The
urban
flood
of
rainstorm
has
posed
a
major
threat
to
human
life
and
property
in
recent
years,
which
seriously
affected
the
sustainable
development
society.
numerical
model
can
simulate
how
an
develops
moves,
then
guide
reduce
disaster’s
impact.
This
study
tight
coupling
named
STUFMS
based
on
Storm
Water
Management
Model
(SWMM)
TELEMAC-2D
for
simulation.
Theoretical
practical
cases
were
respectively
applied
verify
reasonability
accuracy
model.
theoretical
case
demonstrates
that
better
process
water
exchange
between
surface
rainwater
drainage
systems
while
suggests
change
level
flooded
area
simulated
by
are
basically
consistent
with
typical
historical
rainfall
events.
uncertainty
analysis
revealed
terrain
resolution
temporal
series
have
significant
impact
inundation
scopes.
Generally,
behaved
well
simulating
process,
offering
novel
tool
modeling
flood.
Engineering Applications of Computational Fluid Mechanics,
Journal Year:
2025,
Volume and Issue:
19(1)
Published: March 25, 2025
Efficient
and
accurate
flood
inundation
mapping
is
essential
for
risk
assessment,
emergency
response,
community
safety.
The
deep
learning-enabled
rapid
simulation
demonstrates
superior
computational
efficiency
compared
to
traditional
hydrodynamic
models.
However,
most
learning-based
models
currently
focus
on
predicting
the
maximum
water
depth
face
challenges
in
generalizing
rainfall
events
of
different
durations.
This
paper
proposes
a
fast
method
based
image
super-resolution,
utilizing
novel
DenseUNet
architecture
predict
velocity
temporal
events.
proposed
integrates
physical
catchment
characteristics
enhance
resolution
maps
generated
by
coarse-grid
model
using
deep-learning
model.
applied
rural-urban
Shenzhen
River
southern
China.
effectively
reproduces
test
against
fine-grid
model,
achieving
root
mean
square
errors
below
0.06
0.07
m/s,
respectively,
with
percentage
bias
within
±5%.
For
prediction,
exhibits
Nash-Sutcliffe
Pearson
correlation
coefficient
exceeding
0.99.
Similarly,
both
metrics
exceed
0.94.
outperforms
over
2800
times.
developed
this
study
regression
classification
performance
commonly
used
ResUNet
UNet
architectures.
robust
wide
range
super-resolution
scale
factors.
presents
an
efficient
surrogate
mapping,
providing
valuable
insights
applying
methods
simulation.
