HydroFrame Infrastructure: Developments in the Software Behind a National Hydrologic Modeling Framework

The HydroFrame project combines cutting-edge environmental modeling approaches with modern software principals to build an end-to-end workflow for regional and continental scale scientific applications, by enabling modelers to extract static datasets from continental datasets and execute them using high performance computing hardware hosted at Princeton University. In prior work weve provided the capability for users to extract domain data for the ParFlow model at local scales and execute them using freely accessible cloud computing services (i.e. MyBinder.org). We extend this functionality to support larger study areas and timescales by leveraging Kepler for workflow management, Flask for web interface design, and Slurm for asynchronous job execution. Users are currently able to select their desired watershed(s) of interest across version one of the ParFlow-CONUS model. A dashboard was developed to display the status of submitted jobs which allows users to: (1) launch and run a pre-generated, customizable ParFlow-CLM model using their clipped domain files as model inputs; (2) visualize and interact with model outputs using interactive plots; (3) review previous runs and their parameter specifications; (4) re-execute simulations after modifying model inputs; and (5) launch Jupyter notebooks to perform model configuration and data analysis. In this presentation we will discuss the design and infrastructure of the HydroFrame platform, its integration with open-source software packages, local and regional modeling workflows using ParFlow-CLM, and future plans for the platform. Open-source sustainability and software best practices are emphasized in the design of the HydroFrame platform in order to meet the flexible modeling needs of hydrologists, policy makers and planners, and educators. Future iterations of the platform will aim to provide further modeling flexibility through integration with the second generation ParFlow-CONUS model (CONUS2).