Building 2D Earth Models
Learning objectives
- Build a section as many independent traces, one per model column
- Shape a layered earth with dip and a fold
- See that structure in the earth becomes structure in the section
- Recognise that 2D convolution still has no lateral coupling
From One Trace to a Section
Part 2 built a single trace from a single column of earth. A seismic section is nothing more grand than many of those traces placed side by side, one for every column of a two-dimensional earth model. Nothing new happens in the physics: each column is converted to impedance, differenced to reflectivity, and convolved with the wavelet, exactly as before, entirely on its own.
Structure In, Structure Out
Shape the earth on the left and the section on the right follows. A regional dip tilts the layers and the reflectors tilt with them. A fold raises the layers into an anticline and the reflectors arch. Because the model is the truth and the wavelet is fixed, the section is a faithful, band-limited picture of the geology you drew. This is the everyday workhorse of interpretation training and of generating structured synthetic data: paint a structure, convolve, and you have a labelled section.
But keep one limit in clear view. A 2D convolutional section is still trace-by-trace, with no communication between neighbours. It faithfully warps horizons, but it cannot make anything that requires a wave to travel sideways. Fold the model as sharply as you like and you will never see a bowtie; the reflectors just follow the geology. That honesty about what 2D convolution can and cannot do is the theme this part returns to at its close. First, in the next section, you cut this layered model with a fault.