WEBVTT
Kind: captions
Language: en-US

00:00:05.551 --> 00:00:09.834
[silence]

00:00:09.834 --> 00:00:13.059
- Hi, everyone. Welcome to the
USGS Landslide Hazards seminar.

00:00:13.059 --> 00:00:17.460
I’m Stephen Slaughter, and this meeting
is hosted by the Landslide Hazards Program.

00:00:17.460 --> 00:00:20.000
The seminar is a presentation
by researchers, academics,

00:00:20.000 --> 00:00:22.500
students, and professional
geologists from private industry

00:00:22.500 --> 00:00:26.000
and government agencies who are
working on some aspect of landslide

00:00:26.168 --> 00:00:28.700
and landslide science.
The seminar is organized by

00:00:28.868 --> 00:00:31.512
Matt Thomas,
Jaime Kostelnik, and me.

00:00:31.680 --> 00:00:33.800
The presentation is about
50 minutes long, and at the end,

00:00:33.800 --> 00:00:36.138
you can submit questions
via the chat window or use

00:00:36.139 --> 00:00:39.329
the raise-your-hand feature in
combination with your microphone

00:00:39.329 --> 00:00:42.800
and video camera. We typically wait
until the end of the presentation to

00:00:42.800 --> 00:00:46.159
take questions. During the presentation,
please remember to keep your

00:00:46.159 --> 00:00:50.100
microphone and cameras turned off.
Today’s speaker is Rex Baum

00:00:50.100 --> 00:00:55.700
from the USGS, and he is introduced
by Bill Schulz from the USGS. Bill?

00:00:55.700 --> 00:00:57.820
- Thanks, Stephen.
Good afternoon, everyone.

00:00:57.820 --> 00:01:01.339
It’s my pleasure to
introduce today’s speaker.

00:01:01.339 --> 00:01:04.200
Stephen already took that away
from me – Rex Baum. [chuckles]

00:01:04.201 --> 00:01:07.300
Rex is a research geologist in
the Landslide Hazards Program

00:01:07.301 --> 00:01:11.140
of the U.S. Geological Survey, and
he’s currently chief of the Landslide

00:01:11.140 --> 00:01:15.400
Initiation Processes and Probability
Project, but I’d argue that he’s

00:01:15.400 --> 00:01:18.600
essentially led the landslide project,
at least during the past two decades

00:01:18.600 --> 00:01:22.240
that I’ve been with the USGS.

00:01:22.240 --> 00:01:25.100
Rex was educated as a geologist
and an engineering geologist

00:01:25.100 --> 00:01:28.399
with a focus on landslides.
So he’s perfect for this role.

00:01:28.399 --> 00:01:32.400
He [audio garbled] graduate studies at
the University of Cincinnati where he

00:01:32.400 --> 00:01:37.208
studied with the late Arvid Johnson,
receiving his Ph.D. in 1988.

00:01:37.208 --> 00:01:42.380
While at USGS, Rex has performed important
research on landslide processes, landslide

00:01:42.380 --> 00:01:46.880
forecasting and warning, and he has a very
broad education skill set that ranges from

00:01:46.880 --> 00:01:51.819
geotechnics to detailed engineering geologic
field studies, monitoring, and modeling.

00:01:51.819 --> 00:01:56.317
I’m particularly fond of his research into
landslide kinematics, internal stresses,

00:01:56.317 --> 00:01:59.895
and surface deformation that reveals
those other aspects that are important

00:01:59.895 --> 00:02:05.213
in landslide controls and also his studies
on deformation of landslides as they

00:02:05.213 --> 00:02:08.800
move and consequently force groundwater
circulation that partly controls

00:02:08.800 --> 00:02:12.800
landslide speed. Of course, these days,
I think he’s very widely known around the

00:02:12.800 --> 00:02:18.100
world for his research on transient rainfall
infiltration effects on pore water pressure

00:02:18.100 --> 00:02:22.600
and slope stability, both at the local
and regional scale, especially with

00:02:22.600 --> 00:02:24.700
the model triggers
that he’s developed

00:02:24.700 --> 00:02:31.540
for forecasting effects of rainfall on
pore water pressure and slope stability.

00:02:31.540 --> 00:02:36.400
And that model has been used internationally
for regional landslide hazard assessments.

00:02:36.400 --> 00:02:40.549
In addition, Rex has also performed and led
many hazard assessments from essentially the

00:02:40.549 --> 00:02:45.720
day he started at the USGS, where he focused
on snowmelt-induced landslide disasters in

00:02:45.720 --> 00:02:50.352
central Utah. But, since then, he’s
studied all over the Rocky Mountain region

00:02:50.352 --> 00:02:53.626
as well as back in the Midwest
in Ohio, in California,

00:02:53.627 --> 00:02:58.925
Hawaii, the Pacific Northwest,
more recently in Puerto Rico and,

00:02:58.925 --> 00:03:01.725
in the past, in Poland
and El Salvador as well.

00:03:01.725 --> 00:03:05.600
I’ve really been fortunate to work with
Rex for the past 20 years and watch as he

00:03:05.600 --> 00:03:09.100
set a perfect example for
landslide scientists at the USGS,

00:03:09.100 --> 00:03:11.100
essentially checking
all of the boxes

00:03:11.101 --> 00:03:14.069
for supporting the mission of
the Landslide Hazards Program.

00:03:14.069 --> 00:03:17.381
Today he’ll talk about some important
improvements to his triggers model that

00:03:17.381 --> 00:03:21.719
I’m sure will have people excited for
future hazard assessment application.

00:03:21.719 --> 00:03:24.099
Thanks. Rex?

00:03:24.099 --> 00:03:26.400
- Thank you, Bill.

00:03:29.702 --> 00:03:33.509
I don’t quite know how
to follow that introduction.

00:03:33.509 --> 00:03:37.500
Thank you.
That was very generous.

00:03:40.586 --> 00:03:48.800
I want to take the opportunity today
to talk about developments in landslide

00:03:48.800 --> 00:03:55.300
assessments over the
last roughly five decades.

00:03:55.300 --> 00:04:01.500
So starting from before my
career started, actually, and …

00:04:03.313 --> 00:04:09.094
[silence]

00:04:09.094 --> 00:04:11.300
- Hey, Rex?
You may want to turn your video off.

00:04:11.300 --> 00:04:14.749
It’s kind of herky-jerky right now.

00:04:14.749 --> 00:04:19.209
- Okay. Well, I was a little herky-jerky
there trying to figure out where …

00:04:19.209 --> 00:04:20.810
- [laughs]

00:04:20.810 --> 00:04:23.220
- … how to get things to work right.

00:04:23.220 --> 00:04:25.536
So let me see if I can …

00:04:29.905 --> 00:04:32.650
Okay.
Got the camera turned off.

00:04:32.650 --> 00:04:35.317
- Okay. Yeah, and you’re still
in full screen. It looks good.

00:04:35.317 --> 00:04:37.827
You’re in slide – 
Historical perspectives.

00:04:37.827 --> 00:04:41.417
- Okay. Let’s see.
Let me go back. All right.

00:04:41.417 --> 00:04:43.120
So, yeah.

00:04:43.120 --> 00:04:49.425
My talk is in three parts. I’m going to
get some historical perspectives first.

00:04:49.425 --> 00:05:00.669
I’m really just looking at some of the
early developments in landslide assessments

00:05:00.669 --> 00:05:03.668
and then going to
jump up to the present.

00:05:05.723 --> 00:05:08.900
And I’m sorry. I’ve got an echo,
so I’m going to close my door.

00:05:08.900 --> 00:05:11.200
I’ll be right back.

00:05:12.590 --> 00:05:17.316
[silence]

00:05:17.317 --> 00:05:19.700
Okay.
That’s better.

00:05:21.919 --> 00:05:29.159
So wanting to take a look back in the
past to see – kind of help us see

00:05:29.159 --> 00:05:35.000
why we’re where we’re at today
and maybe take a peek a little bit

00:05:35.000 --> 00:05:38.750
into the future as to
directions we might want to go.

00:05:38.750 --> 00:05:43.116
So this is not a
comprehensive review.

00:05:43.116 --> 00:05:47.190
And it’s certainly not
a summary of my career.

00:05:47.190 --> 00:05:52.832
I’m mainly going to be talking about
other people’s work during this

00:05:52.832 --> 00:05:57.687
historical review here.
Want to start with a few definitions.

00:05:57.688 --> 00:06:02.670
So, first of all, an assessment –
what I mean by that is an estimate

00:06:02.671 --> 00:06:12.793
of the prospect, or in other words,
the probability or possibility that

00:06:12.895 --> 00:06:18.058
something will happen.
Susceptibility, with regard to

00:06:18.059 --> 00:06:26.763
landslides, is a state of being likely
to be affected or harmed by an event.

00:06:26.763 --> 00:06:32.858
And zonation is a term that I’ll
use on some upcoming slides.

00:06:32.858 --> 00:06:41.330
And so that has to do with dividing land up
into areas and ranking them by degree.

00:06:41.330 --> 00:06:45.150
And so here 
we have an example.

00:06:45.150 --> 00:06:51.000
There’s a little piece of
Earl Brabb’s map from 1972 –

00:06:51.000 --> 00:06:55.048
one of the early landslide
susceptibility maps.

00:06:55.048 --> 00:07:06.502
And it’s broken down numerically –
number 1 being the least susceptible area

00:07:06.502 --> 00:07:12.200
and Roman numeral VI being the most
susceptible with the exception of L,

00:07:12.200 --> 00:07:22.256
which is landslide deposits, and those are,
of course, absolutely the most susceptible

00:07:22.256 --> 00:07:32.340
in this map that he made.
As far as the assessment or estimate

00:07:32.419 --> 00:07:40.700
of likelihood that landslides would
occur, that was based on percentages

00:07:40.700 --> 00:07:56.300
of the rocks being affected by landslides,
and so the tall figure over to the right

00:07:56.300 --> 00:08:00.645
of the screen shows the
breakdown percentage

00:08:00.645 --> 00:08:06.300
of outcrop area 
affected by landslides.

00:08:09.728 --> 00:08:14.502
And that corresponds to the
different zones –

00:08:14.502 --> 00:08:21.000
Roman numerals I through VI plus L.
So that’s what we’re –

00:08:21.000 --> 00:08:24.039
that’s part of what we’re
going to be talking about today.

00:08:24.039 --> 00:08:32.400
Also going to review some concepts
about hazard and risk.

00:08:32.400 --> 00:08:36.900
So when we talk about the term
“natural hazard,” we’re really

00:08:36.900 --> 00:08:43.703
talking about a probability of a
potentially damaging phenomenon,

00:08:43.703 --> 00:08:49.700
and it’s going to be limited to a specific
time period and a specific area.

00:08:51.390 --> 00:08:58.000
Vulnerability is the degree of loss that’s
going to result given the magnitude

00:08:58.000 --> 00:09:01.900
of whatever this
damaging phenomenon is.

00:09:01.900 --> 00:09:14.745
Elements at risk are things like people,
buildings, utilities, and so on.

00:09:14.745 --> 00:09:22.700
And then the total risk is the product
of the elements at risk times the

00:09:22.700 --> 00:09:27.500
vulnerability and 
the natural hazard.

00:09:27.500 --> 00:09:32.000
And there’s an implied summation there
because the vulnerability is going to vary

00:09:32.000 --> 00:09:34.762
for each element at risk.

00:09:34.762 --> 00:09:47.234
And this information is simplified
from David Varnes’ 1984 paper

00:09:47.234 --> 00:09:56.190
in which he reviewed landslide zonation and
state-of-the-art up to that point in time.

00:09:57.326 --> 00:10:00.341
So let’s look at some early ideas.

00:10:00.341 --> 00:10:08.270
I’m going to first talk about some
in a related field of landslide warning.

00:10:08.270 --> 00:10:12.670
And then I’m going to – just on one slide,
and then I’m going to talk to some parallel

00:10:12.670 --> 00:10:19.800
developments for 
landslide zonation.

00:10:21.610 --> 00:10:29.842
So about 1969, there were a series of
storms that affected the Los Angeles area.

00:10:29.842 --> 00:10:37.325
In 1975, Russ Campell of the USGS
published a professional paper,

00:10:37.325 --> 00:10:42.785
and part of his findings
in there were that

00:10:42.785 --> 00:10:48.517
rainfall intensity seemed to have
something to do with initiation of

00:10:48.517 --> 00:10:53.480
landslides that turned into
deadly debris flows.

00:10:53.480 --> 00:11:03.480
Russ also proposed some early ideas
about a landslide warning system

00:11:03.480 --> 00:11:13.713
that could be based on weather radar
and using this idea of intensity.

00:11:13.713 --> 00:11:19.100
In 1980, Nel Caine published his
well-known paper on the intensity and

00:11:19.100 --> 00:11:27.029
duration relationship between
precipitation and landslides.

00:11:27.029 --> 00:11:32.900
And then, later on, various people
have proposed ideas related to

00:11:32.900 --> 00:11:39.100
hydrometeorological thresholds where
they proposed looking either at stream flow

00:11:39.100 --> 00:11:49.772
or other hydrological measures in
addition to the rainfall measures.

00:11:49.772 --> 00:11:56.400
And so some of these ideas that are
still in use today with regards to

00:11:56.400 --> 00:12:06.344
landslide warning have their origins
in the early 1970s –

00:12:06.344 --> 00:12:12.717
early to mid-1970s and so forth.
So just wanted to mention that

00:12:12.717 --> 00:12:16.079
in connection with the zonation
ideas that we’re going to be

00:12:16.079 --> 00:12:19.300
talking about 
most of the time here.

00:12:19.300 --> 00:12:25.980
So let’s look at some of the drivers
for landslide zonation.

00:12:25.980 --> 00:12:37.013
1952, Los Angeles had some storms that
produced landslides to development of

00:12:37.013 --> 00:12:45.473
a grading ordinance,
which was later updated in 1963

00:12:45.473 --> 00:12:49.900
after they had some more
storms and realized that they

00:12:49.900 --> 00:12:53.700
needed to make 
some improvements.

00:12:54.915 --> 00:13:02.050
Part of enforcing such an ordinance
is knowing where it needs to apply.

00:13:02.050 --> 00:13:11.350
1969 mudflows were added to the
National Flood Insurance Program.

00:13:11.350 --> 00:13:21.380
And around the same time, in New Zealand,
the idea of landslide insurance –

00:13:21.380 --> 00:13:25.040
and I have that in quotes – 
was developed.

00:13:25.040 --> 00:13:30.600
New Zealand had a national – has a national
fund that was started after World War II

00:13:30.600 --> 00:13:41.000
to pay for recovery from
damages caused by the war.

00:13:41.000 --> 00:13:49.195
Later, earthquakes were added to it,
and then in 1970,

00:13:49.195 --> 00:13:55.459
they started funding recovery
from landslides as well.

00:13:55.459 --> 00:14:03.300
France had a geologic hazards decree
requiring disclosure of geologic hazards

00:14:03.300 --> 00:14:07.653
in relation to plans to occupy land.

00:14:07.653 --> 00:14:12.899
I couldn’t find out the year, but Japan
had a landslide prevention law.

00:14:12.899 --> 00:14:22.129
Austria passed a forestry law
that had similar requirements to

00:14:22.129 --> 00:14:25.626
what was done 
in France in 1970.

00:14:25.626 --> 00:14:33.400
So there were these various legal
developments that helped to motivate

00:14:33.400 --> 00:14:42.694
the need for and use of landslide
zonation in regulating land use

00:14:42.694 --> 00:14:51.300
or in helping to identify
eligibility for benefits such as

00:14:51.300 --> 00:14:59.700
the Flood Insurance Program.
So the earliest landslide susceptibility

00:14:59.700 --> 00:15:05.660
map that I was able to
find was published in 1959.

00:15:05.660 --> 00:15:11.900
It was a map covering
the 48 United States.

00:15:11.900 --> 00:15:23.860
And it was based on numbers of landslides,
and the country was divided up according to

00:15:23.860 --> 00:15:30.564
Nevin Fenneman’s physiographic
regions of the United States.

00:15:30.564 --> 00:15:37.800
Another very early one, from 1968,
where the abstract is depicted there

00:15:37.800 --> 00:15:45.400
on the lower left, that was California
Divisions of Mines and Geology –

00:15:45.400 --> 00:15:53.100
a couple of geologists there
published an early map.

00:15:54.082 --> 00:16:02.393
Earliest known probabilistic landslide
susceptibility map was published in 1978.

00:16:03.550 --> 00:16:11.019
I wasn’t able to follow up on this
suspicion, but I think it maybe was

00:16:11.019 --> 00:16:22.139
a forerunner of a probabilistic program
that was used by a Forest Service later on.

00:16:22.139 --> 00:16:26.920
If I remember correctly, 
it was called [Lisa].

00:16:26.920 --> 00:16:39.276
And then 1984, UNESCO published
a review of landslide hazard zonation.

00:16:39.276 --> 00:16:45.569
Most of the work was done actually in the
late 1970s, but by the time the report was

00:16:45.569 --> 00:16:51.779
compiled and published, 
it came out in 1984.

00:16:51.779 --> 00:16:59.897
David Varnes, formerly of the USGS,
was the lead on that project,

00:16:59.897 --> 00:17:03.490
and so he’s 
credited as the author.

00:17:03.490 --> 00:17:10.436
An interesting side note is that Dave
was recognized by the French government

00:17:10.436 --> 00:17:21.775
and awarded what translates into English
as Knight in the Order of Academic Palms

00:17:21.775 --> 00:17:27.600
for this work that led to
the UNESCO publication.

00:17:27.600 --> 00:17:33.281
So it’s definitely worthy of reading.

00:17:33.281 --> 00:17:36.880
There’s a lot that’s
still applicable today.

00:17:36.880 --> 00:17:41.125
Even though the methods he describes
are somewhat out of date,

00:17:41.125 --> 00:17:51.030
a lot of the principles
are still worth reviewing.

00:17:51.030 --> 00:17:57.680
So I’m going to show some examples
of some of these early maps.

00:18:00.129 --> 00:18:07.780
You know, this – the maps shown here
are from an area in Cincinnati, Ohio,

00:18:07.780 --> 00:18:10.410
along the Ohio River.

00:18:10.410 --> 00:18:18.500
The mapping was done by a master’s
student at the University of Cincinnati,

00:18:18.500 --> 00:18:23.550
one of my classmates,
and so this was in the early 1980s.

00:18:25.128 --> 00:18:31.475
But it was based on a method that was
developed at Stanford University in the

00:18:31.475 --> 00:18:34.630
mid- to late 1960s.

00:18:34.630 --> 00:18:42.705
The first map made according to this,
I think, was dated about 1968,

00:18:42.705 --> 00:18:46.290
at least the first one
I’ve seen referenced.

00:18:46.290 --> 00:18:54.900
So the process was to first create an
engineering geologic map, which is depicted

00:18:54.900 --> 00:19:02.800
on the left. And so it shows the geology,
including the bedrock geology and

00:19:02.800 --> 00:19:09.513
the surficial geology from a
engineering/geologic standpoint.

00:19:09.513 --> 00:19:15.400
And then it shows 
landslide features.

00:19:15.400 --> 00:19:24.700
Other maps of this type might also
show faults or other geologic structures

00:19:24.700 --> 00:19:28.460
that could be important to the
stability of the slopes.

00:19:28.460 --> 00:19:34.725
It also shows evidence
of creep and seepage.

00:19:34.725 --> 00:19:38.920
You can see those features
depicted on the map.

00:19:38.920 --> 00:19:48.040
So then the landslide assessment
is depicted on the right.

00:19:48.040 --> 00:19:54.634
And it’s a map that classifies
the ground and the stable ground –

00:19:54.634 --> 00:19:58.800
potentially unstable
ground and moving ground.

00:19:59.517 --> 00:20:03.010
And then further subdivide it.

00:20:03.010 --> 00:20:11.010
And so this process, as you can imagine,
required some – a lot of field work and

00:20:11.010 --> 00:20:20.330
then the interpretive map required quite
a bit of geologic and engineering judgment.

00:20:23.689 --> 00:20:29.420
So now I will go back to Earl Brabb’s map,
which we talked about briefly earlier.

00:20:29.420 --> 00:20:33.930
So we saw this on an earlier slide.

00:20:33.930 --> 00:20:37.750
So this was published in 1972.

00:20:37.750 --> 00:20:39.992
This is a different approach.

00:20:39.992 --> 00:20:44.600
It’s more statistical in nature.

00:20:46.300 --> 00:20:48.290
They had a landslide inventory.

00:20:48.290 --> 00:20:56.400
There was also a slope map and a geologic
map were used in developing this.

00:20:58.030 --> 00:21:00.251
And so …

00:21:03.743 --> 00:21:15.700
This table describes the record
of landslides from the different

00:21:15.700 --> 00:21:20.800
geologic formations,
which are shown in the …

00:21:22.152 --> 00:21:29.820
[silence]

00:21:29.820 --> 00:21:32.337
… which are shown in this column.

00:21:32.337 --> 00:21:47.664
And then they used the density of
landslides to – or percent of area affected

00:21:47.664 --> 00:21:54.300
to subdivide it into different
susceptibility classes.

00:21:56.780 --> 00:22:06.100
And then, based on the different slope
intervals, an area might get a lower or

00:22:06.100 --> 00:22:11.840
a higher susceptibility ranking.

00:22:11.840 --> 00:22:18.180
So this was quite a tedious process,
as you might imagine.

00:22:18.180 --> 00:22:26.700
But it was also very systematic and
really a forerunner of many of the

00:22:26.700 --> 00:22:36.322
statistical methods that are applied today,
including some of the deep learning

00:22:36.322 --> 00:22:45.600
and other types of methods. It really comes
down to counting densities of landslides

00:22:45.600 --> 00:22:52.900
and correlating it with other data
that you’re able to collect.

00:22:52.900 --> 00:23:01.666
And so you can see from the little
blue rectangles that, for example,

00:23:01.666 --> 00:23:07.100
Zone 5 could consist of a number
of different formations and

00:23:07.100 --> 00:23:15.089
different slope categories.
So then it was – then it was one of the

00:23:15.089 --> 00:23:21.685
early forerunners of various statistical
types of mapping that we have today.

00:23:22.483 --> 00:23:31.800
So here we have the very earliest
probabilistic assessment that

00:23:31.800 --> 00:23:38.952
I was able to identify. It was
published in 1978 and was done up at

00:23:38.952 --> 00:23:47.100
Colorado State University in Fort Collins
under contract to the Forest Service,

00:23:47.100 --> 00:23:51.600
and the study area was at the
H.J. Andrews Experimental Forest

00:23:51.600 --> 00:24:00.570
in Oregon, where the USGS now
has an experimental debris flow flown.

00:24:00.570 --> 00:24:05.520
Just briefly about the
methodology applied here.

00:24:05.520 --> 00:24:15.800
You can see from the diagram here with
the various dimensions and so forth

00:24:15.800 --> 00:24:25.520
labeled on it, this approach
applied the infinite slope analysis.

00:24:25.520 --> 00:24:28.300
You can see the map 
over here on the right

00:24:28.300 --> 00:24:36.100
that the areas to which
it was applied were fairly large.

00:24:36.100 --> 00:24:43.200
I don’t know exactly what the pixel size
was, but the computer graphics

00:24:43.200 --> 00:24:45.050
that were used
were quite interesting.

00:24:45.050 --> 00:24:51.100
There was a different letter
for the different areas.

00:24:51.100 --> 00:24:58.080
So factor of safety less than or equal
to 1.2, typed a bunch of W’s in there.

00:24:58.080 --> 00:25:03.590
The printer put I’s in the
intermediate one for 1.2 to 1.7.

00:25:03.590 --> 00:25:10.010
And then factor of safety
greater than 1.7, it looks like a dot.

00:25:10.010 --> 00:25:19.300
So they ran their model for
a lot of different scenarios,

00:25:19.300 --> 00:25:26.410
different groundwater depths,
different strength parameter values.

00:25:26.410 --> 00:25:32.750
Differences in how much clear-cutting
was done and so forth.

00:25:32.750 --> 00:25:42.000
So that’s really the forerunner of
probabilistic assessments that have

00:25:42.000 --> 00:25:48.500
been done in more recent times,
even up to the present.

00:25:50.678 --> 00:26:07.154
So, in Varnes’ 1984 review, he has
a section about zonation principles.

00:26:07.154 --> 00:26:10.000
And it boiled down to these.

00:26:10.000 --> 00:26:14.400
First, the past and present
are keys to the future.

00:26:14.400 --> 00:26:23.732
That’s really going back to Charles Lyell
with the principle of uniformitarianism.

00:26:23.732 --> 00:26:28.509
It’s just stated inversely.

00:26:28.509 --> 00:26:32.890
He said the present is
the key to the past.

00:26:32.890 --> 00:26:44.110
So we’re trying to make geologic
predictions, but we go the other direction.

00:26:44.110 --> 00:26:49.500
Next, landslide-causing conditions
can be identified.

00:26:52.160 --> 00:26:59.400
Basic causes of landslides were already
well-known by 1950, as suggested by

00:26:59.400 --> 00:27:08.300
this paper by Karl Terzaghi published
in 1950, part of a GSA volume.

00:27:08.300 --> 00:27:13.890
And then the idea that degrees
of hazard can be estimated.

00:27:13.890 --> 00:27:28.000
And so this diagram showing factor of
safety and probability density compares

00:27:28.000 --> 00:27:31.078
a couple of cases.

00:27:32.367 --> 00:27:43.056
The blue one with the narrow curve
indicates the very low mean and

00:27:43.056 --> 00:27:48.657
nominal factor of safety. Also
low uncertainty because it’s narrow.

00:27:48.657 --> 00:27:55.725
And there’s a low failure probability
because the area below the factor of safety

00:27:55.725 --> 00:27:58.800
equal to 1 is very small.

00:27:58.800 --> 00:28:03.520
The red one,
much higher uncertainty.

00:28:03.520 --> 00:28:06.200
Even though the
factor of safety is higher,

00:28:06.200 --> 00:28:12.870
there’s much higher probability
of failure for this one.

00:28:12.870 --> 00:28:21.600
So being able to estimate the degrees of
hazard was a principle that was recognized

00:28:21.600 --> 00:28:26.960
very early on and still applies today.

00:28:28.575 --> 00:28:30.950
Let’s talk a little bit about progress.

00:28:30.950 --> 00:28:36.700
Without going into any specifics,
some of the things that have come along

00:28:36.700 --> 00:28:47.780
since those early – we now have methods
for validation of susceptibility maps.

00:28:47.780 --> 00:28:55.691
We have new tools, better tools
for remote sensing, for computing,

00:28:55.691 --> 00:29:00.120
for field data collection, so on.

00:29:00.120 --> 00:29:04.550
We have better data and more
and better geologic maps.

00:29:04.550 --> 00:29:07.040
Better topographic data.

00:29:07.040 --> 00:29:10.900
More data about
past landslides.

00:29:10.900 --> 00:29:18.760
And so those have led to increased
repeatability of the assessments.

00:29:18.760 --> 00:29:25.250
Being able to have other people – you know,
different people assess the same area and

00:29:25.250 --> 00:29:33.881
come up with similar results or
being able to apply methods in new areas.

00:29:33.881 --> 00:29:38.720
Objectivity has been on the increase,
I would say.

00:29:38.720 --> 00:29:48.664
I don’t think it’s likely that any
method will ever become totally objective,

00:29:48.664 --> 00:29:55.800
but certainly we can reach an
acceptable standard of objectivity.

00:29:55.800 --> 00:30:04.956
And our ability to assess landslides has –
and landslide hazard has expanded to

00:30:04.956 --> 00:30:14.338
a wider variety of landslide processes and
greater ability to assess the consequences,

00:30:14.338 --> 00:30:20.000
such as runout of landslides,
for example.

00:30:22.390 --> 00:30:26.626
There are some
continuing challenges, though.

00:30:28.064 --> 00:30:32.483
Uncertainty, although we’ve been
able to reduce some uncertainties,

00:30:32.483 --> 00:30:39.430
others persist and are
likely to for a long time.

00:30:39.430 --> 00:30:47.600
A lot of what’s under the
ground surface is unknowable,

00:30:47.600 --> 00:30:51.500
at least in the detail that
we’d like to know it.

00:30:53.785 --> 00:30:58.500
Slope is often a very good predictor of
landslides, and even though we have some

00:30:58.500 --> 00:31:02.800
very sophisticated methods now,
and we have more data, and so forth,

00:31:02.800 --> 00:31:07.610
it’s still challenging to produce
a map that does a better job

00:31:07.611 --> 00:31:12.420
than slope in predicting where
landslides are going to occur.

00:31:12.420 --> 00:31:17.870
I know there are
some exceptions to that.

00:31:17.870 --> 00:31:26.930
I’m well aware of some in various
places where steeper slopes are

00:31:26.930 --> 00:31:31.380
not necessarily the
more susceptible ones.

00:31:31.380 --> 00:31:38.370
But, in many places,
slope is hard to beat.

00:31:38.370 --> 00:31:47.700
Another challenge is dealing with
risk tolerance, and that really requires

00:31:47.700 --> 00:31:52.817
a lot of interaction with the
end users for our assessments.

00:31:52.817 --> 00:32:04.700
And they need to be well-educated and
well-prepared to talk about those concerns.

00:32:06.591 --> 00:32:12.690
So I’m going to briefly run through an
example of an ongoing assessment.

00:32:12.690 --> 00:32:18.000
I don’t have time to go into a lot of
detail on this, but Corina talked

00:32:18.000 --> 00:32:22.050
a little bit about some of this
a couple of weeks ago.

00:32:22.050 --> 00:32:33.700
So the project is in Puerto Rico, and doing
an assessment for three municipalities –

00:32:33.700 --> 00:32:39.400
Naranjito over here and
then Utuado and Lares.

00:32:41.332 --> 00:32:48.300
And I’ll just mention that there are
some specific geologic terranes that are

00:32:48.300 --> 00:32:52.406
of interest because they’re the
dominant ones in these areas.

00:32:52.406 --> 00:33:00.580
Granitoid rocks, there’s a submarine basalt
unit, and then marine volcaniclastics.

00:33:00.580 --> 00:33:08.920
So that’s those three rock types and soils
developed on them are what we’re dealing

00:33:08.920 --> 00:33:13.200
with mainly.
So those are depicted here.

00:33:16.184 --> 00:33:22.359
[silence]

00:33:22.359 --> 00:33:25.290
I’m going to talk just briefly
about the workflow.

00:33:25.290 --> 00:33:28.200
So we conducted field studies.

00:33:28.200 --> 00:33:35.360
Also did literature surveys to help compile
data from a couple of different sources –

00:33:35.360 --> 00:33:38.670
or, from many different sources,
I should say.

00:33:39.403 --> 00:33:46.250
Then there was a development stage where we
were experimenting with some of our ideas

00:33:46.250 --> 00:33:52.400
for how to conduct the assessment and
also doing some calibration work.

00:33:52.970 --> 00:34:02.140
And then final assessment
doing modeling and validation.

00:34:02.141 --> 00:34:13.505
Model steps involve modeling soil depth
then doing estimates of pressure head

00:34:13.505 --> 00:34:17.280
and one-dimensional slope stability.

00:34:17.280 --> 00:34:20.020
So that was done with triggers.

00:34:20.020 --> 00:34:28.210
And then doing some quasi-three-dimensional
slope stability analysis as the final step.

00:34:30.203 --> 00:34:37.745
I’ll just mention that the calibration
involved all three of these – of the –

00:34:37.745 --> 00:34:45.850
most of the work was focused on the soil
depth models and on the slope stability.

00:34:45.850 --> 00:34:56.242
We had some interaction between
these because soil depth is maybe

00:34:56.242 --> 00:35:04.605
not a precise term. We were trying to
model depth of shallow landslides.

00:35:04.605 --> 00:35:08.500
We didn’t actually go out and measure
soil depth, but in most cases,

00:35:08.500 --> 00:35:11.490
soil depth was quite similar
to the depth of the slides.

00:35:11.490 --> 00:35:15.650
And so that’s really
what we were modeling.

00:35:15.650 --> 00:35:19.903
And so there was some interaction between
the slope stability modeling and the soil

00:35:19.903 --> 00:35:29.300
depth modeling to come up with depths that
were going to give us the best predictions

00:35:29.300 --> 00:35:33.500
for landslide locations.

00:35:35.950 --> 00:35:44.940
So one of the earliest steps with
calibrating the strength parameters.

00:35:44.940 --> 00:35:49.860
We had some
information from the field.

00:35:49.860 --> 00:35:52.360
We knew depths of a lot
of these landslides.

00:35:52.360 --> 00:35:55.911
We also know the slopes
they occurred on.

00:35:55.911 --> 00:36:05.730
And we had some idea of ranges
of strength parameters.

00:36:05.730 --> 00:36:14.700
And so we ran a series
of models for dry conditions

00:36:14.700 --> 00:36:23.038
and for fully saturated conditions.
Used a synthetic grid of

00:36:23.039 --> 00:36:32.300
different slope angles and depths and
computed factor of safety for all of these.

00:36:32.300 --> 00:36:37.900
So I’m showing a couple of
different parameter combinations

00:36:37.900 --> 00:36:43.660
in these four charts. And the upper ones
are dry conditions, zero pore pressure.

00:36:43.660 --> 00:36:49.500
And then the lower ones are pore pressure
with water table at the ground surface.

00:36:50.560 --> 00:37:02.450
Taking all of these and then finding which
combinations were successful for our three

00:37:02.450 --> 00:37:09.180
different geologic terranes,
came up with these charts.

00:37:09.180 --> 00:37:16.350
And so, for the granitoid terrane,
those slides were generally very thin,

00:37:16.350 --> 00:37:18.731
shown by these green X’s.

00:37:18.731 --> 00:37:26.700
And so we tended to have the highest
success with rather high friction angles

00:37:26.700 --> 00:37:37.300
and low cohesion. Other slide types tended
to be – tended to have more that were

00:37:37.300 --> 00:37:39.355
as much as 15 meters deep.

00:37:39.356 --> 00:37:47.430
And so our results for those tended to have
higher cohesion and somewhat lower angles

00:37:47.430 --> 00:37:52.910
of friction but still fairly high because
the slopes were quite steep there.

00:37:53.942 --> 00:38:00.290
Here’s an – here are some
examples of our soil model.

00:38:00.290 --> 00:38:04.400
This is for a small
test area in Naranjito.

00:38:05.770 --> 00:38:13.400
And so the little squares show kind of a
comparison between – these were kind of

00:38:13.400 --> 00:38:16.880
our four best models
for soil depth.

00:38:16.880 --> 00:38:21.300
I had a graduate student from
the Colorado School of Mines

00:38:21.300 --> 00:38:26.200
who did the calibration
exercises for me.

00:38:27.598 --> 00:38:31.500
And his name is Matt Tello.

00:38:33.180 --> 00:38:41.100
So anyway, these were our best estimates
on what soil depths looked like.

00:38:41.670 --> 00:38:52.300
And then, for these different ones – this
is the one-dimensional factor of safety.

00:38:53.150 --> 00:38:56.090
Got some kind of surprising results.

00:38:56.090 --> 00:39:04.100
Overall, our best-performing model for
landslide susceptibility turned out to be

00:39:04.100 --> 00:39:13.600
this one, which assumed a constant
average depth for the soil and

00:39:13.600 --> 00:39:19.005
1D slope stability analysis. And so
it had an area under the curve of,

00:39:19.005 --> 00:39:24.800
I think, about 0.88.
But close behind were these other

00:39:24.800 --> 00:39:33.540
well-performing models which
we prefer because we think the

00:39:33.540 --> 00:39:40.900
soil depths are a little more true
to what we actually see in the field.

00:39:41.950 --> 00:39:46.960
So even though they didn’t get
quite as good a result as constant depth,

00:39:46.960 --> 00:39:48.920
they’re still very good.

00:39:51.209 --> 00:40:00.100
And then, with the quasi-3D method,
so we use a – kind of a –

00:40:00.100 --> 00:40:10.109
something sort of like a sliding gold
pan to represent our failure surface.

00:40:10.109 --> 00:40:14.081
And we try it at the center of
each grid cell throughout the

00:40:14.081 --> 00:40:21.240
digital elevation model. Results are shown
over here, again for our calibration area.

00:40:21.240 --> 00:40:29.400
We had one other parameter,
and that’s kind of the size of the search –

00:40:29.400 --> 00:40:32.807
or, of the trial surface.

00:40:32.808 --> 00:40:37.828
And let’s see here.

00:40:40.187 --> 00:40:47.618
Here, the bottom one and the top one
compare the same soil model but with

00:40:47.618 --> 00:40:54.300
different radii
for the trial surface.

00:40:55.170 --> 00:41:00.500
The larger trial surface
gives maybe a prettier map,

00:41:00.500 --> 00:41:07.553
but it does not perform as well,
so that curve is shown by

00:41:07.554 --> 00:41:11.920
the blue one, and then the
brown one is up there at the top.

00:41:11.920 --> 00:41:16.200
That’s our best-performing model,
which is up here.

00:41:17.956 --> 00:41:27.870
And so the area under the curve
was somewhat less for these,

00:41:27.870 --> 00:41:35.055
closer to about 0.8 for the
best-performing one compared to

00:41:35.055 --> 00:41:40.818
a higher value on the
one-dimensional factor of safety.

00:41:40.818 --> 00:41:47.363
But I’ll show you here a close-up
to compare these, why we prefer

00:41:47.363 --> 00:41:54.000
the quasi-three-dimensional output.

00:41:55.140 --> 00:42:00.596
You can see the edges of these zones
using the one-dimensional factor of safety

00:42:00.596 --> 00:42:02.801
are quite ragged.

00:42:02.801 --> 00:42:08.950
It’s very hard to regulate land use
with something like that.

00:42:08.950 --> 00:42:19.000
Having smoother boundaries to the zones
with high landslide susceptibility

00:42:19.000 --> 00:42:25.870
is much easier to use in a
regulatory or planning situation.

00:42:25.870 --> 00:42:34.800
And so we prefer that even though the
performance metrics are slightly less.

00:42:35.651 --> 00:42:42.720
Okay, so just to wrap up now
with some closing thoughts.

00:42:42.720 --> 00:42:59.579
So we’ve seen these early landslide
zonation principles are still valid –

00:42:59.579 --> 00:43:11.882
this idea of uniformitarianism or that we
can use past and present conditions to …

00:43:16.125 --> 00:43:20.200
… predict what’s
likely in the future.

00:43:20.614 --> 00:43:26.830
Causes are known, although we
continue learning more about that.

00:43:27.400 --> 00:43:34.657
They’re definitely capable of being
determined in nearly every case.

00:43:34.657 --> 00:43:43.900
And we can make estimates of hazard,
and we can assess which areas

00:43:43.900 --> 00:43:47.170
are more hazardous than others.

00:43:47.170 --> 00:43:54.960
We’ve also seen that new tools can
add value to our assessments.

00:43:54.960 --> 00:44:03.780
They can extend our vision and capacity and
certainly reduce error from the days when

00:44:03.780 --> 00:44:08.876
we did things manually,
and tedium as well.

00:44:10.103 --> 00:44:15.106
So I think we have some
opportunities for innovation.

00:44:15.106 --> 00:44:24.260
I think we have some opportunities to take
risk into consideration a little bit more

00:44:24.260 --> 00:44:28.700
in our assessments.

00:44:30.040 --> 00:44:39.217
So this slide is rather busy.

00:44:39.217 --> 00:44:41.400
But it …

00:44:46.380 --> 00:44:55.700
It shows some information about landslides
or landslide areas that I have known.

00:44:56.380 --> 00:44:59.285
So these are all
in the United States.

00:44:59.285 --> 00:45:14.517
And some of them are larger metropolitan
areas, and some are more rural areas

00:45:14.517 --> 00:45:19.650
or maybe single landslides.

00:45:19.650 --> 00:45:21.800
So …

00:45:24.380 --> 00:45:26.000
So we see a wide range here.

00:45:26.000 --> 00:45:32.440
Now, I want to mention something
about these two red lines.

00:45:32.440 --> 00:45:43.100
So if you look in the Schuster and
[inaudible] book on landslide investigation

00:45:43.100 --> 00:45:50.942
and mitigation, there’s a chapter in there
by Wu and others about hazard

00:45:50.942 --> 00:46:01.200
and risk assessment. And they’ve got
a chart in there that they reproduced

00:46:01.200 --> 00:46:07.600
from another publication which
shows levels of acceptable

00:46:07.600 --> 00:46:18.290
risk for various activities,
including open pit mine slopes

00:46:18.290 --> 00:46:20.190
and what have you.

00:46:20.190 --> 00:46:24.640
And so those red lines are
taken from that chart.

00:46:24.640 --> 00:46:34.000
So accepted risk and marginally accepted
risk in various engineering endeavors.

00:46:35.268 --> 00:46:43.548
And so, based on information I had about
likely range of annual probabilities

00:46:43.548 --> 00:46:50.470
for landslides for these various locations,
get the vertical error bars.

00:46:50.470 --> 00:46:54.000
Horizontal ones have to do with
consequences, and they’re based

00:46:54.000 --> 00:47:03.500
either on lives lost or on cost in
dollars or some combination thereof.

00:47:03.953 --> 00:47:08.690
And I just used the
scaling from their chart.

00:47:08.690 --> 00:47:13.366
I don’t really believe it’s possible
to put a dollar value on a human life,

00:47:13.366 --> 00:47:19.100
but that’s the scale that they were using,
so I’ve used their scale,

00:47:19.100 --> 00:47:23.200
except I’ve updated it
to 2020 dollars.

00:47:26.190 --> 00:47:37.098
So I think we have opportunity as we go
into areas to do assessments to identify …

00:47:39.333 --> 00:47:43.659
[silence]

00:47:43.660 --> 00:47:51.300
… zones with the
greatest hazard and hopefully

00:47:51.300 --> 00:47:55.350
separate those from areas
with lower hazard.

00:47:55.350 --> 00:48:01.785
So these – so the purple here
shows Seattle.

00:48:01.785 --> 00:48:14.160
And then the purple squares along here
show zones that Bill Schulz mapped.

00:48:16.726 --> 00:48:24.402
We did a assessment of landslide
susceptibility several years ago,

00:48:24.402 --> 00:48:36.594
and his zones can clearly separate
areas with acceptable risk

00:48:36.595 --> 00:48:40.740
from those that
have unacceptable risk.

00:48:42.225 --> 00:48:48.489
In that particular case, it’s a –
the areas with unacceptable risk are

00:48:48.489 --> 00:48:52.980
relatively small. And so that’s
an ideal we want to strive for.

00:48:52.980 --> 00:49:00.290
But, in cases like Puerto Rico that I was
just showing you, because there is so much

00:49:00.290 --> 00:49:09.150
mountainous area, that area of
unacceptable risk is much larger.

00:49:12.028 --> 00:49:15.100
I just want to conclude
with a few cautions.

00:49:15.100 --> 00:49:20.125
We’ve got a lot of great tools, a lot
of great technologies available to us,

00:49:20.125 --> 00:49:23.109
and new developments.

00:49:24.804 --> 00:49:34.300
But I hope that we don’t let our
enthusiasm for those take us away from

00:49:34.300 --> 00:49:40.598
doing field work and staying in touch
with the actual processes going on.

00:49:40.598 --> 00:49:45.800
Modeling and remote sensing are
no substitute for field work.

00:49:46.469 --> 00:49:50.634
Computer models,
no matter how sophisticated they are,

00:49:50.634 --> 00:49:54.973
are only as good as the
data that we put into them.

00:49:54.974 --> 00:50:04.700
And finally, just because things are
correlated doesn’t mean that there’s

00:50:04.700 --> 00:50:08.060
actual causation involved.

00:50:08.060 --> 00:50:14.152
And so I hope that we can always
keep those in mind going forward

00:50:14.152 --> 00:50:23.060
so that we can keep our assessments
relevant and accurate and meaningful.

00:50:25.434 --> 00:50:35.130
I just want to close with Dave Varnes’
recipe for success in doing assessments.

00:50:35.130 --> 00:50:42.000
This is from a section titled something
about operational principles.

00:50:42.900 --> 00:50:45.820
So defining our purpose.

00:50:45.820 --> 00:50:50.100
Letting the purpose and the processes –
meaning the landslide processes –

00:50:50.100 --> 00:50:52.704
find what should be done.

00:50:52.704 --> 00:50:56.376
Identifying and
involving the users.

00:50:56.376 --> 00:51:03.672
And doing the investigation in phases
and using the best skills and

00:51:03.672 --> 00:51:12.127
best tools obtainable. I think
if we keep those things in mind,

00:51:12.127 --> 00:51:16.312
we can be successful in the future.
Thank you.

00:51:18.285 --> 00:51:25.117
[silence]