What Synthetics Are For
Learning objectives
- Name the four everyday jobs synthetic seismic does
- State the common thread: the true earth is known by construction
- See why noise-free labels make synthetics ideal training data
- Use a feasibility test to ask whether a survey can see a target at all
Four Jobs, One Trick
Modeled seismic earns its keep in four ways, and they all rest on the same trick: you know the answer. The earth that produced the data is the one you built, so you can measure exactly how well anything else recovers it. Real data never offers that. The four cards below are those four jobs, and they share one knob, the field-noise level, so you can watch the known answer survive or drown.
The Four Jobs
- Calibration. Build a synthetic from a well log and lay it against the seismic at that well. A good match confirms your velocities, your density, and your time-depth picks. A poor one tells you something upstream is wrong.
- Interpretation. Before you hunt for a pinchout or a bright spot in real data, model it and learn its signature. The tuning wedge is the classic case: watch a thinning bed brighten and merge so you recognise it later.
- Training data. A machine-learning model needs thousands of labeled examples. Draw the earth, generate the section, and the label comes free, because the label is the geology you drew. Noise degrades the image, never the label.
- Feasibility. Before spending on a survey, ask whether it can even see the target. Model the earth with and without the target, take the difference, and compare it to the noise floor. If the anomaly hides under the noise, redesign before you shoot.
Turn the noise up and each card responds honestly. The well tie loosens, the difference anomaly sinks toward the noise floor, and the tuning signature blurs. Only the training labels hold perfectly still, because they were never measured from the data; they were declared when you drew the model. That asymmetry, perfect labels over imperfect data, is why synthetic seismic and machine learning fit together so well, and it is the thread running through the rest of this course.