Natural Resources Research,
Journal Year:
2024,
Volume and Issue:
unknown
Published: Nov. 18, 2024
Abstract
The
surging
demand
for
Ni
and
Co,
driven
by
the
acceleration
of
clean
energy
transitions,
has
sparked
interest
in
Lachlan
Orogen
New
South
Wales
its
potential
lateritic
Ni–Co
resources.
Despite
recent
discoveries,
a
substantial
knowledge
gap
exists
understanding
full
scope
these
critical
metals
this
geological
province.
This
study
employed
machine
learning-based
framework,
integrating
multidimensional
datasets
to
create
prospectivity
maps
deposits
within
specific
segment.
framework
generated
variety
data-driven
models
incorporating
(rock
units,
metamorphic
facies),
structural,
geophysical
(magnetics,
gravity,
radiometrics,
remote
sensing
spectroscopy)
data
layers.
These
ranged
from
comprehensive
that
use
all
available
layers
fine-tuned
restricted
high-ranking
features.
Additionally,
two
hybrid
(knowledge-data-driven)
distinguished
between
hypogene
supergene
components
mineral
systems.
implemented
augmentation
methods
tackled
imbalances
training
samples
using
SMOTE–GAN
method,
addressing
common
learning
challenges
with
sparse
data.
overcame
difficulties
defining
negative
translating
into
proxy
employing
positive
unlabeled
bagging
technique.
revealed
robust
spatial
correlation
high
probabilities
known
occurrences,
projecting
extensions
sites
identifying
greenfield
areas
future
exploration
Orogen.
high-accuracy
developed
utilizing
Random
Forest
classifier
enhanced
mineralization
processes
promising
region.
Journal of Geophysical Research Solid Earth,
Journal Year:
2025,
Volume and Issue:
130(5)
Published: May 1, 2025
Abstract
Quantitative
analysis
of
crustal
thickness
evolution
across
deep
time
poses
critical
insights
into
the
planet's
geological
history.
It
may
help
uncover
new
areas
with
potential
mineral
deposits
and
reveal
impacts
elevation
changes
on
development
atmosphere,
hydrosphere,
biosphere.
However,
most
existing
estimation
methods
are
restricted
to
arc‐related
magmas,
limiting
their
broader
application.
By
mining
extensive
geochemical
data
from
present‐day
subduction
zones,
collision
orogenic
belts,
non‐subduction‐related
intraplate
igneous
rock
samples
worldwide,
along
corresponding
Moho
depths
during
magmatism,
we
have
developed
a
machine
learning‐based
mohometry
linking
depth,
which
is
universally
applicable
in
reconstructing
ancient
systems'
paleo‐crustal
tracking
complex
tectonic
histories
both
spatial
temporal
dimensions.
Our
novel
model
demonstrates
robust
performance,
achieving
an
R
2
0.937
Root
Mean
Squared
Error
4.3
km.
Feature
importance
filtering
highlights
key
proxies,
allowing
for
accurate
even
when
many
elements
missing.
Model
validation
southern
Tibet
South
China
Block,
regions
characterized
by
well‐constrained
processes,
its
broad
applicability.
Reconstructed
records
strong
correlation
between
thickening
events
formation
porphyry
ore
deposits,
offering
exploration
orogens
subjected
significant
surface
erosion.
enabling
reconstruction
timescales,
this
enhances
our
understanding
Earth's
internal
dynamics
interactions
thereby
advancing
comprehension
Minerals,
Journal Year:
2024,
Volume and Issue:
14(10), P. 1021 - 1021
Published: Oct. 10, 2024
Mineral
resources
are
of
great
significance
in
the
development
national
economy.
Prospecting
and
forecasting
key
to
ensure
security
mineral
supply,
promote
economic
development,
maintain
social
stability.
The
methods
for
prospecting
prediction
have
evolved
from
qualitative
quantitative
prediction,
empirical
research
mathematical
analysis.
In
recent
years,
deep
learning
algorithms
gradually
entered
attention
geologists
due
their
robust
simulation
ability
application
prediction.
Deep
can
effectively
analyze
predict
data,
which
improving
efficiency
accuracy
exploration.
However,
there
not
many
specific
examples
exploration
researchers
yet
conducted
a
comprehensive
discussion
on
advantages,
disadvantages,
prospectivity
mapping
applications.
This
paper
reviews
discusses
highlighting
challenges
faced
by
data
preprocessing,
enhancement,
system
parameter
adjustment,
evaluation,
puts
forward
suggestions
these
aspects.
purpose
this
is
provide
reference
practitioners
field
The
demand
for
critical
raw
materials
is
growing
exponentially
as
the
world
rapidly
evolves
technologically
towards
use
and
production
of
renewable
clean
energy.
To
mitigate
consequences
climate
change
move
away
from
conventional
fossil
fuels,
an
increasing
supply
critical,
economically
important,
rare
heavily
import-dependent
essential.
These
mineral
are
key
components
a
sustainable
future
with
low
CO2
emissions
indispensable
resource
development
wide
range
modern
technologies,
such
as,
electric
vehicles,
solar
panels,
wind
turbines,
batteries,
drones,
military
equipment,
etc.
For
many
years,
processing
has
been
crucial
to
meeting
industrial
social
energy
metals.
evolving
green
transition
primarily
about
not
only
world's
needs,
but
also
society's
expectations
zero
by
2050
or
earlier.
Renewable
will
play
role
in
achieving
transition,
it
require
minerals.
Minerals,
Journal Year:
2024,
Volume and Issue:
14(12), P. 1281 - 1281
Published: Dec. 17, 2024
In
geology
and
mineralogy,
optical
microscopic
images
have
become
a
primary
research
focus
for
intelligent
mineral
recognition
due
to
their
low
equipment
cost,
ease
of
use,
distinct
characteristics
in
imaging.
However,
close
reflectivity
or
transparency,
some
minerals
are
not
easily
distinguished
from
other
background.
Secondly,
the
number
background
pixels
often
vastly
exceeds
individual
particles,
different
particles
image
also
varies
significantly.
These
led
issue
data
imbalance.
This
imbalance
results
lower
accuracy
categories
with
fewer
samples.
To
address
these
issues,
flexible
ensemble
learning
semantic
segmentation
based
on
multiple
optimized
Res-UNet
models
is
proposed,
introducing
dice
loss
focal
functions
incorporating
pre-positioned
spatial
transformer
networks
block.
Twelve
were
used
construct
learnings
using
heterogeneous
strategies.
The
demonstrate
that
system
integrated
five
learners
weighted
voting
fusion
method
(RUEL-5-WV)
achieved
best
performance
mean
Intersection
over
Union
(mIOU)
91.65
across
all
nine
an
IOU
84.33
transparent
(gangue).
indicate
this
scheme
outperforms
models.
Compared
classical
Deeplabv3
PSPNet,
exhibits
significant
advantages.
Rare
earth
elements
(REEs)
have
gained
significant
global
importance
due
to
their
critical
role
in
supporting
the
transition
towards
reduced
carbon
emissions
through
industrial
applications.
REEs
serve
as
essential
raw
materials
for
various
components
modern
infrastructure,
defense
systems,
and
technological
advancements.
Geochemical
geophysical
data
are
pivotal
assessing
potential
of
REEs.
provide
direct
insights
into
elemental
composition
rocks
soils,
offering
valuable
information
on
presence
dispersion
However,
complex
geological
processes
that
influence
distribution
often
exhibit
intricate
spatial
patterns
may
not
be
fully
captured
by
geochemical
alone.
Geophysical
data,
such
gravity
magnetic
offer
indirect
but
complementary
subsurface
structures
mineral
potential.
The
integration
geochemical,
gravity,
can
aid
identifying
exploration
targets
with
increased
confidence
levels.
While
each
source
individually
provides
information,
combination
allows
identification
areas
where
multiple
anomalies
coincide,
indicating
a
higher
likelihood
mineralization.
This
approach
helps
reduce
uncertainties
prioritizing
consistent
characteristics
across
datasets,
thereby
enhancing
chances
discovering
economically
viable
REE
reserves.