Polarity, phase, and wavelet conventions

Part 2, The Interpreter's Toolkit

Learning objectives

Distinguish SEG normal from reverse polarity for zero-phase and minimum-phase data
Recognize zero-phase, minimum-phase, and phase-rotated wavelet shapes on a seismic trace
Identify the convention of an unfamiliar dataset by inspection and by well-to-seismic tie
Explain why a misread polarity inverts amplitude interpretation and a misread phase shifts picks

Nothing on a seismic volume, neither a pick nor an amplitude, means anything until you know how the volume displays a rise in impedance. Two processing teams can take the same field records and deliver volumes that look inverted or shifted relative to each other over an earth that has not moved. Reading those conventions comes before any amplitude work.

Polarity: what a positive number means

Polarity answers one question: does a rise in acoustic impedance going down, a positive reflection coefficient $R$ , appear on the display as a positive number (a peak) or a negative one (a trough)? The SEG standard (Thigpen, Dalby and Landrum, 1975) was written for the recorded pulse, which is close to minimum phase, and the zero-phase rule was added later, so the two read differently:

SEG normal polarity. A compression is recorded as a negative number, so on minimum-phase data a rise in impedance begins with a trough at the boundary. On zero-phase data the convention is that a rise is a central peak (Sheriff and Geldart, 1995).
Reverse polarity, often called European. The opposite sign of either: on zero-phase data a rise in impedance is a central trough. The name is informal and both conventions are used on both sides of the Atlantic, so ask for the picture of a rise, not the label.

Neither convention is wrong; each is consistent. The danger is using one while believing you have the other. On SEG normal zero-phase data a bright trough at the top of a sand is a fall in impedance, the classic gas signature; on reverse-polarity data the same trough is a rise, a hard event. Misread the polarity and you invert the hydrocarbon call.

Phase: where the energy sits

Phase describes how the wavelet arranges its energy in time. Three cases matter to an interpreter:

Zero phase. The wavelet is symmetric about its central lobe, which sits exactly on the boundary. This is what processing aims to deliver, because the brightest lobe then marks the boundary’s time and, on SEG normal data, its sign is the sign of $R$ .
Minimum phase. The wavelet is causal: nothing arrives before the boundary, and the energy is packed as early as its spectrum allows. Explosive and air-gun sources are close to minimum phase before processing. The onset marks the boundary, but the brightest lobe need not be the first: for the figure’s minimum-phase version of a 30 Hz Ricker (its spectrum floored at −60 dB) it is the second lobe, of opposite sign to the onset, about 22 ms after the boundary. The exact delay depends on how the low frequencies are treated; that the bright lobe comes late does not.
Rotated phase. A constant phase rotation $\varphi$ turns the phase of every frequency by the same angle. At 90° the boundary sits on a zero crossing between a peak and a trough. A rotation is measured in degrees, not in time: for a 30 Hz Ricker, 90° moves the lobes 6.3 ms either side of the boundary, less than the quarter period of 8.3 ms that a single frequency would move. A rotation of 180° is exactly a polarity reversal.

Hold one boundary fixed, $R = +0.15$ , and change only the convention it is displayed in: first the phase, then the polarity, then the wavelet.

A 90° rotation alone costs 6.3 ms. The figure opens on data delivered as zero phase but rotated +90°: the boundary sits on a zero crossing between a peak 6.3 ms above it and a trough 6.3 ms below, equally bright at 0.83 of the unrotated peak, while the envelope, dotted in (b), does not move. An interpreter who tracks the peak places every horizon on that reflector 6.3 ms too shallow: a constant bias rather than one that grows, but one that goes straight into depth conversion. Rotate back to 0° and the peak returns to the boundary; past 90° either way the brightest lobe is a trough and the sign reads wrong, the shaded bands of (d).

That 6.3 ms is a timing error, and a well tie would expose it. A polarity error is different. At 0° switch the polarity to Reverse: the trough sits on the boundary, so the time is right and the sign is wrong. No timing check can catch a polarity error; only a boundary whose sign you already know can. Choose Minimum phase with no rotation and the event starts with a trough on the boundary, as the SEG standard defines, yet its brightest lobe is a peak about 22 ms later.

A fair question at this point is whether an interpreter simply has to live with whatever phase the data arrives in. Largely, no. Getting the data to zero phase is a processing job: deconvolution compresses the minimum-phase wavelet and whitens its spectrum, and a wavelet estimated from the source signature or from a well then sets the remaining phase to zero. The deconvolution side is treated properly in the Seismic Data Processing book, Section 2.6 on spiking deconvolution, alongside predictive and surface-consistent deconvolution. What stays with the interpreter is the obligation to find out what was actually done, because a volume can arrive labelled zero phase and not be, and no amount of careful picking recovers from that.

Identifying the convention of an unknown dataset

When a new dataset lands on your desk without clear documentation, there are three ways to find its convention:

Processing report. A competent processing report states polarity and phase, ideally as a picture of how a rise in impedance appears. Read it first.
Known reflector. If the section contains a boundary of known sign, the sea floor on marine data or the top of a known salt body, check how it appears. Water over soft mud gives $R \approx +0.2$ , still strong enough to read at a glance. On zero-phase data a sea-floor peak means SEG normal and a sea-floor trough means reverse. On minimum-phase data the sea floor begins with a trough under SEG normal polarity, so this test needs the phase first.
Well-to-seismic tie. The definitive test: at a well, compute a synthetic seismogram from the sonic and density logs and compare it with the real trace there. The synthetic shows what the seismic should look like under each convention; the one that matches is the one your dataset uses, and the tie also measures any residual phase rotation, which a sea-floor check cannot. The book builds one in Section 11.5, Tying the well.

Why misreading convention is serious

A misread polarity inverts every amplitude inference. The top of a soft reservoir reads with the sign expected of a hard boundary. A gas bright spot stays bright but on the stack reads as a hard, high-impedance event, and on an AVO crossplot the top of a Class III gas sand (a soft, low-impedance sand) moves to where the base of such a sand plots, because a polarity flip negates both the intercept and the gradient. A misread phase moves every pick instead: by 6.3 ms at 90° on 30 Hz data, and by twice that at 15 Hz.

The saving grace is that convention errors are systematic: if you have it wrong, you have it wrong everywhere in the same way, so one check on a boundary of known sign exposes it. Make that check before the first pick.

References

Bacon, M., Simm, R., & Redshaw, T. (2003). 3-D Seismic Interpretation. Cambridge University Press.
Sheriff, R. E., & Geldart, L. P. (1995). Exploration Seismology (2nd ed.). Cambridge University Press.
Thigpen, B. B., Dalby, A. E., & Landrum, R. (1975). Report of the subcommittee on polarity standards. Geophysics, 40(4), 694-699.
Simm, R., & White, R. (2002). Phase, polarity and the interpreter’s wavelet. First Break, 20(5), 277-281.
Brown, A. R. (2011). Interpretation of Three-Dimensional Seismic Data (7th ed.). AAPG Memoir 42 / SEG IG13.
Hilterman, F. (2001). Seismic Amplitude Interpretation. SEG/EAGE Distinguished Instructor Short Course.