In collaboration with other government agencies (e.g., NCAR, NCEP, NESDIS) and universities (e.g., University of Miami, University of Oklahoma), RAP branch scientists develop improved data assimilation and modeling methods for use in the RUC. Techniques for assimilating new observational datasets are developed toward the goal of the best possible estimate of current atmospheric and surface conditions, as well as the best possible short-range forecast. The branch also interacts with other FSL staff in implementing optimal computing methods with RUC software, making the model as efficient as possible on modern computing platforms.

A second primary focus of the branch is the development, real-time implementation, and evaluation of a fully coupled atmospheric/air chemistry mesoscale model prediction system. An MM5-based, fully coupled system has been run in support of the 2002 NOAA Temperature and Air Quality (TAQ) Pilot Project in New England (and other previous experiments). An increasingly important focus of research involves regional air pollution studies.

The following improvements have been made to the RUC20 since its implementation at NCEP.

3D Variational Analysis – A 3D variational (3DVAR) analysis has been developed for the 20-km RUC 1-hour cycle and is now being tested at NCEP for an implementation in 2003. The RUC 3DVAR analysis is configured in a generalized vertical coordinate set, which in this case is the RUC hybrid isentropic-sigma coordinate. The use of this coordinate provides an adaptive analysis space, with sloping of background error covariances in the proximity of baroclinic systems, and a stability-dependence for vertical error covariances. The RUC 3DVAR analysis has been shown to give equal or improved forecast skill to those from the current optimal interpolation analysis for forecast projections from 3 – 12 hours. The RUC 3DVAR analysis also assimilates the following new observational datasets:

• GPS ground-based precipitable water values (now over 180 in the United States)
• 915 MHz boundary-layer profilers (about 25 in the RUC domain)
• RASS temperature low-level virtual temperature profiles from selected 405-MHz and 915 MHz profilers
• Mesonet surface observations

Assimilation of Radar Reflectivity – An initial technique has been developed for assimilation of radar reflectivity and lightning data into the hydrometeor and water vapor fields in the RUC analysis to improve short-range precipitation and cloud forecasts. This technique can modify both rain and snow mixing ratios (Figure 19), accounting for the observed reflectivity in the column. National 2-km resolution reflectivity mosaic fields and 10-km Radar Coded Message fields are being inserted into the RUC20, since the former provides higher resolution but the latter provides beam blockage information to help differentiate "no echo" from "no coverage." Lightning data are used as an indication of convection. Verification of precipitation forecasts made in a parallel cycle using this technique have shown a significant improvement in skill.

Improved Version of the Grell-Devenyi Convective Parameterization – The RUC20 includes improved convective (subgrid-scale) precipitation from an ensemble closure/feedback convective parameterization by Grell and Devenyi. Additional research is being carried out to optimize the closure weighting and even to develop data assimilation techniques for probabilistic convective forecasts. An example of the difference in convective forecasts from the RUC is shown in Figure 20.

Support of the Operational RUC at NCEP – FSL monitors performance of the RUC running operationally at NCEP and works with NCEP to make necessary modifications. As part of this work, FSL must maintain expertise on the IBM SP computing system at NCEP and maintain a close, long-term collaboration with many groups in NCEP.

The RUC/MAPS system was the first regional model to cycle its multilevel soil moisture and soil temperature fields, and continues to be the only one to cycle snow water equivalent depth and multilevel snow temperature. This cycling of snow cover has proven beneficial in short-range surface forecasts in which there is change of snow cover over a 24-hour period, as shown in Figure 21, with snow melted over Wisconsin. In this case, the RUC was able to produce more accurate surface temperature forecasts than other models due to its cycling of snow variable in its 1-hour cycle. Paradoxically, although the RUC is used primarily as very short-range forecast guidance, the cycling of these surface fields requires considerable robustness from the land-surface model for what is, in effect, a simulation spanning months to years. In turn, the cycling of soil moisture and snow water equivalent in the RUC leads to improved short-range forecasts. The RUC research for the GEWEX/GAPP project is aimed at using cycling of surface variables combined with data assimilation of radar reflectivity and other cloud/precipitation observations to provide improved land-surface fields which can result in improved seasonal and interannual climate forecasts.

Use of RUC Wind Forecasts for Estimated Wind Power Potential – FSL continues a collaborative wind energy study with the National Renewable Energy Laboratory (NREL, Department of Energy) now using 20-km RUC/MAPS forecasts. Time-lagged ensembles produced from 20-km RUC forecasts out to 48 hours are used to estimate near-surface wind power potential, while variance among forecast ensemble members provides a measure of uncertainty in those forecasts. The high vertical resolution in the RUC near the surface and frequent update cycling makes it well suited to wind energy forecasting.

Special High-Resolution RUC Forecasts for NOAA Experiments (PACJET, IHOP, and TAQ) – FSL ran a special high-resolution version of the RUC model for three different special experiments over the last year (Figure 22). First, FSL distributed forecast fields to t he NWS Western Region Headquarters for real-time AWIPS display at local offices in support of the PACJET 2002 (Pacific Landfalling Jets) experiment. The special PACRUC configuration consisted of a 10-km grid covering all of the NWS Western Region, nested within the CONUS 20-km RUC domain. The 20-km RUC utilized a 1-hour assimilation cycle to ingest all conventional observations. PACRUC 24-hour, 10-km forecasts were produced every 6 hours, and AWIPS files containing selected surface fields were transferred to NWS via automated LDM scripts. Additional RUC fields were provided to the forecasters through a Webpage, and a Web-based forecast evaluation form was used to obtain forecaster feedback. Special 10-km RUC forecasts were also produced for the May – June 2002 Central Plains International H₂O Project (IHOP-2002) to the NWS Storm Prediction Center in Norman, Oklahoma. Finally, the 10-km RUC was moved to a domain covering the northeastern United States in support of the Temperature and Air Quality (TAQ) experiment in summer 2002. Again, AWIPS-compatible NetCDF files were provided to the NWS, this time through the Eastern and Central Regions. The 10-km RUC TAQ runs were continued through the fall and into winter 2002 – 2003. An example of a lake effect snow band forecast from the 10-km TAQ RUC is shown in Figure 23.

Profiler Impact Experiments with RUC – An assessment of the value of data from the NOAA Profiler Network (NPN) on weather forecasting has been completed. A series of experiments were conducted using the RUC20 model in which various data sources were denied to assess the relative importance of the profiler data for short-range wind forecasts. Average verification statistics from a 13-day test period indicate that the profiler data have a positive impact on short-range (3 – 6 hour) forecasts over a central United States subdomain that includes most of the profiler sites as well as immediately downwind of the profiler observations (Figure 24). Averaged over time of day, the profiler data most strongly reduce the overall vector error in the troposphere below 300 hPa where there are relatively few automated aircraft observations. At night when fewer commercial aircraft are flying, profiler data also contribute strongly to more accurate 3-hour forecasts at jet levels. For the test period, the profiler data contributed 20–30% (at 70 hPa) of the overall reduction of 3-hour wind forecast error by all data sources combined.

Several case studies were examined in detail to illustrate the value of the profiler observations for improving weather forecasts. One of the case studies indicated that inclusion of profiler data in the RUC model runs for the 3 May 1999 Oklahoma tornado outbreak improved model guidance of convective available potential energy (CAPE), 850–300 hPa wind shear 0–3 km helicity, and precipitation in southwestern Oklahoma prior to the outbreak of the severe weather. In another case study, inclusion of profiler data improved RUC precipitation forecasts associated with a severe snow and ice storm that occurred over Kansas. Summaries of NWS forecaster use of profiler data in daily operations support the results from these case studies and the statistical forecast model impact study that profiler data contribute significantly to improved forecasts over the central United States, where these observations currently exist. The results of this profiler impact study have been submitted for publication.

Observation Sensitivity Experiments Using RUC to Examine the Impact of GPS Precipitable Water Observations – In collaboration with the Demonstration Division, the RAP Branch continued 60-km RUC parallel cycle experiments with and without assimilation of GPS precipitable water observations. Positive impact (leading to more accurate forecasts) of GPS precipitable water observations on short-range relative humidity forecasts seen in earlier sensitivity experiments has continued to increase as more stations have been added to the network. Results from a new set of 2-week retrospective experiments for both cold and warm season periods with the 20-km version of the RUC show still greater impact from the addition of GPS precipitable water observations than with the 60-km RUC model.

Regional forecast improvements due to the lidar observations can come from three sources: 1) direct assimilation of the lidar observations on the regional domain 2) improved lateral boundary conditions due to the assimilation of lidar observations on the regional domain and 3) improved lateral boundary conditions due to the assimilation of lidar observations from the global model. As expected, results from this study indicate forecast improvement from all sources. Improvement from the direct assimilation of lidar observations is greatest at the analysis time, then slowly diminishes with forecast length, while improvements from the lateral boundary conditions increase with forecast length. Initial OSSE experiments focused on a best-case scenario, in which no lidar observations were lost due to cloud attenuation and no errors were added to the lidar observations. In these experiments, lidar wind observations produce modest short-range (0 – 12-hour) forecast improvements over the data-rich RUC domain. Midlevel wind forecasts benefit the most, with 6-hour forecast error reductions of up to 10% (Figure 26) from the assimilation of lidar observations. Other experiments which accounted for the loss of lidar observations due to clouds and the existence of random errors in the lidar observations, have shown only a slight reduction in the forecast improvement compared to the best-case experiments. As part of this project, the formulation of the RUC 3DVAR was modified to facilitate the direct assimilation of lidar line-of-site wind observations. This modification also allows for the future direct assimilation of Doppler radar radial velocity observations, in accordance with planned RUC analysis enhancements.

Air Quality Forecasts from a Coupled Weather/Air Chemistry Prediction Model – FSL is developing and applying a next-generation coupled weather/air quality numerical prediction system. This system is capable of forecasting ambient ozone concentrations over regional to urban scales and includes the MM5 meteorological model with "online" chemical capability. In this system the chemical kinetic mechansm is embedded within the meteorological model structure, and thus the integration of the chemistry is performed as part of MM5 (MM5/Chem). Biogenic emissions are also integrated online since they are strongly modulated by meteorology. A second coupled modeling system, led by FSL, is based on the Weather Research and Forecasting (WRF) numerical model and uses the same chemical modules as MM5/Chem.

During the summer of 2002, MM5/chem was run in real time at FSL in support of the New England Temperature and Air Quality (TAQ) pilot experiment. The domains for which forecasts were produced were shown in Figure 16, and Table 2 summarizes each of the runs. In addition, during 5 July – 30 August 2002, data from the model forecasts were provided in real time to several OAR laboratories via the Web. NOAA/ETL received data from the numerical forecasts for verification with meteorological observations at profiler locations, and with surface observations at special high energy usage sites. (Figure 27 shows an example of a meteorological comparison that was displayed in real time.) Scientists at the National Severe Storms Laboratory (NSSL) received meteorological data on a common three-dimensional grid for ensemble predictions. Special three-dimensional chemical and meteorological datasets were also provided to the NOAA Aeronomy Laboratory (AL) for verification as well as for real-time forecasting aboard the Ron Brown. An example of a comparison of ozone forecasts with measurements from the Ron Brown as seen in real time on 24 July is shown in Figure 28. Note that for this particular case the model did exceptionally well in predicting the ozone concentrations, especially for the highest resolution forecasts. Finally, the Air Resources Laboratory (ARL) received surface chemistry and meteorology datasets for verification. The completion rate of the model forecast run on the Jet supercomputer at FSL was greater than 95%.

In addition, data from July and August 2002 have since been rerun to create a complete testbed dataset. A statistical comparison with observations over this time period was performed by AL and is now available as a baseline dataset to compare against. (See Figure 29 for an example of a comparison of ozone forecasts and observations at one particular station.) Development of this type of testbed dataset is a powerful tool in model development to aid the assesment and understanding of currently available atmospheric numerical modeling systems, and point to areas where further development is needed as we work toward production of reliable air quality forecasts.

In addition, FSL is taking the lead in developing the next-generation air quality forecast model, WRF/chem. The first version of this model already exists and includes all chemical modules that are a part of MM5/Chem (grid-scale and subgrid-scale transport, biogenic emissions, deposition, photolysis, and the RADM2 chemical mechanism). Evaluation of this model is currently in progress, and it will probably be run in real time during the summer of 2003.

Participation in Development of the Weather Research and Forecast (WRF) Model System – 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 and associated system are being developed as a collaborative effort among NCAR, NCEP, FSL, the Center for the Analysis and Prediction of Storms (CAPS), and other research institutions together with the participation of a number of university scientists.

In collaboration with NCAR, FSL has worked on the development of physics packages and a three-dimensional (3DVAR) analysis system. FSL has contributed two physics components to the WRF forecast model — an alternative land-surface model based on the RUC land-surface model and the Grell-Devenyi ensemble convective parameterization. Both of these schemes have been fully implemented and tested in the WRF model. FSL has also developed the standard initialization package for the WRF model, and has worked with the University of Miami on development of a quasi-isentropic variant of the WRF nonhydrostatic model.

The WRF model is being tested with RUC initial conditions for two different domains, the 20-km CONUS RUC domain; and the 10-km TAQ New England domain. The WRF standard initialization (SI) procedure was modified to fully use RUC native-coordinate initial conditions, including hydrometeor and land-surface fields. In addition, a post-processor was developed from the WRF model to produce RUC-like GRIB output files, facilitating comparisons with RUC model forecasts. FSL will adapt WRF assimilation and model systems over the next several years to include an advanced rapid update capability for operational implementation at NCEP. It is planned that the WRF model will supplant the current RUC forecast model in the Rapid Update Cycle by 2006.

Implementation of Three-Dimensional Variational (3DVAR) Analysis in the 20-kilometer RUC – With tuning to ensure sufficient accuracy for short-range wind forecasts, the RUC 3DVAR analysis will be added to the 20-km RUC running at NCEP. The 3DVAR implementation will provide smoother analyses, more accurate forecasts, and a framework for assimilation of radial wind observations from radar in the future. Development and real-time testing will continue on the 3DVAR analysis, both for incorporation into the RUC and toward the development of the 3DVAR analysis for the WRF model.

Continued Development of a National-Scale Cloud/Hydrometeor Analysis – Development and real-time testing will continue for further improvements to the RUC national-scale cloud analysis, with the addition of radar, lightning, and surface observations to satellite cloud-top data. Experiments will be carried out testing assimilation of a GOES imager-based multilevel cloud product, as described below under the JCSDA plans.

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 RUC model uses some of the MM5 parameterizations, which will be options for the WRF model.

National Observing System – FSL will continue its efforts with a team working toward an initiative to develop a national mesoscale observing system consisting of tropospheric and boundary layer profilers, ground GPS receivers, and radiosondes with ground tracking systems. This is an initiative with great potential impact for mesoscale forecasting. Also under development is an observation system simulation experiment (OSSE) with a practical observation network design and numerical model to verify the budgets and applicability.

Data Assimilation – Work will commence to test the WRF 3DVAR system within the RUC assimilation cycle.

High-Resolution Experiments Using RUC for the New England Temperature Air Quality Experiments – RUC forecasts will continue to be made at 10-km resolution in support of this experiment. In addition, the WRF model will be run over the same domain, also at 10-km resolution, initialized from the RUC. This will allow initial intercomparisons between the current RUC hydrostatic model and the WRF nonhydrostatic model including RUC physical parameterizations, the RUC land-surface model, and the Grell/Devenyi ensemble-closure cumulus parameterization. Objective verification of the model forecasts will be performed as part of these studies.

Participation in GEWEX – Collaboration will continue on the GEWEX/GAPP program, with the focus on development of a coupled atmospheric/land-surface assimilation system that uses an optimal combination of radar and satellite observations to modify clouds and precipitation along with model forecasts in regions where observations are unavailable.

Contribution of RUC Forecasts to the NCEP Short-Range Ensemble Forecast System – As part of an expansion to NCEP’s Short-Range Ensemble Forecast (SREF) project, FSL will complete its effort begun this last year to set up an ensemble version of the RUC model running out to 63 hours on a 48-km grid over the Eta domain. The RUC SREF is spawned from a set of five members bred from the NCEP Eta model, and is a candidate for inclusion in the NCEP SREF set currently composed only of Eta and Regional Spectral Model bred members. Tentative results from the RUC SREF show substantial spread in the ensemble, but it needs to be determined whether the forecast skill for the ensemble mean exceeds that of the operational RUC20 model, and statistical techniques must be developed to assess the results.

Joint Center for Satellite Data Assimilation Activities – Future work will first involve running and testing the Optical Test Transmittance (OPTRAN) radiative transfer model to replace the European Centre for Medium-Range Weather Forecasts' RTTOV code that has been used in all RUC forward model calculations thus far. OPTRAN has been chosen as the community radiative transfer model by the Joint Center for Satellite Data Assimilation. Outgoing radiances from the RUC will then be subjected to OPTRAN forward model calculations to compare with the GOES radiances. The imager data will be used to determine clear-air radiances with greater resolution than using the sounder estimates. Eventually, the goal is to incorporate the adjoint of the forward calculations into the RUC three-dimensional variational (3DVAR) analysis and to begin using this to rapidly update the radiance data in the RUC and, later, the WRF models.

An ongoing data denial experiment provides insight into the impacts of the various data sources on the analysis. In addition, this assessment tool can be used to gauge the strength and weakness of the different data sources, in order to optimize their respective weights in the variational equation. The statistics used in this study are a comparison of analysis output to radiosonde data (taken as referenced truth). This is possible since radiosonde data are not typically used in the operational LAPS system due to their latency (poor timeliness). The goal of the moisture variational application is to provide a complete product that describes the atmospheric water distribution from vapor to cloud droplets to precipitation, both liquid and frozen. This analysis has been used to improve model initialization. This analysis utilizes all conventional data, along with satellite, radar, and GPS data. The routine is based on the LAPS cloud analysis, but then seeks to quantify all water substance. Variational methods are used to impose dynamic balance and continuity on the first-stage analyzed fields to accommodate the "Hot Start" analysis described below.

LAPS Advanced Quality Control – Quality control of observations is a continuing focus of LAPS analysis development. A Kalman filtering scheme is used to improve the quality and timeliness of surface observations. The method allows users to optimally exploit local model output and past station trends and buddy trends to produce check values for surface stations. In conjunction with LAPS support of a three-dimensional 30-minute analysis cycle, the Kalman scheme allows the merging of mesonetworks with varying cycle times. Working exclusively in data space, the Kalman filter scheme is economical for use in the local computing environment and provides a continuously updated and accurate set of observations where all stations appear at each cycle. This is an appropriate approach for instances when a user requires good product time continuity, but has high variability in observation count from cycle to cycle. Since the Kalman scheme still requires more computer storage than is currently available in the local weather office, it has not been widely used.

LAPS "Hot Start" Procedure – The LAP Branch continues to improve the Hot Start procedure for diabatic initialization of mesoscale models. The Hot Start initialization scheme is designed to develop initial conditions for mesoscale models such as the MM5, RAMS, and the mass-coordinate version of the (WRF) model. This scheme is unique in that it runs on small PC clusters with Linux operating systems and is ideal for applications in local weather offices where accurate short-term cloud and precipitation forecasts are needed. This system depends greatly on the accuracy of the background modeling system, currently the NCEP versions of the RUC and Eta models. The Hot Start scheme uses estimates of vertical motion and cloud water and ice mixing ratios from the LAPS cloud analysis. A variational analysis that applies both mass continuity and mass-momentum balance makes small adjustments to the wind and temperature field to accommodate and sustain the clouds in the first few time steps of the model integration. The cloud retrieval algorithm includes a broad range of microphysical species, cloud-type dependent estimates of cloud vertical motion, and saturation of the cloud environment.

Verification with MM5 during the 6-week International H2O Project showed that the Hot Start outperforms other initialization techniques in the 0 – 6-hour time frame in forecasts of precipitation, most state variables, and 3-D cloud and radar fields. Figure 31 shows equitable threat scores for precipitation for the entire experiment. Note that the bias for the LAPS Hot Start 3-hour forecasts is nearly 1.0 across all precipitation categories (i.e. it is nearly unbiased), whereas the other models characteristically overforecast very light precipitation amounts and display a large dry bias for precipitation amounts >0.25 inch. The Hot Start method is used to initialize MM5 over a variety of domains including the local Denver forecast area, where forecasters use it as an operational tool; the U.S. Air Force Western Range at Vandenberg Air Force Base as part of the Range Standardization and Automation (RSA) implementation discussed below, and IHOP field operational areas at 12- and 4-km grid resolution. The Hot Start system has been coupled to the WRF-mass coordinate model for testing in a local weather service office in conjunction with the Coastal Storms Initiative carried out with the National Weather Service and National Ocean Service. It is also under evaluation as part of the MM5 modeling system at the Central Weather Bureau of Taiwan.

GOES Improved Measurements and Product Assurance Plan – The GOES Improved Measurements and Product Assurance Plan (GIMPAP) project has been a key part of the LAPS moisture algorithm development for integrating the high spatial structure of GOES imagery and sounder data into the LAPS system. GIMPAP includes NESDIS cloud-top pressure and layer-precipitable water products.

The total precipitable water analysis over the IHOP domain as generated by the LAPS analysis system incorporated real-time NESDIS product data with greatly reduced latency (a 30-minute cycle instead of the customary 60-minute cycle) and experimental GOES-11 single field-of-view (unsmoothed) radiance data. It also used indirect NESDIS cloud-top product data via the LAPS cloud analysis. The IHOP product stream was accurate and timely throughout the 6-week field project.

The Range Standardization and Automation (RSA) Project – Several years ago, the Air Force initiated the Range Standardization and Automation (RSA) program to modernize and standardize the command and control infrastructure of the two U.S. Space Launch facilities (ranges), located at Vandenberg Air Force Base, California, and Cape Canaveral Air Station, Florida. During this past year in cooperation with Lockheed Martin Mission Systems, an integrated local data assimilation and forecasting system was installed at both ranges. The RSA system runs on Linux "Beowulf" clusters from IBM at each range and a test cluster at FSL for use in system development. The clusters consist of 8 dual-processor Pentium III nodes and 1 dual-processor front-end node, totaling 18 processors. A Myrinet interconnect is used for high-speed message passing between nodes.

The first version of the RSA Data Assimilation and Forecast System, based on LAPS coupled with the NCAR fifth-generation Mesoscale Model (MM5), is in testing mode. The system produces hourly LAPS analyses and a new MM5 forecast run every 6 hours on a triple-nested domain with 10-km, 3.3-km, and 1.1 km grid spacing, respectively. These analyses make use of the AWIPS Local Data Acquisition and Dissemination (LDAD) interface to incorporate data sources unique to the launch facilities in addition to the radar, satellite, and other datasets available via the AWIPS data feed. Every 6 hours, these analyses are used to perform a diabatic initialization of an MM5 forecast run. The forecast model outputs hourly forecast fields out to 14, 12, and 9 hours for the 10-km, 3.3-km, and 1.1-km grids, respectively, using 2-way nested feedback. The entire system is integrated with the Linux version of AWIPS installed at the Air Force ranges. Figure 33 shows an example of a product on the interior (1-km mesh) of forecast surface wind and temperature, verifying surface plots, and comparative wind forecasts from the MesoEta model over the local area surrounding Vandenberg AFB.

The RSA and projects spurred numerous improvements to the LAPS/MM5 prediction system. First, the cloud analysis was adapted to ingest and use narrowband radar reflectivity from multiple WSR-88D sites, available via the Satellite Broadcast Network (SBN) feed into AWIPS. This provides better coverage over the entire domain of the grid used for the RSA, which is spatially larger than a typical NWS Forecast Office domain. Second, the cloud analysis was modified to use a climatological albedo field to increase the utility of the visible satellite imagery. This, combined with the recent integration of the GOES 3.9 micron channel, has greatly improved the system's ability to detect low stratus clouds over the ocean surrounding the Western Range. Third, for the purposes of initializing a numerical weather prediction model, the final three-dimensional concentrations of the hydrometeor species are scaled as a function of horizontal grid spacing. Finally, to ensure compatibility with current numerical weather prediction model microphysics schemes, any grid box volume containing cloud liquid or ice is raised to its saturation level with respect to the phase of the cloud species. This prevents rapid evaporation of the cloud and the concomitant spurious generation of cool downdrafts within the first few time steps of model integration.

These RSA capabilities together represent the first operational installation of a local modeling system completely integrated with AWIPS in a WFO-like environment, the first operational installation of the LAPS diabatic initialization-Hot Start method, and the first operational use of a Linux-based AWIPS system. Ongoing FSL work includes ingesting and optimizing the use of all local meteorological datasets, incorporating new capabilities such as online verification of the forecast grids, enhancing utilization of the satellite data to improve cloud analyses, and improving the LAPS diababic initialization method.

High Performance Computing – The FSL High-Performance Computing System (HPCS) has been a critical resource for all of the numerical modeling activity in the branch, including the unique mesoscale model ensemble used for the Federal Highways Project described below. This experience continues to provide important feedback to the system developers and computer specialists regarding configuration issues and future upgrade plans.

Branch staff visited with Central Weather Bureau forecasters and researchers in Taipei to conduct training on the potential use of LAPS in operational forecasting and discuss ideas for nowcasting. A Webpage...

GAINS PIII Flight – The maiden flight of the 60-ft diameter GAINS Prototype III (PIII) balloon occurred on 21 June 2002. This flight met several development objectives, including launching the PIII balloon, floating it at altitude for more than eight hours, transforming the balloon envelope into a deceleration device, achieving a safe descent rate, tracking the balloon from an aircraft; forecasting balloon trajectory before launch, updating balloon landing position during flight, and recovering the balloon and payload.

Manufactured by GSSL, Inc., of Hillsboro, Oregon, and launched from their Small Balloon Facility at Tillamook, Oregon, the 500-pound balloon carried a 325-pound payload containing packages from four organizations. The FSL payload included GPS for locating the system, two independent radio- and software-controlled termination methods, and environmental sensors to monitor balloon performance. The payload also contained a GPS reflection experiment from NASA/Langley, redundant locating capability based on a design adapted from the Edge of Space Science (EOSS) of Denver, Colorado, and backup location and termination units from the Physical Science Laboratory of New Mexico State University and from GSSL.

Nominal float altitude of 54,000 ft was achieved after reaching a maximum altitude of 57,000 ft. A tracking aircraft kept the balloon within radio line of sight at all times during the 225-mile flight, and personnel on board coordinated the flight with FAA Air Traffic Control. Two additional vehicles tracked the balloon from the ground and recovered the payload. The balloon’s descent was slowed by the GSSL BERSTM (Balloon Envelope Recovery System), in which the balloon envelope transformed to a parachute and the system descended at about 500 ft per minute. The soft landing caused no damage to the payload capsule, and minimal damage to the wheat field south of The Dalles, Oregon, where it landed. The entire system was removed from the landing site and returned to Tillamook within 24 hours of landing. Figure 36 shows the actual flight path and the path predicted using winds from the global AVN numerical model.

GAINS Pump Test Flight – On 17 August 2002 a flight was launched from Meadow Lake Airport, northeast of Colorado Springs in a cooperative effort between the GAINS and EOSS groups. The main objective of the flight was to test the operation of the turbine developed by Advanced Engineering at GAINS operational altitudes. Results from laboratory experiments indicated that the turbine met the requirements for flow rate, power needs, and weight; this experimental flight was intended to affirm these results under true atmospheric conditions. Turbine electronics and control software algorithms were also tested. A 19,000 cubic-ft polyethylene balloon was used for lift, and a 500-gram latex balloon in a GAINS P-I 8-ft nylon shell used as a ballast balloon. The plan was for the turbine to pump ambient air into the ballast balloon until the desired super pressure was achieved. For this experiment, the turbine was to be tested at 50,000, 60,000, and 70,000 ft, with inflation continuing until a super pressure of ambient plus 15% was achieved. The payload, consisting of all control and communication electronics, was located in an insulated ring that encircled the turbine on a lightweight disc with three supporting legs, known as the "lander" (Figure 37).

The gusty nature of the surface winds prior to launch made for complications, the most significant being the introduction of small holes in the polyethylene lift balloon from contact with the ground. Most, but not all, of these were repaired prior to launch, resulting in a flight duration shorter than expected. The balloon reached a maximum altitude of 32,000 ft before descending back to earth. Unfortunately, this altitude was not sufficient to allow testing of the GAINS pump. Although the pump performance could not be tested, the setup and launch procedures as well as tracking and recovery were successful. This will enhance the ability to achieve the goals when the test is repeated later.

The RUC20 model was run for this SCATCAT case, representing the first time that the RUC had ever been positioned to run entirely over the Pacific Ocean. The AVN model was also used for the first time instead of the Eta model for specification of the RUC boundary conditions. Figure 39 shows a cross section of isentropes and isopleths of isentropic potential vorticity taken perpendicular to the jet core but over a longer length than the dropsonde cross section in Figure 38. These model predictions were compared to fields of winds, potential temperature, and ozone (potential vorticity is taken as a surrogate for ozone) measured by the aircraft. This comparison revealed very similar features, including the warm frontal zone, the upper-tropospheric front/jet system zone, and regions of strong vertical wind shear and associated large DTF within these zones. A deep tropopause fold is present along the warm frontal zone down to almost 700 hPa, but of greater interest are multiple tropopause "undulations" in the upper-level front of the model. Also of interest are mesoscale gravity waves in the lower stratosphere directly above the jet core, with vertically varying horizontal wavelengths of ~100 – 160 km. Similar waves in the dropsonde analyses in the lower stratosphere directly above the jet core displayed wavelengths of ~80 km. A higher-resolution version of the RUC, of course, might have produced shorter wavelength features.

Wild fluctuations in ozone measurements from the NOAA/ARL experimental instrument were measured at the 41,000-ft level, but fluctuations in the potential temperature and wind in-flight observations did not correlate highly with the ozone data, nor was much turbulence reported on this flight leg. It was concluded that these rapid fluctuations in ozone at this level represented "fossil turbulence" or remnants of earlier stratosphere-troposphere turbulent exchange processes. By contrast, the correlation between the ozone and in-flight variables, as well as with the RUC model potential vorticity variations, was very high at the 33,000-ft altitude. Also, moderate turbulence was reported at this flight altitude, as the G-IV was penetrating a rather pronounced gravity wave within the upper-tropospheric frontal zone. Thus, active turbulence was occurring in association with gravity wave activity at this altitude, but not at the higher level.

Time series and spectral analyses were performed for each of these flight legs in an attempt to relate the appearance of turbulence to mesoscale gravity wave activity. Potential temperature and longitudinal wind exhibited a high degree of cross correlation, as did the potential temperature and ozone data at the 33,000-ft level. Such a strong "in-phase covariance" is expected of either deep propagating gravity waves or, more likely, decaying (evanescent) waves.

The SCATCAT research revealed that MOG turbulence occurred in conjunction with gravity waves shed within an upper-level fractured front on the cyclonic shear side of the jet core. Besides this major finding, it was concluded that RUC-forecast DTF is a useful diagnostic of turbulence, whereas ozone is not, and that both the MOG turbulence and the high DTF regions occurred in the vicinity of the strongest mesoscale gravity wave activity in both the model forecasts and the aircraft observations.

Diagnostic Algorithm Development – The forecast skill of Integrated Turbulence Forecasting Algorithm (ITFA) and its component algorithms has been evaluated both objectively and by forecasters. These studies show that the best of the algorithms display similar probability of detection (POD) curves, and that there is considerable room for improvement. Research conducted at FSL indicates that these algorithms also typically predict patterns that are similar to one another, and that MOG pilot reports (PIREPs) of turbulence often fall in the margins of the predicted ITFA regions. The best of these algorithms are fundamentally based on the destabilizing dynamics of vertical wind shear.

FSL developed an experimental turbulence prediction scheme based on a radically different dynamical concept, namely that turbulence is generated as mesoscale gravity waves are shed when an unbalanced jet streak propagates toward an inflection axis in the upper-level height field (the SCATCAT analyses discussed above lend additional support to this contention). Diagnosed gravity waves and model flow imbalance have been shown in detailed case studies by Forecast Research Division scientists to relate strongly, not only to each other, but also to MOG turbulence reports. The flow is considered to be unbalanced when there is a pronounced residual in the computed sum of the terms in the nonlinear balance equation from the RUC model. Imbalance typically occurs in essentially the same region as where mesoscale gravity waves develop and upstream of where the turbulence is reported, as demonstrated in one case shown in Figure 40. Note in this example that the imbalance is occurring precisely at the tip of the dry air stream associated with subsidence within a pronounced jet streak (or "potential vorticity streamer," discussed below). The conventional turbulence diagnostic DTF3, which is a major contributor to the ITFA, fails to predict the swath of turbulence reports in the Ohio River Valley region, whereas the imbalance indicator field successfully captures this event. On the other hand, DTF3 does a credible job at delineating the other swath of turbulence reports in the Great Lakes region, not captured by the imbalance field. This complementary nature of the imbalance indicator fields and DTF/ITFA is typically observed to occur.

A Webpage was created this past year to examine the relationships between diagnosed flow imbalance from the RUC20 model and MOG turbulence reports on a daily basis. This more thorough investigation has shown that mountain waves generated by strong flow over rough terrain like the Rocky Mountains are even more highly correlated with turbulence and flow imbalance than are the imbalances associated with the jet stream and cyclonic storm systems. While mesoscale models like RUC can be useful for diagnosing the flow imbalance regions and where generally gravity waves are likely to form, they do not reliably predict the details of the gravity waves themselves, such as their wavelength, phase speeds, and so forth. The new predictive scheme being developed at FSL not only produces patterns systematically different from the current ITFA algorithms but also predicts turbulence regions missed by those methods. Further refinement of forecast turbulence regions might be obtained by adding the requirement that an efficient wave duct must be present downstream of the region of diagnosed flow imbalance to retard the vertical leakage of wave energy, thus allowing coherent waves to persist. The optimum amount of smoothing of the imbalance fields, the proper thresholds, and other numerical issues must all be resolved, before the new imbalance indicator field can be incorporated into ITFA to increase its utility.

The IHOP project offered a unique opportunity to carry out two aircraft missions (led by the Forecast Research Division Chief) to observe the morning LLJ over Oklahoma and Kansas. Each mission utilized airborne dropsonde data, Differential Absorption Lidar (DIAL) data flown on the German Falcon, High-Resolution Doppler Lidar (HRDL) data from NOAA/ETL also flown on the Falcon, and in one of the cases, hyperspectral radiometric data from the NASA Proteus aircraft, to observe a strong LLJ in good atmospheric conditions (i.e., substantially free of clouds). These observations offer an excellent opportunity to prepare detailed three-dimensional meteorological fields of moisture and winds at a multitude of scales and the possibility to compute a moisture budget. The objective is to examine these data to determine the impact of fine-scale moisture observations on the numerical prediction of precipitation. Another ongoing task is combining datasets obtained from the two aircraft missions to compute moisture budgets and perform diagnostic and numerical modeling studies of these cases to test the hypothesis that warm-season QPF skill can be significantly improved by better characterization of the transport of water vapor by the LLJ.

Structure and Dynamics of Gravity Currents and Undular Bores – The IHOP field phase collected a surprisingly large number of events in which either a thunderstorm outflow boundary or cold front, acting as an atmospheric gravity current, intruded into a stably-stratified boundary layer and generated an undular bore (a kind of hydraulic jump) on the top of the inversion. In some cases, deep convection appeared to have been generated by the vertical motions attending this phenomenon, which are quite strong (updrafts of several meters per second magnitude). An unprecedented number of ground-based and airborne remote sensing systems observed the passage and evolution of bores in IHOP, including FM-CW radar, the NCAR Multiple Antenna Profiler (MAPR), Raman lidar, the NASA GLOW and HARLIE aerosol backscatter lidars, refractivity fields obtained from the NCAR S-POL radar, an Atmospheric Emitted Radiance Interferometer (AERI) system, the French Leandre-II DIAL system aboard the NRL P-3 aircraft, and the University of Wyoming King Air aircraft. FSL is collaborating with a team of international scientists to analyze these data. Also, very high-resolution numerical simulations of two bore events are underway to better understand the origin, dynamics, entrainment mechanisms, and influence on convection initiation by undular bores. Much more will be reported on these studies in the next issue of FSL in Review.

Potential Vorticity Streamers – Researchers in Europe and the United States have noted the frequent occurrence on water vapor satellite imagery (GOES and METEOSAT) of pronounced dark filaments, which are mesoscale in width and varying in length up to the largest scales of atmospheric motion (see Figure 41). The most pronounced of these dry filaments have a parallel jet stream immediately to the south, and according to recent studies conducted at FSL, a similarly shaped band of collocated, enhanced potential vorticity (a "PV streamer"). Another satellite-observed feature — enhanced ozone — suggests that a PV streamer is a downward intrusion of stratospheric air along an upper-level front.

Monitoring PV streamers has prognostic value. In Europe, PV streamers have been identified as precursors to flooding along the Mediterranean slopes of the Alpine piedmont. In the United States a number of case studies have documented MCS development near a preexisting PV streamer. Some of these MCSs were accompanied by severe weather and flash flooding, as was the case with one occurring 28 June 1999 over Kansas (described in the 2002 FSL In Review). This storm was one in a series of four MCSs that occurred along a PV streamer that persisted six days over the central and southeastern United States. The total precipitation from these four MCSs produced a significant fraction of the area’s annual precipitation.

In addition to being a valuable forecasting ingredient to organized convection and heavy precipitation, the PV streamer is a fascinating phenomenon, rich in dynamical and thermodynamic implications deserving of serious scientific study. It combines upper-level jet stream dynamics with the flow of midtropospheric, dry, potentially cold air having a low static stability. Some researchers have suggested that the exit region of an upper-level jet streak enhances a transverse low-level jet as a result of mass balance. Others have suggested that upper-level PV passing over a low-level front can organize a developing cyclonic circulation. With the addition of moisture to the low-level jet, the mechanisms for producing organized convection are all present. The details of how the ingredients combine to shape an MCS await results of field experiments (such as BAMEX) and those of mesoscale modeling at FSL.

NCEP Gauge Quality Control Project – A system for the automated screening of hourly gage precipitation observations has been designed to find failing gauges in the Hydrometeorological Automated Data System (HADS). Funded by the USWRP, this system was solicited by National Centers for Environmental Prediction (NCEP) to provide more timely gauge quality indicators to prevent the use of faulty gauges in analyses used to initialize model runs. Following the identification of several characteristic failure types (e.g., gauges that jam on or off for long periods), tests for these failures on individual gauges have been devised. Several of these tests are based on the most recent 30-day distributions of precipitation characteristics such as daily and hourly rainfall frequency. Gauges that fail these tests can then be entered on a reject list to eliminate them from precipitation analyses and model verification. Because old or inaccurate station metadata have often been a weak link, procedures to automatically update relevant station lists have been introduced. An automated procedure for gauge quality control and rejection will be completed during 2003 and delivered to NCEP for implementation.

Assessing the Quality of Real-Time Precipitation Gauge Observations – An ongoing collaboration with the Real-Time Verification System project group seeks to improve verification procedures for model-based precipitation forecasts. Over the past year, this verification effort concentrated on the use of hourly point precipitation observations at sites to which model fields were interpolated. Of major concern was the determination of the quality of hourly observations in order to screen out unsatisfactory observing sites and to reincorporate sites with good observations that in the past have not been included in the set of stations presented in near real time by River Forecast Centers.

ACARS/AMDAR Quality Control System – A computer program to flag and in some cases correct weather data from automated sensors on commercial aircraft (called ACARS in the U.S. and AMDAR in the rest of the world) was upgraded. The quality control system uses temporal and spatial consistency checks along each flight track and altitude-adjusted climatological consistency checks to discover errors. It also interpolates locations and times for high-resolution observations taken during ascent and descent enabling these data to be used in numerical weather prediction model research and weather forecasting. The quality control system was upgraded to ingest and display data from the MDCRS (Meteorological Data Collection and Reporting System) data stream, produced by Aeronautical Radio, Inc. MDCRS data are largely the same as data provided by airlines directly to FSL, and decoded here, but have some differences. Because MDCRS feeds numerical weather prediction models run by NCEP, it is important to understand these differences. The QC system attempts to match each MDCRS observation with one decoded directly by FSL, and report the differences. The QC system was also upgraded to process experimental icing data from Delta Airlines, in support of a research effort led by the National Center for Atmospheric Research and funded by the FAA.

ACARS-RUC Intercomparison Database – A database has been developed that compares ACARS data with 1-hour RUC forecasts. RUC data are interpolated to the location of each ACARS observation. This database, with 9 months accumulated data, is used to develop error statistics for longitudinal and transverse winds, and for other RUC and aircraft error analyses.

North American Radiosonde Dataset – For years FSL has been providing a CD-ROM archive of quality-controlled radiosonde data for use by the research community. During 2002, updates were made to the global and North American station history files to reflect changes in the network sites, including moves, station identification changes, and addition of new stations. The Website was modified to correct problems in the generation of duplicate skew-T plots for different stations, and software was incorporated to handle the change to a new year without interrupting service.