The Seismological Observatory at Neumayer Station, Antarctica

Continuous monitoring of regional and global seismicity started in 1982. In all these years an enormous amount of seismological data has been collected. Thousands of digital seismogramms have been recorded with different local networks of seismological stations around GvN and Neumayer Station. In the beginning the remote stations were located only some 5 to 10 km away from the base. With growing experience the extensions of the network became subsequently larger. The first networks were entirely located on the Ekström Ice Shelf. Despite the fact that a floating ice shelf is not a very suitable place for seismological observations most of the recordings showed quite satisfying data quality. In 1987 and 1988 two 3-comp onent seismometers were installed on the ice rises Søråsen and Halvfar Ryggen where the ice is laying on solid rock. At these locations shear waves (S waves) can directly be observed. This is not possible on a floating ice shelf because shear wave cannot propagate in water.

The picture above shows the epicenters of recorded earthquakes from 1997 to 2001 which could be associated to well localized events.  The main seismically active regions of the Earth are clearly depicted, e.g. plate boundaries and subduction zones. More than two third of the recorded events are in the teleseismic range with epicentral distances greater than 2500 km and are well localized by NEIC (National EarthquakeInformationCenter, USA), IRIS (Incorporated Research  Institutes for Seismology, USA) and ISC (International Seismological Center, UK). But there are numerous recordings of earthquakes in the local and regional distance range between some 100 km up to 2500 km. Regional earthquakes with epicentral distances of more than approx.1000 km occur predominantly in the Southern Atlantic Ocean, in an area ranging from the Antarctic Penninsula and the Scotia Sea in the West to Bouvet Island in the East and even furter more

eastward. Especially the area around the South Sandwich Islands is a seismically very active region where the South American plate is subducted below the small South Sandwich plate. Many of these earthquakes have magnitudes too low to be recorded at stations outside Antarctica. Because there are only very few seismological stations in Antarctica weaker earthquakes in this region are not localized by international agencies as they require a relatively high minimum number of reported onset times. Of special interest is the occurrence of earthquakes within the Antarctic plate and especially within the Antarctic continent itself. It is generally believed that Antarctica is almost aseismic, at least regarding stronger earthquakes. But is this also true for earthquakes with low magnitudes? Only a sensitive low-treshold monitoring can proof if Antarctica is really almost completely aseismic. However, if some significant seismicity can be detected this will give a better understanding of recent neotectonic processes in Antarctica.

The Seismological Stations

Besides the S-13 seismometers in the seismic observatory (VNA1) there are at present two remote seismographic stations operating. The former stations on the Ekström Ice Shelf  were abandoned in 1999 because they did not significantly improve the accuracy in localizing local and regional events. The remote stations are called "Watzmann" (VNA2) and "Olymp" (VNA3) and are located on the ice rises Halvfar Ryggen and Søråsen. Both stations are equipped with 3-component seismometers with an eigenperiod of 20 seconds (Lennartz Le-3D / 20s). The direct distances from the base to these stations are 43 km and 83 km. As the location of VNA2 proofed to be a good place for seismological observations, with relatively easy access also during winter, a small aperture, short period detection array array with 15 vertical-component seismometers was installed there in February 1997 (see below). For data acquisition a Lennartz PCM-5800 system is used. It uses gain ranging and has a dynamic range of 120 dB. The sample frequency is 62.5 Hz. Digital data from both remote stations are telemetered continuously to the base. The inserted picture shows the PCM unit at station VNA3. It is installed in a wooden box which must be recovered every summer from a depth of almost 3 meters below the snow surface.

The Short Period Array at VNA2

In February 1997 a small aperture, short period array was installed at VNA2 on Halvfar Ryggen with the 3-component seismometer in the array center. It is believed to be the first array of this kind in Antarctica. Arrays are an important tool for the detection and location of very weak earthquakes. A great advantage of array recordings in contrast to single station recordings is that the velocity and the direction of seismic waves crossing the array can directly be measured. Further more the signal-noise ratio in the recordings of weak events is increased substantially by appropriate stacking or beamforming.

The array design is similar to that of the SPITS array in Svalbard which was also designed for the detection of regional earthquakes. The array consists of 15 vertical-component sensors (Mark  L-4) which are arranged on three concentric rings. The diameter of the outer ring is 1960 meters. The diameter or aperture of the array determines how accurate the apparent velocity and the incident direction (backazimuth) of the waves passing the array can be measured. A larger aperture makes these estimations generally more accurate. But only within some limits because with increasing lateral extension of the array the seismic signals may become incoherent.

and stacking may not improve the signal-noise ratio anymore. It may even happen that very low amplitude signals are even badly degraded by stacking. The smallest distance between any of the sensors determines the smallest wavelength which can be resolved by the array. With the inner ring's radius of 200 m the array can resolve wavelengths of approx. 400 m which corresponds to a wavenumber of 2.5 in the wavenumber domain. Waves with shorter wavelenghts cannot be resolved by the array. This can be clearly seen in the array response function. The array response function in the wavenumber domain is the 2-dimensional fourier transform of the sensor coordinates and represents the array's "antenna characteristic". The sharp rise of the central maximum for wavenumbers less than approx. 3 represents the sensitive wavenumber region of the array. Outside this region, e.g. for shorter wavelenghts, the array reacts as a notch-filter. However, aliasing effects may occur at the weaker side lobes at wavenumbers greater than 10. Due to its radial symmetric geometry the antenna characteristic of the array shows also radial symmetry.

Detection and Event Reporting

With the installation of the array in February 1997 the number of detected events increased already by a factor 2. Until August 1998 recording on magnetic tape was still started by a relatively simple STA / LTA trigger which evaluated the ratio of the short term and the longterm average of the ground motion at selected stations. In September 1998 continuous recording started and a more sophisticated detection algorithm was implemented which evaluated all trigger status bits of all seismic channels. The number of automatically detected events rose again. One year later the array processing software "dp/ep" from NORSAR was installed and tuned for this array. This software performs autmatically beamforming within azimuth intervalls of 30°, calculates the STA / LTA ratio and performs a subsequent frequency-wavenumber analysis (fk analysis) if the trigger criteria are met. With this automatic process the number of detected events almost doubled again. The detection results are written to daily event-files which contain beside the triggered first onset time also the determined apparent velocity, the backazimuth and the signal-noise ratio of the detected event. The event-files are completed the next day by the geophysicists with manually picked onset times and other phase readings. If "quick epicenter determinations" (QED) disseminated by NEIC contain preliminary localizations of earthquakes which can be associated to the recorded events these source parameters are added to the event lists. The completed event lists are sent on a daily schedule to NEIC for further evaluation by this agency.

Additionally event-maps showing the epicenters of recorded eartquakes are generated and disseminated by e-mail. In this maps red stars denote localized earthquakes by NEIC which have been recorded at Neumayer Station, blue stars earthquakes which could not be identified in the recordings and green stars represent the estimated localizations of local and regional earthquakes which appear not in the QED listings of NEIC.

This plot shows all the results of the automatic detection algorithm for the period from 01.10.1999 to 19.12.2001. Instead of the measured apparent velocity the correspending inverse value, the "slowness" is used. The slowness is expressed in seconds per degree (in seismology epicentral distances are commonly expressed in degrees rather than in km and 1 degree is 111 km). Each point in this plot represents one detected event in the array recordings with the corresponding values of determined slowness and backazimuth. Several distinct clusters can be identified. All clusters with slowness values less than approx. 10 sec/degree represent certain seismically active regions in the teleseismic distance range. For example the cluster at a backazimuth of 285° and slowness values between  5 and 10 sec/degree comprises almost all earthquakes in South America. The cluster with backazimuths around 180° and slowness values of approx. 5 sec/degree represents the Fiji / Tonga Islands region. Of special interest are clusters with higher slowness values (lower apparent velocities) as they depict seismically active regions which are nearer to Neumayer Station. The most striking accumulation of such events can be observed at an backazimuth of approx. 315° and a mean slowness of 13 sec/degree. All these events are predominantly earthquakes in the South Sandwich Islands region. There are two more clusters with rather high slowness values around 15 sec/degree, one at a backazimuth of 60° and another at approx. 100°. These clusters represent seismically active areas in Antarctica,e.g. depict the major zones of local or regional seismicity at closer distances.

Local Seismicity

This map shows the localizations of some selected local earthquakes in the wider area around Neumayer Station. Two different zones with high seismicity can be identified. The seismically most actice area is in the East of Neumayer Station which is also clearly expressed in backazimuth-slowness representation. The zones of high seismicity can be subdivided into two seperate regions. Both regions show a common South-North alignment of epicenters. They are seperated by a distinct offset and the events in the South are grouped more easterly. The northern actice zone is located in the western part of Fimbulisen, the ice shelf north of SANAE IV. The other region with distinct seismicity is lying more in the South and covers the area of the Jutulstraumen, which is on of the major ice streams in East Antarctica. From geological investigations the Jutulstraumen area is known as a distinct graben structure. There are several indications that this structure might be an active rift system. At a first glance the seismological observations may support this thesis. But great care must be taken for this conclusion as long as the predominant source mechanisms of these events are not known. It may also be the case that these events are icequakes instead of tectonic earthquakes.

Jutulstraumen glacier is a huge, fast flowing ice stream with large and heavily crevassed marginal shear zones. Here the ice is subjected to very strong deformations, thus fracture of ice within these shear zones may be one reason for the observed seismicity. The area at the same latitude as  SANAE IV is approx. the grounding line of Jutulstraumen ice stream. South of the grounding line the glacier is lying on bedrock, north of it the glacier is floating on the sea. Following the tidal changes of water level the glacier executes a vertical movement with changing tides. By this at low tides the glacier is lying furthermore towards North on its bedrock compared to high tides when the ice masses are lifted up to some extent. This results in peridic changes of the flow conditions at the grounding line. Glacier flow over bedrock is generally more a sequence of small "stick-slip" movements instead of a continuous steady creeping. As at low tides the frictional resistance at the grounding line increases the "stick-slip" movements will become more pronounced with larger offset amplitudes during the slip phase. A sudden release of ice that was blocked due to friction for longer time would therefore generate a seismic signal. The same will happen when the ice's movement is suddenly stopped again. Thus at low tides a higher seismicity should be the consequence. Icequakes may also result from the tide-induced periodic bending of the glacier at the grounding line due to fatigue failure of the ice.

To proof if the observed seismicity in these areas shows some tidal influence a spectral analysis of the time series of the detected events in both areas was performed. For the time series of events near the Jutulstraumen grounding line a statistically significant peak was obtained exactly at the frequency of the M2 partial tide, which is the predominant oceanic tide in the South Atlantic.The analysis of the time series of events in the western Fimbulisen area showed absolutely no tidal dependence. These results show that a large part of the detected quakes are of glacial origin and are triggered by oceanic tides. However, they do not exclude the absence of any tectonic seismicity. A strong indicator also for tectonic seismicity is that the recorded events can be divided into two different groups. The seismograms of these groups differ strongly in signal form, in frequency content and in shear wave amplitudes. However, the discrimination between glacial and tectonic events remains still difficult.

Unfortunately reliable determinations of source depths are not possible. The epicenter estimations field only reasonable results for fixed, surface-near source depths. Even additional readings from SANAE recordings do not improve the results. Reliable depth estimations require some more other stations located around the seismically active areas. For a better understanding of the source mechanisms in this region a temporary seismological network of five additional station was installed in January this year. With these stations, distributed at both sides of the Jutulstraumen ice stream, reliable depth estimations and determinations of the predominant source mechanisms should be possible.

Another striking accumulation of local earthquakes is in the West of Neumayer Station, off coast of Kapp Norvegia and near the continental margin. The epicenters of these events are located too far away from the ice edge to be associated with a glacial origin. This type of event can also be observed along the continental margin furthermore to the East. These earthquakes are interpreted as as result of a postglacial rebound of the lithosphere following an earlier glaciation maximum. The corresponding seismicity is also observed in Scandinavia and other regions with an earlier glaciation.

A very special event is the earthquake from December 25 in 1997. This earthquake occured near the Polarstern seamount, far away from the continental margin. The red star in the map above denotes the epicenter derived with data from Neumayer Station and SANAE IV. The white star marks the localization calculated by ISC. The magnitude of this event was 3.9, too low to be recorded at many other stations. Besides one station in central Africa this earthquake was only detected at four other seismological arrays, three in Australia and one in Canada.