A web-based seismic monitoring system for volcanoes, including a rapid spectrogram browser
IceWeb is a near-real-time monitoring tool used at the Alaska Volcano Observatory (AVO) since 1998. Its main products are:
- mosaics of 10-minute and daily spectrograms, each image file containing spectrograms for multiple stations
- reduced displacement plots (think of them as calibrated RSAM plots) with a sampling rate of 1 sample per minute
- hourly and daily pseudo-helicorder plots
All these plots are updated every 10 minutes and linked to the AVO Internal Page.
IceWeb was designed primarily as a tool for the AVO Duty Seismologist, to facilitate alarm response. Given the bandwidth limitations of dial-up modems that were commonplace in 1998 in Fairbanks, static web content was the only way to go. Yet the data needed to be near-real-time, i.e. a delay of no more than 10 minutes. This would mean that by the time the Duty Seismologist received an alarm and fired up his home computer, the spectrograms and reduced displacement plots related to that alarm would be available.
IceWeb also includes a tremor alarm system, based on reduced displacement thresholds manually set for each station/channel. This sends email and pager alerts to the AVO Duty Seismologist when a manually set number of station/channels trigger simultaneously. A web-based GUI allows the Duty Seismologist to alter which volcanoes and which station/channels for each volcano are being monitored, and to adjust the thresholds for each station/channel. In the middle of the night in case, for example:
- As seismicity continues to escalate at a particular volcano, it is sometimes desirable to raise the threshold from (say) 5cm^2 to (say) 10cm^2, rather than have alarms repeat at the same level.
- If a particular station became noisy (e.g. due to telemetry problems) and is causing false alarms to be sent, it can be turned off, allowing the Duty Seismologist to sleep.
IceWeb GUIs were developed to enable AVO staff to generate spectrograms and reduced displacements plots for any volcano for any time period at will.
While designed primarily as an alarm response system, IceWeb is also an exceptionally useful laboratory tool, with many AVO staff favouring the spectrograms over traditional helicorder-style displays. Indeed, IceWeb spectrograms quickly became established as the most efficient way to analyze 24-hours of seismicity at up to 30 volcanoes, each with up to 8 displayed stations.
I wrote IceWeb while a postdoc at the Alaska Volcano Observatory beginning in March 1998, motivated by work John Benoit and Kent Lindquist did during the Pavlof eruption in 1996. IceWeb was forked to other observatories, including CVO, HVO and Montserrat.
The IceWeb helicorders were made redundant by Earthworm helicorders in 1999 and then by SWARM in 2005. But the spectrogram browser continues to be a core monitoring tool and is presently hosted by the Alaska Earthquake Center at http://www.aeic.alaska.edu/spectrograms.
Upon rejoining AVO in November 2008, I wrote IceWeb version 2 (aka "TreMoR"). This was a complete rewrite, replacing the Perl-CGI web interface with PhP, and refactoring the MATLAB code based on GISMO. This enables IceWeb to read waveform data from IRIS/FDSN web services, EW/WinstonWS, Antelope databases, Miniseed/SAC/Seisan files etc.
IceWeb's weakest link is it uses MATLAB to generate the image files and this is slow compared to compiled languages. While MATLAB is great for rapid prototyping, the license requirement limited IceWeb's wider adoption for the USGS Volcano Disaster Assistance Program. So back as far as 1999 I planned to rewrite it in C (hence my interest in the Earthworm sgram module), but that project never rose to the top of my to-do list. However, as of December 2015, IceWeb spectrograms have also become largely redundant thanks to Tom Parker's "Pensive" application. This is essentially a Java rewrite of the IceWeb spectrograms...I encourage you to use that instead for real-time operations! However, IceWeb spectrograms are instrument corrected and properly calibrated, so that the color bar is meaningful. This allows spectrograms from different stations at the same volcano or from different volcanoes to be directly compared. The Pensive spectrograms do not make these corrections, so you should instrument correct the data before they go into Pensive.
-- Glenn Thompson
IceWeb has been running at UAFGI since May 1998. The version here is a copy of what I was running at UAFGI in early 2013. This used the Antelope rtexec framework to run, monitor and autorestart various modules and thereby improve overall robustness. A pipeline approach was taken where one program created 10-minute waveform files, another processed those waveform files into spectrogram plots, etc. Multiple copies of each job could be fired up to parallelize the work, with lockfiles used to prevent any collisions. This is quite different to the simpler but less robust way spectrograms are implemented in GISMO, or in Pensive.
Codes exist to periodically query waveform data servers for existing operational seismic stations, and update the IceWeb configuration files accordingly. If run, these prevent blank data channels appearing on spectrograms.
Various data quality checks are also done to eliminate poor waveform data.
The original Perl-CGI interface was implemented because back in the late-1990s, AVO had relatively powerful Sun workstation servers but the Duty Seismologist at home would have only a low power desktop computer, so it was important to keep as much processing as possible on the server-side. In 2008, PhP was chosen for the same reason. Today, it is much better to use client-side processing (e.g. Java, Javascript).
A new Python version of the full IceWeb project is under construction at https://github.com/gthompson/icewebPy. Israel Brewster at UAFGI is tackling a Python version of the spectrogram browser component at https://github.com/ibrewster/seismic_spectrogram.
For completeness, the legacy version of the code that I created during my 1998-2000 postdoc is included under the "legacy_iceweb" directory. This is prompted by a request from Sylvie Vergniolle to be able to cite the code for a new JGR paper on the Shishaldin 1999 eruption. This also contains instrument response data (not in a modern format), computed reduced displacement data (see Appendix in https://link.springer.com/article/10.1007/s00445-002-0228-z for a detailed explanation of the reduced displacement algorithm used), wind data, log files, and a snapshot of all the files on the webserver (when I left in early January 2000 to take up position as Seismic Network Manager in Montserrat).