Annual report for the Norwegian National Seismic Network

Transcription

1 Annual report for the Norwegian National Seismic Network 2016 Supported by University of Bergen and Norwegian Oil and Gas Association Prepared by Department of Earth Science University of Bergen Allegaten 41, N-5007 Bergen March 2017

2 2

3 CONTENTS 1 Introduction Operation NNSN NNSN field stations maintenance The NORSAR stations and arrays NNSN achievements and plans NNSN achievements in NNSN plans for Projects related to NNSN Seismicity of Norway and surrounding areas for Velocity models and magnitude relations Events recorded by the NNSN The seismicity of Norway and adjacent areas Scientific studies QLg tomography Relocation of seismicity in southern Norway and the North Sea using a Bayesian hierarchical multiple event location algorithm Testing of a method for distinguishing between earthquakes and explosions Towards the integration of NNSN stations in automatic network event location Publications and presentations of NNSN data during Publications Master degree thesis, UiB Oral presentations Poster presentations References... 51

4

5 1 Introduction This annual report for the Norwegian National Seismic Network (NNSN) covers operational aspects for the seismic stations contributing data, presents the seismic activity in the target areas and the associated scientific work carried out under the project. The report is prepared by the University of Bergen with contributions from NORSAR. The NNSN is supported by the oil industry through the Norwegian Oil and Gas Association and the University of Bergen (UiB). All the data stored in the NNSN database are available to the public via Internet, or on manual request. The main web-portal for earthquake information is It is possible to search interactively for specific data and then download the data from ftp://ftp.geo.uib.no/pub/seismo/data. Data are processed as soon as possible and updated lists of events recorded by Norwegian stations are available soon after recording. These pages are automatically updated with regular intervals. 2 Operation 2.1 NNSN The University of Bergen (UiB) has the main responsibility to run the NNSN and operates 34 of the seismic stations that form the NNSN located as seen in Figure 1. NORSAR operates 3 seismic arrays, which also include broadband instruments, and three single seismometer stations (JETT, JMIC and AKN). In addition to the NNSN stations, waveform data from selected stations in Finland (University of Helsinki), Denmark (GEUS), Sweden (University of Uppsala) and Great Britain (BGS) are transferred in real time and included in the NNSN database. More than 20 stations located in or operated by neighbouring countries are recorded continuously in Bergen and can be used for locating earthquakes, see Figure 1 and Figure 2. Phase data from neighbour countries and from arrays in Russia (Apatity), Finland (Finess), Sweden (Hagfors) are also included. In total, NORSAR provides data from 12 broadband stations to the NNSN. One station with real-time data is provided from the Ekofisk field by ConocoPhillips. The station HSPB is operated jointly between NORSAR and the Geophysical Institute, Polish Academy of Sciences, Warsaw, Poland and the stations BRBA and BRBB, both located in Barentsburg, Svalbard, are a collaboration between NORSAR and the Kola Science Centre, Russian Academy of Sciences, Apatity, Russia. The seismicity detected by the network is processed at UiB, however NORSAR also integrate their results into the joint database at UiB. At NORSAR the parameters of analyst-reviewed events are converted into parameter files in Nordic format and forwarded via ftp to UiB on a weekly basis. The magnitude threshold is set to about M=1.5 for regional events of potential interest for the NNSN. After transferring the parameter files, the NORSAR analyst logs into the UiB database using SEISAN and integrates the events. Integration means to merge 1

6 NORSAR and UiB events, which may require to repick seismic phases, to include new phase readings, to edit double phase readings and to relocate the seismic event with the new parameters. Figure 1. Stations contributing to the Norwegian National Seismic Network (NNSN). UiB operates 34 stations (red) and NORSAR operates the stations marked in blue, including the three arrays and stations AKN and JMIC. Seismic data recorded at stations located on Greenland and operated by GEUS are included in the NNSN real-time processing, Figure 2. These data are important for the location of earthquakes west of Jan Mayen and at the northern part of the Knipovich ridge to the Gakkel ridge. 2

7 Figure 2. Seismic stations in the arctic area. UiB is in the process of upgrading the NNSN by changing short period (SP) to broadband (BB) seismometers. The current status of this upgrade is shown in Table 1. As of today the numbers of SP, BB stations and stations with real time transmission are listed in Table 1. Table 1. Overview of UiB seismic stations Short Period Broadband Real time Number of stations 7 27 (24 with natural period greater than 100 sec) 31 (not real time are 2 short period and 1 broadband stations on Jan Mayen) The operational stability for each station is shown in Table 2. The down time is computed from the amount of data that are missing from the continuous recordings at UiB. This is done as the goal is to obtain as complete continuous data from all stations as possible. The statistics will, therefore, also show when a single component is not working. Also, communication or 3

8 computing problems at the centre will contribute to the overall downtime. In the case of communication problems, a station may not participate in the earthquake detection process, but the data can be used when it has been transferred. Thus, the statistics given allow us to evaluate the data availability when rerunning the earthquake detection not in real-time. The data completeness for the majority of the stations is above 95%, except for the following stations HYA and the three Jan Mayen stations (see technical service overview for details). Table 2. Data completeness in % for 2016 for all stations of the NNSN operated by UiB. Station Data completeness % Station Data completeness % Askøy (ASK) 100 Kongsberg (KONO) 98 Bergen (BER) 100 Konsvik (KONS) 100 Bjørnøya (BJO) 96 Lofoten (LOF) 100 Blåsjø (BLS) 100 Mo i Rana (MOR8) 99 Dombås (DOMB) 100 Molde (MOL) 99 Fauske (FAUS) 100 Namsos (NSS) 100 Florø (FOO) 97 Odda (OOD1) 99 Hammerfest (HAMF) Homborsund (HOMB) 100 Oslo (OSL) Skarslia (SKAR) 98 Hopen (HOPEN) 96 Snartemo (SNART) 99 Høyanger (HYA) 94 Stavanger (STAV) 100 Jan Mayen (JMI) Ca. 85 Steigen (STEI) 99 Jan Mayen (JNE) Ca. 50 Stokkvågen (STOK) 99 Jan Mayen (JNW) Ca. 50 Sulen (SUE) 99 Karmøy (KMY) 98 Blussuvoll (TBLU) 99 Kautokeino (KTK1) 100 Tromsø (TRO) 99 Kings Bay (KBS) 98 Vadsø (VADS) 100 4

9 2.2 NNSN field stations maintenance The technical changes for each seismic station are listed below. It is mentioned when these changes are carried out by the respective local contact and not by the staff of UiB. When a station stops working, tests are made to locate the problem. Sometimes the reason cannot be found and the cause of the problem will be marked as unknown. Major changes during this reporting period of 2016 were: Ask (ASK) Bergen (BER) Bjørnøya (BJO1) Blåsjø (BLS) Blussuvoll (TBLU) Dombås (DOMB) Fauske (FAUS) Florø (FOO) Hammerfest (HAMF) Homborsund (HOMB) Hopen : Station inspected to find fault with seismometer east component : Visit. Cable to EW component severed, a spare wire pair was used to repair. No visit or technical changes : Communication down since Data from this time period could not be retrieved because the filesystem on the USB memory was corrupted. All data from this time period has been lost. No visit or technical changes No visit or technical changes : Station down since at 00:08 (UTC). Unknown reason. Power on/off restarted the station : Station down since at 22:30 (UTC). Unknown reason. Power on/off restarted the station : Vault checked by local contact : El. enclosure was inspected by local contact. No humidity or other problems were detected : Station was visited by UiB staff. Small amount of water was removed from the vault : The station vault was inspected by the local contact. No visit or technical changes. No visit or technical changes : Station down since due to power-loss. All data lost : UPS installed by local personnel. 5

10 (HOPEN) Høyanger (HYA) Jan Mayen Trolldalen (JMI) Jan Mayen Ulla (JNE) : The vault was inspected by local personnel. No problems found : Station down since due to PC problems after heavy storm. PC replaced by local contact. Data lost : Station down since due to power failure. Data lost. No technical changes. The station is visited by local personnel. The cable to the equipment developed a fault during November. August 2016: The digitizer for JNE and JNW started showing signs of malfunction. November 2016: The station is visited by local personnel who found that the windmill did not produce any power due to a broken cable. The problem was fixed locally. Jan Mayen (Liberg) JNW (Photo: Local staff at Jan Mayen) September 2016: No technical changes. The station is visited by local personnel. (Photo: Local staff at Jan Mayen) Karmøy (KMY) Kautokeino (KTK) Kings Bay : Visit. New broad band sensor (Trillium 120QA) installed : USB memory connected by local contact. Formatted remotely and put into use : Seismometer not working since Local contact visited and repowered the station, which resolved the problem. Data lost. No visit or technical changes. 6

11 (KBS) Kongsberg (KONO) Konsvik (KONS) Lofoten (LOF) Mo i Rana (MOR8) Molde (MOL) Namsos (NSS) Odda (ODD1) Oslo (OSL) Skarslia (SKAR) Snartemo (SNART) Stavanger (STAV) Steigen (STEI) Stokkvågen (STOK) Sulen (SUE) The USGS plans to move the station to a different site within the mine. Currently they are still in the process of arranging for the work that needs to be done : Station visited, no changes. No visit or technical changes : New UPS installed by local operator. No visit or technical changes. No visit or technical changes : New broad band seismometer (Trillium 120QA) installed. The digitizer (Guralp) was not changed : Visit by local contact. USB memory connected to digitizer. Trees near the sensor were cut to prevent noise. No visit or technical changes : Vault inspection. A small amount of water was removed : Station down since due to problem with the digitizer. Data lost : Station down since at 04:30 UTC. A digitizer problem. Power off/on and station ok from 08: : Station down from Problem with digitizer. The digitizer was replaced by local contact. A small amount of water was removed. No visit or technical changes. No visit or technical changes : Station was shortly visited by UiB staff who had fieldwork in the area : Station visited, no changes : Station down some hours due to power failure. Data lost : Visit. A new Guralp digitizer was installed, and old digitizer and PC were removed. There had been timing problems for some time and a 7

12 new GPS antenna was installed. Tromsø (TRO) Vadsø (VADS) : Station down since Power reset remotely. Data lost : The station was installed with the following equipment: Trillium 120PA sensor and a Guralp CMG-DM24 digitizer. 2.3 The NORSAR stations and arrays NORSAR is operating the following installations: NOA (southern Norway, array, 42 sites, 7 3C broadband sensors and 35 vertical broadband sensors) ARCES (Finmark, array, 25 sites, 25 3C broadband seismic sensors, 9 infrasound sensors) SPITS (Spitsbergen, array, 9 sites, 6 3C broadband sensors and 3 vertical broadband sensors) NORES (Hedmark, array, 12 3C short-period sensors, 9 infrasound sensors) JMIC (Jan Mayen, 3C broadband sensor) AKN (Åknes, Møre og Romsdal, 3C broadband sensor) TROLL (Antarctica, 3C broadband sensor) IS37 (Bardufoss, infrasound array, 10 sites) JETT (Jettan, Troms, 3C broadband sensor) I37H0 (Bardufoss, 3C broadband sensor) In addition NORSAR receives and processes data in near realtime from: FINES (southern Finland, array, 16 sites, 2 3C broadband sensor, 1 3C short-period sensor and 15 short-period vertical sensors, operated by Institute of Seismology, Helsinki, Finland) HFS (Hagfors, Sweden, 10 sites, 1 3C broadband sensor and 9 short-period vertical sensors, operated by the Swedish Defence Reseearch Agency, Stockholm, Sweden) EKA (Eskdalemuir, United Kingdom, 20 sites, 1 3C broadband sensor and 20 shortperiod vertical sensors, operated by the United Kingdom National Data Centre, AWE Blacknest, UK) BRBA, BRBB (Barentsburg, 2 3C broadband sensors and 3 infrasound sensors) APA (Apatity seismic array, parametric data) All NORSAR waveform data and parametric data are openly available and can be accessed through web-interfaces or direct means. The NORSAR webpage provides access to general station information, to automatic and reviewed seismic bulletins, to realtime plots of short and long-period data, and to an AutoDRM request form for waveform data retrieval. The seismic array data are automatically processed and analysed. The fastest near realtime process Automatic Alert is based on single array detection and provides event locations within a few (1-3) minutes delay. The alerts with event and location details are published immediately on (which is also integrated 8

13 into the NNSN website). A second automatic process called GBF (Generalized Beam Forming) awaits for automatic phase picks from all arrays and delivers more reliable/accurate results within up to a few hours delay. Automatically processed seismic events with magnitude larger than 2 (or 1.5 if the event is of special interest) are manually analysed and reviewed. In this step all available waveforms (also from single stations) are utilized. Graphical displays and parametric event data and for Automatic Alert, GBF and Reviewed bulletins can be found on Figure 3. NORSAR seismic arrays/stations (NOA, NORES, ARCES, SPITS, JMIC, AKN, I37H0) and contributing arrays/stations (HFS, FINES, EKA, BRBA, APA). Changes during 2016 a) In 2016 we completed the renovations of the ARCES array. Sensors, digitizers and central acquisition system had been upgraded in The main tasks in 2016 was the refurbishment of the Central Recording Facility (CRF), the installation of an uninterruptable power supply (UPS), finalizing of the central acquisition rack and the 9

14 official revalidation of the ARCES/PS28 primary station of the international monitoring sytem (IMS). Figure 4 shows the renovated CRF with satellite, cell-network communication antenna, GPS-antenna for central timing on the gable and the mains power transformer in the background. Figure 4. The renovated Central Recording Facility (CRF) of the ARCES array b) We installed three more sites of the NORES array. It consists now of 12 3-component seismometers and 9 infrasound sensors The A-and B-ring are equipped with 9 seismometers and 9 infrasound sensors. Three out of the 7 sites of the C-ring got seismometers and the remaining 4 sites will be installed during The research facility Stendammen in the centre of the NORES array has been equipped with a small workshop and it provides accommodation for 2 persons. c) NORSAR s IMS infrasound station in Bardufoss IS37 has been upgraded with a second infrasound sensor at each site in order to facilitate remote calibration of the entire system (i.e. sensors and wind noise reduction system). As a consequence, we could not use anymore a common digitizer for infrasound sensor and seismometer, and had to install a separate digitizer for the seismic sensor at site H0 (see picture below). Figure 5. The instruments at the Bardufoss site. 10

15 Data availability All data recorded at NORSAR are continuous. The following table provides a monthly overview on the data availability of 13 main data streams provided by NORSAR to NNSN. Table 3. Systems recording performance (in % of data completeness) for 14 main data streams provided from NORSAR to NNSN. ARA0 JMIC NAO01 NBO00 NB201 NC204 NC303 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec NC405 NC602 SPA0 AKN JETT HFC2 I37H0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Detections The NORSAR analysis results are based on automatic phase detection and automatic phase associations which produce the automatic bulletin. Based on the automatic bulletin a manual analysis of the data is done, resulting in the reviewed bulletin. The automatic bulletin for northern Europe is created using the Generalized Beam Forming (GBF) method. This bulletin ( is subsequently screened for local and regional events of interest in Fennoscadia and in Norway, which in turn are reviewed by an analyst. Regional reviewed bulletins from NORSAR are available from 1989 and from 1998 onwards they are directly accessible from via internet 11

16 ( Table 4 gives a summary of the phase detections and events declared by GBF and the analyst. Table 4. Phase detections and event summary. Jan. Feb. March April May June Phase detections Associated phases Un-associated phases Screened GBF events for Fennoscandia/Norway No. of events defined by the analyst July Aug. Sep. October Nov. Dec. Phase detections Associated phases Un-associated phases Screened GBF events for Fennoscandia/Norway No. of events defined by the analyst

17 3 NNSN achievements and plans The overall purpose of the NNSN is to provide data both for scientific studies, but equally important for the routine observation of earthquakes. This in principle means that broadband seismometers are desired at all sites. However, in areas where additional stations are deployed for local monitoring, short-period seismometers are sufficient. The number of broadband seismometers in the network will be increased to replace existing short period instruments. A general goal for the future development has to be to achieve better standardization in particular with the seismometers and digitizers. The total number of stations for now should remain stable, but it is important to improve the overall network performance. 3.1 NNSN achievements in 2016 The new broadband station near Vadsø (VADS) on the Varanger peninsula has been completed and the first event was recorded October 27 th at 03:53 (UTC). The two short-period stations ODD1 and KMY have been upgraded with broadband seismometers. The archiving procedure at UiB has been modified, which together with improved station robustness has resulted in increased data completeness. The UiB research has focussed on Lg wave attenuation tomography. The UiB magnitude study for the North Atlantic is being finalized and a paper will be submitted in NORSAR have continued their studies on event re-location and source discrimination. Under the EPOS project, planning and preparation for the 6 Svalbard, 7 Nordland and 3 Jan Mayen stations has started, the seismic instruments have been purchased. Stations deployed under the NEONOR2 project have been uninstalled, detailed processing of the earthquake swarm near Jektvik that started in April 2015 has been carried out. The data are integrated with the NNSN stations and are part of the NNSN database. 3.2 NNSN plans for 2017 Upgrade three more stations with broadband seismometers. The Kongsberg station may be upgraded by the USGS. Implement the new magnitude scale into the processing routines. UiB plans to carry out research on source parameters, automatic detection and determination of fault plane solutions. A research workshop will be held between UiB and NORSAR early in 2017 (done). The research and development activity will continue in close collaboration between UiB and NORSAR. Provide data through IRIS and through a European EIDA node at UiB under the EPOS-Norway project. 13

18 Strengthen the collaboration with NORSAR and the other Nordic countries on data processing through technical visits. Integrate more real-time continuous data from stations located in Sweden and Iceland if available. Improve macroseismic questionnaire in collaboration with other Scandinavian countries. Make the macroseismic data available on the website through the EPOS-Norway project. Jan Mayen data will be send to UiB in real-time. Continue collaboration with other Nordic countries to collect information about historical felt earthquakes. 3.3 Projects related to NNSN Status update for EPOS-IP and EPOS-N European Plate Observing System Implementation Phase (EPOS-IP) ( EPOS is currently well underway in the Implementation Phase (EPOS-IP: ), heading towards the Operational Phase (EPOS-OP) from 2020 onwards. The ongoing Implementation Phase aims to implement key structural pillars of the project. During 2016, it was decided that the EPOS-ERIC legal seat will be at the INGV headquarters in Rome, Italy, and will be operational in Considerable progress has been made in the IT architecture and development of the interoperable services. ICS developments (software for the EPOS portal) include an integration of the web components and metadata harvesting. The thematic core services have undergone a community development by adopting their metadata to the baseline model and providing services for their data. In June 13-15, 2017 a dedicated Nordic EPOS Conference is planned in Helsinki hosted by the University of Helsinki. The purpose of this conference is to bring together various EPOS contributors from the Nordic countries and find possible collaboration points and provide synergies. European Plate Observing System Norway (EPOS-N) ( The EPOS-Norway (EPOS-N) project shares the goals and visions of the EPOS project about addressing key challenges in Earth science, but with special attention to the Arctic. In January 2017, EPOS-N arranged its 2 nd annual workshop in Bergen as a conclusion of the first year for the EPOS-N project. Considerable progress has been made on a number of fields. The Enlighten software has been chosen as the engine behind the Norwegian EPOS web-portal. Implementation of macroseismic data is ongoing through the MIDOP service. EPOS-N has applied for EIDA membership to provide the NNSN and EPOS data. The Solid Earth Science Forum (SESF) is in the establishment phase, and a committee (SESC) consisting of representatives from six partner institutions has been established to coordinate this forum. An External Advisory Board (EAB) has been established, with five members from both Norway and other European countries. EAB also had its first meeting during this year s EPOS-N Annual Workshop, where they provided comments to the project progress and the Annual Report, and selected a chair. The first Annual Report was submitted in January

19 Upcoming challenges include construction of the monitoring stations in the Arctic. A reconnaissance survey around Svalbard is planned this summer, where potential sites for new stations will be inspected. Procurement of seismological equipment is already completed and the instruments are shipped to locations in Norway, Svalbard and Jan Mayen Historical earthquake database for Norway Norwegian earthquakes have been documented at least back to the mid-17th century. A rich dataset on historical earthquakes is available in terms of eye-witness reports, newspaper articles, macroseismic questionnaires and published and unpublished macroseismic intensity maps. However, this data has not been compiled and assessed systematically in terms of macroseismic intensities. In 2016, an effort was initiated to compile all available intensity data for Norway and make it available through an online system. Currently, all data which were available in the NNSN database have been included in the new system, and a prototype is available online ( The online database is developed using the MIDOP tool which has been developed in relation to the European Archive of Historical Earthquake Data (AHEAD) to encourage the adoption of common compilation standards and compatible data formats among macroseismic intensity data compilers. All available data for events before 1900 will be integrated in the European AHEAD database when ready. The database allows searching for information by earthquake (showing all intensities observed for a given earthquake) or by locality (showing all intensities observed in different earthquakes at a given locality). Selected features of the data access system are shown in Figure 6 and Figure 7. Figure 6. Entry page when accessing information by earthquake. The upper left table presents a list of events for which intensity data is available. The event locations are shown in the map. 15

20 Figure 7. Intensity data for a specific earthquake. The lower left table presents the intensity data for each locality, which is also shown in the map. Currently, the focus is on including all available data from other sources. We are systematically working through all maps and tables with intensity data available in the published literature and the UiB archive, and digitizing the information. This will add a significant amount of new data to the database which will, when finalized, contain information at least back to In addition to making the valuable historical earthquake dataset available to the wider community, this effort will help improving the completeness of the NNSN database for historical events. 4 Seismicity of Norway and surrounding areas for 2016 The earthquake locations presented have been compiled from all available seismic stations as described above. All phase data are collected by UiB and all located local and regional earthquakes recorded on NNSN stations are presented on the web pages. The largest are also ed to the European-Mediterranean Seismological Centre (EMSC) to be published on the EMSC web pages. When all available data is collected, a monthly bulletin is prepared and distributed. A brief overview of the events published in the monthly bulletins is given in this annual report. Macroseismic data for the largest felt earthquakes in Norway are collected, and macroseismic maps are presented. Local, regional and teleseismic events that are detected by the UiB network are included. The merging of data between NORSAR and UiB is based on the following principles: i) All local and regional events recorded by NORSAR that are also detected by the NNSN network are included. ii) Local and regional events with local magnitude larger than 1.5 detected by NORSAR and not recorded by the NNSN are included. However, probable explosions from the Kiruna/Malmberget area are not included. 16

21 iii) All teleseismic events recorded by NORSAR and also detected by the NNSN are included. iv) All teleseismic events with NORSAR magnitude M b 5.0 are included even not detected by the NNSN. Data from the British Geological Survey (BGS) and the Geological Survey of Denmark and Greenland (GEUS) are included in the database in Bergen following similar criteria as mentioned above, however only events located in the prime area of interest, N and 15 W-35 E, and with magnitude 2.0 are included. From the Greenland area only earthquakes recorded on NNSN stations are included. Phase data and locations from University of Helsinki and University of Uppsala are included NNSN database to improve NNSN locations for events in the eastern parts of Norway or possibly for larger events elsewhere. Many of the recorded events are explosions. To discriminate between natural earthquakes and manmade explosions, spectrograms are used in the daily routine processing. This was implemented in the processing in Bergen during spring Velocity models and magnitude relations The velocity model used for locating all local and regional events, except for the local Jan Mayen events, is shown in Table 5 (Havskov and Bungum, 1987). Event locations are performed using the HYPOCENTER program (Lienert and Havskov, 1995) and all processing is performed using the SEISAN data analysis software (Havskov and Ottemöller, 1999). Table 5. Velocity model used for locating all local and regional events, except for the local Jan Mayen events (Havskov and Bungum, 1987). P-wave velocity Depth to layer (km/sec) interface (km) Local magnitude M L is computed for all earthquakes based on measuring instrument corrected ground amplitudes A (nm) and applying the M L scale by Alsaker et al. (1991): 17

22 M L = log (A) log(d) D where D is the hypocentral distance in km. The moment magnitude M w is calculated for selected earthquakes on mainland Norway from the seismic moment M 0 using the relation (Kanamori, 1977) M w = 0.67 log(m 0 ) 6.06 The unit of M 0 is Nm. The seismic moment is calculated from standard spectral analysis assuming the Brune model (Brune, 1970) and using the following parameters: Density: 3.0 g/cm 2 Q = 440 f 0.7 P-velocity = 6.2 km/s S velocity = 3.6 km/s In the analysis, the seismic moment is measured from attenuation corrected source displacement spectra (Havskov and Ottemöller, 2003). For the Jan Mayen area, a local velocity model (see Table 6) and coda magnitude scale is used (Andersen, 1987). Table 6. Velocity model used for locating local Jan Mayen events. P-wave velocity Depth to layer (km/sec) interface (km) The regional and teleseismic events recorded by the network are located using the global velocity model IASPEI91 (Kennett and Engdahl, 1991). Body wave magnitude is calculated using the equation by Veith and Clawson (1972): Mb = log(a/t) + Q(D,h) Here h is the hypocentre depth (km), A is the amplitude (microns), T is period in seconds and Q(D,h) is a correction for distance and depth. Surface wave magnitude Ms is calculated using the equation (Karnik et al., 1962): Ms = log(a/t) log(d) where A is the amplitude (microns), T is period in seconds and D is the hypocentral distance in degrees. Starting from January 2001, the European Macroseismic Scale, EMS98, (Grünthal, 1998) has been used. All macroseismic intensities mentioned in the text will refer to the EMS98 instead 18

23 of the previously used Modified Mercalli Intensity scale. The two scales are very similar at the lower end of the scale for intensities less than VII. 4.2 Events recorded by the NNSN Based on the criteria mentioned above, a total of 8,182 local and regional events, were detected by the NNSN during Of these local and regional events, 34% were large enough to be recorded by several stations and hence could be located reliably, and are not classified as explosions (LP or LE). The numbers of local/regional and teleseismic events, recorded per month in 2016 are shown in Figure JAN MAR MAY JUL SEP NOV Figure 8. The number of recorded local/regional (blue) and teleseismic (red) events during The average number of local and regional events recorded per month is 682 (570 in 2015). A total of 1159 teleseismic events were recorded in 2016 and the monthly average of teleseismic earthquakes in the NNSN database, is 96. In addition to the locations determined at UiB and NORSAR, also preliminary locations published by the USGS (United States Geological Survey) or the EMSC (European Mediterranean Seismological Centre) based on the worldwide network are included for earthquakes registered by NNSN stations. During the years there has been an increase in the number of local/regional events recorded into the NNSN database. As can be seen from Figure 9 the number of teleseismic (D) earthquakes recorded are relative stable while the number of local/regional (L/R) events have increased the last four years. The temporary stations in the NEONOR project can explain some of the increase. The number of recorded earthquakes is expected to continue to increase with the planned installation of new stations on Svalbard. 19

24 L/R D Figure 9. Number of local/regional (blue) and teleseismic (pink) events recorded in the NNSN database since 2000 UiB, as an observatory in the global network of seismological observatories, reports local and teleseismic phases to the International Seismological Center (ISC). All events (teleseismic, regional and local) recorded from January to December 2016 with M 3 are plotted in Figure 10. Figure 10. Epicentre distribution of earthquakes with M 3.0, located by the NNSN from January to December Teleseismic events recorded only by NORSAR have M 5.0. Monthly station recording statistics from January to December 2016 are given in Table 6 and 7. This table shows, for each station, local events recorded on more than one station and recorded teleseismic events. The statistics are based on the analysed data and are taken from the database. Table 6 and 7 show both earthquakes and explosions. Identified or suspected explosions will only be located with a minimum number of stations. Therefor some stations (e.g. KTK, MOR8, VADS, FAUS) will have a higher number of detections. The following was observed from Table 6 and 7: At Jan Mayen there was a problem with the digitizer which reduced the number of earthquakes triggered since summer TBLU and OSL are recording mostly teleseismic earthquakes, which is as expected due to their location in noisy environment. Stronger local earthquakes will, however, be detected. 20

25 The new station in Vadsø (VADS) was operational since October The stations KONS, STOK and MOR8 continue to record a relatively large number of small earthquakes and explosions in the area. There are no teleseismic detections on JMI, JNE and JNW as currently the system on Jan Mayen is only detecting local events, and realtime data is not available at UiB. Table 7. Monthly statistics of events recorded at each station for January-June Abbreviations are: L = Number of local events recorded at more than one station and D = Number of teleseismic events recorded at the station. JANUARY FEBRUARY MARCH APRIL MAY JUNE STATION L D L D L D L D L D L D ASK BER BJO BLS DOMB FAUS FOO HAMF HOMB HOPEN HYA JMI JMIC JNE JNW KBS KMY KONO KONS KTK LOF MOL MOR NSS ODD OSL SKAR SNART STAV STEI STOK SUE TBLU TRO VADS AKN JETT NORSAR ARCES SPITS

26 Table 8. Monthly statistics of events recorded at each station for July-December Abbreviations are: L = Number of local events recorded at more than one station and D = Number of teleseismic events recorded at the station. JULY AUGUST SEPT OCT NOV DEC STATION L D L D L D L D L D L D ASK BER BJO BLS DOMB FAUS FOO HAMF HOMB HOPEN HYA JMI JMIC JNE JNW KBS KMY KONO KONS KTK LOF MOL MOR NSS ODD OSL SKAR SNART STAV STEI STOK SUE TBLU TRO VADS AKN JETT NORSAR ARCES SPITS The seismicity of Norway and adjacent areas This section first gives an overview of the seismicity in the monitoring area before presenting the activity in specific areas in more detail. The main area of interest is defined as 54-82N and 15W-35E, Figure 11. We also show the seismicity for the Arctic region including the Barents Sea defined by coordinates N and 25 W-50 E. A total of 5074 of the recorded events are located inside the NNSN prime area. During analysis and using the explosion filter (Ottemöller, 1995), 54% of these events were identified as confirmed or probable explosions, or induced events. Figure 11 shows all local/regional events in the prime area, analysed and located during Among these, 199 are located in the vicinity of the Jan Mayen Island. 22

27 Figure 11. Epicentre distribution of events analysed and located in Earthquakes are plotted in red. Probable and confirmed explosions and induced events are plotted in blue. For station locations, see Figure 1. It should be emphasized that the magnitude calculation for the earthquakes located on the oceanic ridge in the Norwegian Sea uses the same formula as for mainland Norway. As the scale is not appropriate for this region, the magnitudes for these earthquakes are underestimated. A new magnitude scale has been developed, but is not yet applied routinely. Most of the recorded earthquakes in this area have magnitudes above 3.0 if they are recorded on Norwegian mainland stations. Figure 12 shows the location of earthquakes (induced events, known and probable explosions removed) located within the prime area with one of the calculated magnitudes above 3. Table 9 lists the same earthquakes with all earthquakes located close to the Mid-Atlantic ridge removed. 23

28 Figure 12. Epicentre distribution of located events with one of the calculated magnitudes above or equal to 3.0. For station location, see Figure 1. The largest local or regional earthquake in 2016, recorded on Norwegian stations and within the prime area, was a double event that occurred on March 29 th at 10:32:10 (UTC) and 10:32:38 (UTC) in Storfjorden, west of Edgeøya. The earthquakes has magnitudes of M L(BER) =5.2 and M L(BER) =4.8, respectively. Seismograms for the two earthquakes recorded at Barentsburg (BRBA), Hornsund (HSPB), Hopen (HOPEN) and Kings Bay (KBS), are shown in Figure 13. The P-wave onset for the first earthquake and the S-wave onset from both earthquakes are clearly seen. These earthquakes were followed by several aftershocks, where six had an estimated magnitude above

29 Table 9. Earthquakes located in the vicinity of mainland Norway and in the Svalbard area (grey background) with any reported magnitude above or equal to 3.0 for the time period January through December In cases where all BER magnitudes are below 3 but the event still is included in the list, NORSAR (NAO), GEUS- Geological Survey of Denmark and Greenland (DNK), University of Uppsala (UPP), University of Helsinki (HEL) or the British Geological Survey (BGS) has reported a magnitude of 3.0 or larger. Abbreviations are: HR = hour (UTC), MM = minutes, Sec = seconds, L = distance identification (L=local, R=regional, D=teleseismic), Latitud = latitude, Longitud = longitude, Depth = focal depth (km), F = fixed depth, AGA = agency (BER=Bergen), NST = number of stations, RMS = root mean square of the travel-time residuals, Ml = local magnitude and Mw = moment magnitude. Year Date HRMM Sec L Latitud Longitud Depth F AGA NST RMS Ml Mw ML ML NAO L BER L BER L BER L BER L BER L BER L BER L BER L BER UPP L BER L BER L F BER L BER L BER L BER L BER L BER L BER HEL L BER L BER L BER BGS L BER L BER L BER L F BER L F BER L BER L BER L F BER L F BER L F BER L F BER L BER L F BER L F BER EMSC L BER L BER L BER L BER L BER L BER L BER L BER L F BER L BER L F BER BGS L F BER L BER L BER HEL 25

30 Figure 13. Seismogram of the two earthquakes felt at Svalbard in March The P-phases for the first earthquake are marked. Another significant earthquake occurred March 19 th at 21:55 (UTC) in the inner western part of Bottenvika. Sweden. The University of Uppsala reported a magnitude M L(UPP) =4.1 and the calculated magnitude from NNSN is M L(BER) =4.2. Along the mid-atlantic ridge a large number of earthquakes occurred with magnitude above 3. Of the 19 located earthquakes with calculated magnitude above M L =4.0, 13 were located along the Mohns ridge and at the northern part of the Knipovich ridge. The magnitudes calculated for earthquakes in the Norwegian-Greenland Sea are expected to be underestimated. 26

31 With the mainland stations on the Lofoten, in Tromsø and Hammerfest, the network on Jan Mayen, and stations on Bjørnøya, Hopen and Svalbard, the network detection capability in the arctic area is relatively good. We define the arctic area as the region N and 25 W-50 E. Most of the activity falls into three areas: Jan Mayen, the Mid-Atlantic ridge and Storfjorden southeast of Svalbard, as can be seen in Figure 14. Since 2014 data from Danish stations on Greenland (see Figure 2) were included in the daily processing which has increased the location capability for earthquakes west of Jan Mayen and northwest of Svalbard. The number of earthquakes recorded on enough stations to be located, has increased. Figure 14. Seismicity in the Norwegian arctic area during A total of 2027 located earthquakes Seismicity in Nordland Figure 15 shows the seismicity in Nordland since The area includes the locations of earthquake swarm activity such as Meløy (66.8N, 13.5E), Steigen (67.8N, 15.1E) and Stokkvågen (66.3N, 13.1). The Steigen area was more active between 2007 and 2008, then it was relatively quiet until 2015 when 44 earthquakes appeared. During spring 2016 the temporary stations deployed during the NEONOR2 project were stopped. The decreased station-density in the area has resulted in fewer events with low magnitude being located (Figure 16). 27

32 Figure 15. Seismicity in the Nordland area. Blue circles show seismicity for 2016, red circles show seismicity for , and yellow triangle is NNSN seismic stations. Only probably earthquakes are included. In the Jektvik (66.6N,13.5E) area there has been a decrease in seismic activity during 2016 compared to 2015 (Figure 16). During 2016, 295 earthquakes were located here, compared to 476 in The largest earthquake in the area occurred December 29 th, 2016 at 04:29 (UTC) with a magnitude of M L(BER) =3.0. The earthquake was felt. The yearly distribution of earthquakes located in the area is presented in Figure

33 Figure 16 Time distribution (upper) and location (lower) of the earthquake swarm located in the Jektvik area, southwest of the Svartisen glacier. Earthquakes occurring in 2016 are marked in blue. Note! The events shown are limited by N and E, a smaller area than show on the map. The location of the permanent NNSN station (KONS) and the NEONOR2 temporary stations (N2VG, NBB13, NBB 15, NBB17) are marked on the map. 29

34 Figure 17 The yearly number of the earthquakes in the area N and E as seen in Figure Seismicity in the Jan Mayen area Jan Mayen is located in an active tectonic area with two major structures, the Mid Atlantic ridge and the Jan Mayen fracture zone, interacting in the vicinity of the island. Due to both tectonic and magmatic activity in the area, the number of recorded earthquakes is higher than in other areas covered by Norwegian seismic stations. During 2016 a total of 199 earthquakes were located as seen in Figure 18 and of these, 6 had a magnitude equal or above 3.0. During the fall of 2016 the digitizer used to register data from JNW and JNE started to malfunction. The number of located earthquakes in the Jan Mayen area are therefore reduced, as can be seen from the monthly distribution of earthquakes in Figure 18. The largest earthquake in the Jan Mayen region in 2016 occurred on 3 rd September at 06:55 (UTC) and the magnitude is estimated to 4.2L(BER) and 5.4L(NAO). The earthquake is located slightly southeast of the Beerenberg volcano and was felt by the personnel at Jan Mayen. By the personnel at Jan Mayen the earthquake was described as ( «Klokken i morges merket de fleste på stasjonen at det begynte å riste. Jordskjelvet var merkbart i rundt 30 sekunder og ble registrert av seismografene som vi har rundt på øya.» 30

35 Figure 18. Earthquakes located in the vicinity of Jan Mayen during The time distribution is shown in the upper part. The reduced seismic activity observed for the last part of 2016 might be due to a malfunctioning digitizer. The number of recorded earthquakes in the Jan Mayen area has varied over the last years (Figure 19). The number of relatively strong earthquakes (M 3) shows smaller time variation than for the smaller earthquakes. The increases in 2004 and 2005 were due to the M=6.0 earthquake in 2004 and its aftershocks (Sørensen et al., 2007). The same is true for 2011, where the M=6.0 earthquake on 29 January was followed by a sequence of aftershocks. The 30 August 2012 earthquake (M=6.3) with its fore and aftershocks clearly increases the number of recorded events in 2012 compared with previous years, making it the largest number of recorded events yearly for more than 10 years. For the following years after 2012, 31

36 the number of located smaller earthquakes has increased slightly, while the number of larger (M 3.0) earthquakes is relative stable Total number of recorded earthquakes Number of earthquakes with magnitude larger or equal to Figure 19. Yearly distribution of earthquakes located in the Jan Mayen area since The area is as shown in Figure Seismicity at Svalbard The seismicity in the Svalbard area is presented in Figure 20, showing both a map with the seismicity since 2000 and the distribution of events over time. There are several seismically active areas in this region. This report will focus on three main areas: the Storfjorden area including Heer Land (on the northwest side of Storfjorden) and Diskobukta (the area west of Egdeøya), Sørkappland (the area at the very southwest coast of Spitsbergen) and Nordaustlandet. These will now be discussed in more detail. 32

37 Figure 20. Seismicity in the Svalbard area. Bottom: Earthquakes occurring in 2016 are plotted in red circles. Yellow circles show seismicity for The blue triangles give the station locations. Top: Seismicity in the same area is plotted as latitude as function of time. 33

38 Storfjorden The Storfjorden area southeast of Svalbard, defined by latitude N and longitude 16-22E, has been more seismically active since the M w =6.0 earthquake on 21 February The earthquake was the starting point of a prolonged earthquake sequence. The yearly variation in the number of detected earthquakes in Storfjorden area is shown in Figure 20. An increase in the number of located earthquakes is clearly seen in 2008 explained by the M6 earthquake and its aftershocks M 3.0 All earthquakes Figure 21. Yearly number of earthquakes located to the Storfjorden area. The increase in 2010 is mostly explained by an increase in the number of smaller earthquakes as seen by the almost constant levels of earthquakes with M>3.0 since The better detection was due to usage of the data from the Hornsund (HSPB) and Barentsburg stations, from 2010 and 2012, respectively. A total of more than 39 earthquakes with magnitude larger than M=4 have occurred in the area since There is a clear increase in 2016 related to the activity in both Storfjorden and the Heer Land region. Heer Land has been seen to be active in the past, going back to the 1970s. The total number of events located in the Storfjorden area in 2016 was 505, which presents an increase from 2015 when the corresponding number is 176. Figure 20 shows both a map with the seismicity since 2000 and the distribution of events over time. One area where an increase in seismic activity is clearly seen is Diskobukta ( N, E), where not much seismicity had been seen previously. Searching in the NNSN database, there are only located 12 earthquakes, with one of the reported magnitudes above 3.0, in this area. The first located in Eight of these 12 earthquakes are recorded in During 2016, 121 earthquakes are located to this small area compared to 11 in The largest earthquake occurred March 28 th at 10:32 (UTC) with a magnitude M L =5.2 (BER), followed 28 sec later by a slightly smaller earthquake with M L =4.8. These earthquakes were felt throughout Svalbard, and in Longyearbyen Næringsbygget was evacuated. In Svalbardposten one can read: «Svalbard opplevde to rystelser tidlig tirsdag ettermiddag, den første cirka klokka 12.31, den neste sekunder etter. Inne i Post- og bankbygget, hvor Svalbardposten holder til, var rystelsene merkbare, sammen med en slags buldring. 34

39 Det var nesten som det smalt. Stolen flyttet seg centimeter, forteller en kilde Svalbardposten snakket med like etter. I Næringsbygget valgte flere av de ansatte i Longyearbyen lokalstyre å evakuere bygningen i noen minutter, og oppe i andre etasje i Coop begynte bordet å riste. Det var skikkelig ekkelt. Vi satt oppe og siste lunsj da bordet begynte å disse. Først forsto vi ikke hva so skjedde, og noen tittet ut for å se om det var gått ras. Vi ble rent uvel, forteller butikksjef Karin Mella.» The main earthquake was followed by more than 40 aftershocks the next 28 hours, when the activity slowly decreased as can be seen in Figure Monthly number of earthquakes M Figure 22. Monthly distribution of earthquakes in Diskobukta. The peak is the 48 earthquakes located to the area in March Sørkappland At Sørkappland, located at the southwestern coast of Spitsbergen ( N, 14-17E), the earthquake activity has increased the last two years as can be seen from Figure Yearly nunber of earthquakes M 3.0 Figure 23. Yearly number of earthquakes recorded in the Sørkappland area. The earthquakes located the last ten years are plotted in Figure 24. Digital data from the seismic station at Hornsund, Svalbard (HSPB) was routinely included in the NNSN data prosessing from November This increased the number of small earthquakes located to 35

40 the Sørkappland area. Since July 2015 there were 9 earthquakes with one calculated magnitude above 3.0. Figure 24. Earthquakes in the NNSN database since Note that for some time periods the location capability might been reduced and the number of located small earthquakes are lower. The number of larger earthquakes is expected to be reliable. 36

41 Nordaustlandet Nordaustlandet is a well-known seismically active area. Earthquakes located since January 2007 are presented in Figure 25. In 2016, an independent small cluster developed to the west of the regular activity on the western side of the Hinlopenstretet. It should be noted that the location accuracy in this area is rather sensitive to the seismogram interpretation and small changes may change the epicentre by tens of kilometers. However, the separation of this new cluster from the regular pattern is likely to be real. Figure 25. Earthquakes located at Nordaustlandet during 2016 (red) and between (blue) Earthquakes in the southern North Sea During 2016, 14 earthquakes were detected and located in the North Sea as shown in Figure 26 and listed in Table 10. The yearly distribution of earthquakes in the area is presented in Figure 27. The largest of the events occurred on 3 rd November with a magnitude of M L =3.3(BER) and M L =3.9(BGS). The earthquake was located using data from NNSN, BGS, HEL and NORSAR. 37

42 Figure 26. Time distribution over time (upper) and location (lower) of the earthquake located within 54-60N and 1W-5E in the southern North Sea (Note that the map is larger than the area used for selection of earthquakes). Earthquakes recorded during 2016 are marked in red. Seismic stations are marked with blue triangles. 38

43 Figure 27. Number of recorded earthquakes in the area 54-60N, 1W-5E. recorded earthquakes M>3.0 Table 10. Earthquakes recorded during 2016 and located in the area limited by 54-60N and 1W-5E. Year Date HRMM Sec L Latitud Long Depth AGA NST RMS ML MW ML(BGS) L BER L BER L BER L BER L BER L BER L BER L BER L BER L BER (NAO) L BER L BER L BER L BER

44 4.3.5 Felt earthquakes In total, 14 earthquakes were reported felt and located within the target area during 2016 (see Table 11 and Figure 28). For the Jan Mayen Island and the area southwest of the Svartisen glacier, the number of felt earthquakes is expected to be larger than reported. Figure 28. Location of the 14 earthquakes reported felt during Large felt earthquakes are mostly reported to UiB shortly after the origin time, and location information and questionnaires are available for the public on the site Smaller felt earthquakes may be reported by the public to local newspapers or other institutions and then reported to UiB. Depending on the time-delay for these reports to be available at UiB, the information on the web might be accordingly delayed. For any felt earthquake the public has to be made aware of the questionnaire, which is done by informing on web and when UiB is contacted by media, private persons or other institutions. Earthquakes large enough to be felt and occurring in heavily populated areas increases the number of people using the web reporting the intensities. 40

45 Table 11. Earthquakes reported felt in the BER database in Abbreviations are: M L = local magnitude and M w = moment magnitude, W: questionnaires received on web (Y/N). The largest felt earthquakes are marked in red. Earthquakes marked in blue is not located in Norway. Due to technical problems, earthquake no 13 is only recorded at JMIC and could therefore not be located. Nr Date Time (UTC) Max. Intensity (MMI) Magnitude (BER) Instrumental epicentre location :41 IV M L =2.4, M W =2.8, M L =2.6(NAO) 62.71N / 5.53E :55 V M L =4.2, M W =4.2, M L =4.1(UPP) 65.03N / 22.55E :32:10 V M L =5.2, M W = N / 21.05E Y :32:38 V M L = N / 20.67E :31 III M L =2.2, M W =2.4, M L =2.1(NAO) 62.22N / 5.79E N :01 IV M L =2.1, M L =2.0(NAO) 59.59N / 10.59E Y :06 IV M L =2.7, M L =2.5(NAO) 61.94N / 4.94E Y :12 III M L = N / 5.99E N :07 II M L =1.5, M L =2.1(NAO) 67.21N / 14.56E N :55 III M L =4.2, Mb=5.1(EMSC) 71.08N / 7.98W N :43 IV M L = N / 13.56E Y II M L =2.0, M L =1.8(NAO) 61.66N / 4.56E N :50 IV - Jan Mayen N :29 IV M L =3.0, M W =3.3, M L = 4.0(NAO) 67.05N / 13.22E N W The largest felt earthquakes during 2016, are the earthquakes occurring 29 th March at 10:32 and 10:32:10 (UTC time). The magnitude 5.2 earthquake was followed by a slightly smaller earthquake 28 sec later. These earthquakes were reported felt strongly at Svalbard (see section 4.3.3). The largest felt earthquake close to the Norwegian mainland occurred 29 th December. This earthquake was felt in the Meløy area and was located offshore, slightly south-west of Bodø. 41

46 5 Scientific studies This section gives an overview of research work that is carried out under the NNSN project in The main objective of this work is to improve the understanding of earthquakes and the seismological models in the region, mostly by using data recorded by the NNSN. Results will be used to improve the NNSN monitoring service. 5.1 QLg tomography By Andrea Demuth, UiB We analyze the attenuation of Lg waves in Norway and adjacent areas. Attenuation is described by the quality factor Q and is a basic parameter to characterize the crust and mantle. Our goal is to gain a better understanding of the geological structures and tectonic processes in Norway. In a first step, we use spectral displacement amplitude ratios of Lg-waves and P-waves to determine the lateral variation in wave attenuation in a frequency range of 2 Hz till 5 Hz. In this approach, we assume that the main attenuation of the ratio is due to Lg wave attenuation. Thus, lower ratios are interpreted as high Lg wave attenuation. We used all earthquakes recorded by the NNSN since 1990, which have a local magnitude higher than 2.5 and recordings by 4 or more stations. The ratio for one earthquake station pair was assigned to its corresponding travel path. Figure 29 Spectral displacement amplitude ratios of Lg-waves to P waves. Figure 29 shows our Lg to P amplitude ratios for the areas in Norway with path coverage. It is visible that Lg waves are higher attenuated in offshore areas than onshore. Furthermore, we observe stronger attenuation in northern parts of Norway. For a more detailed analysis of the Lg attenuation in Norway, we only use spectral displacement amplitude decay of Lg waves to derive the corresponding quality factor. This is done in a tomographic approach. The tomographic code is built on the theory of Barmin et al. 42

47 2001. The code inverts for source and site terms of each observed source receiver pair as well as for the quality factor. This is done with a damped least square approach. Additionally we implemented a spatial smoothing matrix. In order to test the code, we generated a checker board test and used our real source receiver configuration Figure 30a. Figure 30 (a) Synthetic Q Lg checker board input model with real path coverage (black lines). Light blue triangles represent stations and dark blue stars earthquakes. (b) Inversion result for Q Lg with the new derived tomographic code. The checkerboard pattern is well resolved in areas with high path coverage (Figure 30 b). In areas with low to no coverage the Q value is set to the background value of 400. We observe some smearing on the edges of the path covered areas. The input values for the source terms and site terms alternate in a checker board pattern as well and are well resolved. The next step is to run the tomographic code with the real amplitude values for various frequencies. In order to do that, an average Q Lg value is first determined which is used as starting value for the inversion. The average Q Lg values for various frequencies are going to be used to find a general frequency dependence of Q Lg for Norway. 5.2 Relocation of seismicity in southern Norway and the North Sea using a Bayesian hierarchical multiple event location algorithm By Steven Gibbons, NORSAR We are continuing our reassessment of seismicity in and around southern Norway involving a critical re-evaluation of waveform data and arrival picks, exploitation of as yet unused seismic data, and application of a probabilistic multiple event location algorithm. We have been combining datasets from NORSAR, the University of Bergen, additional regional networks (for example Denmark, the Netherlands, and the United Kingdom), and in a few exceptional cases teleseismic data. For each event, a preliminary location estimate is made using the extended set of seismic arrivals; phases with significant time-residuals, or other indications of poor quality, are removed. The cleaned sets of arrivals are then processed by the Bayesloc multiple event location program ( which has been demonstrated to provide enhanced epicenter distributions for clustered seismicity on both regional and global scales. 43

48 In classical single-event location algorithms, the traditional measure of uncertainty is an error ellipse calculated from the formal uncertainties surrounding the arrival times used in the inversion. Systematic bias in the applied velocity models is often not accounted for and event location estimates are frequently presented with unrealistically small error ellipses. Bayesloc calculates joint probability distributions both for hypocenters and parametric information for multiple events simultaneously. In providing implicit corrections to traveltime estimates, Bayesloc can be demonstrated to provide more realistic estimates of location uncertainty. For example, an event for which the applied velocity model provides a poor representation of the traveltimes may have a large formal error ellipse due to the high residuals. The uncertainty indicated by Bayesloc may be significantly smaller if these traveltimes are correctly calibrated. Similarly, an event with very few observations may have a very small formal error ellipse, since there exists a location for which these few constraints can be satisfied very precisely. Bayesloc searches a huge parameter space using a Markov Chain Monte Carlo algorithm and can identify that such event locations have a very broad probability distribution. The attributed uncertainty is consequently far larger for the poorly constrained events. A current snapshot of the multiple event probability distribution is shown in Figure 31. This indicates at a glance those events which appear very well constrained and these appear to cluster in distinct structures and those events with poorer constraints which may need a reassessment of the associated seismic data. The dataset is being increased continually and special attention is being paid to events which may have far tighter prior constraints. These may be due to large magnitudes (hence recorded on a far greater number of stations) or events that have been very tightly constrained by temporary deployments. A feature of Bayesloc of special interest for this dataset is the probabilistic attribution of phase identifications. In classical event location, a phase might be attributed a label which doesn t actually correspond well with the true path traveled from source to receiver. We provide Bayesloc with multiple path models and, in the case of erroneous phase identification, Bayesloc will attribute a higher probability to those solutions with the correct phase identification rather than derailing the location estimate by forcing the error on the final solution. 44

49 Figure 31. Epicenter estimates for around 300 seismic events in and around southern Norway between 1992 and 2016 located using the Bayesloc program. The blue triangles indicate the locations of permanent seismic stations of the Norwegian National Seismic Network, the NORSAR and Hagfors seismic arrays, station MUD of the Danish national network, and temporary stations of the MAGNUS and NEONOR deployments. Bayesloc returns a joint probability distribution of event locations, corrections to traveltime estimates, precision of arrival-time estimates, and phase labels. For each event, the center of the probability distribution is displayed together with the lateral standard deviation; the darkest symbols indicate the events with the best constrained 5.3 Testing of a method for distinguishing between earthquakes and explosions By Ilma Janutyte, NORSAR We have continued testing of a method which helps to objectively distinguish between earthquakes (EQs) and explosions (Janutyte, 2017). The method was first developed at the Institute of Seismology, University of Helsinki, Finland (Kortström et al., 2016), and is successfully used there to help in the daily data analysis. During this reporting period we have made a reevaluation of the method for the stations FOO and HYA using an extended dataset as well as using different partitioning of the training and validation data. In addition, we have applied the method to datasets at the stations LOF and MOR8 in northern Norway. The datasets for the stations were compiled from the University of Bergen (UiB) catalogs, and we made attempts to obtain examples of both EQs and explosions originating in different 45

50 directions from the selected stations. The limit for distance was from 15 to 270 km (Figure 32). A B Figure 32. Seismic events used to develop and verify the reference spectral models for the stations: A) HYA and FOO in the south, and B) LOF and MOR8 in the north. The seismic events are marked as circles and stars, while the seismic stations are marked as green triangles. The results are shown in Figure 33 and Table 12. The prediction shows possible EQs as positive values and possible explosions as negative values. The more the value is positive, the more it tends to be the EQ-like, while the more the negative, the more the explosion-like, while around the zero the prediction is weaker and uncertain. The test case for FOO and HYA stations in the south shows prediction precision of 100 %, i.e. all the seismic events both EQs and explosions were evaluated as correct. For LOF and MOR8 stations in the north the testing precision is 83 and 87 %, respectively. This might be due to too small datasets and not a sufficient number of training examples for the reference model. Especially for LOF station the total number of reference explosions available in the bulletin is low. Table 12. Number of seismic events used in the study for obtaining (training) and validation (testing) of the spectral reference models, and the obtained reference model precision (error) and precision of the test dataset (test precision). training testing test Station in total model dataset dataset precision code error [%] EQ EX EQ EX EQ EX [%] FOO HYA LOF MOR

51 A B C D Figure 33. Predictions using the validation datasets for the different reference models: A) for HYA station EQ data are up to event number 27; B) the common events for FOO and HYA stations; EQ data are up to event number 15; C) for MOR8 station EQ data are up to event number 21; D) the common events for LOF and MORB stations; EQ data are up to event number Towards the integration of NNSN stations in automatic network event location By Steven Gibbons, NORSAR Network detection at NORSAR of seismic events in Fennoscandia has for many years been performed by the Generalized Beamforming (GBF) method (Ringdal and Kværna, 1989) in which phase arrivals on seismic arrays are associated based upon rules relating to the arrival times and other parameters such as the backazimuth and the apparent velocity. This system has not so far been able to utilize many of the high quality seismic signals on 3-component stations in the region like those from the NNSN network. A significant effort is now being made to run detectors at low detection threshold on the NNSN 3-component stations, and to generate output which can be utilized by the GBF process. In a study aimed at detecting and locating aftershocks following major earthquakes, Gibbons et al. (2016) demonstrated considerable success using a detection statistic based upon a continuous calculation of the Auto Regressive Akaike Information Criterion (AR-AIC). The high frequency content of regional signals in Fennoscandia makes this method attractive for the detection of regional P-phases (see the uppermost two traces of Figure 34). The detection statistic attains local maxima very close to the best manual estimates of the signal onset and, once a detection has been declared, the backazimuth and apparent velocity are estimated using polarization analysis. Work is now being carried out on optimizing the detection of S-phases. 47