Thank you. I’m honored to address this group and the broader KLCC audience. Let me start by saying that there is little doubt amongst people like myself who study seismic hazards that the Cascadia subduction zone can produce an earthquake LARGE enough, or that there is an unknown crustal fault in the Cascades that can produce an earthquake CLOSE enough, to significantly damage our dams upstream.

That said, we live with many FAR more likely risks and there are substantial benefits to having the dams, so we need to understand how likely it is that an earthquake will damage 1 or more dams, and what it would cost to significantly change those odds.

Forecasts, of strong ground motion at a site due to earthquakes, are usually calculated probabilistically. This is what I and my students and colleagues at the USGS do to make the National Seismic Hazard Earthquake Source Model, from which we generate maps of hazard and risk that are used to SET building codes, and to GUIDE insurance rates, risk reduction, and emergency preparedness efforts. You can find these products from links at my UO website or by searching “USGS Earthquake Hazards Program” and click on “Hazards” and then “Maps and Site specific data”.

The products are probabilistic for 2 reason. First, we don’t care if a magnitude 9+ earthquake off the coast or a 5.9 right under our feet brings down the roof above us; we want to know the chance of the roof coming down from all possible sources in a given period of time, usually the design lifetime of the structure. Second, we have extremely large uncertainties in the information that goes into the ground motion models, which, in a probabilistic model we can include in our forecasts.

For example, the geologic evidence for Cascadia earthquakes includes a range of sizes, locations, intervals between events, etc. The length, location, rupture direction and depth, and especially how shaking decreases with distance (called attenuation) affect how strongly and at what frequencies the ground will shake at a given site. We estimate and include a range of values for all these factors to determine how likely it is to exceed some threshold of shaking at a given frequency in a given period of time at the sites we are considering.

Okay, I have 2 reasons for boring you with a lecture in how to make a seismic hazard model. 1^st, the principle use of these models is to inform building codes, so we usually present the results as the level of shaking that will be exceeded 2% of the time in a 50- year interval. This seems reasonable as we don’t want to spend the money to completely earthquake proof all our structures but we want the chance of failure that results in loss of life to be pretty small. Unfortunately, this challenging level of shaking has been portrayed to the public as “likely” to happen or is even “overdue” in some

media reports. To get to the 2% in 50 year probability level (which is equivalent to a 98% chance shaking will be less in the next 50 years) you need the largest possible subduction zone earthquake, a 9.2+ that spans the entire subduction zone, and a very low attenuation or rate of decrease in shaking with distance. The most likely subduction zone earthquake in the next 50 years is actually a mid magnitude 8 off the SW Oregon coast that will cause no damage in the region of our dams.

Now, don’t over-react to this statement! There is a big difference between what is LIKELY to happen in a short period of time and what CAN happen. Unlikely events happen; people win the lottery. And for critical structures like dams whose failure have catastrophic consequences we want a much smaller probability than 2% in 50 years.

The second reason for my lecture is that there is a lot we can do to reduce the uncertainties in our estimates and improve our forecasts, and because low probability events come from the tails of our uncertainty distributions, reducing the uncertainties will almost certainly diminish our estimate of the hazard.

I’d like to end my presentation by shamelessly promoting the work of some of my UO colleagues who are working to better characterize and mitigate this hazard.

1^st Doug Toomey, who runs the Oregon part of the USGS’s Pacific Northwest Seismic Network is working to add an early warning component that can give us 10s of seconds to minutes of warning after a major earthquake occurs, that could be used to reduce the consequences of the shaking before it reaches the site.

2^nd, Amada Thomas is working on empirically measuring attenuation or how shaking decreases with distance and site conditions to reduce the very large uncertainty in our estimates of strong ground motion.

3^rd, Josh Roering is mapping and dating recent landslides to determine if they are earthquake triggered.

To be frank, I’m as worried about landslides as I am about ground motion from Cascadia. Either a Cascadia earthquake or an unknown crustal fault near a dam could rupture and trigger a large landslide into one of our reservoirs, which could lead to overtopping and subsequent erosion, as we came close to seeing by extreme flooding at the Oroville dam in California. Better mapping and dating of the faults and landslides in the Cascades with modern tools that allow us to see through the trees, like LIDAR, will help address this possibility.

Thank you for your attention and I look forward to answering your questions.