►
Description
Abstract
The use of airborne electromagnetic (AEM) data for geoscience applications is rapidly increasing. For instance, in California USA, there is an ongoing AEM project led by the California Department of Water Resources (CDWR), which plans to map out most of the Central Valley of California and some water basins in California. All acquired AEM data and resulting interpretation of the data, which are resistivity models of the subsurface will be publicly available. There are many big initiatives like this project throughout the world. Therefore, it is critical to understand how the resulting resistivity models are obtained from the acquired AEM data, and further equipped with an ability to download and explore the available resistivity data. In this talk, I will first introduce how a resistivity model is obtained from the AEM data then introduce open-source tools that can be used to explore this resistivity model.
Bio
Seogi completed his PhD in Geophysics at the University of British Columbia, Canada, in 2018. His thesis work focused on computational electromagnetics and its application to mining problems. Currently, he is a Postdoctoral Researcher in the Geophysics Department at Stanford. His research focus is on advancing the use of airborne electromagnetic and other remote sensing methods for groundwater management and groundwater science. He continues to contribute to the development of open source software, SimPEG, and educational resources, GeoSci.xyz, for geophysics.
A
Oh
there
we
have
it
so
big,
thanks
to
soggy
for
being
willing
to
jump
in
and
give
the
seminar
this
month,
as
we've
talked
about
it'll,
be
quite
informal,
so
jump
in.
Ask
your
questions
and
we'll
have
this
be
quite
conversational,
so
soggy
is
a
postdoc
at
stanford.
Many
of
you
know
him
he's
a
very
active
contributor
and
one
of
the
founders
of
the
sim
card
project
and
the
title
of
his
talk
today
is
how
do
we
plot
a
resistivity
model
from
airborne
em
data,
so
go
ahead.
Sagi.
B
Yeah
so
as
as
I
kind
of
briefly
said
before
the
title
is
the:
how
do
you
plot
the
redistributable
from
airborne
mechanical
data?
So
it's
much
low-key
kind
of
talk,
so
it's
not
going
to
be
very
technical
but,
as
I
said,
I
kind
of
want
to
share
just
the
bits
and
pieces
of
tools
that
I
use
when
I
am
visualizing
and
analyzing
sort
of
the
result
of
the
inversion,
which
is
the
residue
model,
so
urban
electromagnetic
method,
just
in
case
people,
there
are
probably
some
people
who
haven't
heard
of
the
airborne.
B
So
this
is
like
a
typical
airborne
electromagnetic
experiment.
We've
got
a
helicopter
here
and
then
we
have
a
loop.
So,
aside
of
the
typical
loop
of
airborne
system,
it's
about
11
to
12,
meter
or
10
to
12
meter.
So
it's
a
big
loop.
If
you
think
about
it's,
probably
bigger
much
bigger
than
your
room
and
then
what
we
do,
we
inject
the
current.
B
Okay.
So
that's
the
typical
aem
experiment
and
induced
voltage
that
we
measure,
depending
upon
the
the
type
of
the
survey,
so
sometimes
you
could
use
the
frequency
domain
system
or
time
domain
system,
but
the
one
that
I
explained
was
the
time
domain
system
and
that's
the
we
met.
We
have
the
multiple
time
channels
of
the
data
that
we
can
obtain
information
about
the
subsurface
activity
and
the
only
time
usually
includes
information
about
the
near
surface,
and
the
later
time
includes
includes
information
about
the
deeper
part.
B
So,
to
start
the
analogy
and
frequency
domain,
if
you
have
high
frequency
data,
that
kind
of
contains
information
about
the
near
surface
and
if
you
have
a
low
frequency
data,
it
contains
information
about
that
deeper
part
of
the
surface,
and
once
you
obtain
the
data,
this
is
a
sort
of
the
model
that
you
can
obtain.
So
this
was
sort
of
one
of
my
latest
paper,
and
this
is
somewhere
in
the
north.
California,
northern
california
and
the
scale
is
about
50
kilometer
in
the
length
scale
and
that
depth.
B
B
The
purpose
of
this
survey
was
in
the
context
of
like
mapping
hydrogeology
of
the
subsurface,
so
people
were
interested
in
what
is
a
sediment
type
of
the
surface,
so
blue,
which
is
a
low
resistivity,
corresponds
to
fine
grained
and
materials
like
clay
and
silt
red
means
coarse
and
green,
coarse,
grained
material
like
sand
and
gravel.
So
of
course,
crane
material
can
be
sort
of
the
pathway
of
the
water.
Fine
grain
could
be
a
barriers
of
water
flow,
which
is
an
important
information
to
understand
the
hydrologic
system.
B
B
It
will
be
much
bigger
than
100
kilometer,
like
thousands
of
kilometer
length
scale,
which
is
huge,
and
this
is
the
amount
of
data
that
they
covered
with
the
airborne
em,
and
you
can
always
like
you
already
see
some
sort
of
spatial
patterns
of
the
resistivity,
which
is
pretty
cool.
So
this
is
a
nationwide
survey.
The
typical
line
spacing
of
the
am
data.
B
So
when
we're
talking
about
airborne
data,
we
usually
talk
about
what
is
the
line
space
because
you
fly
over
a
large
area
but
depending
upon
the
line
spacing
the
amount
of
data
that
you
acquire
might
be
quite
different.
And
then
I
think
the
the
typical
line
spacing
used
for
this
nationwide
survey
is
55
to
20
kilometer
line
spacing,
which
is
pretty
large,
okay
and
it
was
for
various
purposes.
I
think
the
major
purpose
was
minerals,
but
I
think
there
are
some
purposes
for
waters
and
other
stuff
as
well
and
what's
really.
B
Is
all
of
this
data,
so
the
measured
airborne
data
in
the
form
of
voltages,
as
well
as
the
resulting
resistivity
model
by
interpreting
the
airborne
of
data
is
available
through
this
website.
That
means
we
have
lots
of
resistive
model
that
can
potentially
use
almost
for
free,
okay,
and
this
is
another
example-
a
little
bit
smaller
scale.
This
is
led
by
usgs
and
bert
minsley,
it's
group,
so
this
is
an
ironian
project
in
the
mississippi
allevial,
mississippi,
liberal
plane.
This
was
very
specific
for
hydrogeologic
mapping
and
this
is
more
like
a
statewide
survey.
B
Here's
like
at
this
boundary.
It
shows
the
states
in
in
the
u.s-
and
this
is
I
think
this
is
the
mississippi
river
and
then,
if
you
look
at
the
scale,
it's
about
a
couple
hundred
kilometers
by
a
couple
hundred
kilometers
scale,
again
really
really
large
and
then
typical
light
spacing
here
with
what
was
about
three
to
six
kilometer
and
the
total
lines
line
kilometer.
So
when
we're
talking
about
the
amount
of
data
that
we
acquire,
usually
people
use
like
a
line
kilometer.
So
that's
the
actual
amount
of
data.
B
B
So
it's
about
40
43
000
line
kilometers,
which
is
a
lot
and
that's
sort
of
the
typical
size
that
that
we
do
it's
about
like
800
line,
kilometers
1
000
line
kilometer.
So
it's
about
43
times
of
those
surveys
which
is
a
huge
survey.
Okay
and
the
other
example
is
cdwr.
It
means
the
california
department
of
water
resources.
B
So
this
is
the
governmental
sector,
basically
kind
of
managing
the
water
groundwater,
as
well
as
surface
water
in
that
in
california,
and
they
are
running
a
relatively
large-scale
airborne
yen
project,
basically
covering
the
entire
central
valley
of
california
and
some
of
the
water
basins
in
in
california.
So
this,
like
a
red
color,
is
the
scheduled
aem
survey
for
the
next
three
years,
and
this
green
color
means
it's
already
acquired
data
and
I
think
the
black
line
shows
the
flight
line.
So
this
is
where
the
salinas
valley
you
can
see.
B
This
is
the
monterey,
not
sure
you
have.
If
you
haven't
been
there,
it's
a
it's
a
beautiful
place
to
go
and
but
that
they
have
quite
a
bit
of
groundwater
issues
because
with
the
sea
water
intrusion
anyways,
there
are
lots
of
airborne
data
coming
as
well
as
well.
There
are
already
lots
of
ironing
data
available,
but
there
are
always
that
there
are
much
more
iranian
data
coming
and
emerging
and
that
will
be
available
to
the
public
and
that's
sort
of
the
bottom.
That's
sort
of
my
peach.
Okay.
Why
why?
B
B
So
that's
sort
of
the
kind
of
like
my
motivation
of
today
and
I
kind
of
want
to
start
with,
like
a
the
first
question
was,
is
how
do
we
obtain
a
resistivity
model
from
airborne
data?
And
this
is
for
this
we're
using
what
we
call
inversion
technique
and
there
there
can
be
many
different
types
of
inversion
and
I'm
going
to
introduce
a
specific
technique
which
basically
use
a
one-dimensional
assumption,
layered
assumption
and
suppose.
B
Like
a
so
here's
a
simple
example
suppose
we
got
the
true
resistivity
model
red
again
high
resistivity
blue
means
the
low
resistivity,
so
red
you
can
think
as
a
coarse
grained.
Material
blue
is
a
fine
grain
of
materials,
there's
some
sort
of
layered
structure
and
then,
if
you
have
a
helicopter
here,
doing
an
experiment,
that's
sort
of
the
data
form
of
the
data
that
you
can
obtain.
B
Since
this
is
in
voltage
at
multiple
time
channels
and
typically
you
got
20
to
30
channels
and
this
data
by
looking
at
the
data,
you
may
not
have
much
idea.
So
what
we
want
to
do
from
this
data.
We
want
to
find
a
resistivity
model
of
the
subsurface
in
one
dimension
that
can
explain
this
data
and
that's
what
we
call
inversion
process
and
for
a
single,
a
sounding
point.
B
You
can
have
1d
resistivity
model
like
this,
and
if
you
move
to
another
place,
then
you
can
have
another
1d
resistivity
model
and
you
have
many
sounding
location
and
by
doing
so
by
repeating
the
one,
the
inversion
you
can
have
a
stitch.
It
1d
resistivity
model
like
this,
but
doing
this.
This
inversion
separately
is
somewhat
cumbersome
and
then
sort
of
you're
ignoring
the
possibility
to
sort
of
like
regularize
this
resistivity
value
in
space
suppose
like
if
you're
interested
in
hydrogeology
application,
and
if
you
think
about
the
general
like
a
structure
of
hydrogeology,
is
layered.
B
So
you
would
prefer
sort
of
the
smooth
variation
of
resistivity
in
lateral
dimension
and
that's
sort
of
the
major
idea
of
what's
called
spatially
constrained.
Inversion
developed
by
solid
at
all
2008
and
the
the
idea
is
simple.
Rather
than
doing
as
one
like
a
separate
1d
version
we're
going
to
invert
all
of
them
together.
But
but
the
regularized,
like
I
put
a
sort
of
prior
information
that
the
resistivity
value
between
the
sounding
are
smoothly
changing.
B
And
if
you,
if
you're
part
of
data
abortion
data
and
then
apply
the
inversion
now,
you
can
actually
obtain
a
resistivity
model
and
potentially
that
would
be
sort
of
the
first
kind
of
first
point
that
you
can
be
exposed
to
use
this
airborne
electromagnetic
data.
And
the
question
is
like
how
do
we
plot
a
resistivity
model
and
was
sort
of
like
kind
of
kind
of
intro?
Of
of
this
talk,
and
now
I'm
going
to
do
a
demo.
B
B
You
guys
all
know
my
calendar
schedule,
so
it's
fine
yeah.
So
this
is
a
notebook
that
I
put
together
and-
and
I
think
it's
actually
available
through
this
github
repository-
that
I
will
post
in
in
slack
channel
later-
and
I
think
this
is
this-
video
is
available
later,
so
you
can
actually
use
it
as
well.
B
Okay,
so
yeah.
So
what
I'm
going
to
do
within
this
notebook,
I'm
going
to
kind
of
show
various
different
ways
how
to
plot
it
up
the
resistivity
model
and
then
I'm
I'm
very,
focusing
on
the
specific
form
of
the
inversion
because,
as
I
said
like,
we
are
inverting
for
basically
one
d
for
each
sounding
and
then
so.
Each
sounding
location
have
an
x
number
of
layers
and
then
like
x,
number
of
resistivity
values
for
each
layer.
Okay,
so
that's
sort
of
yeah
and
I'm
loading
up
a
couple
packages
and
for
simpac
users.
B
B
So
some
like
a
again
there's:
no
universal
form
people
are
exporting
their
data
set.
So
this
is
also
a
very
specific.
But
what's
the
typical
tool
used
in
in
california
is
what's
called
aarhus
workbench
and
that's?
This
is
the
the
function
that
can
read
the
export
exported
form
of
their
resistivity
model.
B
B
B
B
B
You
can
actually
think
that
as
a
sounding
number,
but
the
sounding
number
is
includes
the
certain
order
of
the
operation
over
the
survey.
So
this
is
actually
important,
sometimes
because
if
you
want
to
like
a
plot,
a
resistivity
model,
let's
say
in
the
direction
of
a
survey
and
then
actually
you
need
to
know
which
one
was
the
first.
Let's
say
you
have
a
line
and
then
you
want
to
know,
oh,
which
one
was
the
starting
point
and
that's
actually
important
information
to
know.
Okay,
the
direction
of
time
was
like
that.
B
B
B
The
line
number
the
shape
of
the
line
number
is
the
size
like
number
of
something:
okay
and
you
can
calculate
the
unique
line
and
there
are
lines
so
there
was
unique
to
43
unique
lines
and
as
a
sempac
kind
of
style,
we
factorize
a
model.
So
what
we
like,
what
I
do
resistivity
was
a
matrix
and
the
shape
was
that
and
I
flatten
and
then
vectorize.
B
So
this
is
where
there's
a
bit
of
inconsistency
in
the
1d
code,
because
the
typical
ordering
of
simpac
is
x
y,
then
z,
but
here
I
like,
I
first
use
the
z
then
than
the
radical
like
a
radial
direction
or
sounding
number
okay,
but
later
we
can
change
this,
but
this
is
sort
of
the
current
structure.
Okay
and
then
I
also
put
timestamp
and
that
now
you
can
actually
plot
it
up.
B
So
this
is
just
a
scatter
plot
and
x
and
y
and
I'm
plotting
a
topography.
So
the
blue
color
means
the
low
and
yellow
color
means
highs.
As
you
go
to
the
east,
you
can
see
the
topography
is
going
up.
This
is
where
actually
the
mountain
was
located.
This
data
is
from
somewhere
in
california,
close
to
the
mountain.
B
B
So
you
can
see
the
time
as
well,
which
one
was
the
first.
I
think
they
probably
started
somewhere
in
the
blue,
color
and
then
slowly
moving
to
the
west,
and
then
I
think
this
tie
line
or
this
like
so
the
horizontal
line
was
acquired
at
the
last
okay.
So
you
can
just
take
a
look
at
that
now,
I'm
going
to
use
what's
called
stitched,
1d
model,
and
this
is
the
function,
sorry
that
you
can
import
from
here.
B
And
I'm
not
still
still
debating.
What's
the
name
of
this
sort
of
class
like
currently,
I
named
it
as
a
station
1d
model
and
then
it's
it's
a
it's
a
function
that
kind
of
allows
you
to
just
explore
this
resistivity
model
and
then
apply
that
and
I'm
importing
two
functions,
stitch
2d
model
and
the
plot
layer
and
then
stitch
1d
model.
Input
of
this
object
is
topography
line
and
h
was
the
vertical
discretization.
B
E
B
Like
we
have
now
a
model,
and
so,
as
I
said
for
each
sounding
point,
we
have
a
resistivity
model,
so
sort
of
natural
pathway.
Okay,
let's
look
at
a
sounding
location
and
plot
it
up
as
a
function
of
vertical
profile.
Okay,
so
I
could
start
with
some
number
zero
and
then
stivity
was
ordered
as
like
a
number
of
sounding
by
a
number.
D
B
F
B
And
you
can
see
different
resistivity
values
and
well,
if
you
want
to
look
at
kind
of
many
a
resistivity
model
together,
then
you
could
not
together
but
as
an
interactive
way.
So
you
can
actually
just
change
to
generate
that
little
app
and
you
can
see
different
resistivity.
It
is
changing.
So
it
means
the
earth
is
not
uniform
and
it
looks
smooth.
B
Yeah,
so
you
can
probably
spend
hours
to
look
at
this,
but
let
me
just
move
on
okay,
so
that
was
sort
of
the
the
plotting
in
a
one-dimensional
like
a
vertical
direction
and
that's
fine
and
but
the
the
typical
like
the
way
to
plot
the
resistivity
model.
Is
the
2d
section
plot
or
so
that's
what
I
am
going
to
do
then,
like
you
need
to
know.
Where
is
the
line
like
which
line
is
where
okay?
So
that's?
This
kind
of
the
plot
shows
how
to
do
that
like
so
we
do.
B
We
have
a
unique
line.
I
think
we
got
43
lines
and
then
line
was
including
it
line.
Has
the
number
of
sounding
locations
so
same
as
the
number
of
locations
we
have
so
we're
finding
an
indices
for
each
line.
So
if
I
put
line
0
it's
here,
the
line
number
was
one
blah
blah.
One
was
two:
if
you
change
that
you
can
look
at
this,
so
you
can
see.
Where
is
the
line?
B
B
C
B
B
B
So
so
one
point
before
I
go
there:
it's
like
if
I
need
to
choose
the
distance,
so
it's
a
it's
a
relative
matrix,
so
it
I'm
counting
the
starting
point
of
that
line
and
then
calculating
sort
of
the
effective
distance
and
then
in
the
direction
of
the
fly
okay
flight.
So
that
was
the
reason
why
I
need
to
have
the
timestamp
when
we're
instant,
like
when
they're
initiating
their
model
object
anyway.
So
this
question
was
located.
How
do
we
decide
that.
B
That's
probably
a
little
bit
better,
so
lindsay
what
I'm
actually
doing
under
the
hood,
so
I'm
basically
using
the
p
color
mesh
for
each
of
the
sounding
location
and
then
setting
setting
the
dx.
So
this
is
the
dx
that
you
need
to
set
it
up.
There's
a
default
value,
and
currently
I
think,
it's
probably
like
50
or
something
like
that
and
that's
a
relevant
number
for
aem
to
fill
it
so
like.
If
you
don't
like
kind
of
having
this
space,
then
you
can
basically
fill
in
by
increasing
this
number
right.
So,
let's
say
15.
B
B
So
if
you
look
at
that,
it's
almost
like
a
2d
plot,
but
what
I
was
doing
was
basically
a
whole
bunch
of
like
a
1d
plot
and
and
then
stitching
together,
okay,
so
this
is
kind
of,
and
I
think,
as
the
data
gets
bigger,
you
may
not
want
to
do
every
sounding.
So
I
probably
need
to
change
the
code
a
little
bit
to
do
a
little
bit
smart.
If
that
that's
not
greater
than
x
number,
then
we
just
downsampled,
but
I
don't
really
do
it
at
this
point.
B
B
So
now
that
was
a
sort
of
2d
section,
but
how
about
the
plan
view
map
so
we've?
Well,
what
we
did
we
just
look
at
the
single,
sound
location,
look
at
the
vertical
profile
and
then
we
look
at
the
align
and
then
we
can
plot
it
up
as
a
section
of
the
resistivity
and
what
you
can
think
about.
So
for
each
layer
we
can
plot
up
the
resistivity
values.
So
here
I'm
starting
with
the
I'm
using
the
scatter
plot.
B
This
is
the
first
layer,
so
you
can
see
some
resistors
here
and
some
conductors,
and
this
is
actually
where
the
river
flows
so
along
the
river
you've
got
lots
of
coarse
grained
materials.
It
is
kind
of
makes
sense.
Okay,
if
you
have
a
river
we're
looking
we're
expecting
to
have
a
bit
of
like
lots
of
coarse
green
resistive
sediments,
so
it
was
kind
of
reasonable
and
that
if
you
want
to
see
delta
one.
G
B
B
I
I
don't
handle
the
type
of
power,
so
so
there's
no
like
a,
although
that
the
code
actually
takes
the
topography,
but
when
you're
regularizing
like
we
just
like
that
lateral
constraint
it
through
the
layers,
okay,
so
which
I
think
is
still
okay
but
you're
right.
But
if
there's
a
structure
that
are
not
aligned
with
the
topography,
then
I
think
we're
putting
a
wrong
prior
information.
C
B
B
Okay,
so
that
was
the
2d,
but
as
a
point,
so
we
just
plotted
out
the
each
sounding
location
as
a
point,
but
some
people
doesn't
like
it
and
then
some
people
want
to
see
that
as
a
surface.
So
how
do
we
like
interpolate
these
points
into
a
2d
surface
and
again,
I'm
using
a
very
simple
form
like
which
is
inverse
distance
interpolation?
So
there
could
be
a
whole
bunch
of
different
interpolation
that
you
could
potentially
use
it.
B
I
personally
like
it
because
it's
it's
simple
but
but
kind
of
aligned
with
what
we're
doing,
because
what
the
geophysics
does
is
an
average
of
resistivity
value
in
the
subsurface,
so
inverse
distance
rating
is
basically
an
average,
a
whole
bunch
of
residue
value
close
to
a
whole
bunch
of
sounding
close
to
that
location.
I
kind
of
like
that
and
again
like
there
could
be
many
different
tools
that
you
could.
You
could
use
what
the
inverse
distance
interpolation
takes.
B
Is
the
x
y
location
of
the
sounding,
so
topography
was
basically
the
x
y
location
of
each
sounding
and
the
elevation.
So
I'm
not
using
the
elevation.
Elevation
was
the
third
column,
and
so
you
can
actually
input
the
resistive
value
for
each
layer
and
and
the
dx,
and
the
y
is
the
discretization
in
space.
So
the
output
of
this
interpolation
is
a
two-dimensional
grid,
okay
and
that
we're
using
sort
of
the
uniform
uniform
mesh.
B
B
B
I
think
it
was
actually
100
that
I
used
it
looks
not
bad
and
they're,
probably
like
it's
covers
much
bigger
region
than
that
we
have
the
data.
So
this
was
this
max
distance
is
handling
that
so
rather
than
using
10
kilometer.
So
what
it
does
it
it
calculates
for
each
of
the
points.
10
000
is
the
the
distance
that
we
want
that
we're
interested
in.
That's
what
we're
seeing
so
here.
Let's
say:
if
you
use
a
thousand,
then
that'll
look
like
that.
B
So
now
it
looks
much
more
reasonable,
like
it
can
increase
a
little
bit,
let's
say
2000,
and
that
looks
like
that
so
compared
to
acid
it,
but
it
looks
much
more
pleasing.
So
it's
just
a
matter
of
how
you
set
the
parameters
and
make
it
kind
of
a
reasonable
plot.
So
let's
go
back
to
2000
and
now
you
can
play
with
the
again
the
k
news
point
start
two
looks
quite
choppy,
but
then,
if
you
increase
that
200,
that
looks
a
little
bit
more
pleasant.
B
G
So
yeah,
so
it
may
be
a
side
question
and
sorry
if
I
missed
it,
if
you
said
it
earlier,
but
so
so
for
this
inversion
result
you
you,
you
did
use
a
laterally
constrained
strategies
to
invert
the
sounding
correct
correct,
but
so
my
question
here
is
like
some
of
your
sounding
are
very
far
apart.
So
did
you
when
you
stitch
things
together,
it
was
was
it?
B
B
I
am
using
it
sometimes
mostly
for
visualization
so
like
this
is
a
tool,
a
function
that
you
can
basically
output,
that
resistivity
model
into
a
format
of
the
leapfrog
and
and
then
I
just
got
to
show
you
a
quick
example
and
the
model
that
that
we
plotted
looks
like
this.
So
this
is
the
type
of
resistivity
model
like
a
3d
view
of
the
resistivity
that
you
can
generate
with
the
leaf
rock.
B
I'm
pretty
sure
we
can
do
this
like
using
the
pi
vista
like
an
open
source
tool,
but
I
haven't
haven't
tried
yet
anyway.
Sometimes,
like
commercial
tools
has
pretty
visualization,
so
I
use
it
not
always
and
but
the
more
so
that
was
like
a
basically
just
plotting
the
resistivity
model
in
the
3d
space,
but
the
the
typical
way
that
that
we
used
to
plot
in
in
simpac
or
in
ubc
was
using
a
3d
graph.
So
you
can
actually,
rather
than
like,
applying
a
visual
value
at
each
sounding
location.
B
You
can
interpolate
and
plot
it
up
as
a
3d
grid,
so
that
that
model
object
has
an
internal
function
that
you
can
generate
the
3d
mesh.
So
if
you
do
run
the
get
3d
mesh,
it
runs.
I
think
the
default
number
of
x
and
y
is
100,
and
so
this
actually
is
fixed,
but
you
can
actually
change
if
you
like.
So
let's
say
if
you
said
that
the
x
is
400
and
ui
is
100
and
it'll,
not
40,
it'll
change
so
number
of
x
and
y.
B
So
you
can
kind
of
change
the
the
the
dimension
or
size
of
the
mesh.
So
I'm
just
going
to
use
the
400,
and
here
it
has
the
function
that
you
can
actually
generate
an
interpolation
matrix.
So
what
I'm
doing
I
generate
an
interpolation
matrix,
then
you
can
actually
reuse
it
for
many
resistivity
models.
So
again
I
think
it
actually.
It
uses
the
inverse
distance
rating.
A
B
B
So
if
you
got
so,
you
got
each
layer,
so
you
do
a
2d
interpolation
using
the
same
function
that
I
have
shown.
So
that's
what
was
that's
the
reason
why
why
it
asks
number
of
points,
so
it
calculates
like
an
inverse
distance
weight
and
then
average
over
to
that
2d
surface,
and
then
you
do
that
for
each
layer,
then
that
I
do
a
1d
interpolation
for
vertical
direction.
So
that's
that's
how
it's
set
up
and
then
you
can
actually
generate
some
sort
of
sparse
matrix
that
you
could
do
that
process.
B
B
Once
that
interpolation
matrix
is
generated,
then
you
just
need
to
pass
this
resistivity
vector
to
this
function
and
then
it'll
just
operate
that
matrix
vector
multiplication.
Then
you
got
the
resistivity
3d
model.
Okay,
and
this
function
is
the
same
purpose
of
masking
the
the
cells
that
are
far
away
from
the
sounding
location.
So
if
you
put
the
maximum
distance
to
3000,
it
will
only
plot
straight,
like
a
resistivity,
the
voxels,
within
the
3000
meter
distance
from
the
sound
location.
B
B
So
it
was
kind
of
like
a
it
for
the
plotting.
So
I
started
from
the
one
dimension
in
the
vertical
direction
and
expanded
that
to
the
two
dimension,
which
was
the
vertical
section,
and
then
we
plotted
it
as
a
plan
view
map
and
then
finally
plotted
up
the
3d
model,
just
the
actual
resistivity
value
that
we
recovered
from
the
like
from
the
inversion
and
then
interpolated
that
onto
a
three-dimensional
grid.
B
E
B
Yes,
so
so
here's
the
kind
of
the
the
github
repository
that
I
put
together
and
I
think
it's
slightly
different,
but
I
think
because
I
gave
a
talk
to
our
group,
but
basically
it
kind
of
answers.
Your
question
like
how
to
put
together
an
environment
that
could
replicate
sort
of
the
jupiter
notebook
that
I
have
shown
today.
So
yeah,
I
think
all
you
need
to
do-
is
go
into
this
github
repository
and
look
at
the
environment
profile,
and
then
you
can
actually
use
the
conda
to
replicate
the
version.
B
B
B
So,
but
what's
provided
from
this,
whatever
the
past
project
is
a
single
resistivity
model,
that's
often
the
case,
but
the
fact
that
inversion
process
is
non-unique
means
there
are
many
many
different
ways
to
do
the
model
we
can
fit
the
data,
but
what
we
have
seen
is
a
single
existing
model,
so
there
is
always
chance.
You
can
actually
go
back
to
the
raw
data
or
process
data
and
reinvert
with
the
different
sets
of
prior
information
and
more
different
types
of
code
and
regenerate
the
resistivity
model
for
the
purpose
of
your
project.
B
So
that's
where
you
need
to
actually
use
some
of
the
numerical
code
that
can
invert
the
data,
then
that's
what
syntag
em1d
code
serves
for
it.
So
here's
our
am
data
and
then,
when
it
supports
it,
can
actually
invert
and
obtain
a
resistivity
model
and
again
what
it
provides
is:
what's
called
spatially
constrained,
inversion
which
basically
use
the
1d
and
1d
simulation
but
stitch
them
together,
but
regularize
in
the
direction
like
in
radial
direction
of
each
sound
okay,
so
I'm
gonna
show
so
that's
like.
B
B
B
It's
a
so
we
got
132
datum,
and
so
we
only
got
single
line,
but
we
still
do
this
triangulation
so
tebow.
This
is
where
kind
of
what
you
kind
of
asked:
okay,
okay,
so
we
got
many
simplexes
or
vertices
or
edges
that
were
kind
of
like
kind
of
connecting
to
each
point
and
if
you
don't
put
anything
that
is
sort
of
we're
regularizing
through
each
of
these
edges,
we're
minimizing
the
difference
between
these
points.
Okay,
that's
what
so,
that's
the
regularization
that
we're
putting
in.
B
B
B
B
I'm
just
gonna
kind
of
the
line
of
triangulation
looks
a
little
bit
more
reasonable
and
the
tibous
point
was
okay.
Sometimes
you
may
not
want
to
kind
of
like
connect
these
these
lines.
Okay,
because,
depending
upon
the
scale
like
100
meters
way
too
far,
so
you
don't
want
to
recognize
it
say
the
structure
the
bit
like
a
the
correlation
length
of
the
structure
is
less
than
100
meter.
You
may
not
want
to
put
the
regularization
in
between,
so
this
is
the
function
that
you
can
use.
Let's
say
if
you
set
the
maximum
distance
to.
B
C
B
D
B
B
B
C
B
And
yeah
we
can
also
generate
this
three-dimensional.
There's
a
student
amount
that
I
would
plot
today,
but
that
that's
that's
how
it
is
and
final
thing
that
I
want
to
to
comment
so
that
the
sympathy
m1d
code
is
actually
now.
It
looks
like
almost
like
a
3d
code,
so
you
can
actually
use
the
same
survey
structure
that
you
used
for
a
3d
simulation
or
inversion
code
in
1d,
which
was
sort
of
the
major
kind
of
improvement
that
we
made
anyway.
That
was
it
for
the
inversion
and
yeah.
B
So
that
was
it
and
I'm
kind
of
for
the
context
of
developing
the
that
en1d
code
yeah,
I'm
planning
to
develop
kind
of
a
functionality
to
address
the
arbitrary
shape
of
the
loop
source.
So
currently
it
only
it
can
handle
the
circular
loop
and
the
vertical
magnetic
dipole.
So
I'm
going
to
expand
that
to
I
can
handle
the
arbitrary
shape
for
majorly
kind
of
for
a
ground
loop,
em
survey,
and
another
thing
that
I
want
to
do
is
inverting
for
the
layer
thickness
as
well
as
the
skill
value
and
the
third
one.
B
D
A
Dougie,
do
folks
have
questions
and
we're
we're
on
the
hour,
but
if
people
want
to
stick
around
and
chat
you're
more
than
welcome
to
fair
enough
well
I'll
stop
the
recording,
but
if
people
do
want
to
stick
around
and
chat
you're
welcome
to
thanks
everyone
for
coming
and
big.
Thank
you
sagi.
That
was
a
really
nice
walk
through
and
really
practical.
I
really
enjoyed
this
style
of
seminars,
so
I
hope
that
inspires
others
to
maybe
maybe
show
us
some
share.
Some
work
pull
up
some
notebooks
and
code.