Seismic_Imaging26

This is a set of course materials for a graduate-level seismic imaging class taught at the University of Pisa, developed by a PhD student named Felipe Rincón. The repository contains a single interactive Jupyter notebook and supporting Python code that walk through two computational techniques used in geophysics to create images of underground structures from seismic data.

The first technique is Full Waveform Inversion, or FWI. The basic idea is to generate a computer simulation of how sound waves travel through the ground, compare the simulated recordings with real recorded data, and then adjust the underground model until the two match as closely as possible. This iterative adjustment process lets you infer what the underground looks like from surface measurements. A known difficulty with this technique is called cycle skipping, where the algorithm gets confused and converges to a wrong answer. The 2026 edition of these materials addresses this by starting the inversion with low-frequency data to get a rough picture first, and then progressively introducing higher frequencies for finer detail.

The second technique is Reverse Time Migration, or RTM, which takes the final velocity model produced by FWI and uses it to produce a reflectivity image of the subsurface, showing where rock layers and boundaries are.

The notebook gives students several options to experiment with: two choices of source wavelet (the shape of the input signal), three ways to measure the difference between simulated and observed data, and four optimization algorithms for running the inversion. All exercises use synthetic data, meaning the ground truth is known in advance, which makes it easier to evaluate how well each configuration performs.

Installation requires Devito, an open-source framework for seismic simulations, which can be run through Docker. The README includes step-by-step setup instructions.