NAR Website header

Goal 1, Priority 4: Developing Community Models

Developing numerical models and making them available to the scientific community is at the heart of NCAR's research and service to the community. Key activities in this priority are creating and adding to community models, research models, as well as progressing toward creation of an Earth system model.

FY2007 Accomplishments

Discussion of some of the CCSM-related research has already been touched upon, however, details on CCSM development–including development of component models, and associated models of intermediate complexity–will be covered here.

Click to enlarge. One image from an animation depicting global surface warming as simulated by the Community Climate System Model (CCSM), version three. It shows the surface temperature anomalies relative to an 1870-1899 average for each month. Notice the greater difference from average at the poles. This is known as “polar amplification“.

View Animation FY2007, CCSM successes and development include the extensive number of modeling experiments designed to study the effects on various aspects of the climate system, such as the impacts of releasing varying amounts of chemicals into the atmosphere. These experiments were included as part of the IPCC AR4. Another significant accomplishment was creation of a preliminary version of the next-generation CCSM, referred to as CCSM3.5, which includes improved physics in all component models, including:

  • Ocean: Moved to the POP 2 code-base, incorporated a new advection scheme, a revised horizontal viscosity that allows for the correct amplitude of Tropical Instability Waves, an improved mesoscale eddy parameterization, and movement toward higher vertical resolution
  • Sea Ice: Moved to the CICE 4 code base, incorporated a revised ridging scheme, improved snow treatment, improved the shortwave radiation scheme, improved boundary layer exchange, and developed a new melt pond parameterization
  • Land: Created new surface datasets, incorporated an improved canopy integration scheme and scaling of canopy interception, implemented new surface and sub-surface runoff models, developed a simple groundwater model, and new frozen soil scheme, improved the description of soil water availability, introduced a resistance term to reduce excessive soil evaporation, introduced nitrogen limitation on plant productivity, and improved snow age calculations
  • Atmosphere: Adopted the finite volume core, incorporated modifications to deep convective processes, started assessments of new aerosol formulations, new microphysics, a new planetary boundary layer, and shallow convection scheme and new deep convection schemes, revised gravity wave and orographic wave-breaking formulations, hybrid isentropic coordinates, new radiation packages, and horizontal and vertical resolution sensitivities
  • Biogeochemistry: Completed the land model intercomparison, which resulted in adoption of CLM-CN (Community Land Model with Carbon and Nitrogen cycles), a new ocean ecosystem component, and new control simulations that demonstrate a stable, coupled carbon-climate cycle in both ocean and land
  • Software engineering: Made many improvements and progress (with CISL) on the scaling of CCSM to massively parallel architectures

When coupled together, CCSM3.5 demonstrates significant improvement to many of the system biases that plagued earlier model versions. Additionally, and finally, it is worth noting that scientists from around the globe continue to download CCSM data and the source code. Downloads of the CCSM3 source code, for instance, are nearing 900.

ESSL/MMM’s Weather Research Forecasting (WRF) model is also community built; the powerhouse in the suite of WRF models is the Advanced Research WRF (ARW) system. ARW provides users with a data assimilation capability, a variety of physics packages, and the WRF Software Environment, together with variants, such as WRF-Chem (for atmospheric chemistry modeling). In FY2007, ARW went from strength to strength, and is now the most popular (as defined by registration numbers and use) mesoscale atmospheric model in the world. As of July 2007, ARW had more than 5,300 registered users in more than 90 countries around the world, and is in operational use in approximately 20 locations. In 2007, the ARW was chosen as the basis for the new version of the NCEP Rapid Refresh modeling system.

Issued to the community in December 2006, Version 2.2 is the most recent major release of the ARW. In 2007, MMM conducted three tutorials for WRF system users (two in Boulder, Colorado, one in Korea), and two major WRF workshops. The team is currently looking at establishing an external advisory board in order to provide community input on WRF code management and community support activities.

Click to enlarge. One frame from an animation of modeled precipitable water -- day 10 of 50-km Global ARW simulation, valid 1200 UTC 22 July 2007.

View Animation The ARW continued to be updated and expanded in capability throughout this past year. Its nesting capability and adaptability have proven robust, with up to five nests being run in applications, and with the HYCOM (Hybrid Coordinate Ocean Model) being coupled to the ARW. Both of these projects were done in collaboration with the University of Miami’s Rosenstiel School of Marine and Atmospheric Science (RSMAS). In another collaborative project involving CalTech, WRF was adapted for use as a global model. This version is currently undergoing testing, but is projected to be included in the next major community release, slated for March 2008. Please see the ESSL LAR for further ARW and WRF highlights.

The Developmental Testbed Center (DTC), within RAL’s Joint Numerical Testbed, is a national facility created in 2003 to facilitate interaction between the operational and research communities in accelerating improvement of Numerical Weather Prediction (NWP) for the U.S. In 2007 the DTC:

  • Click to enlarge. 3-h total precipitation (shaded), mean sea level pressure, and 1000-500 mb thickness fields for 60-h forecasts valid at 12 UTC on 3 May 2006. Right panel shows the ARW forecast and left panel shows the NMM forecast. Both WRF configurations used NAM initial and lateral boundary conditions and the same suite of physics parameterizations. For this particular forecast cycle, the ARW and NMM forecasts show rather different evolutions of the cyclone for this extended lead time. Conducted an extended Core Test to determine whether the small differences in forecast skill between the two dynamic cores (AWR and NMM) for 24-hour lead times also pertained to longer lead times (i.e., 60 hours). The actual runs are underway, and results will be available in early FY2008. The DTC is also performing a platform comparison to test whether forecast skill is dependent on the computing platform used to generate the forecasts.
  • Continued to develop the concept of Reference and Operational Configurations for the community by creating a written document describing the process for designating these configurations, the information and support that will be provided for these configurations, and how decisions will be made with respect to retiring configurations.
  • Brought a number of visitors to NCAR to provide support in the areas of physics parameterizations, ensembles, verification techniques, and idealized capability for the NMM dynamic core.
  • Conducted the first joint WRF Tutorial in July 2007, covering both the NMM and ARW dynamic cores. These tutorials include lectures on the pre-processor, model, and post-processing tools, as well as practical sessions that allow the participants to gain hands-on experience building and running each component of the end-to-end system.
  • Developed a state-of-the-art verification toolkit in collaboration with the NWP community. A beta release of the new Meterorological Evaluation Tools (MET) was made available in July 2007, along with a website offering extensive information on how to setup and run the package. MET uses components already present in verification systems developed at NCEP, GSD, and AFWA, as well as new components for the routine calculation of confidence intervals and significance statistics and object-based verification methods.

The Data Assimilation Testbed Center, organized within RAL’s Joint Numerical Testbed in 2006, provides a home for testing and evaluating new data assimilation techniques, inputs, and strategies. In 2007 scientists:

  • Click to enlarge. 36hr WRF forecast verification against Antarctic sondes: Conventional observations only (red), conventional plus COSMIC (green), and retuned conventional plus COSMIC (blue). Performed a month-long evaluation of the impact of COSMIC local refractivity observations on Antarctic weather forecasts. The study found a significant positive impact of COSMIC refractivity assimilation on wind forecasts, tropospheric temperature and surface pressure. This study provided a rational and scientific basis for the operational implementation of COSMIC data in AMPS and also indicated areas of the WRF model that require further attention.
  • Supported the Air Force Weather Agency’s anticipated worldwide regional implementation of WRF-ARW and WRF-Var. Initial studies are focused on optimal data assimilation/forecast configurations on a South-West Asia regional domain. Currently, the data assimilation is performed every six hours, with the cycle being broken every twelve hours to blend back to NCEP’s global forecast system. The possibility of full-cycling (i.e. continuously cycling WRF-Var and WRF-ARW without reverting back to the global model data) has been tested in DATC in 2007. Results indicate that both cycling permutations produce superior forecasts to those run without regional data assimilation, with update-cycling producing the most accurate forecasts. These results provide a benchmark for further full-cycling experiments to assess the impact of AMSU and AIRS radiances, COSMIC, and tuned error covariance.

Click to enlarge. U-wind analysis increment response to a single temperature observation at 50N, 150W for both the “pure 3D-Var“ technique (left) and the hybrid (right). Unlike the 3D-Var, the hybrid response is flow-dependent. Other progress made in FY2007 by ESSL/MMM scientists and university collaborators included developing efficient assimilation of radiances from NOAA polar-orbiting satellites in WRF-Var, focused initially on data from the Advanced Microwave Sounding Unit (AMSU). In addition, development of unified variational/ensemble data assimilation capability for WRF was furthered by incorporating an Ensemble Transform Kalman Filter (ETKF) and flow-dependent forecast error covariances in WRF-Var.

A polar version of WRF is being implemented and tested in the Antarctic Mesoscale Prediction System (AMPS), which is run by ESSL/MMM in support of the U.S. Antarctic Program. The polar WRF modifications were originally developed by MMM collaborator, The Ohio State University, and the code reflects adaptations to better represent sea ice, and features unique to extensive ice sheets. Results thus far, from real-time runs over Antarctica, show improvement over the standard WRF.

Click to enlarge. Shown is a 60 h forecast of Hurricane Felix, predicted at Category 5 just prior to landfall. The plot shows contours of windspeed at 10 m elevation along with wind barb symbols every 40 km. The observed Felix was also Category 5 at this time, making landfall about 80 km south of the predicted location. Hurricane investigation is a growing aspect of WRF-system research and development. A new variant of the ARW, the Advanced Hurricane WRF (AHW), was further developed and tested throughout FY2007. AHW includes specialized data assimilation processes (including 3DVar and an ensemble Kalman filter) and a mixed-layer oceanic model. Current efforts are to run a reference forecast, including the simple ocean model, and two additional forecasts with experimental data assimilation for the 2007 hurricane season, all in near-real time.

The incorporation of interactive chemistry capability by ESSL/ACD scientists in the Community Atmosphere Model has made considerable progress over the last year, and now encompasses a variety of options to accommodate the needs of the coupled climate model. In particular, using the implemented MOZART framework, CAM-Chem can now be configured to combine prognostic and diagnostic variables. As a result, aerosols can either be prescribed, simulated using simple input oxidant fields, or simulated using the full MOZART-4 aerosol parameterization. More details and further MOZART-3 and MOZART-4 accomplishments are available in the ESSL LAR.

In FY2007, the upper boundary of the Whole Atmosphere Community Climate Model (WACCM), another community-derived model, was extended upward to about 500 km, which is also the upper boundary of the TIME-GCM. In the process of achieving this upward extension, several modifications to the model were made, including:

  • Implementation of modules to resolve the major species diffusion, which becomes increasingly important above 110 km.
  • Implementation of modules and revision of codes to reflect the constituent-dependency of the specific heats, gas constant, and mean molecular weight.
  • Revision of the treatment of the vertical diffusion equations for minor species and the heat conduction equation to a formulation appropriate for the upper atmosphere.
  • Further WACCM and TIME-GCM details are available in the ESSL LAR.

NCAR continued hosting the IGBP’s Earth System modeling project, AIMES. AIMES activities relevant to this priority include the Coupled Carbon Cycle-Climate Model Intercomparison Project (C4MIP), which investigates model benchmark and evaluation exercises to explore mechanisms that influence the response of the terrestrial carbon cycle including soil moisture and net primary production, particularly in the tropics, effects of CO2 fertilization, and disturbance and land cover. Find more details on AIMES in ESSL’s LAR.

Models, whether created by the community, individuals, or organizations, are becoming computationally ever more complex. Running such models will require petascale systems, which have between 100,000 and 1,000,000 processors. Such systems are expected to be in production by the end of this decade. It is of supreme strategic importance that the geosciences be positioned to use these systems. For the application developer, the massive levels of parallelism required to exploit these systems present daunting challenges at the algorithm level to achieve scalability.

In FY2007, CISL continued to lead the effort to understand these trends, prepared NCAR community models for the petascale, and also led the broader geoscience community toward petascale computing. In particular:

  • CISL, the National Center for Supercomputing Applications (NCSA), and the San Diego Supercomputing Center (SDSC) partnered to put on the second Geosciences Application Requirements for Petascale Architectures (GARPA-2) in San Diego, California in February 2007. This meeting brought computer scientists, application experts, and vendors together for two days to report results and exchange ideas and information.
  • John Dennis has continued work to further parallelize CCSM components, including POP, CICE, and, in collaboration with Tony Craig, CLM. John has demonstrated POP and CICE running on 30,000 processors, has created a prototype Parallel I/O (PIO) library for use with these components, and won time on the IBM Blue Gene/L system to do a global eddy-resolving ocean climate experiment with Frank Bryan and others.
  • CISL continues to work with researchers from NCAR, University of Colorado, and IBM to integrate the High Order Method Modeling Environment (HOMME) with the Community Atmospheric Model. Accomplishments in FY2007, which includes introducing hyper-diffusivity and local mass conservation to the model, as well as completing preparations for aqua planet simulations.


Click to enlarge. Societal Resilience System of Systems (SRSS) diagram. Lastly, because the interface between atmospheric processes and human wellbeing is mediated through a multifaceted array of natural and social systems, a pressing need exists to develop a conceptual framework that can guide analyses and case studies of examples (as well as counter-examples) of societal resilience. Toward this end, SERE staff are developing a "societal resilience system of systems" (SRSS) framework that will begin to unravel the complex interactions among the myriad contributors within and across the systems involved. This system was conceptualized in FY2007.

FY2008 Plans for Strategic Priority 4

The current, long-term CCSM project plan is to develop and freeze the next version of the model, CCSM4, by the end of 2008. In addition to CCSM3.5’s improved physics, other component model improvements (e.g., introduction of dynamic vegetation) likely will be included, and new components for the carbon cycle and interactive atmospheric chemistry will be finalized. CCSM4 will be the model used to contribute to the next IPCC report.

CISL plans to continue work on CCSM enhancements to include the flux coupler, CAM, and input/output subsystem components. An important research and development objective for the latter two efforts is integration of a scalable, conservative discontinuous Galerkin-based HOMME dynamical core into CAM. The PIO library will be ported/integrated with the Earth System Modeling Framework (ESMF) to assist in the adoption of that framework in petaflops projects in general, and NCAR community models in particular.

During FY2008, it is expected that the governing framework for SRSS will be completed along with a minimum of four prototype case studies, including one from the energy industry led by Plymouth State University, which demonstrates the applicability and need for the SRSS network.

Other FY2008 plans include:

  • Analyzing initial NRCM data from–and continuing experiments using–the NASA Columbia computer for another six years of climate simulations (2000-2005). This effort will include taking the next critical step of embedding ARW into CAM to undertake both current and future climate simulations.
  • Quantifying the role of convective organization in the Madden-Julian Oscillation (MJO) through dynamical and numerical modeling by:
    • Designing new mesoscale parameterizations for climate models.
    • Multi-scale simulations of natural precipitating systems observed by Tropical Rainfall Measuring Mission (TRMM)/CloudSat.
    • Analysis of multi-scale convective organization in the aforementioned tropical channel model.
  • Having examined the historical record of tropical cyclones in the North Atlantic to identify trends in frequency and long-period variability in FY2007, this work will be extended to other ocean basins. NCAR scientists will also utilize the new NCAR NRCM to examine environmental changes associated with these observed trends and variations of tropical cyclones.
  • A full trial of the AHW has been proposed to NOAA as a collaborative effort between NCAR, NOAA, and RSMAS.
  • The range of instruments that can be used within WRF-Var will be expanded to include the Advanced Infrared Spectroradiometer (AIRS) and the Special Sensor Microwave Imager/Spectroradiometer (SSMI/S). In addition, the hybrid approach for the variational/ensemble data assimilation capability for WRF will be tested in a variety of applications, including the full Ensemble Kalman Filter (EnKF) within WRF-Var to create a unified variational/ensemble data assimilation for WRF. And, to accommodate the growing user community and its needs, WRF Users Workshops and tutorials will be conducted in January and July 2008.
  • The DTC will plan and execute another Core Test comparing the forecast skill of the two dynamic cores for higher resolution forecasts (e.g., grid spacing on the order of one km); add data assimilation to its end-to-end system for testing and evaluation. the DTC plans to offer the first official release of MET to the community; and continue to conduct tutorials as well as its visitor program.
  • Work will continue to extend the capabilities of MET to include a broader spectrum of verification capabilities, development of an online tutorial, and inclusion of MET in the WRF Tutorial in July 2008. Members of the verification community will also be invited to join DTC staff for a workshop to be held in spring 2008 to discuss new capabilities for MET and the development of a verification system in which MET could reside.
  • The DATC’s Antarctic, Korean, and Taiwanese WRF NWP testbeds will be supplemented by the first reanalysis testbed, a 10-year Arctic system reanalysis based on a 10-year (2000-2010) period and the WRF-ARW enhanced with polar physics. The group will also begin to test and support the JCSDA’s GSI algorithm in 2008; this s effort is part of the larger plan to work with the JCSDA (including NASA, the Navy, NOAA, and AFWA) in developing next-generation assimilation algorithms suitable for both research and operational communities.WACCM capabilities will be extended to include the top of the thermosphere, and to implement additional thermospheric and ionospheric processes in the model.
  • WACCM capabilities will be extended to include the top of the thermosphere, and to implement additional thermospheric and ionospheric processes in the model.
  • In anticipation of full incorporation of ionospheric processes into WACCM, a community release of the TIE-GCM will be issued, and the CMIT v. 2.0 will be transitioned to the Community Coordinated Modeling Center (CCMC).
  • The massively parallel version of CMIT now in development–v. 2.5–will be tested at the Space Weather Prediction Center and at Boston University in partnership with the Center for Integrated Space Weather Modeling. These activities will set the stage for coupling space weather modeling elements into the ESM.
  • CAM-Chem plans include continued evaluation of the model performance under the different options now available. Additional work will include continued use of MOZART-4 for analysis of the MIRAGE and NASA/INTEX-B field experiments, and MOZART-3 and MOZART-4 data will be published and released to the public.

Related Lab Annual Report Sections:
Goal 1, Priority 4