►
From YouTube: SimPEG meeting March 25
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
Basically,
over
the
last
week,
I've
been
I
after
everything's
dead
settled
in
here.
I've
done
a
few
like
I
hit
up
that
I've
been
trying
to
take
steps
to
get
the
to
serialize
the
surveys
for
DC.
The
DC
survey
object
it's
still
involved,
but
part
of
that
I've
taken
a
few
steps
forward
with
it
just
trying
to
unify
the
2d
and
3d
dipole
receivers
and
sources,
because
there's
there
was
essentially
nothing
separating
them
in
the
code
base.
So
I
just
did
work
to
combine
them.
B
D
B
C
A
D
B
E
B
So
I've
been
thinking
about
and
playing
around
with,
actually
doing
a
3d
by
taking
into
account
three
locations.
If
you
do
a
2d
simulation
there,
just
a
few
little
places
in
the
code
that
need
to
be
modified,
some
sort
of
them
would
also
have
to
add
an
option
about
the
profile
like
the
direction
of
the
profile
or
just
automatically
coming
up
with
that
direction.
E
C
E
B
B
B
E
D
Height
I'm,
like
so
like
Daniel
Garth,
is
what
they
pull
like
a
four-month
slipping
regression
and
we're
absolutely
having
a
hard
time
so
he's
like
staying
up
until
three
to
four
to
threes
and
and
just
stick
through
until
night,
eleven
to
twelve.
So
we're
trying
like
what
we
could
do
to
change
that
so
far.
It's
not
working
well,
but
he's
going
well
and
like
you
just
be
healthy,
so
we're
not
waiting
too
much
in
additional
fee.
D
D
So
yes
yeah
mostly
just
looking
at
the
data
and
then
figure
out
whether
that's
connected
or
not,
but
later
I
think
I'm
planning
to
kind
of
quantify
like
and
make
it
a
little
bit
better.
More
humanik
'el
way
to
figure
that
out.
But
at
the
moment
I'm.
Just
learning.
Look
at
the
hydrological
data
because.
B
D
B
D
Yes,
that's
that's
the
next
step
at
the
moment,
we're
focusing
on
the
what
elbonian
can
do
so
we're
we're
like
a
really
digging
in
like
we're
drilling
down
to
what
we
can
really
tell
by
using
a
bunch
of
physics,
considering
the
all
the
assumption
that
we've
made
and
comparing
that
with,
like
the
ground,
real
ground,
truth
data
from
the
what
like
the
flow
data,
then
next
step
is
coupling
the
airborne
geophysics
with
that
with
the
guatemalan
or
remote
sensing.
So
that's
that's
the
next
phase,
but
we're
we're
haven't,
got
there
yet.
A
A
F
So
then
I
went
to
same
page
two
and
a
half
D
with
just
inverting
for
the
voltages
and
everything
seems
to
be
coming
out.
Pretty
nice,
so
yeah
I
kind
of
spent
the
week
on
that
just
well
in
my
solutely,
when
I
would
didn't
have
to
do
the
running
around
setting
that
up
for
everybody
at
the
office
and
yeah.
D
D
John,
that's
an
interesting
example,
because
the
typical
data
type
for
the
IP
is
a
parent
chargeability.
So
that's
where
you
have
a
problem,
because
if
the
clarity
is
different
and
like
from
DC
and
IP,
which
would
not
happen
that
much
more
there,
probably
some
cases
like
very
few
cases
like
that
in.
D
F
D
E
Like
the
coat
part
I'm
just
doing
more,
gooeys
I'm
gonna
give
a
demo
next
week,
I
think
on
the
9th
next
week
or
in
two
weeks
to
do
their
1d,
the
stitch,
1d
inversions,
so
yeah
and
now
I'm
still
involved
with
Diane
on
the
on
the
carbon
sequestration
project,
and
then
they
have
like
they
wanted
to
invert.
So
many
things
so
now
I'm
working
on
GUI
so
that
they
can
invert
or
aromatic
and
gravity
themselves.
E
D
E
G
E
G
E
G
And
there's
also
a
it's
like
a
magnesium
and
an
iron,
so
there's
kind
of
some
places
in
that
that
that
mineral
that
can
have
magnesium
and
have
iron,
and
if
you
have
more
of
the
iron
in
a
certain
oxidation
state,
you
get
a
strong
magnetic
signature.
And
then,
if
it's
more
magnesium,
you
don't.
You
know.
E
G
E
That's
the
idea
and
I
posted
a
link.
If
you
guys
wanna
I,
want
to
see
what
the
what
kind
of
degrees
for
this
in
Peggy
I'm
Wendy's,
going
to
look
like
I
had
to
record
that
for
for
training,
and
then
you
can
download
the
notebooks
and
and
the
data
set.
If
you
guys
want
to
spin
up
and
then
yeah,
that's
it
so
knowing
you
dev,
but
what's
up
synthetic
still.
F
B
E
G
G
A
G
Is
a
be
secondary
and
I
was
trying
to
go
with
that
I
haven't
I've
been
I
kind
of
run
into
some
hiccups.
So
the
least
what
I've
got
going
so
far
is
in
the
tutorials
folder
under
temporary.
So
it's
available
to
anyone
to
to
maybe
see
if
I
made
an
easy
mistake,
but
I
guess
one
thing
I
was
noticing
was.
A
She
actually
make
I
mean
a
huge
difference.
The
where
you'll
see
differences
is
things
are
discretized
opinion,
so
looking
at,
for
example,
an
EB
formulation.
If
you
were
looking
at
the
magnetic
flux,
if
you're
measuring
beef
you're
gonna
be
sensor,
it's
gonna
be
most
accurate
if
you're
closest
to
the
faces
yeah.
As
with
the
HJ
formulation,
the
B
field
will
be
most
accurate
when
you
are
close
to
edges
yeah,
and
so,
if
you
actually
use
the
same
mesh
and
measure
in
the
same
exact
locations,
there
will
be
numerical
interpolation
differences
between
those
two
solutions.
A
Yeah
now
which
source
is
discretized,
where
also
is,
is
slightly
different,
but
there's
no
sort
of
it
doesn't
strictly
mean.
You
know
if
you're
working
with
conductivity
or
resistivity,
that
you
need
to
be
using
one
of
the
other
yeah,
all
the
machineries
built
in,
so
that
you
can
use
either
either
with
each.
The
biggest
difference
is
just
where
stuff
is,
is
discretized
and
that
really
matters.
Unless
so
much
for
the
3d
case,
it
shouldn't
be
hugely
significant.
A
If
you're
using
grounded
sources
it'll
be
a
bit
different
because
like
where
the
current
density
is
flowing
is
different.
So
like
in
the
eb
formulation,
you
have
basically
a
wire
that
is
defunct
on
edges
to.
C
A
Finite,
so
a
footprint
of
your
source,
whereas
if
you
have
you're
using
the
HJ
formulation,
you're,
actually
defining
current
density
going
through
phases.
So
it's
basically
in
a
sense
of
bigger
footprint
for
your
source.
A
G
A
G
H
G
You
know
it
sounds
like
there's
some
interest
that
people
have
in
actually
using
MT
in
the
near
future
and
using
simpie
to
do
it
and
so
yeah
testing
and
seeing
if
we
can
find
a
way
to
have
it
be
very
usable
for
anyone
else
in
the
near
future.
We
have
a
lot
of
stuff
that
we
can
test
it
against
now,
I
mean
I've,
I've
got
field
data
sets
for
Mt
and
for
Zed
Tim.
G
H
Okay,
yeah,
that's
that's
great
I,
think
empty
is
surfacing
actually
from
a
lot
of
different
perspectives.
I
was
just
talking
to
deke
on
yesterday
and
he's
actually
very
interested
in
in
using
MT
and
they're
trying
to
develop.
You
know
their
own
kind
of
empty
codes.
Well,
actually
what
they
were
trying
to
do
was
to
use
ma
diem
and
then
modify
that,
but
he's
found
that
you
know
particularly
frustrating
and
so
I
was
asking
about
whether
he
was
fine
to
use
simp
Akande.
He
I
guess
he.
H
H
C
H
I
think
I
think
pushing
forth
on
that
would
be
really
good
and
Lindsay
has
put
out
a
document.
She'll
probably
talk
about
that
in
a
bit.
I
think
I
think
focusing
on
that
MT,
which
we've
been
talking
about
now
for
for
years
and
years.
But
you
know
especially
now
with
Jo
here
and
you
know,
sniffing
up
the
codes,
the
serialized
things
and
to
look
at
you
know:
combinations
of
different
meshes
for
different
frequencies
and
just
putting
all
the
stuff
that
we
kind
of
have
in
bits
and
pieces,
but
just
putting
that
together.
I
think
it's.
G
And
I
think
we
need
to
we'll
need
to
I
think
develop
and
add
a
couple
of
utilities.
One
thing
is
that,
if
you're
working
in
areas
that
have
significant
topography,
you
have
to
go
and
shift
the
locations
of
all
of
your
your
stuff
so
that
it's
defined
on
the
surface
of
the
disparate
eyes
topography.
Otherwise
you
start
getting
electric
field
measurements
in
in
air
cells
and
that's
going
to
cause
some
problems.
G
G
The
forward
modeling
is
I
would
like
to
simulate
some
kind
of
geological
scenario
so
that
I
can
learn
about
it
and
I'm
going
to
come
up
with
a
survey
and
then
I'm
going
to
have
to
create
a
mesh
and
I
might
have
topography
and
just
how
would
I
go
through
all
of
those
things
and
each
each
went
as
you.
You
go
from
what
you're,
starting
with
to
what
you'd
like
to
end
up
with
you
find
out.
There
is
a
certain
set
of
consistent
steps
that
you
need
to
do
to
be
successful.
G
G
So
I
guess
I
was
going
to
start
with
yeah
start
with
some
basic
typography
start
with
some.
You
know
basic
idea
of
a
survey
that
I'd
like
to
do
and
then,
when
I
hit
a
hiccup,
where
I
find
a
that
I
don't
have
a
utility
that
I
need
then
come
up
with
a
solution.
We
figure
out
a
solution.
Okay,
move
to
the
next
thing.
Hopefully
things
go
smoothly:
oh
maybe
there
we
could
think
about
this.
That's
sort
of
how
is
planning
on
attacking
that
in,
like
the
immediate
or
moderate
future
perfect.
C
B
You
have
to
develop
bunch
of
time
to
like
oh
I
need
to
come
up
with
this
on
my
own,
like
oh,
there
might
be
something
somewhere
else
like
I
know,
domme
is
kind
of
had
that
experience
with
me
a
couple
of
times
ago.
Is
there
somebody
like
yeah
so
yeah
the
other
comments
about
empty
stuff,
I
figured
Lindsey
was
going
to
talk
about
it
later.
So,
okay.
H
A
Excellent
yeah
you
give
a
couple
updates:
I've
been
working
on
getting
an
abstract
together
for
s
e-g,
just
based
on
the
ux
old
machine
learning
work
that
I
did
for
the
UBC
presentation,
so
I'll
just
share
that
once
it's
in,
like
a
semi
complete
form
right
now,
it's
in
very
draft
stage,
so
yes
I'm
grateful
that
there's
sort
of
extra
time
on
that
and
then
the
other
thing,
as
many
people
have
alluded
to,
is
the
empty
side
of
things.
So
I
just
dropped
a
link.
A
I
think
I
shared
this
last
time,
but
Johnny
weren't
weird
on
the
colas
time
you
were
away
in
the
field
and
so
okay
I'm.
A
A
They
could
serve
any
sort
of
aspect
of
simple
work.
So
if
there's
something
that's
like,
for
example,
you
know
having
a
simple
data
viewer
for
four
different
types
of
data,
or
had
that
you
could
just
like
double
click
on
from
Jupiter
lab,
rather
than
having
to
you
know,
launch
a
notebook
or
things
like
what
sorts
of
things
would
you
actually
like
to
sort
of
watch
as
an
inversion
runs?
Joe
Hammond
has
done
some
really
cool
work,
basically
connecting
czar
and
Redis.
So
Redis
is
like
a
real-time
data
server.
A
So
actually,
as
an
inversion
is
running,
you
can
basically
say
just
like
hang
on
to
these
live
data
for
me
and
that
can
be
displayed
so
things
like
looking
at
gradient
steps,
for
example,
or
if
there's
any
other
intermediate
information
when
an
inversion
is
running.
That
would
just
be
like
helpful
to
actually
have
have
a
look
at
and
potentially
even
make
decisions
down
the
line
it's
like.
Oh,
it
looks
like
I
set.
My
you
know.
A
My
smoothness
is
not
high
enough
that
it
step
that
it
just
took
is
crazy,
like
let's
just
kill
it
now
and
and
revisit
yeah.
So,
if
there's
anything
like
that,
that
you
kind
of
come
up
with
or
think
about,
feel
free
to
add
that
to
this
document
and
we
can
try
and
sort
of
prioritize
what
some
of
those
things
would
be.
The
other
thing
I
would
really
like
to
see
is
like
the
idea
sort
of
prototyping
gooeys.
So
we've
done
some
of
that
with
the
widgets
work.
A
But
thinking
about
you
know
what
else
could
we
be
doing
to
basically
be
prototyping
or
like
building
sort
of
research
who
is
but
then
also
potentially
prototyping,
some
stuff
that
it
like
Don
you'd,
include,
maybe
an
analyst
or
something
like
that
yeah?
So
that's
sort
of
like
the
Jupiter
connection,
but
then
also
there
is
just
the
work
of
improving
the
code
base.
So
I
was
kind
of
thinking
just
based
on
this
conversation
today
that
you
know
translating
some
of
this
into
a
discourse
post
might
be
a
good
first
step.
A
C
B
I
just
want
to
say,
I
didn't
working
on
another
project
that
was
using
the
Mt
canola
like
Devin
was
talking
about
like
it
was
rough
getting
going
with
it,
because
there's
such
a
lack
of
the
documentation
on
it
I
feel
like
I
got
a
good
handle
of
it.
I
was
kind
of
doing
some
local
mesh,
interpolation
and
volume
averaging
interpolation
to
do
some
big
3d
modeling
on
this
fun.
B
Putting
day
a
bunch,
they
built
there's
really
nice
like
3d
model
and
they're,
doing
a
bunch
of
tests
and
modeling
with
that
and
seeing
if
we
can
get
more
information
out
of
it.
I
use
everything
together,
so
I
was
doing
MT
modeling
on
that
and
kind
of
playing
off
that
and,
as
part
of
the
inversion
side
of
it,
I
did
get
some
first
pass
scripts
at
building
the
sensitivity
matrix
already
it's
kind
of
efficient
way
to
do
it.
It
was
I
without
modifying
the
code
too
much.
B
F
B
A
B
I
do
think
that
let's
mention
another
thing
that
implementing
boundary
conditions
another
approach,
besides
a
primary
secondary
formulation-
probably
a
good
way
to
go
about
that
part.
If
we
can
get
2d
implementations
in
there,
I
think
that
would
be
nice.
If
maybe
we
can
look
at
switching
the
1d
implementation
over
to
more
of
an
semi
analytic
approach,
the
same
way
that
the
1d
these
simulation
is
done
right
now,
I
think
I'd,
be
nice
you're.
H
B
To
2d
inversions,
like
that,
like
we
can
always
I'm
building
these
tools,
we'll
be
able
to
use
them
to
generate
potential
boundary
conditions
for
higher
order
things
as
well.
All
right,
so
I
think
that
they're
all
kind
of
it'd
be
nice
to
have
them
in
there
and
things
that
we
can
add.
But
right
now,
like
it's
a
primary
secondary
approach
to
it.
So
using
it
being
switching
it's
like
a
better
boundary
condition.
H
G
Okay,
so
I've
I've
I'm
actually
really
quite
happy
with
how
the
UBC
jiff
Fortran
codes,
work
and
I've
documented
the
inner
workings
of
how
how
they
do
their
modeling,
how
they
implement
the
boundary
conditions,
and
so
maybe
I'll
send
that
to
you.
That's
that's
it's
available
freely
for
anyone,
but
there
is
a
section
that
has
all
of
that
would.
A
Well-
and
one
thing
we
could
think
about
doing
with
the
primary/secondary
is
also
it
would
be
nice.
We
should
be
able
to
actually
have
that
as
a
source,
so
have
the
source
keep
track
of
if
its
primary
secondary
versus
not
and
then
go
from
there
like
that's
what
we
do
in
the
frequency
domain
code,
which
is
kind
of
cool
so
like
because
it
would
be
nice
to
hang
on
to
that
implementation.
Just
to
have
like
you.
C
A
G
I
think
this
is
kind
of
how
it
works
like
it
goes
in
it.
It
will
solve
a
whole
bunch
of
1d
problems
with
the
model
at
the
boundary,
and
then
it
will
then
use
that
turn
that
into
a
source
term
and
then
use
that
to
to
go
and
compute
the
fields
in
the
rest
of
the
domain.
I
think
that's
the
general
idea.
H
D
B
It's
a
it
essentially
goes
through
and
searches
for
overlapping
square.
You
know
rectangular,
prisms
and
rectangles
and
2d.
That
should
be
able
to
generalize
nicely
to
treat
a
tensor
mesh
I.
Think
I've
also
I
had
like
a
volume
averaging
branch
going
up,
discretized
that
has
current
implementations
of
tensor
master
tensor
mesh.
B
Which
I
kind
of
been
building
on
that
as
well,
then
of
the
only
step
like
I
said
it
just
get
hot
treated
actually
working
generally
Dom
has
a
small
example
of
that
with
the
with
his
tiled
map.
It's
just,
though,
that's
a
very
specific
one,
where
the
every
cell
in
right
and
the
local
mesh
has
to
be
in
the
global
mesh
I'm
trying
to
just
make
it
a
little
bit
more
general.
D
F
B
It
was
I
think
I
was
probably
toward
the
higher
range
of
frequencies.
They
were
more
concerned
with
characterizing
near-surface
anomalies.
Walk'n
I
mean
there
were
only
maybe
looking
the
model
I
think
had
a
maximum
depth
of
about
five
kilometers.
So
it
wasn't
that
I
didn't
have
to
go
that
deep
or
that
such
a
long,
wavelengths,
okay.
F
C
B
B
F
G
Yeah
I
seen
some
results
of
some
codes
where
they
say:
oh
yeah,
we
just
we
hooked
everything
in
and
I,
don't
know
what's
happening
under
the
hood.
Maybe
they've
got
something
really
really
fancy
and
spiffy,
but
if
you're,
just
gonna
invert,
everything
on
a
single
mesh
kind
of
I
always
thought
two
to
three
orders
of
magnitude
of
frequencies
and
subsample.
It
pick
pick
some
good
ones,
assign
proper
uncertainties
to
them.
G
A
G
Actually,
just
posted
the
building
that
jiff
tools
cookbook
website
right
now,
but
I've
actually
just
posted
a
a
full
comprehensive
tutorial
for
how
to
start
with
standard
MT
data
and
how
to
go
from.
Having
that
to
getting
your
final
inversion
result
and
all
of
the
things
you
would
do
to
understand
the
anomalies
and
interpret
the
data
and
what
decisions
you
should
make.
So
it's
gonna
take
maybe
like
20
or
something
minutes
to
build,
but
I'll
drop
that
link
as
well,
because
it's
kind
of
a
standard
practice
for
how
you
would
invert
empty
data.
C
D
G
G
E
D
D
Our
current
approach
is
like
basically
use
the
3d
code
and
then
let
the
near-surface
like
really
have
like
very
complicated
structures
to
fit
best
static
shift,
but
there's
another
approach,
but
people
are
using
like
in
like
so
your
size,
simultaneous
inverting
for
those
like
some
sort
of
coefficient
and
yeah
and
I.
Don't.
A
D
H
G
H
H
You
can
put
in
little
bits
of
conductors
here
and
there
in
the
near
surface
and
that's
going
to
sort
of
absorb
the
static
shift
effects,
so
I
I'm
not
sure
where
people
stand,
but
my
understanding
more
and
more
as
that's
what
people
are
doing
and
they're
not
actually
trying
to
compute
what
the
static
shift
parameters
are.
Alan
Jones
might
have
a
different
view
on
that,
but.
H
D
G
Then
you
also
you're
one
of
the
other
things
we
think
about
is
how
sensitive
the
these
cells
next
to
the
receivers,
are
that
sensitivity,
and
so
we
usually
do
this
interface
waiting
on
the
near
surface
cells
to
kind
of
I,
don't
know,
reduce
the
freedom
of
the
the
near
surface
to
be
able
to
be
somewhat
pixelated,
so
you're
kind
of
those
two
things
are
fighting.
You
want
those
near
surface
cells
to
be
able
to
get
these
small
scale
variations
in
electrical
conductivity
to
explain
the
static
shift
or
sort
of
to
account
for
that.