Atmosphere,
Год журнала:
2022,
Номер
13(6), С. 971 - 971
Опубликована: Июнь 15, 2022
Reference
evapotranspiration
(ET0)
is
an
essential
component
in
hydrological
and
ecological
processes.
The
Penman–Monteith
(PM)
model
of
Food
Agriculture
Organization
the
United
Nations
(FAO)
requires
a
number
meteorological
parameters;
it
urgent
to
develop
high-precision
computationally
efficient
ET0
models
with
fewer
parameter
inputs.
This
study
proposed
genetic
algorithm
(GA)
optimize
extreme
learning
machine
(ELM),
evaluated
performances
ELM,
GA-ELM,
empirical
for
estimating
daily
Southwest
China.
Daily
data
including
maximum
temperature
(Tmax),
minimum
(Tmin),
wind
speed
(u2),
relative
humidity
(RH),
net
radiation
(Rn),
global
solar
(Rs)
during
1992–2016
from
stations
were
used
training
testing.
results
FAO-56
formula
as
control
group.
showed
that
GA-ELM
(with
R2
ranging
0.71–0.99,
RMSE
0.036–0.77
mm·d−1)
outperformed
standalone
ELM
0.716–0.99,
0.08–0.77
testing,
both
which
superior
0.36–0.91,
0.69–2.64
mm·d−1).
prediction
accuracy
varies
different
input
combination
models.
using
Tmax,
Tmin,
u2,
RH,
Rn/Rs
(GA-ELM5/GA-ELM4
ELM5/ELM4)
obtained
best
estimates,
0.98–0.99,
0.03–0.21
mm·d−1,
followed
by
(GA-ELM3/GA-ELM2
ELM3/ELM2)
involved
Rn
those
Rs
when
quantity
parameters
was
same.
Overall,
GA-ELM5
(Tmax,
RH
inputs)
other
thus
recommended
estimation.
With
estimation
accuracy,
computational
costs,
availability
accounted,
GA-ELM2
determined
be
most
effective
limited
Water,
Год журнала:
2023,
Номер
15(3), С. 486 - 486
Опубликована: Янв. 25, 2023
Modeling
potential
evapotranspiration
(ET0)
is
an
important
issue
for
water
resources
planning
and
management
projects
involving
droughts
flood
hazards.
Evapotranspiration,
one
of
the
main
components
hydrological
cycle,
highly
effective
in
drought
monitoring.
This
study
investigates
efficiency
two
machine-learning
methods,
random
vector
functional
link
(RVFL)
relevance
machine
(RVM),
improved
with
new
metaheuristic
algorithms,
quantum-based
avian
navigation
optimizer
algorithm
(QANA),
artificial
hummingbird
(AHA)
modeling
ET0
using
limited
climatic
data,
minimum
temperature,
maximum
extraterrestrial
radiation.
The
outcomes
hybrid
RVFL-AHA,
RVFL-QANA,
RVM-AHA,
RVM-QANA
models
compared
single
RVFL
RVM
models.
Various
input
combinations
three
data
split
scenarios
were
employed.
results
revealed
that
AHA
QANA
considerably
methods
ET0.
Considering
periodicity
component
radiation
as
inputs
prediction
accuracy
applied
methods.
Technologies,
Год журнала:
2024,
Номер
12(6), С. 77 - 77
Опубликована: Июнь 1, 2024
Irrigation
is
crucial
for
crop
cultivation
and
productivity.
However,
traditional
methods
often
waste
water
energy
due
to
neglecting
soil
variations,
leading
inefficient
distribution
potential
stress.
The
stress
index
(CWSI)
has
become
a
widely
accepted
assessing
plant
status.
it
necessary
forecast
the
estimate
quantity
of
irrigate.
Deep
learning
(DL)
models
forecasting
have
gained
prominence
in
irrigation
management
address
these
needs.
In
this
paper,
we
present
comparative
study
between
two
deep
models,
ConvLSTM
CNN-LSTM,
using
remote
sensing
data.
While
DL
architectures
been
previously
proposed
studied
various
applications,
our
novelty
lies
studying
their
effectiveness
field
time
series
images.
methodology
involves
meticulous
preparation
data,
where
calculate
Landsat
8
satellite
imagery
through
Google
Earth
Engine.
Subsequently,
implemented
fine-tuned
hyperparameters
CNN-LSTM
models.
same
processes
model
compilation,
optimization
hyperparameters,
training
were
applied
architectures.
A
citrus
farm
Morocco
was
chosen
as
case
study.
analysis
results
reveals
that
excels
over
long
sequences
(nine
images)
with
an
RMSE
0.119
0.123,
respectively,
while
provides
better
short
(three
than
0.153
0.187,
respectively.
