Haiti continues to rebuild from the devastating earthquake of 11 years ago and is also dealing with the fallout from its assassination president in July. The magnitude 7.2 earthquake that struck its western region on August 14th was therefore a disaster for Haiti. The official death toll is currently at around 2000, but media reports for the first few days following the earthquake listed a death rate in the hundreds.
According to a U.S. Geological Survey scientific model, the actual number of deaths could be between five and 50 times higher. The tool, called Prompt Assessment for Global Earthquakes (PAGER), automatically integrates information about earthquakes with data from affected regions to determine the probable scale of the disaster. This includes the potential economic impact and deaths.
PAGER was created in 2007 and does not forecast the exact number. It estimates the likelihood that the number will be within a specified range. PAGER estimates that there is a 35% chance of between 10,000- 100,000 deaths in the earthquake in Haiti. There is a 32% chance that 100,000 will die.
This earthquake was likely less destructive than the Haitian earthquake of 2010, which PAGER accurately predicted at more than 100,000 deaths. It also estimated that there was a 66% chance that the damage to this year's earthquake would be between $100 million-$10 billion. Comparatively, the 2010 earthquake was estimated to have cost $7 billion to $14 billion.
It is important to have an accurate estimate of the number of deaths and the extent of damage before the response can be scaled. The official death count is the total number of bodies that have been found, and this is what news organizations report. It can take several months to determine the true toll in a country like Haiti with its poor infrastructure and restricted access to remote areas. This presents a problem for international aid organizations, countries, and other nations that need to assess the extent and decide the best response.
Scientific American spoke to David Wald, a USGS seismologist who helped create PAGER and other earthquake modeling tools.
[The transcript of the interview is below.]
Why is Haiti's official death rate so low compared to PAGERs?
You may not see many things that the models are predicting. You are missing small towns, villages, and isolated buildings that can be affected by earthquakes and landslides. Access to these areas will be difficult, as well as governance and civil infrastructure. As we expect, these [fatality] numbers are likely to increase over time. It is certain that roads will be affected by landslides. The roads traverse steep terrain, which was shaken.
What data is used by PAGER for modeling fatalities?
It is impossible to determine the exact number of deaths. There are too many unknowns and uncertainties. To make a prediction, you need three ingredients: the shaking, population exposed to each shaking level, and vulnerability of that population, based upon the buildings people are living in.
We know that larger earthquakes can produce complex patterns of shaking. ShakeMap attempts to capture this as accurately as possible. You could have thousands of stations to see the shaking, making it a very well-constrained map.
There are hundreds of stations in California. Additional uncertainty is present if you travel to Haiti, which has fewer seismic stations. It will always be more uncertain until Haiti is equipped with more instruments. The population that is exposed to the different shaking levels is the second ingredient. It is easy to take a population grid, and figure out the population that will be exposed to all the shaking intensity levels. The last question is what will happen to that population. You have extremely vulnerable buildings in China and Haiti. You can sustain many more fatalities from the same shaking than in California, New Zealand, or Japan. These countries have stronger building codes.
Haiti has a large population, strong shaking, and vulnerable buildings. This leads to the conclusion that PAGER will be a red alert, meaning that we are likely to have at least 1,000 deaths.
Why is there so much uncertainty?
There are not many instruments. Haiti has a seismic network that can detect earthquakes. We can't use these instruments for ShakeMap. They must be strong motion instruments. We only know data about two of the six to seven Haitian strong motion instruments.
Another source of uncertainty in our calculations comes from trying to determine [the shape and location of] the fault. This also depends on the data, how reliable the data are, and how complex the earthquake was. It is a difficult and complicated fault to identify in this instance. Sometimes these things can take months to resolve. We do our best to get it done within the first few days or hours.
The problem with modeling is that we don't know how many people died in 2010. These estimates vary from 100,000 to 320,000. The majority of countries have excellent reporting and you can assume that the estimates are accurate if there are three, four, or five fatalities. The truth can be elusive once you start to look at the really large losses like in Haiti.
While we cannot be perfect in this business, we can still be helpful. It was an international response to the Red Alert for Haiti.
How does PAGER detect an earthquake?
The National Earthquake Information Center determines the magnitude and epicenter. It is available 24/7 to send out information and trigger ShakeMap. This will create a map of shaking using the seismic data. PAGER starts as soon as ShakeMap has been created. It overlays the population and takes the shaking. It also estimates fatalities using the country's model. It sends the ShakeMap to other countries, where it can be used to estimate damage and shaking.
Because they are near the ocean or in a low-populated area, most earthquakes occur in the green zone. Even magnitude 8.0 earthquakes can be found in the subduction zones, which are collisions between tectonic plates that cause one to sink beneath the other. These collisions don't affect anything. PAGER's most important feature is the ability to state that nothing actually happened, rather than that something did.
We will be able to watch the model being generated if it is an orange or red alert, such as this one for Haiti. Before sending out the models, we will be reviewing the results. It is something we do for about 10 to 20 minutes, while we review the information coming in.
The official numbers are used in news reports, but they may be lower than the actual numbers. Is this a sign that people will not pay as much attention to the disaster?
There have been many cases in which an earthquake occurs at night and no one is paying attention. It will then be another news cycle with Afghanistan and COVID, and many other things that the media is paying attention to.
We were saying that it would be worse than what was initially reported. These uncertainties may lead to us overestimating the losses. Uncertain numbers can be a challenge for media. It is a challenge for agencies, financial institutions, and urban search-and rescue teams. We are pushing them to the point of worse than what we have seen so far.
Different PAGER users may have different timelines. Search and rescue in urban areas should be completed within hours. You can delay sending $1 billion of aid. Things tend to stabilize over time, so you get an answer that is a bit less constrained than your initial estimates. You can use the model as an estimate and add in the ground events to make the appropriate weight.
