A Multi-Master Event Approach to Earthquake Relocation: Insights from the West Bohemia Swarm Zone (Czech Republic, 2000–2025)

A Multi-Master Event Approach to Earthquake Relocation: Insights from the West Bohemia Swarm Zone (Czech Republic, 2000–2025) | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Multi-Master Event Approach to Earthquake Relocation: Insights from the West Bohemia Swarm Zone (Czech Republic, 2000–2025) Diana Konrádová, Jana Doubravová, Bohuslav Růžek, Jan Burjánek This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7800475/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 9 You are reading this latest preprint version Abstract The intraplate West Bohemia region represents one of the most prominent areas of recurrent earthquake swarm activity in Central Europe, with ongoing seismicity observed through 2025. We relocated this activity using the new multi-master event (MME) method and compared the results with GrowClust (GC) relative locations. The MME approach provides precise relative locations even for small events and can be applied efficiently to long-term, large catalogs. Its key advantage lies in the ability to incrementally incorporate new events without reprocessing the entire dataset, making it suitable for continuous and real-time monitoring. The comparison demonstrates that the newly developed MME approach produces results that are consistent with those of the established GC method. Both methods produce a consistent image of the southern and northern parts of the swarm, while the central segment of the active Nový Kostel seismic zone shows the strongest methodological differences. The MME results are sensitive to the quality of master-event locations, occasionally leading to scattered clusters where input locations are less reliable. In contrast, GC systematically sharpens the seismicity distribution and produces more compact structures. Overall, these results provide the first comprehensive view of the relocated seismicity in the West Bohemia region, extended through 2025, and reveal characteristic migration trends in the activity. Relative location methods West Bohemia/Vogtland Master-event method GrowClust method Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Article Highlights A new multi-master event method improves earthquake relocation for large and long-term seismic catalogs. The study provides the first detailed view of seismic activity in West Bohemia, updated through 2025. Comparison with GrowClust shows consistent fault structures but highlights key methodological differences. 1 Introduction Location of earthquakes is a primary step of any seismological studies. Its accuracy and quality is crucial for next processing from magnitude evaluation to focal mechanism estimation. Earthquake relative location is used to refine the positions of seismic events beyond the accuracy of routine catalogs. While absolute locations are often affected by uncertainties in the velocity model and errors of onset picking, relative relocation techniques exploit differential arrival times between nearby events to achieve much higher spatial resolution. This is essential for resolving fine-scale fault structures, seismic clusters, detecting migrating fluid fronts along active faults, and understanding the underlying seismogenic processes (e.g., Got et al., 1994 ; Rubin et al., 1998 ; Waldhauser & Ellsworth, 2000 ; Shelly et al., 2006 ). The West Bohemia/Vogtland region is well known for recurrent earthquake swarms, which are generally interpreted as a result of fluid-driven processes at mid-crustal depths interacting with local tectonics (Horálek & Fischer, 2008 ). The study area, seismic network (WEBNET), and major tectonic structures are shown in Fig. 1 . Studies applying relative locations have played a crucial role in the understanding of these swarms. For example, Fischer and Horálek ( 2000 ) used master-event method to refine the locations of foci of 1997 swarm. Bouchaala et al. ( 2013 ) applied both the master-event and double-difference (hypoDD; Waldhauser & Ellsworth, 2000 ) relative location methods, using catalog picks as well as cross-correlation differential times of P- and S-waves, to the 2008 West Bohemia swarm in order to refine the cluster structures and to reveal the detailed geometry of the active fault zone. Jakoubková et al. ( 2018 ) refined the locations of the 2014 mainshock–aftershock sequence and several previous swarms (1997, 2000, 2008, 2011) using the double-difference relative location method (HypoDD), and compared them with typical swarm-type seismicity in the area, highlighting key differences in their spatial and temporal evolution. In this study, we develop and test an alternative relative location approach, the multi-master event (MME) method. Standard relocation codes such as hypoDD are most effective when applied to dense localized earthquake clusters, but it becomes computationally expensive for very large catalogs and is not well suited for incremental processing when new events are continuously added. The MME method was designed here to address these limitations: instead of requiring manual selection of master events typically needed for master event relative locations, it automatically identifies suitable master events according to predefined criteria (e.g., number of high-quality phase picks). This approach saves considerable processing time and takes advantage of manually picked phases in the WEBNET catalog, while still allowing user-defined master events if needed. As a result, the method is well suited for large earthquake catalogs with events of varying data quality. For instance, well-recorded events with many clear phase picks as well as small earthquakes with only a few, noisier arrivals, while also enabling incremental, near real-time catalog updates. One of our goals is to assess its performance in the West Bohemia region and compare it to results from GrowClust, a modern and widely used double-difference relocation tool which was developed to reduce the main limitations of hypoDD (Trugman & Shearer, 2017 ), offering improved computational efficiency for large datasets and greater robustness against outliers. 2 Methods All relative location methods are based on the assumption that if two earthquakes occur very close to each other compared to their distance from the recording stations, their seismic waves travel along nearly identical paths. In this case, the effects of velocity model inaccuracies along the path can be largely canceled out, and the relative arrival time difference can be attributed solely to the relative offset between the event locations. This relative location can be resolved with much higher precision than their absolute positions. However, uncertainties in phase picking remain. These can be reduced by calculating differential times from waveform cross-correlation instead of relying solely on catalog picks (e.g., Got et al., 1994 ; Waldhauser & Ellsworth, 2000 ). The concept of determining relative event positions with respect to a well-located reference (master) event originates from the joint hypocenter determination approach introduced by Douglas ( 1967 ) and was later successfully applied to larger seismically active areas, such as in northern California (Stoddard & Woods, 1990 ). These studies demonstrated the method’s ability to enhance location precision and reveal fine-scale deformation structures. Nowadays, the most widely used method is the double-difference relocation algorithm (Waldhauser & Ellsworth, 2000 ), which has proven highly effective for resolving detailed fault-zone structures and relative event distributions (Richards et al., 2006 ). The double-difference method extends the master-event concept by treating all events symmetrically as both “masters” and “targets”. This means that the relative locations of nearby earthquakes are adjusted so that observed differential travel times between pairs of events match those predicted by the velocity model. The first widely distributed code implementing this approach was hypoDD, and a more recent implementation GrowClust (Trugman & Shearer, 2017 ) introduces hierarchical clustering to improve computational efficiency for large catalogs. Both hypoDD and GrowClust can make use of differential times obtained from catalog picks or from waveform cross-correlation (XC). In hypoDD, XC can be used to improve the precision of catalog picks, but it is not a requirement, the code can be run on catalog data alone. XC significantly reduces errors caused by inaccurate onset picking, because correlation lags provide more precise relative timing between similar events. However, Bachura and Fischer ( 2019 ) showed that the use of XC is not always beneficial. For events with large magnitude differences or dissimilar waveforms, the correlation may align peak amplitudes instead of the true first arrivals. This effect can lead to an apparent shift of the largest events away from their absolute locations. In contrast, using carefully selected manual picks can be still a reasonable option (Bachura & Fischer, 2019 ). 2.1 Multi-Master Event Relative Location Method To improve the precision of earthquake locations and enable processing of large seismic datasets, we developed and applied a custom multi-master event (MME) relocation approach. This method is based on differential travel times relative to one of a list of well-constrained master events and supports incremental processing, allowing new events to be added without the need to re-run the entire relocation procedure. This makes the method particularly suitable for operational seismic monitoring and long-term studies in seismically active regions. A schematic flowchart of the MME relative relocation procedure, summarizing the inputs, iterative processing steps, and outputs, is shown in Fig. 2 . The algorithm works iteratively. In each iteration, a master event is selected from the catalog according to predefined criteria. A key condition is the number of available arrival-time picks. The master event should have as many readings as possible to ensure that associated events (targets) can be successfully matched. Based on several preliminary tests, we found that requiring at least 20 high-quality P- and/or S-phase picks provided stable and reliable relocation results for the data and station configuration used in this study. A high number of picks typically corresponds to a good signal-to-noise ratio, which assures low onset picking error. A subset of events in the vicinity of the master event is selected according to a predefined maximum hypocentral distance. In this study we used the limit of 1 km. For each of these events, the classical master-event location problem is solved (e.g., Stoddard & Woods, 1990 ; Bouchaala et al., 2013 ). To determine the optimal hypocenter for each event, the algorithm minimizes the residuals between observed and synthetic differential travel times relative to the selected master event. For a given target event and master, the differential travel time at each station i is defined as: $$\:\delta\:{t}_{i}={\left({t}_{i}^{e}-{t}_{i}^{m}\right)}_{obs}-{\left({t}_{i}^{e}-{t}_{i}^{m}\right)}_{syn}$$ where t i e and t i m are the arrival times for the event and the master, respectively. The observed differential times are computed directly from the phase picks, while the theoretical differential times are calculated using a 1-D constant velocity layered model with the ray-theory approximation. The algorithm then searches for the hypocenter location and origin time (x,y,z,t) of the event that minimizes the sum of squared residuals between observed and theoretical differential times: $$\:min{\sum\:}_{i}{\left(\delta\:{t}_{i}\right)}^{2}$$ The optimization in our MME implementation is performed using a non-linear least-squares approach, which iteratively adjusts the event’s coordinates to minimize the overall misfit between observed and theoretical differential travel times (Tarantola, 2005 ). The inversion is constrained to a three-dimensional volume around the initial location of the event, with boundaries defined by user-specified limits of 1 km horizontally and 2 km vertically, as chosen in this study. After the relocation, the event is saved and removed from the set, and the algorithm looks for remaining events in the catalog that have not been relocated yet. If the remaining events cannot be linked to the current master, a new master event is selected from the remaining set and the relocation process is repeated. This cycle continues until no more events satisfy the criteria for master selection. The inputs and outputs of the algorithm are summarized in Fig. 2 . The algorithm can either continue from a previous relocation using an existing set of master events or start from scratch, choosing the first master event automatically. We tested the sequential processing option to simulate incremental catalog updates, and the procedure is described in detail in section Results. 2.2 GrowClust Relocation Method For comparison, we applied the GrowClust relative relocation algorithm, which uses a hierarchical clustering approach combined with double-difference inversion to improve the relative locations of earthquakes (Trugman and Shearer, 2017 ). GrowClust identifies clusters of similar events based on waveform similarity quantified by cross-correlation, and relocates them by minimizing differential travel-time residuals. The relocation is based on the double-difference method, which reduces the influence of velocity model errors by focusing on differential arrival times between event pairs recorded at common stations: $$\:\delta\:{t}_{kj}={\left({t}_{k}-{t}_{j}\right)}_{obs}-{\left({t}_{k}-{t}_{j}\right)}_{syn}={G}_{k}\cdot\:\delta\:{x}_{k}-{G}_{j}\cdot\:\delta\:{x}_{j}$$ where t k and t j are the arrival times of events k and j . δt kj represents the observed minus theoretical differential arrival time between the two events. δx k and δx j denote the corresponding location correction vectors, and G k and G j are the partial derivatives of travel time with respect to event coordinates (i.e., travel time sensitivity to the small changes in the event location). The system of equations is solved iteratively to minimize the residuals across all available event pairs. The GrowClust algorithm uses a hierarchical approach: it begins by identifying tightly clustered event groups and relocates them, progressively adding larger or more distant clusters in later iterations. This strategy improves scalability and allows GrowClust to process large datasets more efficiently. The relocation considers both P- and S-wave picks and incorporates cross-correlation-derived differential times to improve precision. The details can be found in Trugman and Shearer ( 2017 ). 3 Data We used an earthquake catalog with manual picks of P- and S-wave onsets, recorded by the WEBNET seismic network (Institute of Geophysics, 1991; Horálek & Fischer, 2008 ), operating in the West Bohemia/Vogtland region. WEBNET is a local seismic network designed for monitoring local seismicity, and it has played a key role in documenting swarm activity and fluid-induced processes in the area (Fischer et al., 2014 ; Hainzl et al., 2012 ; Hainzl et al., 2016 ). The WEBNET network geometry (Fig. 1 ) provides dense station coverage in the vicinity of the Nový Kostel seismic zone, ensuring high-quality arrival times and good azimuthal resolution for the majority of events. The dataset includes approximately 62,000 local earthquakes recorded between January 2000 and October 2025. We selected this time period because the catalog coverage is most complete and consistent since 2000. All events have manual picks of P- and S-wave arrival times. The absolute locations were computed using NonLinLoc (Lomax et al., 2000 ) with a 1-D layered velocity model based on Málek et al. ( 2005 ), who parameterized the crust as a constant gradient velocity layers. The catalog includes events across a wide range of magnitudes M L from − 2 to 4.1. The quality of the input absolute locations was evaluated from NonLinLoc covariance matrices (see Appendix A). This analysis confirmed that horizontal and vertical uncertainties are mostly below 1 km, ensuring sufficient accuracy for subsequent relative relocation. Due to computational limitations in applying GrowClust to the full dataset, we restricted the relocation to events with M > 0, yielding a representative subset of ~ 30,000 events. In contrast, the multi-master method was applied to the entire dataset of 62,000 events, including lower-magnitude events. However, for visualization and comparison purposes, only events with M > 0 are shown in the final maps. The comparison between the two methods is based on events relocated up to June 2024, corresponding to the temporal extent of the GrowClust dataset. 3.1 Original Absolute Locations The original absolute locations for the most active Nový Kostel seismic zone are plotted in Fig. 3 . Coordinates are shown in UTM zone 33N (easting in km on the x-axis, northing in km on the y-axis). They show a vertically elongated seismic zone stretching north–south, centered around the Nový Kostel region. Despite several spatial clusters and alignments can be identified, the resulting absolute locations exhibit significant vertical and horizontal scatter, likely due to imperfection of the velocity model and observational uncertainties in arrival times. This scatter obscures finer fault-related structures and limits the ability to interpret structural segmentation from absolute locations alone. 4 Results 4.1 Multi-master Event Results The relocation results and their comparison with the absolute locations are shown in Fig. 4 and Fig. 5 . The MME relative locations are shown in Fig. 4 g-l and Fig. 5 k-t for the most active Nový Kostel seismic zone. All coordinates are given in UTM zone 33N, expressed in kilometers. In the map view (Fig. 4 g-l), seismicity is concentrated within a narrow NNW–SSE trending fault system, extending over ~ 15 km in length. Several parallel strands can be recognized, with the highest density of activity in the central segment (5566 < Y < 5572 km), while the southern (Y 5572 km) segments are less active. Overall the relocated events closely follow the main fault orientation. Figure 4 g-l and Fig. 5 k-t confirm that most seismicity is confined to a depth band between 8 and 10 km. Within this interval, events align into elongated clusters with minor lateral variations. These may reflect subtle internal segmentation of the fault zone, although some methodological effects cannot be fully excluded. In particular, the multi-master approach is sensitive to the choice of master events: if the absolute location of a selected master is biased, this bias can propagate to its associated cluster. Moreover, the imposed 1 km association threshold and the variable number of available phase picks per master may produce apparent breaks or scatter in the relocated structures. Above 6.5 km and below 10.5 km only a sparse seismicity is observed. Depth-sliced projections of the multi-master event relative locations (Fig. 4 g-l) further illustrate the internal segmentation of the Nový Kostel seismic zone. Even within narrow 0.5 km depth intervals, the relocated events consistently follow the NNW–SSE fault orientation. Seismicity is not homogeneously distributed but forms elongated clusters and locally parallel strands, with the highest density concentrated between 8 and 9.5 km. Some depth levels (Fig. 4 h-j) are dominated by compact, well-populated clusters, while others (Fig. 4 l) appear more fragmented and diffuse. The complementary E–W cross-sections, Fig. 5 k-t, emphasize the vertical confinement of the activity. Most events occur between ~ 8 and 10 km, where they form vertically elongated clusters with slight lateral variations. These may reflect internal segmentation of the fault zone, although part of the pattern may also result from the relocation procedure and master-event selection (see later the Discussion section for details). Toward the margins (e.g., panels k-m at shallower depths and s-t at greater depths in Fig. 5 ), seismicity becomes more diffuse, whereas above 7.5 km and below 11 km only the sparse activity is observed. The central segment (Y = 5566–5572 km) is characterized by several subparallel strands. At depths around ~ 9 km, the main structure bends eastward, and the relocated events form clusters separated by ~ 400–500 m. These representative slices highlight the overall structural coherence of the relocated swarm, yet reveal that some internal complexity remains. The complete set of depth slices is available in the electronic supplement (Figure S2), and the vertical cross-sections are presented in Figures S3a and S3b. The MME method, enabled the relocation of nearly all events in the dataset. Because the MME method follows an iterative scheme, its results can be sensitive to the initial conditions, in particular, to the order of events in the catalog. To test the possible sensitivity to input order, we processed the catalog in three ways: 1) sequentially from the oldest to the newest events; 2) in reverse order from the newest to the oldest; 3) randomly shuffled events in the catalog. In all cases, the resulting catalogs were consistent, with differences below the resolution of the mapping scale (Figures S4 and S5 in the electronic supplement). 4.2 GrowClust results The GrowClust (GC) relocations are shown in Fig. 4 m-r and Fig. 5 u-dd. Compared to the diffuse absolute locations, GC relocations reveal a strongly focused seismicity pattern. By exploiting waveform similarity and hierarchical clustering, the method produces narrow, sharply defined strands and a dense central core of activity at depths between 8 and 11 km, where the majority of swarm earthquakes are concentrated. In both map and cross-sectional views, the relocated events align into thin, fault-parallel strands that are in places only a few hundred meters wide, demonstrating GC’s ability to extract fine-scale structural details from the dataset. Clusters that previously appeared as broad clouds in the absolute locations (Fig. 3 , Fig. 4 a-f and Fig. 5 a-j) are transformed into elongated, coherent segments. The depth-sliced map views (Fig. 4 m-r) and E–W cross-sections (Fig. 5 u-dd) confirm this focusing: seismicity is concentrated between 8 and 10 km, forming vertically continuous and internally coherent structures, while only sparse activity is observed above 7 km and below 11 km. These results emphasize that GC captures the central fault zone with high resolution but suppresses more diffuse or peripheral seismicity. Considering the individual segments, the southern part (Y 5572 km) shows a robust and coherent structure, with seismicity forming a narrow, well-aligned fault-parallel band. The central part (5566 < Y < 5572 km) is the most complex, with multiple overlapping clusters, branching features, and local bends in the seismicity distribution. GC relative locations reveal several compact and continuous vertical planes, with seismicity bending slightly eastward at depths of 9 km. Within the 8.5–9.5 km interval, two closely spaced subparallel strands can be recognized, separated by 200–250 m. Overall, the GC relative locations provide a sharply focused image of the Nový Kostel seismic zone, characterized by compact fault strands and continuous vertical planes. 4.3 Incremental Relocation Tests simulating for real-time operation To evaluate the capability of the multi-master method for incremental relocation, we designed a sequential test simulating an operational scenario. The relocation process is carried out in two steps: Initial relocation: In the first step, the algorithm starts with no predefined master list. Master events are selected automatically from the whole available catalog (in this case 2000–2014) as those with the highest number of picks, i.e., the events that best satisfy the predefined criteria. Incremental update: In the second steps, the master events from the initial run are used as input for processing a new daily batch of events. The existing master events are retained and new events are relocated relative to them. If no suitable master is available for a given new event, the event is either selected as a new master event (if it meets the master event criteria) or left temporarily not relocated, awaiting the occurrence of a new master to which it can be linked. The results show that sequential relocation yields comparable spatial resolution to a single-pass relocation, with improved scalability and the ability to integrate new data without re-running the entire process. The key difference between the two approaches lies in how master events are selected. In batch relocation (which still runs iteratively), the best possible master event is always selected from the pool of not-yet-relocated events, the one with the highest number of phase readings. In contrast, the sequential approach selects any event that meets the master criteria at the moment it appears in the data stream. This better reflects real-time processing, where events are relocated progressively as they become available. This test demonstrates the method’s flexibility for handling dynamic catalogs and its suitability for long-term seismic monitoring. 4.4 Relocation of the entire seismic activity from 2000 to 2025 using the MME method We relocated the complete seismicity recorded between 2000 and 2025 using the Multiple-Event relocation (MME) method to obtain an up-to-date and internally consistent catalog. As the WEBNET seismic network configuration was stabilized in 2000, relocation results are shown only for the activity recorded since 2000. Seismic activity up to 2018 has been documented in detail in earlier studies (e.g., Fischer & Horálek, 2003 ; Fischer et al., 2010 ; Vavryčuk, 2011 ; Hainzl et al., 2012 ; Bachura et al., 2021 ), Here we focus primarily on the most recent activity, with locations from 2018 onward reported for the first time, except for the 2024 Klingenthal swarm already described by Büyükakpınar et al. ( 2025 ). The complete set of relocated seismicity between 2000 and 30 September 2025 is shown in Fig. 6 . Events are displayed in 2 km depth intervals to highlight the vertical distribution of activity. All relative locations were obtained using the multi-master event (MME) relocation approach. The Nový Kostel seismic zone remains the dominant feature, but its activity shows a gradual spatial expansion, most prominently northward, and southward, now covering a total length of ~ 20 km. The most recent swarms (red to brown colors) highlight this migration trend. In 2024, a pronounced swarm occurred near Klingenthal (station BUBD) analyzed by Büyükakpınar et al. ( 2025 ), shortly followed by activity near Františkovy Lázně. The Klingenthal swarm follows depth patterns similar to those in the main zone, with smaller clusters in 7–9 km (Fig. 6 b) but the most events and the strongest ones are concentrated in 9–11 km (Fig. 6 c). Notably, all earthquakes with magnitudes M L ≥ 2.5 in the area of Nový Kostel seismic zone are confined to the 9–11 km depth range. By contrast, the Františkovy Lázně activity is distinct in its depth distribution: it is confined mainly to 11–13 km (Fig. 6 d), making it deeper than the main Nový Kostel zone. Moreover, several earthquakes with magnitudes exceeding M L 2.5 occurred there, which contrasts with the main zone. Additional deep activity has emerged near Luby (station LBC), at depths of 13 km and deeper (Fig. 6 e). These events lie outside the main Nový Kostel seismic zone and suggest localized activity in an area that was previously less seismically active. The occurrence of these new deep clusters, together with the Františkovy Lázně swarm, is noteworthy given that both regions coincide with well-documented CO₂ emanations and fluid discharge in the Cheb Basin and its surroundings (Weinlich et al., 1999 ; Geissler et al., 2005 ; Babuška et al., 2016 ). This supports the widely discussed interpretation that mantle-derived fluids play a crucial role in triggering and sustaining earthquake swarms in western Bohemia/Vogtland (e.g., Weinlich et al., 1998 ; Bräuer et al., 2003 ). In the shallowest interval 0–7 km, Fig. 6 a, seismicity is comparatively sparse but present in the Nový Kostel zone, mainly in its northern part. Outside the main zone, several scattered shallow clusters are observed, many of which are related to quarry blasts rather than tectonic activity. Overall, the comprehensive dataset reveals that while the Nový Kostel zone remains persistently active, the recent swarms in Klingenthal and Františkovy Lázně, together with deep events near Luby, indicate that seismic activity now also occurs at greater depths and in regions outside the main zone. These observations point to a continuing north-south migration and an increasing diversification of swarm activity in the western Bohemia/Vogtland region. 5 Discussion The results demonstrate that the multi-master method (MME) provides precise relative locations even for small events, while remaining simple and computationally feasible for long-term and high-volume catalogs. Its ability to incrementally incorporate new events without the need to reprocess the entire dataset makes it particularly suitable for operational monitoring and real-time updating of seismicity patterns. At the same time, the performance of the MME approach is strongly dependent on the quality of the chosen master events. Inaccuracies in the absolute locations used as input can propagate through the relocation process, leading to biased or scattered clusters. This limitation contrasts with the GrowClust (GC) method, which systematically sharpens the seismicity distribution and often produces clearer and more compact fault-parallel structures. However, it remains an open question whether this increased sharpness necessarily reflects the true geometry, or whether it partly results from the intrinsic smoothing tendencies of the algorithm. In some cases, MME relative locations preserve variability that may be methodological artifacts, but which might also capture small-scale structural complexity that GC suppresses. Since the WEBNET catalog already provides high-quality, manually picked P- and S-phases, we did not apply waveform cross-correlation in the MME relative locations. While waveform XC can further improve relative timing, its computation is time-consuming, and for monitoring applications the ability to obtain rapid and consistent solutions is often more important. The differences between the two methods should therefore be interpreted with caution. The contrasting images are likely methodological rather than reflecting genuine structural segmentation. In the case of the MME approach, the relocation is sensitive to the absolute position and quality of the selected master events. If a master is poorly constrained, this bias propagates to its associated cluster. In addition, the imposed 1 km association threshold can artificially separate events into multiple strands, and differences in the number and quality of picks per master can lead to uneven relocation precision. These factors can produce apparent scatter or fragmented clusters that may not reflect true fault geometry. For GrowClust, on the other hand, the use of waveform similarity and hierarchical clustering tends to emphasize coherence and compactness. While this improves the sharpness of fault-parallel structures, it also introduces an intrinsic smoothing effect: events with weaker correlations or located at the margins of the network may be pulled toward the main clusters, reducing the visibility of diffuse or secondary features. Thus, the compact strands in GC images may partly reflect algorithmic focusing rather than genuine absence of complexity. Taken together, these methodological effects explain why the central segment (5566 < Y < 5572 km) appears more scattered in MME relative locations and more sharply defined in GC results. The quantitative comparison of residuals and event shifts further supports this interpretation, confirming that both methods provide stable solutions with small residuals and relative locations mostly below 1 km (Appendix B). To further illustrate the relocation process, events were color-coded according to their assigned master event (Fig. 7 ). Each color represents one master event and its associated group of relocated events. The map-view distribution shows that individual master groups overlap slightly but together form a continuous NNW–SSE-trending structure that coincides with the main seismic zone. Figure 7 shows that the automated master-event selection generally produces spatially coherent clusters within the 6.0–10.5 km depth range. Between 8.0 and 10.0 km, different masters capture compact but still distinguishable clusters, in places forming subparallel strands that reflect the internal structuring of the swarm. Toward shallower depths (6.0–7.5 km), only a few master groups are active, consistent with the overall lower number of events in this interval. This distribution indicates that the automated master-event selection generally produces spatially coherent clusters that align with the main fault zone. At the same time, the method may also contribute to apparent scatter or artificial breaks, particularly if a poorly constrained master propagates its bias to the associated events. Thus, while masters capture much of the structural variability of the swarm, some features may still reflect methodological effects rather than true fault geometry. These differences highlight the complementarity of the two approaches. MME offers scalability and flexibility: it can handle essentially unlimited datasets and can be applied iteratively as new events are recorded, which is crucial for long-term monitoring of active regions. The robustness of the MME approach under partial network coverage is further supported by a sensitivity test excluding the NKC station (Appendix C), which showed that the overall spatial patterns and structural resolution were preserved despite the removal of this key station. GC, in contrast, is computationally more demanding, especially for large catalogs, and in this study it was not feasible to relocate all events; earthquakes below magnitude 0 were excluded. A combined strategy may therefore be most effective: MME can be used to build and maintain a comprehensive, continuously updated catalog, while subsets of interest can be reprocessed using GC to resolve finer-scale structures. One potential workflow would be to initialize a master-event catalog using GC for high-quality reference locations, followed by incremental MME relative locations to track ongoing activity in near real time. Importantly, the present results do not allow us to determine which method provides the most accurate absolute image of the fault system. Both MME and GC produce geologically reasonable and internally consistent structures, yet the true geometry of the swarm likely lies between the two images. 6 Conclusions The multi-master event (MME) method provides precise relative locations even for small earthquakes and is computationally efficient for long-term, large catalogs. Its performance depends on the quality of the selected master events and the accuracy of input absolute locations, whereas GrowClust (GC) systematically sharpens the seismicity distribution. Both methods yield consistent images of the southern and northern parts of the active Nový Kostel seismic zone, while the central segment shows the strongest methodological differences. Overall, the comparison demonstrates that the newly developed MME approach produces results consistent with the established GC method, confirming the robustness of both techniques. In addition, the updated catalog of relocated seismicity for 2000–2025 provides the first comprehensive view of West Bohemia since the 2018 swarm, documenting not only repeated activity in the Nový Kostel zone but also swarms in peripheral areas such as Klingenthal and Františkovy Lázně. This highlights the importance of continuous relocation efforts for understanding the temporal and spatial evolution of seismicity. Declarations Author Contribution D. K. conducted the research, developed the relocation code, prepared the figures 1,2,4,5,6,S1-10 and wrote the manuscript. J. D. contributed to the code development and revised the manuscript. B. R. assisted with the programming. J. B., the PhD supervisor of Diana Konrádová, revised the text and contributed to the preparation of figures 3,4,5,6,7. Acknowledgement The authors would like to thank the late Josef Horálek, who served as PhD supervisor and played a key role in shaping this research through his expertise and guidance. His dedication to understanding the West Bohemia/Vogtland region and his mentorship were an invaluable inspiration. The authors also wish to express their gratitude to Prof. RNDr. Tomáš Fischer, Ph.D., and Mgr. Josef Vlček, Ph.D., for their valuable assistance and discussions regarding the use of the GrowClust relocation method.The study was supported by the Grant Agency of Charles University in Prague, GAUK no. 291423 and by the Czech Science Foundation under grant GACR 25-16408X. References Babuška, V., Růžek, B. & Dolejš, D., 2016. Origin of earthquake swarms in the western Bohemian Massif: Is the mantle CO₂ degassing, followed by the Cheb Basin subsidence, an essential driving force? Tectonophysics, 668–669, pp.42–51. doi:10.1016/j.tecto.2015.12.008 Bachura, M. & Fischer, T., 2019. Waveform cross-correlation for differential time measurement: Bias and limitations . Seismological Research Letters , 90(5), 2005–2014. doi:10.1785/0220190096 Bachura, M., Fischer, T., Doubravová, J. & Horálek, J., 2021. From earthquake swarm to a main shock–aftershocks: the 2018 activity in West Bohemia/Vogtland. Geophysical Journal International, 224(3), pp. 1835–1848. doi:10.1093/gji/ggaa523 Bouchaala, F., Vavryčuk, V. & Fischer, T., 2013. Accuracy of the master-event and double-difference locations: synthetic tests and application to seismicity in West Bohemia, Czech Republic. Journal of Seismology, 17, pp. 841–859. doi:10.1007/s10950-013-9357-4 Bräuer, K., Kämpf, H., Niedermann, S., Strauch, G. & Weise, S. M., 2003. Evidence for a nitrogen flux directly derived from the European subcontinental mantle . Nature, 424, 544–547. doi:10.1038/nature01824 Büyükakpınar, P., Dahm, T., Hainzl, S., Isken, M., Ohrnberger, M., Doubravová, J., Wendt, S., & Funke, S., 2025. Recurring earthquake swarms: The role of magmatic fluids in the crust beneath northwestern Bohemia/Vogtland . Communications Earth & Environment (in revision). Douglas, A., 1967. Joint epicentre determination. Nature, 215, pp. 47–48. doi:10.1038/215047a0 Fischer, T. & Horálek, J., 2000. Refined locations of the swarm earthquakes in the Nový Kostel focal zone and spatial distribution of the January 1997 swarm in Western Bohemia, Czech Republic. Studia Geophysica et Geodaetica, 44, pp. 210–226. doi:10.1023/A:1022162826079 Fischer, T. & Horálek, J., 2003. Space-time distribution of earthquake swarms in the principal focal zone of the NW Bohemia/Vogtland seismoactive region: period 1985–2001. Journal of Geodynamics, 35(1–2), pp. 125–144. doi:10.1016/S0264-3707(02)00058-3 Fischer, T., Horálek, J., Michálek, J. & Boušková, A., 2010. The 2008 West Bohemia earthquake swarm in the light of the WEBNET network. Journal of Seismology, 14, pp. 665–682. doi:10.1007/s10950-010-9189-4 Fischer, T., Horálek, J., Hrubcová, P., Vavryčuk, V., Bräuer, K. & Kämpf, H., 2014. Intra-continental earthquake swarms in West-Bohemia and Vogtland: A review. Tectonophysics, 611, pp.1–27. doi:10.1016/j.tecto.2013.11.001 Geissler, W. H., H. Kämpf, R. Kind, K. Bräuer, K. Klinge, T. Plenefisch, J. Horálek, J. Zedník, and V. Nehybka, 2005, Seismic structure and location of a CO2 source in the upper mantle of the western Eger (Ohře) Rift, central Europe , Tectonics, 24, TC5001. doi:10.1029/2004TC001672 Got, J.L., Fréchet, J. & Klein, F.W., 1994. Deep fault plane geometry inferred from multiplet relative relocation beneath the south flank of Kilauea. Journal of Geophysical Research: Solid Earth , 99(B8), pp.15375–15386. doi:10.1029/94JB00577 Hainzl, S., Fischer, T. & Dahm, T., 2012. Seismicity-based estimation of the driving fluid pressure in the case of swarm activity in Western Bohemia. Geophysical Journal International, 191(1), pp.271–281. doi:10.1111/j.1365-246X.2012.05610.x Hainzl, S., Fischer, T., Čermáková, H., Bachura, M. & Vlček, J., 2016. Aftershocks triggered by fluid intrusion: Evidence for the aftershock sequence occurred 2014 in West Bohemia/Vogtland. Journal of Geophysical Research: Solid Earth, 121(4), pp. 2575–2590. doi:10.1002/2015JB012582 Horálek, J. & Fischer, T., 2008. Role of crustal fluids in triggering the West Bohemia/Vogtland earthquake swarms: Just what we know (a review). Studia Geophysica et Geodaetica, 52, pp. 455–478. doi:10.1007/s11200-008-0032-0 Institute of Geophysics, Academy of Sciences of the Czech Republic, 1991. West Bohemia Local Seismic Network [dataset]. International Federation of Digital Seismograph Networks. Jakoubková, H., Horálek, J. & Fischer, T., 2018. Mainshock–aftershock activity versus earthquake swarms in West Bohemia, Czech Republic. Pure and Applied Geophysics, 175, pp. 109–131. doi:10.1007/s00024-017-1679-7 Lomax, A., Virieux, J., Volant, P. & Berge-Thierry, C., 2000. Probabilistic Earthquake Location in 3D and Layered Models. In: C.H. Thurber & N. Rabinowitz, eds. Advances in Seismic Event Location. Modern Approaches in Geophysics, vol. 18. Dordrecht: Springer, pp. 101–134. doi:10.1007/978-94-015-9536-0_5 Málek, J., Horálek, J. & Janský, J., 2005. One-Dimensional qP-Wave Velocity Model of the Upper Crust for the West Bohemia/Vogtland Earthquake Swarm Region. Studia Geophysica et Geodaetica, 49, pp. 501–524. doi:10.1007/s11200-005-0024-2 Richards, P.G., Waldhauser, F., Schaff, D. & Kim, W.-Y., 2006. The applicability of modern methods of earthquake location. Pure and Applied Geophysics, 163, pp. 351–372. doi:10.1007/s00024-005-0019-5 Rubin, A.M., Gillard, D. & Got, J.L., 1998. A reinterpretation of seismicity associated with the 1989 dike intrusion at Kilauea Volcano, Hawaii. Journal of Geophysical Research: Solid Earth, 103(B5), pp. 10003–10015. doi:10.1029/97JB03513 Shelly, D.R., Beroza, G.C., Ide, S. & Nakamula, S., 2006. Low-frequency earthquakes in Shikoku, Japan, and their relationship to episodic tremor and slip. Nature, 442, pp. 188–191. doi:10.1038/nature04931 Stoddard, P.R. & Woods, M.T., 1990. Master event relocation of Gorda Block earthquakes: Implications for deformation. Geophysical Research Letters, 17(7), pp. 961–964. doi:10.1029/GL017i007p00961 Tarantola, A., 2005. Inverse problem theory and methods for model parameter estimation . Philadelphia, PA: Society for Industrial and Applied Mathematics (SIAM). doi:10.1137/1.9780898717921 Trugman, D.T. & Shearer, P.M., 2017. GrowClust: A hierarchical clustering algorithm for relative earthquake relocation, with application to the Spanish Springs and Sheldon, Nevada, earthquake sequences. Seismological Research Letters, 88(2A), pp. 379–391. doi:10.1785/0220160188 Vavryčuk, V., 2011. Principal earthquakes: Theory and observations from the 2008 West Bohemia swarm. Earth and Planetary Science Letters, 305(3–4), pp. 290–296. doi:10.1016/j.epsl.2011.03.002 Waldhauser, F. & Ellsworth, W. L., 2000. A double-difference earthquake location algorithm: Method and application to the northern Hayward Fault, California. Bulletin of the Seismological Society of America, 90(6), pp.1353–1368. doi:10.1785/0120000006 Weinlich, F. H., Bräuer, K., Kämpf, H., Strauch, G., Tesař, J. & Weise, S. M., 1999. An active subcontinental mantle volatile system in the western Eger rift, Central Europe: Gas flux, isotopic (He, C, N) and compositional fingerprints . Geochimica et Cosmochimica Acta, 63(21), 3653–3671. doi:10.1016/S0016-7037(99)00187-8 Weinlich, F.H., Tesař, J., Weise, S.M., Bräuer, K. & Kämpf, H., 1998. Gas flux distribution in mineral springs and tectonic structure in the western Eger Rift. Journal of the Czech Geological Society, 43(1–2), pp. 91–110. Additional Declarations No competing interests reported. Supplementary Files electronicsupplement.docx Appendix.docx Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 06 Jan, 2026 Reviews received at journal 16 Dec, 2025 Reviews received at journal 28 Nov, 2025 Reviewers agreed at journal 13 Nov, 2025 Reviewers agreed at journal 05 Nov, 2025 Reviewers invited by journal 13 Oct, 2025 Editor assigned by journal 09 Oct, 2025 Submission checks completed at journal 09 Oct, 2025 First submitted to journal 07 Oct, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7800475","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":533859846,"identity":"5e4f418e-841b-4ff3-a1dc-4d507cbd21f2","order_by":0,"name":"Diana Konrádová","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAv0lEQVRIiWNgGAWjYDCCwyDigA0PyVrSSNFyAEwcJkEH33HmZw9+nDkvI+/A/PDjDwY7e4JaJA+zmRv23LjNY3iAzViahyGZmaAWg8MMZhI8H4BaGngYpIFOZCNCC/s3yT8fzoG0MP/8wXCAcDgYHOYxk+a5cYBHnoGHTYKH4YAEEX7hKZOWOZPMY8DMZmbNY5BsQFAL3/nj2yTfHLOzl29vfnzzRwURIYbkQjBJvAYGBvkGUlSPglEwCkbBiAIACkE1AAWEGKcAAAAASUVORK5CYII=","orcid":"","institution":"Charles University","correspondingAuthor":true,"prefix":"","firstName":"Diana","middleName":"","lastName":"Konrádová","suffix":""},{"id":533859847,"identity":"68dd704b-e829-4398-872f-252bcd2b6835","order_by":1,"name":"Jana Doubravová","email":"","orcid":"","institution":"Institute of Geophysics of the Czech Academy of Science","correspondingAuthor":false,"prefix":"","firstName":"Jana","middleName":"","lastName":"Doubravová","suffix":""},{"id":533859848,"identity":"b5db3309-1cca-43f8-b344-02fa8bffe60f","order_by":2,"name":"Bohuslav Růžek","email":"","orcid":"","institution":"Institute of Geophysics of the Czech Academy of Science","correspondingAuthor":false,"prefix":"","firstName":"Bohuslav","middleName":"","lastName":"Růžek","suffix":""},{"id":533859849,"identity":"e6da8ed5-1488-4019-a9fa-20fedf263849","order_by":3,"name":"Jan Burjánek","email":"","orcid":"","institution":"Institute of Geophysics of the Czech Academy of Science","correspondingAuthor":false,"prefix":"","firstName":"Jan","middleName":"","lastName":"Burjánek","suffix":""}],"badges":[],"createdAt":"2025-10-07 14:23:21","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7800475/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7800475/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":94473478,"identity":"b40e92b9-73e7-489e-afa4-0413320092f9","added_by":"auto","created_at":"2025-10-27 15:44:34","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10667421,"visible":true,"origin":"","legend":"","description":"","filename":"Paper.docx","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/1f03c0013abcf29ec889e7ec.docx"},{"id":94473814,"identity":"e5536681-0e52-46d0-bd43-f2e52caa913c","added_by":"auto","created_at":"2025-10-27 15:45:48","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6414,"visible":true,"origin":"","legend":"","description":"","filename":"a03c5cf62e6b4a18990ed04a6db8ca30.json","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/f9f39408846843bb91bcab39.json"},{"id":94473459,"identity":"ceeab68b-b3ec-4fee-99dc-015ec7d602f6","added_by":"auto","created_at":"2025-10-27 15:44:24","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2928474,"visible":true,"origin":"","legend":"","description":"","filename":"electronicsupplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/4f07f9c4de1772ab1164b88f.docx"},{"id":94473246,"identity":"aa30cfef-f0c5-42ee-a6ac-3b857377fdd9","added_by":"auto","created_at":"2025-10-27 15:43:36","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":106166,"visible":true,"origin":"","legend":"","description":"","filename":"a03c5cf62e6b4a18990ed04a6db8ca301enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/a1979c422269dc2f3bcb3cbb.xml"},{"id":94473249,"identity":"93c861ba-12a3-4745-a724-051f19411cab","added_by":"auto","created_at":"2025-10-27 15:43:38","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":1989460,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/8536d915adb77c000f0a866e.png"},{"id":94473451,"identity":"b67ee46e-b396-42ed-ac4c-27f3b7c5d527","added_by":"auto","created_at":"2025-10-27 15:44:22","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":92572,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/6bc32b6557ac90419270f671.png"},{"id":94473694,"identity":"3b20329d-da20-4e4e-bb69-b2ebe3dbf6cf","added_by":"auto","created_at":"2025-10-27 15:45:15","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":212906,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/c489901958e1903e190c56f7.png"},{"id":94473455,"identity":"9f9ae9d8-0c46-475a-ad74-1b48be57c5b0","added_by":"auto","created_at":"2025-10-27 15:44:23","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":805915,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/9bff32b09aad748189845909.png"},{"id":94473515,"identity":"4f3442e6-3b61-4135-ab83-f2790e1e7209","added_by":"auto","created_at":"2025-10-27 15:44:43","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":641128,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/6832bfa90735e0f5d167205c.png"},{"id":94473753,"identity":"2c506c66-81da-4629-9a77-c8c7ac58b335","added_by":"auto","created_at":"2025-10-27 15:45:30","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6554759,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/696e886f2426b691d08ed4ef.png"},{"id":94473700,"identity":"7ee23e9b-40ed-4f86-a66c-5235070d06fa","added_by":"auto","created_at":"2025-10-27 15:45:17","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":543581,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/4c823e329c5a4b46d69e1867.png"},{"id":94473758,"identity":"91147037-0676-4ddb-9a8e-f081ee863c0d","added_by":"auto","created_at":"2025-10-27 15:45:32","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":372785,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/db3fc8fcd0b9a4dc64360d24.png"},{"id":94473448,"identity":"23c6901b-401e-4dee-85a4-f90e4325bb28","added_by":"auto","created_at":"2025-10-27 15:44:21","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19930,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/54bb4bf035db14edbb232a49.png"},{"id":94473124,"identity":"9f050c3e-caf6-4b0a-9bff-f8531c47f0aa","added_by":"auto","created_at":"2025-10-27 15:42:44","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":61278,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/9a395b8d442dfd5c3c3fc47e.png"},{"id":94473579,"identity":"56ba318d-206a-48a5-b4cc-19ac6be258fa","added_by":"auto","created_at":"2025-10-27 15:44:54","extension":"png","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170966,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/e6008bdd4425fa7fe6e24e76.png"},{"id":94473767,"identity":"25ed3a6f-6fd4-42d8-b446-d3d23cd0e474","added_by":"auto","created_at":"2025-10-27 15:45:36","extension":"png","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":129201,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/dbb11b937522c85f5466f3dc.png"},{"id":94473754,"identity":"aba7d4bf-70a5-4dd6-b74d-8acfe88362b5","added_by":"auto","created_at":"2025-10-27 15:45:31","extension":"png","order_by":23,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":967644,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/2192fa9548bef6eb2d7504e1.png"},{"id":94473218,"identity":"c5f497c2-e50c-4437-8f41-03ba4dcc9582","added_by":"auto","created_at":"2025-10-27 15:43:23","extension":"png","order_by":24,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":136218,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/ee22252e3d34e667f72df730.png"},{"id":94473376,"identity":"5c7e3e64-5d2d-4048-8ef5-4ce81cba5874","added_by":"auto","created_at":"2025-10-27 15:44:10","extension":"xml","order_by":25,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":107185,"visible":true,"origin":"","legend":"","description":"","filename":"a03c5cf62e6b4a18990ed04a6db8ca301structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/a4a5689aa8ef73eeb8b60486.xml"},{"id":94473507,"identity":"791e074e-f165-42ec-afde-f94524598e2a","added_by":"auto","created_at":"2025-10-27 15:44:40","extension":"html","order_by":26,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":116396,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/29477b27bcea067cf161a771.html"},{"id":94473723,"identity":"c03861ba-3458-4975-a215-da3566f0fe83","added_by":"auto","created_at":"2025-10-27 15:45:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2483176,"visible":true,"origin":"","legend":"\u003cp\u003eSeismically active area of the West Bohemia/Vogtland region. Black dots mark earthquakes (ML ≥ 0) recorded between 2000 and 2024, orange triangles indicate WEBNET seismic stations, and major tectonic structures such as the Mariánské Lázně Fault and the Eger Rift are highlighted.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/5a873fca1732e36862f81c3b.png"},{"id":94473510,"identity":"4ec5fc62-6cd7-42e2-b52d-49ae40082588","added_by":"auto","created_at":"2025-10-27 15:44:42","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":84500,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the multi-master event (MME) relocation algorithm. Inputs include the seismic event catalog, station list, velocity model, and previously selected master events (if available). The algorithm iteratively selects new master events, relocates associated events, and produces as output both relocated and non-relocated events.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/98151408a02385f805686ec6.png"},{"id":94473365,"identity":"62bb1039-ef6e-4919-99d9-e80b00e6e841","added_by":"auto","created_at":"2025-10-27 15:44:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":213308,"visible":true,"origin":"","legend":"\u003cp\u003eAbsolute locations of seismic events in the Nový Kostel area. Map view (A), depth versus latitude section (B), and depth versus longitude section (C) illustrate the diffuse distribution of hypocenters prior to relocation. Red dots represent events with M \u0026gt; 1.5 and black dots correspond to smaller magnitudes down to M=0. Coordinates are given in UTM zone 33N.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/3e8e145d3b9860cc0bf436f4.png"},{"id":94473755,"identity":"a31c39ef-8f7f-48f7-ad84-b5507e8ac113","added_by":"auto","created_at":"2025-10-27 15:45:31","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":834772,"visible":true,"origin":"","legend":"\u003cp\u003eDepth-slice map-view comparison of absolute locations (top row), multi-master locations (middle row), and GrowClust locations (bottom row). Each panel represents a 0.5 km depth interval.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/81d72f2afd6a775a1781acea.png"},{"id":94489324,"identity":"908d6e3b-7c8a-4178-aa0f-b30b782662b6","added_by":"auto","created_at":"2025-10-27 17:04:11","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":671078,"visible":true,"origin":"","legend":"\u003cp\u003eVertical cross-section E-W comparison of absolute locations (top row), multi-master relative locations (middle row), and GrowClust relative locations (bottom row).\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/82d18aa367d02c497e096d73.png"},{"id":94489651,"identity":"699843f7-55d1-46d9-893a-5c4952cbac94","added_by":"auto","created_at":"2025-10-27 17:05:31","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":8098297,"visible":true,"origin":"","legend":"\u003cp\u003eRelocated seismicity in the West Bohemia/Vogtland region between 2000 and 30 September 2025, shown in 2 km depth intervals: (a) 0–7 km, (b) 7–9 km, (c) 9–11 km, (d) 11–13 km, and (e) 13–15 km. Circle size scales with magnitude, and colors indicate event time (see legend). Triangles denote WEBNET seismic stations, and blue lines mark the main faults in the region.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/8338070ded92ce4836e77608.png"},{"id":94473375,"identity":"075c1153-7196-4446-883e-d7cf9816c914","added_by":"auto","created_at":"2025-10-27 15:44:10","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":545553,"visible":true,"origin":"","legend":"\u003cp\u003eDepth-sliced projections of relocated seismicity color-coded by assigned master event, shown in 0.5 km intervals between 6.0 and 10.5 km: (a) 10.0–10.5 km, (b) 9.5–10.0 km, (c) 9.0–9.5 km, (d) 8.5–9.0 km, (e) 8.0–8.5 km, (f) 7.5–8.0 km, (g) 7.0–7.5 km, (h) 6.5–7.0 km, and (i) 6.0–6.5 km. Each color represents one master event and its associated group of relocated events. The distribution shows that masters generally control spatially coherent subsets aligned with the main NNW–SSE fault trend, while at shallower depths only a few groups are active due to the overall lower number of events. Triangles denote WEBNET seismic stations.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/4bab25e8d613cbecedc7c484.png"},{"id":94491218,"identity":"22bd31a2-5ce2-4fa2-b81d-aaee4e8af973","added_by":"auto","created_at":"2025-10-27 17:23:55","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":11736244,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/4ab63e71-8572-4ce5-8567-073598dbd7cc.pdf"},{"id":94473311,"identity":"9073602c-4d70-4634-b52b-c0d94cf6b39a","added_by":"auto","created_at":"2025-10-27 15:43:57","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2928474,"visible":true,"origin":"","legend":"","description":"","filename":"electronicsupplement.docx","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/0c5b64b672c533083b7cc80b.docx"},{"id":94473827,"identity":"ca359cb8-8efc-4d57-9f9e-daa734db2075","added_by":"auto","created_at":"2025-10-27 15:45:51","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":15919,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-7800475/v1/73f59aee836ec0ec70c5038d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Multi-Master Event Approach to Earthquake Relocation: Insights from the West Bohemia Swarm Zone (Czech Republic, 2000–2025)","fulltext":[{"header":"Article Highlights","content":"\u003cul\u003e\n \u003cli\u003eA new multi-master event method improves earthquake relocation for large and long-term seismic catalogs.\u003c/li\u003e\n \u003cli\u003eThe study provides the first detailed view of seismic activity in West Bohemia, updated through 2025.\u003c/li\u003e\n \u003cli\u003eComparison with GrowClust shows consistent fault structures but highlights key methodological differences.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"1 Introduction","content":"\u003cp\u003eLocation of earthquakes is a primary step of any seismological studies. Its accuracy and quality is crucial for next processing from magnitude evaluation to focal mechanism estimation. Earthquake relative location is used to refine the positions of seismic events beyond the accuracy of routine catalogs. While absolute locations are often affected by uncertainties in the velocity model and errors of onset picking, relative relocation techniques exploit differential arrival times between nearby events to achieve much higher spatial resolution. This is essential for resolving fine-scale fault structures, seismic clusters, detecting migrating fluid fronts along active faults, and understanding the underlying seismogenic processes (e.g., Got et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1994\u003c/span\u003e; Rubin et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1998\u003c/span\u003e; Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e; Shelly et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe West Bohemia/Vogtland region is well known for recurrent earthquake swarms, which are generally interpreted as a result of fluid-driven processes at mid-crustal depths interacting with local tectonics (Hor\u0026aacute;lek \u0026amp; Fischer, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). The study area, seismic network (WEBNET), and major tectonic structures are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Studies applying relative locations have played a crucial role in the understanding of these swarms. For example, Fischer and Hor\u0026aacute;lek (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) used master-event method to refine the locations of foci of 1997 swarm. Bouchaala et al. (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) applied both the master-event and double-difference (hypoDD; Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) relative location methods, using catalog picks as well as cross-correlation differential times of P- and S-waves, to the 2008 West Bohemia swarm in order to refine the cluster structures and to reveal the detailed geometry of the active fault zone. Jakoubkov\u0026aacute; et al. (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) refined the locations of the 2014 mainshock\u0026ndash;aftershock sequence and several previous swarms (1997, 2000, 2008, 2011) using the double-difference relative location method (HypoDD), and compared them with typical swarm-type seismicity in the area, highlighting key differences in their spatial and temporal evolution.\u003c/p\u003e\u003cp\u003eIn this study, we develop and test an alternative relative location approach, the multi-master event (MME) method. Standard relocation codes such as hypoDD are most effective when applied to dense localized earthquake clusters, but it becomes computationally expensive for very large catalogs and is not well suited for incremental processing when new events are continuously added. The MME method was designed here to address these limitations: instead of requiring manual selection of master events typically needed for master event relative locations, it automatically identifies suitable master events according to predefined criteria (e.g., number of high-quality phase picks). This approach saves considerable processing time and takes advantage of manually picked phases in the WEBNET catalog, while still allowing user-defined master events if needed. As a result, the method is well suited for large earthquake catalogs with events of varying data quality. For instance, well-recorded events with many clear phase picks as well as small earthquakes with only a few, noisier arrivals, while also enabling incremental, near real-time catalog updates. One of our goals is to assess its performance in the West Bohemia region and compare it to results from GrowClust, a modern and widely used double-difference relocation tool which was developed to reduce the main limitations of hypoDD (Trugman \u0026amp; Shearer, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), offering improved computational efficiency for large datasets and greater robustness against outliers.\u003c/p\u003e"},{"header":"2 Methods","content":"\u003cp\u003eAll relative location methods are based on the assumption that if two earthquakes occur very close to each other compared to their distance from the recording stations, their seismic waves travel along nearly identical paths. In this case, the effects of velocity model inaccuracies along the path can be largely canceled out, and the relative arrival time difference can be attributed solely to the relative offset between the event locations. This relative location can be resolved with much higher precision than their absolute positions. However, uncertainties in phase picking remain. These can be reduced by calculating differential times from waveform cross-correlation instead of relying solely on catalog picks (e.g., Got et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e1994\u003c/span\u003e; Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe concept of determining relative event positions with respect to a well-located reference (master) event originates from the joint hypocenter determination approach introduced by Douglas (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1967\u003c/span\u003e) and was later successfully applied to larger seismically active areas, such as in northern California (Stoddard \u0026amp; Woods, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1990\u003c/span\u003e). These studies demonstrated the method\u0026rsquo;s ability to enhance location precision and reveal fine-scale deformation structures.\u003c/p\u003e\u003cp\u003eNowadays, the most widely used method is the double-difference relocation algorithm (Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), which has proven highly effective for resolving detailed fault-zone structures and relative event distributions (Richards et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). The double-difference method extends the master-event concept by treating all events symmetrically as both \u0026ldquo;masters\u0026rdquo; and \u0026ldquo;targets\u0026rdquo;. This means that the relative locations of nearby earthquakes are adjusted so that observed differential travel times between pairs of events match those predicted by the velocity model. The first widely distributed code implementing this approach was hypoDD, and a more recent implementation GrowClust (Trugman \u0026amp; Shearer, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) introduces hierarchical clustering to improve computational efficiency for large catalogs.\u003c/p\u003e\u003cp\u003eBoth hypoDD and GrowClust can make use of differential times obtained from catalog picks or from waveform cross-correlation (XC). In hypoDD, XC can be used to improve the precision of catalog picks, but it is not a requirement, the code can be run on catalog data alone. XC significantly reduces errors caused by inaccurate onset picking, because correlation lags provide more precise relative timing between similar events. However, Bachura and Fischer (\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) showed that the use of XC is not always beneficial. For events with large magnitude differences or dissimilar waveforms, the correlation may align peak amplitudes instead of the true first arrivals. This effect can lead to an apparent shift of the largest events away from their absolute locations. In contrast, using carefully selected manual picks can be still a reasonable option (Bachura \u0026amp; Fischer, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Multi-Master Event Relative Location Method\u003c/h2\u003e\u003cp\u003eTo improve the precision of earthquake locations and enable processing of large seismic datasets, we developed and applied a custom multi-master event (MME) relocation approach. This method is based on differential travel times relative to one of a list of well-constrained master events and supports incremental processing, allowing new events to be added without the need to re-run the entire relocation procedure. This makes the method particularly suitable for operational seismic monitoring and long-term studies in seismically active regions.\u003c/p\u003e\u003cp\u003eA schematic flowchart of the MME relative relocation procedure, summarizing the inputs, iterative processing steps, and outputs, is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The algorithm works iteratively. In each iteration, a master event is selected from the catalog according to predefined criteria. A key condition is the number of available arrival-time picks. The master event should have as many readings as possible to ensure that associated events (targets) can be successfully matched. Based on several preliminary tests, we found that requiring at least 20 high-quality P- and/or S-phase picks provided stable and reliable relocation results for the data and station configuration used in this study. A high number of picks typically corresponds to a good signal-to-noise ratio, which assures low onset picking error.\u003c/p\u003e\u003cp\u003eA subset of events in the vicinity of the master event is selected according to a predefined maximum hypocentral distance. In this study we used the limit of 1 km. For each of these events, the classical master-event location problem is solved (e.g., Stoddard \u0026amp; Woods, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e1990\u003c/span\u003e; Bouchaala et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). To determine the optimal hypocenter for each event, the algorithm minimizes the residuals between observed and synthetic differential travel times relative to the selected master event. For a given target event and master, the differential travel time at each station \u003cem\u003ei\u003c/em\u003e is defined as:\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\:\\delta\\:{t}_{i}={\\left({t}_{i}^{e}-{t}_{i}^{m}\\right)}_{obs}-{\\left({t}_{i}^{e}-{t}_{i}^{m}\\right)}_{syn}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cem\u003et\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e\u003cem\u003ee\u003c/em\u003e\u003c/sup\u003e and \u003cem\u003et\u003c/em\u003e\u003csub\u003e\u003cem\u003ei\u003c/em\u003e\u003c/sub\u003e\u003csup\u003e\u003cem\u003em\u003c/em\u003e\u003c/sup\u003e are the arrival times for the event and the master, respectively. The observed differential times are computed directly from the phase picks, while the theoretical differential times are calculated using a 1-D constant velocity layered model with the ray-theory approximation. The algorithm then searches for the hypocenter location and origin time (x,y,z,t) of the event that minimizes the sum of squared residuals between observed and theoretical differential times:\u003cdiv id=\"Equb\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equb\" name=\"EquationSource\"\u003e\n$$\\:min{\\sum\\:}_{i}{\\left(\\delta\\:{t}_{i}\\right)}^{2}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe optimization in our MME implementation is performed using a non-linear least-squares approach, which iteratively adjusts the event\u0026rsquo;s coordinates to minimize the overall misfit between observed and theoretical differential travel times (Tarantola, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). The inversion is constrained to a three-dimensional volume around the initial location of the event, with boundaries defined by user-specified limits of 1 km horizontally and 2 km vertically, as chosen in this study.\u003c/p\u003e\u003cp\u003eAfter the relocation, the event is saved and removed from the set, and the algorithm looks for remaining events in the catalog that have not been relocated yet. If the remaining events cannot be linked to the current master, a new master event is selected from the remaining set and the relocation process is repeated. This cycle continues until no more events satisfy the criteria for master selection.\u003c/p\u003e\u003cp\u003eThe inputs and outputs of the algorithm are summarized in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e. The algorithm can either continue from a previous relocation using an existing set of master events or start from scratch, choosing the first master event automatically. We tested the sequential processing option to simulate incremental catalog updates, and the procedure is described in detail in section Results.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 GrowClust Relocation Method\u003c/h2\u003e\u003cp\u003eFor comparison, we applied the GrowClust relative relocation algorithm, which uses a hierarchical clustering approach combined with double-difference inversion to improve the relative locations of earthquakes (Trugman and Shearer, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). GrowClust identifies clusters of similar events based on waveform similarity quantified by cross-correlation, and relocates them by minimizing differential travel-time residuals.\u003c/p\u003e\u003cp\u003eThe relocation is based on the double-difference method, which reduces the influence of velocity model errors by focusing on differential arrival times between event pairs recorded at common stations:\u003cdiv id=\"Equc\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equc\" name=\"EquationSource\"\u003e\n$$\\:\\delta\\:{t}_{kj}={\\left({t}_{k}-{t}_{j}\\right)}_{obs}-{\\left({t}_{k}-{t}_{j}\\right)}_{syn}={G}_{k}\\cdot\\:\\delta\\:{x}_{k}-{G}_{j}\\cdot\\:\\delta\\:{x}_{j}$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cem\u003et\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003et\u003c/em\u003e\u003csub\u003e\u003cem\u003ej\u003c/em\u003e\u003c/sub\u003e are the arrival times of events \u003cem\u003ek\u003c/em\u003e and \u003cem\u003ej\u003c/em\u003e. \u003cem\u003eδt\u003c/em\u003e\u003csub\u003e\u003cem\u003ekj\u003c/em\u003e\u003c/sub\u003e represents the observed minus theoretical differential arrival time between the two events. \u003cem\u003eδx\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eδx\u003c/em\u003e\u003csub\u003e\u003cem\u003ej\u003c/em\u003e\u003c/sub\u003e denote the corresponding location correction vectors, and \u003cem\u003eG\u003c/em\u003e\u003csub\u003e\u003cem\u003ek\u003c/em\u003e\u003c/sub\u003e and \u003cem\u003eG\u003c/em\u003e\u003csub\u003e\u003cem\u003ej\u003c/em\u003e\u003c/sub\u003e are the partial derivatives of travel time with respect to event coordinates (i.e., travel time sensitivity to the small changes in the event location). The system of equations is solved iteratively to minimize the residuals across all available event pairs.\u003c/p\u003e\u003cp\u003eThe GrowClust algorithm uses a hierarchical approach: it begins by identifying tightly clustered event groups and relocates them, progressively adding larger or more distant clusters in later iterations. This strategy improves scalability and allows GrowClust to process large datasets more efficiently. The relocation considers both P- and S-wave picks and incorporates cross-correlation-derived differential times to improve precision. The details can be found in Trugman and Shearer (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Data","content":"\u003cp\u003eWe used an earthquake catalog with manual picks of P- and S-wave onsets, recorded by the WEBNET seismic network (Institute of Geophysics, 1991; Hor\u0026aacute;lek \u0026amp; Fischer, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), operating in the West Bohemia/Vogtland region. WEBNET is a local seismic network designed for monitoring local seismicity, and it has played a key role in documenting swarm activity and fluid-induced processes in the area (Fischer et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Hainzl et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Hainzl et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). The WEBNET network geometry (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) provides dense station coverage in the vicinity of the Nov\u0026yacute; Kostel seismic zone, ensuring high-quality arrival times and good azimuthal resolution for the majority of events.\u003c/p\u003e\u003cp\u003eThe dataset includes approximately 62,000 local earthquakes recorded between January 2000 and October 2025. We selected this time period because the catalog coverage is most complete and consistent since 2000. All events have manual picks of P- and S-wave arrival times. The absolute locations were computed using NonLinLoc (Lomax et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) with a 1-D layered velocity model based on M\u0026aacute;lek et al. (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), who parameterized the crust as a constant gradient velocity layers. The catalog includes events across a wide range of magnitudes M\u003csub\u003eL\u003c/sub\u003e from \u0026minus;\u0026thinsp;2 to 4.1.\u003c/p\u003e\u003cp\u003eThe quality of the input absolute locations was evaluated from NonLinLoc covariance matrices (see Appendix A). This analysis confirmed that horizontal and vertical uncertainties are mostly below 1 km, ensuring sufficient accuracy for subsequent relative relocation.\u003c/p\u003e\u003cp\u003eDue to computational limitations in applying GrowClust to the full dataset, we restricted the relocation to events with M\u0026thinsp;\u0026gt;\u0026thinsp;0, yielding a representative subset of ~\u0026thinsp;30,000 events. In contrast, the multi-master method was applied to the entire dataset of 62,000 events, including lower-magnitude events. However, for visualization and comparison purposes, only events with M\u0026thinsp;\u0026gt;\u0026thinsp;0 are shown in the final maps. The comparison between the two methods is based on events relocated up to June 2024, corresponding to the temporal extent of the GrowClust dataset.\u003c/p\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Original Absolute Locations\u003c/h2\u003e\u003cp\u003eThe original absolute locations for the most active Nov\u0026yacute; Kostel seismic zone are plotted in Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Coordinates are shown in UTM zone 33N (easting in km on the x-axis, northing in km on the y-axis). They show a vertically elongated seismic zone stretching north\u0026ndash;south, centered around the Nov\u0026yacute; Kostel region. Despite several spatial clusters and alignments can be identified, the resulting absolute locations exhibit significant vertical and horizontal scatter, likely due to imperfection of the velocity model and observational uncertainties in arrival times. This scatter obscures finer fault-related structures and limits the ability to interpret structural segmentation from absolute locations alone.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Multi-master Event Results\u003c/h2\u003e\u003cp\u003eThe relocation results and their comparison with the absolute locations are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e. The MME relative locations are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg-l and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ek-t for the most active Nov\u0026yacute; Kostel seismic zone. All coordinates are given in UTM zone 33N, expressed in kilometers. In the map view (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg-l), seismicity is concentrated within a narrow NNW\u0026ndash;SSE trending fault system, extending over ~\u0026thinsp;15 km in length. Several parallel strands can be recognized, with the highest density of activity in the central segment (5566\u0026thinsp;\u0026lt;\u0026thinsp;Y\u0026thinsp;\u0026lt;\u0026thinsp;5572 km), while the southern (Y\u0026thinsp;\u0026lt;\u0026thinsp;5566 km) and northern (Y\u0026thinsp;\u0026gt;\u0026thinsp;5572 km) segments are less active. Overall the relocated events closely follow the main fault orientation.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg-l and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ek-t confirm that most seismicity is confined to a depth band between 8 and 10 km. Within this interval, events align into elongated clusters with minor lateral variations. These may reflect subtle internal segmentation of the fault zone, although some methodological effects cannot be fully excluded. In particular, the multi-master approach is sensitive to the choice of master events: if the absolute location of a selected master is biased, this bias can propagate to its associated cluster. Moreover, the imposed 1 km association threshold and the variable number of available phase picks per master may produce apparent breaks or scatter in the relocated structures. Above 6.5 km and below 10.5 km only a sparse seismicity is observed.\u003c/p\u003e\u003cp\u003eDepth-sliced projections of the multi-master event relative locations (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eg-l) further illustrate the internal segmentation of the Nov\u0026yacute; Kostel seismic zone. Even within narrow 0.5 km depth intervals, the relocated events consistently follow the NNW\u0026ndash;SSE fault orientation. Seismicity is not homogeneously distributed but forms elongated clusters and locally parallel strands, with the highest density concentrated between 8 and 9.5 km. Some depth levels (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh-j) are dominated by compact, well-populated clusters, while others (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003el) appear more fragmented and diffuse.\u003c/p\u003e\u003cp\u003eThe complementary E\u0026ndash;W cross-sections, Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ek-t, emphasize the vertical confinement of the activity. Most events occur between ~\u0026thinsp;8 and 10 km, where they form vertically elongated clusters with slight lateral variations. These may reflect internal segmentation of the fault zone, although part of the pattern may also result from the relocation procedure and master-event selection (see later the Discussion section for details). Toward the margins (e.g., panels k-m at shallower depths and s-t at greater depths in Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e), seismicity becomes more diffuse, whereas above 7.5 km and below 11 km only the sparse activity is observed. The central segment (Y\u0026thinsp;=\u0026thinsp;5566\u0026ndash;5572 km) is characterized by several subparallel strands. At depths around ~\u0026thinsp;9 km, the main structure bends eastward, and the relocated events form clusters separated by ~\u0026thinsp;400\u0026ndash;500 m. These representative slices highlight the overall structural coherence of the relocated swarm, yet reveal that some internal complexity remains. The complete set of depth slices is available in the electronic supplement (Figure S2), and the vertical cross-sections are presented in Figures S3a and S3b.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe MME method, enabled the relocation of nearly all events in the dataset. Because the MME method follows an iterative scheme, its results can be sensitive to the initial conditions, in particular, to the order of events in the catalog. To test the possible sensitivity to input order, we processed the catalog in three ways: 1) sequentially from the oldest to the newest events; 2) in reverse order from the newest to the oldest; 3) randomly shuffled events in the catalog. In all cases, the resulting catalogs were consistent, with differences below the resolution of the mapping scale (Figures S4 and S5 in the electronic supplement).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e4.2 GrowClust results\u003c/h2\u003e\u003cp\u003eThe GrowClust (GC) relocations are shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003em-r and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eu-dd. Compared to the diffuse absolute locations, GC relocations reveal a strongly focused seismicity pattern. By exploiting waveform similarity and hierarchical clustering, the method produces narrow, sharply defined strands and a dense central core of activity at depths between 8 and 11 km, where the majority of swarm earthquakes are concentrated. In both map and cross-sectional views, the relocated events align into thin, fault-parallel strands that are in places only a few hundred meters wide, demonstrating GC\u0026rsquo;s ability to extract fine-scale structural details from the dataset.\u003c/p\u003e\u003cp\u003eClusters that previously appeared as broad clouds in the absolute locations (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea-f and Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea-j) are transformed into elongated, coherent segments. The depth-sliced map views (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003em-r) and E\u0026ndash;W cross-sections (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eu-dd) confirm this focusing: seismicity is concentrated between 8 and 10 km, forming vertically continuous and internally coherent structures, while only sparse activity is observed above 7 km and below 11 km. These results emphasize that GC captures the central fault zone with high resolution but suppresses more diffuse or peripheral seismicity.\u003c/p\u003e\u003cp\u003eConsidering the individual segments, the southern part (Y\u0026thinsp;\u0026lt;\u0026thinsp;5566 km) displays relatively low activity but a stable pattern that is consistently imaged by GC. The northern part (Y\u0026thinsp;\u0026gt;\u0026thinsp;5572 km) shows a robust and coherent structure, with seismicity forming a narrow, well-aligned fault-parallel band. The central part (5566\u0026thinsp;\u0026lt;\u0026thinsp;Y\u0026thinsp;\u0026lt;\u0026thinsp;5572 km) is the most complex, with multiple overlapping clusters, branching features, and local bends in the seismicity distribution. GC relative locations reveal several compact and continuous vertical planes, with seismicity bending slightly eastward at depths of 9 km. Within the 8.5\u0026ndash;9.5 km interval, two closely spaced subparallel strands can be recognized, separated by 200\u0026ndash;250 m. Overall, the GC relative locations provide a sharply focused image of the Nov\u0026yacute; Kostel seismic zone, characterized by compact fault strands and continuous vertical planes.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e4.3 Incremental Relocation Tests simulating for real-time operation\u003c/h2\u003e\u003cp\u003eTo evaluate the capability of the multi-master method for incremental relocation, we designed a sequential test simulating an operational scenario. The relocation process is carried out in two steps:\u003c/p\u003e\u003cp\u003e\u003col\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eInitial relocation: In the first step, the algorithm starts with no predefined master list. Master events are selected automatically from the whole available catalog (in this case 2000\u0026ndash;2014) as those with the highest number of picks, i.e., the events that best satisfy the predefined criteria.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003cspan\u003e\u003cli\u003e\u003cp\u003eIncremental update: In the second steps, the master events from the initial run are used as input for processing a new daily batch of events. The existing master events are retained and new events are relocated relative to them. If no suitable master is available for a given new event, the event is either selected as a new master event (if it meets the master event criteria) or left temporarily not relocated, awaiting the occurrence of a new master to which it can be linked.\u003c/p\u003e\u003c/li\u003e\u003c/span\u003e\u003c/ol\u003e\u003c/p\u003e\u003cp\u003eThe results show that sequential relocation yields comparable spatial resolution to a single-pass relocation, with improved scalability and the ability to integrate new data without re-running the entire process. The key difference between the two approaches lies in how master events are selected. In batch relocation (which still runs iteratively), the best possible master event is always selected from the pool of not-yet-relocated events, the one with the highest number of phase readings. In contrast, the sequential approach selects any event that meets the master criteria at the moment it appears in the data stream. This better reflects real-time processing, where events are relocated progressively as they become available. This test demonstrates the method\u0026rsquo;s flexibility for handling dynamic catalogs and its suitability for long-term seismic monitoring.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Relocation of the entire seismic activity from 2000 to 2025 using the MME method\u003c/h2\u003e\u003cp\u003eWe relocated the complete seismicity recorded between 2000 and 2025 using the Multiple-Event relocation (MME) method to obtain an up-to-date and internally consistent catalog. As the WEBNET seismic network configuration was stabilized in 2000, relocation results are shown only for the activity recorded since 2000. Seismic activity up to 2018 has been documented in detail in earlier studies (e.g., Fischer \u0026amp; Hor\u0026aacute;lek, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Fischer et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Vavryčuk, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Hainzl et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2012\u003c/span\u003e; Bachura et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), Here we focus primarily on the most recent activity, with locations from 2018 onward reported for the first time, except for the 2024 Klingenthal swarm already described by B\u0026uuml;y\u0026uuml;kakpınar et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThe complete set of relocated seismicity between 2000 and 30 September 2025 is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. Events are displayed in 2 km depth intervals to highlight the vertical distribution of activity. All relative locations were obtained using the multi-master event (MME) relocation approach.\u003c/p\u003e\u003cp\u003eThe Nov\u0026yacute; Kostel seismic zone remains the dominant feature, but its activity shows a gradual spatial expansion, most prominently northward, and southward, now covering a total length of ~\u0026thinsp;20 km. The most recent swarms (red to brown colors) highlight this migration trend.\u003c/p\u003e\u003cp\u003eIn 2024, a pronounced swarm occurred near Klingenthal (station BUBD) analyzed by B\u0026uuml;y\u0026uuml;kakpınar et al. (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), shortly followed by activity near Františkovy L\u0026aacute;zně. The Klingenthal swarm follows depth patterns similar to those in the main zone, with smaller clusters in 7\u0026ndash;9 km (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eb) but the most events and the strongest ones are concentrated in 9\u0026ndash;11 km (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ec). Notably, all earthquakes with magnitudes M\u003csub\u003eL\u003c/sub\u003e \u0026ge; 2.5 in the area of Nov\u0026yacute; Kostel seismic zone are confined to the 9\u0026ndash;11 km depth range. By contrast, the Františkovy L\u0026aacute;zně activity is distinct in its depth distribution: it is confined mainly to 11\u0026ndash;13 km (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed), making it deeper than the main Nov\u0026yacute; Kostel zone. Moreover, several earthquakes with magnitudes exceeding M\u003csub\u003eL\u003c/sub\u003e 2.5 occurred there, which contrasts with the main zone.\u003c/p\u003e\u003cp\u003eAdditional deep activity has emerged near Luby (station LBC), at depths of 13 km and deeper (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee). These events lie outside the main Nov\u0026yacute; Kostel seismic zone and suggest localized activity in an area that was previously less seismically active. The occurrence of these new deep clusters, together with the Františkovy L\u0026aacute;zně swarm, is noteworthy given that both regions coincide with well-documented CO₂ emanations and fluid discharge in the Cheb Basin and its surroundings (Weinlich et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1999\u003c/span\u003e; Geissler et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Babuška et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). This supports the widely discussed interpretation that mantle-derived fluids play a crucial role in triggering and sustaining earthquake swarms in western Bohemia/Vogtland (e.g., Weinlich et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1998\u003c/span\u003e; Br\u0026auml;uer et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn the shallowest interval 0\u0026ndash;7 km, Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea, seismicity is comparatively sparse but present in the Nov\u0026yacute; Kostel zone, mainly in its northern part. Outside the main zone, several scattered shallow clusters are observed, many of which are related to quarry blasts rather than tectonic activity.\u003c/p\u003e\u003cp\u003eOverall, the comprehensive dataset reveals that while the Nov\u0026yacute; Kostel zone remains persistently active, the recent swarms in Klingenthal and Františkovy L\u0026aacute;zně, together with deep events near Luby, indicate that seismic activity now also occurs at greater depths and in regions outside the main zone. These observations point to a continuing north-south migration and an increasing diversification of swarm activity in the western Bohemia/Vogtland region.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"5 Discussion","content":"\u003cp\u003eThe results demonstrate that the multi-master method (MME) provides precise relative locations even for small events, while remaining simple and computationally feasible for long-term and high-volume catalogs. Its ability to incrementally incorporate new events without the need to reprocess the entire dataset makes it particularly suitable for operational monitoring and real-time updating of seismicity patterns.\u003c/p\u003e\u003cp\u003eAt the same time, the performance of the MME approach is strongly dependent on the quality of the chosen master events. Inaccuracies in the absolute locations used as input can propagate through the relocation process, leading to biased or scattered clusters. This limitation contrasts with the GrowClust (GC) method, which systematically sharpens the seismicity distribution and often produces clearer and more compact fault-parallel structures. However, it remains an open question whether this increased sharpness necessarily reflects the true geometry, or whether it partly results from the intrinsic smoothing tendencies of the algorithm. In some cases, MME relative locations preserve variability that may be methodological artifacts, but which might also capture small-scale structural complexity that GC suppresses.\u003c/p\u003e\u003cp\u003eSince the WEBNET catalog already provides high-quality, manually picked P- and S-phases, we did not apply waveform cross-correlation in the MME relative locations. While waveform XC can further improve relative timing, its computation is time-consuming, and for monitoring applications the ability to obtain rapid and consistent solutions is often more important.\u003c/p\u003e\u003cp\u003eThe differences between the two methods should therefore be interpreted with caution. The contrasting images are likely methodological rather than reflecting genuine structural segmentation. In the case of the MME approach, the relocation is sensitive to the absolute position and quality of the selected master events. If a master is poorly constrained, this bias propagates to its associated cluster. In addition, the imposed 1 km association threshold can artificially separate events into multiple strands, and differences in the number and quality of picks per master can lead to uneven relocation precision. These factors can produce apparent scatter or fragmented clusters that may not reflect true fault geometry.\u003c/p\u003e\u003cp\u003eFor GrowClust, on the other hand, the use of waveform similarity and hierarchical clustering tends to emphasize coherence and compactness. While this improves the sharpness of fault-parallel structures, it also introduces an intrinsic smoothing effect: events with weaker correlations or located at the margins of the network may be pulled toward the main clusters, reducing the visibility of diffuse or secondary features. Thus, the compact strands in GC images may partly reflect algorithmic focusing rather than genuine absence of complexity.\u003c/p\u003e\u003cp\u003eTaken together, these methodological effects explain why the central segment (5566\u0026thinsp;\u0026lt;\u0026thinsp;Y\u0026thinsp;\u0026lt;\u0026thinsp;5572 km) appears more scattered in MME relative locations and more sharply defined in GC results. The quantitative comparison of residuals and event shifts further supports this interpretation, confirming that both methods provide stable solutions with small residuals and relative locations mostly below 1 km (Appendix B).\u003c/p\u003e\u003cp\u003eTo further illustrate the relocation process, events were color-coded according to their assigned master event (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). Each color represents one master event and its associated group of relocated events. The map-view distribution shows that individual master groups overlap slightly but together form a continuous NNW\u0026ndash;SSE-trending structure that coincides with the main seismic zone.\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e shows that the automated master-event selection generally produces spatially coherent clusters within the 6.0\u0026ndash;10.5 km depth range. Between 8.0 and 10.0 km, different masters capture compact but still distinguishable clusters, in places forming subparallel strands that reflect the internal structuring of the swarm. Toward shallower depths (6.0\u0026ndash;7.5 km), only a few master groups are active, consistent with the overall lower number of events in this interval.\u003c/p\u003e\u003cp\u003eThis distribution indicates that the automated master-event selection generally produces spatially coherent clusters that align with the main fault zone. At the same time, the method may also contribute to apparent scatter or artificial breaks, particularly if a poorly constrained master propagates its bias to the associated events. Thus, while masters capture much of the structural variability of the swarm, some features may still reflect methodological effects rather than true fault geometry.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThese differences highlight the complementarity of the two approaches. MME offers scalability and flexibility: it can handle essentially unlimited datasets and can be applied iteratively as new events are recorded, which is crucial for long-term monitoring of active regions. The robustness of the MME approach under partial network coverage is further supported by a sensitivity test excluding the NKC station (Appendix C), which showed that the overall spatial patterns and structural resolution were preserved despite the removal of this key station. GC, in contrast, is computationally more demanding, especially for large catalogs, and in this study it was not feasible to relocate all events; earthquakes below magnitude 0 were excluded.\u003c/p\u003e\u003cp\u003eA combined strategy may therefore be most effective: MME can be used to build and maintain a comprehensive, continuously updated catalog, while subsets of interest can be reprocessed using GC to resolve finer-scale structures. One potential workflow would be to initialize a master-event catalog using GC for high-quality reference locations, followed by incremental MME relative locations to track ongoing activity in near real time.\u003c/p\u003e\u003cp\u003eImportantly, the present results do not allow us to determine which method provides the most accurate absolute image of the fault system. Both MME and GC produce geologically reasonable and internally consistent structures, yet the true geometry of the swarm likely lies between the two images.\u003c/p\u003e"},{"header":"6 Conclusions","content":"\u003cp\u003eThe multi-master event (MME) method provides precise relative locations even for small earthquakes and is computationally efficient for long-term, large catalogs. Its performance depends on the quality of the selected master events and the accuracy of input absolute locations, whereas GrowClust (GC) systematically sharpens the seismicity distribution. Both methods yield consistent images of the southern and northern parts of the active Nov\u0026yacute; Kostel seismic zone, while the central segment shows the strongest methodological differences. Overall, the comparison demonstrates that the newly developed MME approach produces results consistent with the established GC method, confirming the robustness of both techniques. In addition, the updated catalog of relocated seismicity for 2000\u0026ndash;2025 provides the first comprehensive view of West Bohemia since the 2018 swarm, documenting not only repeated activity in the Nov\u0026yacute; Kostel zone but also swarms in peripheral areas such as Klingenthal and Františkovy L\u0026aacute;zně. This highlights the importance of continuous relocation efforts for understanding the temporal and spatial evolution of seismicity.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\n\u003cp\u003eD. K. conducted the research, developed the relocation code, prepared the figures 1,2,4,5,6,S1-10 and wrote the manuscript. J. D. contributed to the code development and revised the manuscript. B. R. assisted with the programming. J. B., the PhD supervisor of Diana Konr\u0026aacute;dov\u0026aacute;, revised the text and contributed to the preparation of figures 3,4,5,6,7.\u003c/p\u003e\n\u003ch2\u003eAcknowledgement\u003c/h2\u003e\n\u003cp\u003eThe authors would like to thank the late Josef Hor\u0026aacute;lek, who served as PhD supervisor and played a key role in shaping this research through his expertise and guidance. His dedication to understanding the West Bohemia/Vogtland region and his mentorship were an invaluable inspiration. The authors also wish to express their gratitude to Prof. RNDr. Tom\u0026aacute;\u0026scaron; Fischer, Ph.D., and Mgr. Josef Vlček, Ph.D., for their valuable assistance and discussions regarding the use of the GrowClust relocation method.The study was supported by the Grant Agency of Charles University in Prague, GAUK no. 291423 and by the Czech Science Foundation under grant GACR 25-16408X.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBabu\u0026scaron;ka, V., Růžek, B. \u0026amp; Dolej\u0026scaron;, D., 2016. \u003cem\u003eOrigin of earthquake swarms in the western Bohemian Massif: Is the mantle CO₂ degassing, followed by the Cheb Basin subsidence, an essential driving force?\u003c/em\u003e Tectonophysics, 668\u0026ndash;669, pp.42\u0026ndash;51. doi:10.1016/j.tecto.2015.12.008\u003c/li\u003e\n\u003cli\u003eBachura, M. \u0026amp; Fischer, T., 2019. \u003cem\u003eWaveform cross-correlation for differential time measurement: Bias and limitations\u003c/em\u003e. \u003cem\u003eSeismological Research Letters\u003c/em\u003e, 90(5), 2005\u0026ndash;2014. doi:10.1785/0220190096\u003c/li\u003e\n\u003cli\u003eBachura, M., Fischer, T., Doubravov\u0026aacute;, J. \u0026amp; Hor\u0026aacute;lek, J., 2021. \u003cem\u003eFrom earthquake swarm to a main shock\u0026ndash;aftershocks: the 2018 activity in West Bohemia/Vogtland.\u003c/em\u003e Geophysical Journal International, 224(3), pp. 1835\u0026ndash;1848. doi:10.1093/gji/ggaa523\u003c/li\u003e\n\u003cli\u003eBouchaala, F., Vavryčuk, V. \u0026amp; Fischer, T., 2013. \u003cem\u003eAccuracy of the master-event and double-difference locations: synthetic tests and application to seismicity in West Bohemia, Czech Republic.\u003c/em\u003e Journal of Seismology, 17, pp. 841\u0026ndash;859. doi:10.1007/s10950-013-9357-4\u003c/li\u003e\n\u003cli\u003eBr\u0026auml;uer, K., K\u0026auml;mpf, H., Niedermann, S., Strauch, G. \u0026amp; Weise, S. M., 2003. \u003cem\u003eEvidence for a nitrogen flux directly derived from the European subcontinental mantle\u003c/em\u003e. Nature, 424, 544\u0026ndash;547. doi:10.1038/nature01824\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eB\u0026uuml;y\u0026uuml;kakpınar, P., Dahm, T., Hainzl, S., Isken, M., Ohrnberger, M., Doubravov\u0026aacute;, J., Wendt, S., \u0026amp; Funke, S., 2025. \u003cem\u003eRecurring earthquake swarms: The role of magmatic fluids in the crust beneath northwestern Bohemia/Vogtland\u003c/em\u003e. \u003c/strong\u003e\u003cem\u003eCommunications Earth \u0026amp; Environment\u003c/em\u003e\u003cstrong\u003e (in revision).\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eDouglas, A., 1967. \u003c/strong\u003e\u003cem\u003eJoint epicentre determination.\u003c/em\u003e\u003cstrong\u003e Nature, 215, pp. 47\u0026ndash;48. doi:10.1038/215047a0\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eFischer, T. \u0026amp; Hor\u0026aacute;lek, J., 2000. \u003cem\u003eRefined locations of the swarm earthquakes in the Nov\u0026yacute; Kostel focal zone and spatial distribution of the January 1997 swarm in Western Bohemia, Czech Republic.\u003c/em\u003e Studia Geophysica et Geodaetica, 44, pp. 210\u0026ndash;226. doi:10.1023/A:1022162826079\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eFischer, T. \u0026amp; Hor\u0026aacute;lek, J., 2003. \u003c/strong\u003e\u003cem\u003eSpace-time distribution of earthquake swarms in the principal focal zone of the NW Bohemia/Vogtland seismoactive region: period 1985\u0026ndash;2001.\u003c/em\u003e\u003cstrong\u003e Journal of Geodynamics, 35(1\u0026ndash;2), pp. 125\u0026ndash;144. \u003c/strong\u003edoi:10.1016/S0264-3707(02)00058-3\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eFischer, T., Hor\u0026aacute;lek, J., Mich\u0026aacute;lek, J. \u0026amp; Bou\u0026scaron;kov\u0026aacute;, A., 2010. \u003c/strong\u003e\u003cem\u003eThe 2008 West Bohemia earthquake swarm in the light of the WEBNET network.\u003c/em\u003e\u003cstrong\u003e Journal of Seismology, 14, pp. 665\u0026ndash;682. doi:10.1007/s10950-010-9189-4\u003c/strong\u003e\u003c/li\u003e\n\u003cli\u003eFischer, T., Hor\u0026aacute;lek, J., Hrubcov\u0026aacute;, P., Vavryčuk, V., Br\u0026auml;uer, K. \u0026amp; K\u0026auml;mpf, H., 2014. \u003cem\u003eIntra-continental earthquake swarms in West-Bohemia and Vogtland: A review.\u003c/em\u003e Tectonophysics, 611, pp.1\u0026ndash;27. doi:10.1016/j.tecto.2013.11.001\u003c/li\u003e\n\u003cli\u003eGeissler, W. H., H. K\u0026auml;mpf, R. Kind, K. Br\u0026auml;uer, K. Klinge, T. Plenefisch, J. Hor\u0026aacute;lek, J. Zedn\u0026iacute;k, and V. Nehybka, 2005, \u003cem\u003eSeismic structure and location of a CO2 source in the upper mantle of the western Eger (Ohře) Rift, central Europe\u003c/em\u003e, Tectonics, 24, TC5001. doi:10.1029/2004TC001672\u003c/li\u003e\n\u003cli\u003eGot, J.L., Fr\u0026eacute;chet, J. \u0026amp; Klein, F.W., 1994. Deep fault plane geometry inferred from multiplet relative relocation beneath the south flank of Kilauea. \u003cem\u003eJournal of Geophysical Research: Solid Earth\u003c/em\u003e, 99(B8), pp.15375\u0026ndash;15386. doi:10.1029/94JB00577\u003c/li\u003e\n\u003cli\u003eHainzl, S., Fischer, T. \u0026amp; Dahm, T., 2012. \u003cem\u003eSeismicity-based estimation of the driving fluid pressure in the case of swarm activity in Western Bohemia.\u003c/em\u003e Geophysical Journal International, 191(1), pp.271\u0026ndash;281. doi:10.1111/j.1365-246X.2012.05610.x\u003c/li\u003e\n\u003cli\u003eHainzl, S., Fischer, T., Čerm\u0026aacute;kov\u0026aacute;, H., Bachura, M. \u0026amp; Vlček, J., 2016. \u003cem\u003eAftershocks triggered by fluid intrusion: Evidence for the aftershock sequence occurred 2014 in West Bohemia/Vogtland.\u003c/em\u003e Journal of Geophysical Research: Solid Earth, 121(4), pp. 2575\u0026ndash;2590. doi:10.1002/2015JB012582\u003c/li\u003e\n\u003cli\u003eHor\u0026aacute;lek, J. \u0026amp; Fischer, T., 2008. \u003cem\u003eRole of crustal fluids in triggering the West Bohemia/Vogtland earthquake swarms: Just what we know (a review).\u003c/em\u003e Studia Geophysica et Geodaetica, 52, pp. 455\u0026ndash;478. doi:10.1007/s11200-008-0032-0\u003c/li\u003e\n\u003cli\u003eInstitute of Geophysics, Academy of Sciences of the Czech Republic, 1991. \u003cem\u003eWest Bohemia Local Seismic Network\u003c/em\u003e [dataset]. International Federation of Digital Seismograph Networks.\u003c/li\u003e\n\u003cli\u003eJakoubkov\u0026aacute;, H., Hor\u0026aacute;lek, J. \u0026amp; Fischer, T., 2018. \u003cem\u003eMainshock\u0026ndash;aftershock activity versus earthquake swarms in West Bohemia, Czech Republic.\u003c/em\u003e Pure and Applied Geophysics, 175, pp. 109\u0026ndash;131. doi:10.1007/s00024-017-1679-7\u003c/li\u003e\n\u003cli\u003eLomax, A., Virieux, J., Volant, P. \u0026amp; Berge-Thierry, C., 2000. \u003cem\u003eProbabilistic Earthquake Location in 3D and Layered Models.\u003c/em\u003e In: C.H. Thurber \u0026amp; N. Rabinowitz, eds. \u003cem\u003eAdvances in Seismic Event Location.\u003c/em\u003e Modern Approaches in Geophysics, vol. 18. Dordrecht: Springer, pp. 101\u0026ndash;134. doi:10.1007/978-94-015-9536-0_5\u003c/li\u003e\n\u003cli\u003eM\u0026aacute;lek, J., Hor\u0026aacute;lek, J. \u0026amp; Jansk\u0026yacute;, J., 2005. \u003cem\u003eOne-Dimensional qP-Wave Velocity Model of the Upper Crust for the West Bohemia/Vogtland Earthquake Swarm Region.\u003c/em\u003e Studia Geophysica et Geodaetica, 49, pp. 501\u0026ndash;524. doi:10.1007/s11200-005-0024-2\u003c/li\u003e\n\u003cli\u003eRichards, P.G., Waldhauser, F., Schaff, D. \u0026amp; Kim, W.-Y., 2006. \u003cem\u003eThe applicability of modern methods of earthquake location.\u003c/em\u003e Pure and Applied Geophysics, 163, pp. 351\u0026ndash;372. doi:10.1007/s00024-005-0019-5\u003c/li\u003e\n\u003cli\u003eRubin, A.M., Gillard, D. \u0026amp; Got, J.L., 1998. \u003cem\u003eA reinterpretation of seismicity associated with the 1989 dike intrusion at Kilauea Volcano, Hawaii.\u003c/em\u003e Journal of Geophysical Research: Solid Earth, 103(B5), pp. 10003\u0026ndash;10015. doi:10.1029/97JB03513\u003c/li\u003e\n\u003cli\u003eShelly, D.R., Beroza, G.C., Ide, S. \u0026amp; Nakamula, S., 2006. \u003cem\u003eLow-frequency earthquakes in Shikoku, Japan, and their relationship to episodic tremor and slip.\u003c/em\u003e Nature, 442, pp. 188\u0026ndash;191. doi:10.1038/nature04931\u003c/li\u003e\n\u003cli\u003eStoddard, P.R. \u0026amp; Woods, M.T., 1990. \u003cem\u003eMaster event relocation of Gorda Block earthquakes: Implications for deformation.\u003c/em\u003e Geophysical Research Letters, 17(7), pp. 961\u0026ndash;964. doi:10.1029/GL017i007p00961\u003c/li\u003e\n\u003cli\u003eTarantola, A., 2005. \u003cem\u003eInverse problem theory and methods for model parameter estimation\u003c/em\u003e. Philadelphia, PA: Society for Industrial and Applied Mathematics (SIAM). doi:10.1137/1.9780898717921\u003c/li\u003e\n\u003cli\u003eTrugman, D.T. \u0026amp; Shearer, P.M., 2017. \u003cem\u003eGrowClust: A hierarchical clustering algorithm for relative earthquake relocation, with application to the Spanish Springs and Sheldon, Nevada, earthquake sequences.\u003c/em\u003e Seismological Research Letters, 88(2A), pp. 379\u0026ndash;391. doi:10.1785/0220160188\u003c/li\u003e\n\u003cli\u003eVavryčuk, V., 2011. \u003cem\u003ePrincipal earthquakes: Theory and observations from the 2008 West Bohemia swarm.\u003c/em\u003e Earth and Planetary Science Letters, 305(3\u0026ndash;4), pp. 290\u0026ndash;296. doi:10.1016/j.epsl.2011.03.002\u003c/li\u003e\n\u003cli\u003eWaldhauser, F. \u0026amp; Ellsworth, W. L., 2000. \u003cem\u003eA double-difference earthquake location algorithm: Method and application to the northern Hayward Fault, California.\u003c/em\u003e Bulletin of the Seismological Society of America, 90(6), pp.1353\u0026ndash;1368. doi:10.1785/0120000006\u003c/li\u003e\n\u003cli\u003eWeinlich, F. H., Br\u0026auml;uer, K., K\u0026auml;mpf, H., Strauch, G., Tesař, J. \u0026amp; Weise, S. M., 1999. \u003cem\u003eAn active subcontinental mantle volatile system in the western Eger rift, Central Europe: Gas flux, isotopic (He, C, N) and compositional fingerprints\u003c/em\u003e. Geochimica et Cosmochimica Acta, 63(21), 3653\u0026ndash;3671. doi:10.1016/S0016-7037(99)00187-8\u003c/li\u003e\n\u003cli\u003eWeinlich, F.H., Tesař, J., Weise, S.M., Br\u0026auml;uer, K. \u0026amp; K\u0026auml;mpf, H., 1998. \u003cem\u003eGas flux distribution in mineral springs and tectonic structure in the western Eger Rift.\u003c/em\u003e Journal of the Czech Geological Society, 43(1\u0026ndash;2), pp. 91\u0026ndash;110.\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"journal-of-seismology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jose","sideBox":"Learn more about [Journal of Seismology](http://link.springer.com/journal/10950)","snPcode":"10950","submissionUrl":"https://submission.nature.com/new-submission/10950/3","title":"Journal of Seismology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Relative location methods, West Bohemia/Vogtland, Master-event method, GrowClust method","lastPublishedDoi":"10.21203/rs.3.rs-7800475/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7800475/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe intraplate West Bohemia region represents one of the most prominent areas of recurrent earthquake swarm activity in Central Europe, with ongoing seismicity observed through 2025. We relocated this activity using the new multi-master event (MME) method and compared the results with GrowClust (GC) relative locations. The MME approach provides precise relative locations even for small events and can be applied efficiently to long-term, large catalogs. Its key advantage lies in the ability to incrementally incorporate new events without reprocessing the entire dataset, making it suitable for continuous and real-time monitoring. The comparison demonstrates that the newly developed MME approach produces results that are consistent with those of the established GC method. Both methods produce a consistent image of the southern and northern parts of the swarm, while the central segment of the active Nov\u0026yacute; Kostel seismic zone shows the strongest methodological differences. The MME results are sensitive to the quality of master-event locations, occasionally leading to scattered clusters where input locations are less reliable. In contrast, GC systematically sharpens the seismicity distribution and produces more compact structures. Overall, these results provide the first comprehensive view of the relocated seismicity in the West Bohemia region, extended through 2025, and reveal characteristic migration trends in the activity.\u003c/p\u003e","manuscriptTitle":"A Multi-Master Event Approach to Earthquake Relocation: Insights from the West Bohemia Swarm Zone (Czech Republic, 2000–2025)","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-27 14:28:46","doi":"10.21203/rs.3.rs-7800475/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-01-06T21:05:35+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-12-16T14:09:45+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-11-28T14:44:55+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"201547205188536558452193318966443277516","date":"2025-11-13T08:43:09+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"202153091130442212444489276932029561761","date":"2025-11-05T10:16:31+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-10-13T09:16:08+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-10-09T04:46:38+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-10-09T04:46:31+00:00","index":"","fulltext":""},{"type":"submitted","content":"Journal of Seismology","date":"2025-10-07T14:13:46+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"journal-of-seismology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"jose","sideBox":"Learn more about [Journal of Seismology](http://link.springer.com/journal/10950)","snPcode":"10950","submissionUrl":"https://submission.nature.com/new-submission/10950/3","title":"Journal of Seismology","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"ffc7c347-fe1a-4f05-932e-0bdb8ee0e476","owner":[],"postedDate":"October 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-04-28T14:09:36+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-27 14:28:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7800475","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7800475","identity":"rs-7800475","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}