Global Mesoscale Modeling with GEOS-5: Global Weather at Local Scales

Overview

NASA's Global Modeling and Assimilation Office (GMAO) has developed the Goddard Earth Observing System Model, Version 5 (GEOS-5) with the capability to produce global weather at local scales much finer than the highest-resolution global weather models used around the world. This capability enhances our ability to maximize the impact of satellite observations in climate, weather, and atmospheric composition prediction. Exploring the impact of new observations is a key element of the GMAO's mission. Through a process known as observing system simulation experiments (OSSEs), scientists at NASA and the National Oceanic and Atmospheric Administration (NOAA) can estimate the usefulness of various configurations of observing systems in an entirely simulated context. We have produced a 2-year, 7-kilometer (km), non-hydrostatic global mesoscale simulation using GEOS-5 as a reference Nature Run for OSSEs. We refer to this simulation as the 7-km GEOS-5 Nature Run or 7km-G5NR. The 7km-G5NR serves as a virtual Earth, producing synthetic observations for OSSEs to explore the potential of future observing systems.

Project Details

By incorporating a non-hydrostatic, finite-volume dynamical core with scale-aware physics parameterizations, the 7km-G5NR produces organized convective systems and robust weather systems ideal for producing observations for existing and new remote sensing instruments. In addition to standard meteorological fields, the 7km-G5NR includes a suite of aerosol types (dust, sea salt, sulfate, and black and organic carbon) and trace gas concentrations (ozone, carbon monoxide, and carbon dioxide), with emissions downscaled to 10 km using ancillary information such as power plant location, population density, and night-light information. This free-running 7km-G5NR develops its own weather around the Earth in a purely synthetic environment constrained only by surface boundary conditions for sea-surface temperatures and sea ice, daily volcanic and biomass burning emissions, as well as high-resolution inventories of anthropogenic sources. This produces a long, dynamically consistent atmospheric state that matches the climatology and variability of the real atmosphere.

Results and Impact

The fidelity of physical processes represented in the 7km-G5NR resembles the details of the observational data, including eye-wall structure and reformation in intense tropical cyclones, towering cumulus convection within mesoscale convective complexes, and thunderstorm activity along frontal boundaries. Several seasonal simulations with GEOS-5 at 3.5-km global resolution will produce more detailed synthetic observations approaching the resolution of most modern satellite observing systems. These global mesoscale simulations can be applied to "downscale" coarse-resolution climate simulations to explore local impacts. This downscaling has been demonstrated with GEOS-5 over the observational period 2005–2006. Unlike the 7km-G5NR, this downscaling "replay" is constrained by the observations, producing local details of the specific weather events observed over this period within the context of a comprehensive global simulation.

Why HPC Matters

Recent computing advances have permitted experimentation with global atmospheric models at scales typically reserved for limited-area regional models. However, production applications like the 7km-G5NR have remained limited. Utilizing the Discover supercomputer (specifically 480 2.8-GHz, 16-core Xeon Sandy Bridge nodes) at the NASA Center for Climate Simulation (NCCS), the 7km-G5NR produced over two years of high-resolution weather and chemistry data at 30-minute intervals for the period May 2005–June 2007. This massive computation created nearly 4 petabytes of output in just over 75 days of dedicated computational time on Discover, consuming nearly 13 million processor-hours.