International Journal of Digital Earth,
Journal Year:
2023,
Volume and Issue:
16(1), P. 593 - 619
Published: March 1, 2023
Drainage
pattern
recognition
is
crucial
for
geospatial
understanding
and
hydrologic
modelling.
Currently,
drainage
methods
employ
geometric
measures
of
overall
local
features
river
networks
but
lack
basin
unit
shape
features,
so
that
potential
correlations
between
segments
are
usually
ignored,
resulting
in
poor
results.
In
order
to
overcome
this
problem,
paper
proposes
a
supervised
graph
neural
network
method
considers
the
networks.
First,
based
on
hierarchy
networks,
confluence
angle
units,
multiple
classification
extracted.
Then,
typical
samples
from
multi-scale
NSDI
USGS
databases
used
complete
training,
validation
testing
steps.
Experimental
results
show
indexes
proposed
can
describe
characteristics
different
patterns.
The
effectively
sample
adjacent
segments,
flexibly
transfer
associated
among
segment
neighbours,
aggregate
deeper
thus
improving
accuracy
relative
other
reliably
distinguishing
Environmental Earth Sciences,
Journal Year:
2024,
Volume and Issue:
83(14)
Published: July 1, 2024
Abstract
Flash
floods
stand
as
a
substantial
peril
linked
to
climate
change,
imposing
severe
menace
both
human
existence
and
built
structures.
This
study
aims
assess
compare
the
effectiveness
of
four
distinct
machine
learning
(ML)
methodologies
in
production
flood
susceptibility
maps
(FSMs)
Ibaraki
prefecture,
Japan.
Additionally,
investigation
examine
influence
excluding
plan
profile
curvature
factors
on
accuracy
resulting
maps.
The
dataset
comprised
224
spots,
consisting
112
flooded
non-flooded
locations,
11
environmental
factors.
models
were
trained
using
70%
dataset,
while
remaining
30%
was
utilized
for
model
evaluation
ROC
curve
method.
results
indicated
that
ANN-MLP
SVR
achieved
notable
accuracy,
with
area
under
values
95.23%
95.83%
respectively.
An
intriguing
observation
made
when
excluded,
it
led
an
improvement
model,
96.7%.
Furthermore,
generated
FSMs
classified
into
five
hazard
levels.
northern
region
predominantly
exhibited
very
low
levels,
areas
located
southern
region,
closer
main
streams,
demonstrated
considerably
higher
levels
categorized
high
high.
Ultimately,
this
marks
novel
endeavor
investigate
impact
factor
precision
algorithms
creation
FSMs,
which
serve
fundamental
tools
subsequent
investigations.
