Hi I'm Lisa Stright. I'm an Assistant Professor in the Department of
Geosciences at Colorado State University. I want to thank AAPG Foundation for
inviting me to talk to you about my research today through the DL program.
So today I'm going to talk about template-based modeling, bridging the gap
between quantitative outcrop studies and subsurface reservoir characterization.
My background is a mix of engineering and geology, but what I focused my
research on is in reservoir modeling. In particular, I focused a lot on
converting what we see in an outcrop into models for the subsurface. So we have
better understanding of fluid flow and conductivity and seismic responses in
So as such, it's really important to have a good understanding of the sedimentology
and stratigraphy and the outcrops. And so my co-PIs on the project, the Chile
Slope Systems and Joint Industry Project Steve Hubbard and Brian Romans have
done the foundational work for this study and helping with the outcrop
characterisation that I use in my models.
Furthermore I'm going to be presenting a lot of work for my students and the
students of our research consortium today who are listed here past and present.
And finally, I'd like to thank the sponsors of this research. We're currently
finishing up phase two, or year six, and we're heading into phase three next
So what you're looking at is a drone photo that the Cretaceous Tres Pasos
formation in the Magallanes Basin of Chile. This is a series of stacked deep
water slope channels. So how did they get there? So what we can see in the next
slide here is the process of turbidity currents which carve and deposit sands
in these deep water slope channels.
Turbidity currents are sediment laden gravity flows supported by turbulence.
And they organize into channel levees over bank lobe deposits. In the left
image, what you can see is a Google Earth image of offshore Monterey,
California. So you can see Monterey Bay stepping back into the land there.
And you see a channel cutting into the shelf and then continuing out into
the oceanic basin. And then on the right, we see a shallow seismic image from
offshore Gulf of Mexico from Sylvester et al. in 2012. And the bottom right is
an artist's rendition of what a turbidity current looks like.
So the reason we're interested in them is because the ultimate deposition of
sand in deep oceanic basins eventually becomes a place of storage of oil that
we explore for these days. But one of the challenges of exploring for oil in
oceanic basins is that we have to use remote imaging and so oftentimes we're
dealing with low resolution seismic data, which is difficult to really help us
understand not only the architecture of the slope channels, but also to really
be able to see the bedding and the controls on fluid flow and connectivity in
So again, you see an image of seismic-- a cross section across a deep water
slope channel system in the upper image and a line drawing of that below. And
then on the right, you see an amplitude map showing where the cross section is.
So from the amplitude map, you can note that we can see a width of the channel
belt but not a tremendous amount of detail in channel location or even what's
inside the channel fill.
So we're going to be focusing in hierarchically on the basal building block
of these systems, which is called a channel element. Now the scale is about 250
meters wide by about 14 meters thick. And those channel elements stack into
channel complex, and multiple channel complexes stack into complex sets.
So the challenge again with seismic data is that we have a difficult time
really seeing what's inside these channel elements and even from the seismic
exactly how they're stacking. So we head to the outcrop and we spend some time
looking at the bed scale architecture and stacking patterns of channel elements
to get a better understanding of how to interpret the seismic data.
So this image on the lower right is of the same outcrops that we saw on the
first slide, the Tres Pasos Formation in the Magallenas Basin, with a student
looking at a correlation panel of those rocks. If we zoom in and look at what a
single channel element fill looks like, you can see it here, where we've got
thick amalgamated sandstones in the axis, or on the right, over to thin inter
bedded sandstone and mud stone on the left.
And then at the base of the channel, we have a channel base drape, which
tends to be a flow barrier. So let's take a look at a digital outcrop modeling
workflow for flow simulation and understanding how bed scale information
impacts fluid flow. So this is from the Williams Fork Formation, the Piceance
Basin of Colorado.
In the upper left, you see a LIDAR image and a photo mosaic of a fluvial
outcrop. Below that, you see a line drawing sketch of that fluvial outcrop
that's been gridded as an outcrop model. And on the right, you see varying
degrees of heterogeneity added into the model, with A being the lowest amount
of heterogeneity, and D including the most.
In order to understand how fluid flows impacted by this increasing degree of
heterogeneity, this author put a injector on the left and a producer on the
right to do a synthetic water flood through this 2-D model. And what you see is
that with the increasing amount of heterogeneity, and in this case in the
increasing continuity of shale drapes along each one of the surfaces in the
model, that there is a decrease in the ability of the water to flow from the
left to the right.
So that the take home from this is that more heterogeneity impacts fluid
flow, more bed scale heterogeneity. Furthermore, many authors do forward
seismic modeling from outcrop models to get a better handle on how bed scale
architecture impacts the seismic response. So this is an example from the
[INAUDIBLE] Formation in the Pyrenees.
And so what you see on the left is a model that's at the sedimentary unit or
the geobody scale. So it's showing different geobodies with homogeneous fill in
the geobodies. And then the model below that has more heterogeneous fill. And
so the idea with forward seismic modeling studies from outcrop is trying to
understand how that bed scale heterogeneity propagates throughout the model.
So at 75 Hertz, the images on the right, you see a significant difference
between a homogeneous and a heterogeneous model. But the challenge is, once we
start moving into lower and lower resolutions, that we often see-- our low and
lower frequencies that we often see in subsurface seismic data-- the images
start to look very similar. And not only that, but the ability to definitively
interpret that fine scale bed information is very difficult to nearly
So as a community, we're moving more and more toward thinking about outcrops
in 3D and doing 3D constructions of outcrops. And so this is an example where
we have a 3D digital outcrop model and surfaces interpolated between those to
generate a really detailed Facies model in three dimensions. Some of the
conclusions from these studies, whether it be 2D or 3D is for flow, that more
heterogeneity matters, and for seismic, that lower resolution-- there's lower
resolution or lower ability to see bed scale information on the low frequency
One of the challenges, however, is that the lengths to subsurface models is
largely qualitative. And it's difficult to make direct links to subsurface
fluid flow and subsurface seismic interpretations. So some of the questions
that I'm really interested in this type of research is understanding more about
that scale heterogeneity, when it matters, and then if it really does matter,
how do we include it in our models?
So that leads me to my objective. So the objective of this research is to
gain a foundational understanding of how bed to geobody skill architecture i.e.
the facies distributions, controls fluid flow and seismic response. We're going
to do this by investigating the fundamental building blocks of deep water
channel systems and these are the channel elements that I mentioned in the
first couple of slides. We're going to be using our observations, quantitative
observations directly from outcrops, into three dimensional outcrop models.
So where's the study area? So the study area is in the southern tip of South
America, in the Magallenas Basin of Chile. This is a Retroarc Foreland Basin
that was formed by the Andean orogeny. Now its sediment is fed into the basin
from the north to the south. And over the period of the lower to upper
Cretaceous, the basin filled with four to six kilometers of deep water
Now you see a 3D diagram showing the Tres Pasos and genetically linked
Dorotea Formation in the upper right hand side. So the focus of this project is
the Tres Pasos Formation. And we're looking at the slope channels of the
pro-grading slope system of the Tres Pasos Formation. So the full thrust belt
continued to propagate into the basin and uplifted the basin's and sediments on
their side, exposing the deltaics and the slope clinoforms and outcrop.
So what you see here is a DEM with the Landsat image draped on top of it.
The paleo currents are from north, the top of the image, to the south, the
bottom of the image. Now the blue lines are the deltaic top sets that directly
feed into the pro grading slope clinoforms. The clinoform we're going to be
focusing on today is the clinoform in red, which is called the Figueroa
clinoform. An analogous subsurface system is the Brookian succession in
offshore Alaska, with similar high-relief clinoforms and deep water slope
So this is a correlation panel from north to south again showing in green,
the deltaic top sets, undifferentiated because they have not been the focus of
this study, and then each of the interpreted slope clinoforms heading into the
basin. Now the brown is early Tres Pasos formation, where there's a lot of mass
transport deposits and ponded slope turbidites. We won't be talking about those
at all in this talk today.
So the red line there shows the Figueroa Clinoform, as I told you we're
going to talk about today. So we can see the top left of that clinoform goes
from the deltaics to the shelf break, down the slope to the base. And above it,
you see various sand packages. Now as we go from proximal to distal, Ben
Daniels has mapped out a different channelized outcrops from the north to the
And what I want to point out is that we see even along this length of the
clinoform, variable stacking patterns of each of these different channel
elements. Now the one benefit of having such a great seismic scale and regional
scale outcrop is that we can always put the channelized deposits in context of
position on slope and also evolution of the slope system.
So the outcrops we're going to be looking at today are in the lower most
image along the Figueora Clinoform. So we're going to start thinking about how
do channel stacking patterns in the internal architecture control flow and
connectivity as well as the seismic response?
OK. So what I'm going to show you is the construction now of the lower
Laguna Figueroa Outcrop Model. So what you're looking at here is a DEM with an
ortho-photo draped on top of it. So the dashed box is the lower Laguna Figueroa
Outcrop, and this is where we have vertically stacked, low sinuosity channels.
And so some of you familiar with Petrel might notice that this is a model that
we have in Petrel.
Within this box, we have thousands of meters of measured section. So as you
see in the photo pan below, in white lines and the colored lines on the DEM
model. In the beginning, I showed a picture of a student looking at the outcrop
with a correlation panel. This is the correlation panel where each of the faint
vertical lines are measured sections, and the yellow shows sandy channel
element fills. And the surfaces are correlations of bases of channel elements
from outcrop location to outcrop location.
Well you see there's 18 channel elements that were interpreted into three
channel complexes. Now this is critical for the modeling, and also for
understanding the evolution of the system, that from outcrop plate location to
the outcrop location, using paleo currents, this student mapped out plan forms
of each of the individual 18 channel elements.
Now these are the foundation for the 3D outcrop model. These were digitized
and put in to three dimensions as channel forms, as well as tied to the outcrop
data that we have. So this is an image of one of those channel element
surfaces. The colors on the surfaces at this location are simply just
elevation, where purple is the deepest and red is the shallowest. Each channel
element as I mentioned the beginning, is about 14 meters thick.
When we put all 18 channel elements together, and I'm now calling this a
model, so this is 18 channel elements as geobodies. So we have four channel
elements in the bottom most channel complex, 10 in the middle, and four in the
top most complex, which is in blue.
So the question is, how do we actually fill those geobodies? Well that's where
we go to the statistics of the data. We've gone in and digitized all the
measured sections from this outcrop location and worked on coming up with a
digital model that's representative of this outcrop. Now what we see is, it's a
relatively low sinuosity system, about 1.05 in sinuosity.
And we see relatively constant fill throughout the outcrop. So the
architecture-- the internal architecture of the channels-- isn't really
changing. With the digital data, we created something called vertical facies
proportioned curves. And so you see these over on the right hand side, where
you can see categories of axis off-axis, and margin. Axis is the center of the
channel and off axis is between axis and margin. And margin is the edge of the
So the vertical axis shows elevation off of the base of the channel. And the
horizontal axis shows the proportion of facies encountered in the measured
sections at each of those locations. And so what you can see as the axis is
primarily filled with highly amalgamated sandstone facies. But there are some
finer grained facies at the base, which we would call a draping facies.
So there's two different kinds. There's a facies that's kind of more gray
and red, that would be a barrier to flow, or the channel base drape. And something
that's a little more green and orange, which would be more of a baffle to flow.
And you can see that the predominance of these finer grained inter-bedded
facies increases as you move out into the margin.
So the model on the bottom left is a template that we generated using this
digital data to represent what we see from these outcrops statistics. And this
is going to be the template going forward to fill that outcrop model. And
you'll find that this the channel base drapes, so the facies that's in black,
plays a leading role in the results in the modeling that you'll see coming up.
So this is an example of what that channel base drape looks like. So this is
a particularly sandy section. We've got a lot of very high net to gross. But in
the middle of it, you see a rather thick section of inter-bedded sand shale,
but very, very, very fine grained. So in a flow setting, this would be a flow
barrier. And again, you can see it's actually rather thick compared to the two
people for scale on the bottom right hand side. So we want to capture this as a
flow barrier in our models, and make sure that we're understanding how it's
going to impact flow and connectivity between successive channel elements.
So this is what the final model looks like once we put that internal
architecture template into the model. Now the final model is actually quite
large. It's 600 million cells. And the grid cells are two meters by two meters
by 25 centimeters. And they're that fine to be able to capture that bed scale
information in the model.
So before going into this entire, complicated, 600 million cell model and
starting to try and understand connectivity and fluid flow and seismic
response, what we're going to do is deconstruct it to its very basic building
blocks, the channel element. And first of all, we're gonna look at fluid flow.
So we'll look at a two channel element case, and we're going to look at fluid
flow in this two channel element case. With those two channel elements and
they're detailed internal architecture, we'll look at stacking in a laterally
offset configuration, bottom right, up to a vertically aligned configuration,
And then, in the flow simulation experiment, we're going to inject water in
the lower left of the model, and produce oil in the upper right hand side of
the model. Now this isn't a realistic fluid flow experiment. But the idea is we
want to intersect the internal architecture and the stacking patterns with as
much of the fluids as we can to get an understanding of whether it, architecture,
or stacking patterns are controlling more of the fluid flow.
What we'll look at is a cross section at the location of the oil producer at
water break through time. Now we're going to look at the six different stacking
patterns of these two channel elements next. So first of all, this is the
laterally offset stack case at water break through time where blue is water and
red is oil.
So what you see is that drape facies plays a predominant role in controlling
where the fluids go or where they don't go in this flow simulation model.
There's a very small connection between the left channel and the right channel.
Now as we swing that right channel up into more of a vertically aligned case,
so what we see is that the axial facies, once juxtaposed right one on top of
another, they act as a thief zone.
So we have less of a production from the marginal facies, and more rapid
production from the injector, connection from the injector to the producer and
bypass of those marginal facies. Interestingly enough, this laterally offset
stack case has the highest recovery efficiency, because we're producing more of
the oil that's stored in those marginal facies.
However, the water break through time is a little bit slower than this
vertically aligned case. So the thing that's more and most important for
thinking about subsurface modeling is this, that when we start thinking about
building coarser grid cell models, that heterogeneity matters more and that
therefore, potentially finer grid cells in the laterally offset stacked case,
where those fine grained facies play a larger role in controlling fluid flow
than in the vertically aligned case.
So now if we look at all of the fluid flow results, with the case that we
just talked about in the left column here, where we've got water breakthrough
time in days, recovery efficiency, cumulative oil production, and pore volumes
injected shown. The heterogeneous models, so the models with the variable
internal architecture as observed from outcrop and from the template that we generated,
the results are shown in red. In comparison, there is a homogeneous model shown
in gray with similar or exact pore volumes as what we have in the other models.
What you see, again, is very similar to the results that I showed on the
last slide, where in the vertically offset case, or the vertically aligned
case, that the fluid flow simulation results are not that dissimilar to a
homogeneous case and that as we start to move to a position where the channels
are laterally offset stacked, and the marginal facies begin to play a stronger
role, we start to diverge in are our simulation results from the homogeneous
case. Similarly, in the middle column, we can start thinking about facies
So again, these channels aren't completely straight. And as we add a little
bit of sinuosity to the channels, we can imagine that the higher energy
portions of the flow would push sand over to one edge of the channel versus
another. So this is looking at the question of where are those marginal facies
positioned? Does the position of the marginal facies matter, even with the same
proportion of marginal facies? What we see is with asymmetry, we're
exacerbating the difference in the laterally offset stacked case, between the
homogeneous model and the heterogeneous model. So the positioning of those
marginal facies matters.
OK, so now if we can go look at the full field model and start thinking
about more complex stacking patterns and facies juxtapositions across channel
elements. What we're going to do is look at that 600 million cell model using a
static statistic after Funk et al. In 2012. The idea behind this statistic is
that we're going to look at the surface area at the base of a single channel
element and see how the facies are juxtaposed across one element to another.
Where we've got sand to sand juxtaposition, that represents the best
connectivity across two channel elements.
So we'd like that proportion to be actually pretty high to represent great
opportunity for fluid flow between two channel elements. Where the brown facies
exist between those two channel elements, we can think of it as baffled fluid
flow. Where the black facies exist, that's the drape facies, that's where you'd
have a flow barrier. And then we can also think about what the transition of
fluids, or transmission of fluids from the channel element to the background
facies might be.
So what I'm going to be sharing you are a series of plots like you see on
the right, where the x-axis shows the surface area proportion that is
represented as a proportion of sand to sand connections, baffled connections in
brown, barrier connections in gray, and then the potential for fluid flow to
the background, which is the cross hatched section. To give you an example that
you've already seen to better understand the statistic, what you'll see is the
simple two channel element fluid flow experiments from laterally offset stack
to vertically aligned. And the statistic in the bottom left.
And so, as I mentioned, the laterally offset stack, there's a very small
connection between the two channel elements. And that's represented in the
statistic, where we have a very small surface area proportion to sand to sand
connection. And it increases as we move up into the vertically aligned
scenario, where we have a higher connectivity and also as we showed, a very
much faster water breakthrough. So higher connectivity may not always be the
Now if we look at the full 3D model, so I'm showing you a cross section
through the model. But imagine that these channels are going into the plane,
and you have channel center lines that are migrating together and away from one
another into the 3D plane. And the reason we think about this is connectivity
is a 3D problem. We want to understand connectivity in three dimensions.
So you see Complex 1 in the lower most part of the model, where you see more
of the laterally offset stacked channels. Complex 2 is more vertically aligned,
and Complex 3 we start migrating back and forth a little bit more. And so if we
look at the surface area proportions statistic, you can see kind of a lower
connectivity in that lower channel complex with a pinch point. So the point
where it goes back to zero, meaning there's no connectivity between channel
Complex 1 and Complex 2 all along the length of the channel belt.
And then you see the statistic increase quite a bit toward the top of
channel Complex 2. What I want to point out is if you look at the box plot
graph to the right of it, this is a graph that captures the range of lateral
migration between channel element pairs. And so at that point at the top of
channel Complex 2, what you see is minimal lateral migrations. So that means
that the channels are tracking one another, and they're staying close together.
What that does is that promotes connectivity of sand to sand connections
across the axes. And in the areas where you've got large degrees of lateral
migration you see that connectivity metric drop off back to the left. So the
take home from this is that this is a statistic we can tie back to seismic
data, is from seismic we could actually see the width of a channel belt.
So narrow and narrower channel belts mean more vertically aligned channels
and a higher probability of sand to sand connections across those channels. So
what we can do with this is play some sensitivity games. So we may not know
exactly what the channel width is. So the base case in this model the channel
width was 250 meters. We can drop that back to 200 meters and see how it
impacts the connectivity.
What it does is it has an overall effect of reducing the channel surface
connectivity overall. But proportionally, it reduces the amount of sand to sand
connectivity. And then that change is shown in red. And it's a negative change,
so a loss of connectivity.
More importantly, however, is a change in the internal architecture, and
specifically the drape coverage. So if we start to bring the drape coverage
down and increase the amount of coverage of the drape over the channel, we
really start to shut off that sand to sand connectivity. So this is a first
order control between channels. So it's really important to understand and
represent in our models.
And then finally we look at the potential of how grid cell sizes can impact
connectivity. So we took the model and up scaled it to a 10 by 10 by 1, so not
significant upscaling from the two by two by 0.25. But enough to show that four
times aerial upscaling-- five times aerial upscaling and a four times vertical
upscaling will reduce the sand to sand connectivity. So the decision of
gridding can impact that connectivity in our models.
Finally as I mentioned, connectivity is a 3D problem. So what you see on the
top is a net sand map from the model, where purples are high net sand thickness
and reds are low net sand thickness. And the black line shows the location of
the cross section in the middle where green is the lower channel complex,
yellow is the middle channel complex, and brown is the upper channel complex.
And then on the bottom, and what I'm showing for 10 equally spaced segments
along the model, these connect vertical connectivity diagrams. Now the idea
here was to think about, are there distinct flow units we could potentially
delineate within this model, and are they interpretable flow units from seismic
data. So if we can interpret at the complex level, would we be able to
interpret the flow units?
And the answer is for this particular case, in the lower channel complex, we
have laterally offset stacked channels, very low amounts of connectivity. But
pretty much everywhere, as with the composite statistic I showed on the last
slide, we've got a pinch point back at the top of the complex showing a break
in connectivity between Complex 1 and Complex 2. So that surface, potentially
interpretable from seismic data, does coincide with a flow unit boundary.
However as we look up into the upper section, it's variable. And what we see
on the southernmost, rightmost side of the model, is that as the channels
spread out, there is a breaking connectivity. But it's not at a complex
boundary. And it doesn't carry out across the entire model. So the flow unit is
in constant thickness package, and it varies across the channel model.
OK, so we've talked now about fluid flow simulation. The next step is to
think about using these outcrop models to think about seismic response. So we
took the entire 600 million cell model and forward seismic model that using 1D
convolution. And what you see on the left is the results of that published in
Pemberton et al. This year. And what you see, similar to what I showed in the
introduction, is that with decreasing frequency from 90 Hertz down to 15 Hertz
we lose our ability to resolve individual channel elements, and we move into
potentially at 30 Hertz being able to resolve complexes, and at 15 Hertz just
the entire channel complex set.
So the question we started to ask in looking at this is how would-- with
what we've learned so far-- would we guide people in being able to interpret
subsurface seismic responses from our outcrop data. And we can come up with
qualitative interpretations and qualitative guidelines from these type of full
3D outcrop modeling studies. But the idea is can we go back to those
fundamental building blocks of the channel elements, a single channel element
and multiple channel elements stacked in different orientations, to generate
some sort of set of guidelines or understanding of what the seismic responses
are that we're interpreting.
So what you see is a 60 Hertz response of a single channel and multiple
channel element template models. And what we're going to be looking at is all
phase rotated data. So it's 90 degree phase rotated so that the peak is within
the sedimentary body, and the top and the base of the sedimentary body is at
the zero crossing. And what we're going to analyze is attributes from these
small template models, thinking about what is the apparent thickness, so the
thickness as interpreted from the seismic data, and then the composite
amplitude, the amplitude between the top 0 crossing and the base 0 crossing of
So first, before we jump into that, let's think about some basic theoretical
geophysics that we can base this work on. So this is a tuning wedge model,
where we have a constant high acoustic impedance in the background, then a
constant low acoustic impedance in the wedge. Now on the bottom graph you see
amplitude versus wedge thickness. And we're looking at the amplitude of the top
reflector across the wedge.
Now, at the thickest part of the wedge, you have a constant amplitude, where
there's no interference between the top reflector and the bottom reflector. But
as the wedge begins to thin, you have constructive interference that creates an
increase in amplitude, and then destructive interference as the wedge thins and
the signals is lost.
So that brightening in amplitude is one attribute that we'll look at. But
the other attribute is the apparent thickness and the apparent thickness change
as interpreted from seismic versus the actual thickness. And so the graph on
the left, you see apparent thickness on the vertical axis and actual thickness
on the bottom axis. Now pink is the one to one line, showing that the thickness
that you're interpreting from seismic is identical to the true thickness. And
when, with the onset of tuning, the apparent thickness begins to diverge. And
so, you start interpreting a thicker wedge than actually what exists in
OK, so what we're going to do is we're going to look at these two
attributes, amplitude and apparent thickness, with our little single channel
element model. And just to set it up for you, we're going to look at a high acoustic
impedance axis and a low acoustic impedance margin. And so you see the change
in acoustic impedance in the red line from axis is high, dropping off to low
acoustic impedance in the margin, and a constant background acoustic impedance
OK, so first, let's look at apparent thickness. We'll start with our base
fundamental case, the single channel element. And what you see there is
variable seismic frequency from 180 Hertz down to 20 Hertz. And you see at 180
Hertz, we're able to resolve the top and the base of the channel element. But
when we drop down to 20 Hertz, our apparent thickness is almost twice that of
that true element thickness.
What we're going to do is we're going to focus in on 30 Hertz. All the
examples from here forward I'll show you are 30 Hertz examples. So there is a
graph on the lower left now that's apparent thickness versus true stratigraphic
thickness. And this is all of our model data. Again the one to one line shows
us that the point at which our apparent thickness, the thickness we're
interpreting from seismic, matches identically to the true thickness of that
And so where you see the black line is the apparent thickness of a single
channel element and it ranges between 22 and 23 meters thick. Whereas we know
the true channel element thickness is 14 meters thick. So we've got a
significant overestimation in that channel thickness.
So we're going to look at the two channel element case, and see how it
compares to the single channel element case. So what I'm showing in the lower
left is a laterally stacked case. And we see it tracks the one to one line with
a slightly thicker apparent thickness in certain locations. And now as we swing
up to a more vertically stacked case, we're going to talk about the case in
The first thing you'll notice is that we actually hit the one to one line.
So we're getting a thick enough stacking pattern to really start to be able to
resolve these two channel elements in seismic. So apparent thickness can tell
us here that we have to channel elements. And we also see a deviation of
variable apparent thickness off of the single channel element lines.
So now if we think of amplitude responses. So we're going to be looking at
the left most part of the tuning wedge and the amplitude response of that
tuning wedge, because everything we're looking at in this 30 Hertz model is
fully tuned. So for a single channel element, RMS amplitude versus net sand
thickness creates a line that's got relatively constant slope.
When we add in a second channel element laterally offset stacked, we see a
slight change in that RMS amplitude due to the stacking patterns in those two
channel elements. And again, as we bring in another channel element to a more
vertically offset stacked configuration, we see not only that difference in net
sand thickness, but also a change in the RMS amplitude, a distinct pattern. But
what's most interesting is that we can bound these different cases between a
single channel element and two vertically stacked elements between two
differentially sloped lines.
So there's distinct RMS amplitude patterns for these different
configurations of channel elements stacking. So the tuning wedge results that
I've shown you so far have mainly focused on the impact in amplitude based on
change in thickness of a sedimentary body. But what about the internal
heterogeneity? So I'll take you back to our initial model of the half element
of the channel showing the decrease in acoustic impedance from the access to
Now, if this were a homogeneously filled channel element, what we would see,
so the orange line in the bottom graph, is our theoretical tuning response,
where we have that increase in amplitude as that natural wedge, in this
particular case, thins. However, when we add in the heterogeneity, the decrease
in acoustic impedance is proportional to that increase in amplitude due to
tuning. And we have a dampened tuning effect. And so, we can almost bound the
two end members of a sandy filled channel element and a heterogeneous, more
marginal facies filled channel element to see, again, the importance of where
those marginal facies are placed within these channel elements and what their
So finally, what we want to be able to do ultimately, is to tie this back to
subsurface seismic responses and be able to interpret whether or not we have
one channel element, two channel elements, and ultimately how they're stacked.
And so we can go back and think about, what are the attributes we can interpret
from seismic? So RMS amplitude between two 0 crossings, and then the apparent
thickness of that, the distance between those two 0 crossings.
So again, what you see in the bottom graph is RMS amplitude versus apparent
thickness, with the black line showing a single channel element. Now, if we add
in a second channel element laterally offset stacked, you don't see a
significant change or significant difference between these two channel elements
and one. But what you would see in seismic is a reflector that's much longer. So
the length of two channel elements versus one. So you'd be able to detect this,
even though it's not showing in RMS amplitude versus apparent thickness
So then again, when we start to incorporate a second channel element we see
a strong RMS amplitude versus apparent thickness response. And then finally,
the vertically stacked case where the apparent thickness increases
dramatically. And so, we can think about, again, using these template models as
a way to tie back to seismic, and potentially have a foundation for a better
seismic interpretation and to understand what the amplitude and thicknesses
So in conclusion, I talked about fluid flow, and how the internal
architecture of channels and how the channel stack can impact fluid flow. So just
in summary, when channels are stacked vertically we juxtapose the axial facies
and the active seep zones. And we bypass a lot of the production, or a lot of
the oil in the marginal facies. And in the laterally stack case, the marginal
facies play a strong role in being baffles or barriers. And both of these
effects, particularly in the laterally offset stack case, are exacerbated when
we add in asymmetry and start to drop the net to gross, or increase the
proportion of these fine-grained marginal facies.
Furthermore, I showed how narrow channel belts give us a higher probability
of connectivity, because we're vertically aligning those channel elements and
promoting sand to sand connections. And then finally, when we're trying to
think about flow units that stratigraphic packages that we could potentially
interpret from seismic don't always correlate to flow units, and that flow is a
And so the summary of that is when we start thinking about trying to take
what we've learned from these outcrop studies and apply it to subsurface
modeling, we could potentially get away with larger grid cells and homogeneous
fill for vertically stacked channels. And then as far as the seismic response
goes, we're able to show how RMS amplitude and apparent thickness could
potentially with these little template models help us to differentiate multiple
channel elements, and then how they're stacked, which again is really critical
for understanding fluid flow.
Furthermore, the internal architecture of the channel element heterogeneity
dampens tuning. So there is an amplitude response from this heterogeneity,
which again, that heterogeneity is one of those things that's critical for
understanding fluid flow in these stack channel element systems. And the
templates in member cases, single versus multiple, homogeneous versus
heterogeneous, can aid in the sub-seismic scale interpretation in the
So the way forward. As a community, we're starting to collect more and more
digital data from outcrops. And the question is, how can we start using this
digital data as a better quantitative analog to the subsurface, to help us
better understand the subsurface and connectivity and seismic interpretation?
What I propose is making sure that we can use this quantitative data in a
way that's architecturally sound. So that is translating the outcrop data and
statistics, for example, vertical facies, proportion curves, as shown on the
right, into templates that capture the architecture that we observe in the
outcrop in the lower left. And we can use these in, for example, machine
learning workflows, in seismic model-based inversion workflows, well log 2D
modeling around the wellbore interpretations.
And then finally, the question always looms is what makes a good subsurface
analog? With digital data we can spend some time really comparing outcrop
statistics to subsurface statistics for a better analog selection. So I thank
you for your attention and listening to my research presentation. And again I
want to thank AAPG for inviting me to share my research with you, and also for
all the co-authors on this work and the sponsors who supported it. Thank you.