To play devil's advocate here, because as you know I'm also a numerical probability skeptic: the numerical probability aficionado might say something like "at some point, with limited resources, you have to decide whether to build that infrastructure in Seattle to hedge against Cascadia or in St. Louis to hedge against New Madrid, and what's your basis for decisions about allocation?"
I guess, but how does the numerical probability help?
You can say in Seattle: we have evidence that 41 insanely big quakes have occurred on this fault at fairly regular intervals over the past 10000 years.
Beyond this clear statement about the historical prevalence, what does numerical probability add?
I am wondering whether "probability" a place holder for commensurable units for earthquake hazard/risk? Is the main gripe here that it should be called "extrapolated frequency"? I'd be curious to hear more thoughts about problems where inaction is an action. We know earthquakes will happen, we can't retrofit all every building, where do we start? A related problem is earthquake insurance. If you can't come up with a number, you can't provide insurance. That is also a bad outcome.
This resonates so closely with the professional geoseismic research efforts of the government seismic entity in New Zealand named 'Af8' which focusses on the probability of the next major ( 8.5 rating) earthquake on the 'Alpine Fault' , which runs the length of most of the South Island of New Zealand. Both 'Cascadia' and the 'Alpine Fault' sit on the 'Ring of Fire' which has to be one of the most active friction areas on this planet - the 'Alpine Fault is also well 'overdue' based on the many quite regular movements during the past 2000 years!
I lived in Valparaíso, Chile shortly after their magnitude 9 in 2010. We felt a few aftershocks. It was surprising to me how little damage was done there. Due to building codes, most structures were undamaged and what was visible a few weeks later were some cracked walls and busted facades. The main casualties were from the tsunami.
Now I live in the PNW, and the infrastructure is much worse. At least my house is built of wood on a rocky hill, so it should remain mostly intact
Its important to also understand the broader relevance of this issue. Earthquakes are just the tip of the engineers' universe of extreme/catastrophic/existential risk v. minuscule/extremely low exceedance probability and likelihood impact estimation e.g. how do you turn 'how long is piece of string' stats into decision inflection points ie. 'acceptable risk' and thence what management is 'reasonable', how much to spend, and how many DALYs are institutionally tolerable.
There are many hazards like earthquakes whose predictability is similarly constrained for practical purposes albeit with significantly different circumstances e.g. Coronal Mass Ejections/Solar flares known from the Carrington Event which threaten any deep space travel and Moon/Mars based viability/fantasy but where the stats are even scarcer; the mundane area of floods v. building recommendations. Local catastrophic floods are so frequent as to be unavoidable. So what risk should be tolerable? Pandemics? Even modern water quality standards are derived from such statistical inference.
I think about Taleb’s point… you can, um, predict your exposure to an event much better than you can predict the event itself
Reducing your exposure to the event… even if it’s mass relocation… is much more important and in your control than the probability
Which Taleb book is this from?
Sometimes themes reoccur in all his books and technical papers
But it is certainly in Anti Fragile
To play devil's advocate here, because as you know I'm also a numerical probability skeptic: the numerical probability aficionado might say something like "at some point, with limited resources, you have to decide whether to build that infrastructure in Seattle to hedge against Cascadia or in St. Louis to hedge against New Madrid, and what's your basis for decisions about allocation?"
I guess, but how does the numerical probability help?
You can say in Seattle: we have evidence that 41 insanely big quakes have occurred on this fault at fairly regular intervals over the past 10000 years.
Beyond this clear statement about the historical prevalence, what does numerical probability add?
I am wondering whether "probability" a place holder for commensurable units for earthquake hazard/risk? Is the main gripe here that it should be called "extrapolated frequency"? I'd be curious to hear more thoughts about problems where inaction is an action. We know earthquakes will happen, we can't retrofit all every building, where do we start? A related problem is earthquake insurance. If you can't come up with a number, you can't provide insurance. That is also a bad outcome.
This resonates so closely with the professional geoseismic research efforts of the government seismic entity in New Zealand named 'Af8' which focusses on the probability of the next major ( 8.5 rating) earthquake on the 'Alpine Fault' , which runs the length of most of the South Island of New Zealand. Both 'Cascadia' and the 'Alpine Fault' sit on the 'Ring of Fire' which has to be one of the most active friction areas on this planet - the 'Alpine Fault is also well 'overdue' based on the many quite regular movements during the past 2000 years!
I lived in Valparaíso, Chile shortly after their magnitude 9 in 2010. We felt a few aftershocks. It was surprising to me how little damage was done there. Due to building codes, most structures were undamaged and what was visible a few weeks later were some cracked walls and busted facades. The main casualties were from the tsunami.
Now I live in the PNW, and the infrastructure is much worse. At least my house is built of wood on a rocky hill, so it should remain mostly intact
Its important to also understand the broader relevance of this issue. Earthquakes are just the tip of the engineers' universe of extreme/catastrophic/existential risk v. minuscule/extremely low exceedance probability and likelihood impact estimation e.g. how do you turn 'how long is piece of string' stats into decision inflection points ie. 'acceptable risk' and thence what management is 'reasonable', how much to spend, and how many DALYs are institutionally tolerable.
There are many hazards like earthquakes whose predictability is similarly constrained for practical purposes albeit with significantly different circumstances e.g. Coronal Mass Ejections/Solar flares known from the Carrington Event which threaten any deep space travel and Moon/Mars based viability/fantasy but where the stats are even scarcer; the mundane area of floods v. building recommendations. Local catastrophic floods are so frequent as to be unavoidable. So what risk should be tolerable? Pandemics? Even modern water quality standards are derived from such statistical inference.