Ecological Indicators,
Год журнала:
2023,
Номер
146, С. 109898 - 109898
Опубликована: Янв. 12, 2023
Eutrophication
of
lakes
harms
aquatic
organisms,
changes
the
function
and
causes
water
pollution,
which
has
affected
global
public
health.
As
an
internationally
important
wetland
reserve,
Honghu
Lake
is
polluted
to
varying
degrees
eutrophication
trend
accelerated.
The
objective
this
study
quarterly
monitor
assess
trophic
status
from
2000
2021
using
level
index
(TLI).
TLI
was
retrieved
by
developing
a
semi-empirical
model
based
on
radial
basis
neural
network
(RBFNN),
air
temperature
data
were
added
as
input
parameters,
forming
control
group
with
that
did
not
include
data.
Based
178
Landsat
images
2021,
over
last
20
years
successfully
determined.
main
findings
are
follows:
1)
accuracy
retrieval
model,
included
(R2
=
0.723,
RMSE
4.971),
significantly
higher
than
without
0.197,
10.453);
2)
waters
in
northwestern
part
those
observed
other
lake
water;
3)
There
significant
seasonal
fluctuations
Lake.
highest
values
summer
autumn,
while
lowest
winter
spring;
4)
accelerated
2013
onwards,
showing
change
status,
light
eutrophic
mid-eutrophic.
Our
results
indicate
method
more
predictive.
Moreover,
for
entire
determined,
may
contribute
protection
environment.
Ecological Informatics,
Год журнала:
2024,
Номер
81, С. 102608 - 102608
Опубликована: Апрель 21, 2024
Reservoir
eutrophication,
caused
by
human
activities
and
climate
change,
has
emerged
as
a
critical
environmental
concern
that
attracted
both
governmental
public
attention.
However,
accurate
measurement
of
water
quality
parameters,
such
chlorophyll
(CHL-a),
clarity
(Secchi
depth;
SD),
total
suspended
solids
(TSS),
in
inland
waters
is
challenging
due
to
the
optical
complexity
individual
bodies,
which
impedes
optimization
conventional
bio-optical
algorithms.
The
aim
this
study
was
demonstrate
viability
harmonizing
Sentinel-2
Multi-Spectral
Imager
(MSI)
Landsat-8
Operational
Land
(OLI)
satellite
imagery
surface
reflectance
(SR)
products
facilitate
monitoring
reservoir
CHL-a,
SD,
TSS
using
Google
Earth
Engine
(GEE)
platform
machine
learning
Machine
models
were
trained
OLI
MSI
identify
bands
combinations
predicting
TSS.
Among
algorithms
tested,
random
forest
(RF)
(S-2
MSI:
R2
=
0.61,
mean
absolute
error
[MAE]
6.56%,
root-mean-square
[RMSE]
12.51
μg/L,
L-8
OLI:
0.56,
MAE
8.44%,
RMSE
16.01
μg/L)
yielded
best
results
test
set
for
CHL-a
prediction
from
OLI,
outperforming
k-nearest
neighbor
(KNN),
AdaBoost,
artificial
neural
network
(ANN)
models.
It
also
showed
superior
performance
SD
prediction.
feature
importance
analysis
revealed
specific
band
ratios,
(red/red
edge1)*red
edge2
red/blue
significant
predictors
ratio
green
highly
predictive
respectively.
fall
predictions
varying
trophic
levels
reservoirs.
indicated
2%
reservoirs
oligotrophic,
while
46%,
43%,
9%
mesotrophic,
eutrophic,
hypertrophic,
Meanwhile,
51%
6%,
35%,
8%
Overall,
demonstrates
effectiveness
estimating
parameters
This
approach
potential
yield
valuable
insights
aiding
assessment
management
at
regional,
national,
global
levels.
Environmental Science & Technology,
Год журнала:
2025,
Номер
unknown
Опубликована: Янв. 16, 2025
Accurate
prediction
of
chlorophyll-a
(Chl-a)
concentrations,
a
key
indicator
eutrophication,
is
essential
for
the
sustainable
management
lake
ecosystems.
This
study
evaluated
performance
Kolmogorov-Arnold
Networks
(KANs)
along
with
three
neural
network
models
(MLP-NN,
LSTM,
and
GRU)
traditional
machine
learning
tools
(RF,
SVR,
GPR)
predicting
time-series
Chl-a
concentrations
in
large
lakes.
Monthly
remote-sensed
data
derived
from
Aqua-MODIS
spanning
September
2002
to
April
2024
were
used.
The
based
on
their
forecasting
capabilities
March
August
2024.
KAN
consistently
outperformed
others
both
test
forecast
(unseen
data)
phases
demonstrated
superior
accuracy
capturing
trends,
dynamic
fluctuations,
peak
concentrations.
Statistical
evaluation
using
ranking
metrics
critical
difference
diagrams
confirmed
KAN's
robust
across
diverse
sites,
further
emphasizing
its
predictive
power.
Our
findings
suggest
that
KAN,
which
leverages
KA
representation
theorem,
offers
improved
handling
nonlinearity
long-term
dependencies
data,
outperforming
grounded
universal
approximation
theorem
algorithms.
