►
From YouTube: SimPEG meeting July 30
Description
Progress on incorporating the simulation class into the potential fields codes, discussion on receiver classes, parallelizing the natural source EM codes and generating an example that uses GemPy to construct a geologic model.
A
A
C
B
Well,
maybe
we
can
start
with
quick
reports
from
bolts,
I
guess
I
can
go
first,
just
a
quick
update
is
I
showed
out
first
to
dong
thanks
for
reviewing
this
PR.
This
is
so
the
release
of
simple
Oh
point
eleven
point:
six.
This
includes
updates
just
basically
breaking
up
the
regularization
into
multiple
different
files.
So
originally
it
was
one
ginormous
file
and
now
we've
broken
it
up
into
basically
a
folder
with
multiple
files.
So
it's
a
little
easier
to
navigate
and
then
that
makes
it
a
little
easier
for
new
contributions.
B
So
J
who
joined
us
last
time
is
gonna,
be
starting
a
pull
request
soon
with
some
of
his
work
for
including
cross
gradient
terms
in
the
regularization,
as
well
as
some
directives
to
deal
with
to
deal
with
joint
inversions.
So
hopefully
this
will
help
simplify
that
a
bit
for
him,
so
that's
kind
of
cool.
That's
basically,
my
my
only
update
at
this
point
are
there
other
folks
who
want
to
jump
in.
A
Also,
I
guess:
I've
also
tried
to
run
a
little
couple
simulations
with
with
a
single
like
or
yeah
electric
electric
field
receiver,
and
then
mobile,
magnet
or
magnetometer,
and
it
seems
that
the
inverted
then
the
data
inverts,
but
it
seems
that
it
pulls
the
anomaly
like
I
just
have
a
block
and
it
pulls
it
to
whatever
direction
that
you
feel
the
sensor
is
our
station
is.
Is
that
something
that
I
would
adjust
in
sensitivities?
Or
would
that
be
a
form
or
like
the
way?
The
forward
modeling?
Is
it.
B
Shouldn't
be
before
modeling
I
mean
we
can.
We
can
run
some
extra
tests
and
things
like
that
to
try
and
make
sure
that
it's
it's
doing
what
it
should
be
doing,
but
assuming
that
the
forward
modeling
is
correct,
then
that's
something
that
you
could
do
like
incorporating,
potentially
a
sensitivity
weight
in
the
version
like
in
the
regularization.
A
lot
like
the
Dom
has
been
up
to
you.
Okay,.
A
B
A
D
It's
a
bit
convoluted
though,
because
you
kind
of
need
to
create
them
and
kinda
need
to
create
them
by
default.
Otherwise,
I
love
it
check.
So
is
it
okay,
if
I
always
instantiate
dictionaries
with
rules
are
all
true
by
default
on
the
receiver
class?
Does
that
make
sense?
So
if
you
you'd
like
a
list
of
strings,
they
will
just
rather
create
a
dictionary
with
all
active
receivers.
I
think.
B
That's
fair,
yeah
I
think
that's
fair.
At
the
very
least,
it's
a
good
starting
point,
and
if
we
want
to
refactor
later
we
should,
we
sure
can
but
I
think
that,
like
working
under
this
sort
of
the
most
general
assumption
of
like
we
expect
the
dictionary
coming
in
of
active
receivers,
always
that's
that
seems
reasonable.
D
B
D
B
Think
it's
like
so
there's
nothing
right
now
that
I
think
is
critical
to
that
it
could
be
helpful
down
the
line
for.
B
D
D
B
That
should
always
be
the
same
order
as
what
you've
provided
in
the
survey,
so
we
should
always
be
able
to
unwrap
it
if
we've
got,
for
example,
single-source
than
we
have
a
receiver
list,
and
then
each
receiver
has
like
a
component
dictionary.
Oh,
this
is
confusing,
so
this
they're
the
two
things
going
on.
If
we
just
had
a
component
list,
then
then
this
all
holds
because
that
list
is
ordered
right,
and
so
you
would
unwrap
it
based
on
the
order.
What
we
actually
need
to
be
doing
for
this
is
relying
on
order
Dix.
C
B
Update
we
will
have
to
actually
be
strictly
on
the
most
recent
release
of
Python
or,
like
sorry,
3,
6
or
3.
7
I
think
it's
3
thanks
3
6.
So
in
this
case,
if
we
have
an
ordered
dictionary,
then
each
of
the
keys
is
is
in
whatever
order
you
provided
them
in.
So
if
we're
doing
an
e/m
survey
and
I've
got
X,
Y,
I'm,
sorry
X,
real
X
image,
Y
real
Y
image,
that's
all
I'm
gonna,
listen
to,
and
then
each
of
those
has
its
own
set
of
rules.
B
D
D
D
B
B
B
Yeah
I
think
we
should
sketch.
This
out
is
maybe
just
like:
let's,
let's
sketch
out
what
we
actually
want
to
be
doing,
cuz
I
don't
actually
know
it's
like
in
the
code.
We
want
to
be
using
and
P
dot
wear
or
not,
but
conceptually
that's
that's
what
should
be
going
on.
Okay,.
D
C
D
B
Then
I'm
dumb
do
you
have
any
updates
at
all
on
the
work
that
you've
been
doing
with
Jim
I
cuz
I
know
you
were
poking
around
trying
to
get
that
up
and
running?
Were
you
successful
or
is
that
still
under
construction
or
yeah
yeah.
D
D
B
B
You
be
willing
to
start
a
pull
request
like
with
an
example
in
this
sim
peg
docks,
so
that
one
like
we
those
examples
we
sort
of
expects
like
run
as
a
Python
script
as
well.
So
we
wouldn't
necessarily
like
need
to
use
I,
PI
widgets
and,
like
turn
it
into
a
fancy
simulator
just
to
show
how
to
make
the
dot
initial
connection.
Do
you
think
that
would
be
a
easy
thing
to
do?
Yeah.
B
Not
into
simple
but
just
in
the
docs,
so
we
have
like
a
we
can
parse
this
out.
I,
don't
know
if
it
actually
exists
right
now,
but
just
a
Doc's
requirement.
You
can
make
a
note
at
the
top.
That
just
says
this
example
requires
gem
pie.
Yeah
so
like
it
won't
become.
Simply
won't
depend
on
that
at
all,
but
it's
just
here's
an
example
of.
If
you
want
to
use
gem
pie
to
build
up
your
geologic
model,
you
can
do
that
ya.
D
B
E
D
C
E
E
E
Survey
campaign
so
yeah,
it
really
depends.
I
mean
there
are
you
know
things
that
could
go
wrong,
that
that
delay,
but
I
think
in
theory
at
least
we
should
have
what
a
few
results
by
then
to
present
yeah.
So
we
have
yeah
Lindsey
and
soggy
and
Doug
and
a
whole
bunch
of
other
people
on
on
the
team.
Of
course,
signing
off
on
that
I
guess
idea:
I
was
wondering
you,
you
have
that
small
comment,
but
I
think
it
merits
some
discussion
actually
I.
F
Yes,
there's
like
usually
it's
more
resistive
like
okay,
compared
to
like
an
aqua
toilet
like
a
clay,
so
that's
sort
of
like
a
usual
conception,
so
I
think
they're,
probably
some
exceptions
but
they're
okay,
somebody
would
read
like
Oh
was
a
conductor
than
they
people
might
scratch
like
their
heads.
Oh,
why
is
that?
Okay.
E
F
E
E
F
B
E
B
B
B
C
Have
a
few
PM
inversions
running
and
they
seem
to
be
reasonable
things
at
this
point.
Most
of
the
testing
I've
done
with
that
grounded
transmitter
is
just
with
collecting,
like
various
components
of
the
e
field,
data
converting
that
I
started
just
through
a
couple
examples
where
I'm
started
to
try
to
include
both
you
field
and
me
field
data,
so
we'll
see
kind
of
how
much
improvement
we
get
with
that
and
how
it
works.
C
With
the
one
inversion
that
I've
done,
it
seems
like
I've
kind
of
had
to
increase
and
I'm
like
floor
uncertainty
values
a
little
bit
on
both
the
E
and
the
B
components
to
be
able
to
kind
of
roughly
kind
of
fit
them
both
together.
So
I
haven't
really
dug
in
to
kind
of
wipe
that
is
like
I
can
fit
them
both
fairly
easily
on
their
own
put
the
even
if
you
feel
dated
and
together
and
birth
them
together,
it's
a
little
bit
more
challenging
to
fit
the
data.
C
C
Yeah
we're
getting
really
good
agreement
for
those
grounded
translators,
both
between
like
the
UBC,
the
octree
one
and
with
like
a
simple
analytic
expression
that
I
coded
up
to
be
able
to
handle
just
kind
of
this
finite
length
wire
in
a
full
space
kind
of
all
components.
So
B&B,
so
that's
I'm,
pretty
sure
we're
pretty
sure
the
port
model
is
in
reasonably
good
shape.
Excellent,
let's.
E
C
E
C
F
Had
a
I
wrote
that
my
shop
update
but
I
had
a
chance
to
go
USGS
in
Menlo
Park,
oh
yeah,
so
the
connection
that
I
have
was
named
Mickey
and
Daniel,
so
those
are
sort
of
the
potential
field.
People
but
I
think
they're
more
like
they're
more
on
that
geology
side,
so
they're,
geologists,
but
using
lots
of
gravity
and
magnetic
they
don't
they're
very
kind
of
suspicious
and
I.
Don't
like
critical
about
the
inversion
but
anyway
and
has
had
a
chance
to
meet
them.
F
We
got
some
airborne
data
which
shows
like
a
really
strong
remnant,
but
they
were
interested
to
have
our
aromatic
data.
So
they
got
regional
aromatic
data
but
much
like
a
much
greater
height.
So
it's
like
115
year
height,
but
what
we
have
was
like
15
meter
height,
so
it's
kind
of
cool
for
them
to
have
much
higher
resolution
so
had
a
chance
to
visit
and
yeah
I
got
a
little
bit
of
plan
to
collaborate
with
them.
B
F
F
That's
all
I
grabbed
net,
but
they
think
this
is
like
a
simple
dipole,
but
that
well
we're
interested
in
this
part
where
you
got
like
a
kind
of
low
anomalies
and
that's
where
the
basalt
flows
so
anyway
in
general,
that
those
are
their
interest
like
they
kind
of
they
think
by
sort
of
unraveling.
How
hope
is
all
flow
was
happen?
F
E
F
B
F
F
F
Of
the
Grady
unit,
which
one
is
great-
oh
oh
here,
yeah,
yes,
yes,
I
think
there
are
a
couple
of
seismic
lines
where
we
actually
know
where
the
base
is
yes,
but
not
everywhere.
So
yeah
I
was
thinking
just
like.
Do
a
very
coarse
grid,
inversion
just
to
fit
that
large-scale
features
then
basically
like
kind
of
using
that
model.
Remove
this
from
the
local
grid.
E
F
Sounds
those
that
was
the
thought
and
then
what
they
did
was
what
they
called
match
filter
so
that
looked
interesting,
so
I'm
not
sure
you've
heard
about
this,
so
they
that's,
basically
what
they
get,
what
they
called
match
filter.
So
this
is
like
large-scale
picture,
smaller
scale,
smaller
scale,
smaller
scale,.
F
C
F
That's
that's
the
northern
Sacramento
Valley,
this
northern
part
of
the
California,
and
then
this
is
the
remnant
like
duration.
So
most
of
them
are
here
and
then
there
are
a
few
outliers,
but
so,
but
we
actually
know
the
random
direction,
which
is
kind
of
cool
and
also
we
know,
there's
life
above
about
the
general
value
of
that
right.
I
like
a
moment.
B
B
Yeah
no
I
think
that's
it.
That's
a
good
call
is
we
can
set
up
like
a
discourse
and
then
maybe
stick
an
auto-reply
or
something
like
that
on
the
mailing
list.
If
we
decide
to
get
rid
of
it,
but
we
might
want
to
keep
it
around
for
a
bit.
It
seems
like
you
know,
having
that
as
an
avenue
of
communication
might
be
a
positive
thing
for
for
folks.
B
D
E
E
E
I'm
sure
there
must
be
some
publicly
available.
You
know
Sat
data
from
from
that
area
and
if
it
could,
could
it
all
be
useful,
then,
as
you
know,
my
friend
nominees
leaders,
just
like
the
GIS
analysts
and
that
that
tech
and
I'm
sure
she
like
she'd,
be
willing
to
devote
some
time
to
try
to
do
something
with
it.
And
it's
just
the
question
of
like
in
theory.
Is
it
could
any
of
that
be
usable,
I.
E
F
Yeah,
so
that's
like
it's
more
like
a
kind
of
machine
learning
types
of
problem,
I
guess
like
where
you,
but
then
you
need
some
labeled
data
like
okay,
where
where
did
you
find
a
she
aquifer
right?
If
you
don't
actually
know
like,
you
can
have
like
a
multiple
data
layer
and
then
you
need
some
sort
of
label
data.
Then
you
can
tell
oh
here:
I
got
like
a
low
soil.
Moisture
high
ends
are
a
large
conductivity
or
something
like
that.
I'll
give
you
an
aquifer.
F
Don't
honestly
I
don't
know
because,
like
there
are
tons
of
possibility-
and
this
could
be
very
like
a
site
bent
and
I
mean
it's
the
same
problem
for
mining
vectors
right,
like
you
got
a
bunch
of
maps,
can
we
find
the
mining
war,
but
like
that
really
depends
I,
guess
and
I'm,
not
sure
how
successful
it
is
at
the
moment
on
how
many
people
are
choosing
not
not
quite
sure,
but
that
usually
uses
yes
like
of
the
satellite
data?
Is
that
water
balance
like
they
want
to
know?
F
But
I
think
that
looking
at
the
in
SAR
and
the
soil
moisture
could
could
actually
give
you
some
good
ideas.
I
guess
like
some
ideas,
I'm,
not
sure
you
can
defined
an
aquifer,
but
it
still
can
you
do
lots
of
information
cuz.
They
particularly
like
in
Sardinia
that
liked
it,
because
if
you
look
at
the
monitoring
data
of
insert,
then
that
gives
you
how
much
like
a
deformation
happen
and
that
could
be
correlated
with
some
sort
of
clay
and
the
pumping
so
yeah
that
dead
girl.
B
Then,
in
that
case,
to
I
think
it's
a
bit
of
a
question
of
scope
and
like
what
also
other
tools
in
the
ecosystem
are
doing
so
right
now,
actually
like
within
the
Pangaea
ecosystem.
There
are
a
lot
of
tools
for
for
remote
sensing
and
so
like
taking
a
look
at
some
of
those
and
seeing
how
we
can
better
interface
to
them.
B
Think
that
that's
that's
a
really
good
place
to
look
for
how
we
might
sort
of
interface
and
it's
great,
because
it's
also
all
like
core
Python
and
they're
deaf,
friendly
x-ray,
friendly
and
so
as
long
as
we're
sort
of
playing
in
that
same
space.
Then
it
should
be
pretty
easy
to
like
work
with
those
types
of
datasets
and
also
plug
into
simpler,
whatever
pieces
you
need.
There.
E
B
Start
I
mean
so
I
dropped
a
blog
link
in
in
there.
I
would
just
start
poking
around
the
pen,
geo
blog
and
their
website,
and
then
a
lot
of
those
will
point
potentially
to
other
papers
or
to
the
software
into
the
examples
that
beareth
up
there
using
which
are
actually
like.
You
know,
I'm
useful
for
immediately
getting
up
and
running.
B
Any
other
discussion
topics
or
things
that
need
to
be
prioritized.
I
was
offline
for
a
few
days,
go
back
late
sunday
after
some
backpacking,
and
then
I've
been
in
a
workshop,
so
I
sort
of
trying
to
regroup
on
on
where
we're
at.
So,
if
there's
any
like
items
that
you
view
is
high
priority
for
issues
that
are
at
the
moment.
D
B
B
D
F
A
Really
I'm,
just
just
paralyzing
the
forming
of
the
jitter
getting
J
I,
see
so
I
kind
of
just
used
similar
to
what
Tom
was
doing
and
just
computing
every
source
in
parallel
kind
of
thing,
so
every
column
of
the
back,
sorry
every
vector
for
that
hole,
mate
receipt.
And
then
you
can
do
whatever
you
want
after
we're.
To
sum
it
up,
get
jtj
or
anything
like
that.
A
F
F
B
A
B
A
D
A
D
D
B
Excellent
any
last
thoughts,
comments,
question.