(07-03-2015, 12:16 PM)Island Chris Wrote: [ -> ]The following is only mildly interesting to me:
https://eos.org/project-updates/seismic-...the-models
What is a little more interesting is the comments about how they did their science discussions and reports.
Chris
Thank you for that. I really like that "Powell Blood Oath". Kinda ties in with some recent "passionate arguing" of my own on another board about the need for appropriate moderation of chat forums and the elimination of those who can't conform to commonly accepted societal norms for appropriate interaction. I even backed up my arguments with references. All for naught, I'm afraid.
Moving on, though, I still found something of interest in the subject of this article:
Quote:PSHA modeling typically seeks to strip out aftershocks, foreshocks, and swarms to isolate main shocks. This “declustering” is highly uncertain, leaving anywhere from 80% to 20% of the earthquakes as “main shocks.” The gathered participants agreed that there should be standardized declustering algorithms and tests of whether the declustered catalog exhibits Poissonian behavior...
I've thought about this myself in observing Roger's analyses of various quake predictors over the years. This topic always seems to be a point of contention. It is absolutely appropriate to remove certain events from a catalog for testing purposes, but those being tested always argue that Roger is being arbitrary, and thus flawed in his analysis. I agree only to the extent outlined in this article, that there is no standard method of declustering. But declustering must nevertheless be done to attain a fair review of predictive success.
And taking my thoughts a step further, I've often struggled with the absolutism of prediction analysis. Some have pointed out that even a .1 magnitude or a few kilometers makes the difference between a hit or a miss, potentially with radical changes in significance of results. If only there were some reasonable and logical method of making a 'fuzzy' analysis that would still impart meaningful results. But alas, I simply don't have the knowledge to even know if such is possible. Intuitively, it seems possible to me. But there could be very sound reasoning why it's not. (see, recognizing the weakness of my own argument
)
Brian
Over the years I have said that even not getting location at all could be interesting, if the magnitude range and time/date were significant. This discussion occurred most lately with Duffy's work. Speaking of which, he has not posted in a long time.
I have tended to ask Roger (and probably Brian) to make simple plots of events vs. date, so I/we can get a visual look at the data.
Chris
(07-04-2015, 02:12 PM)Island Chris Wrote: [ -> ]Over the years I have said that even not getting location at all could be interesting, if the magnitude range and time/date were significant. This discussion occurred most lately with Duffy's work. Speaking of which, he has not posted in a long time.
I have tended to ask Roger (and probably Brian) to make simple plots of events vs. date, so I/we can get a visual look at the data.
Chris
More specific please. Mag and date range?
Roger
(07-04-2015, 04:42 PM)Roger Hunter Wrote: [ -> ] (07-04-2015, 02:12 PM)Island Chris Wrote: [ -> ]Speaking of which, he has not posted in a long time.
I think he still lurks from time to time. His efforts are interesting. I like his approach. But unfortunately big quakes don't happen often enough so it will take time to gather more data.
(07-04-2015, 02:12 PM)Island Chris Wrote: [ -> ]I have tended to ask Roger (and probably Brian) to make simple plots of events vs. date, so I/we can get a visual look at the data.
More specific please. Mag and date range?
I think(?) Chris means in general that even getting just the magnitude and date right would be meaningful - "mag 7.5+ within 5 hours of date/time XXX". Do that enough times and I thing significance would be high. My skeptical mind though has trouble thinking of a way that there could be some precursory phenomenon that could be detected that could give size and time information without location. All forms of energy radiate from specific locations and directional information can be obtained with suitable instruments.
On the other hand, maybe Chris would like to see analyses of quake predictors without regard to location and see how they do. The software run through the catalogs should be very quick then.
Brian
(07-04-2015, 09:38 PM)Skywise Wrote: [ -> ] (07-04-2015, 04:42 PM)Roger Hunter Wrote: [ -> ] (07-04-2015, 02:12 PM)Island Chris Wrote: [ -> ]Speaking of which, he has not posted in a long time.
I think he still lurks from time to time. His efforts are interesting. I like his approach. But unfortunately big quakes don't happen often enough so it will take time to gather more data.
(07-04-2015, 02:12 PM)Island Chris Wrote: [ -> ]I have tended to ask Roger (and probably Brian) to make simple plots of events vs. date, so I/we can get a visual look at the data.
More specific please. Mag and date range?
I think(?) Chris means in general that even getting just the magnitude and date right would be meaningful - "mag 7.5+ within 5 hours of date/time XXX". Do that enough times and I thing significance would be high. My skeptical mind though has trouble thinking of a way that there could be some precursory phenomenon that could be detected that could give size and time information without location. All forms of energy radiate from specific locations and directional information can be obtained with suitable instruments.
On the other hand, maybe Chris would like to see analyses of quake predictors without regard to location and see how they do. The software run through the catalogs should be very quick then.
Brian
Oh, I see. Look at Amit's evaluation above. It's date vs prediction only.
Roger
(07-05-2015, 04:42 AM)Roger Hunter Wrote: [ -> ] (07-04-2015, 09:38 PM)Skywise Wrote: [ -> ]On the other hand, maybe Chris would like to see analyses of quake predictors without regard to location and see how they do. The software run through the catalogs should be very quick then.
Oh, I see. Look at Amit's evaluation above. It's date vs prediction only.
I should clarify myself. I meant even for predictors who do give locations. Do an analysis that doesn't consider their locations. How do they do on time and mag alone?
Egad, I've not been communicating well lately.
I think my brain is turning to mush. I've been stymied for the past week trying to learn how to use something mathematical called quaternions.
Brian
Roger,
Brian is mostly correct. I was just talking in general about the past on this page. I'm not asking for you to do an evaluation now. And, yes, there probably has not been anyone here who has gotten date and magnitude range correct. Instead, hypothetically, let's say someone did have a highly statistically significant record for date and magnitude ranges. In the past, you have posted that this is not useful, because presumably you can't save lives that way.
From my perspective, it would be useful and interesting; I would want to know the method, and it could eventually tell you something about the process of quakes being triggered or cascading from some creep or smaller quakes or whatever.
Chris
(07-05-2015, 12:25 PM)Island Chris Wrote: [ -> ]Roger,
Brian is mostly correct. I was just talking in general about the past on this page. I'm not asking for you to do an evaluation now. And, yes, there probably has not been anyone here who has gotten date and magnitude range correct. Instead, hypothetically, let's say someone did have a highly statistically significant record for date and magnitude ranges. In the past, you have posted that this is not useful, because presumably you can't save lives that way.
From my perspective, it would be useful and interesting; I would want to know the method, and it could eventually tell you something about the process of quakes being triggered or cascading from some creep or smaller quakes or whatever.
Chris
Yes, I agree. If someone is getting the date right then something is going on and deserves closer examination. In my evaluation of Amit I look only at date. I use mag 6+ quakes because the smaller ones happen so often any date would be a hit.
Roger