Remote Sensing,
Год журнала:
2024,
Номер
16(15), С. 2760 - 2760
Опубликована: Июль 28, 2024
Aboveground
biomass
(AGB)
serves
as
a
crucial
indicator
of
the
carbon
sequestration
capacity
coastal
wetland
ecosystems.
Conducting
extensive
field
surveys
in
wetlands
is
both
time-consuming
and
labor-intensive.
Unmanned
aerial
vehicles
(UAVs)
satellite
remote
sensing
have
been
widely
utilized
to
estimate
regional
AGB.
However,
mixed
pixel
effects
hinder
precise
estimation
AGB,
while
high-spatial
resolution
UAVs
face
challenges
estimating
large-scale
To
fill
this
gap,
study
proposed
an
integrated
approach
for
AGB
using
sampling,
UAV,
Sentinel-2
data.
Firstly,
based
on
multispectral
data
from
vegetation
indices
were
computed
matched
with
sampling
develop
Field–UAV
model,
yielding
results
at
UAV
scale
(1
m).
Subsequently,
these
upscaled
(10
Vegetation
calculated
establish
UAV–Satellite
enabling
over
large
areas.
Our
findings
revealed
model
achieved
R2
value
0.58
0.74
scale,
significantly
outperforming
direct
modeling
(R2
=
−0.04).
The
densities
Xieqian
Bay,
Meishan
Hangzhou
Zhejiang
Province,
1440.27
g/m2,
1508.65
1545.11
respectively.
total
quantities
estimated
be
30,526.08
t,
34,219.97
296,382.91
This
underscores
potential
integrating
accurately
assessing
regions,
providing
valuable
support
conservation
management
Forests,
Год журнала:
2024,
Номер
15(10), С. 1696 - 1696
Опубликована: Сен. 25, 2024
Mangroves
play
a
crucial
ecological
and
economic
role
but
face
significant
threats,
particularly
on
Hainan
Island,
which
has
the
highest
mangrove
species
diversity
in
China.
Remote
sensing
AI
techniques
offer
potential
solutions
for
monitoring
these
ecosystems,
challenges
persist
due
to
difficult
access
field
sampling.
To
address
issues,
we
propose
novel
model
combining
Mangrove
Rough
Extraction
Decision
Tree
(MREDT)
Dynamic
Attention
Convolutional
Network
(DACN-M).
Initially,
used
drones
surveys
conduct
multiple
observations
Dongzhaigang
Nature
Reserve,
identifying
boundaries
of
mangroves.
Based
features,
constructed
MREDT
mitigate
failure
caused
by
light
instability,
simplifying
transfer
other
study
areas
without
requiring
annotated
samples
or
extensive
surveys.
Next,
developed
DACN-M
model,
refines
rough
extraction
features
from
incorporates
contextual
information
more
accurate
detection.
Experimental
results
demonstrate
that
our
proposed
method
effectively
differentiates
mangroves
vegetation,
achieving
F1
Scores
above
75%
IoU
values
greater
than
60%
across
six
areas.
In
conclusion,
not
only
accurately
identifies
monitors
distribution
also
offers
advantage
being
transferable
need
This
provides
robust
scalable
solution
protecting
preserving
critical
ecosystems
supports
effective
conservation
efforts
various
regions.
