Understanding earthquake hazard maps 1/18/18 1

Understanding earthquake hazard maps

Americans have become used to color-coded warnings that purport to predict the risk of terrorist attacks. The warnings, which vary from terrifying red to sedate green, produce both fear and ridicule because we don’t know how the alert level is set. Presumably at any time it reflects a mix of information of varying reliability and U.S. politics.

The Department of Homeland Security terror alert scale.

Their geologic counterparts are colored hazard maps that predict the risk of ground shaking due to earthquakes. These are used worldwide to set insurance rates and choose building codes that specify standards for earthquake resistant construction. Because these impact millions of people, it’s worth understanding how the “earthquake alert” level is set. The best way to do this without going into the math is to look at how maps change depending on what mapmakers assume.

Most of what’s on the maps isn’t surprising and stays the same as maps are updated. For example, in U.S. maps the high hazard in California along the San Andreas Fault has been on maps for years and makes sense. That’s because we know enough about the earthquake history to feed into the computer programs to make sensible predictions about the hazard.

On the other hand, some features change dramatically from map to map. This occurs when we don’t know much, so the predicted hazard changes dramatically depending on the assumptions. A prime example is the New Madrid zone seismic zone in the central U.S., where large earthquakes occurred in 1811 and 1812 and small earthquakes continue today. Until 1996 maps showed the hazard in the New Madrid zone much smaller than in California, and then it became the Understanding earthquake hazard maps 1/18/18 2 same or higher. Similarly, Canadian maps used to have hazard “bulls eyes” along the east cost, but now have it smeared out.

We can understand what’s going on using the famous newspaper questions: how, where, when, and what?

The first question is: how do mapmakers define the hazard? Hazard isn’t a physical thing we measure. Instead, it’s something mapmakers define and then use computer programs to try to predict. Specifically, the hazard is defined as the shaking predicted to occur over a certain time. The shaking is measured in terms of the acceleration of the ground, because buildings respond to that. To see why, consider that putting a house on a railroad flatcar going 60 miles per hour wouldn’t hurt it – but a sudden stop would.

Worldwide, the usual definition of hazard is the shaking that there is a 10% of exceeding once in 50 years, or once every 50/10% or 500 yr. Maps made this way showed the earthquake hazard in the central U.S. much less than in California because large earthquakes are rarer. However in the 1990s the hazard was redefined through a complex process involving the U.S. Geological Survey (USGS), the Federal Emergency Management Agency (FEMA), and earthquake engineers working through a FEMA-sponsored group. The new definition, shaking that there is a 2% of exceeding once in 50 years, or once every 2500 years, made the Midwest hazard as big as California’s. Understanding earthquake hazard maps 1/18/18 3

Seismic hazard maps for the U.S., showing the expected shaking. A new definition in USGS maps changed the predicted seismic hazard in the New Madrid area in the central U.S. from much less dangerous than California (top) to as or more dangerous (bottom).

Although the predicted high hazard resulted simply from changing assumptions in a computer program, the new result had a major impact. It caused major issues for the insurance industry and pressure from FEMA on communities to adopt building codes as strong – and expensive – as in California. However, because 2500 years is much longer than the 50-year life of typical buildings, it’s not clear if the new definition makes economic sense. We don’t know whether the increased costs are justified by the increased benefits, or whether it would make sense to use a less stringent standard and free up resources for other uses. For example, it’s tricky to decide how to balance making hospitals stronger with treating uninsured patients. Understanding earthquake hazard maps 1/18/18 4

$A 100M$ seismic retrofit of the Memphis Veterans’ Administration hospital, removing nine floors, brought it to the California standard.

The second question is: where to assume earthquakes will happen? Some places, like the San Andreas Fault, are obvious. The tricky question is in places where we don’t have a good earthquake history. For example, along the east coast of Canada older maps showed high hazard only where there had been historic earthquakes. These “bulls eyes” didn’t make much sense, since we only have a few hundred years of earthquake data and so can’t really appreciate where on a longer time scale earthquakes may happen. Hence coauthors and I favored assuming that earthquakes could occur anywhere along the coast on the old faults remaining from when the continents split up, as is done in the new map. Understanding earthquake hazard maps 1/18/18 5

An older Geological Survey of Canada map assumed that the earthquake hazard along Canada’s east coast was concentrated where historic earthquakes had occurred, whereas a new map assumes the hazard is more uniform.

The third question is when to assume earthquakes will happen. Traditionally, hazard maps assume that the probability of a large earthquake on a fault is constant with time, such that a future earthquake is equally likely immediately after the past one and much later. An alternative is to use time-dependent models in which the probability is small shortly after the past one, and then increases with time. For times since the previous Understanding earthquake hazard maps 1/18/18 6 earthquake less than about 2/3 of the assumed average time between earthquakes, time-dependent models predict lower probabilities. Eventually, if a large earthquake hasn’t occurred by this time, a time-dependent model predicts higher probabilities and thus higher hazard. Using this model for the New Madrid seismic zone and Charleston, South Carolina areas predicts significantly lower hazards because Charleston and New Madrid are "early" in their cycles. In contrast, this model predicts that we’re “overdue” for an earthquake on the southern portion of the San Andreas Fault that broke in 1857 in the Fort Tejon earthquake.

Top: Comparison of earthquake probabilities on faults as a function of the average time T between earthquakes. Bottom: Assuming that earthquakes in the Charleston area occur on average 550 years apart, we are early in the cycle since the last one in 1886, so a time- dependent hazard map predicts much lower hazard than a time-independent map.

The last question is what to assume will happen when a large earthquake occurs. Mapmakers chose magnitudes for future earthquake and models that predict how much shaking will occur as a function of distance from the epicenter. Comparing maps for the New Madrid area shows that the assumed maximum magnitude of the largest earthquake on the main faults affects the predicted hazard especially near Understanding earthquake hazard maps 1/18/18 7 those faults, whereas the assumed ground-shaking model affects the hazard all over the area, because the model also includes smaller earthquakes off the main fault.

Comparison of hazard maps for the New Madrid area. The columns show the effect of changing the assumed magnitude of the maximum earthquake from 8 to 7, and the rows show the effect of two different ground motion models.

These examples show that different assumptions give quite different hazard maps. Because what to assume is often unclear, especially in regions inside plates like North America where large earthquakes are rare, hazard Understanding earthquake hazard maps 1/18/18 8 estimates have major uncertainties. Hence when we look at a hazard map, it’s good to remember that this is just one of a large number of quite different and equally likely maps one could make.

A good way to think about this is how it’s done in the atmospheric sciences. Weather forecasts often show a range of predictions, for example tracks of a major snowstorm, which show the uncertainties involved. Similarly in assessing the possible effects of global warming, the Intergovernmental Panel on Climate Change reports compare a wide range of different models.

Ultimately, we want to bear in mind seismologist Cinna Lomnitz’s observation that predicting earthquake hazards involves playing “a game of chance against nature of which we still don't know all the rules."