Seismic geomorphology: reading paleo-landscapes from 3D volumes

Three ways to cut a plan view

Every slice is a surface through the volume, and the map shows an attribute, here amplitude, where the surface passes. The three common slices differ in what that surface follows.

• Time slice. A flat surface, $t_s = t_0$. It follows a depositional surface only where the strata are flat and parallel. Across dipping strata it passes from one stratum to the next and shows the layering as bands that run parallel to the structure contours; across a thickening interval it drifts too, and over a flat base it climbs into younger rock where the interval thins. Time slices are quick and good for faults, footprint and structure, and they are where channels are often first spotted, but the age of everything on one has to be checked.
• Horizon slice. A surface hung from one mapped horizon, $t_s = t_H + \Delta t$, which is what flattening the volume on that horizon gives. On the horizon itself it follows a depositional surface exactly, and it stays on one as long as the strata run parallel to the horizon. Where the interval thickens or thins, a fixed $\Delta t$ drifts across the strata.
• Stratal slice. A surface that keeps the same fraction of the interval between two mapped horizons, H1 below and H2 above (Zeng et al., 1998a, b).

Here $t_1$ and $t_2$ are the times of H1 and H2 on each trace. A stratal slice follows a depositional surface wherever the strata between the horizons thicken in proportion to the interval, often a fair assumption between two regional flooding surfaces in a steadily subsiding basin. It fails where they do not: onlap, downlap, erosion, or thickening concentrated early or late in the interval. Check every stratal slice against the reflections in section before reading it as a landscape.

Exercise, read the figure

1. The figure opens on a time slice through the well at the level of the channel. Plate (c) shows it passing through rock that spans about 1.5 Myr in age, and plate (a) shows pieces of the channel beside build-ups and a mass-transport deposit that never shared its landscape. Plate (b) shows why: the slice is flat and the strata are not.
2. Switch to Stratal. The age span falls to zero and the whole channel appears, with its levees and the crevasse splay: one landscape.
3. Press Sweep the level, or drag it, and watch the landscapes follow one another: the build-ups, the channel, the mass-transport deposit, each whole on its own surface; away from it, an element shows only where its reflections come within reach of the slice.
4. Choose Horizon at the level of the channel. The horizon slice follows the dip, so it does better than the time slice, but where the interval thins and thickens it drifts into younger and older rock. Set the thickening to zero and it becomes a stratal slice.
5. Set both the dip and the thickening to zero. With flat, parallel strata the time slice, too, follows one surface.
6. Turn on Uneven thickening. Now even the stratal slice drifts: proportional slicing is only as good as its assumption.
7. Back on the stratal slice at the channel, drag the channel fill from 30 m down to 6 m and watch the brightest lines of the channel move from its margins to its axis, then fade.

Amplitude and tuning in plan view

Amplitude on a slice is set by impedance contrasts, $R = (Z_2 - Z_1)/(Z_2 + Z_1)$ with $Z = \rho V_P$, and by how the reflections from the top and base of a bed interfere. A bed thicker than about a quarter of the dominant wavelength shows its top and base as separate reflections. As it thins toward that thickness the two interfere constructively and the amplitude peaks: this is the tuning thickness (Widess, 1973; Kallweit and Wood, 1982). For a Ricker wavelet of peak frequency $f_{\mathrm p}$ the peak comes at a two-way thickness of

half the trough-to-trough period $T_{\mathrm d}$ of the wavelet, which is $\lambda_{\mathrm d}/4$ for the wavelength $\lambda_{\mathrm d} = V_P T_{\mathrm d}$ at the dominant frequency: 13 ms at 30 Hz, or 15 m in sand of 2350 m/s (Section 1.7). Below it the amplitude falls roughly in proportion to thickness, so a thin bed is detected even though its top and base are not resolved.

In plan view this makes an amplitude map something other than a thickness map. A channel whose fill is no thicker than the tuning thickness is brightest along its axis. One whose fill is thicker is brightest along two lines near its margins, where the lens thins through the tuning thickness, with a dimmer axis between them. Spectral decomposition uses the same effect, since each thickness tunes at its own frequency (Partyka et al., 1999; Part 6).

Recognizing the elements in plan view

• Channels. Sinuous ribbons of nearly constant width, often with bright margins from tuning; meander loops, cutoffs and scroll bars in fluvial and some submarine systems (Section 4.4). Their width and sinuosity scale with the flow that cut them.
• Levees. Wedges that thin away from both banks, thickest on the outer banks of bends: in plan, a halo of amplitude that fades over a few hundred metres; in section, the gull-wing profile of a leveed channel.
• Crevasse splays. Lobes that break through a levee, usually at an outer bend, and fan out with distributary fingers; in section, a thin sheet that thins away from the channel.
• Mass-transport deposits. Lobate bodies with chaotic or transparent internal reflections, an erosional base and a rugose top. Their kinematic indicators show which way they moved: an arcuate headscarp upslope, striations and grooves along the path, and arcuate pressure ridges at the toe, convex downslope (Posamentier and Kolla, 2003; Bull et al., 2009).
• Carbonate build-ups. Rounded or elongate mounds with a strong positive reflection from the top of the hard carbonate over a transparent or chaotic interior, the strata around them onlapping or draping their flanks. Beneath a build-up the faster carbonate pulls the reflections up in time.

What plan-view geometry predicts for reservoirs

The shape of an element, read on the right surface, is a forecast of whether the reservoir is there and how it connects.

• A channel fill is a ribbon: well connected along its course and not across it, so a well placed off the ribbon misses it. Its levees are thin-bedded and poorer, but connected to the channel.
• A crevasse splay, or a lobe at a channel mouth, is a sheet attached to its feeder through the breach or the mouth: extra volume, and a path for fluids.
• Build-ups are isolated bodies: each can hold its own accumulation, with its own contacts and pressure, and each needs its own well.
• A mass-transport deposit is usually muddy and chaotic: a seal or a baffle above older reservoirs, and an eroder that can cut them out or split them into compartments. Sandy blocks inside it are rarely connected to anything.

Read connectivity only on a slice that follows the surface. A time slice that cuts a continuous channel into pieces makes it look compartmentalized, and one that puts a build-up beside a channel invites a connection that never existed.

The workflow

1. Map the bounding horizons. Pick regional surfaces that were once a land surface or a sea floor, such as sequence boundaries and flooding surfaces (Section 4.2), tied to wells, across the whole volume.
2. Slice between them. Flatten on one horizon for horizon slices close to it, and build stratal slices between two for the interval between them.
3. Extract and step. Display amplitude or another attribute on each slice (amplitude on the surface, RMS or maximum amplitude in a short window, coherence, spectral decomposition) and step through the slices in order, watching each landscape appear and give way to the next.
4. Interpret the landforms. Name each element by its geometry in its setting: a sinuous body on a slope is a submarine channel, and the same shape on a coastal plain is a fluvial one.
5. Confirm. Check every feature in section, on neighbouring slices and against wells: a real element persists over several slices and has a shape in section.
6. Predict. Combine the landforms with well data and rock physics (Part 5) into reservoir presence, quality and connectivity, and revise as new wells come in.

Pitfalls and limits

• Resolution. Vertically, a bed is resolved down to about a quarter of the dominant wavelength (8 to 40 m in most surveys) and detected, but not resolved, below it. Laterally, after migration, features narrower than about half a wavelength or a few bins (typically 25 to 100 m) blur or vanish.
• The wrong surface. A slice that crosses strata mixes landscapes. Where the strata onlap, downlap or are eroded, even a stratal slice drifts; check it against the section.
• Amplitude is not geology. Tuning, fluids and the side lobes of a strong reflector within half a wavelet of the slice all change amplitude with no change of landform. Confirm with the section and with rock physics.
• Noise. Acquisition footprint, migration artifacts and multiples can draw lines and patches; a real feature persists over neighbouring slices and has a shape in section (Section 2.1).
• Present-day bias. The paleo-landscape may have had a different orientation, climate and setting from any modern analog: let the analog test the interpretation, not dictate it.