Soil
erodibility
(K)
refers
to
the
resistance
of
soil
erosion
and
is
an
important
factor
in
forecasting
erosion.
The
accuracy
K
determines
predictions
loss
effective
deployment
measures
for
conserving
water.
China
has
no
high-resolution
map
distribution
at
national
scale
due
uncertainty
obtaining
limitation
complex
diverse
topographic
conditions.
We
used
most
recent
soil-sampling
data
(4710
profile
points),
calculated
point-scale
using
erosion-productivity
impact
calculator
(EPIC),
random-forest
method
predict
across
by
combining
soil-landscape
relationships
environmental
variables
determined
theory
formation.
mean
predicted
was
0.035
t
ha
h
ha-1
MJ-1
mm-1,
with
a
range
from
0.015
0.061.
small
Northwest
sandstorm
region
Qinghai-Tibet
Plateau
(means
0.032
0.031,
respectively)
large
Loess
0.040
0.042,
respectively),
which
were
different
natural
geographic
conditions
soil-forming
environments
each
region.
highly
accurate,
10-fold
cross-validation
model
0.49,
root
square
error
(RMSE)
0.0077,
absolute
(MAE)
0.0059.
represented
feature
details
spatial
continuity
better
than
traditional
polygon-linking
had
higher
spatial-modeling
did
ordinary-kriging
(R2random
forest
=
0.49
>
R2ordinary
kriging
0.42).
Elevation,
solar
radiation,
wind
speed,
surface
reflectance
primary
affecting
K,
increase
(%IncMSE)
32.98,
30.69,
30.03,
28.33%,
respectively,
indicating
influence
factors
on
evolution
formation
current
physicochemical
properties.
This
study
provides
first
national-scale
China,
can
provide
basis
predicting
regional
planning
conservation
The
Great
Wall
of
China,
one
the
most
emblematic
and
historical
structures
built
by
humankind
throughout
all
history,
is
suffering
from
rain
wind
erosion
largely
colonized
biocrusts.
However,
how
biocrusts
influence
conservation
longevity
this
structure
virtually
unknown.
Here,
we
conducted
an
extensive
biocrust
survey
across
found
that
cover
67%
studied
sections.
Biocrusts
enhance
mechanical
stability
reduce
erodibility
Wall.
Compared
with
bare
rammed
earth,
biocrust-covered
sections
exhibited
reduced
porosity,
water-holding
capacity,
erodibility,
salinity
2
to
48%,
while
increasing
compressive
strength,
penetration
resistance,
shear
aggregate
37
321%.
We
further
protective
function
mainly
depended
on
features,
climatic
conditions,
types.
Our
work
highlights
fundamental
importance
as
a
nature-based
intervention
Wall,
protecting
monumental
heritage
erosion.
The Science of The Total Environment,
Год журнала:
2023,
Номер
908, С. 168249 - 168249
Опубликована: Окт. 31, 2023
USLE-type
models
are
widely
used
to
estimate
average
annual
soil
loss
at
large
scales,
with
the
erodibility
factor
(K)
being
sole
component
that
accounts
for
soil's
susceptibility
erosion.
The
includes
information
on
permeability
in
equation,
however,
most
definitions
of
K
consider
hydrological
influence
only
very
crudely
and
indirectly.
Thus,
direct
impact
surface
runoff
infiltration
drainage
erosion
is
largely
neglected.
objective
this
study
incorporate
hydraulic
properties
map
by
merging
available
global-scale
measured
saturated
conductivity
(Ksat)
data
texture
organic
carbon
into
a
modified
factor.
To
achieve
this,
Wischmeier
Smith
(1978)
texture-
permeability-based
equation
(KWischmeier
factor)
was
include
Ksat,
called
Kksat
Using
Random
Forest
machine
learning
algorithm,
KWischmeier
were
each
correlated
remote
sensing
covariates
spatial
extrapolation
two
independent
maps
1
km
resolution.
We
noted
clear
decrease
mean
value
(0.023
t
ha
h
ha-1
MJ-1
mm-1)
compared
(0.027
mm-1).
reduction
values
pronounced
tropical
regions
reflecting
difference
(e.g.,
clay
iron),
whereas
other
climate
showed
relatively
minor
changes
comparison
as
well
recent
global
modeling
Borrelli
et
al.
(2017)
(KGloSEM
maps).
As
many
studies
discussed
an
overall
overestimation
(R)USLE
based
rates
measurements,
might
improve
modeled
right
direction.
marks
important
initial
step
integrating
can
prove
their
significance
future
studies.
Geoderma,
Год журнала:
2024,
Номер
444, С. 116853 - 116853
Опубликована: Март 12, 2024
Soil
erodibility
(K)
is
the
intrinsic
susceptibility
of
a
soil
to
water
erosion.
Currently,
its
detailed
and
accurate
spatial
distribution
information
especially
over
large
areas
urgently
required
for
national
regional
erosion
assessment
conservation
decision
making.
This
study
combined
pedotransfer
function
with
digital
mapping
techniques
develop
high-resolution
map
across
China.
The
First,
based
on
recent
survey,
we
adopted
erosion-productivity
impact
calculator
(EPIC)
calculate
values
at
4710
sampling
points.
Then,
caclulated
points,
used
five
including
polygon
linking
(PL),
ordinary
kriging
(OK),
Cubist,
extreme
gradient
boosting
(XGBoost),
random
forests
(RF)
generate
erodibility.
three
latter
machine
learning
modeled
quantitative
relationships
between
set
environmental
covariates.
results
showed
that
methods
exhibited
much
more
details
than
PL
OK
did.
Among
RF
achieved
highest
accuracy
R2
0.49
RMSE
0.0077
t
ha
h
ha−1
MJ−1
mm−1
10-fold
cross-validation.
Spatial
uncertainty
analysis
predictions
high
uncerntainty
occurred
in
northwestern
China
low
center
southeast.
We
found
topographical
climatic
variables
are
major
factors
indirectly
controlling
variation
while
particle
composition
SOC
contents
directly
influence
variation.
International Soil and Water Conservation Research,
Год журнала:
2024,
Номер
13(1), С. 15 - 26
Опубликована: Май 29, 2024
Soil
erodibility
is
a
measure
of
soil
susceptibility
to
water
erosion
and
serves
as
an
essential
element,
also
known
the
K-factor,
in
empirical
prediction
models,
such
USLE,
RUSLE,
CSLE.
The
currently
available
map
K-factor
for
China
was
generated
based
on
conventional
polygon
linkage
method
species
survey
1980s.
For
update,
investigation
4,262
samples
from
series
2010s
random
forest
regression
model
were
used
generate
new
China.
A
digital
at
250
m
spatial
resolution
by
calculating
K
values
points
training
data
using
environmental
information
predictive
variables.
comparison
results
between
maps
show
that
there
has
been
decreasing
trend
recent
decades.
value
decrease
mainly
due
update
(the
mean
changed
0.03193
t·ha·h/(MJ·mm·ha)
database
0.02988
series)
less
influenced
replacement
mapping
methods
0.03197
forest).
This
study
quantified
sources
change
previous
updated
national
demonstrated
values,
which
consistent
with
increasing
organic
matter
improved
ecological
environment
Remote Sensing,
Год журнала:
2024,
Номер
16(16), С. 3017 - 3017
Опубликована: Авг. 17, 2024
Soil
erodibility
(K)
refers
to
the
inherent
ability
of
soil
withstand
erosion.
Accurate
estimation
and
spatial
prediction
K
values
are
vital
for
assessing
erosion
managing
land
resources.
However,
as
most
K-value
models
empirical,
they
suffer
from
significant
extrapolation
uncertainty,
traditional
studies
on
focusing
individual
empirical
have
neglected
explore
pattern
differences
between
various
models.
This
work
proposed
a
universal
framework
selecting
an
optimal
soil-erodibility
map
using
enhanced
by
machine
learning.
Specifically,
three
models,
namely,
erosion-productivity
impact
calculator
model
(K_EPIC),
Shirazi
(K_Shirazi),
Torri
(K_Torri)
were
used
estimate
values.
Random
Forest
(RF)
Gradient-Boosting
Decision
Tree
(GBDT)
algorithms
employed
develop
which
led
creation
maps.
The
distribution
associated
environmental
covariates
also
investigated
across
varying
Results
showed
that
RF
achieved
highest
accuracy,
with
R2
K_EPIC,
K_Shirazi,
K_Torri
increasing
46%,
34%,
22%,
respectively,
compared
GBDT.
And
distinctions
among
variables
shape
patterns
been
identified.
K_EPIC
K_Shirazi
influenced
porosity
moisture.
is
more
sensitive
moisture
conditions
terrain
location.
More
importantly,
our
study
has
highlighted
disparities
in
Considering
data
distribution,
measured
values,
outperformed
others
estimating
plateau
lake
watershed.
aimed
create
maps
offered
scientific
accurate
method
assessment
Land Degradation and Development,
Год журнала:
2025,
Номер
unknown
Опубликована: Янв. 20, 2025
ABSTRACT
The
task
of
soil
erosion
estimation
received
a
significant
push
by
integrating
remote
sensing
and
geographical
information
systems
(GIS)
with
the
Revised
Universal
Soil
Loss
Equation
(RUSLE)
in
early
1990s
due
to
its
ease
applicability.
Topographic
(LS)
factor
played
quintessential
role
loss
determination,
especially
for
undulating
regions.
In
most
worldwide
studies,
topographic
extracted
from
Digital
Elevation
Model
(DEM)
using
“LS
equations”
failed
account
varying
slopes
before
material
joins
stream
or
river.
this
study,
slope
length
(L)
steepness
(S)
derived
without
cutoff
are
compared
analyzed
hilltop
mine.
results
reflect
that
LS
and,
ultimately,
over‐estimated
owing
absence
any
limits
on
terrains
when
used
conventionally
GIS
environment.
mean
estimated
is
252.26
ton
ha
−1
year
,
whereas
332.81
conventional
application
same
equation.
overestimation
was
reduced
35%
as
per
volume‐based
validation
study.
Thus,
study
proves
usefulness
factor,
which,
date,
has
mostly
been
neglected
research
studies
terrains.
pattern
also
highlights
negating
impact
vegetation
steep
slopes,
cementing
their
Nature
based
Solution
(NbS)
dynamic
landscapes
like
Mines.