Journal of Forest Science,
Год журнала:
2024,
Номер
70(5), С. 209 - 222
Опубликована: Май 6, 2024
Tree
volume
is
a
characteristic
used
in
many
cases,
such
as
determining
fertility,
habitat
quality,
growth
size,
allowable
harvesting,
and
the
principles
of
forest
trade.
It
imperative
to
develop
methods
that
predict
stand
obtain
this
extensive
information
quickly
cost-effectively.
This
study
supervised
self-organising
map
(SSOM),
multi-layer
perceptron
(MLP),
radial
basis
function
(RBF)
neural
networks
based
on
physiography,
topography,
soil,
human
factors.
A
sensitivity
analysis
method
called
importance
prediction
was
determine
how
input
variables
influenced
network
output.
First,
homogeneous
units
prepared
with
ArcMap
(Version
10.3.1,
2015)
by
combining
digital
layers
measure
tree's
per
hectare.
Then,
separate
tree
species
different
diameter
classes
were
measured
circular
grid
200
m
×
150
m,
0.1
ha
coverage,
3.3%
sampling
intensity,
at
breast
height
(DBH)
greater
than
7.5
cm
using
systematic
unit
regular
random
method.
The
modelling
results
showed
SSOM,
MLP,
RBF
predicted
most
accurately
according
Furthermore,
found
altitude
above
sea
level,
soil
depth,
slope
are
influential
variables.
In
contrast,
texture
least
effective
predicting
volume.
Agriculture,
Год журнала:
2024,
Номер
14(11), С. 1876 - 1876
Опубликована: Окт. 24, 2024
Precision
agriculture
aims
to
improve
crop
management
using
advanced
analytical
tools.
In
this
context,
the
objective
of
study
is
develop
an
innovative
predictive
model
estimate
yield
and
morphological
quality,
such
as
circularity
length–width
ratio
potato
tubers,
based
on
phenotypic
characteristics
plants
data
captured
through
spectral
cameras
equipped
UAVs.
For
purpose,
experiment
was
carried
out
at
Santa
Ana
Experimental
Station
in
central
Peruvian
Andes,
where
clones
were
planted
December
2023
under
three
levels
fertilization.
Random
Forest,
XGBoost,
Support
Vector
Machine
models
used
predict
quality
parameters,
ratio.
The
results
showed
that
Forest
XGBoost
achieved
high
accuracy
prediction
(R2
>
0.74).
contrast,
less
accurate,
with
standing
most
reliable
=
0.55
for
circularity).
Spectral
significantly
improved
capacity
compared
agronomic
alone.
We
conclude
integrating
indices
multitemporal
into
estimating
certain
traits,
offering
key
opportunities
optimize
agricultural
management.
Journal of Forest Science,
Год журнала:
2024,
Номер
70(5), С. 209 - 222
Опубликована: Май 6, 2024
Tree
volume
is
a
characteristic
used
in
many
cases,
such
as
determining
fertility,
habitat
quality,
growth
size,
allowable
harvesting,
and
the
principles
of
forest
trade.
It
imperative
to
develop
methods
that
predict
stand
obtain
this
extensive
information
quickly
cost-effectively.
This
study
supervised
self-organising
map
(SSOM),
multi-layer
perceptron
(MLP),
radial
basis
function
(RBF)
neural
networks
based
on
physiography,
topography,
soil,
human
factors.
A
sensitivity
analysis
method
called
importance
prediction
was
determine
how
input
variables
influenced
network
output.
First,
homogeneous
units
prepared
with
ArcMap
(Version
10.3.1,
2015)
by
combining
digital
layers
measure
tree's
per
hectare.
Then,
separate
tree
species
different
diameter
classes
were
measured
circular
grid
200
m
×
150
m,
0.1
ha
coverage,
3.3%
sampling
intensity,
at
breast
height
(DBH)
greater
than
7.5
cm
using
systematic
unit
regular
random
method.
The
modelling
results
showed
SSOM,
MLP,
RBF
predicted
most
accurately
according
Furthermore,
found
altitude
above
sea
level,
soil
depth,
slope
are
influential
variables.
In
contrast,
texture
least
effective
predicting
volume.
