California is Safe?

According to a study of deaths from natural hazard "events"* across the U.S., earthquakes, volcanoes, and other spectacular geophysical hazards are much less deadly than common weather events like heat waves, floods, and thunderstorms. The study was published in the open-access International Journal of Health Geographics, so you should all be able to follow the link - but if you would rather read the summary version, the Los Angeles Times noticed that California is mostly safe, and ran with that angle.

The study's authors looked at county-level data on natural fatality hazards during the period 1970-2004 to produce a map of the most naturally hazardous areas in the United States:


Fans of earthquake doooooooooom will immediately see a problem in this methodology. There were no "Big Ones" during the period 1970-2004. There were tragic earthquakes - San Fernando, Whittier Narrows, Loma Prieta, Landers, Northridge, San Simeon - but nothing that resembled seismologists' doomsday scenarios for Los Angeles and San Francisco. No full-length ruptures of any of the faults that run immediately underneath densely-populated urban cores.

If a real Bay Area "Big One" on the Hayward or San Andreas fault, immediately underneath Oakland or San Mateo, strikes tomorrow afternoon - and it could - anywhere from several hundred to a few thousand people will probably die. In contrast, the SHELDUS database (used for the study) lists about 100 natural hazard deaths in the Bay Area for the entire study period, and 1,072 for all of California.**

I haven't crunched the numbers for how much a Big One would shift California's position relative to the national average. It's clear, though, that very large, very infrequent earthquakes are a significant part of the overall picture of natural hazard fatalities in California, and this study doesn't take them into account.

There are good reasons to suspect that we over-worry about rare catastrophes - they dominate our imaginations, and the news cycle, in a way that a steady trickle of deaths from summer thunderstorms can't. But we simply can't rely on data gathered during a relatively seismically quiet 34-year period to tell us how to compensate for this phenomenon.


*I put "events" in scare quotes because deaths related to what people consider "normal" weather conditions, rather than a noteworthy heat wave or cold snap, don't make it in to the databases used in this study. That includes, for example, many heat-related deaths in the Southwest.

**I don't know quite how to adjust for the weird way in which SHELDUS distributes deaths for multi-county or poorly located events. The Bay Area number is definitely undercounting deaths from Loma Prieta, and could be over- or under-counting deaths from other events as well, but I think the statewide number should be trustworthy.

Full reference: Kevin A Borden, Susan L Cutter (2008). Spatial patterns of natural hazards mortality in the United States International Journal of Health Geographics, 7 (1) DOI: 10.1186/1476-072X-7-64

More like this

I wonder what will happen to the map when they add in the 2005 numbers from Katrina -- New Orleans looks absolutely safe on this map.

Not sure if the authors corrected for this, but the kind of pattern shown here is very typical of county by county studies. Low population density counties seem to be more extreme. Arizona and Wyoming, for instance, have counties at both the high and low extremes adjacent to each other. California, having more high population density counties, tend more towards white.

Variation goes up (squared) as sample size goes down. In other studies, people have drawn spurious conclusions because they forgot to account for this.

The worst earthquake in the continental US was in 1812 near New Madrid, Louisiana. An earthquake of that magnitude in the same region could kill thousands but there's little earthquake preparation there (everyone is expecting the Big One to happen in California).

Lab Lemming: 1906 (+fire) killed over 3,000 people. Compare to this chart for hurricanes - roughly the same order of magnitude.

Zen: I admit to only skimming that bit, but they did attempt to correct for the small sample size. They run a geographical clustering test and show it in the next figure - but in the absence of data there's really only so much statistical chicanery you can do.

California is "mostly safe." Earth is "mostly harmless." Just a coincidence? Hmmm.

You're absolutely right in everything you said, of course. But do you ever recall seeing a news story that did *not* misunderstand and mis-use statistics?