Environmental Data Science,
Journal Year:
2024,
Volume and Issue:
3
Published: Jan. 1, 2024
Abstract
In
environmental
science,
where
information
from
sensor
devices
are
sparse,
data
fusion
for
mapping
purposes
is
often
based
on
geostatistical
approaches.
We
propose
a
methodology
called
adaptive
distance
attention
that
enables
us
to
fuse
heterogeneous,
and
mobile
predict
values
at
locations
with
no
previous
measurement.
The
approach
allows
automatically
weighting
the
measurements
according
priori
quality
about
device
without
using
complex
resource-demanding
assimilation
techniques.
Both
ordinary
kriging
general
regression
neural
network
(GRNN)
integrated
into
this
their
learnable
parameters
deep
learning
architectures.
evaluate
method
three
static
phenomena
different
complexities:
case
related
simplistic
phenomenon,
topography
over
an
area
of
196
$
{km}^2
annual
hourly
{NO}_2
concentration
in
2019
Oslo
metropolitan
region
(1026
).
simulate
networks
100
synthetic
six
characteristics
measurement
spatial
resolution.
Generally,
outcomes
promising:
we
significantly
improve
metrics
baseline
models.
Besides,
Nadaraya–Watson
kernel
provides
as
good
system
enabling
possibility
alleviate
processing
cost
sparse
data.
encouraging
results
motivate
keeping
adapting
space-time
evolving
isolated
areas.
Sustainability,
Journal Year:
2024,
Volume and Issue:
16(10), P. 4124 - 4124
Published: May 14, 2024
This
study
introduces
a
developed
environmental
quality
assessment
system,
detailing
its
hardware,
software,
and
comparative
analysis
against
publicly
available
system.
While
showing
larger
deviations
in
particulate
matter
air
humidity
parameters,
the
proposed
system
demonstrates
sufficient
accuracy
other
characteristics.
It
establishes
standardized
operating
procedure
evaluates
uncertainty
assurance
measures,
ensuring
reliability
measurements.
The
offers
comprehensive
capabilities,
measuring
parameters
like
total
volatile
organic
compounds,
carbon
dioxide,
temperature,
humidity,
matter,
noise,
nitrogen
oxides,
sulfur
ozone,
monoxide,
with
real-time
monitoring
functions
for
detecting
changes.
Its
user-friendly
interfaces,
scalability,
potential
integration
existing
systems
enhance
versatility
cost-effectiveness
across
diverse
settings.
underscores
need
future
research
to
accuracy,
reliability,
operability
explore
smart
city
initiatives
management
systems.
Overall,
represents
promising
advancement
technology,
facilitating
management.
Research Square (Research Square),
Journal Year:
2024,
Volume and Issue:
unknown
Published: Aug. 9, 2024
Abstract
Particulate
Matter
(PM)
air
pollution
poses
significant
threats
to
public
health.
Existing
models
for
predicting
PM
levels
range
from
Chemical
Transport
Models
statistical
approaches,
with
Machine
Learning
(ML)
tools
showing
superior
performance
due
their
ability
capture
highly
non-linear
atmospheric
responses.
This
research
introduces
a
novel
methodology
leveraging
ML
predict
PM2.5
at
fine
spatial
resolution
of
30
metres
and
temporal
scale
10
seconds.
The
aims
demonstrate
its
proficiency
in
estimating
missing
measurements
urban
areas
that
lack
direct
observational
data.
A
hybrid
dataset
was
curated
an
intensive
aerosol
campaign
Selly
Oak,
Birmingham,
UK,
utilizing
citizen
scientists
low-cost
Optical
Particle
Counters
(OPCs)
strategically
placed
both
static
mobile
settings.
Spatially
resolved
proxy
variables,
meteorological
parameters,
properties
were
integrated,
enabling
fine-grained
analysis
distribution
along
road
segments.
Calibration
involved
three
approaches:
Standard
Random
Forest
Regression,
Sensor
Transferability
Evaluation,
Road
Evaluation.
Results
demonstrated
high
predictive
accuracy
(R2
=
0.85,
MAE
1.60
µg
m−³)
the
standard
RF
model.
transferability
evaluations
exhibited
robust
generalization
capabilities
across
different
sensors
(best
R2
0.65,
2.76
types
0.71,
2.46
m−³),
respectively.
has
potential
significantly
enhance
beyond
regulatory
monitoring
infrastructure,
thereby
refining
quality
predictions
improving
exposure
assessments.
findings
underscore
importance
ML-based
approaches
advancing
our
understanding
dynamics
implications
paper
important
science
initiatives,
as
it
suggests
contributions
small
number
participants
can
local
patterns
many
1000s
residents.
Environmental Data Science,
Journal Year:
2024,
Volume and Issue:
3
Published: Jan. 1, 2024
Abstract
In
environmental
science,
where
information
from
sensor
devices
are
sparse,
data
fusion
for
mapping
purposes
is
often
based
on
geostatistical
approaches.
We
propose
a
methodology
called
adaptive
distance
attention
that
enables
us
to
fuse
heterogeneous,
and
mobile
predict
values
at
locations
with
no
previous
measurement.
The
approach
allows
automatically
weighting
the
measurements
according
priori
quality
about
device
without
using
complex
resource-demanding
assimilation
techniques.
Both
ordinary
kriging
general
regression
neural
network
(GRNN)
integrated
into
this
their
learnable
parameters
deep
learning
architectures.
evaluate
method
three
static
phenomena
different
complexities:
case
related
simplistic
phenomenon,
topography
over
an
area
of
196
$
{km}^2
annual
hourly
{NO}_2
concentration
in
2019
Oslo
metropolitan
region
(1026
).
simulate
networks
100
synthetic
six
characteristics
measurement
spatial
resolution.
Generally,
outcomes
promising:
we
significantly
improve
metrics
baseline
models.
Besides,
Nadaraya–Watson
kernel
provides
as
good
system
enabling
possibility
alleviate
processing
cost
sparse
data.
encouraging
results
motivate
keeping
adapting
space-time
evolving
isolated
areas.
