Improving numerical weather prediction systems depends critically on the ability to transition innovations from research to operations (R2O) and to provide feedback from operations to research (O2R). This R2O2R cycle, sometimes referred to as "crossing the valley of death", has long been identified as a major challenge for the U.S. weather enterprise.
As part of a broader effort to bridge this gap and advance U.S. weather prediction capabilities, the Developmental Testbed Center (DTC) with staff at NOAA and NCAR has developed the Common Community Physics Package (CCPP) for application in NOAA's Unified Forecasting System (UFS). The CCPP consists of a library of physical parameterizations from NOAA, NCAR and other organizations, and of a software framework, which connects the physics with atmospheric models based on metadata and standardized interfaces. The range of physics options in the CCPP physics library enables the application of the UFS - as well as every other model using the CCPP - across scales, from nowcasting to seasonal and from high-resolution regional to global.
While the initial development of the CCPP was centered around the FV3 (Finite-Volume Cubed-Sphere) dynamical core of the UFS, its focus has since widened. The CCPP is also used by the DTC Single Column Model to support a hierarchical testing strategy, and by the next generation NEPTUNE (Navy Environmental Prediction sysTem Utilizing the Numa - Nonhydrostatic Unified Model of the Atmosphere - corE) model of the Naval Research Laboratory. Further, NOAA and NCAR recently signed an agreement to jointly develop the CCPP framework as a single, standardized way to interface physics with their models of the atmosphere (and other components of the Earth system).
Improving the interoperability and portability of model physics requires a generalized and model-agnostic development for both the framework and the physical parameterizations. At the same time, advances in high performance computing systems that increase the computational performance in times of stagnating processor speeds and power limitations mandate a high degree of optimization and customization from future numerical modeling systems. The need to scale out to orders of magnitude larger numbers of cores and to run efficiently on a wide range of architectures embedded in a variety of modeling systems requires a careful design and implementation of interoperable components such as the CCPP.
In this contribution, we will provide a brief overview of the general concept of the CCPP, present the technical design chosen to meet these requirements and report on the progress towards meeting those. We will describe the integration of the CCPP in the host model and touch upon the challenges in creating a flexible modeling framework while maintaining high computational performance. We will also provide information on how to obtain, use and contribute to the CCPP, as well as on the future development of the CCPP framework and upcoming additions to the CCPP physics library.
