Waveform data processing and analysis

Every single day, several gigabytes of waveform data from IMS stations arrive at the IDC in Vienna. Without processing, this bulk of information would be of little use to most States.  Quick answers to the most pressing questions are needed.

Was an event detected?  Where did it take place? What characteristics does it display? How big was the event?  Was it a natural or a man-made event? Targeted processing of monitoring data is needed to give States the necessary information to answer these questions and enable them to make decisions concerning the nature of an event.

Waveform data are derived from the monitoring of seismic and acoustic waves that move through the Earth, the oceans and the atmosphere. Their analysis answers crucial questions about an event.

Three of the four monitoring technologies are called waveform technologies, comprising seismic, hydroacoustic and infrasound technologies. They are used to monitor and record the movement of energy that is generated by certain events and propagates as seismic waves or acoustic waves through the Earth, the oceans or the atmosphere. 

Monitoring data recorded at stations using these three technologies are called waveform data. Typically, waveform data are displayed as traces moving across a computer screen with the x-axis showing time and the y-axis representing the movement of the medium that is being monitored, i.e. ground, air or water.

Station processing

Once the data are stored at the IDC, data from each single station undergo independent analysis to detect signals which originate from seismic or acoustic disturbances. This process is called station processing and is entirely automatic.

Once a disturbance is detected, the characteristics of the relevant signals are measured and recorded in a large database. These characteristics include time, size and the azimuth, i.e. the direction from which the signal arrived at a station. As with all processing and analysis of data, an important component of the processing is to determine the accuracy and reliability of each parameter.

After storage at the IDC, data from each single monitoring station is analysed independently in an automatic process to detect relevant signals.
Seismic waves generated by the Indonesian earthquake in December 2004.

Somewhat different methods of automatic processing are used for seismic, infrasound and hydroacoustic data. The principle, however, is the same — to detect any signal that may have originated from a disturbance and could possibly have been a Treaty violation.

Network processing
It can be expected that an event is recorded at more than one IMS monitoring station. So the next step then is to sort out which signals from different stations originate from the same event. This phase is called network processing and is a very complex task. There may be many events in the course of a day, typically including more than 100 earthquakes.

In network processing, data from different stations but originating from the same event are examined to establish a location estimate.
Hydroacoustic waves generated by the Indonesian earthquake in December 2004.

By examining all available data and associating them to an event, a clearer picture of what actually occurred begins to emerge. A location estimate can now be established for a potential event.  A good location estimate of an event is important to measure the size of an event since the size of the signal generally decreases with distance from the event.

Standard Event Lists

Infrasound waves generated by the Indonesian earthquake in December 2004.

From this automatic processing, lists of events emerge and, as events accumulate, they are made available to Member States. The first of these lists, Standard Event List 1 (SEL1) includes events that have been recorded at primary seismic and hydroacoustic stations.  At this point, infrasound data are not yet available for the processing because the signals travel more slowly through the atmosphere, creating an additional delay in the recording process.

Standard Event Lists cont.

Example of Standard Event List SEL3.

Based on the events listed in SEL1, additional data from auxiliary seismic stations are requested to refine the location of events that have already been listed. Requests are formulated automatically and sent to selected auxiliary seismic stations. With this additional information and with incoming infrasound data the next, more comprehensive and higher quality list of events is created, Standard Event List 2 (SEL2).

Automatic processing results in three consecutive event lists, the Standard Event Lists 1, 2 and 3. Each of them includes additional information on observations made.
Events listed in the Reviewed Event Bulletin have been examined by analysts.

The last of these event lists is Standard Event List 3 (SEL3), which includes additional information resulting from data that arrive late and later recorded signals. SEL3 is the most refined automatic list of events issued by the IDC.

This process is entirely automatic, carried out by specially designed computer programs. No human intervention in the data processing has taken place until this point.

Although the process of automatic processing is quite sophisticated, analysts need to review the results to ensure that reliable information is provided to Member States.

Interactive analysis and the Reviewed Event Bulletin

The automatic processes to compile the Standard Event Lists from the original waveform data are quite sophisticated.  It is still necessary, though, for the results of automatic processing to be reviewed by an analyst to provide reliable and comprehensive information to Member States.

Sometimes signals are associated to the wrong event. Occasionally, events are listed which are not real, and sometimes events are missed. Analysts at the IDC therefore review every single event listed in SEL3. In 2010, there were an average of 161 events each day.

Analysts review every single event listed in Standard Event List 3. The daily average in 2007 was 126 events.

The analyst discards events which are not real, adds signals which have not been associated to an event, and corrects and improves the location estimates of those events which are real.

Analysts reviewed an average of 126 events per day in 2007.

Analysts must also disentangle cases where the automatic processing has associated signals from two events together — a so-called mixed event — or where signals from one event have been interpreted automatically as two events — a so-called split event. They also scan the data for events which may have been missed entirely.

The role of the analyst in the overall analysis process is of crucial importance.  A lot of responsibility lies on the shoulders of analysts who sift through all automatically generated event lists. The work is like an investigation that must be carried out based on scattered information and against tight deadlines. Demands are placed on the analyst’s experience, judgement and detecting skills.

Analysts carry great responsibility. They work against tight deadlines and need experience, good judgement and sharp detecting skills.

After interactive analysis, the number of events is reduced approximately by half. The confirmed and corrected events are listed in the Reviewed Event Bulletin (REB).

The events contained in the REB are of higher quality in view of the information on which they are based. The reason is that the IDC analyst has looked at all the available information, assigned it correctly and, thus, prepared the event for the next step in the processing, the automatic screening of events.

The result of the interactive review is the Reviewed Event Bulletin which is of higher quality than the previous event lists.

Automatic waveform event screening

The map shows all events listed in Reviewed Event Bulletins from February 2000 to November 2010.

The purpose of this process is to identify events in the REB that are natural.  Several criteria are applied in this process, looking at specific characteristics of the listed events in order to distinguish them as natural or man-made.  The criteria or characterization parameters are listed in the protocol to the Treaty.

As most entries in the REB are seismic events, the values measured for the various seismic waves are of particular interest. There are different kinds of seismic waves: body waves that travel through the interior of the Earth, and surface waves that travel along its surface.

Automatic event screening leads to the identification of events as natural or man-made. The criteria for this screening process are listed in the Treaty.

Each type of seismic event creates these seismic waves in a different and recognizable pattern. These patterns can provide information on the nature of an event - whether it is natural or man-made.

Seismograms of the announced nuclear explosion in the Democratic People's Republic of Korea in 2006 and of a nearby earthquake in 2002.

For example, for events occurring near a monitoring station the two types of body waves are examined: P-waves and S-waves. An earthquake typically generates small P-waves and big S-waves. With an explosion the ratio is usually reversed – the P-waves are considerably larger than the S-waves. As these waves travel at different speeds — P-waves move faster than S-waves — an easy distinction between them can be made in the respective seismograms.

The comparison of seismograms from the 2006 nuclear explosion in the Democratic People’s Republic of Korea and an earthquake in the same region in 2002 provides a good case study. The explosion’s seismogram (top) recorded at a seismic station in the Republic of Korea shows large P-waves followed by considerably smaller S-waves. In contrast, the earthquake generated smaller P-waves and larger S-waves (bottom).

Seismic waves generated by earthquakes and explosions display distinct properties, the examination of which helps to identify the nature of an event as natural or man-made.
Example of a T-phase - spectrogram of an earthquake. Click image to listen.

Other very important screening criteria are the magnitude ratios of body waves and surface waves. Measurements taken more than 2000 kilometres away from an event form the basis for the calculation of the body wave magnitudes. The magnitude ratio between body waves and surface waves is known to be comparatively larger for man-made events such as explosions than for natural events.

Information on the depth of an event is also used as a criterion to distinguish between natural and man-made events. Technology limits the depth at which an explosion can be detonated. Hence, great depths clearly disqualify an event from being of human origin.

The depth of an event can also exclude it from being of man-made origin as technology limits the depth at which an explosion can be detonated.
Example of an H-phase - spectrogram of an explosion. Click image to listen.

Some events may have hydroacoustic data associated with them. Particular properties of hydroacoustic waves are examined to help identify an event as natural or man-made.

Sound waves in the ocean can be one of three types of signal: H-phase, T-phase or N-phase. H-phase signals come from in-water explosions caused either by man-made devices or underwater volcanic eruptions. T-phases are waves generated by distant earthquakes. They are formed when seabed vibrations caused by the earthquake send sound into the water. N-phase signals are noise signals from a wide range of physical sources such as iceberg-generated noise, airgun surveys and whale song.

Some events may have hydroacoustic data associated with them. Particular properties of hydroacoustic waves are examined to help identify an event as natural or man-made.
Example of a N-phase - spectrogram of whale song. Click image to listen.

Following further analysis of their frequency properties, the association of T-phases serves to exclude an in-water explosive event during the event screening process, as these waves are considered to be generated exclusively by underground events.

Should the screening criteria indicate that the event is natural, it will be screened out. Only those events that were clearly identified as man-made or where screening did not provide clear answers are kept and listed in the Standard Screened Event Bulletin (SSEB) as potentially suspicious events.

Next Chapter:
Radionuclide data processing and analysis