The major accomplishment for RAP this year was the conduct of two major field programs, LANTEX and WISP94 .
The Research Applications Program (RAP) conducted a wide variety of activities during 1994 including a major winter storms field program (WISP94), a major field and modelling study of terrain-induced windshear and turbulence at the site of the new Hong Kong Airport (LANTEX), and the development of a new prototype weather display system. These activities reflect the emphasis of RAP on focussed research and its application. Primary sponsors of RAP activities during this year were the Federal Aviation Administration (FAA), the National Aeronautics and Space Administration (NASA) and the Hong Kong government. In the following, we present brief summaries of RAP activities during 1994 .
The Winter Icing and Storms Project (WISP) is a cooperative research effort designed to study the structure and evolution of winter storms. Its fourth field effort, WISP94, took place between 25 January and 25 March 1994 in northeastern Colorado.
The primary goals of WISP are: 1) to improve our understanding of processes involved in the production and depletion of supercooled liquid water (SLW) in winter storms and 2) to improve forecasts of aircraft icing. Previous field programs have clarified processes which contribute to liquid production but have shed little light on ice initiation -- nucleation and early growth. Understanding these processes is vital to completing our knowledge of supercooled liquid production and depletion in the atmosphere. WISPIT (WISP Instrument Test) was conducted in winter 1992-93 to test both concepts and models for ice nucleation and to test instrumentation for the measurement of small ice particles. Based on the success of that limited field program, WISP94 was planned to focus on ice initiation studies.
The initiation of ice crystals both in wave clouds over the Rocky Mountains and in upslope storms in eastern Colorado received highest priority during WISP94. Secondary priority was given to the studies of cold surges, the formation of large supercooled droplets and precipitation evolution.
The scientific goals of the ice initiation portion of WISP94 were:
The NCAR Electra was the primary airborne ice nucleus and aerosol sampling platform. The University of Wyoming King Air provided cloud physics and aircraft performance measurements. During the latter half of the project, the University of Massachussetts' W-band (3 mm) radar was installed on-board the King Air for detailed observations of cloud structure.
The MM5 model, with a cloud microphysics package developed as part of previous WISP research efforts, was run twice daily. Nested grids of 60, 20 and 6.7 km horizontal resolution were available. These runs were used in operations planning as well as in an assessment of overall model performance and the microphysical parameterization.
During WISP94 thirty-seven weather events were studied. Of these, fifteen events were wave clouds, twelve events were upslope clouds, and the remainder consisted of cold surges, snowbands, lee cyclones and other winter weather events of interest to the project. A comprehensive set of ice nucleus filters and bulk aerosol samples were collected in the air and at the surface and are being processed for comparison with ice crystal measurements to determine factors controlling ice formation. Large supercooled droplets were found in stratiform upslope clouds on at least eight flights, and flight patterns were flown to determine their association with shear layers near cloud top as well as other turbulence-generating mechanisms. The real-time MM5 runs helped forecasters determine the arrival time of cold surges in order that the research aircraft, radars, CLASS soundings and mobile microphysics van could be deployed to document their passage. The CSU CHILL and NOAA-K radars collected excellent data sets which, when combined with the airborne and surface hydrometeor measurements, will help interpret polarization measurements in terms of ice crystal type.
WISP94 was a multiagency program, which included scientists from the following organizations:
Roy Rasmussen is the WISP scientific coordinator and the chairman of the WISP Scientific Steering Committee and Marcia Politovich is the WISP field operations coordinator. A data catalog is available from the Research Applications Program.
A parametric radiative transfer model, developed by Li Li and J. Vivekanandan, focusses on the detection and quantification of supercooled water, vapor, and ice in winter clouds. The model can be easily adapted to both ground-based and spaceborne radiometry applications. A typical mean vertical structure of the atmosphere is assumed. The water vapor density profile is approximated by an exponential function defined by integrated water vapor amount and scale height. A linear temperature profile is described by an effective lapse rate and an effective near-surface temperature. A layer of homogeneous liquid cloud profile is specified by its cloud base height, cloud thickness, and liquid cloud water content. A layer of ice cloud above the liquid cloud is defined in a similar way except that the scatterers' mean size and bulk density are also specified. A pressure profile can be derived from the surface pressure and the vertical temperature profile. To avoid unreasonable atmospheric structures, the following constraints are imposed:
It should be noted that meteorological parameters defined above form a complete set of input to our physical model. No other intermediate optical parameters are used; therefore, an intermediate retrieval algorithm is unnecessary.
NOAA ground-based dual-channel radiometers measure downwelling radiation in the zenith direction at 20.6 and 31.65 GHz. In addition, one of the radiometers has a 90 GHz channel. The radiative transfer processes are determined basically by molecular oxygen, water vapor, liquid water and ice. The 20.6 GHz channel, which is offset from a weak water vapor resonant line at 22.235 GHz, senses mainly the integrated vapor and is less sensitive to the pressure and water vapor profiles. The 31.65 GHz channel is primarily sensitive to liquid water, and the 90 GHz channel is sensitive to the absence of ice in the scattering regime. Both emission and scattering should be considered at 31.65 GHz and 90 GHz channels. In this work, we use Liebe's millimeter-wave propagation model (MPM) to calculate the gaseous absorption at all three channels. The Rayleigh approximation is applied to compute the absorption of cloud liquid droplets. The Mie theory is used to obtain extinction and scattering properties of ice particles. The ice particle size distribution is specified by a modified Gamma size distribution. The invariant imbedding method of vector radiative transfer theory is used to simulate absorption, emission, and scattering processes. This model will be used to generate training vectors for the neural network-based retrieval model discussed in the next section.
Li and Vivekanandan have designed a neural network-based retrieval procedure for water vapor and cloud liquid using 20.6 and 31.65 GHz channels. Based on a long-time winter seasonal average, we assume that the liquid water equivalent for the vapor distribution is between 0.4 and 0.96 cm and the total liquid water in the column is between 0 and 0.8. The other variables are fixed at a surface pressure = 83 kPa, vapor scale height = 2 km, cloud thickness = 1 km, cloud base height = 1.5 km, near surface temperature = -2 degrees C and effective lapse rate = 6 degrees per km. Since the NOAA radiometers are well calibrated at the 20.6 and 31.65 GHZ channels, we ingest NOAA's retrieved results into our physical models. By comparing the simulated brightness temperatures with radiometer measurements, the bias in the physical model is estimated. Then the dual input and dual output neural network is trained by a set of bias-free simulated data. A dry adiabatic atmospheric sounding and the corresponding three-channel brightness temperatures are used to calibrate the 90 GHz channel in the absence of any ice clouds.
The actual brightness temperature measurements of the two and three-channel radiometers are used to retrieve vapor, liquid and ice quantities. Results of two and three-channel neural network models are compared with the radiosonde observations of vapor and to NOAA's retrieval of vapor and liquid. In the case of a two-channel technique, downwelling brightness temperatures were measured by a NOAA radiometer located at Platteville, Colorado, on 15 March 1991. Results show excellent agreement between neural network-based inverted quantities and NOAA's retrieval techniques. Any deviation between remotely sensed vapor amounts and radiosonde observed values is primarily due to the spatial variation of vapor.
Three-channel radiometer measurements collected by the Erie radiometer were also processed. Comparison between the three-channel radiometric retrieval of water vapor and liquid water with NOAA's model results showed good agreement. Water vapor values retrieved by NOAA and from the neural networks have been compared against independent radiosonde measurements. Assuming the radiosonde data is accurate, the three-channel model-produced results are about the same as, or better than, those of NOAA's dual-channel model. The additional 90 GHz channel enabled the retrievals of mean ice particle size. It is worth mentioning that the forward model can match accurately the measured brightness temperature when it is initialized with the retrieved microphysical quantities of the three-channel model. In other words, the retrieved quantities are indeed a model solution. Therefore it is feasible to retrieve ice information using ground-based three-channel radiometers. Thus, a well-trained neural network is capable of retrieving water vapor, cloud liquid, and cloud ice. The performance of these techniques can be verified by comparing with radar and aircraft estimated values of the same.
As part of the WISP94 field program, a comprehensive evaluation program was conducted to evaluate a set of diagnostic icing and turbulence algorithms as applied to different numerical models. The program, WISP94 Real-time Icing Prediction and Evaluation Program (WRIPEP), ran concurrently with the WISP94 project from 25 January to 25 March 1994. WRIPEP consisted of a real-time display for visualizing the icing and turbulence forecasts, and a statistical package to evaluate the forecast quality of a number of model-algorithm couplings. Roelof Bruintjes, Gregory Thompson and Frank Hage designed and developed WRIPEP while Barbara Brown and Randy Bullock developed the statistical analysis package.
The overall purpose of WRIPEP is to evaluate (in real time and post-analysis) the present icing and turbulence algorithms, as applied to the most up-to-date numerical forecast models, and to conduct a near real-time verification exercise using pilot reports and in situ measurements from the research aircraft.
The WRIPEP display is quite flexible and has many features. An example shown in Figure 1 , consists of four X-windows and a Graphical User Interface (GUI) for responding to users' requests. Three of the windows are used to show model-generated forecasts while the fourth (lower-right) is used to show symbolic products such as pilot reports (PIREPs), AIRMETs, and SIGMETs. The display is designed for icing and turbulence forecasts, but it could be easily modified to include ceiling and visibility or other aviation products.
The primary numerical forecast models used to calculate icing and turbulence potential were the Eta and MAPS systems run at the National Meteorological Center (NMC). The Eta model is a recent NMC development; one of its fundamental aspects is the incorporation of step topography and the vertical eta coordinate. The Mesoscale Analysis and Prediction System (MAPS) was developed at NOAA's Forecast Systems Laboratory (FSL) and now runs operationally at NMC as the Rapid Update Cycle (RUC). In addition, the fifth-generation Penn State/ NCAR mesoscale model (MM5) was run at NCAR in real time during the field program and has been utilized in this study as well.
Three diagnostic icing and two turbulence algorithms were applied to output from each model configuration. These icing algorithms, in various stages of development and testing by several organizations cooperating in WRIPEP, use temperature and relative humidity thresholds to diagnose cloudy environments conducive to icing. The 40 km version of the Eta model and all resolutions of the MM5 model contained water and ice parameterizations that explicitly predicted cloud liquid water and ice fields. These, too, were used in predicting locations of aircraft icing and were verified statistically.
The statistical analyses do not indicate that one algorithm is superior to the others; however, the algorithm developed by RAP does provide operational forecasters the opportunity to assess its skill and also refine the automated diagnostic output with additional data sources such as satellite data and surface observations.
An overview of WRIPEP and results from the performance of the algorithms together with the statistical analyses are summarized in a series of three conference papers to be presented at the 1995 American Meteorological Society (AMS) Annual Meeting in Dallas.
Christopher Davis has studied a mesoscale cold surge observed during the WISPIT field experiment. This was a transient feature marked primarily by a sudden increase in windspeed and brief heavy snow over northeastern Colorado during 11-12 March 1993. Abrupt changes in surface pressure and temperature indicative of a gravity current were not noted. Gradual temperature falls and pressure rises were observed behind the velocity surge; however, there was no cyclonic wind-shift nor did the feature last for more than a few hours, suggesting that it was not a classical front. Data indicates that the surge originated near the peak of the Cheyenne ridge, a 400 m, east-west elongated hill near the border between Colorado and Wyoming. Simulations using an idealized, two-dimensional isentropic model show that the transient surge results from impulsively started flow over heated terrain. Heating creates convergence that is in-phase with the velocity surge at the leading edge of the topographic wave. The result is a coherent disturbance in velocity which moves downstream at nearly the mean-flow speed.
Simulations with the PSU/NCAR nonhydrostatic model (MM5) using a horizontal resolution of 6.7 km captured the evolution of the observed surge. These simulations identify the surge as a topographic wave that is modified by heating, but they also indicate an important role played by latent heating in the updraft at the leading edge of the surge. Because the latent heating occurs mainly downstream from the Cheyenne Ridge, the formation of the topographic wave is unaltered by it; however, the updraft at the leading edge of the surge greatly intensified, leading to a narrow, propagating band of heavy snow.
Ben Bernstein and Politovich analyzed a WISP91 freezing drizzle case in which mesoscale weather features played an important role in the development of a hazardous icing situation. During 16-17 March 1991, a weak low pressure system moved across south-central Colorado and brought a complicated weather scenario to the Denver area. In advance of the system, strongly stratified, moist, sub-freezing air was drawn into eastern Colorado. South-southeasterly winds forced this air to rise along the southern slope of the Palmer Lake Divide (PLD), a 2500 m high ridge which extends eastward from the Rocky Mountains, just north of Colorado Springs. The additional lift caused by upslope conditions formed an extensive area of moderate freezing drizzle on the south side of the PLD. These conditons persisted for several hours during the early morning of 16 March (approximately 3:00 a.m. to 7:00 a.m. LT) from Colorado Springs to Limon, Colorado. The freezing drizzle turned to snow after 1500 UTC, 16 March (8:00 a.m. LT).
At that time, the center of the low pressure system was approaching Colorado from the southwest. Satellite imagery showed that an area of cold cloud tops, indicating a deep cloud layer, was associated with the area of snowfall. A series of soundings was launched from a network of seven CLASS (Cross-Chained LORAN Atmospheric Sounding System) sites across northeastern Colorado as the system approached. These soundings revealed the existence of a deep, dry layer above the strongly stratified, subfreezing layer of air near the surface. This dry layer appears to have been caused by a long period of downward motion, as westerly winds descended the eastern slopes of the Rocky Mountains in advance of the center of the low. Snow falling from the upper-level cloud sublimated in this dry layer; after several hours the layer had sufficiently moistened so that crystals survived the fall to the ground. Thus, the dry layer served to delay the initial northward advance of snowfall at the surface behind the leading edge of the cold air, as was evidenced by a lag in the northward progression of reflectivity recorded by WISP research radars. As the low pressure center moved across the WISP domain, winds aloft switched from southerly to easterly. The area of snow moved toward the foothills from the eastern end of the network (near the Kansas-Colorado border). Again, the elevated dry layer delayed the progression of snowfall reaching the ground and kept it from falling into a low-level shallow supercooled liquid cloud in the western part of the network. By 0300 UTC, 17 March, an eastward-moving low pressure system had passed Denver. This caused the westward progression of deep cloud and snowfall to cease, leaving the shallow supercooled cloud layer intact. The University of Wyoming King Air research aircraft performed a sounding near Greeley at 0305, 17 March, and found large supercooled liquid water droplets near the top of the cloud. Large droplets are particularly hazardous to aircraft, because they may rapidly create rough accretions of ice on the airframe and significantly hamper aircraft performance. Data from a nearby microwave radiometer and wind profiler indicate that the supercooled liquid water dissipated just after the aircraft flight and that shear conditions necessary for the formation of large droplets (as described by Pobanz et al. 1994) had only formed about an hour before the flight. The shallow liquid cloud was limited in extent, with clear skies to the west and snowfall to the east of Greeley. This case shows the transient nature of some icing events and the interplay between upper and lower-level cloud features in producing freezing drizzle.
Measurements from ground-based remote sensors, including dual-channel microwave radiometers, ceilometers, radio-acoustic sounding systems (RASS), and a K-band (.86 mm wavelength) radar were combined to produce vertically-resolved water vapor and liquid water profiles through the atmosphere. This work was conducted by Politovich in collaboration with the NOAA ETL.
At present, there is no single instrument that can reliably, accurately and remotely determine the distribution of cloud liquid water in the atmosphere. This information would be useful for several applications, the most notable being aircraft icing, numerical model initialization, and weather modification. In order to estimate altitudes at which supercooled liquid resides in the atmosphere, the following information is needed:
The most difficult measurement to obtain is the liquid cloud top height. One method tested uses a simple adiabatic parcel lifting model initiated at a height determined from a ceilometer, temperature and pressure from RASS or rawinsonde, along with total integrated liquid water (ILW) from a ground-based dual-channel microwave radiometer to obtain an estimate of the liquid cloud top. This tends to underestimate the true liquid cloud top; for two cases examined in detail, 61% of icing pilot reports in the area were from above this cloud top estimate. Vertical cloud boundaries from a K-band radar were also used in the study; these often indicated thicker clouds than were observed from the research aircraft, possibly due to the ambiguity of the ice/liquid phase distinction. In addition to defining vertical limits for cloud liquid, initial attempts at estimating the liquid water profile were tested. The simplest assumption is that liquid water is distributed uniformly through the vertical extent of the cloud. Comparisons with actual soundings through clouds show this to be a a poor assumption, but it serves as a useful first guess.
The simple adiabatic model described above provides another estimate of the liquid water profile. As air parcels rise and vapor condenses, liquid increases with height in a prescribed manner assuming no depletion by entrainment or ice processes. The actual amount of liquid condensed depends upon the temperature and, to a secondary extent, on the pressure at cloud base; but the shape of the profile is similar for most temperatures. Layers of air may also be lifted to cool and form clouds; the resulting liquid profiles depend on the initial temperature and moisture profiles of the lifted air. This information is generally not available in real time, however. In addition, along with the lifting process, entrainment or interaction with the ice phase (e.g., preferred deposition or riming) may alter the liquid profile from an adiabatic one.
If data are available to characterize the profile of SLW at a location of interest, these data may be used to construct a profile which may be used as a "climatological" mean. Such a profile was derived from thirty-six vertical soundings through cloud obtained during WISP. After retrieving cloud base and top altitudes and measuring the ILW, a climatological profile may be applied to provide an estimate of liquid water as a function of height.
These methods for estimating the liquid profile are shown in the figure as direct comparisons between estimated profiles and measurements obtained from research aircraft. The adiabatic and climatological profiles are in general agreement with the aircraft measurements and are definitely more realistic than the uniform profile. Several sources of error influence the comparisons. There is some degree of error inherent in the liquid water measurements from the aircraft-borne instruments and the radiometer. Additional differences may be contributed by horizontal variations of liquid and the fact that the aircraft sounding locations were 20-32 km away from the radiometer site.
The clouds used in this study were probably "best-case" candidates for these tests, that is, they were totally supercooled and had little to no precipitation. Our studies show that these methods hold some promise for estimating liquid water by remote sensing and provide a basis for additional studies in this area.
Two papers were recently completed describing details of this study and a companion study which included water vapor profiling: "Moisture Profiling of the Cloudy Winter Atmosphere Using Combined Remote Sensors" by Stankov et al. (1994) and "Determination of Liquid Water Altitudes Using Combined Remote Sensors" by Politovich et al. (1994).
Kevin Manning, David Gill and Davis carried out a real-time forecasting exercise using the nonhydrostatic, PSU/NCAR model (MM5) in support of WISP94 field operations. The central aims of the exercise were (1) to guide WISP94 field operations, with special attention to the forecast range from 6 to 24 hours and at higher resolution than available from NMC's operational models; (2) to test the new Reisner (Option 4) mixed-phase microphysics for a large sample of simulations of winter weather; (3) to investigate the added value of increasing horizontal resolution to 6.7 km in order to better represent terrain effects and frontal structures. A total of 120 forecasts were made during a 60 day period from 25 January to 25 March 1994. Forecasts were run twice daily, each for 24 hours. Preliminary results from model verification studies indicate that the excessive cloudiness in the upper troposphere was not removed by mixed-phase microphysics, suggesting that more sophisticated microphysics is needed or other factors such as the accuracy of the advection of moisture variables should be considered.
Jon Reisner, Roelof Bruintjes and Rasmussen continued to further upgrade the microphysical parameterizations implemented in the NCAR/Penn State MM5 model. In the previous version of the microphysical parameterization only the mixing ratios for the different water species in the model were predicted and the number concentration of the ice species were prescribed. This was changed in the upgraded version of the parameterization to include predictions for the number concentrations of ice, snow and graupel. Observations from WISP cases showed cloud liquid water being depleted too rapidly when a fixed number concentration was prescribed, leading us to implement equations to explicitly predict number concentration of ice species.
In order to test the performance of the parameterization and examine the explicit prediction of supercooled liquid water (SLW), numerical simulations were conducted for three different WISP cases representing different types of storms. These were 1) an anticyclonic storm (13-15 February 1990); 2) a shallow cyclonic storm (13-14 March 1990); and a deep cyclonic storm (5-7 March 1990). The simulations were conducted using four different domains with different horizontal resolutions of 60 km, 20 km, 6.7 km, and 2.2 km. The 20 km domain was run concurrently with the 60 km domain, whereas the 6.7 and 2.2 km domains were run in a one-way nest mode. Initial and boundary data for the 60 km domain were supplied from the NMC's archived forecast model fields. The results from the simulations were compared to data obtained from the WISP 1990 field program.
Comparisons between the model results and the observations revealed that riming and depositional growth of snow and/or graupel were the dominant microphysical processes responsible for the depletion of SLW especially when the number concentrations of the ice species were fixed. When the number concentrations of the ice species were explicitly predicted, riming and depositional growth were limited and the simulated microphysical fields showed better agreement with the observations.
Reisner, Bruintjes and Rasmussen are currently preparing a paper describing the results from these experiments.
Bruintjes and Thompson in collaboration with Tom Lee from the Naval Research Laboratories in Monterey, California are comparing satellite measurements of integrated water vapor (IWV) and integrated liquid (ILW) water with model based values of these quantities.
Ideally, accurate forecasts of aircraft icing should be determined by forecasts of supercooled liquid water and the size and concentration of cloud droplets. However, numerical models operationally in use at the NMC have only begun to incorporate explicitly prognosed cloud liquid water. In these early versions, there are a few inherent problems with model-generated cloud liquid water. First, since there are no measurements of this parameter, models cannot begin with an initial analysis field of cloud liquid water. Second, most analyses smooth out the relative humidity horizontally and vertically since they are produced on horizontal scales of approximately 60 km and vertical scales of 500 m or more; some finer scale variations exist. This tends to de-emphasize important regions of high humidity. Last, combining these two factors results in a "spin-up" problem. Since cloud water is not an analyzed field, a model must produce its own cloud water by increasing the already too low (analyzed) relative humidity to saturation, then converting humidity to cloud liquid water using parameterizations which themselves have limitations.
Validation data for an icing forecast are sparse at best, coming mostly from ground-based upward-looking microwave radiometers which give accurate point or small-area measurements and from experimental aircraft observations. Aircraft pilot reports give qualitative information about icing but cannot be used to quantify cloud liquid water. However, over oceans, the Special Sensor Microwave Imager (SSM/I) aboard the Defense Meteorological Satellite Program (DMSP) spacecraft gives good estimates of integrated liquid water and excellent estimates of integrated water vapor at a spacing of 25 km. Thus, there exists an opportunity to assess model performance in terms of integrated cloud liquid water and integrated water vapor over large oceanic regions.
Comparisons over the eastern Pacific Ocean between the SSM/I data and numerical model output data collected during WISP94 were conducted to evaluate the model performance. A comparison with the SSM/I data was conducted using explicitly predicted cloud liquid water from: 1) NMC's soon-to-be operational mesoscale Eta model and 2) the fifth-generation PSU/NCAR mesoscale model (MM5). Both models include parameterizations to explicitly predict water and ice species. Two parameters, integrated liquid water and integrated water vapor, were compared with the SSM/I measurements. While integrated liquid water is the more critical parameter, integrated water vapor is more accurately obtained from both the models and the SSM/I and also can provide useful comparisons. A total of ten case days were investigated.
The comparisons indicated that numerical model cloud liquid water parameterizations appear to perform best in strong, dynamically forced, synoptic-scale situations. Weakly-forced, mesoscale features provide the biggest challenge to liquid water parameterizations, showing substantial differences between model and satellite fields of ILW and IWV.
The poor correspondence between SSM/I data and model-predicted ILW emphasized the need for improving model initializations of moisture fields. From the case studies, it appears likely that significant improvements to numerical model forecasts can be gained by incorporating satellite data. This is obviously true for 24-hr and longer forecasts but may also be true for short-term forecasting.
A paper describing the preliminary results from the comparisons was submitted for presentation at the 1995 American Meteorological Society Annual Meeting in Dallas.
The overall goal of this FAA-funded effort is to provide accurate real-time and 0-30 minute nowcasts of the snowfall rate in the airport area for ground de-icing purposes. In order to do this, real-time snowgauge and radar data are combined into a user friendly product. Key activities this year included deployment of a snowgauge network by Jeff Cole and Rasmussen to test various snowgauges and wind shielding devices, evaluation of the snowgauge performance by Charles Wade, development of an initial Z-S algorithm by Wayne Adams, development of a user-friendly display by Hage and evaluation of two snow echo extrapolation techniques by Gregory Stossmeister. Rasmussen is the overall project manager for this project.
As part of Project SNOW, Cole and Rasmussen deployed a snowgauge network in order to evaluate the performance of various commercially-available snowgauges and wind shielding devices used to measure the liquid-equivalent water content of winter-time solid precipitation (snow, sleet, snow pellets). (A snowgauge is essentially equivalent to a raingauge, but has an ethylene glycol solution in the catchment container to dissolve the solid precipitation and prevent freezing of the liquid contents. A layer of oil retards evaporation from the solution.) The gauges were evaluated in terms of their ability to provide accurate, reliable, real-time measurements of snowfall liquid-equivalent accumulation with a resolution of 0.25 mm. The data collected from the snowgauges was used to monitor snowfall rates and water content for commercial aircraft deicing operations and to aid in the development of radar-reflectivity algorithms used to monitor and predict snowfall rates at the ground.
Seven combinations of snowgauges and shields were evaluated at three sites. At the National Weather Service site at Stapleton, an 8-inch Belfort weighing snowgauge with a Canadian Nipher shield was co-located next to the NWS Belfort gauge with an Alter shield. Hourly observations of snowfall liquid-equivalent accumulation from the NWS gauge were provided by the NWS. A second site at Stapleton was located adjacent to the United Airlines maintenance hanger on Quebec St. At this site a 12-inch Electronic Techniques, Inc. (ETI) weighing snowgauge was located inside a Wyoming shield, and two 8-inch Belfort gauges (one with a Nipher shield and one with no shield) were located nearby. Data from these two sites were recorded using a Campbell Scientific Data Logger. An automated weather station provided real-time observations of ambient meteorological conditions.
The third site was located at the NWS Automated Surface Observing System (ASOS) site at Denver International Airport. Gauges at this site consisted of an 8-inch ETI gauge with Nipher shield, an ASOS heated tipping bucket gauge and an ASOS Light-Emitting Diode Weather Indicator (LEDWI) sensor.
Ten rain and snowfall events between January and April 1994 provided an opportunity to evaluate the various gauges and shields under a variety of weather conditions. Wade analyzed these data and found that the gauges are capable of accurate, high-resolution precipitation measurements, provided that shielding is used to reduce the effects of wind. The unshielded gauge at Stapleton typically undermeasured precipitation amounts by 50%. The most significant problem encountered was due to snow sticking to the inside surfaces of the gauge orifice and not being recorded in real time. This effect was variable but sometimes amounted to an under estimation of precipitation by up to 30%. The LEDWI sensor provided useful information on when precipitation was falling and on its relative intensity, but the quantitative rates appeared to be too high. The heated tipping bucket on the ASOS consistently underestimated precipitation amounts by a factor of two or more and does not appear to be a useful or reliable measurement.
During the winter of 1994/1995 the gauges will again be deployed in the field in an effort to improve the measurements. An effort will be made to reduce or eliminate the problem with snow and ice which accumulates on the inside surfaces of the gauge or shield.
Adams has continued development of real-time Z-S models. He found that advection effects have a much more significant impact on the Z-S relationship than on radar-rainfall relationships. Three Z-S models -- the familiar power-law model and two neural-network-based approaches -- all perform very poorly in high-wind-speed events unless great care is taken to temporally and spatially shift radar and snowgauge measurements to maximize the correlation between them. One component of this shift is the fall-speed time delay. In this case, snow, with its much narrower spread of fall velocities, is easier to analyze than rain.
Variations in the relationship between snowfall rate and reflectivity factor poses special problems for a real-time Z-S relationship. Ideally, the relationship will be based on recent data; however, a good regression fit requires a minimum number of Z-S pairs, as well as minimum variations in the values of both Z and S. So a trade off must be made between plasticity and stability, and some limits must be placed on the values of the regression coefficients so that noisy Z-S data do not bias the model to predict unrealistic snowfall rates.
One possible solution to this dilemma is to project snowfall accumulation (for, say, 30 minutes) based on the integral of the reflectivity factor (that is, integrate both "sides" of the Z-S relationship). A regression model based on these data should be less vulnerable to Z-S scatter; preliminary results support this hypothesis. In reality, accumulation projections are what is actually needed to support terminal-area ground de-icing anyway. The procedure of obtaining S from snowfall amounts (a numerical derivative of noisy data) and then re-integrating S to obtain snowfall amounts is rather circuitous and is responsible for at least some of the inaccuracies of Z-S-based snowfall projections.
Stossmeister evaluated two algorithms developed at NCAR to extrapolate radar reflectivity patterns. The first algorithm is called TITAN which stands for Thunderstorm Identification, Tracking, Analysis, and Nowcasting. This algorithm is an object-oriented approach to the problem. It is dependent on identification of thunderstorm "cells" in a radar volume which it can trace backward in time and extrapolate forward into the future. The second algorithm developed for winter applications is called REP which stands for Radar Echo Prediction. The REP algorithm was designed for large amorphous areas of reflectivity which contain regions moving in different directions at different speeds. Snowbands with differential motion are often observed in the Front Range area. Such storms are often poorly handled by TITAN. The REP algorithm operates on the principle of pattern matching. The original gridded data representation of the storm is divided into a number of cells which are spatially logged over previous observation of the storm to find the best match with the previous data plane. The motion of the subgrids is then extrapolated into the future to "predict" the reflectivity field.
The two tracking methods are being tested using data obtained from fifteen snowfall events observed near Denver. The goal of the evaluation is to determine the strengths and weaknesses of each algorithm and to decide which algorithm provides the best extrapolation of radar echo during winter storm conditions.
Average probability of detection and false alarm rate scores computed for one winter season suggest that the TITAN algorithm in its present configuration has skill over persistence in the 10-30 minute time period for reflectivities greater than 25 dBZ. TITAN outperformed persistence in the lower reflectivity ranges only in the 10-20 min time frame. This result may be misleading however, due to a significant number of cases where echo covered most of the verification area. As of this writing, the evaluation of the REP algorithm was not sufficiently complete to report skill statistics. A more detailed examination of the evaluation results is ongoing in order to determine the performance characteristics of both TITAN and REP in comparison with persistence in different weather situations. A more detailed analysis of the skill scores on a radar volume-by-volume basis will be performed in an attempt to define each algorithm's performance characteristics as a function of radar echo characteristics. Radar data from other winter seasons may also be used to produce a more varied set of input data since weather characteristics vary greatly from year to year. Testing of the algorithms for real-time operations with NEXRAD radar data will begin this winter.
Rasmussen, Cole, and Kevin Knight have analyzed the meteorological conditions associated with commercial large jet-aircraft accidents in which ice accumulation on the wing prior to takeoff was a primary factor in the accident. Of the ten large jet-aircraft accidents in which wing surface ice contaminants were considered a contributing factor, seven had weighing snowgauge data available. Of these, five were snow events and two were freezing drizzle events. The common factors in the five snow cases were light precipitation rates (0.08-0.1 inches/hr), temperatures of 25 to 31 degrees F, and windspeed 8 to 13 knots. This particular combination of meteorological conditions may be very conducive to the buildup of hazardous ice accumulations on critical aircraft surfaces. The nearly identical liquid equivalent precipitation rate suggests that accidents may be avoided if snowstorm periods with precipitation rates equal to or greater than this value are avoided. This would require a real-time estimate of liquid equivalent snowfall rate to be available to ground operations coordinators and airline station control centers.
In contrast to snowfall rate, no common visibility value was apparent. For instance, visibility was found to range from 0.25 to 2 miles.
Figure 1 presents a plot of visibility versus precipitation rate as well as regression lines from previous studies of snowfall rate and visibility, including recent results from an ASOS visibility sensor comparison with an ETI weighing gauge with a Nipher windshield that we conducted during the winter of 1993-1994 at the Denver ASOS site. Also indicated on the right hand side is the NWS snowfall categories for snowfall rate based on visibility. The results show that visibility has a nearly order of magnitude variation at the critical snowfall rate of 0.1 inch per hr, and that this variation spans the NWS visibility-defined snowfall rate categories of S-, S, and S+. Thus, the use of visibility to determine snowfall rate, especially rates near the critical value of 0.1 inch per hr, is not recommended. Part of the reason for the wide scatter in visibility for a given snowfall rate is the observed order of magnitude variation in snowflake density for a given snowflake size. Since visibility reduction due to a snowflake is proportional to the cross-sectional area of the snowflake and not the density of the snowflake, wide variations in precipitation rates are possible for a given snowflake size due to variations in snowflake density. This can often be very misleading to pilots and ground operations, who often rely on visual estimates of snowfall rate in the absence of other data. In particular, the high visibility during the La Guardia accident was also noted by the first officer of the aircraft. He described the snowfall as "not heavy, no large flakes". Since he did not have access to any real-time liquid equivalent precipitation rate, the only way he could make a judgement that the snowfall rate was not heavy was by visibility; and he even notes that there were no large flakes. This visual estimate of snowfall rate gave him a false sense of security that the conditions were not that bad, when in fact the conditions were actually similar to those in previous ground de-icing accidents. The La Guardia weather observer characterized the snow as wet, which is consistent with the snowflakes having a high density, leading to the high precipitation rates observed. Thus, the use of visibility to estimate snowfall rate for ground de-icing purposes should be avoided. A paper describing these results in more detail will be presented at the Aviation Weather Conference in Dallas, Texas in January.
Hail avoidance is of major concern to the aviation community. Damage to airfoils can significantly degrade aircraft performance by reducing lift. In-flight encounters can shatter windshields and are the primary cause of in-flight engine shutdowns. The accurate detection of hail would have obvious benefits. Although hail detection has been the subject of numerous studies, an independendent testing of suggested techniques has not been done. Consequently, RAP conducted field programs during the summers of 1992 and 1993 to collect the necessary in situ hail measurements and radar observations. Research activities during 1993 were concerned primarily with the evaluation of algorithms that identify hail storms based on radar reflectivity measurements alone. Future work will determine the added improvement, if any, that multiparameter radar measurements provide for hail detection. Also, a preliminary investigation was conducted in which the radar signatures for hail were verified by in situ particle measurements made by research aircraft.
Cathy Kessinger has completed an evaluation of three radar reflectivity-based hail algorithms that included the Next Generation Radar (NEXRAD) hail algorithm, currently used with the NWS WSR-88D radars, and two algorithms from the National Severe Storms Laboratory (NSSL) that are planned replacements for the NEXRAD hail algorithm: the Probability of Hail (POH) and the Probability of Severe Hail (POSH). During the 1992 and 1993 field projects, surface hail reports were collected by hail intercept crews, a volunteer observing network, the Mountain States Weather Services, and the NWS. These reports formed the verification database for the evaluation effort. The hail reports were edited to correct time or space errors.
The hail algorithms were run in real time using data from the Mile High Radar (MHR) located 15 km northeast of Denver Stapleton Airport. A co-location program was used to match the verification data (hail and rain reports) with hail storm designations made by the algorithms. The matching was accomplished by defining storms as either 30 or 40 dBZ envelopes and by defining influence areas with radii of 5, 10 and 15 km about a reporting site. Contingency tables were created for computation of the Critical Success Index (CSI), the Probability of Detection (POD), the False Alarm Ratio (FAR), the Miss Rate (MR), and the Heidke Skill Score. To evaluate algorithm performance at increasing hail sizes, these quantities were computed for all hail reports, and for hail greater than or equal to 6 mm, greater than or equal to 13 mm, and greater than or equal to 19 mm. Similar to an evaluation done by NSSL, this study found that the NEXRAD "hail" algorithm performed with the least skill (a CSI of 0.47 for any size hail and a storm definition of 30 dBZ). Inclusion of the NEXRAD "probable hail" category with the "hail" designations improved skill levels considerably (a CSI of 0.72).
Because the algorithm detections by POH and POSH are expressed as percentages, a search for a "hail"/"no hail" threshold was conducted. Although designed to detect severe (large) hail the best performances with the NSSL algorithm for hail in Colorado were found for hail of any size. The best performance for the POH parameter was a CSI of 0.90 at an algorithm designated hail probability of 10%. This result was achieved for verification areas with 10 and 15 km radii and for the 40 dBZ storm outline. The best performance for the POSH parameter was 0.86 for verification areas with 15 km radius and the 40 dBZ outline. As hail size increased the performance of all algorithms decreased.
As a starting point for the multiparameter hail detection work, Edward Brandes, Vivekanandan, John Tuttle and Kessinger began a study that compared radar measurements from the CP-2 radar to in situ particle observations made on 24 June 1992 by the T-28 aircraft operated by the South Dakota School of Mines and Technology. Excellent agreement was found between multiparameter radar signatures of hail, raindrops, and mixed-phase precipitation and in situ observations. Radar reflectivity estimates determined by direct (radar) measurement and estimates computed from observed particle distributions generally agreed within 5 dB. Maximum values of differential reflectivity (ZDR) and the fractional contribution of liquid water to total reflectivity differed by less than 0.8 dB and a factor of two, respectively.
The storm had a positive ZDR column that extended more than 2 km above the melting level. Similar columns have been thought to play an important role in hail formation. The column of the 24 June storm was nearly coincident with the storm updraft and contained mixed-phase precipitation. The aircraft data revealed that the ZDR measurement was dominated by a small number of very large raindrops (some exceeding 5 mm in diameter). Trajectories computed with a precipitation growth model suggest that the drops originated primarily with partially or totally melted particles from a quasi-stationary feeder band within the inflow air to the storm and from the storm's reflectivity core. Particle observations and radar measurements at approximately the -2 degrees C level in the column revealed that the fractional contribution of drops to radar reflectivity was roughly 0.5 to 0.8. However, the concentration of supercooled water represented by the drops (a maximum of 0.7 g per cubic meter and an average of 0.2 g per cubic meter) was about half that associated with cloud water. Hence, the relative importance of the large drops and consequently the ZDR column to hail production may be minor.
The goal of this effort sponsored by the FAA is to develop automated techniques for place specific 0-2 hr forecasts of convective weather phenomena that are a hazard or efficiency concern to the aviation weather community. This includes the use of numerical modelling and rule-based methods to forecast initiation, growth, movement and dissipation of thunderstorms. Radar-based techniques for automatically detecting and extrapolating convergence lines, convective storms and cumulus clouds are essential for the thunderstorm forecasting effort.
NCAR has developed a rule-based automated thunderstorm forecast algorithm . Sandra Henry is currently utilizing the automated thunderstorm forecast algorithm to evaluate the accuracy of 30-minute thunderstorm forecasts. The algorithm also provides a scientific tool that allows continued development and fine-tuning of the forecast rules. Results from the algorithm show that the automated forecast rules currently implemented into the algorithm are capable of producing very good forecasts of thunderstorm initiation. However, the quality of the forecast is very dependent on the accuracy of automatically detecting and extrapolating boundary-layer convergence lines. The automatic boundary detections are generated by an algorithm called MIGFA .
Forecasts of thunderstorm extrapolation and dissipation are based largely on the performance of TITAN (an algorithm that automatically detects and extrapolates storms). One of the primary weaknesses identified with the algorithm thus far is the need to identify cumulus clouds and their various stages of development. The algorithm's inability to distinguish between radar echoes associated with cumulus clouds and those associated with other clouds (e.g. mid-level stratiform clouds) can lead to significant over forecasting of thunderstorms.
Integral to the automated thunderstorm forecast algorithm being developed at RAP is the radar detection and forecast position of surface convergence boundaries. MIT/Lincoln Laboratories has developed a Machine Intelligent Gust Front Detection Algorithm (MIGFA) that has been found to perform reasonably well in the detection of convergence lines and reflectivity thin lines using TDWR data. A version of this algorithm was brought to RAP and modified to run off the Mile High Radar (MHR) data. Rita Roberts has evaluated the performance of MIGFA on a limited number of MHR cases which resulted in an average POD of 66% and FAR of 6%. Problems in performance were related to 1) difficulty in boundary detection in regions of high clutter and weak convergent flow, 2) the use of only one elevation scan for identification of boundary locations, 3) the inability to consistently detect stationary, radially-aligned boundaries or spatially diffuse reflectivity thin lines, and 4) false detections due to range ambiguous data from MHR.
To enhance the detection of thin lines on radar reflectivity images, a continuous, 2-D wavelet transform technique was tested by Jason Helland and Carl Hagelberg and incorporated into the MIGFA code. The spectrum of the wavelet consists of two localized Gaussian-shaped peaks which enables the wavelet transform to act as a very selective filter. Use of the wavelet transform then lends itself to both the reduction of noise (specifically ground clutter and data outliers) and multiscale analysis of images to improve detection of a wider range of thin line features. Tests of the wavelet transform technique involved a two-step process: 1) projecting the reflectivity image onto a directionally selective wavelet basis, and 2) combining the multiscale-multiorientation information into an enhanced interest image. The intensity of pixels on this image is proportional to the potential that the pixel lies on a thin line feature. Preliminary tests of this technique using MIGFA indicate that the multiscale analysis improves the range of detection capability; however, thorough testing is warranted.
Additional efforts are underway to ingest NEXRAD data into MIGFA, as the first step toward getting the autonowcaster algorithm to run off of any NEXRAD dataset.
Research continues on methods to forecast winds in the terminal environment. In collaboration with Lincoln Labs/MIT, Andrew Crook has examined the ability of a small-scale numerical model to forecast the motion of gust fronts from Lincoln's Terminal Winds analysis system. The analysis system produces winds at 2 km resolution every 5 minutes, using data from TDWR and WSR-88D radars and LLWAS. A thermodynamic retrieval uses wind data over several time levels. Once the buoyancy field has been retrieved, the model can be integrated forward to produce a forecast. Figure 1 , shows a 40 min forecast of a gust front that moved over Memphis airport on 9 June 1994. A verification analysis showed that the numerical forecast improved over persistence by approximately 30% after 40 minutes.
Research has also been performed on other methods to improve these forecasts. Jenny Sun has examined the ability of the adjoint method to retrieve the buoyancy field. Tests on various datasets have shown that the adjoint method can retrieve the buoyancy field with greater accuracy than the traditional retrieval technique, because of the way it handles noise in the data. Tuttle has also examined the performance of the TREC algorithm (Tracking Reflectivity Echoes by Correlation) in retrieving the horizontal windfield from WSR-88D data. The TREC algorithm could substantially improve the winds produced by the Terminal Winds Analysis System.
This study also includes an examination of the causes of the along-roll features, called "pearls on a string" by Kuettner (1971). These along-roll features are not only seen in the satellite imagery as periodic cumulus clouds but also within the boundary layer as periodic kinematic features derived from dual-Doppler analysis. Figure 3 shows the northwest-southeast rolls, as delineated by the convergence field and time changes in the vertical vorticity field.
During the past few decades, a great deal of pure and applied research has been performed to better understand atmospheric turbulence. These wide ranging efforts dealt with turbulence phenomenology, in-situ and remote detection, numerical modelling, operational forecasting and the response of aircraft in turbulence. The direct application of the resultant knowledge to operational meteorology and aviation has been somewhat limited. This is primarily the result of an all too common problem: inadequate real-time measurements. The only direct measurement of turbulence intensity currently available is from pilot reports (PIREPs). The lack of adequate atmospheric measurements also limits the ability of aviation meteorologists and numerical weather models to generate accurate forecasts of turbulence.
Turbulence has a significant impact on flight efficiency in the tactical and strategic use of airspace. This impact is directly related to the aircrew's desire to avoid impending turbulence encounters and the lack of definitive planning information to know where to expect these encounters. Results include the introduction of delays as crews execute reroutes and altitude changes to avoid encounters, accompanying increases in fuel consumption, and additional workload for pilots and controllers as they coordinate these changes.
Automated reporting of meteorological data from commercial aircraft is an extremely valuable source of information for the operational aviation and meteorological communities. Currently, the Aircraft Communications Addressing and Reporting System (ACARS) provides access to wind and temperature measurements from commercial aircraft. Advances in data-link technologies, combined with the future implementation of the Automatic Dependent Surveillance (ADS) system will provide ever wider access to this data. Augmenting the qualitative, intermittent and subjective AIREPs with quantitative, automated and aircraft-independent turbulence measurements is a high priority within both the air traffic and meteorological communities.
Recognizing the need to ensure the standardization, applicability, and accuracy of aircraft-reported meteorological data, the Air Navigation Commission of the ICAO has created the Automatic Air Reporting Study Group (ATAR SG) to provide advice in this area. Early on, an ad-hoc group of meteorological experts, the predecessor to the ATAR SG, recognized that efforts were needed to determine a method for generating a real-time turbulence algorithm suitable for automated air-reports. Larry Cornman responded to this requirement by developing a fully automated, computationally efficient algorithm designed to be installed on commercial aircraft. In order to maximize the cost effectiveness of generating these automated air reports, the algorithms have been designed to use currently available on-board data and computational resources.
Operational demonstrations of the NCAR turbulence algorithm on commercial aircraft will occur over the next several years in conjunction with the NOAA/FAA/NCAR Commercial Aircraft Sensing of Humidity (CASH) program. Researchers from Transport Canada are also participating in validation efforts involving the NCAR in-situ turbulence algorithm. Discussions with the ICAO are underway regarding adoption of the NCAR algorithm as an international standard for the automated reporting of turbulence from commercial transport aircraft.
A new airport is currently under construction in Hong Kong to replace the aging and saturated Kai Tak International Airport. The new airport is situated on a small island called Chep Lap Kok (CLK) just a few kilometers west of the much larger and relatively mountainous Lantau Island. Since the prevailing boundary-layer flow in the Hong Kong region is generally from the east, the new airport lies in the lee of Lantau Island. This situation creates the possibility that, under certain circumstances, terrain-induced windshear and turbulence (TIWT) produced by the flow over Lantau Island may be strong enough to impact aircraft operations at CLK. Indeed, previous studies have shown that significant TIWT does occur in this area. Therefore, the Research Applications Program in collaboration with the Royal Observatory of Hong Kong has begun to assess the feasibility of an operational system to detect and warn of the presence of significant TIWT. A scientific examination of the nature of the flow over Lantau Island has been emphasized during the early phases of this program, including climatological, numerical and field studies.
A review and reexamination of previous studies by Peter Neilley, Teddie Keller and Brant Foote has revealed that significant turbulence at the new airport site can be expected to occur during any of the basic climatological weather regimes common to Hong Kong. Further, these studies suggested that boundary-layer windspeeds in excess of about 10 meter per sec may be sufficient to generate significant TIWT in the lee of Lantau. The degree of turbulence was found to decrease with height and distance from the Islands.
In order to obtain a better understanding of the nature of the flow over Lantau Island, a series of fine-scale numerical modelling experiments has been conducted by Keller and Terry Clark. Since the parameter space that governs the flow over the island is large (there essentially is an infinite combination of wind and stability profiles possible), two separate approaches to the design of the model experiments have been taken in order to simplify this study. The first approach was to build upon the current knowledge of flow over terrain and the behavior of terrain-induced gravity waves by initializing the model with relevant idealized profiles of wind and stability that have been studied previously. The changes in the perturbation response of the flow over Lantau Island to prescribed changes in the ambient profile parameters that are known to affect that response (e.g. Froude number, critical-level height, etc.) are then observed. The second approach was to initialize the model with observed profiles during times known to be associated (or not associated) with turbulence at CLK.
The results from the idealized profile simulations have shown considerable sensitivity of the response to some of the details of the idealized profiles. For example, runs in which a low-level easterly flow of 15 m/s reversing to a westerly flow aloft (through a tanh profile) with a constant static stability of 0.01 per sec showed that the height of the critical level played an important role in determining the magnitude of the perturbed response with maximal response found for critical levels near 2.5 km. Similar sensitivity to the ambient windspeed was also found with the largest perturbations found for flows (using the same static stability) around 10 meter per sec. Lesser sensitivity was found for changes in the wind direction and the degree of shear near the critical level. One set of experiments in which an orthogonal component of wind was added to the basic flow revealed that the nature of the downstream response strongly depended on the details of this orthogonal flow profile. In these cases, the definition of a critical level is called into question and the dynamics of gravity wave behavior in such three-dimensional shear flows is largely unexplored. These results also suggest that accurate predictions of the terrain-induced turbulence may require frequent and precise information on the structure of the boundary layer flow.
The numerical modelling studies using actual sounding data performed to date have concentrated on a few summer situations in which significant turbulence was known to have occurred. During these cases, no critical-level was present in the ambient flow and the low-level static stability was generally quite small. Initial modelling efforts on these cases suggest that very high horizontal resolution is needed to capture the turbulence flow response. This suggests that a small-scale dynamical mechanism such as mechanical vortex shedding in the surface layer of the terrain may be responsible for the observed turbulence in these cases.
The Hong Kong Program has also embarked on an extensive field program called LANTEX
centered on the observations taken with NCAR's instrumented King Air aircraft. The King Air
has been routinely sampling the turbulence environment in the vicinity of CLK since April
1994. The field study program also deployed a Doppler lidar at CLK, an ATD integrated
sounding system based at an upstream site, and a large array of automated surface weather
observing stations scattered throughout the Hong Kong Territories. As of August 1994, fourteen
significant turbulence cases have been observed including several cases qualitatively described
as being "severe enough to cause a 747 to go around." In these severe cases, vertical
velocity variations of about 5 meter per sec were observed all along the proposed landing and
takeoff corridors. Preliminary analysis of the field program cases to date, suggest that mean
boundary layer windspeed is a skillful predictor of the turbulence intensity produced by Lantau
Island. This suggests that for these summertime cases with low stability and no significant wind
shear or critical levels that mechanical mixing induced by the details of the terrain may be the
principle mechanism responsible for the generation of the turbulence. Verification of the high
amplitude gravity wave breaking mechanism explored with the numerical modelling studies
awaits upcoming flights in the fall and winter when conditions more conducive to gravity wave
amplification are common.
An important aspect of the work to assess the feasibility of an operational windshear warning
system will be the construction of a functional prototype. Preliminary work has been carried out
by Gerry Wiener and Zhonqi Jing to define a general system architecture
to support TIWT detection and forecast
algorithms. It is envisioned that the OWWS algorithms
will be based upon fuzzy-logic methodology. This technique combines the qualitative
knowledge of an expert system with the quantitative methods of a numerical algorithm. It
incorporates all available information and builds upon consensus of this information (wind
profiler data, anemometer data, MM5 model output, and Clark model lookup tables, etc.). Each
input data source will be pre-processed to a usable stage (e.g., wind and/or turbulence
measurements from the profiler) and then passed along with quality control information to the
fuzzy-logic algorithm. Using this framework, the algorithms will be robust and accurate
producing the best possible detection probabilities and lowest false alarm rates, given the
available data. Furthermore, this type of algorithmic system will be flexible in terms of
including new data sources, if they become available.
One important data source to the OWWS algorithms will be the wind profilers. Wind profilers
have been shown to produce reliable and accurate wind measurements in steady-wind and
minimal-clutter environments. However, in strongly varying winds, clutter and/or rain data
from these devices can become erroneous. In order to use profilers for automated TIWT
detection, these problems must be overcome. Cornman and Cory Morse have been
working together in developing
better algorithms to provide the required operational utility including thorough analysis of the
theoretical wind measurement capabilities of boundary layer profilers. From a mathematical and
numerical analysis, the specific parameters of a linear windfield that can be reliably estimated
have been determined. Algorithms that build upon this information have been developed and, in
this preliminary stage, show great potential in the ability to produce high temporal resolution (1-2
min), wind vector and turbulence estimates. Further research is needed to solve the
contamination effects of clutter, variable signal-to-noise ratios, velocity folding and falling
The Terminal Area Surveillance System (TASS) program within the Federal
Aviation Administration is developing a radar system that will perform
both aircraft surveillance and weather hazard detection functions. This
system is early in the research and development process and will
probably utilize electronically steered beam technology. The
electronic beam steering should provide greater flexibility in
dynamically allocating radar resources between aircraft surveillance
and weather observations than has been possible in previous systems
with mechanically scanned antennas. RAP is providing expertise in
determining scan strategies appropriate for the detection of hazardous
weather phenomena, developing automated methods for detection of the
phenomena and creating three-dimensional graphical displays for their
In order for the TASS system to perform its meteorological functions,
Brandes and Vivekanandan have proposed that a full hemispherical
surveillance scan be obtained at regular intervals. By necessity, this
scan (and others developed for monitoring weather hazards) is
interlaced among aircraft surveillance scans. The surveillance scan
will consist primarily of low resolution radar reflectivity information
that will be used to identify regions with possible weather hazards.
Special high resolution scans that sample only in regions of potential
hazard will then be implemented. The special scans may be made of
detailed reflectivity information and radial velocity measurements.
This information will be processed by other algorithms to quantify the
hazard and to monitor trends in intensity. Strategies have been
produced for a wide variety of phenomena.
Several radar designs with widely different sampling characteristics
have been proposed for use in the TASS system. To aid in the design
and evaluation of potential radar configurations, as well as evaluation
of the proposed scan strategies, Vivekanandan and Charlie Le are
developing a TASS radar simulator. This simulator will use observed or
modelled data of weather phenomena as input and produce data
representative of the radar design and scan strategy as output.
Through the use of the simulator, the system designs and scan
strategies will be tested and optimized without the need to build a
series of costly hardware prototypes.
Dave Albo has developed a new microburst detection algorithm for use in
demonstrating the weather detection capabilities of the TASS radar.
This algorithm uses fuzzy logic techniques to combine a variety of
information to determine the existence of microburst events. The
information used consists of an estimation of windshear, an estimate
for the presence of precipitation and its effect on the likelihood of
microbursts, and an estimate of the effect of clutter contamination on
microburst detection. These estimates are combined and used to build
areas most likely to be microbusts, which are matched against expected
microburst size and shape criteria, with final microburst shapes drawn
around the appropriate regions. The algorithm performs comparably to
the existing TDWR microburst detection algorithm on the datasets
evaluated, and performance is maintained through the lower data
resolutions that are anticipated for TASS scanning.
Bill Myers has led the effort to enhance the 3DTV, the
Three-Dimensional Terminal Viewer
, which provides the user with a perspective view of real-time
aviation weather hazards in the terminal area from a variety of
viewpoints, e.g., from the cockpit of any aircraft or the airport
control tower. Enhancements included a high resolution airport model,
a configurable air traffic simulation system and further developments
of prototype 3-D aviation weather hazard glyphs (icons) all integrated
into a standard simulation environment. The inclusion of the high
resolution airport model along with positional data from all ground
vehicles could provide a means for airport operation under poor
visibility conditions. The air traffic simulation is the first step in
development of an air traffic study or training system to examine the
effects of different terminal area air traffic management strategies in
the presence of aviation weather hazards. In contrast to simply making
the glyphs cognitive extensions of the the 2-D graphical icons
currently in use in the aviation weather system today, as was done in
the previous version of 3DTV, the development of new glyphs to
represent aviation hazards examines the use of suggesting atmospheric
dynamics in the presentation of a hazard region to more intuitively
communicate the nature of the hazard.
The representation of data and extracted features as an aviation
weather hazard "virtual world" in which one is
"immersed" was also explored. In a collaborative effort with
NCAR's Scientific Computing Division (SCD), the software was ported to
an immersive environment called the CAVE developed by NCSA. This
environment is characterized by projecting the application onto three walls
and floor of a room in stereo. When the users, wearing shutter glasses,
view this scene, they have the impression of being immersed in a 3-D
environment complete with depth perception. Navigation paradigms for
exploring this world and previewing flight paths were also explored.
This immersive system was presented by Myers at the ACM/IEEE SIGGRAPH conference.
IX. Terminal Area Surveillance System
A. Scan Strategy
B. Microburst Algorithm
C. Three-Dimensional Display
X. Aviation Weather Products Generator (AWPG)
An important aspect of the work to assess the feasibility of an operational windshear warning system will be the construction of a functional prototype. Preliminary work has been carried out by Gerry Wiener and Zhonqi Jing to define a general system architecture to support TIWT detection and forecast algorithms. It is envisioned that the OWWS algorithms will be based upon fuzzy-logic methodology. This technique combines the qualitative knowledge of an expert system with the quantitative methods of a numerical algorithm. It incorporates all available information and builds upon consensus of this information (wind profiler data, anemometer data, MM5 model output, and Clark model lookup tables, etc.). Each input data source will be pre-processed to a usable stage (e.g., wind and/or turbulence measurements from the profiler) and then passed along with quality control information to the fuzzy-logic algorithm. Using this framework, the algorithms will be robust and accurate producing the best possible detection probabilities and lowest false alarm rates, given the available data. Furthermore, this type of algorithmic system will be flexible in terms of including new data sources, if they become available.
One important data source to the OWWS algorithms will be the wind profilers. Wind profilers have been shown to produce reliable and accurate wind measurements in steady-wind and minimal-clutter environments. However, in strongly varying winds, clutter and/or rain data from these devices can become erroneous. In order to use profilers for automated TIWT detection, these problems must be overcome. Cornman and Cory Morse have been working together in developing better algorithms to provide the required operational utility including thorough analysis of the theoretical wind measurement capabilities of boundary layer profilers. From a mathematical and numerical analysis, the specific parameters of a linear windfield that can be reliably estimated have been determined. Algorithms that build upon this information have been developed and, in this preliminary stage, show great potential in the ability to produce high temporal resolution (1-2 min), wind vector and turbulence estimates. Further research is needed to solve the contamination effects of clutter, variable signal-to-noise ratios, velocity folding and falling precipitation.
The Terminal Area Surveillance System (TASS) program within the Federal Aviation Administration is developing a radar system that will perform both aircraft surveillance and weather hazard detection functions. This system is early in the research and development process and will probably utilize electronically steered beam technology. The electronic beam steering should provide greater flexibility in dynamically allocating radar resources between aircraft surveillance and weather observations than has been possible in previous systems with mechanically scanned antennas. RAP is providing expertise in determining scan strategies appropriate for the detection of hazardous weather phenomena, developing automated methods for detection of the phenomena and creating three-dimensional graphical displays for their depiction.
In order for the TASS system to perform its meteorological functions, Brandes and Vivekanandan have proposed that a full hemispherical surveillance scan be obtained at regular intervals. By necessity, this scan (and others developed for monitoring weather hazards) is interlaced among aircraft surveillance scans. The surveillance scan will consist primarily of low resolution radar reflectivity information that will be used to identify regions with possible weather hazards. Special high resolution scans that sample only in regions of potential hazard will then be implemented. The special scans may be made of detailed reflectivity information and radial velocity measurements. This information will be processed by other algorithms to quantify the hazard and to monitor trends in intensity. Strategies have been produced for a wide variety of phenomena.
Several radar designs with widely different sampling characteristics have been proposed for use in the TASS system. To aid in the design and evaluation of potential radar configurations, as well as evaluation of the proposed scan strategies, Vivekanandan and Charlie Le are developing a TASS radar simulator. This simulator will use observed or modelled data of weather phenomena as input and produce data representative of the radar design and scan strategy as output. Through the use of the simulator, the system designs and scan strategies will be tested and optimized without the need to build a series of costly hardware prototypes.
Dave Albo has developed a new microburst detection algorithm for use in demonstrating the weather detection capabilities of the TASS radar. This algorithm uses fuzzy logic techniques to combine a variety of information to determine the existence of microburst events. The information used consists of an estimation of windshear, an estimate for the presence of precipitation and its effect on the likelihood of microbursts, and an estimate of the effect of clutter contamination on microburst detection. These estimates are combined and used to build areas most likely to be microbusts, which are matched against expected microburst size and shape criteria, with final microburst shapes drawn around the appropriate regions. The algorithm performs comparably to the existing TDWR microburst detection algorithm on the datasets evaluated, and performance is maintained through the lower data resolutions that are anticipated for TASS scanning.
Bill Myers has led the effort to enhance the 3DTV, the Three-Dimensional Terminal Viewer , which provides the user with a perspective view of real-time aviation weather hazards in the terminal area from a variety of viewpoints, e.g., from the cockpit of any aircraft or the airport control tower. Enhancements included a high resolution airport model, a configurable air traffic simulation system and further developments of prototype 3-D aviation weather hazard glyphs (icons) all integrated into a standard simulation environment. The inclusion of the high resolution airport model along with positional data from all ground vehicles could provide a means for airport operation under poor visibility conditions. The air traffic simulation is the first step in development of an air traffic study or training system to examine the effects of different terminal area air traffic management strategies in the presence of aviation weather hazards. In contrast to simply making the glyphs cognitive extensions of the the 2-D graphical icons currently in use in the aviation weather system today, as was done in the previous version of 3DTV, the development of new glyphs to represent aviation hazards examines the use of suggesting atmospheric dynamics in the presentation of a hazard region to more intuitively communicate the nature of the hazard.
The representation of data and extracted features as an aviation weather hazard "virtual world" in which one is "immersed" was also explored. In a collaborative effort with NCAR's Scientific Computing Division (SCD), the software was ported to an immersive environment called the CAVE developed by NCSA. This environment is characterized by projecting the application onto three walls and floor of a room in stereo. When the users, wearing shutter glasses, view this scene, they have the impression of being immersed in a 3-D environment complete with depth perception. Navigation paradigms for exploring this world and previewing flight paths were also explored. This immersive system was presented by Myers at the ACM/IEEE SIGGRAPH conference.
Scientific activities at NCAR/RAP include research on thunderstorm hazards (hail, turbulence, windshear, and heavy rain), thunderstorm prediction and lifecycle evolution, in-flight icing, ground de-icing, snowfall, clear air turbulence, terrain induced windshear and turbulence, and ceiling and visibility. The end result of these research activities will be a new aviation weather products, display and system concepts. A critical step in the development of new products is to ensure that the final product and display concepts meet the user needs without impacting work load.
A number of system prototypes are being developed by Diedre Garvey, Tom Wilshire, Steve Delp, Nancy Rehak, Mike Dixon and Paul Burry that will allow users to view high resolution weather products tailored for their particular need. New high resolution weather sensing systems (radars, wind profilers, aircraft measured data, etc.) allow for the development and distribution of four-dimensional (three space dimensions plus time) datasets. The new data are processed into high resolution, 4-dimensional grids of aviation impact (icing, turbulence, ceiling, etc) and state of the atmosphere (wind, temperature, pressure) variables. NCAR/RAP is developing display capabilities that utilize the four-dimensional data and present the information to users in a tailored format. A system titled the "Aviation Weather Products Generator (AWPG)" has been developed that allows aviation users to graphically view weather at specific altitudes and along user selectable routes of flight.
During the past year, RAP has been extensively involved in working with the FAA user working groups to define the user needs and requirements for advanced aviation weather products. Input from flight service specialists and enroute traffic managers was used to design and tailor AWPG weather products particular to these functions.
An AWPG software system was developed and documented. System products include a high resolution 2-dimensional national radar mosaic, national lightning mosaic, flight advisories, and flight category (IFR, VFR, etc.). Four dimensional products (three-space dimensions plus a forecast capability) include icing potential, turbulence potential, winds, temperature, and freezing level. The four-dimensional products are derived from the state variables from both the NOAA-developed Eta and RUC models. The algorithms required for deriving the aviation impact variables (icing, turbulence, wind, temperature and freezing level) were the results of scientific research being conducted at NCAR and NOAA/FSL.
Initial product and display concepts developed for the prototype AWPG system were transferred to industry during the year. Cooperative Research and Development Agreements (CRDAs) were developed between the FAA and industry and NCAR/RAP served as the FAA's agent. The CRDAs were established to accelerate the development process with the hope that the NCAR-developed technology will make it to the commercial market quickly. CRDA participants during 1994 included Kavouras, WSI, GTE, Lockheed and Harris Corporation.
Scientific research and the development of advanced product prototype systems coupled with industry CRDA's makes a powerful technology transfer capability within RAP.
Located within the National Center for Atmospheric Research's Research Application Program (NCAR/RAP) is the Aviation Weather Development Laboratory (AWDL) . The AWDL was created as a long term research, development and demonstration laboratory to support the development and evaluation of advanced aviation weather products and other technology transfer activities. The AWDL facility is shared amongst RAP's research programs and sponsors include the FAA, the NSF, the Government of Hong Kong and NASA.
The components that make up the AWDL include science, engineering, user interface, validation, and demonstration. The AWDL concept allows for experimental weather products to be field tested at user sites such as FAA facilities (Air Traffic Control Tower, Terminal Radar Approach Control, Automated Flight Service Station), NWS facilities (Weather Forecast Office, National Aviation Weather Advisory Unit) and industry facilities (Airline Meteorology Office, Dispatch).
The AWDL rapid prototyping capability allows users to gain experience with new systems, test them in a real working environment and provide input for future development and refinement. End system users are involved with product development from concept definition through implementation. End users include pilots, controllers, traffic managers, supervisors, and airline and airport operators.
During the summers of 1991 and 1992, the NASA/Langley 737 flew microburst penetration missions in Denver, Colorado and Orlando, Florida, both sites for TDWR testbed radar operations. Part of this work centered around developing better wind shear detection methods for ground-based TDWR systems.
F-factor has become an industry standard when working with airborne systems and discussing aircraft performance. Both divergent shear and downdrafts contribute to F-factor, and each contributes depending upon how fast an event is traversed. The current TDWR algorithms do not use a performance-based metric for determining windshear related hazards; rather they use head-wind loss as the hazard criteria. While a significant and largely successful effort has been made towards stable headwind loss estimates, until recently little effort had been targeted at performance-based windshear hazard estimates from ground-based radar. NCAR, along with MIT/Lincoln Laboratory, has recently addressed this need.
Based on simulation studies by Albo and Kent Goodrich, efforts have been focussed exclusively on two-dimensional least-squares methods for estimating shear with single-Doppler radar data. While this work does not constitute a performance-based windshear detection algorithm, it is the basis for one.
By far the most difficult problem in algorithm development is "truthing". The NASA deployments offer an unprecedented opportunity to compare radar measurements of hazardous shear to what actually existed. There are many pitfalls and problems when comparing results of two such disparate systems as Doppler radar and in situ aircraft measurements: sampling volumes and sampling intervals are vastly different. Common sampling volumes and simultaneous sampling times are the exception rather than the norm. But with sufficient care, meaningful results can be extracted and such is the case here.
Kim Elmore (the primary investigator) found that 2-D F-factor estimates compare well with the NASA in situ F-factors -- correlation coefficient of 0.69 and slope of 0.74, all significant at the 99% confidence limit. One interesting result is that while the vertical wind may make up as much as 50% (about 20% on average) of the total F-factor, improving the vertical wind estimate does not significantly improve the overall statistics. Most poor statistical comparisons can be attributed to four effects: phase-type misalignments, excessive distance from the radar, disparate spatial resolution and/or excessive aircraft path and radar beam separation. In fact, results from this work indicate the radar derived F and in situ F compare better than the raw ensemble statistics at first indicate.
Dates refer to visitor's stay at NCAR during FY94. No dates are given for collaborators who did not visit NCAR.
Gunnar Aaro; Luftartsverket, Ost-Norge - Oslo ATCC; Applied Science and Engineering Groups
Joan Bauerlein; Luftartsverket, Ost-Norge - Oslo ATCC; Applied Science and Engineering Groups
Stan Benjamin; National Oceanic and Atmospheric Administration; Applied Science Group
E.A. Betterton; University of Arizona; Applied Science Group
Howie Bluestein; University of Oklahoma; Applied Science Group
V.N. Bringi; Colorado State University; Applied Science Group
John Brown; National Oceanic and Atmospheric Administration; Applied Science Group
Mary Cairns; National Oceanic and Atmospheric Administration; Applied Science Group
John Cardwell; University of Manchester Institute for Science and Technology; Manchester, England; Applied Science Group
C.H. Chan; University of Washington; Applied Science Group
Kin Sang Chim; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Jay-Chung Chen; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Don Chisholm; Phillips Lab/GPAP; Applied Science Group
Young-Il Choe; Korea Civil Aviation Bureau; Applied Science and Engineering Groups
Zou Chunshen; China Meteorological Administration; Applied Science and Engineering Groups
Chris Clarke; Colorado State University; Applied Science Group
Rodney Cole; MIT Lincoln Laboratories; Applied Science Group
Harry Colella; Martin Marietta; Applied Science and Engineering Groups
W.R. Cotton; Colorado State University; Applied Science Group
R.R. Czys; Illinois State Water Survey; Applied Science Group
Richard Delanoy; MIT Lincoln Laboratory; Applied Science Group
Paul DeMott; Colorado State University; Applied Science Group
Geoff DiMego; National Oceanic and Atmospheric Administration; Applied Science Group
Jim Evans; MIT Lincoln Laboratory; Applied Science Group
Warren Fellner; Federal Aviation Administration; Applied Science and Engineering Groups
Henry Fields; NAWAU; Applied Science Group
Cecilia Girz; National Oceanic and Atmospheric Administration; Applied Science Group
Paul Gluhosky; Yale University; Applied Science Group
Georg Grell; National Oceanic and Atmospheric Administration; Applied Science Group
Zhang Guocai; China Meteorological Administration - Beijing; Applied Science and Engineering Groups
M. Hagen; Inst. f. Physik d. Atmosphare; Applied Science Group
Art Hansen; Federal Aviation Administration; Applied Science and Engineering Groups
He Hao; Civil Aviation Administration of China; Applied Science and Engineering Groups
Hui He; Yale University; Applied Science Group
Rich Hodur; Naval Research Laboratories; Applied Science Group
Yan Hong; China Meteorological Administration - Beijing; Applied Science and Engineering Groups
C. Hreberach; University of Oklahoma; Applied Science Group
Li Huibin; Civil Aviation Administration of China; Applied Science and Engineering Groups
George Hunger; Seagull Technology; Applied Science and Engineering Groups
J.N. Hwang; University of Washington; Applied Science Group
Seiji Itoh; Matsushita Communications Industrial Co., Ltd - Yokoham, Japan; Applied Science Group
Max Ivey; Hong Kong University of Science and Technology; nology; Applied Science and Engineering Groups
Rich Jesuroga; National Oceanic and Atmospheric Administration; Applied Science Group
Xu Jia Qi; Civil Aviation Administration of China; Applied Science and Engineering Groups
Wang Jinggui; China Meteorological Administration - Beijing; Applied Science and Engineering Groups
Xue Jishan; China Meteorological Administration - Beijing; Applied Science and Engineering Groups
R.H. Johnson; Colorado State University; Applied Science Group
Sheldon Katz; Martin Marietta; Applied Science and Engineering Groups
John Keller; MIT Lincoln Laboratory; Applied Science Group
Gene Kingsbury; Federal Aviation Administration; Applied Science and Engineering Groups
Kenneth Klasinski; Federal Aviation Administration; Applied Science Group
Steven E. Koch; North Carolina State University; Applied Science Group
S.C. Kot; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Michael Kraus; National Oceanic and Atmospheric Administration; Applied Science Group
T.W Krauss.; National Hydrology Research Centre; Applied Science and Engineering Groups
Jean-Louis Laforte; University du Quebec a Chicoutimi; Applied Science Group
C.Y. Lam; Royal Observatory, Hong Kong; Applied Science and Engineering Groups
Sharon Lau; Royal Obseratory, Hong Kong; Applied Science and Engineering Groups
Chang-Soo Lee; Korea Airports Authority; Applied Science and Engineering Groups
Jean T. Lee; University of Oklahoma/CAPS; Applied Science Group
Ophelia Lee; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Thomas Lee; Naval Research Laboratories; 8 August-12 August; Applied Science Group
Kenneth Leonard; Federal Aviation Administration; Applied Science and Engineering Groups
Jennifer Mahoney; National Oceanic and Atmospheric Administration; Applied Science Group
Adrian Marroquin; National Oceanic and Atmospheric Administration; Applied Science Group
Stephen Martin; Sandia National Laboratories; Applied Science Group
Brooks Martner; National Oceanic and Atmospheric Administration; Applied Science Group
John Marwitz; University of Wyoming; Applied Science Group
G.K. Mather; Cloudquest - Nelspruit, South Africa; Applied Science Group
Warren McEvoy; Federal Aviation Administration; Applied Science and Engineering Groups
John McGinley; National Oceanic and Atmospheric Administration; Applied Science Group
Peter Meischner; Institute fur Physik d'Atmosphare; Applied Science Group
P.W. Mielke, Jr.; Colorado State University; Applied Science Group
Joe Miller; Seagull Technology; Applied Science and Engineering Groups
Andrei A. Monakov; St. Petersburg Institute of Aviation Instrument Making; Applied Science Group
Stu Muench; Phillips Lab/GPAP; Applied Science Group
Everett Nickerson; National Oceanic and Atmospheric Administration; Applied Science Group
Don Norquist; Phillips Lab/GPAP; Applied Science Group
Anders Nystrom; Swedish Institute of Space Physics; Applied Science Group
Harold Ochs; Illinois State Water Survey; Applied Science Group
B.M. Pobanz; Lawrence Livermore Laboratory; Applied Science Group
Eugene Poolman; South African Weather Bureau; Applied Science Group
Hoi-To Poon; Hong King Royal Observatory; 28 August-8 September; Applied Science Group
Kim Yong Pyo; Korean Civil Aviation Bureau; Applied Science and Engineering Groups
Jim Ramer; National Oceanic and Atmospheric Administration; Applied Science Group
Peter Ray; Florida State University; Applied Science Group
Robert Rauber; University of Illinois; Applied Science Group
Al Rodi; University of Wyoming; Applied Science Group
Paul Ruscher; Florida State University; Applied Science Group
R.W. Russell; University of California, Irvine; Applied Science Group
Tom Schlatter; National Oceanic and Atmospheric Administration; Applied Science Group
Nelson Seaman; Pennsylvania State University; Applied Science Group
Mimi Shen; Seagull Technology; Applied Science and Engineering Groups
Danny Sims; Raytheon Service Company; Applied Science Group
Sung Ki Soo; Korean Civil Aviation Bureau; Applied Science and Engineering Groups
John Sorensen; Seagull Technology; Applied Science and Engineering Groups
Boba B. Stankov; National Oceanic and Atmospheric Administration; Applied Science Group
Anders Stjernman; Swedish Institute of Space Physics; Applied Science Group
Yoon Tae Suk; Korean Civil Aviation Bureau; Applied Science and Engineering Groups
Ki-Soo Sung; Korea Aviation Bureau; Applied Science and Engineering Groups
Mark Tepper; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Merhala Thurai; 14 August-18 August; Rutherford Appleton Laboratory, Chilton, U.K.; Applied Science Group
L. Tsang; University of Washington; Applied Science Group
Francis Kwan-Leung Tse; Hong Kong University of Science and Technology; Applied Science and Engineering Groups
Joseph Turk; Colorado State University; Applied Science Group
Koji Ukena; Matushita Communications Industrial Co., Ltd. - Yokoham, Japan; Applied Science and Engineering Groups
Kenneth Van Sickle; National Science Foundation; Applied Science and Engineering Groups
J. Verlinde; Pennsylvania State University; Applied Science Group
R.M. Wakimoto; University of California, Los Angeles; Applied Science Group
Chi Wenjiang; China Meteorological Administration - Beijing; Applied Science and Engineering Groups
Wesley Wilson; MIT Lincoln Laboratory; Applied Science Group
Marilyn Wolfson; MIT Lincoln Laboratory; Applied Science and Engineering Groups
Vince Wong; University of Oklahoma; Applied Science Group
Wang Xiaomin; China Meteorological Administration; Applied Science Group
Quan Xungang; China Meteorological Administration; Applied Science Group
Tae-Seok Yoon; Korea Civil Aviation Bureau; Applied Science and Engineering Groups
Choo Young; Korea Civil Aviation Bureau; Applied Science and Engineering Groups
Xu Yu Dun; Civil Aviation Administration of China; Applied Science and Engineering Groups
Kim Jong Yui; Korea Aviation Bureau; Applied Science and Engineering Groups
He Cheng Zi; Civil Aviation Administration of China; Applied Science and Engineering Groups
Jing-Meng Zou; China Meteorological Administration; Applied Science and Engineering Groups
Peter Zwack; Ecole Nationale de la Meteorologie; Applied Science Group