I-WRF: Containerized WRF, MET, and MET Plus for Portability, Scaling, and Outreach

The I-WRF project for the NSF Cyberinfrastructure for Sustained Scientific Innovation program provides a framework using application containers that allow WRF, MET, and METplus to be run on a wide range of resources without intensive software installation and management requirements. I-WRF is intended to lower the bar for multi-disciplinary researchers to facilitate wider use of atmospheric modeling and simulation, providing the ability to run WRF easily on desktops, cloud resources, and supercomputers, without the need for configuring and deploying individual elements on their own. By using an application container approach, deployment of the framework on new resources and in new contexts will be simplified.

I-WRF is supporting three science use cases to test the efficacy of the application container in multiple environments. The initial use case for the I-WRF application container is Land Use/Land Cover (LULC) analysis using a ``storyline'' approach. Recent work has examined the impact of land use on changing
climate factors. The inital I-WRF use case incorporates two deep convection events: March 28, 2017 and July 4, 2017. Both events included an organized line of deep convection that moved over Dallas-Fort Worth (DFW) from west to east. To investigate urban influence, we performed simulations using no LULC, current day LULC
and then again with modified surface conditions, which increase DFW to twice, four, and eight times its original size. The resulting influence of increased urban land use on the simulated deep convection events exhibits rainfall rate which scales with the size of the simulated DFW area, both in magnitude and spatial extent. Both simulations utilized very high grid cell counts in order to present a sufficiently intensive problem to fully test the capabilities of the containerized WRF software. Further use cases are analyses of climate change impacts on wind and solar renewable energy resources and extending the I-WRF framework to include the WRF-Chemistry model in order to simulate periods of poor air quality during possible future climate conditions.

I-WRF Framework development can be followed on the NCAR WRF GitHub site and on its own Docker Hub registry. The Framework is developed by a joint team between Cornell Center for Advanced Computing and the NCAR Research Application Lab. Current testing is taking place on the NCAR Cheyenne supercomputing system and Cornell Red Cloud OpenStack system. Planned targets for execution include the Jetstream 2 research cloud and Stampede3 supercomputer resources, as well as public cloud provider infrastructure.

With the goal of a portable container framework that can be executed on laptop, cloud, or supercomputer, the I-WRF project will include execution scripts and terraform/ansible cloud scripts to aid researcher adoption and extend the impact of the WRF software stack.