Tying the well: proving the wiggle is the rock

You have been calling something top reservoir for three sections. Nobody has demonstrated it.

The well knows where the reservoir is, in metres. The seismic knows where its reflectors are, in milliseconds. The only bridge is a synthetic seismogram, and the workstation builds it from the files you imported and nothing else: sonic and density from the LAS, multiplied into impedance; the checkshot to place every impedance step at its two-way time; a zero-phase Ricker at a frequency you choose; and the recorded traces at the well location beside the result.

In the workstation

Open the Well Tie viewport. One unit trap first: the LAS delivers sonic as slowness in microseconds per foot, as real logs do, so the velocity is @@V_p = 304800 / \Delta t@@ in metres per second; the workstation does the conversion and says so. Sweep the wavelet frequency until the synthetic carries the same character as the data, then slide the bulk shift and watch the correlation panel, not the wiggles. When the peak is where you believe it, ACCEPT THE TIE: the accepted shift follows your maps into depth conversion.

Read the correlation curve

The panel on the right is correlation against shift, and it is more informative than the traces beside it. A dense reflectivity gives a sharp single spike, and a tie you can defend. A thick clean sequence gives the opposite: few strong reflectors, a broad correlation function, and side lobes almost as tall as the true peak. That second case is where cycle skips come from: a tie one wavelet period out still correlates well, still looks convincing, and puts every horizon you subsequently name one loop wrong, which at this depth is tens of metres. The defence is not a better eye; it is the shape of the correlation function, and the habit of asking whether the peak you chose is really the only candidate.

And this is why the tie belongs before the picking in any workflow you run outside this lab. A horizon picked on the wrong loop is internally consistent: it tracks beautifully, closes its loops, grids into a perfectly plausible map of the wrong surface. Nothing downstream will catch it. The tie is the only place in the whole workflow where the seismic is compared against something measured rather than interpreted.