Sunday, February 22, 2015

The Future of GeoNet Revisited - Part 4: Early Warning

Early Warning....
Let me state up front that I have mixed feelings about early warning for geological hazards events. It is very hard to do well, and the consequences of false alarms can be severe. Also, technology is a very small part of the end-to-end process. Detecting the hazard is in many cases the easy part - getting the message to affected communities in forms that can easily be acted upon is the really hard part of the complete package. Finally, any system has to be super robust. It may be in place for decades before a devastating event occurs. And when disaster strikes, will the early warning system (end-to-end) still be fully functional?

 Another issue I will get off my chest at this stage is that there is a danger that early warning is used as a funding opportunity by researchers (often with the best of intentions), particularly those not involved in operational systems. It is easy to say that someday this wonderful research I am doing will lead to early warning. This is the "I can cure cancer" syndrome we all know about. How many times have the media announced a cure for cancer? What is obvious is we are slowly moving to be able to treat (or at least delay) many cancers - but there is no one silver bullet. I believe it is the same with early warning for geological hazards.

GeoNet and Early Warning....
GeoNet has much of the infrastructure and technology necessary to contribute to a forecasting and early warning capability for New Zealand for some perils. GeoNet is currently set up to collect research data and report on geological hazards and would require considerable reconfiguration for short term early warning. The lack of a fully staffed 24/7/365 warning centre is a major component not currently available. GeoNet operates a duty system rather than being staffed 24 hours a day every day of the year. Sensor network expansion and increased robustness is required, and research and development is needed to take the outcomes of scientific research and transform these into operational tools if GeoNet’s role was expanded to include warning centre capabilities.

For the record, forecasting is a form of time-dependent hazard assessment, whereas early warning requires the identification of an imminent peril and the likely time of impact.  Some geological hazards are easier to forecast than others, and the benefits of the forecasts can vary considerably. Let's briefly look at the perils we face.

Volcano Early Warning....
The GeoNet volcano monitoring programme already provides a level of volcano forecasting which would be enhanced by a 24/7/365 warning centre, improved remote data collection systems and additional research and development. Let's look at the example of the eruption of Te Maari craters on 6 August 2012, following nearly 120 years of inactivity of Tongariro volcano. It followed three weeks of unrest including an increase in earthquake activity and changes in fluid chemistry, leading  GeoNet to raise the alert level for the volcano. Although the alert level is not designed to be predictive, the increased activity triggered increased community and land owner (Department of Conservation) consultation and resulted in a better prepared community when the eruption took place. This is an example of effective volcano forecasting in practice.

Tsunami Early Warning ...
The compelling case for early warning capability in New Zealand is the potential for local or near-regional source tsunami. The 2013 update of the 2005 tsunami hazard assessment for New Zealand demonstrated that a regionally generated tsunami from a Kermadec earthquake could impact highly populated parts of the North Island from Bay of Plenty through the Auckland and Northland regions with travel times of between one and two hours. Further, it is likely the causal earthquake would not be strongly felt because the volcanic region reduces the earthquake shaking, negating the effectiveness of using natural warning signs. Local-source tsunami caused by an earthquake on the Hikurangi subduction zone offshore of the east coast of the North Island also poses a threat, making tsunami the most crucial of the perils requiring early warning capability. 

Figure 1. A scene from within the Pacific Tsunami Warning Centre (PTWC) in Hawaii. PTWC currently provides tsunami advice on ocean-wide tsunami to all countries in the region (including New Zealand) using internationally available seismic and sea level data. The alerts provided by PTWC come around 10 minutes after a tsunami-generating earthquake so can only be used for distant and regional source tsunami warning.
Landslide Potential ...
Landslide potential is site specific, but forecasting can be based on rainfall rates, earthquake shaking and volcanic activity (lahars and other forms of debris flows) and the severity of likely landslides reported. This is an area of active research which is likely to bear fruit in the next decade.

Earthquake Early Warning....
Earthquake early warning, on the other hand, although already operational in places like Japan, is probably the lower priority for New Zealand because of the very short warning times, marginal outcome improvements and the much higher requirements for robustness and sensor densities for its effectiveness. Earthquake early warning is fundamentally different to the other early warning capabilities discussed above. We cannot predict the location or size of future earthquakes. We can only detect the start of an earthquake near where it ruptures and warn at a distance because seismic waves travel slower than electronic signals. Earthquake early warning times are measured in seconds, unlike the tens of minutes to hours and days possible with the other perils discussed above. New Zealander's live on top of our earthquakes! And we often have earthquakes in unexpected places making earthquake early warning very difficult. The Alpine Fault is the only structure in New Zealand where someday we may be able to deploy a cost effective earthquake early warning system.

Resources and Priorities....
For GeoNet to take on a leading role in event forecasting and early warning would require considerably more resources. Such an undertaking would be a step up in capability (people, expertise and resources) at least as large as when GeoNet was established in 2001. There is a compelling case for the establishment of a New Zealand tsunami early warning system for near-regional and local source events. But GeoNet can only provide part of the solution. Education, evacuation zone and route planning and a very effective public alerting system are also requirements for an effective end-to-end tsunami warning system.

Wednesday, February 18, 2015

The Future of GeoNet Revisited - Part 3: Impact Reporting

In my last blog I talked about the GeoNet Community - the large and growing group of people who rely on, use and are interested in GeoNet, our operations, data and other outputs. In this blog I will introduce the first of two fundamental changes I see happening to GeoNet and our community over the next decade.

Potential Impact reporting ....
Our vision is that GeoNet will be able to provide near-real-time potential impact reporting not just for earthquakes, but also for volcanic eruptions, tsunami impacts and landslide potential. The potential impact reports can then feed directly into systems designed to estimate the likely levels of damage given the people and infrastructure at risk. This is a major move from event reporting (where, when, how big) to impact reporting (what will be the likely effects where people or infrastructure reside). This reporting will use instrumental data, community reporting (citizen science) and effective modelling.

If we consider earthquakes, then felt intensity is a form of impact reporting. The magnitude of an earthquake estimates the physical size of the event where it ruptured – whereas the intensity relates to its impact on people, landforms, buildings and infrastructure. So reporting an earthquake location, depth and size is event reporting, but providing intensity estimates at multiple locations where people live and work is impact reporting. For a volcano, stating it has erupted is event reporting, but giving ballistic and ash fall damage estimates is impact reporting. You get the picture.

So why are we not doing impact reporting now? In short, because it is hard to get it right! Let’s consider earthquakes (again) as being a seismologist I find it easier. If we had enough sensors, then we could just use the sensor network to give an accurate estimate of the shaking level where you are for any earthquake. But the nearest sensor to you may be tens of kilometres away so we have to make an estimate based on the earthquake location, size and depth, the ground type below where you are and various other factors (like if you are in a multi-storied building). If you are lucky there may be a sensor near where you are, but it may be on different ground (soft rather than hard rock for example). We need either a huge increase in the number of sensor sites, or we can use known science to estimate the likely felt intensity anywhere. We can also supplement the physical measurements  and modelling with reports from people as explained in my last blog. 

ShakeMap NZ ....
The approach taken by the USGS ShakeMap, which we are in the process of implementing in New Zealand, is to use modelling and all available data. For example, Figure 1 shows the ShakeMap for the most recent large earthquake in New Zealand, the M 6.0 Wilberforce earthquake of 6 January 2015. In this case the nearest strong motion stations were a long distance from where the earthquake was centred so the maximum recorded shaking was less than 5% the force of gravity. However, ShakeMap estimated shaking levels of more than 20% of the force of gravity near the epicentre. 

Figure 1. ShakeMap of the Wilberforce earthquake of 6 January 2015 showing shaking intensity at the surface. The maximum accelerations indicated in the yellow and orange zones (around 0.2 g) could potentially have caused minor damage if the location was not so remote.
This shows both the strength and weakness of ShakeMap - it gives us an estimate of the maximum shaking levels but we can not confirm this value because we have no nearby instruments (see Figure 2). This event was also originally mis-located because of the influence of a small foreshock in a similar location which confused the automatic location system (an issue which was not identified by the Duty Officer immediately). Because ShakeMap requires the location, depth and magnitude as well as any actual shaking data to estimate the overall pattern of shaking, the mis-location caused the shaking pattern to be also wrongly-estimated. This was not a big problem in this case because of the remoteness of the earthquake from population centres, but this would not have happened if we had more sensors in the region. The remoteness of the location also meant we had few felt reports from close to the earthquake location.

Figure 2. The strong motion recordings for the Wilberforce earthquake of 6 January 2015. Note that the indicator bar are 10% the force of gravity and that there are no recording close to the earthquake epicentre.
The best choice we have is to improve our models of shaking AND to increase the number of sensors over time as I advocated in my original GeoNet technological blog series. Improving our knowledge of shaking requires more data on the earthquake source, the effects of the earth the earthquake waves travel through and the near-surface damping and amplification effects near where you require the shaking information. In many ways putting in more sensors is easier!

By providing improved potential impact reporting outputs like ShakeMap directly into systems providing damage and harm estimates GeoNet can make a major positive difference. In the modern world this is becoming more and more important making this development essential for the future of GeoNet and our community.

Future Event Scenarios
Before I move on to forecasting (or early warning) let's consider another step on that road. For recent volcano and earthquake events GeoNet has published a short list of what the most likely future scenarios may be along with the probabilities (chances) of what may happen. For example, for the Wilberforce earthquake discussed above we estimated that a normal aftershock sequence was by far the most likely future scenario, but other possibilities could not be totally ignored. In future GeoNet will provide this information following all geohazards events.