In 1989, the Center for Analysis and Prediction of Storms was established at the University of Oklahoma as one of the National Science Foundation's first 11 Science and Technology (S&T) Centers. Its formal mission is to demonstrate the practicability of storm-scale numerical weather prediction and to develop, test, and validate a regional forecast system appropriate for operational, commercial, and research applications. Its ultimate vision is to make available a fully functioning stormscale NWP system around the turn of the century.
Central to achieving this goal is an entirely new three-dimensional, nonhydrostatic model system known as the Advanced Regional Prediction System (ARPS). It is a entirely new and complete numerical prediction system designed for the explicit representation of convective and cold-season storms. It includes a data ingest, quality control, and objective analysis package known as ADAS (ARPS Data Analysis System), a single-Doppler radar parameter retrieval and assimilation system known as ARPSDAS (ARPS Data Assimilation System, of which ADAS is a component), the prediction model itself, which is the topic of this paper, and a post-processing package known as ARPSPLT. These components are illustrated in the following figure.
In planning for its development, the ARPS was required to meet a number of criteria. First, it had to accommodate, through various assimilation strategies, new data of higher temporal and spatial density (e.g., WSR-88D data) than had traditionally been available. Second, the model had to serve as an effective tool for studying the dynamics and predictability of storm-scale weather in both idealized and more realistic settings. It must also handle atmospheric phenomena ranging from regional scales down to micro-scales as interactions across this spectrum are known to have profoundly important impacts on storm-scale phenomena. These needs required that the model have a flexible and general dynamic framework and include comprehensive physical processes. The system should also run efficiently on massively parallel computers. In short, it was our goal to develop a model system that can be used effectively for both basic atmospheric research and operational numerical weather prediction, on scales ranging from regional to micro-scales.
The numerical forecast component of the ARPS is a three-dimensional, nonhydrostatic compressible model in generalized terrain-following coordinates that has been designed to run on a variety of computing platforms ranging from single-processor scalar workstations to massively parallel scalar and scalar-vector processors. This highly modular code is extensively documented and has been written using a consistent style throughout to promote ease of learning and modifications well as maintainability. The present version contains a comprehensive physics package and has been applied successfully during the past few years to real-time operational prediction of storm-scale weather over the Southern Great Plains of the United States.
Principal elements of the Advanced Regional Prediction System (ARPS)
Current Features and Capabilities of ARPS
After almost six years of development and testing, the ARPS model now contains physics and numerical solution options consistent with most other non-hydrostatic codes. It does, however, offer a number of unique capabilities in documentation, code structure, scalability on parallel platforms, and ease of use, and thus we summarize below the current features of the system and highlight with underlining those which, in our judgement, are unique to the ARPS. Specific accomplishments for 1997 are shown in italics.
- Computer Language - Fortran-77 with Fortan-90 extensions.
- Documentation - Extensive in-code documentation along with a comprehensive user’s guide.
- Code Design - Fully self-contained codes that are completely portable among both conventional vector-scalar machines (e.g., Cray J90, C90, T90, and workstations and PCs) as well as massively parallel architectures (e.g., T3E, SP2, distributed homogeneous or heterogeneous clusters). The model system is written with a single consistent coding style using industry-standard practices to ensure readability, maintainability, and ease of modification.
- Code Structure - The ARPS subroutines are organized by functionality, and the entire software system is divided into sub-directories based on code type and purpose.
- Availability - All source code and documentation are available via the CAPS web site (http://www.caps.ou.edu) or an anonymous ftp server (ftp.caps.ou.edu), including PDF and postscript versions of the user’s guide.
- User Support - An e-mail based user support system has been in place for several years and continues to be an effective mechanism for dealing with user questions and for reporting bugs in the code. An FAQ link has been added to the ARPS page, as well as a posting of user questions and responses. Finally, a training and applications support group has also been established for those users requiring support beyond what is otherwise made available.
- Dynamic Framework - Nonhydrostatic and fully compressible with Boussinesq option.
- Coordinate System - Generalized terrain-following coordinate on the Arakawa C-grid with equal-spacing in the horizontal and user-specified stretching in the vertical.
- Map Projections - Polar stereographic, Lambert Conformal, and Mercator options.
- Domain Geometry - 1-D, 2-D, and 3-D configurations.
- Prognostic Variables - Cartesian wind components, perturbation potential temperature and pressure, subgrid-scale turbulent kinetic energy, mixing ratios for water vapor, cloud water, rainwater, cloud ice, snow and graupel/hail.
- Spatial Discretization - Options for second-order quadratically-conservative, fourth- order quadratically-conservative, Zalesak’s multi-dimensional flux corrected transport (FCT; positive definite), and multidimensional positive definite centered difference (MPDCD) finite difference schemes for advection. Second-order centered differences are used for all other terms.
- Temporal Discretization - Second-order leapfrog scheme for large time steps with Asselin time filter option. First-order forward-backward explicit with second-order centered implicit option for small (acoustic mode) time steps.
- Solution Technique - Split-explicit (mode-splitting) with vertically-implicit option.
- Initial State - Options for horizontally-homogeneous initialization using a single sounding or analytic functions, or a three-dimensional horizontally inhomogeneous state.
- Lateral Boundary Conditions - Options for periodic, rigid, zero-gradient, wave- radiating, externally-forced, and user-specified conditions. All can be mixed and matched.
- Top & Bottom Boundary Conditions - Options for rigid, zero-gradient, periodic, Durran-Klemp radiation, and Rayleigh sponge layer.
- Divergence Damping - The model provides an option for divergence damping to control acoustic oscillations.
- Reference Frame Rotation - Options for inclusion of some or all Coriolis terms.
- Domain Translation - Options for user-specified or automated (based on feature- tracking algorithms) translation of the computational domain for horizontally homogeneous environments.
- Adaptive Mesh Refinement (AMR) - The Skamarock AMR interface is available on shared memory machines for using unlimited levels of grid nesting at arbitrary locations and orientations specified at run time. One-way interactive self-nesting is also available.
- Subgrid Scale Turbulence - Options include Smagorinsky-Lilly diagnostic first-order closure, 1.5-order turbulent kinetic energy formulation, and Germano dynamic closure. The model also provides options for isotropic and anisotropic turbulence based upon grid aspect ratio.
- Spatial Computational Mixing - 2nd- and 4th-order options.
- PBL Scheme - Convective PBL turbulence based on TKE scheme.
- Cloud Microphysics - Options for Kessler warm-rain, Lin-Tao 3-category ice, and Schultz simplified ice NEM parameterizations. The Lin-Tao scheme is now almost as computationally efficient as the Schultz scheme due to the use of look-up tables and other optimization strategies.
- Cumulus Parameterization - Options for Kuo and Kain-Fritsch schemes separately or in combination with other microphysics options.
- Surface Layer Parameterization - Surface momentum, heat, and moisture fluxes based on bulk aerodynamic drag laws as well as stability-dependent formulations.
- Soil Model - Two-layer diffusive soil model with surface energy budget equations. Options are provided for multiple soil types in a single grid cell. An API initialization option is also now available.
- Longwave and Shortwave Radiation - Full long- and short-wave radiation capabilities including cloud interaction, cloud shadowing, and terrain gradient effects.
- Surface Data - 1 km resolution (over US) USDA surface characteristics database (soil type, seasonal vegetation type) and pre-processing software.
- Terrain - 5 minute global terrain database, 30 second database for 70% of the earth, and 3 second data for the US. A package is provided for processing these data.
- Real Data Ingest and Analysis - The ARPS Data Analysis System (ADAS) provides the capability to ingest, quality control, and objectively analyze (using the Bratseth or Barnes schemes) virtually any type of observations including WSR-88D Level II data. CAPS currently ingests: NIDS data from over 20 WSR-88D radars; surface and wind profiler observations, rawinsonde observations, Level II data from the Oklahoma City WSR-88D radar, conventional and Oklahoma Mesonet surface observations, output from several NCEP models, and GOES satellite data.
- Links to External Models - Using GRIB and GEMPAK readers, the EXT2ARPS package allows users to initialize and force the inner domain and lateral boundaries of the ARPS with data from other models including the RUC and Eta.
- Data Assimilation - A forward-variational four-dimensional data assimilation system, augmented by a single-Doppler velocity retrieval package, is available. These systems may be used with ARPS or as stand-alone software.
- ARPS Adjoint - The adjoint and tangent linear versions of the warm-rain-option ARPS are available, with the adjoint including the LBFGS minimization package.
- History Dumps - The ARPS supports the following formats: unformatted binary, formatted ASCII, packed binary, NCSA HDF, NetCDF, packed NetCDF, GrADS, GRIB, AVS, Savi3D, and Vis5D. These formats can be read by post-processing programs provided with the model or by user-created programs based on a template provided.
- Restart Option - Full restart capability is available at intervals selected by the user.
- Compilation - The compilation of all programs is handled by a single Unix shell script that invokes the Unix make command. Computer system dependencies are automatically handled by the script to facilitate easy migration among platforms and operating systems.
- Execution - Interactive (via a motif X-windows interface) and batch execution are supported for ARPS and its post-processing packages.
- Parallel Processor Options - The ARPS utilizes the PVM and MPI message-passing libraries and a system-independent translator for execution on distributed memory computers and clusters.
- User Interfaces - ARPS and its post-processing packages utilize namelist input files which can be edited manually or configured using a motif X-windows interface that is particularly helpful to new users. In 1997, a web-based ARPS browser was implemented using Pearl scripts.
- System Automation - The entire forecast system, including data acquisition, quality control, analysis, retrieval, assimilation, forecast model execution, and graphical product generation and display (on the Web) is 100% automated by Unix shell scripts.
- Code Validation - A suite of code validation tests is available, ranging from basic advection and symmetry tests to analytic Navier-Stokes solutions and 3-D storm- and meso-scale simulations.
- Sample Datasets - CAPS provides a complete horizontally inhomogeneous sample dataset for users interested in exploring the full capabilities of the model.
- Graphical Post-Processing and Analysis - A vector graphics post-processing package known as ARPSPlt is available for generating color plots, 3-D wire frames, and profiles of basic and derived fields using model-generated history data. The package supports overlays, color filling, user-specified contour intervals and annotation, and multiple picture formats. It is based on ZXPLOT, a vector graphics package similar to NCAR Graphics that performs a variety of graphics functions and supports X-windows, GKS, and postscript functionality. The ZXPLOT object code (only) is currently available free of charge and is required for using ARPSPlt.
- Decision Support System - A web-based decision support system known as ARPSView is available for the display of basic and derived quantities from the model forecasts. This system is fully automated with Unix shell scripts, and can be examined at http://throttle.ou.edu.
- Additional Analysis Tools - A combination of software packages supplied by both local and external users is available in ARPSTools. Capabilities include time-dependent trajectories, thermodynamic diagrams and hodographs, and various statistics.
Recently Added Features and Improvement in ARPS
1997 Annual Report and 1996 Annual Report contains more detailed information on CAPS and ARPS.