ObjectivesThe primary focus of the Regional Analysis and Prediction Branch is research and development of the Rapid Update Cycle (RUC) and its development version, the Mesoscale Analysis and Prediction System (MAPS), which provide high-frequency, hourly analyses of conventional and new data sources over the contiguous United States and short-range numerical forecasts (out to 12 hours) in support of aviation and other mesoscale forecast users. The RUC runs operationally at the National Centers for Environmental Prediction (NCEP) at the highest frequency among its suite of operational models. In addition to developing and testing improvements to the RUC at FSL, another key focus of the branch has been the actual transfer of improvements made in MAPS at FSL to the RUC running at NCEP. The RAP Branch works closely with NCEP in implementing these improvements to the RUC. A variety of model and assimilation development, verification, and observational data investigation activities are carried out under the RUC/MAPS focus.
The RUC has a unique role within the National Weather Service (NWS) in that it is the only operational system that provides updated national-scale numerical analyses and forecasts more often than once every 6 hours. It was developed in response to the needs of the aviation community and other forecast users for high-frequency, mesoscale analyses and short-range forecasts covering the conterminous U.S. It is a critical part of the NWS modernization program, and has become widely used in NWS Forecast Offices and other facilities. Evaluations of MAPS and the RUC have clearly demonstrated the RUC's advantage in providing high-frequency, recently initialized forecasts based on the latest observations. The RUC is a key part of the Aviation Weather Program – commercial and general aviation are both critically dependent on accurate short-range forecasts. The RUC will continue to improve over the next several years, thus perpetuating the successful collaboration between FSL and NCEP.
In collaboration with other government agencies and universities, scientists
develop improved data assimilation and modeling methods for use in MAPS and
RUC. New datasets are assimilated as they become available, so that all
potentially useful observations can be incorporated to define the current
atmospheric and surface conditions, thereby producing the most accurate
forecast possible. The branch also interacts with other FSL staff in
implementing optimal computing methods with MAPS/RUC software, making the
model as efficient as possible on modern computing platforms.
A new version of the RUC, RUC-2, was implemented operationally at NCEP on 6 April 1998. Before operational implementation of the RUC-2, a comparison between the RUC-2 and RUC-1 was conducted during a two-month field evaluation. Users were provided access to information and communication with the developers for improving the understanding of the RUC-2 characteristics and capabilities, which was an efficient way to identify and implement needed changes. The RUC-2 is the 40-km version of MAPS, developed and extensively tested at FSL, which produces new three-dimensional analyses and short-range forecasts every hour, compared to every 3 hours in RUC-1. This new version is a significant advance over RUC-1, not just in assimilation frequency, but also in resolution, types of data assimilated, and model physics. These changes allow the RUC-2 to more accurately represent significant weather systems across the United States in all seasons. Subsequent to the RUC-2 initial implementation, FSL has developed smaller improvements to analysis and model physics methods in the RUC and worked with NCEP to implement these changes in the version running there (listed here).
Summary of Characteristics of RUC-2
Computational Grid – RUC-2 covers a geographical domain about 50% larger in area than the 60-km RUC-1, extending farther in all directions, but especially in the southeast, north, and west, and covering considerably more oceanic area. Moving the boundaries slightly farther from the coasts has significantly improved forecasts in these areas. The RUC-2 has 40-km resolution (151 x 113 grid points) and 40 levels, whereas the RUC-1 grid has 60-km resolution (81 x 62 grid points) and 25 levels. The RUC-2 continues to use an isentropic-sigma hybrid vertical coordinate, which follows the terrain at low levels, but follows surfaces of constant potential temperature higher up. The RUC-2 isentropic-hybrid coordinate is able to capture temperature and moisture vertical structures quite accurately, as shown in Figure 2.
Topography – Finer horizontal scale topographic features are depicted in considerably more detail in the RUC-2. The improved topography enables the model to generate many topographically induced weather events, including mountain/valley circulations, mountain waves, sea breezes, and enhanced precipitation on windward slopes. Further improvements in resolution for the RUC are planned, as described below.
Frequency of Assimilating Observations – In the RUC-2, observations are assimilated hourly, as compared to once every three hours in RUC-1, leading to improved short-range forecasts of all variables.
Figure 2. A comparison of rawinsonde and RUC-2 gridded soundings for Lake Charles, Louisiana, 1200 UTC 7 April 1999.
Time of Availability – For the 40-km RUC-2, the data cutoff time is 21 minutes after each hour, meaning that RUC-2 analyses and forecasts are available almost an hour earlier than those from RUC-1, which had a data cutoff time of 1 hour and 20 minutes. As a result of the increased frequency and earlier time availability of RUC-2, users can access forecasts that are produced with the most recent data.
Data Assimilated – The RUC-2 assimilates the following new sources of observations, which were not used in the RUC-1: VAD (Velocity-Azimuth Display) wind profiles, GOES precipitable water estimates, SSM/I precipitable water estimates, and GOES high-density cloud-drift winds (from visible, infrared, and water vapor images).
Moist Physics – In the RUC-2 a comprehensive cloud physics package was imported from the NCAR/Pennsylvania State MM5 mesoscale model. Mixing ratios of five types of hydrometeors are explicitly predicted: cloud water, rain water, snow, ice crystals, and graupel. Even the number density of ice crystals is predicted as part of the microphysical processes. Cloud fields initialize each RUC-2 model run, using the most recent hydrometeor forecast to avoid cloud spin-up problems.
Surface Processes – The RUC-2 features a multilevel soil and vegetation model with evolving soil moisture and temperature fields that are more accurate than climatology. The snow physics package accounts for the processes of snow accumulation on the ground surface and snow melting. The RUC-1 used coarse land use data and derived sea surface temperatures from climatology, and did not consider snow cover. The RUC-2 employs improved land use data, including vegetation class, monthly vegetation fraction from the National Environmental Satellite, Data, and Information Service (NESDIS), and soil type. It utilizes daily detailed fields of sea-surface and lake-surface (for the Great Lakes) temperature. The RUC-2 also cycles snow cover and canopy water along with the soil fields to further improve short-range forecasts.
Turbulence – The RUC-2 calculates the kinetic energy associated with turbulence (TKE) explicitly from equations derived by Burk and Thompson (Level 3.0). Boundary fluxes are improved using the explicit turbulence parameterization, and explicit TKE maxima are commonly found in upper-level frontal zones.
Radiation – Full atmospheric radiation imported from the MM5 model is included in the RUC-2. The radiative heating/cooling is influenced by hydrometeors (clouds) at each model level.
Development of the 40-km RUC
Improved Analysis Techniques (including development of a three-dimensional variational method) – Scientists continued the testing and development of a three-dimensional variational analysis scheme to be implemented in the RUC in Fiscal Year 2000. A method for controlling rotational versus divergent winds in the analysis was developed. Parallel tests show close fit to observations, sufficiently fast speed, and forecasts about equal in accuracy to those from the current optimal interpolation scheme.
Improvements to the Multilevel Soil/Vegetation Model – The performance of the MAPS/RUC land-surface processes component in the cold season was significantly enhanced by the addition of frozen soil physics and a two-layer snow model. With these changes, the evolution of snow cover and soil temperature during the winter season is improved. The goal here is to improve the MAPS/RUC prediction of skin temperature and surface air temperature, and to avoid the significant errors that may result even at short timescales from inaccurate forecasts of snow cover. A comparison of the cycled MAPS snow cover field in January 1999 and a NESDIS snow cover field based on Advanced Very High-Resolution Radiometer (AVHRR) satellite data (Figure 3) indicates very close agreement, even though the MAPS field is based only on snow accumulation and snow melt forecasts. A one-dimensional version of the MAPS/RUC land-surface process model was run for multiyear simulations using observed surface atmospheric conditions as part of the international Project for Intercomparison of Land-surface Parameterization Schemes (PILPS), and showed very good results in comparison to other schemes from various regional and global climate models.
Development of a Cloud Analysis
An initial cloud analysis for the RUC was developed, combining data from GOES cloud-top pressures and from explicit forecasts of cloud water and ice from the RUC using MM5 cloud microphysics. In this initial procedure, hydrometeor mixing ratios from the RUC forecast are modified to be consistent with cloud clearing or cloud building based on GOES cloud information. This RUC cloud analysis has been tested in a parallel 40-km cycle at FSL and has shown positive impact on cloud forecasts from 1 hour and even out to 12 hours in duration. An example of positive impact in a cloud forecast is shown in Figure 4. This technique will be implemented in the operational RUC at NCEP to improve cloud, precipitation, and icing forecasts. Further development of the RUC cloud forecast will be made for the addition of data from radar, surface cloud, lightning, and satellite microwave observations.
Figure 3. (top) Snow water equivalent depth (mm) field from the ongoing MAPS 1-hour cycle on 5 January 1999, and (bottom) the NESDIS snow cover field for the same date.
The cloud analysis effort has also benefitted from comparisons between MAPS/RUC cloud products and various cloud estimates from the ARM/CART site in Oklahoma. In addition, an ongoing intercomparison is being made of surface skin temperatures from MAPS and from GOES.
Production of Integrated Datasets for GEWEX/GCIP – FSL continues its participation in the multiyear Global Energy and Water Cycle Experiment (GEWEX) and its GEWEX Continental-scale Intercomparison Project (GCIP) by providing data from the RUC-2/MAPS model to the GCIP archive of gridded datasets. The goal of GCIP is improved understanding of the continental-scale hydrological cycle components, and ultimately, improved climate prediction capability. Ongoing improvements to all aspects of MAPS, but especially to its land-surface component, contribute toward this goal. MAPS fields integrating various data sources on a high-frequency basis are being used, along with those from the Eta model and the Canadian Regional Finite Element model, to better understand the energy and water cycles over the United States and southern Canada. The continued improvement of the soil/vegetation model in RUC/MAPS contributes strongly to the accuracy of its estimates of critical hydrological fields for GCIP, including surface fluxes and soil moisture. Despite the climate application of GCIP, these same improvements can also be critical for improved short-range forecasts of surface temperature, clouds, and precipitation from the RUC needed by its operational users.
Figure 4. Effect of RUC cloud analysis on RUC cloud forecasts. Cloud-top pressure fields for 1200 UTC 14 May 1999 from (a, top) RUC 1-hour forecast from control cycle with no cloud analysis showing level where combined hydrometeor mixing ratio exceeds 10-6 g g-1, (b, middle) same from parallel cycle with hourly assimilation of GOES cloud-top pressure data with cloud clearing and building, and (c, bottom) NESDIS GOES cloud-top pressure field at verification time.
Testing of a 20-km and higher resolution version of the RUC – Testing of the RUC model was performed at horizontal resolutions of 20 km, 13 km, and 10 km, in expectation of the implementation of higher resolution versions of the RUC at NCEP as NCEP computing capability continues to increase. NCEP has acquired an IBM SP computer in Fiscal Year 1999, and a RUC with approximately 20-km resolution will be implemented on this computer next year. (The 20-km RUC terrain field is shown in Figure 5.) Tests of the RUC model at 13-km resolution (cover) have shown the expected accurate simulation of orographic precipitation and local downslope winds for a significant snowstorm and the forest blowdown storm that occurred in Colorado in October 1997.
Verification of the MAPS/RUC Forecasts – Daily monitoring and verification of the MAPS/RUC temperature, wind, humidity and precipitation analyses and forecasts are continuing as part of an effort to detect coding problems and physical inconsistencies in the analysis scheme and forecast model.
Figure 5. The 20-km RUC terrain field.
The following additional accomplishments were carried out in the Regional Analysis and Prediction Branch during Fiscal Year 1998.
The MAPS/RUC Web Homepage – Continued enhancements were made to the MAPS/RUC homepage, including 36-hour forecasts from the 40-km MAPS model computed four times daily, continuation of the popular RUC user forum, observation counts for MAPS and RUC, and monitoring of NWS "Area and State Forecast Discussions." A RUC users' forum has been continued to provide the latest news on the RUC and answer user questions.
Observation Sensitivity Experiments Using RUC for NAOS – Experiments using the RUC with a 1-hour cycle and forecasts out to 36 hours every 12 hours were conducted for different observation mixes. Three different experiments were carried out for a 1-month test period as part of the multiagency North American Observation System experiment. Similar experiments have been carried out with NCEP's Eta and Aviation model by NCEP.
Verification of Wind Fields for Terminal Airspaces – A collaboration with NASA Ames Research Center was continued to assess RUC wind forecast errors and their impact on use of automated air traffic management in the terminal airspace. As part of this work, a collocation study of ACARS wind and temperature accuracy was completed, resulting in an estimate of horizontal wind component accuracy of 1.1 m s-1 and temperature accuracy of 0.5 K.
Participation in Development of the Weather Research and Forecast (WRF) Model – The overall goal of the WRF Model project is to develop a next-generation mesoscale forecast model and assimilation system that will advance both the understanding and prediction of important mesoscale weather, and promote closer ties between the research and operational forecasting communities. The model is being developed as a collaborative effort among NCAR, NCEP, FSL, and CAPS (Center for the Analysis and Prediction of Storms), together with the participation of a number of university scientists. Because the project aims to improve the forecast accuracy of significant weather on horizontal scales of a few kilometers, the model will be nonhydrostatic. It is anticipated that this model will be a leading candidate to replace the present hydrostatic RUC model to serve the rapid updating function at NCEP.
FSL has continued its work toward a significant option for the WRF forecast model, a version using a generalized vertical coordinate that can move during the model integration. This work is being done in collaboration with scientists from the University of Miami and NCAR. This option allows the use of a smoothed hybrid isentropic/sigma-z coordinate that retains the advantages of the RUC hybrid coordinate on scales of 20 km or greater, but also accommodates local perturbations such as convective clouds or breaking mountain waves with quasi-horizontal coordinates. An example of this is shown in Figure 6, where the quasi-isentropic generalized WRF option with a 2-km resolution successfully simulates a breaking mountain wave.
Development of the Well-Posed Model – The Multiscale Modeling Group has developed a well-posed fourth order accurate limited area model that can be used for any scale of motion anywhere on the globe for the Air Force Office of Scientific Research. In addition, a hot start Bounded Derivative Initialization (BDI) of the multiscale model for the open boundary value problem has been successfully demonstrated with an improved Kuo cumulus parameterization scheme (the BDI theory also allows the use of other diabatic heating schemes). Based on results from the previously published paper entitled "The Role of Gravity Waves in Slowly Varying in Time Mesoscale Motions" and the recently submitted paper entitled "The Role of Gravity Waves in Slowly Varying in Time Equatorial Motions," it is now known that this initialization will provide an accurate gravity wave free approximation of the dominant component of the solution for the large scale, mesoscale, and small scale anywhere on the globe except in the presence of topography. Currently, topographic effects are being included in the initialization package. Once this final case is complete, the initialization theory will be complete and the model/initialization package will be ready for hindcasting and forecasting experiments.
Figure 6. A vertical cross section using the FSL generalized vertical coordinate option for the WRF model for a mountain wave. The mountain height is 2 km, and the horizontal resolution is 2 km. The thick lines are isentropes at 10-K resolution; thin lines are actual model levels: (a, left) sigma-z version, and (b, right) generalized isentropic/sigma-z version.
The Regional Analysis and Prediction Branch will continue to work with the National Centers for Environmental Prediction to improve the Rapid Update Cycle over the next several years. The primary near-term tasks follow.
Transfer of the 40-km RUC to NCEP's new IBM SP Computer – The initial implementation of all models will be at current resolution. Testing of a 20-km RUC to be implemented at NCEP will be done both at FSL and NCEP. This higher-resolution version is expected to improve RUC accuracy in many areas, but especially for cloud, precipitation, and surface forecasts.
Development of a National-Scale Cloud Analysis – Development and real-time testing will continue on the national-scale cloud analysis (also applicable to other domains), combining observations (especially from satellite and radar), and explicit cloud forecasts by the RUC model.
Refinement and Testing of Improved Physical Parameterizations for Soil/Vegetation Processes, Turbulence, Convective Clouds, and Cloud Microphysics – Some of this work will be done in collaboration with NCAR, since the MAPS model uses the MM5 physics.
Development of the Three-Dimensional Variational Analysis – Development and real-time testing will continue on the 3-DVAR analysis with emphasis on its initial implementation with the 20-km RUC.
Testing of the Nonhydrostatic Generalized Vertical Coordinate Model – Testing of this model will continue, in collaboration with the University of Miami and NCAR, as part of FSL's contribution to the overall multiagency development of the Weather Research and Forecast model.
Other ActivitiesThe following activities are also planned during Fiscal Year 1999.
Observation Sensitivity Experiments Using RUC for NAOS – Evaluate data sensitivity tests for GPS and satellite data precipitable water and similar tests for examining different possible configurations for the North American Atmospheric Observing System.
Verification of Wind Fields for Terminal Airspaces – Continue collaboration with NASA-Ames in assessing wind forecast accuracy and developing algorithms for estimating forecast error.
Participation in Development of the Weather Research and Forecast (WRF) Model – Continue collaboration with the GEWEX/GCIP program, including improvement of land surface processes in MAPS, particularly with new high-resolution land-use and soil datasets.
Development of the Well-Posed Model – The multiscale model incorporates topography via the generalized Kasahara transformation of the vertical coordinate, and the corresponding balancing terms currently are being added to the initialization package. Once topographic balancing is included, atmospheric BDI theory will be complete and the multiscale model/BDI initialization package will be ready for hindcasting and forecasting experiments. Initially the BDI vertical velocity will be compared with the vertical velocity computed from the vorticity evolution method which uses the dense temporal data of the National Profiler Network. Because the BDI vertical velocity is essentially only dependent on the heating, any discrepancies between the two vertical velocities should lead to improvements in the physical parameterizations.