Earth system science data,
Journal Year:
2023,
Volume and Issue:
15(1), P. 447 - 464
Published: Jan. 31, 2023
Abstract.
The
Svalbard
Archipelago
represents
the
northernmost
place
on
Earth
where
cryospheric
hazards,
such
as
thaw
slumps
(TSs)
and
thermo-erosion
gullies
(TEGs)
could
take
rapidly
develop
under
influence
of
climatic
variations.
permafrost
is
specifically
sensitive
to
occurring
warming,
therefore,
a
deeper
understanding
TSs
TEGs
necessary
understand
foresee
dynamics
behind
local
hazards'
occurrences
their
global
implications.
We
present
latest
update
two
polygonal
inventories
extent
recorded
across
Nordenskiöld
Land
(Svalbard
Archipelago),
over
surface
approximately
4000
km2.
This
area
was
chosen
because
it
most
concentrated
ice-free
and,
at
same
time,
current
human
settlements
are
concentrated.
were
created
through
visual
interpretation
high-resolution
aerial
photographs
part
our
ongoing
effort
toward
creating
pan-Arctic
repository
TEGs.
Overall,
we
mapped
562
908
TEGs,
from
which
separately
generated
susceptibility
maps
using
generalised
additive
model
(GAM)
approach,
assumption
that
manifest
Land,
according
Bernoulli
probability
distribution.
Once
modelling
results
validated,
patterns
combined
into
first
multi-hazard
map
area.
available
https://doi.org/10.1594/PANGAEA.945348
(Nicu
et
al.,
2022a)
https://doi.org/10.1594/PANGAEA.945395
2022b).
Journal of Rock Mechanics and Geotechnical Engineering,
Journal Year:
2022,
Volume and Issue:
15(5), P. 1127 - 1143
Published: Aug. 11, 2022
To
perform
landslide
susceptibility
prediction
(LSP),
it
is
important
to
select
appropriate
mapping
unit
and
landslide-related
conditioning
factors.
The
efficient
automatic
multi-scale
segmentation
(MSS)
method
proposed
by
the
authors
promotes
application
of
slope
units.
However,
LSP
modeling
based
on
these
units
has
not
been
performed.
Moreover,
heterogeneity
factors
in
neglected,
leading
incomplete
input
variables
modeling.
In
this
study,
extracted
MSS
are
used
construct
modeling,
represented
internal
variations
within
using
descriptive
statistics
features
mean,
standard
deviation
range.
Thus,
units-based
machine
learning
models
considering
(variant
slope-machine
learning)
proposed.
Chongyi
County
selected
as
case
study
divided
into
53,055
Fifteen
original
unit-based
expanded
38
through
their
variations.
Random
forest
(RF)
multi-layer
perceptron
(MLP)
variant
Slope-RF
Slope-MLP
models.
Meanwhile,
without
factors,
conventional
grid
(Grid-RF
MLP)
built
for
comparisons
performance
assessments.
Results
show
that
Slope-machine
have
higher
performances
than
models;
results
stronger
directivity
practical
Grid-machine
It
concluded
can
be
more
comprehensively
reflect
relationships
between
landslides.
research
reference
significance
land
use
prevention.
