Observatoire Multidisciplinaire des Instabilités de Versants





Daily seismological data validation

We present here the 2 approaches used to detect issues with the OMIV seismological data quality such as:

  • Issue with the digitizer
  • Sensor not connected to the digitizer (cable cut or damaged, sensor disconnected)
  • Sensor damaged or 'uprooted', and therefore not recording the surface vibrations

Daily validation is critical to control the quality of the acquisition chain and data continuity in the landslide context: Instruments are particularly exposed (mudflows, rockfalls, ground movement that may tilt a sensor).

The 50m-long cables separating each sensor from the digitizer in the case of the OMIV seismological antennas also make it more likely to have cuts or damage, affecting data quality and completeness.

Both methods showed below are currently in use, and can run with limited connectivity to the stations (transfer of a few minutes of data per day at most).


Daily noise correlation within all antennas with near real-time transmission (La Clapière and Séchilienne sites)

Click any image for a larger view. The images above are automatically updated daily.

In the images above, each line corresponds to the first minute of recording of a day. Each column corresponds to a vertical channel in the antenna. The color displayed is the correlation of the vertical channel with the rest of the antennas' vertical channels.

All data is filtered to retain frequencies between 1 and 20Hz only before correlation.
A correlation above .7 means the signal recorded is very consistent between channels, normally indicating good functioning of the antenna. A correlation very close to zero that occurs consistently for a number of days typically means there is an issue with the channel. Intermediate cases may be explained by intermittent and very local noise (close to a particular sensor). In the particular case of the Séchilienne gallery (station GAL), the correlation is not expected to be high for the channels deep in the gallery (channels 0-6/7) due to the depth variation.

This plot is used daily as part of the seismological data validation process to detect recording issues. It has also, in one occurrence (fast-moving La Clapière landslide), showed that re-placing the sensors (verticality) significantly increased the correlations; see that particular example here. In any case, it allows for an early detection of the problem so issues can be addressed in the field early on.

A limitation of this method is that it excludes 1) all horizontal channels and 2) single-sensor seismological stations. The following method is needed to cover all data.


Visual inspection of recorded signals at the time of regional earthquakes, for all OMIV channels with near real-time transmission (La Clapière, Avignonet and Séchilienne sites)

Click any image for a larger view.

The second method we use routinely for the validation of seismological data is the visual inspection of recorded signals at the time of regional earthquakes. It can be run for all channels of all stations. The routine is run for earthquakes that produce a signal that comes out of the background noise (earthquakes in the Alps with a magnitude of 2 or more, and earthquakes located in Europe with a magnitude of 4 or more). That's typically one or two earthquakes per week.

This method does not allow a daily QC of all channels, but it does cover all channels once or twice a week.

This approach is currently applied to all stations for which the seismological stations are remotely accessible, i.e. 3 out of the 4 OMIV landslides (Avignonet, La Clapière, Séchilienne).

The images above will soon be created and archived automatically.