Wednesday, December 11, 2013

GeoNet 2023 Part 3: The way ahead

Before I start, I would like to point out that forecasting the future is difficult, particularly when it concerns technology and it is likely to lead to a BIG fail. This was expressed very well by Niels Bohr who said (and yes, I know there is dispute about who said this first):

“Prediction is very difficult – especially if it is about the future.”

In previous blogs (Part 1 and Part 2 of this series) I have shown that we have sometimes got it right in the past, so if I restrict myself to the future of GeoNet, perhaps I will increase my chances. So here goes - what will GeoNet (or what GeoNet becomes) look like in 2023?

Sensor networks 2023 ….
I expect sensor site numbers to explode in the coming decade as a whole series of technological advances come together. The number of sensors will increase by at least an order of magnitude, meaning GeoNet in 2023 will have round 6000 sensor sites available. This sounds far-fetched, but remember how few real-time sensor feeds we had 10 years ago.

What will bring about this change? I expect the same technology advances which have revolutionised computer and data communications technology will finally start making its mark on sensor technology. This has been slow to happen, but it will. The trick is to increase the density (number of sensors) while at least maintaining the measurement accuracy. Previous proposals for increased sensor coverage have advocated more but lower quality sensors. What I envisage is a world where just about everything (position, strain, temperature, pressure, chemistry, shaking level, etc.) can be measured to a high level of accuracy.

Where will all these new sensors come from? The answer is from an extension of existing and yet to be utilised techniques. For example, sensors for measuring temperature and pressure can use the changes in the properties of fibre optic cable lengths and rings. Micro-electro-mechanical systems (MEMS) technology has come a long way in the last decade. We all have MEMS in our smartphones and tablets to tell the device which way is up (its orientation). These are low accuracy devices but very good ones exist and are improving all the time. These are already used in some of the strong shaking instruments we use (see the CUSP instruments). Price is the current barrier to widespread use of high accuracy MEMS sensors in very large numbers.

Consider the recent improvements in GPS technology. Again we all have GPS receivers in our smartphones and tablets. Expect the accuracy of GPS devices to increase with time and become part of multi-sensor platforms. In many respects our current smartphones have much of the technology required to act as sensor platforms, although they do not yet have the necessary sensor accuracy.

And I have not even mentioned nanotechnology yet! Nanotechnology is the manipulation of matter on an atomic and molecular scale.This technology is already starting to produce very small sensors, and this trend is likely to continue. In some ways it is an extension of MEMS technology, but much smaller. The impact on sensor technology of nanotechnology is very hard to predict!

One of the real barriers to very good sensor coverage of New Zealand is the sea that surrounds us. It would be so much easier to locate earthquakes and monitor tsunami if we had sensors on the seabed surrounding New Zealand. The problem is that such sensors are currently very (very) expensive to install and maintain. But imagine if they were installed as part of the data communications infrastructure which connects different parts of New Zealand and other countries. An international collaboration I am involved in, which is a joint undertaking between United Nations organisations, scientific institutions and commercial companies is investigating the use of submarine cables as instrument platforms for environmental and hazards monitoring. Cables capable of carrying sensors (usually assumed to be at repeater sites; see Figure 1) are called green undersea cables. It is early days, but the requirements for low data latency, which is not available with most satellite technology, and route diversity will drive terrestrial solutions. It is therefore likely that there will be many more submarine cables installed in coming years. If these cables are utilised for sensor deployment we will end up with a huge number of sensors covering the world’s oceans.

Figure 1: A map of submarine cable routes. Submarine cable repeaters (blue dots) are along the cables although the total number is about four times those shown (40 to 150 km apart). A typical transpacific cable has about 200 repeaters. Current tsunami buoys and other ocean observatories are also plotted. The figure is from an ITU report.

Data communications 2013 ….
This is both the hardest and easiest capability to predict. If the past predicts the future, then data bandwidth will not be a problem for GeoNet in 2023. Predicting exactly how bandwidth will be made available to move the huge amount (by today's standards) of data collected by GeoNet 2013 is difficult. But our data volumes will be tiny compared to super high density 3D video (and virtual reality I assume, having read far too much science fiction). The "last mile" problem will be solved by current rural broadband initiatives and satellite technologies. So I will leave it at that, assuming there will be ample bandwidth available “somehow” for GeoNet in 2023!

GeoNet data 2023 ….
Everything will be in the cloud. The GeoNet data centre will be distributed and very resistant to geological hazards and equipment failures. It will reconfigure automatically and move data and processing capability and capacity around as required. The volumes of data collected each day will be orders on magnitude greater than today, but all data will still be online and easily accessed. The data archive and delivery will come from somewhere in the cloud electronically close to you. And the way it is delivered will be very configurable.

GeoNet outputs 2023 ….
By 2023 GeoNet will be providing very fast impact reports following geological events to a large number of stakeholders as well as the public and media. Much more background will be provided for events, and many new ways to visualise GeoNet data and information will be available in 2023. We have started to move in this direction by reporting likely felt intensity rather than just magnitude for earthquakes.

It will be a very mobile world – almost all data and information delivery will be to mobile devices but these will be closely connected to the cloud. With data, information and compute capability existing in the cloud, the distinction between mobile and fixed devices (like this computer I am typing these words into) will have little meaning. By 2013 GeoNet will be providing not only the data to researchers, but tailored compute capability to allow very detailed data analysis and modeling electronically close to users. 

Summary ….
Overall the development of GeoNet will continue to parallel that of computer and data communications technology. But additionally, expect to see a huge increase in the number and usability of sensor technology.

That's it from me in 2013. Now all I have to do is live long enough to see what happens!

Tuesday, December 3, 2013

GeoNet 2023 Part 2: The here and now

Before launching into what GeoNet may look like in 2023, I will briefly review where we are at now and try to answer the question – is the past a good predictor of the future?

What is GeoNet?
GeoNet is New Zealand’s geological hazards monitoring system – we monitor earthquakes, volcanic activity, tsunami and land stability. As well as monitoring these hazards, GeoNet collects high quality data for research which will lead to better knowledge and therefore mitigation of our geological hazards.

GeoNet can also be viewed as a large, distributed data collection, processing, archiving and delivery system. It is comprised of sensors networks, processing and archiving capability, and data and information delivery functions.

And yet another way to look at GeoNet – it is a New Zealand high technology project that made good!

GeoNet networks ….
GeoNet operates a network of over 600 sensor sites throughout New Zealand, connected by a variety of data communication systems (satellite, radio, landline and mobile) which form a huge computer network. The approximate breakdown of sensor types is:

  • 180 “weak motion” earthquake recorders (both National and Regional networks of sensors) to locate earthquakes
  • 180 continuous GPS sites which record how the land deforms slowly and during earthquakes
  • 250+ “strong motion” sensors which record the shaking levels in felt earthquakes, including sensors in buildings and on bridges
  • 17 tsunami gauge sites to record sea level change caused by tsunami
  • Plus a variety of other sensors to record position, chemistry, water levels and temperatures for volcano and landslide monitoring
The big changes in the GeoNet sensor networks have been in the way we move data around the country. The fundamentals of the sensor and data recording technology have not changed much, but with the spread of the Internet our ability of moving data has grown. In 2001 many places required expensive satellite data communications, but this situation is improving fast. The spread of the Internet was predictable and has paralleled the growth of GeoNet.

GeoNet data ….
The data from the sensor networks feeds into GeoNet’s distributed data centre system. When GeoNet began in July 2001 our plan was to have a main data centre in Wellington with a backup site at GNS Science’s Wairakei campus near Taupo. Over the last few years we have moved away from that concept to a distributed data centre model using compute capacity and storage in external and internal “clouds”. GeoNet now operates around 100 “virtual computers” which are centrally configured and managed allowing fast rollout and quick failure replacement. GeoNet Rapid, which automatically locates New Zealand’s earthquakes is run primarily in a cloud service in Auckland with the backup here in Wellington. The rapid availability and growth of the cloud is something I had not expected, but is now central to GeoNet operations.

In the early days of GeoNet we calculated that if computer hard disk space continued to increase at the (then) current rate, we could keep all data on-line indefinitely. Currently GeoNet collects around 8 GB a day and the total archive is around 30 TB. When GeoNet started, 30 TB of online storage required robotic tape changing systems costing millions of dollars. Now I have around 10 TB of storage at home - this is one technology prediction we got right!

Figure 1: GeoNet sensor network 2013 - Seismographs (big and small red dots); Strong motion (big and small green squares); GPS (black and light blue triangles); Tsunami gauges (upside-down dark blue triangles).

GeoNet outputs ….
The data and information produced by GeoNet is delivered through the GeoNet website, which is itself a distributed system of New Zealand and international computer servers. We also have information available via our smartphone Apps (currently on Android and iOS). Via the website, it is possible to find such things as earthquake information, volcano status and the position changes happening to our GPS stations as New Zealand slowly changes shape as we are buckled by the slow collision between the Pacific and Australian tectonic plates. Researchers can download data on earthquake shaking, the raw data used to measure the slow deformation as New Zealand deforms, and all the time-series data (waveforms) recorded by seismographs and strong motion instruments. All of the information on the sensor networks (sensor locations, types and calibrations, etc.) is available via the website so that the data can be interpreted and used correctly.

To demonstrate how the use of the GeoNet website has grown, lets look at the case of Dino the pink dinosaur. In the early days of GeoNet, Dino appeared in front of the White Island volcano-cam and caused the one and only complete outage of the website when "huge" numbers of admirers arrived to view him (or her?). Traffic to the site reached 10 hits per second! Today a widely felt earthquake drives traffic to 10s of thousands of hits per second.


So, is the past a good predictor of the future? Sometimes! The growth and development of GeoNet has paralleled that of the Internet and computer technology and will probably continue to do so. 

Next blog - GeoNet 2023 Part 3: The way ahead

Tuesday, November 19, 2013

GeoNet 2023 Part 1: Looking back to look forward

I was recently asked to take part in a “navel gazing” exercise as a part of the eResearch2020 project and it got me thinking about both where GeoNet has come from, but more importantly, where we are going over the next decade. What will be the big changes? Where will sensor and data processing be at in another 10 years? Is the past a good predictor of the future? So first let’s look back to look forward in this first part of a short blog series.

In the beginning ….
In 1982 I was employed to investigate the possibility of collecting all New Zealand’s seismograph data centrally and electronically in Wellington. In those days all earthquake recording required those rotating drums and needles that movie sets so love. And most of the recording was done onto film which needed developing before use. I quickly established that the cost of digitally recording and transmitting all of the data to Wellington would climb into the millions of dollars (and that was 1982 dollars!). That could have been the shortest job ever – but I am still working on GeoNet more than three decades later!

Going digital ….
The solution at the time (mid-1980s) was to “go digital” and record the earthquake data on magnetic tapes that were then posted to Wellington for analysis. So I worked on methods of identifying the earthquake signals in the background noise caused by the weather, people and other animals. We could only record 25 MBytes (yes you read that right, mega-bytes not giga-bytes!) on each tape so we had to “throw away” most of the recorded ground signals. The world moved slower in the 1980s, but by around 1990 most of the 30 or so earthquake recording sites around New Zealand had been converted to digital recording.

Figure 1: The EARSS (Equipment for the Automatic Recording of Seismic Signals) digital seismograph which recorded on 25 Mbytes tape cartridges. Software running on a microprocessor automatically detected earthquake signals and recorded segments of data to the tape. 

Fast earthquake location, 1990 style ….
At that stage the tapes were posted to us once a week by the local farmers meaning it could take up to a month to get all the data required to locate an earthquake. The short cut was to ring the farmers who would read off earthquake wave arrival times from a paper printout. Using that information and data from seismographs around the Wellington region, we would be able to (if luck was on our side and the farmers were at home) provide a rough location and size for a well recorded felt earthquake in about an hour. The height of technology and science at the time!

I have just checked - the last tape from those old “tape seismographs” was received and read in mid-2005, only a little over eight years ago. By then we had made the huge change to recording ground shaking continuously at our seismograph sites and transferring the data to our data centres almost instantly for analysis. For many years following 2005 our earthquake processing, although now much faster, still required manual intervention to achieve acceptable results. All locations sent to the GeoNet website were reviewed by a seismologist before publication – a process requiring about 20 minutes.

A new beginning ….
From the beginning of GeoNet in July 2001 we progressively replaced the tape seismographs,  added other sensor technologies and increasing the number of sensor sites from around 60 in 2001 to over 600 in 2012. Then in 2012 we introduced GeoNet Rapid with automatic earthquake processing and reporting including a blow-by-blow record of the “history” of the earthquake location process published directly to the GeoNet website.


Next blog - GeoNet 2023 Part 2: The here and now

Monday, January 7, 2013

GeoNet and Tsunami - Part Two


Introduction

In my last tsunami blog I outlined GeoNet’s role operating the real-time tsunami gauge (sea level) network, and the use of these gauges for tsunami modelling, characterisation and warning.

GNS Science does not operate an official warning centre, but are the science advisors (using the GeoNet capability) to the Ministry of Civil Defence & Emergency Management (MCDEM), the New Zealand agency responsible for tsunami warning. International and New Zealand data are used to characterise the potential of tsunami generated by distant or regional earthquakes to threaten the New Zealand coast.

Distant and Regional Source Tsunami

Distant source tsunami take many hours to reach New Zealand allowing adequate time for warning and evacuation if required. Regional tsunami sources have travel times of between one and three hours and usually originate from the South-west Pacific region. In this case although there is less time official warnings are still possible.  For both distant and regional source tsunami New Zealand relies on the Pacific Tsunami Warning Centre (PTWC), located in Hawaii to alert us to possible tsunami threats. PTWC serves as the operational headquarters for the Pacific Tsunami Warning and Mitigation System (PTWS). The PTWS is governed by Pacific member countries of the Intergovernmental Oceanographic Commission (IOC) which is a body under the United Nations Educational, Scientific and Cultural Organization (UNESCO). In a later blog I will outline how New Zealand contributes to PTWS.

The PTWC monitors an expansive seismic and sea level network (provided by member countries of PTWS) in the Pacific and issues tsunami bulletins which are used to trigger the New Zealand response. Once a notification is received from PTWC (via a variety of communications channels) the likelihood of serious impact in New Zealand can be assessed.  A brief consultation between the GeoNet and MCDEM Duty Officers takes place and this can lead to the issuing of either a “no threat”, “potential threat” or “warning” message. While a “warning” will be issued by MCDEM as a default action if an earthquake exceeds certain thresholds, in most cases no action is required because the event is too distant or small to be a danger to New Zealand. As a first response the GeoNet Duty Officer uses the best available information on the earthquake size and location and a catalogue of tsunami forecast models to quickly estimate the likely tsunami impact in pre-defined coastal zones around New Zealand (see Figure 1). This information is provided to MCDEM as a first estimate of the likely actions required by responding agencies.

If time permits, the GeoNet Duty Officer calls a meeting of the Tsunami Experts Panel to provide a more detailed estimate of the likely impacts on New Zealand. The panel is comprised of New Zealand experts from GNS Science, the National Institute of Water and Atmospheric Research (NIWA), New Zealand universities and private organisations. Extra observations and modelling techniques are employed by the Duty Officer and members of the Tsunami Experts Panel who give continuing updates to MCDEM on the probable impacts of the tsunami. As part of this process a Science Liaison Officer is provided to the National Crisis Management Centre (NCMC, located in the Beehive basement) if the centre has been activated. This provides a seamless connection for science advice to the emergency responders.  This process of review and update continues until the threat posed to New Zealand passes.

Figure 1: The tsunami threat level map produced at the time of the March 2011 Japan Tsunami. Note that the colours used for the threat levels have changed to avoid confusion with evacuation zones. For more details refer to the Tsunami Warning and Advisory Plan (page 13) on the MCDEM website.
Local Source Tsunami

What about local source tsunami warning? Here we mean tsunami with a travel time of less than one hour to the nearest coast. The greatest local source tsunami threat to New Zealand is from the subduction zone along the East Coast of the North Island, where the Pacific and Australian plates meet. This could potentially cause a huge tsunami similar to the one that struck Japan in 2011, but unlike Japan we have very little indication that such a tsunami has ever occurred.

 New Zealand does not have a dedicated local tsunami warning capability.  While MCDEM will issue warnings in the same manner as described above in the case of a nearby large earthquake, these warnings are unlikely to be timely enough for effective response so it is important people know the natural warning signs and act on those. Examples from Indonesia, Samoa, Chile and Japan suggest that people are much more likely to survive a tsunami if they heed the natural warning signs and self-evacuate. Waiting for an official warning often means losing those vital few minutes with fatal results.

So, people in coastal areas should watch out for:
  • strong earthquake shaking (hard to stand up);
  • weak earthquake shaking lasting for a minute or more;
  • strange sea behaviour such as the sea level suddenly rising and falling, or the sea making loud and unusual noises or roaring like a jet engine.
If any or all of these are observed – don’t wait for an official warning – let the natural signs be the warning. Take immediate action to evacuate the predetermined evacuation zones, or if they don’t exist go to high ground or go inland (both is best).

It is important to note that the hardware to provide a dedicated local tsunami early warning system, even when fully developed only provides a small part of what is required for a robust, sustainable, end-to-end local tsunami early warning capability. The warning messages need to reach the community at risk and the community must have pre-planned response procedures if effective local tsunami warning is to succeed. And this must be sustained for decades. Additionally, it is important that any warning system not undermine self-evacuation (mentioned above as so important) triggered by natural warning signs. Education is a cornerstone of a sustained tsunami risk awareness and response programme.

GeoNet and Local Tsunami Early Warning

By its nature GeoNet does have some of the tools required to provide local tsunami early warnings, including a broadband seismograph network, a tsunami gauge (sea level) network, expert staff and access to international data feeds. However, several components required for a robust local warning capability are lacking. For example, New Zealand currently has no offshore deep sea tsunami detection capability, and relies on other countries’ sensors. And further developments of the earthquake systems are required:

  • Improved offshore earthquake location capability. Because of the long thin nature of New Zealand earthquake location and depth estimation accuracy drops off quickly for offshore events;
  • Improved earthquake size (magnitude) estimation (using both seismic and GPS techniques);
  • Fast earthquake source characterisation – is it the kind of earthquake which may cause a tsunami?;
  • Tsunami (slow source) earthquake identification capability – is an earthquake of the kind that appear to be smaller but can cause large tsunami?

These capabilities are being researched or are under development but not yet available. Even with all these capabilities, I believe an effective local tsunami early warning system would require at least some offshore deep ocean sensors off the East Coast of the North Island. This would provide good capability for that region (the most destructive of the possible local sources), with capabilities in other regions mainly limited to warnings based only on earthquake size, depth and location. A further requirement of an effective local tsunami early warning system is a fully staffed 24/7 operations centre. GeoNet Duty Officers are currently “on-call” and can respond from home or work, but are not full time in the role. Automation can be employed as much as possible, but with current and envisaged levels of technology all countries attempting local tsunami early warning have 24/7 staffed operations centres. 

The bottom line is that GeoNet could play a small but significant part in the national effort to establish a fully operational and effective local tsunami warning capability. But an extra zero would need to be added to the GeoNet budget if this were to become a reality, and a coordinated effort by many New Zealand organisations would be required.