Forests,
Journal Year:
2024,
Volume and Issue:
15(12), P. 2222 - 2222
Published: Dec. 17, 2024
For
tropical
rainforest
regions
with
dense
vegetation
cover,
the
development
of
effective
large-scale
soil
mapping
methods
is
crucial
to
improve
management
practices
replace
time-consuming
and
laborious
conventional
approaches.
While
machine
learning
(ML)
algorithms
demonstrate
superior
predictability
properties
over
linear
models,
their
practical
automated
application
for
predicting
using
remote
sensing
data
requires
further
assessment.
Therefore,
this
study
aims
integrate
Unmanned
Aerial
Vehicles
(UAVs)-based
hyperspectral
images
Light
Detection
Ranging
(LiDAR)
points
predict
indirectly
in
two
mountains
(Diaoluo
Limu)
Hainan
Province,
China.
A
total
175
features,
including
texture
indices,
forest
parameters,
were
extracted
from
sites.
Six
ML
Partial
Least
Squares
Regression
(PLSR),
Random
Forest
(RF),
Adaptive
Boosting
(AdaBoost),
Gradient
Decision
Trees
(GBDT),
Extreme
(XGBoost),
Multilayer
Perceptron
(MLP),
constructed
properties,
acidity
(pH),
nitrogen
(TN),
organic
carbon
(SOC),
phosphorus
(TP).
To
enhance
model
performance,
a
Bayesian
optimization
algorithm
(BOA)
was
introduced
obtain
optimal
hyperparameters.
The
results
showed
that
compared
default
parameter
tuning
method,
BOA
always
improved
models’
performances
achieving
average
R2
improvements
202.93%,
121.48%,
8.90%,
38.41%
pH,
SOC,
TN,
TP,
respectively.
In
general,
effectively
determined
complex
interactions
between
hyperparameters
prediction
leading
an
performance
models.
GBDT
generally
outperformed
other
pH
while
XGBoost
achieved
highest
accuracy
SOC
TP.
fusion
LiDAR
resulted
better
each
single
source.
models
utilizing
integration
features
derived
those
relying
on
one
summary,
highlights
promising
combination
UAV-based
advance
digital
property
forested
areas,
monitoring.
Geoderma,
Journal Year:
2024,
Volume and Issue:
442, P. 116798 - 116798
Published: Feb. 1, 2024
Soil
pH
is
one
of
the
critical
indicators
soil
quality.
A
fine
resolution
map
urgently
required
to
address
practical
issues
agricultural
production,
environmental
protection,
and
ecosystem
functioning,
which
often
fall
short
meeting
demands
for
local
applications.
To
fill
this
gap,
we
used
data
from
an
extensive
survey
13,424
surface
samples
(0–0.2
m)
across
cropland
Jiangxi
Province
in
Southern
China.
Using
digital
mapping
techniques
with
46
covariates,
produced
a
30
m
topsoil
We
integrate
different
variable
selection
algorithms
machine
learning
methods.
Our
results
indicate
Random
Forest
covariates
selected
by
recursive
feature
had
best
performance
r
0.583
RMSE
0.41.
The
prediction
interval
coverage
probability
our
was
0.92,
indicating
low
estimated
uncertainty.
Climate
identified
as
most
predicting
contribution
37.42
%,
followed
properties
(29.09
%),
management
(21.86
parent
material
(6.22
biota
(5.39
%)
factors.
mean
5.21,
great
pressure
acidification
region.
high
values
were
mainly
distributed
Northern,
Western,
Eastern
parts
region
while
majorly
located
central
part.
Compared
past
surveys
1980
s,
there
no
significant
change
surveyed
can
provide
important
implications
guidance
decisions
on
heavy
metal
pollution
remediation,
precision
agriculture,
prevention
acidification.
Geoderma,
Journal Year:
2024,
Volume and Issue:
444, P. 116855 - 116855
Published: March 14, 2024
The
accumulation
of
soil
salt
becomes
a
worldwide
widespread
phenomenon,
being
major
threat
to
global
production.
As
an
environmental
stress,
salinity
can
reduce
the
vegetation
photosynthetic
activity.
Solar-induced
chlorophyll
fluorescence
(SIF)
is
electromagnetic
signal
actively
released
by
during
photosynthesis.
SIF
not
only
capture
lower
activity
due
stress
promptly,
but
also
less
affected
atmosphere
and
background.
However,
ability
observation
detect
remains
unclear.
Here,
we
use
standardized
solar-induced
iluorescence
index
(SIFI)
from
time
series
(2000
∼
2020)
OCO-2
based
product
(GOSIF)
develop
model.
results
show
that:
identify
(EC
≥
2
4
dS
m−1)
class
scale.
SIFI
calculated
at
May
August
(hereafter
SIFI5-8)
optimal
sensitivity
indices
for
rainfed
cropland,
herbaceous
cover,
irrigated
shrubland,
grassland.
SIFI10-11
forest
sparse
vegetation;
(2)
By
comparison,
ovrerall
classification
accuracy
predicted
above
70
%.
order
cropland
>
bare
area
grassland
shrubland
cover
(3)
During
least
three-quarters
period
2000
2020,
was
4.9
Mkm2;
(4)
annual
change
rate
content
generally
between
−0.05
0.05
m−1
yr−1.
Soil
in
South
Africa
West
Asia
increased
greatly
with
0.02
0.03
These
demonstrate
estimate
salinity,
providing
new
perspective
explaining
evaluating
variation.
Geoderma,
Journal Year:
2023,
Volume and Issue:
440, P. 116738 - 116738
Published: Dec. 1, 2023
Soil
salinization
is
one
of
the
main
factors
contributing
to
land
degradation,
affecting
ecological
equilibrium,
environmental
health,
and
sustainable
development
agriculture.
Due
spatial
temporal
heterogeneity
soil
properties
conditions
in
a
large-scale
region,
monitoring
accuracy
can
be
challenging.
This
study
investigated
whether
classification
diverse
crop
types
on
time
series
improve
prediction
regional
salinity
levels.
Specifically,
we
evaluated
changes
salt
content
(SSC)
under
vegetation
cover
over
Hetao
Irrigation
District
(HID)
using
multi-phase
Sentinel-2
imagery
ground-truth
data
collected
from
June
September
2021
2022.
Focused
sunflower
maize
fields,
this
analyzed
impact
classifying
these
two
examining
four
distinct
SSC
estimation.
Five
indices
were
selected
as
characteristic
parameters
pool
17
(VIs)
13
(SIs)
derived
satellite
images.
Moreover,
three
machine
learning
algorithms
used
establish
estimation
models.
The
findings
underscored
efficacy
considering
different
enhancing
response
sensitivity
spectral
improving
modeling
accuracy.
Among
indices,
VIs
made
more
contributions
model
than
SIs,
achieving
highest
coefficient
determination
(R2)
0.71.
artificial
neural
networks
algorithm
outperformed
other
terms
stability,
yielding
an
optimal
R2
0.72
Root
Mean
Square
Error
(RMSE)
0.15%.
proposed
mapping
approach
that
considers
various
series,
offering
valuable
insights
for
accurately
assessing
salinization,
guiding
strategies
its
prevention
remediation.
Remote Sensing,
Journal Year:
2024,
Volume and Issue:
16(24), P. 4812 - 4812
Published: Dec. 23, 2024
Soil
salinization
is
a
significant
global
ecological
issue
that
leads
to
soil
degradation
and
recognized
as
one
of
the
primary
factors
hindering
sustainable
development
irrigated
farmlands
deserts.
The
integration
remote
sensing
(RS)
machine
learning
algorithms
increasingly
employed
deliver
cost-effective,
time-efficient,
spatially
resolved,
accurately
mapped,
uncertainty-quantified
salinity
information.
We
reviewed
articles
published
between
January
2016
December
2023
on
sensing-based
prediction
synthesized
latest
research
advancements
in
terms
innovation
points,
data,
methodologies,
variable
importance,
trends,
current
challenges,
potential
future
directions.
Our
observations
indicate
innovations
this
field
focus
detection
depth,
iterations
data
conversion
methods,
application
newly
developed
sensors.
Statistical
analysis
reveals
Landsat
most
frequently
utilized
sensor
these
studies.
Furthermore,
deep
remains
underexplored.
ranking
accuracy
across
various
study
areas
follows:
lake
wetland
(R2
=
0.81)
>
oasis
0.76)
coastal
zone
0.74)
farmland
0.71).
also
examined
relationship
metadata
accuracy:
(1)
Validation
accuracy,
sample
size,
number
variables,
mean
exhibited
some
correlation
with
modeling
while
sampling
type,
time,
maximum
did
not
influence
accuracy.
(2)
Across
broad
range
scales,
large
sizes
may
lead
error
accumulation,
which
associated
geographic
diversity
area.
(3)
inclusion
additional
environmental
variables
does
necessarily
enhance
(4)
Modeling
improves
when
area
exceeds
30
dS/m.
Topography,
vegetation,
temperature
are
relatively
covariates.
Over
past
years,
affected
by
has
been
increasing.
To
further
we
provide
several
suggestions
for
challenges
directions
research.
While
sole
solution,
it
provides
unique
advantages
salinity-related
studies
at
both
regional
scales.
Sustainability,
Journal Year:
2023,
Volume and Issue:
15(18), P. 13996 - 13996
Published: Sept. 21, 2023
Soil
salinization
is
a
serious
global
issue;
by
2050,
without
intervention,
50%
of
the
cultivated
land
area
will
be
affected
salinization.
Therefore,
estimating
and
predicting
future
soil
salinity
crucial
for
preventing
investigating
potential
arable
resources.
In
this
study,
several
machine
learning
methods
(random
forest
(RF),
Light
Gradient
Boosting
Machine
(LightGBM),
Decision
Tree
(GBDT),
eXtreme
(XGBoost))
were
used
to
estimate
in
Werigan–Kuqa
River
Delta
Oasis
region
China
from
2001
2021.
The
cellular
automata
(CA)–Markov
model
was
predict
types
2020
2050.
LightGBM
method
exhibited
highest
accuracy,
overall
prediction
accuracy
had
following
order:
>
RF
GBRT
XGBoost.
Moderately
saline,
severely
saline
soils
dominant
east
south
research
area,
while
non-saline
mildly
widely
distributed
inner
oasis
area.
A
marked
decreasing
trend
salt
content
observed
2021,
with
rate
4.28
g/kg·10
a−1.
primary
change
included
conversion
soil.
generalized
difference
vegetation
index
(51%),
Bio
(30%),
temperature
drought
(27%)
greatest
influence,
followed
variables
associated
attributes
(soil
organic
carbon
stock)
terrain
(topographic
wetness
index,
slope,
aspect,
curvature,
topographic
relief
index).
Overall,
CA–Markov
simulation
resulted
suitable
(kappa
=
0.6736).
Furthermore,
areas
increase
other
levels
continue
decrease
From
2046
numerous
converted
These
results
can
provide
support
control,
agricultural
production,
investigations
future.
gradual
decline
past
20
years
may
have
large-scale
reclamation,
which
has
turned
alkali
into
also
related
effective
measures
taken
local
government
control
Remote Sensing,
Journal Year:
2024,
Volume and Issue:
16(4), P. 642 - 642
Published: Feb. 9, 2024
Remote
sensing
(RS)
technology
can
rapidly
obtain
spatial
distribution
information
on
soil
salinization.
However,
(1)
the
scale
effects
resulting
from
mismatch
between
ground-based
“point”
salinity
data
and
remote
pixel-based
“spatial”
often
limit
accuracy
of
monitoring
salinity,
(2)
same
RS
model
usually
provides
inconsistent
or
sometimes
conflicting
explanations
for
different
data.
Therefore,
based
Landsat
8
imagery
synchronously
collected
ground-sampling
two
typical
study
regions
(denoted
as
N
S,
respectively)
Yichang
Irrigation
Area
in
Hetao
District
May
2013,
this
used
geostatistical
methods
to
“relative
truth
values”
corresponding
pixel
scale.
Additionally,
multispectral
data,
14
indices
were
constructed.
Subsequently,
Correlation-based
Feature
Selection
(CFS)
method
was
select
sensitive
features,
a
strategy
similar
concept
ensemble
learning
(EL)
adopted
integrate
single-feature-sensitive
Bayesian
classification
(BC)
order
construct
an
salinization
(Nonsaline,
Slightly
saline,
Moderately
Strongly
Solonchak).
The
research
results
indicated
that
exhibits
moderate
strong
variability
within
30
m
scale,
heterogeneity
needs
be
considered
when
developing
models;
theoretical
models
variance
functions
S
conform
exponential
spherical
model,
with
R2
values
0.817
0.967,
respectively,
indicating
good
fit
characteristics
suitability
Kriging
interpolation;
(3)
compared
single-feature
BC
identification
constructed
using
EL
demonstrated
better
potential
robustness
effectiveness.
International Journal of Applied Earth Observation and Geoinformation,
Journal Year:
2024,
Volume and Issue:
132, P. 104081 - 104081
Published: Aug. 1, 2024
Understanding
soil
moisture
dynamics
is
crucial
for
crop
growth.
The
digital
mapping
of
field
distribution
provides
valuable
information
agricultural
water
management.
optical
satellite
data
fine
scale
a
region.
However,
these
are
greatly
limited
due
to
cloud
contamination
and
revisit
period.
Despite
the
reported
beneficial
effects
spatiotemporal
fusion
methods,
accurate
estimates
high-resolution
through
still
unclear,
particularly
when
using
Sentinel-2/3
images.
This
study
introduces
new
estimation
framework
integrating
spatio-temporal
spectral
from
images
machine
learning
algorithm,and
thus
provide
spatiotemporally
continuous
estimation.
includes
four
methods
(ESTARRFM,
Fit-FC,
FSDAF
STFMF)
models
(PLSR,
SVM,
RF
GBRT).
feasibility
was
validated
in
Hetao
Irrigation
Area
Inner
Mongolia,
China.
results
showed
that
fused
image
generated
by
Fit-FC
visually
closest
true
image,
followed
ESTARFM,
FSDAF,
STFMF.
fusion-machine
provided
reliable
multi-layer
(0
∼
20,
20
40
60
cm)
irrigation
area.
dense
time
series
facilitated
detection
events
irrigated
farmland.
Our
findings
highlighted
effectiveness
providing
daily
monitoring
farmland
on
large
scale.
These
high
spatial–temporal
resolution
growth
resource
management,
contributing
further
expanding
application
remote
sensing
precision
agriculture.