WEBVTT

00:00:00.000 --> 00:00:05.000
[silence]

00:00:05.000 --> 00:00:06.000
[noise]

00:00:06.000 --> 00:00:12.000
Good morning. And we are back for Day 2 of the Northern California Earthquake Hazards Workshop.

00:00:12.000 --> 00:00:22.000
It is the 35th anniversary of the 1989 Loma Prieta earthquake and when we said what are some of the lessons that we want to learn from the Loma Prieta earthquake, we came up with two.

00:00:22.000 --> 00:00:32.000
And the first one was it really, really, really was a lesson in liquefaction and so to tell us about what we know about liquefaction hazards in Northern California 

00:00:32.000 --> 00:00:39.000
we have a session moderated by the fantabulous Rob Moss and Eric Thompson. Take it away, Rob and Eric.




00:00:39.000 --> 00:00:46.000
Thanks, Sarah. Our session is gonna be very intriguing this morning. We have four speakers.

00:00:46.000 --> 00:01:02.000
Tom Holzer, Jorge Macedo, Kristin Elmer, and Dave Engler.  And so we start off this morning with Tom Holzer.

00:01:02.000 --> 00:01:19.000
Thanks for inviting me to the NorCal Workshop. It's kind of hard to believe it's been 35 years since we experienced the Loma Prieta earthquake, but for those those of us who were active at the USGS in Menlo Park, it was 

00:01:19.000 --> 00:01:36.000
for many of us a career earthquake; we worked on it for years afterwards. So what I'd like to do this morning is briefly talk about some of the subsidence lessons that we learned and what we saw regarding liquefaction in the Loma Prieta earthquake.

00:01:36.000 --> 00:01:45.000
The title slide you're looking at shows another aspect of the earthquake. This was the first earthquake I experienced where many international visitors 

00:01:45.000 --> 00:01:53.000
showed up. This is Kenji Ishihara who is one of the leading geotechs in Japan actually.

00:01:53.000 --> 00:01:58.000
He did a lot of the original research on liquefaction, looking at a sand boil on 

00:01:58.000 --> 00:02:06.000
Treasure island and to Kenji's right, a young USGS Mike Bennett and John Tinsley,

00:02:06.000 --> 00:02:13.000
now retired of course. So what I want to do in the next few minutes is talk about where 

00:02:13.000 --> 00:02:20.000
the liquefaction occurred and what it did and then talk a little bit about it's legacy and then turn to the future 

00:02:20.000 --> 00:02:21.000
of the Bay Area

00:02:21.000 --> 00:02:36.000
regarding liquefaction. This is a map from Professional Paper 1551-B, which if you're really interested in liquefaction is sort of the 

00:02:36.000 --> 00:02:52.000
[indiscernible] of Loma Prieta liquefaction and what you can see is the liquefaction is actually very widespread down here to the south, the Pajaro and Salinas River.

00:02:52.000 --> 00:03:02.000
There was extensive liquefaction and in this image on the right you can see on you're left 

00:03:02.000 --> 00:03:09.000
the river levy along the Pajaro River has actually sunk a little bit and cracked. There was a lot of this 

00:03:09.000 --> 00:03:18.000
along the flood plain of the two rivers the Salinas. The largest most costly 

00:03:18.000 --> 00:03:30.000
damage was to the Moss Landing Marine Facility and it had to be replaced at a cost of about $8 million dollars.

00:03:30.000 --> 00:03:35.000
The other area of extensive liquefaction was in the East Bay and you can see multiple dots here.

00:03:35.000 --> 00:03:46.000
Both the airport and the port facilities incurred damage and then San Francisco actually had a lot of, liquefaction 

00:03:46.000 --> 00:03:54.000
of course, it's famous for the 1906 liquefaction.  Marina district up here on the north-end was 

00:03:54.000 --> 00:03:59.000
probably the most severely damaged in

00:03:59.000 --> 00:04:05.000
San Francisco in the photo you're looking at here; this is a sand boil that erupted in someone's background.

00:04:05.000 --> 00:04:31.000
there weren't many pipeline breakages and the gas pipeline system had to be replaced here. What's interesting about this map though is where there is in there aren't any dots here along Coyote Creek in the South Bay, which suffered, extensive liquefaction in 1906 there was none. So it probably is telling us that we're probably close to a threshold and in

00:04:31.000 --> 00:04:36.000
terms of triggering this liquefaction.

00:04:36.000 --> 00:04:46.000
Now rather than show pictures of the Marina district, I thought I would, or I'm sorry, rather than show pictures of the 

00:04:46.000 --> 00:04:55.000
Loma Prieta, I thought I would show pictures of more severe damage.

00:04:55.000 --> 00:04:57.000
In other words, once you pass that threshold because that's what we're really worried about for the Bay Area.

00:04:57.000 --> 00:05:16.000
In the 1906 San Francisco earthquake there was extensive pipeline damage and it was in fact that, pipeline damage that made it challenging to [indiscernible]

00:05:16.000 --> 00:05:31.000
and you can see in this picture from Branner Library. The severity of the fire in 1906 when it was really burning, this is the Ferry Building which survived.

00:05:31.000 --> 00:05:52.000
Another important thing that can happen from liquefaction is you lose bearing capacity failure so that the buildings that are not built on piles or appropriate foundations can move just just like that levy did in or wherever.

00:05:52.000 --> 00:05:59.000
The interesting thing about the 1964, and the Nigata earthquake, this earthquake in the Alaskan earthquake triggered engineering interest in liquefaction. Before these two earthquakes

00:05:59.000 --> 00:06:18.000
there had not been a lot of work done on liquefaction. So a lot of research conducted after 1964 had been completed by the Loma Prieta earthquake and we could actually test out some of that methodology.

00:06:18.000 --> 00:06:32.000
I'd like to turn now to what I think is probably the most important legacy of the. Loma Prieta earthquake and that is the Seismic Hazard Mapping Act.

00:06:32.000 --> 00:06:44.000
I'm going to say a little bit more at the end about how it came to be, but basically after Loma Prieta the legislature 

00:06:44.000 --> 00:07:11.000
enabled the California Geological Survey to map areas of potential liquefaction. What these maps are is they don't so much identify liquefaction as special studies zones where one is required under certain circumstances to do studies, but perhaps even more important, these maps triggered disclosure at the time of sales.

00:07:11.000 --> 00:07:24.000
So if you've bought a house in any of these areas where they've been mapped for liquefaction or seismic landslides, you've at least been alerted to it.

00:07:24.000 --> 00:07:50.000
And the reason for this was the residents in the Marina district we're surprised by what happened to them and it was their lack of knowledge that led to this system of at least notifying people of a potential hazard so they could do something about it; either not buy the property or address the hazard.

00:07:50.000 --> 00:08:05.000
And what these maps look like, the liquefaction maps, is they just color the area where there's the potential for liquefaction is based on geology and water level and and shaking potential.

00:08:05.000 --> 00:08:13.000
And the state makes the judgment and then delineates these areas. This is the Mountain View quad. You can see it's an extensive area.

00:08:13.000 --> 00:08:23.000
Not all of this area will liquefy, but we would expect that some of it might, and so a special study is required.

00:08:23.000 --> 00:08:28.000
Now, I mentioned that a lot of research had been done since 1964 and one of the primary 

00:08:28.000 --> 00:08:40.000
things that had been research was how to actually improve the ground, how to make it resistant to liquefaction to mitigate it.

00:08:40.000 --> 00:08:47.000
And what we're looking at is two of many methods that are now available, the stone column you 

00:08:47.000 --> 00:09:02.000
vibrate the soil and actually put a stone column into the ground so a permeable stone column so that the water can escape from the liquefying soil; so the soil doesn't

00:09:02.000 --> 00:09:15.000
build up pore pressures in it. Another if you don't have structures on the site is simply to compact the soil by dropping a heavy weight which you can see in the other slide.

00:09:15.000 --> 00:09:23.000
These are images from Hayward Baker a big ground improvement company.

00:09:23.000 --> 00:09:30.000
And there were actually quite a few sites in the Bay Area that had been improved. So it shows you the level of awareness.

00:09:30.000 --> 00:09:33.000
Jim Mitchell and Fred Wentz did this compilation of sites.

00:09:33.000 --> 00:09:44.000
The only shortcoming to the study and when they acknowledged was that the ground motions really weren't to design levels 

00:09:44.000 --> 00:09:46.000
the shaking was a little bit smaller, but the point is these sites tended to work pretty well.

00:09:46.000 --> 00:09:59.000
The other thing that got tested by the earthquake was what is known as the Saint Edres simplified procedure.

00:09:59.000 --> 00:10:05.000
this is the standard tool engineers use to 

00:10:05.000 --> 00:10:11.000
look to test for liquefaction.

00:10:11.000 --> 00:10:24.000
Most of these improvements were on Treasure Island. This is a map of the filled ground in San Francisco and much of this ground suffered liquefaction to some extent.

00:10:24.000 --> 00:10:30.000
I've already mentioned the Marina District, which is up here. This is a post-1906 

00:10:30.000 --> 00:10:33.000
event that really suffered major damage. In fact, it was the most seriously damaged facility

00:10:33.000 --> 00:10:49.000
in San Francisco. But what really was important was the East Bay fills and what I want you to do is to 

00:10:49.000 --> 00:11:02.000
focus in this area here, down here. And I'm gonna place the fills that are there today and their post-1906 fills,

00:11:02.000 --> 00:11:12.000
and you can see there's an extensive amount of fills, largely hydraulic, sandy hydraulic fills which are susceptible to liquefaction.

00:11:12.000 --> 00:11:20.000
and one of the things we did after all, more creative was to develop a methodology to predict the liquefaction on a probabilistic basis.

00:11:20.000 --> 00:11:26.000
And this is what those fills would look like in a repeat of the 1906 earthquake.

00:11:26.000 --> 00:11:31.000
You can see there's a high probability, point 7 to point 8. Most of these fills liquefied.

00:11:31.000 --> 00:11:49.000
Of course, there's going to be areas of ground improvement in here, but the point is for these fields, Loma Pieta was their first serious shaking and about 14% of the fills, liquefied in that earthquake.

00:11:49.000 --> 00:11:56.000
And we can look to much worse, like both action in a 1906 event.

00:11:56.000 --> 00:12:05.000
This is the scenario map for the Hayward earthquake and you can see the, liquefaction is still pretty 

00:12:05.000 --> 00:12:11.000
Severe, greater than point 2 to point 3 over most of the fills,

00:12:11.000 --> 00:12:24.000
but not quite as severe as the 1906 earthquake and it presumably is the magnitude the duration of shaking that is making the difference here.

00:12:24.000 --> 00:12:50.000
Well, I wanted to close with some reflections and I've said refractions to be a little facetious on the Loma Prieta earthquake. As I was just showing you the earthquake really was a reminder of the liquefaction hazard in the Bay Area and this is prompted a lot of public and private entities to mitigate this hazard.

00:12:50.000 --> 00:12:59.000
This and the seismic shaking hazard were estimated by Tom Brocher (USGS) in a 2018 study that probably about $80 billion dollars has been spent since Loma Prieta to mitigate these hazards.

00:12:59.000 --> 00:13:11.000
And I'm sure, I didn't look at it, the study carefully but I'm sure the liquefaction mitigation efforts have been in the billions.

00:13:11.000 --> 00:13:25.000
The other thing I wanted to mention is that there were people responsible for a lot of these public and private responsible.

00:13:25.000 --> 00:13:34.000
And I sort of called them heroes here, but for example, after the earthquake, 

00:13:34.000 --> 00:13:59.000
Congress passed a supplemental and about $20 million dollars came to the USGS for post-earthquake investigations and the reason this happened was Rob Wesson who was the office chief at the time and, Virgil Frizell had educated a lot of Congressional staffers about the seismic hazards

00:13:59.000 --> 00:14:06.000
in the Bay Area and what could be done about it as well as other locations in the United States.

00:14:06.000 --> 00:14:17.000
And Arch Johnson who was the head of the Tennessee earthquake research group had educated Al Gore about it.

00:14:17.000 --> 00:14:21.000
So, my point is, there were people already knowledgeable about things that could be done.

00:14:21.000 --> 00:14:35.000
At the state level, Willie Brown, who represented the Marina District and was in the state assembly, approached Jim Davis, the state geologist,

00:14:35.000 --> 00:14:42.000
and Jim had the Seismic Hazard Mapping Act in his pocket and that was how that came to be, 

00:14:42.000 --> 00:14:54.000
And then after the Loma Prieta earthquake Lloyd Cluff at PG&E developed the CRADA with the USGS that led to a lot of the research.

00:14:54.000 --> 00:15:03.000
on the Loma Prieta earthquake that was relevant to utilities and things that PG&E was interested in.

00:15:03.000 --> 00:15:11.000
And then finally, I can't underestimate the role of project IMPACT. FEMA for the first time had a administrator who was an emergency manager by background, James Lee Witt.

00:15:11.000 --> 00:15:30.000
He instigated a cooperative effort between communities and federal and state agencies that were designed to 

00:15:30.000 --> 00:15:53.000
help communities adapt and mitigate their hazard. And basically this provided for my team unlimited access for borings and soundings and we were able to confirm the liquefaction hazards in the East Bay fills that I was sharing and do it in the context of the City being aware of what we were doing.

00:15:53.000 --> 00:16:01.000
Well with that I'll close and thank you very much for this opportunity to look back 35 years ago.

00:16:01.000 --> 00:16:05.000
Thank you.

00:16:05.000 --> 00:16:11.000
Alright, well thanks so much for that nice overview of Loma Prieta liquefaction 

00:16:11.000 --> 00:16:13.000
Tom. I'll just remind everyone that we'll have time; we have a nice block 

00:16:13.000 --> 00:16:23.000
for discussion and questions. After everybody's spoken, after everyone's given their talk.

00:16:23.000 --> 00:16:36.000
Our next talk will be, Jorge Macedo from Georgia Tech and he'll be focusing on the Harbor Bay Business Park District.

00:16:36.000 --> 00:16:51.000
Hello everyone, I'm Jorge Macedo from Georgia Institute of Technology. I'm gonna be discussing on the Harbor Bay Business Park liquefaction during Loma Prieta earthquake with a critical station mechanics perspective.

00:16:51.000 --> 00:16:58.000
And I want to thank the organizers for inviting me, it's a pleasure to be here and Looking forward to any discussion.

00:16:58.000 --> 00:17:05.000
Something else I wanted to mention is that the content of this presentation is actually published in SD (Science Direct) paper.

00:17:05.000 --> 00:17:12.000
Are you gonna be focusing on some aspect and I'll do my best to cover as much as I can in the 15 minutes I have, okay.

00:17:12.000 --> 00:17:33.000
So, first of all, this is the study area. This is actually on the international telegraph plus and in particular address that we used to reference you know our assessments was at 4020 Harbor Bay Parkway that is located within that area.

00:17:33.000 --> 00:17:42.000
And this is a quite interesting area from a liquefaction perspective because first of all it did experience liquefaction during the Loma Prieta earthquake.

00:17:42.000 --> 00:18:08.000
And also there was subsurface investigations that are shown over here that, were available before the earthquake, after the earthquake, and also they expanded over several years after the earthquake because you can see here for instance there are boreholes from 1979 to 1984 that are the purple circles over here. There are also boreholes in 2019 

00:18:08.000 --> 00:18:17.000
these are relatively reason was that are the green circles over here. There are CPTs from 1984 that are the blue triangles.

00:18:17.000 --> 00:18:31.000
A CPT from 1989 that are the red triangles over here that those were after earthquake and also, CPT is a recent safety relatively recent from 2018 and 2019.

00:18:31.000 --> 00:18:45.000
So this makes, you know, this case history quite interesting because of you know, rich information in terms of subsurface characterization and also the liquefaction observations.

00:18:45.000 --> 00:18:54.000
For example, here you can see examples, this is just a selected example of liquefaction manifestation.

00:18:54.000 --> 00:18:57.000
You can see a little bit of liquefaction features over here in that area. Also, some typical objective features on the back part of the photo.

00:18:57.000 --> 00:19:08.000
I apologize I didn't have a close photo over here. Remember this is 1989. There are not many photos that are available.

00:19:08.000 --> 00:19:21.000
And this one over here, you can see, also some signatures of [indiscernible] even within that area that is having across.

00:19:21.000 --> 00:19:34.000
So, there were a lot of liquefaction observations all over that area after the earthquake. So let me show you in a general sense the stratigraphic features,

00:19:34.000 --> 00:19:43.000
subsurface conditions at this location as well, and I will show your representative CPT for that purpose.

00:19:43.000 --> 00:20:00.000
Remember a CPT is pretty much proof that we, technical engineers we put into the ground in order to collect three main things typically, the deep resistance, the slip friction on the CPT and also the pore pressure,

00:20:00.000 --> 00:20:08.000
that data is interpreted and we infer subsurface information from that data. For instance, here what you are seeing is that the persistence on the left.

00:20:08.000 --> 00:20:15.000
You are seeing the pore pressure here. You're seeing the BQBQ is normalized pore pressure 

00:20:15.000 --> 00:20:21.000
parameter, and we typically use it to get more context for the response of the material.

00:20:21.000 --> 00:20:28.000
Typically, I would say for BQ material higher than 0.4 to 0.6 

00:20:28.000 --> 00:20:42.000
we are seeing I would say a fine grain in material. This is evident, for example, in this plot, the first layers are sandy material, so BQ is 0 and then once you get to the Holocene Bay Mud you see an increment on BQ.

00:20:42.000 --> 00:20:50.000
So, the friction ratio is a normalization between the two persistence and the frictionless lift.

00:20:50.000 --> 00:21:03.000
We use it also to infer compressibility, the higher the value, the more compressible the material for example. And then CPT data can be also used to interpret what is the Andrena strain, which is 

00:21:03..000 --> 00:21:09.000
shown over here on this plot on the right. So anyways from the data that we collected 

00:21:09.000 --> 00:21:25.000
different units when infer that are indicated here from 1 to 9. The first 2.5 m are identified on sand fill. Then from 2.5 m to 6 m approximately there is a fill over here that is a loose sand fill which interbedded in the 

00:21:25.000 --> 00:21:33.000
sand, excuse me, sand, sorry, and also sand/clay mixes, which you can see here by the variation of the friction,

00:21:33.000 --> 00:21:51.000
by the slip friction. So then, you have the Holocene Bay Mud over here, which is followed by the Merritt Sand you can see the Holocene Bay Mud, as I said, is showing BQ values and then the transition to the sand brings the values back to 0.

00:21:51.000 --> 00:22:01.000
And then there is a transition to the place, which are also different units that can be used in [indiscernible] behavior, intermediate behavior or contractive behavior.

00:22:01.000 --> 00:22:07.000
This behavior was inferred from all their observations as I will be discussing.

00:22:07.000 --> 00:22:20.000
Anyways, this is a typical stratigraphy on that particular side. Before going into more details, oh, and something else I wanted to mention as well that I forgot. I just realized I forgot for this presentation,

00:22:20.000 --> 00:22:31.000
I'm going to be focusing on the loose and feel that are the units 2, 3, and 4. And I don't have the time to focus on the other units.

00:22:31.000 --> 00:22:39.000
Those other units are discussed in the paper I mentioned. So, for these units, I would be showing how the characterization was conducted 

00:22:39.000 --> 00:22:47.000
and how critical station mechanic was used. Before going into that, something quickly I wanted to share was the comparisons of CPTs,

00:22:47.000 --> 00:22:58.000
before and after the earthquake. So, the left plot is showing for instance and in a comparisons of CPTs at the ITP-2

00:22:58.000 --> 00:23:08.000
location that is showing the before earthquakes are showing with this stream line, I think, a gray line,

00:23:08.000 --> 00:23:14.000
and then the after are shown with a with a solid blue lines. Okay, the same applies for the three

00:23:14.000 --> 00:23:25.000
plots over here. I just wanted to highlight a couple of features. You know the [indiscernible] that we upset was not necessarily the same for instance the left and the right 

00:23:25.000 --> 00:23:41.000
plots are showing different trends pretty much our general observation from this is that effects can be different and you know depending on particular soil properties at very specific locations.

00:23:41.000 --> 00:23:49.000
Even though CPT is within a relatively narrow area, even then we saw different responses and that may be related also to the fabric 

00:23:49.000 --> 00:24:00.000
or how the particles are arranged at particular locations. There are also similar observations after the contaminated sequence.

00:24:00.000 --> 00:24:22.000
So, you know, it's up to discussion how liquefaction is changing the soil fabric after that change is going to be responding because typically our assessments are conducted with the data collected after the earthquake in many cases or many case history is not necessarily the data before the earthquake.

00:24:22.000 --> 00:24:47.000
On this particular case history interestingly there was data before and after. So, now let me show you how the combination of the normalized deep resistance y axis here and the normalized friction ratio x-axis here how this combination is landing in a CPT chart 

00:24:47.000 --> 00:24:58.000
where you can see on a spectrum of behavior from gravity sands to place another sensitive source and also if you look in this direction you can see different states.

00:24:58.000 --> 00:25:06.000
Typically the negative; this lever over here on the side is pretty much a related expressing the state parameter 

00:25:06.000 --> 00:25:15.000
and a state parameter is one way to define soil state. Typically engineers, for example, use a relative density.

00:25:15.000 --> 00:25:30.000
But whether relative density is, I would say, limited compared to state parameter cause state parameter in addition to the information on density, it's also bringing information on confinement and both are important.

00:25:30.000 --> 00:25:36.000
Anyway, there can be discussions on that. I'll be more than happy to be engaged in those discussions.

00:25:36.000 --> 00:25:51.000
Well, as I said, for this unit this is the field unit. You can see the Deng's materials over here that are on the top and then you can see also some a lightly dilate and some materials within that field.

00:25:51.000 --> 00:26:08.000
And you can see also the interpreter materials are generally contractive. The distinction between dilate and contractive on that plot is pretty much don't based on this line over here that is expressing the state parameter equal to -0.05.

00:26:08.000 --> 00:26:17.000
The points that are below that line will be the dilative and, sorry, contractive and the points that are above will be dilated.

00:26:17.000 --> 00:26:27.000
So, something else that we did for the field materials is to collect samples. And what you can see here is the range of, you know, particle 

00:26:27.000 --> 00:26:47.000
sales distributions for the field material; these are the black lines, and then you can see the one representative degradation on the orange line and how that degradation compares with the well-known sand or the database. In purple you can see also the degradation of the introverted layers.

00:26:47.000 --> 00:26:57.000
So in critical, what we do is pretty much team first, all behavior based on the state of the soil.

00:26:57.000 --> 00:27:13.000
In other words in that case the state parameter and also the mechanical or mechanistic properties of a soil material. The mechanistic properties being evaluated through, typically to laboratory testing or other, other means.

00:27:13.000 --> 00:27:21.000
But the idea is that once you know those properties, the mechanical properties and the state, you can infer soil behavior in general.

00:27:21.000 --> 00:27:29.000
So, in that particular case, we obtain the mechanistic properties based on free action testing.

00:27:29.000 --> 00:27:51.000
And what you are seeing here is the results from that testing, the left plot over here is range ratio versus effective stress and this used to evaluate the critical state line and the parameter is a dark  evaluated through that are the altitude of the critical state line and also there's low being a proxy for compressibility.

00:27:51.000 --> 00:28:00.000
Then we evaluate also the strength, the latency relationship. Every point here represents a triaxel test.

00:28:00.000 --> 00:28:16.000
And it is just representing the maximum state ratio and a maximum delay latency. This is relatively straightforward to be obtained from reaction tests and then this plot here is representing the state parameter and maximum dilatency in the maximum day license.
 
00:28:16.000 --> 00:28:26.000
So from this plot, as I said to a mechanistic parameters are elevated from this one, an order to the critical state friction.

00:28:26.000 --> 00:28:44.000
A ratio and also, the particular biometric coupling that is also bringing compressibility information and then Kai is a kinematic parameter over here that is expressing the potential of a soil to dilate.

00:28:44.000 --> 00:28:53.000
So as something else we do on, critical mechanics is to infer soil state from the mesh or CPT information.

00:28:53.000 --> 00:29:02.000
In that case, the, y-axis is representing the CPT information and x-axis is representing the soil state.

00:29:02.000 --> 00:29:09.000
So, typically, private expansion simulations are conducted in order to get the relationship between those 

00:29:09.000 --> 00:29:18.000
and as you can see here these lines are dependent on I_R support express stiffness and B expressed confinement.

00:29:18.000 --> 00:29:26.000
So depending on that combination there can be different lines. So, in that particular study, we did infer properties.

00:29:26.000 --> 00:29:48.000
Here the y-axis is as a coefficient that is used for the inversion. We did infer properties for 2 particular gradations and those allowed us to relate those properties for the inversion with CPT properties as well and to correct the Plewes method for our particular case because if Plewes method was departing from the trends that we were seeing.

00:29:48.000 --> 00:29:55.000
So we use all of that information to get or to infer what is the state parameter from different CPTs.

00:29:55.000 --> 00:30:05.000
This is shown over here each, you know, entry over here is a different CPT, the y-axis is that and then the x-axis is the state parameter.

00:30:05.000 --> 00:30:29.000
And as you can see here, you know, the state parameter is predominantly positive. What we typically do is to use a 80 or 90% type of this distribution in order to infer or to, what was the soil response because you know it is consider that this person time will be dominating some behavior.

00:30:29.000 --> 00:30:39.000
So anyways, that is how the state parameter was evaluated and now we have the state parameter and we have also the mechanical properties that were shown over here.

00:30:39.000 --> 00:30:56.000
In other words, we can infer behavior. And that's the particular aspect of our criticalization mechanics that it does't matter to say it in a way what is soil, if it is sandy or fine grain material, if you call the magnificent properties and the state parameter you can enforce our behavior.

00:30:56.000 --> 00:31:04.000
So, and now this is for completeness to show you how the behavior of the way math was typically.

00:31:04.000 --> 00:31:25.000
So you can see that it was having significant portions on the contractive side of these charts and one interesting aspect on the very mat is that even though they might this very much put half a positive state per a meter because of its large compressibility it can actually show a hardening behavior.

00:31:25.000 --> 00:31:32.000
And that was particularly important, you know, for processing the system response on that particular site as detailed on the paper.

00:31:32.000 --> 00:31:41.000
But I wanted to highlight that because typically the opposite would be expected, that the positive state parameter is going to be, you know, associated with the brittle response.

00:31:41.000 --> 00:31:52.000
So, and yes, for completeness, this is also the distribution on how the combination of the normalized the persistence and friction ratio look like for the place to send materials.

00:31:52.000 --> 00:32:00.000
So the other part of the budget here is, you know, you need the mechanical properties and you need a state, but you need to also have the demand.

00:32:00.000 --> 00:32:13.000
The demand was estimated based on recordings and the Yerba Island station that was a rock recording and also a recording on the surface based on the A and the NAS station that is showing liquefaction

00:32:13.000 --> 00:32:22.000
features as you can see here. So, that information, the ground motion recordings was used to estimate the variation of the cyclic threat ratio.

00:32:22.000 --> 00:32:45.000
And then once the cyclic threat ratio was estimated, the representative value was selected as typically done and we plotted the cyclic third ratio which is shown over here this is red area with the characteristic as a parameter that was inferred based on the CPT data and the mechanical properties that I described based on the CPT data and the mechanistic, the mechanical properties that I described before.

00:32:45.000 --> 00:32:51.000
So when we plot them in a chart that is, I would say, maybe a bit different from the common chart.

00:32:51.000 --> 00:33:00.000
Because now it is relating the CSR with the correct characteristic state parameter and you know, it was infer that the site was actually liquefiable.

00:33:00.000 --> 00:33:27.000
So this is just an example on how to use criticalization mechanics to certain perfection triggering. More detail discussions about the John Wayne mud behavior and other units are on the paper I mentioned, but I wanted to close the talk by highlighting that in criticalization mechanics it doesn't matter really if it's assigned or a find grain material as far as you have the mechanical properties and the state you can infer the behavior and in particular

00:33:27.000 --> 00:33:43.000
compressibility is one of the key aspects of such behavior. That departs from classical procedures because those are more empirical based and typically formulated for clean sense or signs with low fine contents.

00:33:43.000 --> 00:33:50.000
So happy to discuss more about that. Thank you very much for for your attention.

00:33:50.000 --> 00:34:02.000
Alright, great. Thanks so much Jorge for that fantastic talk. Our next talk will be by Kristen Ulmer.

00:34:02.000 --> 00:34:13.000
Thank you for inviting me to present today. As you can see from the title slide here I am going to present an energy-based framework for evaluating liquefaction.

00:34:13.000 --> 00:34:19.000
I'll begin with the definition of liquefaction, followed by an overview of liquefaction evaluation procedures.

00:34:19.000 --> 00:34:27.000
I'll talk about our proposed energy based procedure and then wrap up with some conclusions. In it's most basic definition 

00:34:27.000 --> 00:34:35.000
liquefaction initiation is the total transfer of overburdened stress from the soil skeleton to the poor fluid under undrained cyclic loading.

00:34:35.000 --> 00:34:45.000
This occurs predominantly in saturated sandy soils with contractive tendencies. As the excess pore water pressure increases or Delta U increases 

00:34:45.000 --> 00:34:56.000
you have a decrease in vertical effective stress. The excess pore pressure ratio that Ru, which is the ratio of the excess poor water pressure over the vertical effect of stress, increases.

00:34:56.000 --> 00:35:04.000
If you get an Ru ratio equal to one, liquefaction initiates. As, I'm sure you are already aware 

00:35:04.000 --> 00:35:16.000
this leads to a lot of issues as the risk to infrastructure. So it's important for us to be able to predict liquefaction initiation in a forward analysis so that we can try to mitigate 

00:35:16.000 --> 00:35:30.000
those issues. So there are several liquefaction evaluation procedures available to help us do that. These tend to provide a way to estimate the demand and imposed by an earthquake and the capacity of a given soil to resist liquefaction.

00:35:30.000 --> 00:35:37.000
Given the demand and the capacity, a practitioner could estimate a factor of safety or the probability of liquefaction.

00:35:37.000 --> 00:35:46.000
There are three main categories of liquefaction evaluation procedures. There's the most commonly used stress based methods or the simplified method.

00:35:46.000 --> 00:35:51.000
Which there are several varieties of that. There's the strain-based method, which is not typically used, but it's out there.

00:35:51.000 --> 00:35:59.000
I won't have time to talk about this one today. And then there are energy based methods, one of which is what we are proposing as a group.

00:35:59.000 --> 00:36:07.000
Let's start first with the stress-based method so I understand the differences between these. To compute demand for stress-based method.

00:36:07.000 --> 00:36:14.000
You compute the cyclic stress ratio, which is the average to your stress divided by the initial effective vertical stress.

00:36:14.000 --> 00:36:26.000
Ideally we would have some sort of share stress time history and we take the maximum sheer stress and multiply it by a ratio of 0.65 which is traditionally used to get an average sheer stress.

00:36:26.000 --> 00:36:40.000
I usually don't have a shared stress time history. We usually have an acceleration time history. That's at some site here's ground surface the water table and some susceptible soil layer 

00:36:40.000 --> 00:36:42.000
and we want to know what's the factor of safety and we want to know what's the factor of safety against the liquefaction for this soil layer.

00:36:42.000 --> 00:36:56.000
We might be able to get an A max at the ground surface and multiply by 0.65 times that A max. Using Newton's laws be able to compute these CSR as the shear stress, average shear stress 

00:36:56.000 --> 00:37:08.000
can then be written now. Instead as the maximum acceleration multiplied by the total vertical stress  multiply by Rd, which is a stress reduction factor accounting for non-rigid response of the soil column.

00:37:08.000 --> 00:37:16.000
So the entire soil column doesn't move rigidly, usually you have some flexibility in there. 

00:37:16.000 --> 00:37:22.000
So then there are a few other correction factors that are typically added on in order to compute what I'm calling here CSR star.

00:37:22.000 --> 00:37:29.000
This is the magnitude and the overburden and sheer stress static shear stress corrected CSR.

00:37:29.000 --> 00:37:38.000
Magnitude scaling factor MSF accounts for duration of ground shaking the K sigma accounts for effects of overburden stress.

00:37:38.000 --> 00:37:48.000
And the K alpha effects for accounts for non-level ground conditions or some other initial sheer stress as being imposed on the soil.

00:37:48.000 --> 00:37:55.000
So you take the CSR star, computed for a bunch of different soil layers in a case history data set 

00:37:55.000 --> 00:38:08.000
and the institute test metric of some kind for those cases histories. So that could be the corrected SPT blow count the corrected content resistance we're corrected chewy velocity 

00:38:08.000 --> 00:38:14.000
And try to draw a line between liquefaction case histories and no liquefaction case histories.

00:38:14.000 --> 00:38:24.000
And several modelers have done this and basically there are several capacity curves that are available to try and compute what we call these cyclic resistance ratio, the CRR.

00:38:24.000 --> 00:38:32.000
So for some institute test metric value, you could compute the CRR and know your resistance against liquefaction.

00:38:32.000 --> 00:38:38.000
And then you could compute factor safety against liquefaction as CRR over CSR start.

00:38:38.000 --> 00:38:40.000
Now let's talk about energy based methods. Dissipated energy per unit volume of soil has been shown to be closely correlated with excess poor pressure generation.

00:38:40.000 --> 00:38:59.000
Perhaps more so than just share stress alone. The dissipated energy is computed as the cumulative area bound by shear stress versus shear strain histories as loops.

00:38:59.000 --> 00:39:09.000
So on a stress-based method, you account for just the sheer stress essentially in the energy based method we're proposing we account for both shear stress and shear strain.

00:39:09.000 --> 00:39:17.000
As you can see the history says loops it starts out as basically the same size, small loops.

00:39:17.000 --> 00:39:27.000
And the loops increase in size as the soil softens due to excess poor water pressure generation. Our method aims to overcome some of the issues with previous energy based methods, most of which the issues were that they were oversimplified.

00:39:27.000 --> 00:39:37.000
Or they weren't simplified enough. They were too crude or they were too complicated.

00:39:37.000 --> 00:39:46.000
But we want to be able to maintain the simplicity of simplified stress based methods. While capturing the benefit of having both shear strain and shear stress

00:39:46.000 --> 00:39:55.000
linked in our energy based method. So what we like to have is to have dissipated energy on the vertical axis 

00:39:55.000 --> 00:40:03.000
and analogous to the CSR star that we were showing previously. Again, some institute test metric for a bunch of case histories.

00:40:03.000 --> 00:40:13.000
And identify boundary line between the liquefaction and liquefaction cases so that we can identify the capacity in terms of dissipated energy to resist liquefaction.

00:40:13.000 --> 00:40:26.000
And then compute factor safety against the liquefaction.  The questions we set out to answer include how do we compute dissipated energy in the first place for all these case histories?

00:40:26.000 --> 00:40:32.000
And then what is the dissipated energy required to liquefaction, which would be defined by some curve like this here?

00:40:32.000 --> 00:40:41.000
So what we needed is to compute dissipated energy for a bunch of case histories and define this boundary curve between the liquefaction and liquefaction-action cases.

00:40:41.000 --> 00:40:51.000
So beginning with available cases history databases, we relied heavily on the [indescrible] in address, 2014-2016 database.

00:40:51.000 --> 00:40:59.000
I'll provide a reference at the end, and this is a snapshot of just a little bit of the data that's available in their compiled database.

00:40:59.000 --> 00:41:06.000
And this covers many different earthquakes for a lot of different locations, different magnitudes, maximum accelerations, depths,

00:41:06.000 --> 00:41:20.000
normalize content persistence, finds contents, and vertical stresses. What we did is we went through and updated the Amax values and that the [indiscernible] database using USGS ShakeMap values.

00:41:20.000 --> 00:41:23.000
As long as we can find them.

00:41:23.000 --> 00:41:31.000
And then we went in and computed dissipated energy for each of these case histories.

00:41:31.000 --> 00:41:43.000
Question is, how do we do that? Dissipated energy delta W. Is defined as the cumulative area bounded by the shear stress strain histories as loops as I mentioned, but we could compute that.

00:41:43.000 --> 00:41:51.000
As the area within one equivalent cycle, that's this delta W one equivalent multiplied by the number of equivalent cycles 

00:41:51.000 --> 00:41:57.000
and equivalent M, which is attuned to some events of magnitude M. We rely on the lastly at all 

00:41:57.000 --> 00:42:08.000
2017 relationships. But how do we compute the area of just one history?

00:42:08.000 --> 00:42:16.000
We have this equation here for damping, which is the disssipated energy divided by 4 pi times the stored energy.

00:42:16.000 --> 00:42:25.000
The stored energy is this shaded red triangle, which is equal to 0.5 times the average shear stress multiplied by gamma.

00:42:25.000 --> 00:42:33.000
Gama is the average true stress divided by the shear modulus, which is the slope of this line here.

00:42:33.000 --> 00:42:46.000
We can rewrite this equation and solve for the area within one history loop as 2 pi times damping times the average shear stress squared divided by this year modulus.

00:42:46.000 --> 00:42:52.000
But then how do we compute each of these factors? As before in the stress-based method 

00:42:52.000 --> 00:42:58.000
the average shear stress is 0.65 times the maximum acceleration times total stress times Rd.

00:42:58.000 --> 00:43:08.000
We just use an equation for WUS or CEUS that could be Western or Central Eastern specific by Lasley et al., 2016.

00:43:08.000 --> 00:43:15.000
The two modulus can be computed as the maximum shear modulus times g over g max for some strain level.

00:43:15.000 --> 00:43:23.000
G over Gmax can be obtained iteratively from modules reduction curves that are available in the literature.

00:43:23.000 --> 00:43:32.000
And Gmax can be computed as a function of the shear wave velocity multiplied by the unit weight of soil.

00:43:32.000 --> 00:43:42.000
The [indiscernible] velocity can be estimated in turn as a function of SPT or CPT data if you are using SPT or CPT as you mean in situ metric.

00:43:42.000 --> 00:43:51.000
So on the right you can see a pair of associated modules reduction and damping curves.

00:43:51.000 --> 00:43:56.000
So damping would be selected from an associated damping curve to be compatible with that geology max value.

00:43:56.000 --> 00:44:09.000
So essentially you need the same strain. So ideally you would iteratively come to the correct strain level at which you are able to use the same tau average, the G over G max and D.

00:44:09.000 --> 00:44:13.000
Adding lot altogether. We now have an equation for dissipated energy normalized by initial vertical effective stress, which is our capacity 

00:44:13.000 --> 00:44:21.000
metric in our energy based framework.

00:44:21.000 --> 00:44:29.000
The parameters that are boxed in blue here are things that you would have needed for the stress-based method anyway.

00:44:29.000 --> 00:44:41.000
So nothing new there. The things boxed in red are new but can all be computed. Based on things that are already in the literature and based on information you already gather as part of the stress-based method.

00:44:41.000 --> 00:44:49.000
So essentially very little additional effort is required in order to use the dissipated energy method compared to using the stress-based methods.

00:44:49.000 --> 00:44:56.000
The only thing you have to do is use a couple extra iterative steps in order to compute some of these 

00:44:56.000 --> 00:45:04.000
additional parameters. So now we know how to compute dissipate energy.

00:45:04.000 --> 00:45:12.000
We did that for all of the case histories in our database and plotted the dissipated energy versus normalized concept resistance 

00:45:12.000 --> 00:45:22.000
and using maximum likelihood as equations to estimate these blue lines here. Which are our probabilistic median 

00:45:22.000 --> 00:45:36.000
curves. We also came up with this green deterministic curve here. The functional form you can see here as well as the coefficients needed to compute the capacity in terms of dissipated energy.

00:45:36.000 --> 00:45:44.000
And I will provide you with a link at the very end in order for you to take a little more time understanding what we mean by understanding what we mean by uncertainties excluded and included 

00:45:44.000 --> 00:45:58.000
in our paper. The next thing we did was we performed laboratory tests. In order to make that very crucial link between how to compute dissipated energy in the field and in the lab and do it consistently.

00:45:58.000 --> 00:46:06.000
I'll talk about what I mean by that. So we just talked about how to compute dissipated energy in the field using a total stress framework.

00:46:06.000 --> 00:46:13.000
Where you have one history since loop that's multiplied by number of cycles. In the lab,

00:46:13.000 --> 00:46:20.000
as the soil softens, those histories loops start to increase. Right, they start to get bigger.

00:46:20.000 --> 00:46:25.000
You can't control or we're not able to control the soil and make sure that it doesn't start to behave that way.

00:46:25.000 --> 00:46:31.000
When we're witnessing liquefaction begin to occur, right. The excess pore pressure ramps up.

00:46:31.000 --> 00:46:35.000
The soil softens and this is what we get, but we want to be able to translate our laboratory data into a total stress 

00:46:35.000 --> 00:46:51.000
framework that we've been using in the field. So how do we do that? We plot dissipated energy versus number of cycles of loading in a lab specimen.

00:46:51.000 --> 00:46:59.000
And initially, as you can see the histories of slips start out small and basically repeat the same size multiple times.

00:46:59.000 --> 00:47:08.000
So you have this linear accumulation of dissipated energy. For a while until those histories loops start to expand.

00:47:08.000 --> 00:47:14.000
And then you start to see this exponential increase in dissipated energy. Total accumulation of that.

00:47:14.000 --> 00:47:30.000
With each cycle. As opposed to following this linear trend. So what we propose is taking this initial linear phase and extrapolate that out until you reach a liquefaction and here we're defining it in terms of the Ru 

00:47:30.000 --> 00:47:44.000
equaling one. So once you've reached that excess poor pressure ratio at that point on the green total stress dissipated energy line that is the total dissipated energy required to trigger liquefaction 

00:47:44.000 --> 00:47:52.000
for this particular soil specimen. So this is the key to consistent interpretation between the field and the lab is to use total dissipated energy.

00:47:52.000 --> 00:47:58.000
I briefly wanted to make a connection to Loma Prieta in 1989. Here we have these liquefaction case histories.

00:47:58.000 --> 00:48:11.000
If you're familiar with the next generation, liquefaction project. You'll know our website has data about all sorts of cases histories including these particular 1989 event case histories.

00:48:11.000 --> 00:48:19.000
You can take all of those cases histories and plot it either in terms of dissipated energy or CSR star versus QC1NCS.

00:48:19.000 --> 00:48:27.000
And as you can see the 1989 Loma Prieta event gives us quite a few of these white liquefaction case histories above.

00:48:27.000 --> 00:48:41.000
Our pass curve in both stress based and energy based methods. And so these are some very important case histories that we need to look at and try to understand. Why is it that we're predicting liquefaction for these cases?

00:48:41.000 --> 00:48:59.000
But no liquefaction was actually observed. I'll conclude with a few things. So estimates of liquefaction demand and so two can be computed in a simplified total stress energy based framework using parameters that are commonly required for stress-based procedures so you don't have to go looking for extra information.

00:48:59.000 --> 00:49:07.000
Our limit state curves that we developed can we use testimony the fact of safety against the liquefaction or the probability of liquefaction.

00:49:07.000 --> 00:49:16.000
And laboratory tests can be interpreted within a total stress energy based framework consistent with the field based estimates of dissipated energy.

00:49:16.000 --> 00:49:23.000
So I'd like to thank you for your attention and your time and look forward to answering some of your questions.

00:49:23.000 --> 00:49:37.000
Thanks Kristin. Our next talk will be Davis Engler, presenting the Bayesian framework, or modeling the liquefaction probabilities.


















00:49:37.000 --> 00:49:48.000
Hi, my name is Davis Engler. I'm going to discuss a Bayesian framework for incorporating surficial geology and CPT records within the USGS ground failure liquefaction product.

00:49:48.000 --> 00:50:03.000
This is work that means several colleagues at the Geologic Hazard Science Center in Golden, Colorado have been doing in collaboration with Brett Maurer out of University of Washington and Mertcan Geyin at NGI.

00:50:03.000 --> 00:50:09.000
A quick outline. We'll start with the motivation and spend a little time on background and existing models.

00:50:09.000 --> 00:50:23.000
Then I'll introduce the framework and discuss the strategies for implementing this stuff. And then finally present our results for the Loma Prieta earthquake and compare them with the geospatial model.

00:50:23.000 --> 00:50:31.000
There's been lots of great work on modeling liquefaction, triggering and surface manifestations and mapping, artificial geology of the years and our contribution 

00:50:31.000 --> 00:50:40.000
in developing a mathematical framework that can bring these different efforts together and can junction with the geospatial models.

00:50:40.000 --> 00:50:48.000
So we integrate all this information through a Bayesian updating framework. There are three components to the Bayesian equation.

00:50:48.000 --> 00:50:55.000
The prior model, the likelihood, and the posterior, which will kind of keep these color schemes throughout.

00:50:55.000 --> 00:51:09.000
We treat the geospatial model as the prior because it's widely applicable and based on the simple factors The likelihood function is what we'll use to update the prior locally using geotechnical and geological data.

00:51:09.000 --> 00:51:16.000
So this animation below is an illustration of how the prior and likelihood act to form the posterior,

00:51:16.000 --> 00:51:25.000
shown in red. The uncertainty of the prior and likelihood to change. This is a plot of the probability density function, the prior and blue, likelihood in black.

00:51:25.000 --> 00:51:37.000
And again, the posterior showed in red. Note that when the likelihood function is more precise, indicated by a narrower distribution, The posterior approaches the likelihood.

00:51:37.000 --> 00:51:47.000
And when it's more uncertain. Indicated by wider distribution, the posterior approaches to prior.

00:51:47.000 --> 00:51:53.000
So let's review the existing models and data that we're going to be integrating in.

00:51:53.000 --> 00:52:06.000
Basically our approach boils down to an extension of the Holzer approach. In this approach first, the exceedance curves are computed for a given geologic unit as a function of LPI.

00:52:06.000 --> 00:52:12.000
These curves are then converted to the probability of liquefaction occurrence as a function of PGA 

00:52:12.000 --> 00:52:17.000
given a relationship observed during the Loma Prieta earthquake.

00:52:17.000 --> 00:52:29.000
In the last plot, the PGA values are normalized by the magnitude scaling factor.

00:52:29.000 --> 00:52:35.000
One of the ways that we're extending upon this approach is through using alternative surface geology maps.

00:52:35.000 --> 00:52:46.000
This includes using the Witter et al. or recently extended Wentworth et al. which is just for the San Francisco Bay Area as well as the Wills et al. map, which is for the entire state of California.

00:52:46.000 --> 00:52:56.000
Of course the Witter map is more detailed spatially and in terms of the unit descriptions, but the Will's map has greater spatial coverage.

00:52:56.000 --> 00:53:06.000
One thing we found with both mappings is that many of the different geologic units have little or no CPT data to get distributions from.

00:53:06.000 --> 00:53:08.000
So it was useful to group them up luckily Witter et al. had already grouped the units into susceptibility classes 

00:53:08.000 --> 00:53:20.000
and these classes account for the age as well as the depositional environment of the different mapped units.

00:53:20.000 --> 00:53:32.000
An important component in the framework is the liquefaction triggering models. These models take the output of the CPT records and given a magnitude PGA and water table 

00:53:32.000 --> 00:53:42.000
calculated the LPI. The models shown here and used here in is the Boulanger and Idriss (2014) model.

00:53:42.000 --> 00:53:56.000
The CPT records we're using are the same as used in the Holzer approach and here they're shown along with the liquefaction susceptibility classes for the Witter and Wentworth mapping on the left as well as the Wills mapping on the right.

00:53:56.000 --> 00:54:11.000
This figure also allows us to briefly see some of the differences and similarities between the liquefaction. Between each liquefaction susceptibility class between these two different mapping schemes.

00:54:11.000 --> 00:54:21.000
One important recent development that we're taking advantage of is a global model for the probability of surface manifestation as a function of LPI.

00:54:21.000 --> 00:54:33.000
LPI is plotted on the horizontal access here and the probability of exceedance curves they gave is along the vertical axis.

00:54:33.000 --> 00:54:41.000
They also provided curves of different severities and these curves were developed from a global data set of service manifestations.

00:54:41.000 --> 00:54:47.000
Hence my mention of global.

00:54:47.000 --> 00:54:56.000
Perhaps the most important contributors in estimating liquefaction probability are the depth of the water table here shown for the Fan et al. model.

00:54:56.000 --> 00:55:03.000
And the PGA estimated from Shake Map for Loma Prieta shown on the right.

00:55:03.000 --> 00:55:12.000
And finally we have the prior model, the geospatial model. This comes from Zhu et al., 2017, and it gives the post-earthquake liquefaction probability anywhere in the world.

00:55:12.000 --> 00:55:23.000
And it's derived from a large set of global earthquakes from which liquefaction inventories are available.

00:55:23.000 --> 00:55:29.000
So I want to take a minute and summarize the ingredients so we can see how everything fits into the Bayesian framework.

00:55:29.000 --> 00:55:36.000
As mentioned, the geospatial model goes into the prior and everything else we discussed goes in. To come up with the likelihood.

00:55:36.000 --> 00:55:45.000
Which will call the LPI based model. Again, we're taking two estimates of liquefaction probability and trying to reconcile them probabilistically.

00:55:45.000 --> 00:55:53.000
As we go through this presentation, we'll see how these different pieces fit into the framework.

00:55:53.000 --> 00:56:06.000
So now we're going to look at a method. To model post-earthquake liquefaction probability given these CBT records maps and models.

00:56:06.000 --> 00:56:15.000
We'll start with a road wrap from the modeling framework that I'm presenting. The first step is to model the distribution of LPI for the different susceptibility classes.

00:56:15.000 --> 00:56:20.000
And for that, these are the ingredients that we make use of.

00:56:20.000 --> 00:56:26.000
The second step is to model LPI as a function of magnitude, water table, and PGA.

00:56:26.000 --> 00:56:32.000
And it makes use of the same ingredients to step one.

00:56:32.000 --> 00:56:42.000
Third step is to model the LPI distribution for a given earthquake so that we need to add in the shaking estimates from ShakeMap and the water table estimates as well.

00:56:42.000 --> 00:56:49.000
Four-step is to compare the likelihood and prior models. So we need to bring in the geospatial model results into our analysis.

00:56:49.000 --> 00:57:02.000
This is also where we convert into our LPI based probability model. And fifth, we describe how we combine the likelihood and prior models to get our posterior.

00:57:02.000 --> 00:57:09.000
So what this shows is a bunch of violin plots. And the point of this is to show The distribution of LPI 

00:57:09.000 --> 00:57:21.000
for different geologic units where the LPI. It's calculated from this basic scenario of magnitude PGA and the number of CPTs is given in parentheses.

00:57:21.000 --> 00:57:32.000
Of the colors on here are given by the liquefaction susceptibility classes. And this just kind of allows us to see how the different classes have similar distributions

00:57:32.000 --> 00:57:39.000
for the units that are inside of them. We use similar susceptibility grouping for the Wills et al. units 

00:57:39.000 --> 00:57:51.000
and develop this to correspond with the Witter groupings in that similar susceptibility classes should have a similar distributions of LPI.

00:57:51.000 --> 00:57:58.000
So the idea is that given an LSC like the "High" Wills class, we could take all of the CPT records from the units within it 

00:57:58.000 --> 00:58:07.000
and given a magnitude PGA and water table depth calculate LPI for each record. This clip shows the histogram and approximate density 

00:58:07.000 --> 00:58:18.000
of the log of LPI as PGA increases. One thing we found is that for many scenarios of MPGA and GWT, a portion of the LPI data is 0.

00:58:18.000 --> 00:58:22.000
And so we show with this ratio on the left.

00:58:22.000 --> 00:58:35.000
As we play the video the PGA increases and then occasionally the groundwater table decreases and we can see that the distribution of LPI increase and conversely the probability of LPI being 0 decrease.

00:58:35.000 --> 00:58:46.000
The fitted distributions are either log normal or gamma, which everyone had a higher likelihood of generating the LPI data for a given scenario.

00:58:46.000 --> 00:58:55.000
Repeating this for all the susceptibility classes we tabulated the mean LPI and the empirical probability of LPI being 0.

00:58:55.000 --> 00:59:04.000
As a function of PGA shown on the horizontal access. As PGA increases, expect to see that the mean LPI should increase.

00:59:04.000 --> 00:59:14.000
The probability of LPI should being 0 should decrease. And we also expect to see that the trends between the different classes should be very distinct.

00:59:14.000 --> 00:59:21.000
But however, we wanna see that, there's consistency for the different mappings like the dotted and dashed lines.

00:59:21.000 --> 00:59:29.000
So using these and other statistical moments of LPI allows us to approximate LPI is either a log normal or gamma distribution.

00:59:29.000 --> 00:59:45.000
By fitting it to these moments. Now, we'll note here that these very low and low curves were not empirically derived, but we assumed distributional parameters based on judgment due to the lack of data for them.

00:59:45.000 --> 00:59:50.000
Okay, so basically what we have is we have a bunch of locations where we want to get LPI.

00:59:50.000 --> 01:00:06.000
Based off of magnitude the PGA at the location and the water table. So basically we can interpolate the statistical moments of LPI between at each location between nearby MPGA and GWT scenarios.

01:00:06.000 --> 01:00:20.000
And then for the distribution, either log normal gamma, to these interpolated moments. Following that, we can plug these LPIs in to this manifestation model, take a probability of liquefaction.

01:00:20.000 --> 01:00:30.000
And at this point we can stop and take a look and compare these two models the one that we just made the LPI based and then the geospatial model shown here for Loma Prieta.

01:00:30.000 --> 01:00:36.000
One large difference is that the LPI based model predicts overall much higher probabilities than the geospatial model.

01:00:36.000 --> 01:00:41.000
Although not everywhere.

01:00:41.000 --> 01:00:55.000
We also looked at several different performance measures to evaluate and compare these if these different models. Whereas the ROC curve and AUC measure relative performance between the positive and negative inventories.

01:00:55.000 --> 01:01:04.000
The Brier score is an absolute performance measure of the predictive probabilities themselves. So we'll look at both of these.

01:01:04.000 --> 01:01:16.000
So we'll start with the ROC curve. Each point along the curve corresponds to a probability threshold between 0 and one, which acts as the dividing line between no liquefaction and yes, liquefaction.

01:01:16.000 --> 01:01:26.000
For a given threshold, we plot the true positive rate on the vertical axis. Which is just the fraction of all the locations where liquefaction actually occurred.

01:01:26.000 --> 01:01:37.000
Which were correctly classified at this threshold. And we plotted against the false positive rate on the horizontal axis, which is the fraction of all the negative observations which were incorrectly classified as positive.

01:01:37.000 --> 01:01:45.000
The ROC is essentially tells us how we can successfully classify the positive and negative observations at this threshold.

01:01:45.000 --> 01:01:51.000
And the curve is formed by looping through multiple thresholds between 0 and 1. With lower thresholds plotted on the upper right.

01:01:51.000 --> 01:01:57.000
Perfect model would plot along the axis at the upper left and a useless model is on the 1 to 1 line.

01:01:57.000 --> 01:02:09.000
The area underneath the ROC curve is also often used measure called AUC. As it gives the probability of correctly classifying any random pair of positive and negative observations when given one of each.

01:02:09.000 --> 01:02:15.000
Perfect model would have an AUC of 1. So here the LPI actually goes a little bit higher.

01:02:15.000 --> 01:02:25.000
The prior score is a measure of how close the predicted probabilities are with the actual probabilities. Which are either 0 or 1 depending on whether or not liquefaction occurs.

01:02:25.000 --> 01:02:31.000
Prior score computer from all observations is shown on the right, but where we see they do about the same.

01:02:31.000 --> 01:02:39.000
And we also show the score for the only the positive and we can do it for only the negative where we see that the different models actually outperform in different regions.

01:02:39.000 --> 01:02:45.000
Now a lower prior score being better, I should mention.

01:02:45.000 --> 01:02:56.000
So unfortunately we don't have good constraints on the prior models uncertainty and the relative uncertainties between the prior likelihood really determines how well you're going to mix between these 

01:02:56.000 --> 01:03:05.000
different, probability thresholds. So the geospatial model performs well in some circumstances like at the higher threshold area.

01:03:05.000 --> 01:03:10.000
The LPI Holzer model performs better at the lower threshold levels down in here as well as at the very high threshold.

01:03:10.000 --> 01:03:19.000
So ideally we would have a model that kind of takes the better features from both of these.

01:03:19.000 --> 01:03:30.000
So we went looking for that model. We looked at a few different methods for modulating the prior geospatial uncertainties and found one which maximized that kind of ROC performance.

01:03:30.000 --> 01:03:37.000
And this is what that posterior model looks like on the right. We can see that it has a nice blend of the two,

01:03:37.000 --> 01:03:43.000
but how do its overall predictions at the inventory stack up?

01:03:43.000 --> 01:04:04.000
Well, let's start again with the ROC on the left. We have the likelihood LPI model shown in the black and again the geospatial in blue and posterior in red. What this shows is that the posterior model is generally a really good mixture of whichever model is out performing in the different threshold levels and in some places it's actually higher than both.

01:04:04.000 --> 01:04:15.000
The AUC again is higher, but only moderately higher. But does anyone who's done this knows it's kind of a difficult to really increase AUC especially when all you're working with is things that aren't perfect.

01:04:15.000 --> 01:04:23.000
But in terms of the Brier score we broke it down into positive versus negative because each model has different downsides for these different cases.

01:04:23.000 --> 01:04:32.000
It's always a trade-off between doing well and the positive and negative. And we can see that the posterior was actually able to do better in the positive without doing worse in the negative, which is really important.

01:04:32.000 --> 01:04:41.000
I'm about ready to wrap it up. So I think I'll just kind of leave it on this summary slide and yeah, thank you all for watching.

01:04:41.000 --> 01:04:49.000
Alright, so I think that is it for our presentations and yeah, thank you Davis.

01:04:49.000 --> 01:04:53.000
Shall we just open up the

01:04:53.000 --> 01:04:54.000
questions. Yeah.

01:04:54.000 --> 01:05:02.000
Open up the phone lines for all the questions. I wanted to kick things off and Tom 

01:05:02.000 --> 01:05:03.000
Are you still here with us?

01:05:03.000 --> 01:05:09.000
Yeah, Tom's here. Tom, I always found the age dating that you all did with your subsurface investigations particularly informative.

01:05:09.000 --> 01:05:21.000
And I was just wondering if you wanted to comment on the impetus for that and thoughts on future using age dating along with subsurface investigations.

01:05:21.000 --> 01:05:34.000
Yeah, well the important observation here is as far as I know there have been no place to scene age deposits liquefied in California.

01:05:34.000 --> 01:05:42.000
Yeah, that stands in contrast, what's happening back east in the, particularly South Carolina.

01:05:42.000 --> 01:06:00.000
Thing and the difference between those 2 areas has never been explained but I've taken it as a fact that the Pleistocene age deposits just don't liquefy here and it may be because they get shaken so much by earthquakes over 10,000 years.

01:06:00.000 --> 01:06:01.000
Okay.

01:06:01.000 --> 01:06:07.000
That, You know, it's just driven out of them. So it was always important for us to recognize where the Pleistocene was.

01:06:07.000 --> 01:06:13.000
And a good artificial geologist, can use a lot of clues. You don't actually, we need age dating many other times.

01:06:13.000 --> 01:06:19.000
For example, the Merit Sand is a good example of that. So.

01:06:19.000 --> 01:06:38.000
Good. And then I have one other comment which was. Early on you showed a slide with Professor Ishihara and one of my favorite memories working with Ishihara was shopping for sledgehammers in Tokyo with him because we're doing some two-hour was shopping for sledgehammers in Tokyo with him because we were doing some two-phase hammers in Tokyo with him because

01:06:38.000 --> 01:06:40.000
we were doing some 2 physical work. Do you have any good memories that you want to share with him because we're doing some 2 physical work.

01:06:40.000 --> 01:06:42.000
Do you have any good memories that you want to share, with some of the other luminaries in our field?

01:06:42.000 --> 01:06:57.000
Oh boy, there were so many of them. I remember John Barrel showing up at my office and volunteering to, to help, One of the early energy guys.

01:06:57.000 --> 01:06:58.000
Yeah.

01:06:58.000 --> 01:07:08.000
Maybe the first, I guess. Likequo Faction. I remember working in a drug needle infested lot with Tom O'rourke.

01:07:08.000 --> 01:07:19.000
We had to be careful we didn't. Catch some weird disease. So, yeah, I mean, they were just a bunch of experiences because they just seemed like everybody came out to capture this earthquake.

01:07:19.000 --> 01:07:32.000
Nobody's mentioned the good year blimp that photographed the. Lower per year earthquake it just gave it a notoriety that no other earthquake has I think.

01:07:32.000 --> 01:07:40.000
Of the blimp. Yeah, nice. Thanks, Tom. Looks like we've got a hand up in the Yosemite room.

01:07:40.000 --> 01:07:45.000
Yeah, so I have a question and, I can't see the chat, so maybe this has already been talked about.

01:07:45.000 --> 01:07:49.000
But this pertains to pretty much all the talks and I'm wondering how Most of the discussions of CSR, it's been a great session by the way.

01:07:49.000 --> 01:08:04.000
Really interesting talks. Most of the discussion of CSR and the energy method as an alternative. Talk about stress and strain response but without reference to frequency of excitation.

01:08:04.000 --> 01:08:10.000
So my question is And when you're talking about drain versus under in response, of course that depends on the permeability structure, whether the fluid pressure can be relieved.

01:08:10.000 --> 01:08:22.000
During cyclic loading or not, that depends on frequency. So my question is basically, is everybody, how do you include frequency dependence in?

01:08:22.000 --> 01:08:30.000
Characterize them and transition from drained to undrained behavior as it would affect cyclic. Hardening or compaction softening.

01:08:30.000 --> 01:08:36.000
So this pertains to. Classic stress and strain based method. I mean stress based methods and the energy based method as well.

01:08:36.000 --> 01:08:42.000
So how is sick look loading? Impact and fluid response, which of course is critical of looking at.

01:08:42.000 --> 01:08:45.000
The perfection potential.

01:08:45.000 --> 01:08:53.000
Woof, that's a heavy question. I'm gonna put that one to Jorge and Kristen and see if you guys want to throw your 2 cents in on that.

01:08:53.000 --> 01:09:07.000
Surely I can throw some, thoughts about that. And I would say that the way that we do, but please help particularly to that is, you know, to numerical modeling.

01:09:07.000 --> 01:09:20.000
These days you have considered models where the are used for liquefaction purposes. So you can include the frequency content of the design your own motion as a variable of the modeling.

01:09:20.000 --> 01:09:28.000
So you can have a suite of models with different frequency contents and then you can assess you know what's the influence of that frequency content.

01:09:28.000 --> 01:09:36.000
And And also, you know, when you are having a medical model, you can really have, I would say, couple conditions.

01:09:36.000 --> 01:09:44.000
Meaning that this is neither drain or and drain. Is going to be potentially partially trained, so permeability would be also a factor.

01:09:44.000 --> 01:09:54.000
So quite complex phenomena, so that's that's that's why I personally think that probably a numerical modeling may be a good way to to look at it.

01:09:54.000 --> 01:10:01.000
And, and ultimately the influence I think cannot be, you know, there is no, I don't think there is a general statement personally.

01:10:01.000 --> 01:10:11.000
Because the influence of the frequency is going to be dependent also on what is the period and other characteristics of the system that you are evaluating and how that is changing.

01:10:11.000 --> 01:10:15.000
So those are my 2 cents on the question.

01:10:15.000 --> 01:10:16.000
Thank you. Anyone else?

01:10:16.000 --> 01:10:31.000
Yeah, so I was just gonna add that if you're looking at simplified methods, you know, which is basically what I was looking at, you won't see much discussion on frequency specifically or any direct, you know, accounting for frequency just because that's a that's a tough one.

01:10:31.000 --> 01:10:40.000
But you can definitely look through literature and see based on laboratory tests, you know, to see the effect of frequency and whether or not it's actually influencing CSR for certain soil types.

01:10:40.000 --> 01:10:46.000
Whether it's sands or maybe, so is contain more finds. So that's what I would look into.

01:10:46.000 --> 01:10:47.000
Yeah, okay, thanks.

01:10:47.000 --> 01:10:58.000
Yeah. Like to add that although frequency isn't directly considered the number of cycles is. And that's a very crude estimate too.

01:10:58.000 --> 01:11:15.000
I mean, it's just an average number of cycles for the thing. And then the other problem you have is as the soil softens the frequency of the shaking that's able to load the soil changes very long period.

01:11:15.000 --> 01:11:21.000
Motion your surface waves are going to be sharing the soil at the end of the process.

01:11:21.000 --> 01:11:22.000
Yeah. And.

01:11:22.000 --> 01:11:29.000
Yeah, there's a cumulative damage. I'm thinking in terms of material science view on this, cumulative damage effects as well as silicon in terms of material science view on this, cumulative damage effects as well as sigma floating effects.

01:11:29.000 --> 01:11:42.000
And this might be the kind of place for inCTu poor pressure monitoring and a place that has susceptibility would be able to basically reveal a transfer function between stress and strain and for the pressure response as a function of load cycling.

01:11:42.000 --> 01:11:45.000
And as a function of frequency. The interesting thing to try. But doing the forward modeling, I seems like that's the best way.

01:11:45.000 --> 01:11:56.000
But then you have to know about. And see to permeability and all sorts of other stuff and I guess into other components and the energy balance.

01:11:56.000 --> 01:11:57.000
Yeah.

01:11:57.000 --> 01:12:05.000
So Come, get a problem, maybe it doesn't have much of an impact. But it's something that I worry about thinking about sharing a brand new

01:12:05.000 --> 01:12:06.000
Okay.

01:12:06.000 --> 01:12:11.000
It's a good argument to keep funding wildlife sites and other downhill sites like that.

01:12:11.000 --> 01:12:12.000
Okay.

01:12:12.000 --> 01:12:13.000
Yes, I think so. Thank you.

01:12:13.000 --> 01:12:27.000
Robbie has very quick comment. Just wanted to add that on the lab, you know, at least some fine rained it's always when you change the frequency you can change you can see a change in the response as well So yes, just wanted to add that comment.

01:12:27.000 --> 01:12:28.000
Thank you.

01:12:28.000 --> 01:12:33.000
Thank you.

01:12:33.000 --> 01:12:42.000
Other questions from the crowd?

01:12:42.000 --> 01:12:55.000
I had one for Davis. As they put in the chat, I thought that the graphics and the figures that you had were very, informative and demonstrative and I I want to commend you on that.

01:12:55.000 --> 01:13:05.000
I myself have been using the browser score. Over ROC numbers just because it gives more of an absolute and your thoughts on that.

01:13:05.000 --> 01:13:24.000
As we see in the literature lots of area under the curve numbers but sometimes that doesn't give you the bigger picture and I I see Eric smiling too so let's have a little side discussion on Briar score when we're talking about comparing models.

01:13:24.000 --> 01:13:39.000
Yeah, yeah, absolutely. I, Sorry, I really agree. I think that was just kind of an interesting demonstration for Looking at ROC and AUC when you're comparing those 2 different models that have, you know, completely different absolute, probabilities.

01:13:39.000 --> 01:13:44.000
I mean, It goes to, you know, factor of 2 to 4, even higher than that.

01:13:44.000 --> 01:13:51.000
So, you know, while it is important to be able to, you know, correctly distinguish your positive from your negatives, I think.

01:13:51.000 --> 01:13:57.000
Yeah, the Bryer score definitely has a A little bit more to say. In terms of what's actually happening.

01:13:57.000 --> 01:13:59.000
Yeah.

01:13:59.000 --> 01:14:11.000
Yeah, I think that one of the key points is what Rob alluded to is that. Area under the curve is really only a evaluation in a sense.

01:14:11.000 --> 01:14:17.000
It doesn't tell you anything about the absolute values. So,

01:14:17.000 --> 01:14:27.000
Well, a while the RRC curve is informative. It's nice to pair with something like prior scored it to make sure you've got the whole picture.

01:14:27.000 --> 01:14:39.000
Yeah, Rob, one question, one follow-up question for David. You know, when you do ROC assessments, it's kind of your getting, I would say an average.

01:14:39.000 --> 01:14:50.000
It's, it's a classification metal so you get an average view. You know, your data.

01:14:50.000 --> 01:14:51.000
Okay.

01:14:51.000 --> 01:14:55.000
But, in our experience, sometimes, you know, a, so many scenarios with higher areas under the core values.

01:14:55.000 --> 01:15:02.000
Good to still have some, you know, if you look at some specific subsets, you can see some physical.

01:15:02.000 --> 01:15:04.000
You know, outliers to say it in a way, for example, in cases of liquefaction in New Zealand.

01:15:04.000 --> 01:15:17.000
And not sure if you guys are also considering to. To put any additional constraints to bring a little bit more physics into the overall assessment.

01:15:17.000 --> 01:15:22.000
I mean, as curious if you are thinking of that as well.

01:15:22.000 --> 01:15:30.000
I'm not entirely sure I haven't, gone down that road. I think there's, multiple areas where we really want to.

01:15:30.000 --> 01:15:44.000
Don't look out. I know that's a Yeah, that's definitely something to look into more in terms of like how do we actually deal with these inventories and how do we deal with like the balancing between them and you know it's

01:15:44.000 --> 01:15:45.000
So yes.

01:15:45.000 --> 01:15:51.000
Well, in terms of physics, I think we're relying on the empirical, triggering model.

01:15:51.000 --> 01:15:59.000
But I think the framework that Davis has developed, you know, one could go in and substitute, you know, like a something that's a little bit more of a constitutive approach.

01:15:59.000 --> 01:16:11.000
Modeling, local faction and, do the same kind of thing. But I think for our purposes at least, we felt like the.

01:16:11.000 --> 01:16:20.000
That, interest. Current model made sense, but I think you could potentially substitute in anything else there as well.

01:16:20.000 --> 01:16:21.000
So.

01:16:21.000 --> 01:16:37.000
Yeah, and, Eric gets very quickly, you know, just to give an example when, when we were using some inversion procedures in New Zealand for a estimating a state perimeter, there were some methods that were giving you a very large a UC value.

01:16:37.000 --> 01:16:46.000
But then when you were going to specific sites, the states that you were in Bertin where he has to, I would say, were not.

01:16:46.000 --> 01:16:57.000
A reasonable from a physical context. So however, that you see was higher statistically. So.

01:16:57.000 --> 01:16:59.000
I think you have something to add.

01:16:59.000 --> 01:17:10.000
Oh, I sorry, I just wanted to change the subject a little bit. We've got a chat from Anne at USGS and she said, what can we say about liquefaction during an earthquake sequence?

01:17:10.000 --> 01:17:17.000
When will it repeat? Awesome question. Do our panelists want to take a stab at that one?

01:17:17.000 --> 01:17:24.000
Well, I'll take a initial shot. I mean, clearly, Liquor Faction can repeat.

01:17:24.000 --> 01:17:35.000
We've seen sites the Wildlife site being the Probably one of the classic examples where we instrumented in it because we saw likequifax and then we got a Liquifaction again.

01:17:35.000 --> 01:17:41.000
But one of the intriguing things of it after Loma Creative, there were a lot of sites.

01:17:41.000 --> 01:17:53.000
That liquefied in both 1906 and But it wasn't the same deposit. Floods on the river had flushed out what had liquefied in 1,906 and replaced it with younger deposits.

01:17:53.000 --> 01:17:57.000
So you have to be a little careful with these sites.

01:17:57.000 --> 01:18:04.000
We also have, with larger layers, this phenomena avoid ratio redistribution. And so the soil liquefy.

01:18:04.000 --> 01:18:16.000
And it will tend to densify at the bottom of the layer, but then loosened at the top of the layer and so The net effect is that we can see, same unit liquor fine again and again.

01:18:16.000 --> 01:18:19.000
When I was sort of why I thought Jorge's plots with the CPT data before and after the earthquake was so interesting.

01:18:19.000 --> 01:18:28.000
And it wasn't clear what the overall trend was if it was making things go up or down.

01:18:28.000 --> 01:18:36.000
Is anybody aware of more data like that? I actually haven't seen previously any like pre and post earthquake.

01:18:36.000 --> 01:18:38.000
CPT data.

01:18:38.000 --> 01:18:39.000
Yep.

01:18:39.000 --> 01:18:48.000
There is some data from New Zealand, from the Contemporary of Queens. But there are also some styles that are looking at that more at the fundamental level, you know.

01:18:48.000 --> 01:18:56.000
Is I think 2 factors are how fabric is changing. And also what is intensity that you're applying to the medium.

01:18:56.000 --> 01:19:08.000
There are some DNM studies that are say us in that it could the trains could be going either way depending on how far it changes and how is intensity that it's coupled with with the fabric change.

01:19:08.000 --> 01:19:24.000
Eric, we have that data for wildlife. We did around the original CPTs, 3 CPTs, were only a meter away from the original.

01:19:24.000 --> 01:19:25.000
Okay.

01:19:25.000 --> 01:19:30.000
And the natural variability just tended to overwhelm the whatever changes they were. There might have been some slight increase in the friction ratio.

01:19:30.000 --> 01:19:35.000
But we couldn't see any change in the tip significant anyway.

01:19:35.000 --> 01:19:45.000
Right, yeah. Oh, go ahead, Rob.

01:19:45.000 --> 01:19:46.000
That's right.

01:19:46.000 --> 01:19:48.000
So, I was gonna there was a study from Loma Prieta as well. I think the author was Shimo that had pre and post CPT investigations that a particular as well.

01:19:48.000 --> 01:19:53.000
I was just gonna say I think it's really interesting because if you just take a very simplistic intuitive.

01:19:53.000 --> 01:20:05.000
Perspective on this like you'd think the process of liquefaction is densifying your material and so you should be you know, lowering your susceptibility for this for any subsequent.

01:20:05.000 --> 01:20:13.000
Earthquake that occurs, but that just doesn't seem to be what actually happens. With real soil.

01:20:13.000 --> 01:20:14.000
So.

01:20:14.000 --> 01:20:15.000
Yeah.

01:20:15.000 --> 01:20:20.000
Yeah, but Eric, you may wanna think also that the soil will need to be deposited again in a very gentle manner.

01:20:20.000 --> 01:20:28.000
So you are kind of erasing the fabric to save them away. And creating a new fabric that is going to be the positive that are and there are very loose structure.

01:20:28.000 --> 01:20:31.000
So. But that that's also is quite complex. Yep.

01:20:31.000 --> 01:20:36.000
Oh yeah, exactly. It's much more complex than our, than a simple intuition.

01:20:36.000 --> 01:20:44.000
Yeah, Eric, I always assumed what it was going on was the the grains weren't losing grain to grain contact.

01:20:44.000 --> 01:20:51.000
Certainly part of the soil is what Rob was referring to would, would probably lose, remove the grain to green contact.

01:20:51.000 --> 01:20:59.000
But if it doesn't lose the grain to green contact, then its density probably isn't going to change too much when the, poor pressure equilibrates again.

01:20:59.000 --> 01:21:00.000
Right.

01:21:00.000 --> 01:21:07.000
And did you, did you wanna follow up on that or you feel like that was, we kicked it around enough?

01:21:07.000 --> 01:21:11.000
Yes, you kicked it around. Thank you.

01:21:11.000 --> 01:21:15.000
Good.

01:21:15.000 --> 01:21:26.000
Let's see. Oh, Jorge, I had a question for you. Given what you learned from that site, did you have any recommendations as the project moved forward?

01:21:26.000 --> 01:21:37.000
Was this in a Was this in this project was in the process of building or were you just analyzing the data post project?

01:21:37.000 --> 01:21:48.000
My scope was purely, you know, scientific driven, so we, we didn't. Provide much input to any project development.

01:21:48.000 --> 01:21:58.000
So if you were asked, to put on your engineering hat. How would you advise the client to proceed?

01:21:58.000 --> 01:22:05.000
Given that you included the critical state aspects of the soil. With the LICKA Faction probability have increased, decreased.

01:22:05.000 --> 01:22:09.000
Changed differently around the site.

01:22:09.000 --> 01:22:28.000
I would say the main thing that on this particular case, the, Great to go, and mechanics is bringing to the table is the fact that was if not all simplified procedures are kind of having formulated for sands with relatively low fine contents.

01:22:28.000 --> 01:22:42.000
So, evaluating, for example, the response of the Yom Beymat, I didn't have time to talk much about the Y, but it's kind of a sealed in material even though it has a name of a play, it behaves more like a play, it behaves more like a seal.

01:22:42.000 --> 01:22:45.000
So you can bring into the table, it behaves more like a seal. So you can bring into the table, you know, how the response of that material is going to be influencing on the system.

01:22:45.000 --> 01:22:54.000
Response of the overall profile. So I think that's what the that's the main thing that.

01:22:54.000 --> 01:23:03.000
Ready to go to the table on that context. It's giving you, you know, tools to get additional insights.

01:23:03.000 --> 01:23:09.000
On how, you know, like in that particular case, find are going to be contributing to the overall response.

01:23:09.000 --> 01:23:31.000
We haven't compared that with the traditional procedures to be able to comment on that particular case. But, the other thing I can say is that I I come a little bit from a background on the on the mining industry where you deal with the materials that are having very fine very high- contents.

01:23:31.000 --> 01:23:37.000
And critical estate was kind of the prevalent, the prevalent, I would say, metalli in those cases.

01:23:37.000 --> 01:23:41.000
So. And I think that's what I can comment.

01:23:41.000 --> 01:23:42.000
Well, Rob has to leave us now. So I'll just say thanks, Rob, for helping out and facilitating the discussion.

01:23:42.000 --> 01:23:43.000
Thanks, Warren.

01:23:43.000 --> 01:23:56.000
I see the hand is up in the assembly room. Do we wanna go there for a question?

01:23:56.000 --> 01:24:01.000
Yeah, hi, this is this is Steve Hayman. I told I have to tell you all my name.

01:24:01.000 --> 01:24:02.000
Yeah.

01:24:02.000 --> 01:24:11.000
I had a question for Kristen and I really like your energy approach to characterizing the perfection in terms of histories, hysteresis loops and as I understand them.

01:24:11.000 --> 01:24:23.000
They are calculated in share stress and sheer strain, space. And we had this discussion just earlier a minute or 2 ago about the influence of compassion and subsidence and change in poor structure.

01:24:23.000 --> 01:24:31.000
If you looked at Looking the non share share strain domain to see if that improves the situation in terms of the energy.

01:24:31.000 --> 01:24:43.000
Approach, you know, compaction or expansion. Blowing metric, compaction, volumetric expansion, doing the work against the effect of confining pressure in addition to sheer strain during work against severe stress.

01:24:43.000 --> 01:24:47.000
Maybe thought about that and what difference do you think it would make?

01:24:47.000 --> 01:25:00.000
I don't think we have to answer the first question. So we've been focusing fairly in this year's share strange your stress space that's an interesting question.

01:25:00.000 --> 01:25:07.000
To think about. We are, we are in our laboratory tests, you know, we're running these.

01:25:07.000 --> 01:25:12.000
Constant volume tests, right? And so we're sort of constraining the volume to make sure we don't have volume.

01:25:12.000 --> 01:25:24.000
Metric change in those particular tests and then and then to seeing how to. So. Interesting question though, something we look at.

01:25:24.000 --> 01:25:25.000
Hmm.

01:25:25.000 --> 01:25:31.000
So it's something you could try in a triaxial. As you, and as a stabilization mechanism, we're collapse as an unstable.

01:25:31.000 --> 01:25:34.000
And it could be characterized in terms of energy. Just something to think about.

01:25:34.000 --> 01:25:36.000
Hmm, yeah, thank you.

01:25:36.000 --> 01:25:39.000
Yeah, thanks.

01:25:39.000 --> 01:25:51.000
Alright, I think we have maybe time for just one more question if anybody has. Remaining thoughts or comments.

01:25:51.000 --> 01:26:04.000
I guess if not, I'll just say I thought it was a really nice, mix of, perspectives from sort of big picture to, you know, sort of.

01:26:04.000 --> 01:26:08.000
Lab scale physical testing of soil samples. So. It was really nice to see the different approaches.

01:26:08.000 --> 01:26:23.000
And how they can kind of inform each other here. So, I think with that we'll just close and thank the, the speakers.

01:26:23.000 --> 01:26:25.000
Yeah.

01:26:25.000 --> 01:26:26.000
Thank you for inviting us.

01:26:26.000 --> 01:26:28.000
Wow.

01:26:28.000 --> 01:26:29.000
Yep, thank you.

01:26:29.000 --> 01:26:30.000
Thank you very much.

01:26:30.000 --> 01:26:35.000
Thank you so much, and thank you, Eric, for moderating. And thank you, Rob, wherever you all.

01:26:35.000 --> 01:26:36.000
That was a great session. There was a lot of fun. We have a very short break coming.

01:26:36.000 --> 01:26:48.000
Please come back at 1110 am Pacific time unless You are in our next Loma Predator session, in which case P stick around.

01:26:48.000 --> 01:26:55.000
That means you, Cynthia, Jeff, Roberts, Tim, Bolt, a completely different USGS, a scientist named Tom and Ruth.

01:26:55.000 --> 01:26:56.000
Perfect.

01:26:56.000 --> 01:27:00.000
Please stay, everyone else