCO2) indicated a mixing of mantle (30%-40%) and crust-derived fluids. The source location of the emission related tremor would represent a step forward in its characterization, and for setting up more advanced automated detection and machine learning classification techniques to exploit the information provided by seismic tremor for an improved automatic monitoring of non-volcanic, CO2 -gas emissions.
High Agri Valley; Mefite d'Ansanto; Induced seismicity; Machine Learning; Environmental Geochemistry
Fluid injections in the subsurface: a multidisciplinary approach for better understanding their implications on induced seismicity and the environment / Panebianco, Serena. - (2022 Mar 10).
File in questo prodotto:
File Dimensione Formato  
PhD_Thesis_Panebianco.pdf

accesso aperto

Descrizione: Tesi di dottorato
Tipologia: Tesi di dottorato
Licenza: DRM non definito
Dimensione 15.41 MB
Formato Adobe PDF
15.41 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11563/154467
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact