Environmental Science & Technology,
Journal Year:
2024,
Volume and Issue:
58(42), P. 18822 - 18833
Published: Oct. 11, 2024
Stream
salinization
is
a
global
issue,
yet
few
models
can
provide
reliable
salinity
estimates
for
unmonitored
locations
at
the
time
scales
required
ecological
exposure
assessments.
Machine
learning
approaches
are
presented
that
use
spatially
limited
high-frequency
monitoring
and
distributed
discrete
samples
to
estimate
daily
stream-specific
conductance
across
watershed.
We
compare
predictive
performance
of
space-
time-unaware
Random
Forest
time-aware
Recurrent
Graph
Convolution
Neural
Network
(KGE:
0.67
0.64,
respectively)
explainable
artificial
intelligence
methods
interpret
model
predictions
understand
drivers.
These
applied
Delaware
River
Basin,
developed
watershed
with
diverse
land
uses
experiences
anthropogenic
from
winter
deicer
applications.
capture
seasonality
first
flush
deicers,
streams
elevated
correspond
well
indicators
application.
This
result
suggests
these
be
used
identify
potential
salinity-impaired
best
management
practices.
Daily
driven
primarily
by
cover
(urbanization)
trends
may
represent
processes
weather
up
three
months.
Such
modeling
likely
transferable
other
watersheds
further
risks
Hydrological Sciences Journal,
Journal Year:
2024,
Volume and Issue:
unknown, P. 1 - 11
Published: Dec. 10, 2024
Urbanization
and
extreme
flows
are
altering
stream
temperature
dynamics,
yet
our
understanding
of
the
impact
urbanization
on
is
limited.
We
deployed
27
water
loggers
in
three
headwater
catchments
over
summers.
categorized
flow
as
low,
high,
or
average
calculated
daily
anomalies.
Comparing
Z
scores
between
conditions
revealed
events
temperature.
used
multiple
linear
regressions
to
identify
landscape
predictors
found
during
low
temperatures
were
significantly
warmer.
Additionally,
urban
linked
reduced
warming
flows.
Our
study
highlights
that
increase
events;
however,
this
effect
was
less
pronounced
more
urbanized
sites.
High
did
not
affect
These
results
underscore
vulnerability
rivers
flows;
may
help
mitigate
these
effects.
Environmental Research Letters,
Journal Year:
2024,
Volume and Issue:
19(9), P. 091001 - 091001
Published: July 30, 2024
Abstract
Poor
water
quality
threatens
human
and
environmental
health,
as
well
the
usability
of
for
sectoral
purposes.
Despite
widespread
recognition
its
importance,
our
knowledge
is
severely
impaired
by
a
lack
information.
However,
global
data
required
to
assess
critical
regions
(hotspots)
where
pollution
poses
risks
safe
use,
economic
development
ecosystem
services
health.
Here,
we
identify
blind
spots
in
current
monitoring
efforts,
elucidating
on
associated
challenges
diagnosing
issues
knock-on
effect
both
science
society.
Furthermore,
provide
recommendations
addressing
these
–
which
strong
emphasis
placed
improved
accessibility
transparency
existing
addition
increasing
efforts.
Water,
Journal Year:
2024,
Volume and Issue:
16(10), P. 1322 - 1322
Published: May 7, 2024
This
work
analyzed
the
nutrient
dynamics
(2011–2022)
and
discharge
(2005–2022)
for
Bahlui
River
at
four
distinctive
locations:
Parcovaci—a
dam-protected
area
that
has
been
untouched
by
agriculture
or
urbanization;
Belcesti—a
primarily
agricultural
area,
also
dam-protected;
Podu
Iloaiei—a
region
influenced
Holboca—placed
after
a
heavily
urbanized
area.
The
analysis
focused
on
determining
series
of
statistical
indicators
using
Minitab
21.2
software.
Two
drought
intervals
one
flood
interval
were
to
highlight
daily
evolution
during
selected
period,
showing
constructed
reservoirs
successfully
control
streamflow.
For
entire
mean
median
values
streamflow
is
consistent,
considering
locations’
positions
from
source
river’s
end.
total
nitrogen
phosphorus
as
representative
quality
indicators.
study
follows
influence
areas’
characteristics
reservoirs’
presence
dynamics.
results
showed
most
influential
factor
impacts
presence,
which
controls
discharge,
creates
wetlands
swamps,
implicitly
concentration.
Environmental Science & Technology,
Journal Year:
2024,
Volume and Issue:
58(42), P. 18822 - 18833
Published: Oct. 11, 2024
Stream
salinization
is
a
global
issue,
yet
few
models
can
provide
reliable
salinity
estimates
for
unmonitored
locations
at
the
time
scales
required
ecological
exposure
assessments.
Machine
learning
approaches
are
presented
that
use
spatially
limited
high-frequency
monitoring
and
distributed
discrete
samples
to
estimate
daily
stream-specific
conductance
across
watershed.
We
compare
predictive
performance
of
space-
time-unaware
Random
Forest
time-aware
Recurrent
Graph
Convolution
Neural
Network
(KGE:
0.67
0.64,
respectively)
explainable
artificial
intelligence
methods
interpret
model
predictions
understand
drivers.
These
applied
Delaware
River
Basin,
developed
watershed
with
diverse
land
uses
experiences
anthropogenic
from
winter
deicer
applications.
capture
seasonality
first
flush
deicers,
streams
elevated
correspond
well
indicators
application.
This
result
suggests
these
be
used
identify
potential
salinity-impaired
best
management
practices.
Daily
driven
primarily
by
cover
(urbanization)
trends
may
represent
processes
weather
up
three
months.
Such
modeling
likely
transferable
other
watersheds
further
risks
