The micro‐seismicity of Co. Donegal (Ireland): Defining baseline seismicity in a region of slow lithospheric deformation

A catalogue of precisely located micro‐seismicity is fundamental for investigating seismicity and rock physical properties in active tectonic and volcanic regions and for the definition of a ‘baseline’ seismicity, required for a safe future exploitation of georesource areas. In this study, we produce the first manually revised catalogue of micro‐seismicity for Co. Donegal region (Ireland), an area of about 50K M2 of on‐going deformation, aimed at localizing natural micro‐seismic events occurred between 2012 and 2015. We develop a stochastic method based on a Markov chain Monte Carlo (McMC) sampling approach to compute earthquake hypocentral location parameters. Our results indicates that micro‐seismicity is present with magnitudes lower than 2 (the highest magnitude is 2.8).The recorded seismicity is almost clustered along previously mapped NE‐SW trending, steeply dipping faults and confined within the upper crust (focal depth less than 10 km). We also recorded anthropogenic seismicity mostly related to quarries' activity in the study area.


| INTRODUC TI ON
The Co. Donegal region is located in the northwestern part of Ireland, and it is part of the NE-SW trending Caledonian fold belt of Lower Palaeozoic age.The major tectonic feature of the area is the presence of the NE-SW trending fault system, of which the Leannan Fault is the most representative example.The Leannan fault is directly in line with the Great Glen fault and the possibility that it is a north-west continuation of this structure is discussed (Pitcher et al., 1964).Ireland is a region of slow lithospheric deformation and at the present Co.Donegal is the only seismically active area in Ireland, with an average rate of 1 Mw = 2-3 event every 3-4 years.
Regions of slow lithospheric deformations are characterized by a low seismicity rate and limited accumulation and release of tectonic energy, leaving open the question if such deformation is 'diffusely' accommodated along a wide fault system (or even across the whole crust) rather than clusterized along specific faults.In particular, the sparsity of the seismic networks (common in region with low seismic risk) makes it appear, in current catalogue, that seismicity is generally diffuse, but it could also be grossly mislocated due to the small amount of seismic data available.Well-planned experiments with an appropriate density of seismic stations are required to register and locate small events.Otherwise, standard national seismic networks with a low density of seismic stations could not be able to register the target events.Precise location can be achieved through visual inspection of seismic waveforms and manual picking of P-and S-direct phases, where automatic methods should be favoured only in case of large (>10,000 events) catalogue, due to their proness in detecting false-positive cases (Münchmeyer et al., 2022).In addition to the detailed seismic monitoring through dense network, the adopted location procedure has strong impact on the location of events and on the discrimination between natural and human-induced events (also based on the hypocentral depth), therefore on the assessment of the ordinary background micro-seismicity within 1D volume.
Co. Donegal is also the site of post-orogenic radiogenic granites partly exposed along the NW Atlantic coast and potentially buried at shallow depth in the eastern part of the study region.For this reason, the area represents a target for deep (0-5 km depth) geothermal investigation (Goodman et al., 2004).Having a high quality of precisely located micro-seismicity is fundamental before the exploitation of geo-resources for two main reasons.First, seismic waves generated by local events bring information useful to reconstruct the 3D seismic properties of the Earth's crust beneath the seismic network and to better constraint the rock volume in which secondary porosity and fluid circulation may be present.Second, the definition of a 'baseline' seismicity for the area where the exploitation will be in place is necessary to keep the seismic risk related to the exploitation itself under strict control.
For these two reasons, the final aim of our work is the precise location of both human-induced and natural micro seismic events

Significance statement
This work provides the first-ever catalogue of manually re- by 12 broadband seismic stations.We compile a first manual-revised catalogue of Donegal micro-seismicity, and we integrate it with the location of the seismic events occurred in a small seismic sequence during January 2012 in the study area (see Figure 1).In order to detect the seismic events we analysed the continuous waveforms by applying a STA/LTA network coincidence trigger algorithm (Team, 2017) and using Jupyter Notebook (Executable Books Community, 2020).Then, we performed a manual picking of P and S waves first arrival times of the events detected by the trigger algorithm (Goldstein & Snoke, 2005) for a total of 35 natural events and 79 human-induced events.In order to recover as many phases/events as possible, we varied the STA/ LTA parameters several times to find the more reasonable ratio between true triggered events and fake events given by the natural seismic noise.We reviewed all the arrival times to exclude erroneous ones.We also integrated seismic data recorded by  (Lomax et al., 2000).Establish realistic uncertainties is fundamental to robustly relate retrieved earthquake locations to other geological/geophysical observations.The strength of

| DATA AND ME THODS
McMC method lies on the fact that the data uncertainties are also considered as part of the unknowns and are robustly estimated through the McMC sampling following a Hierarchical Bayes approach (Malinverno & Briggs, 2004).Moreover, a detailed velocity model of the investigated area does not exist at the moment.This aspect can be easily solved by using our method because we just need to specify the minimum and maximum values of all the priors to define the prior probability distribution (i.e. the velocities of P and S waves).Earthquake locations were represented in map view and cross-section by using GMT (Generic Mapping Tool) and its Python extension PYGMT (Wessel et al., 2019).At these stations, the Ts-Tp time of all the natural events detected lies in the range 1.5-3 s.This means that the source of all the events must be approximately at the same distance from these stations, leading to a clustered behaviour of all the natural events in the near centre of the network, as reported in 9.

F I G U R E 2
The anthropic event is shown in Figure 3a.The typical waveform of the anthropic events detected is characterized by an initial clear P-wave onset, whereas the S-wave onset is less or no visible on the traces.A relatively long coda occurs at high frequency.Two distinct low-frequency arrivals can be found in the coda of such kind of events (Figure 3b).These patterns are widely recognized in anthropic seismicity (Korrat et al., 2022) (Saadalla et al., 2023) and can be used to classify the artificial events.We also performed a comparative spectral analysis between a natural earthquake and an artificial seismic signal.Natural and human-induced seismic events have different spectral content, and this feature has been used for a correct classification in doubt cases.In Figures 4 panel   a) and b) we selected a window of 5 s of noise for both events and in panel c) and d) we selected only the blast and natural waveform to analyse their frequencies content.The spectrum of the blast events is characterized by lower frequencies content and a sharply decrease with increasing frequencies.On the contrary, the natural event is characterized by a more stable pattern of frequencies (Figure 5).Both P ( P ) and S wave uncertainties ( S ) have been well-constrained.

| RE SULTS
Their curves show a Gaussian distribution with a narrow shape.The average value for P is 1.2, for S is 1.4, both with standard deviation lower than 1 (Figure 7g,h).Finally, the mean value for the V P ∕ V S ratio is between 1.68 and 1.70 with a standard deviation of less than 0.1.
Figure 8 shows the seismicity recorded during the small seismic sequence occurred in Fanad during January 2021.Seismic of detected events to be located was 203: 141 were classified as blast, considering the features described in section 'Methods', 58 as natural events and 4 regional events that have not been considered in the location procedure.After a first location, the events with uncertainties on X and Y parameters >1 Km were eliminated from the catalogue.The high uncertainties of these events may be caused by errors in the detection of real events or uncertainty during the manual picking.For this reason, the total number of events in the final catalogue and shown in the maps 9 and 10 decreased from 203 to 114: 79 are blast events and 35 are natural events.Figure 10 shows the distribution of located blast events within the Donegal area.Most of the blast events are concentrated in the proximity of active local quarries in the area.Table 1 resumes the geographical locations for the Donegal's quarries.Quarries locations have been used to cross-validate the location of the recorded blast events.Comparing the position of the quarries and the clusters of localized blast events, the correlation between the two is clearly visible.This means that the majority of the anthropic events in Donegal are quarry blasts and they have been correctly localized by McMC procedure to their correspondent quarry (Table 2).

| Local magnitude
We calculated the local magnitude for the natural events located in Co. Donegal as reported in Table 3.The second member is the empirical calibration function, which in turn is a function of the epicentral distance (Richter, 1935).This formula applies the updated Woods-Anderson parameters (gain = 2080, natural period T0 = 0.8 s and damping constant h = 0.7).We then calculated the magnitude at each station using a calibrated local scale for Ireland, provided in the study of Grannell et al., 2018 following the formula: where A is the maximum amplitude of the earthquake S wave on a  2).We then averaged the individual ML values for each station.However any individual ML measurement that was more than two standard deviations from the mean ML value was discarded, and then the average ML was recalculated.We used this criterion for both approaches.The following Table reports the coefficients of the seismic stations used in this study and applied to Equations ( 1) and ( 2):

Time from OT (s)
the magnitude values reported in Table 3.In some cases, the magnitudes calculated by Equation ( 2) are still slightly lower than the 'MLRI35' values.However, the majority of them are below 2 with small standard deviations and this result is in agreement with the magnitude values expected for Co. Donegal.

| DISCUSS ION
The natural seismicity examined in this study is consistent with the tectonic setting of the region and, in particular, with the mapped NE-SW fault system (Figure 9a).The lower panel shows earthquakes plotted that most events occur within the radiogenic granites buried at shallow depth.The presence of Donegal granites might have created a vulnerable zone in the Lennan Fault System, potentially contributing to its current seismic activity (Hensen et al., 2019).
The natural seismicity is gathered between 55.12 and 55.22 N in latitude and between −7.74 and − 7.60 E in longitude (580-590 Km on the X coordinate and 6110-6120 Km on the Y coordinate) (Figure 9a), both for more recent and sparse seismic events and for a seismic sequence occurred in the area.This fact indicates that tectonic stress, in region of slow tectonic deformation, is still accommodated along pre-existent fault systems which behave as weak zone in the upper crust.Even if the (suggested) sources for such stress are located either far away (i.e.
mid-Atlantic ridge, Hensen et al., 2019) or at the surface (i.e. post-glacial rebound Wu et al., 1999), seismic events pop-up in precisely defined zones, characterized by fault systems mapped at the surface and most probably inherited from previous geodynamic processes.Here, the process that led the development of the Great Glen fault, for which our study area is the southward continuation, was probably active mainly prior to the Upper Carboniferous (300 Ma, Kennedy, 1946).Nevertheless, present-day seismic activity still clusterizes along such weak zones.1. Clusterized micro-seismicity is present with a magnitude generally lower than Mw = 1.0-2.0(i.e.not felt by population) 2. Such events are localized within previously mapped fault systems in Co. Donegal.
3. Anthropogenic seismicity is widely recorded through the entire study area, and it is mostly related to quarries' activity in the studied area.
Through this work we produced the first manually revised catalogue of Donegal micro-seismicity and, more widely, of the entire Ireland.This kind of catalogues is fundamental for both getting new insights into the geodynamic processes on-going in regions of slow lithospheric deformation, and for future exploitation of geo-resources.

ACK N O WLE D G E M ENTS
We would like to thank the Dublin Institute for Advanced Studies for sharing the seismic data.In particular, we are in debt with James Grannel for information about Donegal quarries, the discussion on the magnitude and event catalogue and the station cor-

A. INTRODUCTION ON SEISMIC EVENT LOCATION
Here, we define a seismic event as a sudden release of elastic energy below the topographic surface, which produces seismic waves that propagate across the rock volume.Locating seismic events is a fundamental task for both monitoring the subsurface and studying the Earth's structure.A network of seismic stations at the surface (but also within the rock volume) can be used to locate such events, identifying the P and S waves arriving at the seismic stations, and measuring their arrival times t o P and t o S .Locating seismic events is a standard geophysical inverse problem (Tarantola, 2005), where seismologists indirectly measure the 3D position of a physical process (the seismic energy release) through the measure of related quantities (the arrival times of the seismic waves).Many different methodologies have been developed for solving such inverse problem, either based on linearized methodologies (e.g.Billings et al., 1994;Waldhauser & Ellsworth, 2000) or stochastic approaches (e.g.Lomax et al., 2009).Due to the need of modelling the propagation of the seismic waves, different workflows consider a 1D (e.g.Lahr, 1989) or 3D (e.g.Theunissen et al., 2017) Earth's model.
In this study, we develop a stochastic method based on a Markov chain Monte Carlo (McMC) sampling of parameters (Mosegaard & Tarantola, 1995).In stochastic methods, many randomly generated numerical representations of the physical model (i.e. the so-called models) are collected and, for each model, the predicted observations are generated (this operation is called forward modelling and the predictions are called synthetics).Through the comparison of synthetics and observations each single model is evaluated.Here, our physical model considers the existence of a homogeneous halfspace which represents the rock volume and the seismic event can occur in any place of the rock volume itself with the same probability.
To avoid bias given by mis-estimated uncertainties, we strictly follow a Hierarchical Bayes methodology Malinverno and Briggs (2004).
The McMC sampling is guided by a Metropolis rule (Metropolis et al., 1953) and, thus, sampled solutions are used to make Bayesian inferences on the investigated parameters (e.g.estimating mean and standard deviation of each parameter, but also correlation between different parameters).

B. BACKGROUND ON BAYESIAN INFERENCES, MARKOV CHAIN MONTE CARLO AND HIERARCHICAL BAYES
Bayesian inferences is based on Bayes' theorem (Bayes, 1763) where m is a model (i.e. a point in the model space), d is the vector of the observations, P(m| d) is the posterior probability distribution (see below), P(m) is the prior probability distribution (see Section E), and L(m) is the likelihood value (see Section D).Bayes' theorem is a probabilistic formula which relates a new state of knowledge (so-called vised micro-seismicity of Co.Donegal, a region of ongoing slow lithospheric deformation.Through our work, we tried to give an answer, with a newly developed algorithm, to the open question whether the deformation in these kinds of regions is "diffusely" accommodated within the whole crust rather than clusterized along specific faults.F I G U R E 1 (a) Seismic network deployed in Donegal in the framework of the SIM-CRUST project.12 broadband seismic stations have been deployed for about 2.5 years of continuous recording.Two stations belong to a previous temporary seismic network, ISLE/ISUME projects.Seismic data have been integrated with data from the permanent seismic station IDGL, Irish National Seismic Network.Interstation distance is less than 10 km in the central area.Sources of anthropic noise are present in the area, lowering the quality of the seismic signal recorded.Fault traces in the region are reported as red lines.Colours indicate different terranes.From Agostinetti and Licciardi (2015), courtesy of Andrea Licciardi.(b) Historical and Instrumental seismicity recorded in Donegal region.Seismic stations D34, UFAN and IDGL used to locate the seismic sequence of 2012.[Colour figure can be viewed at wileyonlinelibrary.com] of Co. Donegal, which could give new insights in the tectonic deformation mechanisms and help to find out granite volumes characterized by micro-fractures suitable for fluid circulation.The seismic data provided in this paper have been collected thanks to the SIM-CRUST project: 'Seismic imaging and monitoring of the upper crust: exploring the potential of low-entalphy geothermal resources of Ireland' (sim-crust.dias.ie).The project comprises the development of two high-density (inter-station distance <10 km) seismic networks to explore the Dublin basin and the Donegal granite region, case of study of our research.Additional seismic data, for a small seismic sequence occurred few months before the deployment of SIM-CRUST stations have been analysed for a more complete picture of the seismicity in the area.These data have been provided by a temporary project (Readman & O'Reilly, 2022) operated by the University College Dublin.Our study focuses on the detection and location procedure through a Markov chain Monte Carlo approach of both natural and human-induced seismicity recorded from August 2012 to July 2015 Within the SIM-CRUST project a seismic network of 12 stations has been installed and maintained in the Co.Donegal from August 2012 to June 2015 (1.a) with the inter-station distance between 5 and 20 km.All stations were equipped with broadband seismometers (Guralp CMG-40 T), with a flat response between 50 Hz and 60s.Seven Earth Data PR6-24 Portable Field Recorders have been used, together with four Guralp CMG-EAM Flexible data acquisition modules.Due to the limited space available for storing the digitizer, a Nanometrics Taurus has been installed at station DL13.All stations (except for DL21 and DL13) have been installed within a building allowing for continuous power supply.Continuous waveform data have been archived as 1-day MSEED files at DIAS (Dublin Institute for advanced Studies).
other three temporary seismic stations installed during a small seismic sequence occurred in January 2012 near the Fanad peninsula (Möllhoff & Bean, 2016).The seismic sequence was triggered by a Mw 2.5 event occurred next to Milford and has been felt by the local population (https:// www.irish times.com/ news/ doneg al-earth quake s-homes -1.694869).Finally, a total of 114 earthquakes were located using a hierarchical Bayes Markov chain Monte Carlo (McMC) algorithm.The used McMC algorithm has been developed on purpose for this study and described in detail in Appendix A. The Markov chain Monte Carlo (McMC) method allows us to estimate the realistic uncertainties on the investigated parameters Vertical waveform plot for the natural event # 31 with waveforms sorted by arrival times at the stations.P and S onsets are clearly visible.[Colour figure can be viewed at wileyonlinelibrary.com]The seismic network recorded both natural earthquakes and quarry blast events.In the following paragraph, we report the processing and results of two representative events: natural event # 31 and anthropic event # 15. Figure 2 shows the plot for the natural event # 31 after the picking phase sorted by arrival times at the stations.The event is well recorded on all the seismic stations of the network, where the onset of P and S waves are clearly recognizable from the seismic noise of the traces.Stations DL31, D34, DL10 and DL13 lie on the external perimeter of the seismic network, mostly covering all azimuth directions.

Figure 6
Figure6shows the sampled locations of the ML = 1.5 natural event # 31 in map view (panels a and c) and cross-section (panel b).The pink Figures 9 and 10 show the final maps of the natural and human-induced events that have been located.The initial total number

Figure 9
Figure9shows the located natural events within the Donegal Granite region.The natural seismicity is gathered between 55.12 and 55.22 N and between −7.59 and − 7.74 E (580-590 Km on the X coordinate and 6110-6120 Km on the Y coordinate) (Figure9a).In the lower panel (panel b), the events are shown in cross-section along the track line A-B.The majority of the events is clearly aligned, showing a trend towards SE with a high dip angle.
Due to the scarceness of local seismicity and the consequent difficult calibration of existing magnitude values with respect to ours, we decided to use two different approaches for magnitude comparison.We first calculated the F I G U R E 7 Gaussian distributions of (a) longitude [Km] (b) latitude [Km] (c) depth [Km] (d) origin time [s] (e) P-wave velocity [m/s] (f) S-wave velocity [m/s] (g) uncertainty related to the P-wave picking (h) uncertainty related to the S-wave picking (i) P-and S-wave velocity ratio.[Colour figure can be viewed at wileyonlinelibrary.com] individual magnitude at each station using the following formula by Richter (1935): being A the measured ML Wood-Anderson amplitude in millimetres.

Wood-
Anderson filtered trace, S is a station correction coefficient and Log(A0) is a distance-dependent correction term accounting for geometrical spreading and anelastic attenuation resulting from this equation: with a = 1.095717, b = 0.001552 and c = −2.028571and R is the hypocentral distance.We added a stations correction term for the individual station used by the Irish Seismic Network and provided in the study of Grannell et al., 2018 for both Formulas (1) and ( ural seismic events calculated by the two equations, their standard deviations and the corresponding number of stations used to reach the final ML value.We called 'MLRI35' the magnitudes calculated by the Equation (1) and 'ML_Donegal' the magnitudes resulting from Equation (2).After a first magnitudes calculation, the values of ML_Donegal were substantially different from the mean MLRI35 values, with higher standard deviations.We then discarded the seismic stations D32 and D34 from the ML_ Donegal calculation because they have lower S/N ratio and are the most distant from the epicentres.Finally, we come up with (1) ML = log 10 (A) − log 10 (A0)(2) ML = Log(A) − Log(A0) + S Log(A0) = − a * log(R) − b * R − cF I G U R E 8 Natural event belonging to the 2012 sequence registered at the UFAN seismic station.(a) P-wave onset (red bar) and S-wave onset (blue bar).(b) Zoom on the first P-wave arrivals for all the events.P-wave onset is clearly detectable on all events.[Colour figure can be viewed at wileyonlinelibrary.com] in cross-section.The majority of events shown in Figure occur along steeply SE dipping feature.The upward continuation of the represented tectonic feature could reach the Glen fault, a NE-SW trending sinistral strike-slip deformation widespread throughout the Highlands of Scotland and NW Ireland, that has been interpreted as a response to the subduction of Iapetus and continental collision (Kirkland et al., 2008).The southwestern continuation of the Great Glen Fault zone is marked in NW Ireland by the Leannan Fault (Figure 9) and the presumed fault marked in the cross-section Figure 9 (panel b) may be part of its splay.The Figure also shows that most earthquakes are located at depths of less than 10 km, with a few events occurring at greater depth.The depth distribution of the earthquakes suggests F I G U R E 9 Natural seismic events plotted on the X-Y map and on the Z coordinate.Panel (a) shows the seismic stations and the track of the NW-SE profile.Panel (b) shows the natural events projected on depth and the presence of the presumed fault.[Colour figure can be viewed at wileyonlinelibrary.com]

F
Blast events plotted on the X-Y plane.Clusters of events are related to local active quarries.[Colour figure can be viewed at wileyonlinelibrary.com]TA B L E 1 Donegal quarries' names and coordinates referred to the clusters of blast represented in Figure 10.[Colour observe a relevant number of seismic sources throughout the study area, not consistent with 'standard' natural seismic events (i.e.not consistent with shear events, with unclear or absent S-wave arrival and associated long-period coda waves).Such events have been associated to local human activities, mainly explosions in quarries.Such events form small local clusters in precise locations.Being the positions of the quarries know within a certain precision, such seismic sources could be in future used as 'active sources' for further crustal investigations.5 | CON CLUS IONSSeismic traces registered from August 2012 to June 2015 by stations of the Seismic network deployed in Donegal in the framework of the SIM-CRUST project and by two stations belonging to a previous temporary seismic network, ISLE/ISUME projects have been analysed to detect natural and anthropic seismicity, in a target area for future geothermal exploitation.The position of the located events indicate that: TA B L E 3 Magnitude values at each stations calculated by using Equation (1) ('MLRI35') and Equation (2) ('ML_Donegal'), their standard deviations and the corresponding number of station used to reach the final averaged ML value.

A
rection coefficients, Martin Möllhoff for the maintenance of the seismic stations and the data archives.The seismic data analysed in this thesis have been collected during the SFI funded project SIM-CRUST Grant Number 11/SIRG/E2174 (https:// sim-crust.dias.ie/ ).NPA thanks John O'Raw, Danny Mc Fadden, David Mc Gloin, Michelle Johnstone, Henry and Martin Callaghan, Henry McKinney, Gerard Mc Daid, Michael Tinney, Derek Flanagan and Dr Alessandro Amato for the deployment of the seismic network.Andrea Licciardi and Andrew Schaeffer helped in managing the seismic network in Donegal.Preliminary analysis of the quarry blasts recorded in Donegal has been done by Maite Zabaltza during her internship at DIAS. S U PP O RTI N G I N FO R M ATI O N Additional supporting information can be found online in the Supporting Information section at the end of this article.this article: Riva, F., Agostinetti, N. P., Marzorati, S., & Horan, C. (2024).The micro-seismicity of Co. Donegal (Ireland): Defining baseline seismicity in a region of slow lithospheric deformation.Terra Nova, 36, 62-76.https://doi.org/10.1111/ter.12691 (3)P(m| d) = kP(m)L(m) ,

Table 3
reports the final magnitude values for the 35 nat-