Title: A Framework for Cloud-Resolving OSSEs
Presenting Author: Thomas Clune
Organization: NASA Goddard Space Flight Center
Co-Author(s): Arlindo da Silva

Abstract:
Traditional approaches to performing Observing System Simulation Experiments (OSSEs) become computationally untenable at the cloud-resolving (convection allowing) spatial/temporal resolutions that are required by the end of the decade. Solutions to these challenges will have many beneficiaries, including NASA's Earth System Observatory (ESO) missions currently in development and the Decadal Survey Incubator missions. To this end, we have designed, and begun implementing a framework for performing global cloud-resolving OSSEs. Storage-limitations require traditional OSSE systems to produce output at relatively low temporal resolutions that are inadequate to resolve processes of interest (e.g., convection.) This new framework is based on a number of key innovations: (1) We extend the parallel I/O capabilities in the GEOS Earth system model by including on-line sampling of geophysical variables, allowing the system to achieve very high temporal resolution that is only limited by the model timestep. This capability allow access to all model geophysical variables and diagnostics interpolated to specific geospatial locations and times: ground stations, aircraft trajectories, satellite swaths, etc. (2) Unlike traditional OSSE system where a Nature Run (NR) is performed once and a pre-determined set of output fields is written to disk, we adopt a 2 phase approach. In the first phase, the model is run with very limited output, except for frequent checkpoints, and a number of browse product for case selection. In the second phase, the model is rerun from spun up checkpoints of interest with output sampled at user specified locations using the sampling capility in (1). This OSSE framework leverages the architecture of the open source Goddard Earth Observing System (GEOS) model which is based upon the Earth System Modeling Framework (ESMF), an ESTO initiated project. This architecture greatly simplifies accessing and efficiently interpolating model geophysical quantities to at arbitrary geospatial positions. In this talk, we will first elaborate in more detail the design and implementation of our framework. We then conclude with current status including a demonstration of the technology with a 3km global resolution NR.