Climate change and landslides in subpolar Alaska: Less ice, more water

Video Transcript
Download Video
Right-click and save to download

Detailed Description

Landslide initiation processes in subarctic Alaska are complicated by the prevalence of ice-rich permafrost. Permafrost and permafrost thaw influence landslide type, frequency, and continued ground surface deformation, resulting in positive feedbacks between permafrost thaw and landsliding. Thaw-induced landslides in interior Alaska and rain-induced landslides across the state pose persistent hazards to vulnerable communities throughout Alaska.

Patton A (2021) Climate change and landslides in subpolar Alaska: Less ice, more water. USGS Landslide Hazards Program Seminar Series, 10 November 2021.


Date Taken:

Length: 00:53:04

Location Taken: US

Video Credits

Credit: Annette Patton, University of Oregon.
Caption: A view of snowy Denali and the steep hillslopes of the Alaska Range, viewed from the Denali Park Road in June 2017.



All right. Well, I can see we just
hit one minute after the hour,

so we’re going to go ahead
and get started.

My name is Matt Thomas,
and thank you for tuning in

to the USGS Landslide Hazards
Program seminar series.

For those of you that are new to this
meeting, you have the ability to submit

questions via the chat window or the –
use the raise-your-hand feature

in combination with your
microphone and video camera.

We typically wait until the end of
the presentation to take questions.

And, in the meantime, if you’re
not intending to speak, please just

do your best to make sure your video
is off and your microphone is muted.

Josh, thanks so much for
introducing today’s speaker.

I’ll turn it over to you.

- Thanks, Matt. It’s my pleasure –
I’ve been such a fan of this

lecture series, so it’s been
a fantastic way to stay connected

and learn a ton.
So, thank you, Matt,

for keeping this whole thing going,
and more so next year.

But today we’re here to hear
from Annette – Annette Patton.

Dr. Annette Patton is
a product of Oregon.

She’s proud to be from Oregon,
but I think even more so, Oregon’s

proud to call her one of our own.
She grew up just outside of Corvallis

just up the street here from
Eugene, where she is now.

She went to Whitman College and
was a geology major there before she

went on to Colorado State University.
She did her master’s working on

debris flows in the Front Range.
Did tons of mapping.

Did some – a lot of field work.
Analysis of the debris flow activity

in that part of the world –
even some cosmo dating.

And she moved up far north to
Alaska to work on landslides and

ice and water – some of the stuff
we’re going to hear about today.

That was for her Ph.D. And she
stayed at Colorado State to do that.

All of that was with Sara
Rathburn, who’s a fantastic

fluvial geomorphologist,
but was also really great

at advising Annette through this
more hillslope-oriented work.

And I think that also speaks to
Annette’s sort of ability to work well

and pull together skills and people.
And she’s had some great

collaborations with Denny Capps
and others up in Alaska

that have been
important for her work.

And so that – the products of all
that you’re going to see today.

I had the great fortune of getting to
know Annette when I was advertising

for a postdoc for a brand-new grant to
work on landslide warning systems in

Sitka, Alaska, in the southeast area.
And I got this email from Annette when

she was – not yet finished her Ph.D.,
and she was – you know,

expressed interest in this position.
At the time, I was really sort of

dead set on thinking I need
someone with sensors experience –

someone that knows how to do sensors,
and they can learn the geology.

And she was, you know,
suggesting that maybe I should

think the other way.
And I remember thinking, okay.

Then she showed up at my office and
came in, and she was, you know,

visiting family – not completely
unannounced, I should say.

And within, like, 10 minutes,
it was obvious that this was

the person for the job.
And I can’t tell you how delightful

it’s been to have Annette working on
this project in southeast Alaska.

She was also a science-in-
residence in the town of Sitka.

So the Sitka Sound Science Center has
these science-in-residence programs,

so they bring in a scientist,
and they live there, and they

interact with the community.
So Annette can walk around that

town and, you know, chew the fat
with city managers and K-through-12

students and teachers and others,
and she’s – became a part of that.

And they really trust her and her –
and her science background, and her

prowess is really respected and
recognized in that community.

And that really is incredibly helpful
for the project and the work we’re

doing there helping that community
understand their hazards context.

And so, with that, I’m going to pass
the mic to Annette, and it’s all yours.

- Well, thank you, Josh.
That was very kind.

All I can say is that I am so excited
to be here today to share some of this

really cool work with all of you folks.
This has been, as Josh said, an excellent

seminar series, and I really appreciate
the opportunity to participate.

So Josh mentioned a couple of different
research projects that I’ve been involved

in over the last few years as part of
both my Ph.D. graduate program

and my current postdoc work.
And today, what I really wanted to do

was bring together some conclusions
from the last several years of work that

I’ve done up in Alaska with my mentors
Sara Rathburn and Josh Roering and

pull out a couple of themes,
thinking about landslides in these

high-latitude environments.
And specifically, I wanted to

think about how climate change and
landslides might be interacting up

in high-latitude areas like Alaska.
And you’ll note I have this little

catchy – or, hopefully catchy
theme of, less ice, more water.

And those are some of the
themes that I’m going to

really be focusing on
in this talk today.

Before I get started, though,
I have rather scary-looking

acknowledgement slides here.
Because all of the work that

I’m presenting has been
very collaborative work.

I couldn’t have done it without field
help and logistic help, intellectual

contributions, technical support,
by this long list of individuals

as well as the many agencies that
have partnered on various aspects

of the several projects I’m
going to talk about today.

So I just want to really acknowledge
upfront that this is all very collaborative

work and couldn’t have been
accomplished by any one person alone.

So thank you to these
many contributors to

all of this research that
I’m going to talk about.

Again, give you just a little overview
of where I’m going today.

I want to talk about climate change
and the influences on landslide process

in high-latitude sub-polar climates.
And, to do that, I’m going to focus on

two main factors of climate change that
are likely to interact with landslides.

The first is that increased temperatures
and climate change in general is

resulting in pretty widespread loss
of permafrost, particularly in

sensitive areas where permafrost
is a little bit more fragile.

And then the second big change that
is a result of climate change is changes

to precipitation patterns – possibly
changes in increasing rainfall totals

and rainfall intensity with potentially
more frequent extreme rainfall events.

So, first, we’re going to go to Denali
National Park in interior Alaska to

really hone in on a discontinuous
permafrost region and think about

how landslides and ice interact.
And then, second, I’m going to take you

down to southeast Alaska because
there’s no better place to learn about

very wet climate landslides than
in steep, rainy, southeast Alaska.

So I’ll dive right into geomorphic
process and landsliding

in Denali National Park. And
we’re going to be focusing on ice.

And, to give you a little bit of
background on some of the ice

processes that matter, we’ll first
maybe just highlight the fact that

all climate models that are looking at
climate change and forecasted climate

futures are looking at projecting
increases – or, decreases in permafrost

area in places like Alaska.
And the figure on the left is showing

total permafrost area in the future,
and all of those lines are declining,

showing that, no matter what climate
model you use, there’s likely to be

a pretty dramatic loss in permafrost
by area in places like Alaska.

And, on the right, you see
a map of the Arctic Circle

and some of the areas south of that.
And these are all projections, again,

with different climate models showing
possible future extents of permafrost.

And the different colors – the color
scheme is a little odd, I know.

But the different colors are just telling
you the number of models that agree on

possible future permafrost extents.
And the main takeaway here is that,

while permafrost in the really
high-latitude, polar, Arctic regions

is likely to persist in various
climate futures, there’s a lot of

expected loss of permafrost, particularly
in places like interior Alaska and some

of these areas where permafrost is
already a little bit more sensitive.

And so that’s one of the reasons
why Denali is a great place to study

thawing permafrost landslides.
Currently, Denali is situated in

an area that is considered discontinuous
permafrost, meaning that there’s

patchy distribution of permafrost.
It doesn’t exist everywhere.

But, on some of the places that
are a little bit colder – maybe higher

elevations or north-facing hillslopes,
permafrost is pretty extensive

in this area. So that’s
where we’re going to focus.

But, again, this is a very sensitive
kind of on-the-cusp permafrost area.

And Denali-specific permafrost models
estimate that less than 1% of the park

is going to be underlain by permafrost
by the end of this century.

So we’re talking about
drastic changes to geomorphic

conditions over the
next several decades.

And that has some really important
implications for landslide process.

First of all, changing permafrost
characteristics has really important

implications for hydrologic process.
Permafrost, by definition,

is perennially frozen. And that
may or may not include water.

So ice may or may not be present.
It could just be really cold ground

all the time. But, by nature of
being really cold – below zero degrees

Celsius – permafrost is a very
effective barrier to groundwater flow.

Whether it’s ice itself or just frozen
ground, water really can’t flow through

permafrost very effectively.
So, whenever permafrost thaws,

that means that you increase
the hydraulic conductivity

and connectivity of the subsurface,
meaning that groundwater

is a lot more able to move between
different areas in the subsurface.

Notably, the active layer, or the layer
above permafrost that does thaw

in the summers, that’s likely to
expand and create some of these

increased connections between
different surface water reservoirs.

Now, additionally, many places
where permafrost exist do include

pretty rich ice. There are high
concentrations of frozen water.

And, if ice melts, that results in
a pretty large burst of pore water.

So big temporary increases
in pore water pressure can also

result from
thawing permafrost.

Now, there are some important
physical processes to consider as well.

Ice is a really nice cohesive matrix.
So, when that ice melts, that reduces

the cohesion of the subsurface.
And ice, with a little bit of water on it –

if you’ve ever been walking around on
an icy day, you know that that’s a very

low-friction surface. So that can also
really influence landslide process.

And finally, as these very perennially
frozen conditions transition into a

maybe seasonally frozen climate
instead, you’re going to be transitioning

from a permafrost to a freeze-thaw
system where ice wedging

and cryostatic pressure can
start to make bigger changes

and increase fracture patterns
in the subsurface.

Now, landslides in
permafrost are not new.

And they can look like
a lot of different things.

Some landslides are rockfalls.
You’ve maybe heard the term

“retrogressive thaw slump,”
where these sort of low-angle

hillslopes can kind of increase –
or, the landslide body can sort of retreat

upslope as the scarp continuously fails.
But some permafrost landslides

also happen in steep areas –
changes in glacier process and

alpine permafrost can influence
landsliding in those areas as well.

So landsliding in permafrost
areas is very diverse.

One particular type of landslide
that I’ll talk about a little bit later

is called an active layer detachment.
I’ll give you a little better description

later, but that’s a type of
permafrost-specific landslide

where the perennially thawed portion
of the soil detaches from the

permafrost beneath and slides
along the permafrost boundary.

Now, to summarize some of that sort of
introductory process, permafrost thaw

is likely to be changing both the
hydrologic and physical properties

of soil and subsurface materials.
And that’s likely to increase landslide

frequency and spatial distribution,
particularly in these sensitive areas.

Now, in Denali specifically, again, there
are a lot of different types of landslides.

Some of them are occurring on steep
hillslopes – things like debris flows

and big rock avalanches that happen
on steep hillslopes like you would

see elsewhere in the world.
But others are some of these

more permafrost-specific landslides
like active layer detachments,

which you see here and
in the photo on the top left.

And, in Denali, those
landslides are important.

We can see in human history how
landslides are shaping the landscape.

Back in the early 1900s, a very large
valley-spanning landslide blocked one

of the rivers in Denali National Park.
It resulted in a temporary lake that

was called Bergh Lake.
And that lake appears on some of the

older U.S. topographic maps of the area.
But, a few decades after that landslide

happened, the landslide dam
breached, and the lake drained.

And there is no longer a lake called
Bergh Lake in Denali National Park.

So we can see those landslide process –
sort of controls on the landscape

in human history. We can also see
how landslides are important at

redistributing carbon and sediment
throughout the landscape, and,

in many cases, delivering hillslope
materials to surface water.

So this is an example of a debris flow in
Denali National Park that brought this

beautiful fan of sediment and organic
material down to the McKinley River.

And, if you folks have been regularly
watching this seminar series,

you hopefully tuned into
Denny Capps’ seminar

a few weeks ago where he
talked specifically about

the Pretty Rocks Landslide
in Denali National Park.

It’s a landslide/rock/glacier complex
where this very rich ice in the

subsurface is resulting in pretty
catastrophic complications for

maintaining a road in this area.
So, as you can see in the photo

on the left here, this is a difficult
place to try to take vehicle traffic.

So these hazards are
also very important

to human infrastructure
and human safety.

So the work that I did for my Ph.D.
involved some mapping – geomorphic

mapping and landslide inventory
along the Park Road corridor.

And that’s what you see here
on this map, where the black line

is the Denali National Park Road,
and the green boundary

is our study area
that we defined.

And this is a really important
infrastructure resource for the park.

This is the only road that accesses
the interior of Denali National Park.

So there are some private
stakeholders – or, private landowners

with inholdings in the park.
This is the only way that they can

take ground traffic
into their private property.

And it’s also one of the primary
ways that visitors get to interact

with the park. They take buses out, and
they get to see these classic landscapes.

So this is a really important resource
for the park, and understanding

landslide hazards in this area
is of great practical concern.

So all of these orange dots that you
see are landslides that we inventoried

in the course of this project.
We visited all of them on foot and

verified them and mapped them on
the ground, with a few exceptions,

where some wildlife encounters
prevented up-close [laughs]

investigation. There are a couple of
dashed lines on my geomorphic maps

where we said that there
was a bear in the way.

But, in general, we visited all of
these landslides in person.

So we have a really rich data source
of really precise information

about landslides all along this
Park Road corridor

between Mile 33 and
Mile 69 of the Park Road.

And I mentioned some
geomorphic mapping.

I’m going to just zoom in and show
you an example of what that looks like.

This isn’t the main focus of the
talk today, but I feel strongly that

geomorphic and geologic context is
important for understanding landslide

hazard, so I wanted to share some
of that with all of you folks.

This is just an example of what
some of that mapping looks like.

You’ll see it’s a nice little strip
map that follows the Park Road.

And the geology of Denali National
Park is very notoriously complicated.

There’s a lot of
felsic volcanic rocks.

There’s some Cretaceous sedimentary
rocks, as well as, of course,

a very complex Quaternary
sedimentary sequence.

So you don’t necessarily
need to know exactly

what all of these different
colors mean, but I do want you

to notice how complicated
and diverse this landscape is.

And what we found in this mapping is
that our landslides occurred in many

different geologic units,
but there were two main offenders

in terms of sourcing landslides.
And the first were the felsic volcanic

rocks, which are some of the examples
that you see in, like, the pink units here.

There’s little yellow dots,
and the Pretty Rocks Landslide,

which you’ve already seen.
Those felsic volcanic rocks –

we conducted some petrographic
analysis, and all of the feldspars

are weathering rapidly to various types
of clay and results in a very susceptible

lithology for landsliding. But we
also found that the vast majority

of landslides in Denali National Park
occur in unconsolidated Quaternary

sediments – things like outwash
deposits, old relict landslide deposits,

and other fluvial deposits
on riverbanks and things like that.

And so these are some maybe typical
examples of what that might look like.

Some of these big, old landslide
deposits sourcing new landslides as

well as these shallow-angle landslides
in glacial deposits and things like that.

In terms of the permafrost,
we can look at a couple of

other geomorphic characteristics
of this landslide population

and think about what
they mean as well.

So I’ve got a couple of images that are
going to look a little bit like this,

where the blue histogram is showing
you the distribution of a characteristic

in the entire study area.
So a pixel-by-pixel histogram showing

slope angle, in the case of this top plot,
for the entire study area.

And then the red distribution
is the same characteristic,

but for the
population of landslides.

So what you see here is that there
are places where landslides occurred

on hillslopes that are less common
in the regional study area.

And, in fact, we see a bimodal
distribution of slope angles

where landslides occurred.
Some of them occur on the

very steep hillslopes, which is
what you might expect at, you know,

20-, 25-, 30-degree hillslopes.
But there’s also a population of

landslides occurring at lower slope
angles with a median of about

18 degrees and some landslides
occurring even down on hillslopes

below 10 degrees.
We verified all of these in person.

They are landslides that are
occurring on very shallow hillslopes.

Another characteristic that’s
important to note here is the

mean decadal ground temperature.
So, again, the blue distribution,

which kind of goes in the background
here, that’s showing the entire study

area mean decadal ground temperature.
And the red is the landslide population.

And that mean decadal ground
temperature is important because

it’s a metric telling you
where permafrost exists.

If the decadal ground temperature is
below zero degrees, that’s permafrost.

If it is above zero degrees,
that’s not permafrost.

What you might notice is that
landslides are preferentially occurring

down in the permafrost regions.
There are landslides occurring

in non-permafrost areas,
but they are primarily and

disproportionately occurring in the
colder portions of our study area.

If you want to see what that looks like
on a map, this is that same study area

map that I showed you earlier,
but superimposed on a model of

permafrost distribution in the park.
And I can’t take credit for this

modeling work myself.
Panda and others, in 2014,

produced a Technical Report
with a permafrost model

for Denali National Park.
And you’ll see that, where this

blue area is, those are the places where
ground temperatures are below zero.

That’s permafrost. And where
it’s red, that’s not permafrost.

So, again, you’ll see that these areas
are not continuous permafrost.

It doesn’t exist everywhere in the park.
But the landslides are tending to

cluster in these places that are
a little bit blue in color, meaning that

there is permafrost underneath,
at least according to this model.

What you might also notice is that a lot
of these pixel colors in all of these cells,

they’re very light colors.
That means that the permafrost

temperatures are very close to zero,
and that kind of gives you maybe

a sense of how
fragile this system is.

This permafrost is not really, really,
really cold. It’s just near zero.

So, as climate change is warming,
these are areas that are probably

going to be experiencing a lot of pretty
rapid change to permafrost distribution.

Now, if you look back at slope angle
again, but instead think about

permafrost and seasonally thawed
areas separately, you maybe can

see where these two different
distributions are coming from.

Landslides in permafrost had
a lower median slope angle than

the landslides in the seasonally
thawed areas, which kind of matches

what you might expect – that loss of
cohesion, increased hydrologic

conductivity and connectivity,
and even potentially very low-friction

surfaces along ice boundaries.
That can all influence landslide process

and make low-angle hillslopes
more susceptible to landsliding.

You might notice that I don’t
have any error bars on this plot.

This is a plot just of the population of
landslides inventoried in the study area.

So it’s showing you the entire
distribution of our landslide population.

Now, with all of that together,
we concluded in this project

that there are two main drivers
of landslides in Denali National Park

and other sort of discontinuous
permafrost regions.

And those include rainfall and –
or, rain-induced landslides like you

might find anywhere else in the world,
where, if you get enough rain on

a steep enough hillslope, landslides
of different types can occur.

But the other type of landslide that
we saw were permafrost-induced,

or at least seasonally – seasonal
thaw-induced, where ice complications

and that sort of reduced cohesion
and friction of ice melt is allowing

landslides to occur on low-angle,
more susceptible hillslopes.

Now, to think a little bit more about
the process of how permafrost and

landslides interact, we wanted to
look a little bit more specifically

at some landslides and track
how they change through time.

So, again, we’re looking at the Denali
Park Road, but I’m now going to

highlight three specific landslides
that we looked at in detail at

different points along the road.
And this is what those three landslides

look like. They’re all
active layer detachments.

So, again, that means that the seasonally
thawed portion of the subsurface

is detaching from the permafrost
beneath and sliding along

the permafrost boundary.
They’re also all on south-facing

hillslopes of comparable slope
angles – less than 20 degrees.

And notably, two of these landslides –
the Stony Pass Slide, which you see

on the bottom left, and the
Ptarmigan Landslide, which you

see on the right there,
those are relatively young.

They occurred in summers
of 2015 and 2016, respectively.

This Eielson active layer
detachment on the top left,

that’s a much older landslide.
We don’t know the exact date.

But I’ll show you some of the details.
But we think that it probably initiated

sometime between the late 1960s
and the late 1990s, based on

the amount of vegetation that
has returned to that landslide.

That landslide initiated in
unconsolidated glacial outwash,

and currently, the model that we’re
working with estimates that the mean

decadal ground temperature in that
general area is just above zero degrees.

So currently, it’s probably
not classified as permafrost.

But we expect that,
several decades ago,

when this landslide initiated,
it probably was permafrost.

And I will also mention that
we were able to validate that

permafrost model in a few places.
And so what we did was we conducted

repeat terrestrial Lidar surveys in
the summer of 2017 and 2018

and then used the Wheaton
Geomorphic Change Detection tool

to look for topographic change
over the one-year study period.

What we found at this older active
layer detachment is that there was

pretty little significant change
in elevation at this site.

Here, where you see bright red colors,
that means that there’s elevation loss,

and blue means elevation gain.
But most of this map that you

see here is pretty gray.
There’s not a lot of significant

change beyond the error
of our survey equipment.

In contrast, the two younger
active layer detachments did show

pretty significant change over
the course of our study period.

So this is the Ptarmigan
active layer detachment.

It’s a much more elongate
active layer detachment.

But what you might notice is that
there are places of pretty significant

change here. You can ignore the
really bright red spot up near

the scarp of this landslide. We had a
limited amount of time with the laser

scanner in June of 2017, so we had to
include a little snow field in our survey.

Similarly to the previous landslide,
this occurred in unconsolidated

Quaternary sediments – in this case,
an old landslide deposit.

But here the current projection for
permafrost is that it is technically

permafrost – again, very close
to that zero-degree mark.

So kind of on the cusp of
whether or not you might actually,

in the field,
find permafrost or not.

In terms of what we found in this
elevation change of this landslide,

we saw some lowering within the
channel of the landslide by several

centimeters over the course of the
one year and some increase in elevation

near the toe of the landslide
and along some of the levees here,

which is probably related
to biologic activity.

So we did see pretty
significant change here.

And similarly, at the other
young landslide, we also saw

significant elevation change.
So this is the Stony Pass Landslide.

It initiated in the spring of 2015
in old glacial deposits.

And here, again, the modeled mean
decadal ground temperature is

a little bit below zero degrees,
indicating that permafrost is present.

And, in terms of what we saw
in this one-year study period,

there are a couple of notable patterns.
First of all, landslide elevations lowered

near the scarp with a few centimeters of
scarp retreat, where the scarp dropped

by about – up to a meter – half a meter
to a meter in a few pixels up here

in this area as the scarp
moves back upslope.

And we also saw some increase in
elevation at the toe of this landslide.

And this is another interesting
landslide to think about because,

as you can see in
the photo on the left,

this landslide is right at the
margin of the Denali Park Road.

So this is a classic example of
a location where the landslide

is interacting with human process.
And, notably, the Denali road crew –

we were able to chat with them
a little bit, and we know that they

had to remove some material
from the toe of this landslide in order to

clear the road on several occasions.
We don’t have an exact volume of

how much material they removed,
but we know that there were

at least a couple of occasions
where they had to remove some

material from that road between
2015 and our final survey in 2018.

So we can see that this
landslide is still moving,

and continuing to move,
for possibly a few different reasons.

One of those reasons might be
the fact that permafrost in this area

is continuing to thaw.
So what you see on this map –

we wanted to get a better sense of where
permafrost existed in this landslide.

So we actually used
a ground-penetrating radar

to survey permafrost in the subsurface.
And that’s what you see here

on the bottom right, where there is
a little ground-penetrating radar.

We used a 500-megahertz antenna
and drug it along in grid lines

across this landslide. And, on the top
photo here, this is an example of

what the radargrams look like.
And, on the left, you can see

there are places where there’s
a pretty nice continuous reflector.

That we interpreted as permafrost.
And there are other locations in the

survey where there is no clear reflector,
and we did not interpret that permafrost

existed within the measurement
depth of our radar.

And all of that is plotted here
on this map on the bottom left.

Where you see blue lines, those are
places where we saw a reflector and

considered permafrost to be present.
And the black lines are places where

we did not see evidence of
permafrost in the radar data.

Similarly, we validated some of
these points with a manual probe

that was 2.3 meters in length.
And where you see blue points,

those are places where we
measured a depth to ice.

We were able to kind of feel
what that ice – where that ice

was in the subsurface.
It’s got a very specific tactile sensation.

So we were measuring specific
depths to ice at these blue locations.

But, where there’s a white point,
that means we took a measurement

and did not measure permafrost
within that 2.3 meters’ depth.

So notably, we’re only looking at
shallow permafrost here.

It is possible that there is permafrost
at depth that we just can’t measure.

But what you might see is that there
is permafrost that exists around

the margins of the landslides
and on some of this

undisturbed portion of the hillslope.
But, within the landslide body,

there’s little or no
shallow permafrost present.

And that’s a really interesting
and unique observation.

And, to think about – to sort of validate
whether or not that matches what we

see in the temporal record, we actually
had access to some permafrost

monitoring temperature data
from a nearby location.

This is maybe half a kilometer, quarter
kilometer, away from our landslide site.

So it’s not exactly the same
characteristics in depth to permafrost,

but here you see a profile
of temperature with depth

at various months of the
year for both 2017 and 2018.

And, in the winter, of course,
the surface of the –

of the ground gets much colder.
In the summer, it gets much warmer.

But here, down at depth, there’s a
location where permafrost exists,

meaning that the ground temperature
is always below zero degrees.

At this site, it’s maybe about
3 or 4 meters’ depth.

But at our site
with the landslide,

we were looking more at,
like, a meter depth.

And what I’ll also note here is that,
in 2017, we had a very – a nice cold

winter typical of what we would expect.
But the winter between 2017 and

2018 was a lot less cold.
It was a relatively warm winter.

And that’s the winter between
our two ground surveys.

And so what that leads us to is this
sort of process of continued permafrost

thaw after a landslide occurs.
And there’s a particular mechanism that

we concluded was important here, and
that is the presence of windblown snow.

So this photo on the left is an example
of what we saw all across the national

park when we were out doing surveys
in the spring in 2017 and 2018, where

depressions, like landslide bodies, are
able to accumulate windblown snow.

And it might sound a little
counterintuitive, but in the winter,

when the air temperatures are
really cold, snow is a really effective

insulator that keeps the ground
temperatures from getting cold as well.

And, in contrast, in the summer, when
the air temperatures get warm again,

these landslides are in opposite form.
They have a lot less insulation with

the ground surface – or, with the –
with the air temperatures because

the landslides are characterized
by pretty sparse vegetation.

So it’s kind of the worst of both worlds,
where it’s – the ground can’t get

as cold in the winter, and they can
get quite a bit warmer in the summer,

based on that sort of landslide-related
insulation change.

And that means that what we have
here is a positive feedback loop where

permafrost thaw begets landslides,
but landslides continue to beget

additional permafrost thaw by altering
the thermal structure of the subsurface.

At this site in Stony Pass, and probably
other sites in sensitive permafrost areas,

we expect that this type of process
is likely to lead to ongoing

deformation of some of these
permafrost-induced landslides.

So that’s a lot of talk about ice.
I now want to transition to the second

part of what we wanted to talk
about today, which is more rain.

So there’s no better place, like I said,
to look at rain-induced landslides

than southeast Alaska. So I’m going to
take you to a little town called Sitka.

And, in terms of changes of rain with
climate change, it’s a little bit noisy.

It’s a little bit hard to predict.
But most models agree that climate

change is likely to lead to an
increase in cumulative rainfall

and rainfall intensity,
particularly at high-latitude areas.

So what you see on the left is a figure
from a paper that was looking at

integrated water vapor transport –
that’s the amount of moisture in

the air – and looking at pretty
positive changes, or increases,

in water vapor transport globally with
very high increases up at high latitudes.

Here in Alaska, you might see
a percent increase of 60 to 90% increase

in water content available
in the atmosphere.

Now, that also relates to some models
that project annual rainfall increases.

And what you see on the right here
are different time period projections

with the recent past and the now,
and then a couple of future projections

showing annual rainfall across Alaska,
where these darker green colors are

indicating that more
rainfall is likely to happen.

And that’s probably a combination
of both the sort of increase in water

available in the atmosphere
as well as the transition from

snowfall to rainfall
in some of these places.

In southeast Alaska, and across
a lot of the western United States,

the rainfall to pay attention to is a type
of storm called an atmospheric river.

And, if you’ve ever heard the term
“Pineapple Express,” that’s a type

of atmospheric river. And these are
storms where very high concentrations

of moisture-rich air come in sort of
current-like patterns off of the ocean.

And I’ll show you an example
of what that looks like in

a radar image in the next slide.
But notably, atmospheric rivers

initiate the vast majority of shallow
landslides in the western U.S.

And, in southeast Alaska,
and in Sitka specifically, these ARs

are responsible for the vast majority
of extreme rainfall events.

You’ll see here, in Sitka and other parts
of southeast Alaska, 80, 90, almost

100% of very extreme rainfall events
are brought in by atmospheric rivers.

Back in August of 2015, one of
these types of storms occurred.

Here’s that radar image
that I talked about.

So what you see here is a very
moisture-rich current of air coming off

of the Pacific Ocean and making
landfall right here in southeast Alaska.

And this was a particularly
extreme atmospheric river.

It was – I think, depending on
what time period you looked at,

like, hour rainfall or three-hour or
six-hour, it was on the order of,

like, 10-, 20-year return interval –
in some cases, depending on

the time interval,
even longer.

So this one storm initiated more
than 40 landslides near Sitka.

Most of them
were debris flows.

And one of those debris flows
caused three fatalities in town.

And that’s actually the landslide
that you see here in this photo.

This is called the
South Kramer Landslide.

And this is a pretty small community.
It’s a population of 8,000 or 9,000.

So that impact was very closely felt.
It’s a small community.

They’re very closely connected.
And the community themselves actually

came up with a lot of questions
about landslide hazards

and how to mitigate risk
and reached out to various

scientific organizations
to help them do that.

What I also want to note in this photo –
or, this Google Earth image,

on the bottom left here is that you
might be able to see the very steep

landscape that we’re talking about here.
So this is the town of Sitka.

This right here – this kind of main
area is considered downtown Sitka,

where the more densely
developed area is right at the base

of some really
steep hillslopes.

These are really classic coast glacial –
very steep glacial hillslopes,

as you would find anywhere
in southeast Alaska.

But the town is really nestled
against this really steep topography.

I don’t like going anywhere and
thinking about landslide process,

again, without some geologic
and geomorphic context.

So the first thing we did with this
project was go in and produce

a geomorphic map of the area
near Sitka. So the photo that you saw

on the previous slide was
looking at these hillslopes here.

Again, the exact units that I’ve
mapped out here aren’t too important,

but I wanted to show you that it’s
a pretty complex environment,

where there’s a history of glaciation,
as you can see with these big, beautiful

glacial valleys and some places
where there’s glacial sediment

preserved as well as
glacial topography.

And it’s a pretty active region
in terms of landslides and

debris flows, specifically.
Most of these debris flows

that I’m talking about are
initiated by small landslides.

And so notably, these little red fans
at the base of the steep hillslopes,

those are all debris fans with evidence
of old debris flows coming down out

of these debris channels and
depositing sediment right where

the steep valley wall meets
the shallow valley bottom.

If you zoom out a little bit and wanted
to think about landslides across larger

areas of southeast Alaska, we are
fortunate to have some partners

at the U.S. Forest Service who have
been compiling a landslide inventory

in the Tongass National Forest with
records dating back 80, 85, 90 years.

They’ve been using air photos.
And they’ve created a really rich

data set of landslide occurrence
across southeast Alaska.

So these two maps that you see here
are showing you the distribution of

landslides across our study area.
Sitka, again, is right here in this

little bay called the Sitka Sound
with the town of Sitka right here

where my little red
cursor is indicating.

Now, in addition to the glacial
history that I mentioned earlier,

there’s also a lovely volcanic history.
This little island just off of the coast

of Sitka is called Mount Kruzof –
sorry, Kruzof Island.

And, on Kruzof Island
is Mount Edgecumbe,

which is a recently active volcano.
And, about 11,000 years ago,

it erupted, depositing tephra across
large areas of this region of southeast

Alaska with Sitka, on Baranof Island,
and Chicagof Island to the north.

And these gray lines here are showing
you the isopach maps mapped by

Riehle and others showing tephra
thickness across these areas with

up to a meter being pretty
typical in Sitka itself.

Now, on top of the glacial
history and the volcanic history,

its also important to mention the
human history in southeast Alaska,

and notably,
the logging history.

The U.S. Forest Service has records
of timber harvest dating back to

about 1900. And those you see here
as green polygons on the map.

And so, with all of that context –
all of that – those different histories

of important process in southeast
Alaska, we can then think about the

landslides in context of the various
contributors to landslide process.

And all of these red dots on the
left map are landslides mapped

by the Forest Service. And then here,
on the right map, that’s translated into

a gridded landslide density with
5-kilometer-squared grid cells.

And there are a couple of
pretty notable patterns in terms of

where landslides are
initiating in this landscape.

First of all, landslides are very frequent,
very dense, on Kurzof Island

where these volcanic tephra
deposits are very thick.

That might not surprise you too much.
But we also see some correlation

between landslide occurrence
and timber harvest.

That might be partly a bias in mapping
landslides in logged areas, but there’s

also some pretty good evidence that
various folks – Swanston and others,

and – have linked timber harvest to
smaller, more frequent landslides.

Now, there’s one more pattern that
I want to draw your eye to on some of

these maps, and that is the pattern
of really high landslide density here

on the western coast of Baranof Island.
There are a lot of landslides occurring

right on that coastline. And, if we
think back to the process that’s

initiating most of these landslides –
the big atmospheric rivers that bring

relatively intense precipitation – that
might start to make a little bit of sense.

And, in particular, if you look at the
slope aspect of landslides across the

Tongass National Forest, they are
preferentially occurring on south-

and southwest-facing hillslopes.
And that’s what you see in this

little rose diagram here,
where the blue area is showing you

slope aspects of the entire Tongass
National Forest. And then the gray

distribution on top are showing
you where landslides occur.

And the south- and
southwest-facing aspect

is where a disproportionate number
of landslides are occurring.

So thinking about those big
atmospheric rivers coming off of

the Pacific Ocean, they make landfall
on, first of all, these mostly south-

and southwest-facing hillslopes.
That all starts to check out.

This is where those big
storms are coming from.

That’s a lot of great information about
where landslides are happening in

southeast Alaska, and Sitka specifically,
but it’s also important to start thinking

about when landslides might happen.
And our team on this project are,

of course, not the first people who have
wanted to identify when landslides are

likely to happen and even to think
about the storms that cause landslides.

And, in fact, we’re drawing a lot of
inspiration from previous works that

have looked at creating forecasting
thresholds for landslides based on

rainfall intensity and duration
or rainfall characteristics

and antecedent soil
saturation conditions.

So there’s a rich literature using
techniques like this to think about

predicting, or forecasting,
landslide-prone conditions.

And so we’re building
off of that rich literature.

And, to do so, what we need is some
data about storm occurrence in Sitka.

And, fortunately for us, there is
a pretty long data set of hourly

precipitation records available at the –
there’s a weather station at the

Sitka Airport with 20 years
of hourly rainfall records.

So that’s a really great start.
I’ve also mentioned that we have

a really nice inventory of landslides
that’s been curated by the

Forest Service. And, for some of those
landslides – not all, but for some of

them, we have really detailed timing
information, down to a few hours

in some cases, and even down to,
like, the half an hour in other cases.

So what you see on this map is some of
the monitoring equipment that existing

before our project started,
which includes that weather station.

There’s also a magnetic observatory
with some precipitation records, as well

as some stream gauges operated by a
couple of federal and state agencies.

Now, over the course of this project,
we have expanded that monitoring

network spatially by quite a lot.
And so what you see here on this

map now are some of the new
monitoring equipment that has been

installed in the last two years.
Not all of it is associated with the

NSF project that I am directly a part of,
but we have some partners at the

Alaska State Survey who have installed
a high-elevation meteorological station

on Harbor Mountain. And our lovely
partners at the U.S. Geologic Survey

have also installed a soil hydrologic
monitoring station nearby, also on

Harbor Mountain, where they’re
measuring pore pressure and soil

moisture, as well as gathering
rainfall and other meteorologic data.

In addition to those two state and
federal locations, our project has

also installed a network of
10 satellite-connected rain gauges.

They’re tipping-bucket rain gauges
spread out spatially across town

to start to look at spatial variability
in different storm patterns.

So we know that storms are pretty
variable in space, and the amount of

rainfall that one area receives might
not be the same as somewhere else.

And that is particularly true in Sitka,
where microclimates reign supreme.

And then, additionally, these three
light blue points, those are places

where we have also started
experimenting with a lower-cost

hydrologic monitoring equipment
to expand the network and

hopefully make hydrologic
monitoring a little bit more

accessible to small
communities like Sitka.

And that equipment
looks something like this.

It’s manufactured by
a corporation called Onset.

And, at these particular stations that
we’ve installed, we’ve included

rain gauges as well as soil moisture
sensors that measure electrical

conductivity to estimate volumetric
water content as well as a shallow

groundwater well where a little
pressure sensor at the bottom

of the well is a good estimate
of saturated pore pressures.

So we have these great new monitoring
data starting to collect records.

But, in the meantime –
oh, I should also give you

an example of what that data looks like.
This is what some of the output

of our lower-cost monitoring
equipment looks like.

You’ll see rainfall on the top,
and then the pressure at the bottom

of our groundwater well is here
in the black line on the bottom.

And what you can see is that, as it rains,
the well fills up, pressure increases,

and then it gradually declines.
That all might sound fairly intuitive,

but the great thing about this
data set is that, A, we are getting it

in near real time. So we can
see potentially hazardous conditions

in near real time. And we’re starting
to be able to quantify rates of change

and how much rain results in
what rate of increase of groundwater

saturation and then also how
long it takes to dry back out.

So, in the meantime, while we’re
starting to collect those longer-term

records at these various monitoring
stations, we still want to create some

useful tools for the community of Sitka
to work with, to forecast or predict

landslide conditions and potentially
use that information to mitigate risk.

So what we’ve done is we’ve started
with the rainfall data available from

the Sitka Airport and looked
at the distribution of rainfall intensity

and duration for
the 20 years of record.

And here what you see is
durations of 1, 3, and 6 hours.

I actually just use a moving window
through the entire period of record.

So any non-zero rainfall was
accounted for with total – 1-hour,

3-hour, and 6-hour rainfall intensities.
And then here, these triangles are

showing you the storm events
where we know with temporal

specificity that a landslide occurred.
And what you might see right away

is that those landslide events are
happening at the tail portions of the

distribution where extreme rainfall
intensities are happening.

And, in fact, with the possible outlier
of this little yellow landslide,

all of these events are occurring
during storms with a 1-year-or-greater

return interval, which translates to
a annual probability of 1 or less.

So these are the pretty extreme storms
that are causing landslides in Sitka.

And that’s really good and helpful to
know because, not only can we start to

pick out when to be concerned about
landslides, but we can also look at large

portions of this distribution where
no landslides have occurred during

the period of record, and we
can actually reduce a lot of anxiety

in communities like Sitka where,
right now, it rains a lot.

It rains pretty regularly.
And people don’t know if they

need to be worried or not.
And so they spend a lot of time

being really anxious and worried
about whether or not a particular storm

might cause a landslide.
So we can look at these large portions

of the distribution and give people
maybe a little bit more information

so that they don’t have to
be anxious quite so often.

Now, what we’re working on
right now is building off of

some of that information
about the rainfall intensity

to use multivariate models to
predict landslide occurrence.

What you see here are some examples
of logistic regression models that use

different time periods of rainfall to
try to separate out landslide events

from no-landslide events, where
landslides are up here at Category 1,

and no landslide is down
at the bottom as Category zero.

And we’ve played around with
a lot of different time variables.

I’ve just shown
you a few here.

But what you might see is that the
rainfall characteristic that separates

out landslide events the
best is this 3-hour rainfall.

And so that is a really good key
that this 3-hour rainfall is maybe

a triggering condition. And that’s
what it takes to start a landslide.

But we also know that pre-existing or
antecedent saturation conditions matter.

And building, again, off of some of the
previous literature looking at landslide

thresholds, we can actually compare
different triggering time scales with

different antecedent rainfall time scales.
And that’s what you see here in this

table, which is a little messy – this is
all still in works – where I’ve compared

some different antecedent intensity
time scales in hours to different

antecedent rainfall time scales in days.
And the colors and the numbers here

are a model selection criterion,
where low numbers indicate

that the model is a better fit
for our available data.

And these 3-hour rainfall variables,
as well as potentially a 1-day

antecedent rainfall condition,
provides the best model fit.

And that tells us a lot about the
time scales that rainfall is

influencing hydrologic process
and triggering landslides.

So, the next steps of this project are
working to incorporate some of that

real-time data and some
Weather Service forecast into

statistical models and then putting
all of that on a publicly available

risk mitigation website, where we’re
going to be providing some real-time

information that is geared towards
both community members,

who maybe don’t have a lot of
scientific background, and to some of

our partners at the Weather Service
so that we can create information

that allows people, whether or not
they have scientific background,

to have enough information with
context to make decisions and improve,

potentially, safety and reduce
risk in this community.

So, to pull all of that together,
a couple of conclusions.

The first part of this talk,
we talked about how less ice

is resulting in permafrost thaw.
And that can generate shallow

landslides, even on
relatively low-angle hillslopes.

And then, additionally, positive
feedback between landslide occurrence

and permafrost thaw are likely to
contribute to ongoing deformation

in some of these sensitive areas.
Now, secondly, climate change

is also likely to be changing
patterns of precipitation,

particularly in some of
these high-latitude areas.

So we’re starting to characterize the
fact that extreme bursts of precipitation

are really responsible for triggering
debris flows in this landscape.

And, using forecasting techniques
established by others, we’re building

tools that use rainfall records
and landslide inventories to

create landslide warning thresholds
for the community of Sitka.

And, with that,
thank you all so much.

I really appreciate being here today,
and I’m happy to take any questions.