Moho depth disposition of the contiguous United States: A multi-modal data driven approach
Year: 2025
Type: article
Abstract: This repository supports the manuscript “Moho depth disposition of the contiguous United States: A multi-modal data-driven approach” by providing code, data, and documentation to reproduce all figures, tables, and results. It implements a Random Forest Regressor with von Kármán and Gaussian noise for uncertainty quantification, using geophysical predictors (e.g., gravity, heat flow, sediment thickness). Key Features: Reproducible Pipeline: Scripts in src/ generate all outputs in figures/ and results/ from data in data/. Noise Modeling: Adds spatially correlated (von Kármán) and random (Gaussian) noise, with 50 uncertainty distributions per feature (see automate-noise-addition_uncertainty2.py). Comprehensive Outputs: Includes SHAP values, permutation importance, and sensitivity analyses (e.g., von_karman_sensitivity_results.csv). Quick Start: Run data preparation: python src/gravity_handling.py and python src/automate-noise-addition.py Train model: python src/US_NEW.py Error estimation: python src/automate-noise-addition_uncertainity2.py See README.md for detailed instructions, repository layout, and script descriptions. Contact: Ramees R. Mir: [email protected] Sameem Makhdoomi: [email protected] Please cite our paper if using this repository (citation forthcoming).
Author: Mir, Ramees R.
Institution: Tongji University
Language: English
Subfield: Geology
Field: Earth and Planetary Sciences
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6930725791