How to look at a 3D volume

The two-minute reconnaissance

When you open a new 3D volume, do not start picking. Spend two minutes doing reconnaissance. The goal is not to interpret anything yet — it is to build a mental map of the volume so that when you do start interpreting, you know what normal looks like. A disciplined reconnaissance:

• Scroll through time slices top to bottom. Map-view patterns are where the obvious regional features live — drainage systems, polygonal fault networks, salt margins, major stratigraphic boundaries. Get the big picture before you zoom in.
• Scroll through inlines end to end. Look for major reflectors, fault networks, changes in reflector character. Note depths where the data gets noisy or fades.
• Scroll through crosslines end to end. Same survey, different axis. Features that were foreshortened on the inline view will be clearer here.
• Spot-check a few arbitrary lines. Pick an interesting-looking structure on a time slice and cut a line perpendicular to it. This is how you catch features that are oriented wrong for the acquisition axes.

Two minutes of this saves two hours of re-work later. The interpreter who dives straight into picking a target horizon often discovers — after days of work — that the horizon they have been tracing is actually a processing artifact or a sideswipe event from a feature off to the side.

The "check the display" habit

Before you interpret anything, confirm what you are looking at. This takes 30 seconds and catches a large fraction of the errors junior interpreters make:

• What polarity convention? SEG positive means an increase in acoustic impedance reads as a peak (red on an RWB palette). European convention is the opposite. The same rocks produce opposite visual signatures on the two conventions. If you do not know which your dataset uses, you cannot read amplitudes at all.
• What phase is the wavelet? Zero-phase data has its peaks aligned with the reflectors. Minimum-phase or mixed-phase data has the peak offset from the reflector. Picking the wavelet peak on zero-phase data gives you the reflector; doing the same on minimum-phase data gives you a position that lags the reflector by a quarter of a wavelet cycle or more.
• What palette and gain? RWB makes polarity obvious; grayscale de-emphasizes polarity in favour of geometry. High gain saturates the brightest events into solid blocks; low gain makes weak events disappear. Neither setting is wrong, but you need to know which you are looking at.
• Time or depth domain? Time sections are in ms TWT. Depth sections are in m or ft. Features look different in each domain because layers of different velocity have different time vs. depth ratios. Pick map contour intervals appropriate to the domain (50 ms vs 50 m are entirely different contours).

§1.0's F3 cube viewer lets you toggle palette and gain and see the effect on the same geology. If you have not played with it, do that now before continuing.

Never interpret from one view alone

The single most common pitfall in 3D interpretation is treating a single 2D slice as if it contained the answer. Any feature you pick on an inline should be confirmed on the crossline and on the time slice. Any fault you trace on a cross-section should be visible as a lineament on the time slice within the fault's time window. Any subtle amplitude anomaly that excites you on one slice should be compared with the surrounding slices in both directions.

The practical rhythm is: spot a feature on one view, immediately flip to the other views and confirm. If the feature is coherent across views, it is probably real. If it appears only on one view or disappears when you shift to a nearby slice, it is probably noise or acquisition footprint. Train yourself to flip views automatically, without stopping to think about whether it is worth it. It always is.

Start broad, end narrow

Interpretation workflows invariably spiral inward: start with a regional overview, identify the targets, zoom in to specific features, and refine. Inverting this order — picking a target before understanding the regional context — leads to interpretations that do not fit the larger story.

A rule of thumb: every pick should be consistent with a story that makes sense regionally. A horizon that plunges into a fault on one crossline, then jumps 300 ms on the next crossline without a corresponding fault, is telling you the horizon pick is wrong. A reservoir sand that thins to zero in one direction and thickens in the other without a depositional explanation is telling you the geometry is wrong. The volume is self-consistent; interpretations that appear internally inconsistent are almost always the interpretation's fault, not the data's.