GeoWorld: A Programmable Subsurface World Simulation Engine for Earth Sciences

GeoWorld is an open-source Earth science world engine focused on subsurface geological modeling, seismic simulation, and AI-assisted quantitative analysis. Starting from a controllable 2D subsurface world, the project gradually expands to 3D/4D/5D modeling, providing a reproducible synthetic data environment for geophysics, rock physics, and other fields, addressing the pain points of high cost and difficulty in sharing real geological data.

In fields such as oil exploration and geological reservoir analysis, researchers have long faced difficulties in obtaining high-quality reproducible subsurface data: real data is extremely costly (involving drilling and seismic equipment) and hard to share due to commercial protection. GeoWorld emerged to address this, providing a programmable synthetic subsurface world generation engine that supports the creation of complex geological models including strata and faults, for algorithm validation, model training, and ensuring research reproducibility.

GeoWorld v0.1 has implemented 2D synthetic subsurface world generation, including stratigraphic structures (sedimentary sequences of different ages), fault systems (fracture structures consistent with geomechanics), fold morphologies (rock layer bending caused by tectonic movements), and fluid inclusions (distribution of CO₂ and oil-gas accumulations). It can also generate key rock physics parameters (Vp, Vs, density, porosity, saturation, etc.), providing a foundation for seismic simulation and reservoir evaluation.

GeoWorld integrates simplified synthetic seismic data generation capabilities, simulating seismic wave propagation based on subsurface models and physical parameters, outputting synthetic records similar to real exploration data to assist in the development and testing of seismic interpretation algorithms. It also provides reservoir capacity estimation functions (carbon sequestration/oil-gas storage potential). In addition, the system supports visualization of professional geological profiles and attribute distribution maps, and each simulation generates a detailed execution log (parameters + random seed) to ensure result reproducibility.

GeoWorld supports multi-language interfaces including Python, Fortran, C++, and R, adapting to the needs of researchers from different backgrounds and seamlessly integrating into existing Earth science computing workflows. It also reserves external tool interfaces, allowing models to be exported to professional commercial/open-source software for complex analysis.

GeoWorld has a wide range of application scenarios: 1. Academic research: providing standardized test data for fields such as geophysics; 2. Education and training: helping students understand complex geological concepts; 3. Industrial applications: providing rapid prototype verification tools for oil-gas companies and environmental institutions; 4. AI training: generating large-scale labeled data for training ML models for geological interpretation.

GeoWorld will expand from 2D to 3D modeling, 4D evolution (temporal changes of geological processes), and 5D analysis (probabilistic models for uncertainty quantification) in the future. Technically, it plans to support pressure/stress evolution, geochemical processes, well logging data integration, and agent-assisted workflows. In summary, GeoWorld represents the development of Earth science tools towards programmability, reproducibility, and AI assistance, and is expected to lower research barriers and become an important infrastructure in the field.