Software Architecture Evaluation of Earth System Models

Earth System Models (ESMs) play a vital role in understanding and assessing climate change and other earth system-related issues. They are complex and long-living software systems, mainly programmed in Fortran, that undergo changes as science progresses. They have a myriad of variants that must be maintained to support reproducing experiment results. In a research context, often with contributions from scientists on non-tenured contracts and without a formal software engineering education, this can lead to architecture erosion, hampering further development and, therefore, scientific progress. Furthermore, it harms code comprehension, introducing risks for the quality of the earth system models. To address these challenges, our goal is to design and study methods for improving the maintainability of ESMs implemented in Fortran. In this paper, we assess two widely used earth system models—namely UVic and MITgcm—combining dynamic software profiling with static code analysis for reverse engineering. We introduce a new approach to module interface discovery in Fortran systems. We provide a detailed analysis of the ESMs’ architectures and report quantitative properties.