Operational Implementation of a JEDI-Based Data Assimilation System for The Weather Company’s Global High-Resolution Atmospheric Forecasting (GRAF) System

The Weather Company (TWC) has been providing operational and convective-allowing global MPAS (Model for Prediction Across Scales) forecasts since 2019, in part due to a collaboration with UCAR on high performance computing (HPC) initiatives based on IBM’s POWER9 processor and NVIDIA Tesla V100 GPU acceleration. This framework - the Global High-Resolution Atmospheric Forecasting (GRAF) System – provides hourly-updating 72-hr forecasts and currently operates on a variable resolution 15/4km mesh consisting of 4.79 million cells. GRAF has been providing hourly-updating precipitation forecasts for TWC’s short-term weather forecast platforms and has become a foundational asset and tool across media applications, including widespread use throughout the broadcast television weather industry.

More recently, the HPC system has undergone a significant expansion. Now with over 200 compute nodes, a more advanced TWC weather prediction system is becoming a reality. As planned, data assimilation (DA) continues to play an integral role for next-generation GRAF, with a transition from a partially-cycled GSI (Gridpoint Statistical Interpolation) system to a fully-cycled, rapidly-updating JEDI (Joint Effort for Data assimilation Integration) implementation. TWC JEDI development will progress from 3DVar to Hybrid 3D- and 4D-EnVar, using a wide range of observation datasets such as radar, satellite, conventional, and TWC proprietary sources. The ensemble component of the hybrid DA approach will be based on an in-house MPAS configuration.

TWC’s recent DA development efforts have yielded the first phase of operational MPAS-JEDI based on a partially-cycled environment, which also includes an in-house cloud analysis after the traditional DA component. The assimilation is done on a native 12-km MPAS mesh (with corresponding 12-km B-matrix), and the increments are interpolated back to the 15/4km background mesh. The system has been automated with an NCAR JEDI-based workflow using GPU MPAS for the short-term forecast step.

This presentation will provide an overview of current JEDI-driven GRAF, ongoing and future JEDI-related work, and a roadmap summary including the future of AI for GRAF.