Working With JEDI—An Outside Perspective

We are investigating the data assimilation (DA) requirements and opportunities that exascale computing presents for cloud resolving numerical weather prediction models. This includes requirements for the I/O and performance needed to meet operational time to delivery requirements. It also includes opportunities for revisiting traditional methods that are currently impractical due to their computational cost, as well as exploring the use of new assimilation techniques and use of novel HPC architectures. The Joint Effort for Data assimilation Integration (JEDI) is a core component of this work and provides a flexible platform with a rigorous software development strategy to conduct our research. Our initial step was to familiarize ourselves with JEDI by contributing a simplified parallel toy “bathtub” shallow water model to JEDI. This presentation describes our experience integrating our toy model into JEDI and then using JEDI to run assimilation experiments with that model.