Cloud software for enabling community-oriented integrated hydrologic modeling

Abstract:

We have developed and deployed cloud computing and modeling tools that contribute to a software ecosystem that bridges many of the scientific and technological barriers to continental scale hydrologic modeling. Identified as a grand challenge in hydrology, high-resolution, continental-scale simulations are essential to addressing and predicting hydrologic response to a range of stressors on spatial and temporal scales previously unattainable. Modeling at this scale requires large, labor intensive, input datasets that are curated by teams of interdisciplinary scientists. While hosting these datasets in cloud accessible formats is a necessary first step towards community engagement, we must also provide the ability to execute and collaborate around them at local and regional scales to broaden intellectual contribution. The HydroFrame project (www.hydroframe.org) brings together hydrologists and computer scientists to build an end-to-end workflow for continental scale simulation. In previous work, we provided static domain and parameter datasets for the National Water Model (NWM) and Parflow (PF-CONUS) on demand, at regional watershed scales. We extend this functionality by connecting existing cloud applications and tools into a virtual ecosystem that supports extraction of domain and parameter datasets, execution of NWM and PF-CONUS models, and collaboration. This work leverages existing technologies such as the CUAHSI HydroShare (www.hydroshare.org) data repository, Jupyter Notebooks, Docker containers, Binder computing environments, and model specific subsetting codes. This presentation will discuss the cyberinfrastructure design of our cloud-based ecosystem of tools for working with CONUS model domain and parameter datasets, as well as highlight use cases for research, education, and reproducible science.