A framework for assessing the computational reproducibility of geo-simulation experiments

Abstract

Recent advances in computational technologies have enhanced geo-simulation experiments (GSEs), making computational reproducibility assessments increasingly critical. However, existing methods often focus on isolated aspects, lacking a comprehensive framework. This study proposes an integrated framework for assessing reproducibility in GSEs, structured into two parts: (1) evaluating overall computational workflows, and (2) investigating individual processes to identify inconsistencies. The framework employs a detailed assessment model using hierarchical dimensions and metrics that combine quantitative measures (e.g., output consistency) and qualitative evaluations (e.g., clarity of descriptions). These components address both broad and granular aspects of computational processes. The framework is implemented in a prototype system to support reproducibility assessments and demonstrated through practical applications. This systematic approach provides a robust and adaptable method for assessing reproducibility, promoting the resolution of challenges in existing methods.