## Geological layer-depth sequences

Subsurface geology is inherently uncertain due to the limited availability of direct measurements and the complexity of geological processes over time. Data from wells, seismic surveys, and logging tools provide partial and sometimes ambiguous information, leaving gaps in our understanding of the subsurface. 
Predicting geological scenarios enables us to integrate available data and reduce uncertainty. Thus, predictive methods enable informed decision-making in applications like drilling, resource extraction, and risk mitigation, ensuring more efficient and safer operations.

Most geological systems are composed of thin layers with different properties deposited on top of each other, forming a flat layer cake of constant thickness in the simplest case. As these layers are buried deeper into subsurface, different physical processes might bend the layers or break them, creating geological faults. 


**Your task is to create a data-driven model that predicts likely 1D layer-depth sequences based on formations where lateral wells are typically drilled.** The model should generate the geological deformations based on the **geology in the well up to the current drilling location** (input).


Figure: 1. Geology with parallel horizontal layers.
![Geology with parallel horizontal layers](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F832185%2Fb4d93f531a66de875bce3e526d2c3c2c%2Fimage%20(6).png?generation=1736785314112560&alt=media)

Figure 2. Actual 1D geology with faults and dips.
![Actual 1D geology with faults and dips](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F832185%2F44fc83c8159022d4062502205e3a4248%2Fimage%20(8).png?generation=1736785363139707&alt=media)


If, at a local scale, the assumption of the constant thickness layer cake holds, geology can be defined by delta_z(X), the deformation at each location. 
The data in this competition is delta_z(x) along the lateral (horizontal) wells, creating 1D layer-depth sequences.



Such reconstruction would help with predictions for **geosteering** - the real-time process of directional changes of a well while drilling. The predicted geological realizations can be compared to acquired data to reduce uncertainty, further informing decision-making. One of the possibilities is using this model instead of a proposal distribution in a method similar to a particle filter [Muhammad et al. 2024](https://doi.org/10.2118/218444-MS).

## Competition Financial Sponsor
This competition is supported by [ROGII Inc](https://www.rogii.com/). **ROGII** provides software solutions for geology, drilling, and completions teams to use on a customized cloud-based platform.

We acknowledge ROGII for providing the field data and the monetary prizes.