Understanding seismic wave propagation is essential for earthquake early warning, seismic hazard assessment, and broader geophysical research. While classical numerical solvers such as finite-difference methods offer high accuracy, they are computationally intensive, particularly for large-scale or real-time simulations.

In this challenge, we invite participants to explore machine learning methods for accelerating 2D elastic wave simulations. The goal is to learn the dynamics of wave propagation from synthetic data generated by finite-difference simulations across diverse velocity models, earthquake locations, and focal mechanisms.

In this contest, your goal is to build a model that predicts the future wavefield evolution based on early wavefield snapshots and the spatial velocity model (Vp, Vs). This simulates a realistic forecasting problem, where only partial wave data is available at early times.

You’ll be provided with a dataset consisting of:  
**Training Data**  
--500 simulations of 2D elastic wave propagation  
--Each on a 128×64 spatial grid with 60 time frames (total duration: 0–1.77s)  
--Corresponding P-wave (Vp) and S-wave (Vs) velocity fields for each simulation  

**Test Data**   
--100 new simulations on the same grid and corresponding P-wave (Vp) and S-wave (Vs) velocity fields for each simulation   
--Only the first 15 time frames (0–0.42s) of wavefield data are provided   
--**Your task is to predict the remaining 45 time frames (0.45–1.77s)**   

## Breakdown of Kaggle contest scores (45 points):   
- Leaderboard Performance: 5 points   
If ranked top 5: 5 points   
Everyone else: 3 points   
- Final Paper: Literature review, methodology, and results 40 points
- Code submission: You must submit your code for the Kaggle contest so we can check the reproducibility of your results. If we detect any manual manipulation of the predictions you submit (as detailed on the Kaggle contest page), you will not receive any points for the contest.