i-manager’s Journal on Structural Engineering,
Год журнала:
2024,
Номер
13(3), С. 34 - 34
Опубликована: Янв. 1, 2024
This
study
presents
a
replicable,
cost-efficient
method
for
estimating
forest
biomass
critical
sustainable
structural
material
sourcing
using
Sentinel-2
satellite
imagery
and
Gaussian
Process
Regression.
A
simplified
inventory
method,
coupled
with
spectral
data
in
the
visible
to
mid-infrared
bands,
enables
accurate
quantification
across
diverse
structures
Mediterranean
climates.
Compared
traditional
LiDAR-based
techniques,
this
approach
offers
faster,
lower-cost
deployment
without
significant
trade-off
accuracy,
making
it
suitable
applications
construction
timber
forecasting,
infrastructure
planning,
environmental
assessments.
The
has
been
validated
several
types
is
packaged
freely
accessible
programming
tool
direct
integration
into
engineering
planning
workflows.
Remote Sensing,
Год журнала:
2025,
Номер
17(10), С. 1759 - 1759
Опубликована: Май 18, 2025
Climate
change
is
threatening
the
sustainability
of
crop
yields
due
to
an
increasing
frequency
extreme
weather
conditions,
requiring
timely
agricultural
monitoring.
Remote
sensing
facilitates
consistent
and
continuous
monitoring
field
crops.
This
study
aimed
estimate
alfalfa
height
through
satellite
images
machine
learning
methods
within
Google
Earth
Engine
(GEE)
Python
API.
Ground
measurements
for
this
were
collected
over
three
years
in
four
Canadian
provinces.
We
utilized
Sentinel-2
data
obtain
imagery
corresponding
same
timeframe
location
as
ground
measurements.
Three
algorithms
employed
plant
from
images:
random
forest
(RF),
support
vector
regression
(SVR),
gradient
boosting
(XGB).
The
efficacy
these
has
been
assessed
compared.
Several
widely
used
vegetation
indices,
instance
normalized
difference
index
(NDVI),
enhanced
(EVI),
red-edge
(NDRE),
selected
study.
RF
feature
importance
was
determine
ranking
features
most
least
significant.
selection
strategies
compared
with
situation
where
all
are
used.
demonstrated
that
XGB
surpassed
SVR
when
assessing
test
performance.
Our
findings
showed
could
predict
R2
0.79
a
mean
absolute
error
(MAE)
around
4
cm
indicated
exhibited
lowest
accuracy
among
tested,
0.69
MAE
4.63
cm.
analysis
important
red
edge
(NDRE)
water
(NDWI)
variables
determining
height.
results
also
using
strategies,
can
be
estimated
comparably
high
accuracy.
Given
models
fully
trained
developed
(v.
3.10),
they
readily
implemented
decision
system
deliver
near
real-time
estimations
farmers
throughout
Canada.
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,
Год журнала:
2024,
Номер
17, С. 12969 - 12982
Опубликована: Янв. 1, 2024
Gross
primary
production
(GPP)
measures
the
amount
of
carbon
fixed
by
plants
and,
thus,
plays
a
significant
role
in
terrestrial
cycle
and
global
food
security,
especially
context
climate
change
neutrality.
Currently,
all-sky
high-resolution
(<100
m)
GPP
is
increasingly
needed
for
better
understanding
food–carbon–water–energy
nexus.
However,
previous
studies
usually
used
optical
satellites
to
estimate
clear-sky
at
kilometer-scale
resolution.
Due
missing
estimates
under
cloudy-sky
conditions,
monitoring
spatio–temporal
changes
from
would
suffer
some
uncertainties.
Moreover,
one
issue
that
they
only
satellite
images
or
environmental
data
rather
than
jointly
integrating
them
biome
types.
To
address
these
challenges,
this
study
attempts
use
active
microwave
Sentinel-1
synthetic
aperture
radar
(SAR)
10
m
resolution
GPP.
measurements
across
nine
types
North
America
were
employed
develop
SAR-based
model.
Meanwhile,
an
optical-based
model
with
Landsat-8
was
also
proposed
comparison.
The
results
revealed
that,
first,
SAR
can
be
utilized
By
images,
data,
types,
optimal
showed
high
accuracy
estimating
daily
coefficient
determination
(R
2
)
=
0.764,
root-mean-square
error
(RMSE)
1.976
gC/m
/d,
mean
absolute
(MAE)
1.308
/d.
Second,
had
reasonable
validation
0.809,
RMSE
1.762
MAE
1.165
/d).
Third,
contributed
more
model,
while
contribution
higher
Fourth,
performance
GPP,
two
models
consistency
0.730
1.858
/d)
together.
Therefore,
demonstrated
provides
important
source
advancing
our
cycle,
change.
Research Square (Research Square),
Год журнала:
2024,
Номер
unknown
Опубликована: Дек. 2, 2024
Abstract
Arkansas'
subtropical
climate
nurtures
extensive
forested
regions,
particularly
within
the
Ozark-
St.
Francis
and
Ouachita
National
Forests.
Despite
this,
state
lacks
an
up-to-date,
high-resolution
map
detailing
distribution
of
tree
species
its
forests.
This
study
harnesses
power
machine
learning,
specifically
Random
Forest
(RF),
Gradient
Tree
Boosting
(GTB),
Support
Vector
Machine
(SVM),
K-Nearest
Neighbors
(K-NN)
classifiers,
Google
Earth
Engine
(GEE)
framework.
These
classifiers
are
applied
to
classify
in
forests
by
integrating
data
from
various
sources,
including
Sentinel-1/-2,
Landsat-8,
Agriculture
Imagery
Program
(NAIP).
The
evaluates
classification
accuracy
single-sensor
images
against
fused
composites,
revealing
that
Landsat-8
Sentinel-1
achieve
highest
validation
at
0.8875.
is
closely
followed
which
yield
accuracies
0.8863
0.8859,
respectively.
Among
RF
demonstrates
accuracy,
GTB,
K-NN,
SVM
when
images.
incorporates
Shapley
Additive
Explanations
(SHAP)
elucidate
feature
importance
introduces
a
weighted
ensemble
method,
resulting
remarkably
accurate
with
score
0.9772.
research
highlights
efficacy
combining
learning
algorithms
fusing
satellite
significantly
enhance
accuracy.
Moreover,
capitalizes
on
explainable
AI
(XAI)
principles
leverages
cloud
computing
capabilities
GEE
create
more
precise,
cover
regional
scale.
i-manager’s Journal on Structural Engineering,
Год журнала:
2024,
Номер
13(3), С. 34 - 34
Опубликована: Янв. 1, 2024
This
study
presents
a
replicable,
cost-efficient
method
for
estimating
forest
biomass
critical
sustainable
structural
material
sourcing
using
Sentinel-2
satellite
imagery
and
Gaussian
Process
Regression.
A
simplified
inventory
method,
coupled
with
spectral
data
in
the
visible
to
mid-infrared
bands,
enables
accurate
quantification
across
diverse
structures
Mediterranean
climates.
Compared
traditional
LiDAR-based
techniques,
this
approach
offers
faster,
lower-cost
deployment
without
significant
trade-off
accuracy,
making
it
suitable
applications
construction
timber
forecasting,
infrastructure
planning,
environmental
assessments.
The
has
been
validated
several
types
is
packaged
freely
accessible
programming
tool
direct
integration
into
engineering
planning
workflows.