From topography to process: A Python toolkit for landscape evolution analysis

Abstract

Python can help solve a fascinating inverse problem in geomorphology: inferring the processes that shaped Earth's landscapes from current topography. While machine learning offers a direct approach, limited real-world labeled data poses a significant challenge. To address this, we developed an original research toolkit that integrates physics-based and machine learning models for solving the inverse problem. The toolkit has three components: (1) a landscape generator built with Landlab that creates synthetic landscapes with known process parameters; (2) a feature extraction framework using NumPy, SciPy, and NetworkX; and (3) a machine learning pipeline with Scikit-learn that estimates process parameters and identifies the most influential features. This talk focuses on the domain-specific aspects of landscape evolution modeling and feature extraction. It demonstrates how Python helps in advancing geomorphological theory and in deepening our understanding of landscape evolution.

Description