Building cyberinfrastructure for the reuse and reproducibility of complex hydrologic modeling studies

Abstract

Building cyberinfrastructure for the reuse and reproducibility of large-scale hydrologic modeling studies requires overcoming a number of data management and software architecture challenges. The objective of this research is to advance the cyberinfrastructure needed to overcome some of these challenges to make such computational hydrologic studies easier to reuse and reproduce. We present novel cyberinfrastructure capable of integrating HydroShare (an online data repository), CyberGIS-Jupyter for Water and high performance computing (HPC) resources (computational environments), and the Structure for Unifying Multiple Modeling Alternatives (SUMMA) hydrologic modeling framework through its application programming interface for orchestrating model runs. The cyberinfrastructure is demonstrated for a complex computational modeling study on a contiguous United States dataset. We present and discuss key capabilities of the cyberinfrastructure including (1) containerization for portability across compute environments, (2) Globus for large data transfers, (3) a Jupyter gateway to HPC environments, and (4) Jupyter notebooks for capturing the modeling workflows.