bioRxiv (Cold Spring Harbor Laboratory),
Journal Year:
2024,
Volume and Issue:
unknown
Published: Oct. 18, 2024
Abstract
1.
Species
distribution
models
(SDMs)
are
crucial
tools
for
understanding
and
predicting
biodiversity
patterns,
yet
they
often
struggle
with
limited
data,
biased
sampling,
complex
species-environment
relationships.
Here
I
present
NicheFlow,
a
novel
foundation
model
SDMs
that
leverages
generative
AI
to
address
these
challenges
advance
our
ability
predict
species
distributions
across
taxa
environments.
2.
NicheFlow
employs
two-stage
approach,
combining
embeddings
two
chained
models,
one
generate
in
environmental
space,
second
geographic
space.
This
architecture
allows
the
sharing
of
information
captures
complex,
non-linear
relationships
trained
on
comprehensive
dataset
reptile
evaluated
its
performance
using
both
standard
SDM
metrics
zero-shot
prediction
tasks.
3.
demonstrates
good
predictive
performance,
particularly
rare
data-deficient
species.
The
successfully
generated
plausible
not
seen
during
training,
showcasing
potential
prediction.
learned
captured
meaningful
ecological
information,
revealing
patterns
niche
structure
taxa,
latitude
range
sizes.
4.
As
proof-of-principle
model,
represents
significant
modeling,
offering
powerful
tool
addressing
pressing
questions
ecology,
evolution,
conservation
biology.
Its
joint
hypothetical
niches
opens
new
avenues
exploring
evolutionary
questions,
including
ancestral
reconstruction
community
assembly
processes.
approach
has
transform
improve
capacity
manage
face
global
change.
Journal of Animal Ecology,
Journal Year:
2023,
Volume and Issue:
92(12), P. 2248 - 2262
Published: Oct. 25, 2023
Abstract
Data
deficiencies
among
rare
or
cryptic
species
preclude
assessment
of
community‐level
processes
using
many
existing
approaches,
limiting
our
understanding
the
trends
and
stressors
for
large
numbers
species.
Yet
evaluating
dynamics
whole
communities,
not
just
common
charismatic
species,
is
critical
to
responses
biodiversity
ongoing
environmental
pressures.
A
recent
surge
in
both
public
science
government‐funded
data
collection
efforts
has
led
a
wealth
data.
However,
these
programmes
use
wide
range
sampling
protocols
(from
unstructured,
opportunistic
observations
wildlife
well‐structured,
design‐based
programmes)
record
information
at
variety
spatiotemporal
scales.
As
result,
available
vary
substantially
quantity
content,
which
must
be
carefully
reconciled
meaningful
ecological
analysis.
Hierarchical
modelling,
including
single‐species
integrated
models
hierarchical
community
models,
improved
ability
assess
predict
processes.
Here,
we
highlight
emerging
‘integrated
modelling’
framework
that
combines
integration
modelling
improve
inferences
on
species‐
dynamics.
We
illustrate
with
series
worked
examples.
Our
three
case
studies
demonstrate
how
can
used
extend
geographic
scope
when
distributions
richness
patterns;
discern
population
over
time;
estimate
demographic
rates
growth
communities
sympatric
implemented
examples
multiple
software
methods
through
R
platform
via
packages
formula‐based
interfaces
development
custom
code
JAGS,
NIMBLE
Stan.
Integrated
provide
an
exciting
approach
model
biological
observational
types
sources
simultaneously,
thus
accounting
uncertainty
error
within
unified
framework.
By
leveraging
combined
benefits
produce
valuable
about
as
well
dynamics,
allowing
holistic
evaluation
effects
global
change
biodiversity.
Biological reviews/Biological reviews of the Cambridge Philosophical Society,
Journal Year:
2024,
Volume and Issue:
unknown
Published: Aug. 8, 2024
ABSTRACT
Big
biodiversity
data
sets
have
great
potential
for
monitoring
and
research
because
of
their
large
taxonomic,
geographic
temporal
scope.
Such
become
especially
important
assessing
changes
in
species'
populations
distributions.
Gaps
the
available
data,
spatial
gaps,
often
mean
that
are
not
representative
target
population.
This
hinders
drawing
large‐scale
inferences,
such
as
about
trends,
may
lead
to
misplaced
conservation
action.
Here,
we
conceptualise
gaps
a
missing
problem,
which
provides
unifying
framework
challenges
solutions
across
different
types
sets.
We
characterise
typical
classes
then
use
theory
explore
implications
questions
trends
factors
affecting
occurrences/abundances.
By
using
this
framework,
show
bias
due
can
arise
when
sampling
and/or
availability
overlap
with
those
species.
But
set
per
se
is
biased.
The
outcome
depends
on
ecological
question
statistical
approach,
determine
choices
around
sources
variation
taken
into
account.
argue
approaches
long‐term
species
trend
modelling
susceptible
since
models
do
tend
account
driving
missingness.
To
identify
general
review
empirical
studies
simulation
compare
some
most
frequently
employed
deal
including
subsampling,
weighting
imputation.
All
these
methods
reduce
but
come
at
cost
increased
uncertainty
parameter
estimates.
Weighting
techniques
arguably
least
used
so
far
ecology
both
variance
Regardless
method,
ability
critically
knowledge
of,
on,
creating
gaps.
outline
necessary
considerations
dealing
stages
collection
analysis
workflow.
Nature Conservation,
Journal Year:
2025,
Volume and Issue:
58, P. 11 - 30
Published: Jan. 20, 2025
In
the
last
decade,
databases
of
records
species
observed
at
same
location
different
points
in
time
over
large
spatial
extents
have
been
made
available.
Unfortunately,
these
sources
are
scarce
regions
such
as
Latin
America.
We
present
a
dataset
60,179
point
occurrences
(i.e.
presence-only
data,
PO)
and
45,468
camera-trap
survey
presence-absence
PA)
for
63
carnivores
Neotropical
Region
from
2000
to
2021.
collated
data
various
sources,
including
64
newly-digitised
bibliographic
references.
cleaned,
taxonomically
harmonised
standardised
following
Darwin
Core
Humboldt
standards
them
here
csv
files.
also
fit
analyses
by
aggregating
into
two
periods
(time1:
2000–2013
time2:
2014–2021),
with
PO
grid
cell
counts
100
×
km
PA
polygons
varying
size,
presented
geopackage
These
can
be
used
large-scale
distribution
models,
calculation
population
trends,
extinction
risk
educational
purposes.
Conservation Biology,
Journal Year:
2025,
Volume and Issue:
unknown
Published: March 24, 2025
Species
distribution
models
(SDMs)
are
important
tools
for
assessing
biodiversity
change.
These
require
high-quality
occurrence
data,
which
not
always
available.
Therefore,
it
is
increasingly
to
determine
how
data
choice
affects
predictions
of
species'
ranges.
Opportunistic
records
and
expert
maps
both
widely
used
sources
species
SDMs.
However,
unclear
SDMs
based
on
these
differ
in
performance,
particularly
the
marine
realm.
We
built
233
fish
from
2
families
with
types
compared
their
performances
potential
predictions.
occurrences
were
sourced
field
surveys
South
China
Sea
online
repositories
International
Union
Conservation
Nature
Red
List
database.
generalized
linear
explore
drivers
differences
prediction
between
model
types.
When
projecting
distinct
regions
no
calibrated
using
opportunistic
performed
better
than
those
maps,
indicating
transferability
new
environments.
Differences
predictor
values
accounted
dissimilarity
predictions,
likely
because
included
large
areas
unsuitable
environmental
conditions.
Dissimilarity
levels
among
differed,
suggesting
a
taxonomic
bias
sources.
Our
findings
highlight
sensitivity
distributional
data.
Although
have
an
role
modeling,
we
suggest
researchers
assess
accuracy
reduce
commission
errors
knowledge
target
species.
Journal of Biogeography,
Journal Year:
2023,
Volume and Issue:
50(8), P. 1405 - 1408
Published: July 8, 2023
and
macroecology
originated
in
a
novel
top-down
statistical
view
(Brown
&
Maurer,
1989),
to
name
but
few.We
are
an
age
of
innovationduringwhichtherapidemergenceofnewtechniquescan
provide
unparalleled
information
from
the
smallest
largest
spatial
scales,
individuals
communities,
seconds
tomillennia.Moreover,acquireddataaremorenumerousandmore
accessiblethaneverbefore-evenbeyondtraditionalresearchcommunities,
many
cases
unprecedentedly
vast
increasingly
globalepistemiccommunities-andtheorymorerefined.
Global Ecology and Biogeography,
Journal Year:
2024,
Volume and Issue:
33(5)
Published: Feb. 26, 2024
Abstract
Aim
Climate
change
and
habitat
loss
or
degradation
are
some
of
the
greatest
threats
that
species
face
today,
often
resulting
in
range
shifts.
Species
traits
have
been
discussed
as
important
predictors
shifts,
with
identification
general
trends
being
great
interest
to
conservation
efforts.
However,
studies
reviewing
relationships
between
shifts
questioned
existence
such
generalized
trends,
due
mixed
results
weak
correlations,
well
analytical
shortcomings.
The
aim
this
study
was
test
relationship
empirically,
using
approaches
account
for
common
sources
bias
when
assessing
trends.
Location
Tanzania,
East
Africa.
Time
period
1980–1999
2000–2020.
Major
taxa
studied
57
savannah
specialist
birds
found
belonging
26
families
11
orders.
Methods
We
applied
recently
developed
integrated
spatio‐temporal
distribution
models
R‐INLA,
combining
citizen
science
bird
Atlas
data
estimate
ranges
species,
quantify
predictive
power
traditional
trait
groups,
exposure‐related
sensitivity
traits.
based
our
on
40
years
observations
African
savannahs,
a
biome
has
experienced
increasing
climatic
non‐climatic
pressures
over
recent
decades.
correlated
patterns
linear
regression
models.
Results
find
indications
identified
by
previous
research,
but
low
average
explanatory
from
an
ecological
perspective,
confirming
lack
meaningful
associations.
analysis
finds
compelling
species‐specific
results.
Main
conclusions
highlight
importance
individual
assessments
while
demonstrating
usefulness
approach
analyses
Diversity and Distributions,
Journal Year:
2024,
Volume and Issue:
30(10)
Published: April 12, 2024
Abstract
Aim
Planning
conservation
action
requires
accurate
estimates
of
abundance
and
distribution
the
target
species.
For
many
mammals,
particularly
those
inhabiting
tropical
forests,
there
are
insufficient
data
to
assess
their
status.
We
present
a
framework
for
predicting
species
using
jaguarundi
(
Herpailurus
yagouaroundi
),
poorly
known
felid
which
basic
information
on
is
lacking.
Location
Mesoamerica
South
America.
Time
Period
From
2003
2021.
Taxa
yagouaroundi.
Methods
combined
camera‐trap
from
multiple
sites
used
an
occupancy
modelling
accounting
imperfect
detection
identify
habitat
associations
predict
range‐wide
jaguarundis.
Results
Our
model
predicted
that
probability
positively
associated
with
rugged
terrain,
herbaceous
cover,
human
night‐time
light
intensity.
Jaguarundi
was
be
higher
where
precipitation
less
seasonal,
at
intermediate
levels
diurnal
temperature
range.
camera
also
revealed
additional
detections
jaguarundis
beyond
current
International
Union
Conservation
Nature
(IUCN)
range
distribution,
including
Andean
foothills
Colombia
Bolivia.
Main
Conclusion
Occupancy
low
throughout
much
Amazonian
lowlands,
vast
area
centre
Further
work
required
investigate
whether
this
represents
sub‐optimal
conditions
Overall,
we
estimate
crude
global
population
35,000
230,000
individuals,
covering
4,453,406
km
2
Meso‐
America
0.5
level
occupancy.
allows
initially
detailed,
well‐informed
should
challenged
refined
improved
layers
records
detection.
encourage
similar
studies
lesser‐known
pooling
existing
by‐catch
growing
bank
surveys
around
world.
Scientific Reports,
Journal Year:
2024,
Volume and Issue:
14(1)
Published: Aug. 20, 2024
As
an
important
fishery
resource
and
endangered
species,
studying
the
habitat
of
Coilia
nasus
(C.
nasus)
is
highly
significant.
This
study
used
survey
data
from
southern
Zhejiang
coastal
waters
2016
to
2020,
employing
a
maximum
entropy
model
(MaxEnt)
map
distribution
C.
nasus.
Model
performance
was
evaluated
using
two
metrics:
area
under
curve
(AUC)
receiver
operating
characteristic
for
training
test
sets
true
skill
statistics
(TSS).
aimed
predict
explore
how
environmental
variables
influence
suitability.
The
results
indicated
that
models
each
season
had
strong
predictive
performance,
with
AUC
values
above
0.8
TSS
exceeding
0.6,
indicating
they
could
accurately
presence
In
area,
primarily
found
in
brackish
or
marine
near
bays
islands.
Among
all
factors,
salinity
(S)
bottom
temperature
(BOT)
highest
correlations
distribution,
although
these
varied
across
seasons.
findings
this
provide
empirical
evidence
reference
conservation
management
designation
its
protected
areas.
Methods in Ecology and Evolution,
Journal Year:
2024,
Volume and Issue:
unknown
Published: Oct. 22, 2024
Abstract
Occupancy
models
estimate
distributions
of
imperfectly
detected
species,
but
violations
the
closure
assumption
can
bias
results.
However,
researchers
working
with
mobile
animals
may
find
it
impossible
to
eliminate
such
violations.
Here,
we
tested
hypothesis
that
occupancy
fit
realistic
sampling
data
generate
unbiased
estimates
for
an
itinerant
Wood
Thrush
(
Hylocichla
mustelina
)
population.
In
2013
and
2014,
tracked
movements
41
breeding
males.
We
modelled
territory
shift
probabilities
using
logistic
exposure
within‐territory
continuous‐time
stochastic
process
models.
then
constructed
individual‐based
model,
simulated
(1000
iterations)
spatiotemporal
locations
individuals
these
populations
162
different
point
count
protocols
variable
spatial
(sampling
radius
placement
method),
temporal
(survey
length,
between‐survey
intervals
number
surveys)
characteristics.
compared
true
values
instantaneous,
daily
seasonal
from
simulations.
parameterized
continuous
time
based
on
within
34
unique
territories
estimated
a
probability
0.0099
(95%
CI:
0.0060,
0.0152).
Simulated
indicated
ranged
0.18
(0.06,
1.00)
0.80
(0.71,
0.89)
depending
protocol
increased
increasing
survey
radius,
length
interval.
Protocols
shorter
surveys
were
good
estimators
instantaneous
(low
mean‐squared
error)
poor
occupancy;
longer
generated
underestimated
occupancy.
Logistic
regression
ignored
imperfect
detection
outperformed
estimating
not
or
For
animals,
sites
changes
in
space
time.
Consequently,
aspects
have
strong,
predictable,
effects
model
parameter
estimates.
Our
results
demonstrate
how
factors
interact
is
critical
designing
produce
representative
biological
interest
researcher.