Publications

Low level wind shear has been identified as a cause or contributing factor in a significant number of aviation accidents. Research has shown that the most dangerous type of wind shear is the microburst (Fujita, et al., 1977 and 1979). Briefly, a microburst is an intense local downdraft that results in a strong divergent outflow near the surface. The diameter of the outflow region may vary from 3 to 10 Km. Although many of these accidents were nonfatal, six of them resulted in a total of 550 lives lost. During the past 17 years, the mainstay of the effort by the Federal Aviation Administration (FAA) to provide wind shear warnings to pilots has been the Low Level Wind Shear Alert System (LLWAS). The system has been redesigned, based on extensive operational experience and new knowledge about the nature of the aviation wind shear hazard (Goff and Gramzow, 1989). In parallel development, the Terminal Doppler Weather Radar (TDWR) has provided a capable alternative for ground-based microburst detection (Turnbull, et al., 1989). Recent studies on the integration of LLWAS with TDWR have established the value of a combined TDWR/LLWAS wind shear detection system (Cole and Todd, 1993) The LLWAS system is being developed in four phases, I, II, III, and IV, which reflect the chronology of operational deployments. The original LLWAS, now called LLWAS I, was designed for the detection of frontal shears under the assumption that hazardous wind shear is associated with large-scale meteorological features (Goff and Gramzow, 1989). This system was deployed at 110 airports between 1977 and 1987. LLWAS I had no microburst detection capability and had excessive false alerts. LLWAS II was developed to reduce the false alert rate of LLWAS I and to provide a modest microburst detection capability. It is a direct response to recommendations by the National Research Council (NRS-NAS, 1983), following the 1982 microburst crash in New Orleans. This upgrade, deployed by modifying the software in LLWAS I, provided an improvement that would not suffer the delays and costs of the major construction that is required for off-airport LLWAS III sensors. These upgrades to LLWAS I were installed between 1988 and 1991. LLWAS II will be the operational wind shear detection system at many airports until the late '90s. LLWAS III was developed in response to the requirements that LLWAS have a microburst detection capability (NRS-NAS, 1983). This system was designed by a combination of computer simulation studies (Wilson and Flueck, 1986) and a successful field test of a prototype at Stapleton International Airport, Denver in Augist 1987 (Smythe, et al., 1989 and Wilson et al., 1991). LLWAS III combines a dense sensor network and a sophisticated Wind Shear/Microburst (WSMB) detection algoritohm to provide a substantial microburst detection capability. The prototype LLWAS III has continued to operate at Stapleton International Airport, Denver since 1987 and has been credited with the "save" of a commercial airliner on July 8, 1989. Nine LLWAS IIIs are being installed this year. LLWAS IV will be deployed at 83 airports in the late '90s. The LLWAS IV wind shear and microburst detection algorithms will be identical to LLWAS III. This system features a full hardware upgrade. Major imporvements include an ice-free sensor and hardware that is more reliable and maintainable. This report provides an evaluation of the effectiveness of LLWAS II and LLWAS III. The TDWR operational test bed at Orlando International Airport, Orlando (MCO) provides a unique data set for this evaluation. This test-bed features data from a 14-sensor LLWAS, the prototype TDWR, FL-2C, operated by MIT/LL, and the University of North Dakota meteorolgical radar (UND). Data from this test bed in the summers of 1991 and 1992 are used to provide an evaluation of LLWAS II and LLWAS III. Since LLWAS IV uses the same wind shear detection algorithm, it is expected that LLWAS III and LLWAS IV will have comparable wind shear detection capabilities.

In recent years a number of aircraft accidents have resulted from a small scale, low altitude wind shear phenomena known as a microburst. Microbursts are produced within thunderstorms and are characterized by intense downdrafts which spread out after impacting the earth's surface, displaying strong divergent outflows of wind. They are often associated with heavy rainfall, but can occur without surface rainfall (Wolfson, 1988). The Terminal Doppler Weather Radar (TWDR) program is the first system developed to detect microbursts from a ground-based radar in the airport terminal area. Improving safety is its primary goal, and test operations in Denver, Kansas City, and Orlando have shown it to be highly successful in identifying microbursts. In general, this identification has been performed with a > 90% probability of Detection (POD) and a < 10% Probability of False Alarm (PFA) (Merritt et. al., 1989). The Integrated Terminal Weather System (ITWS) will introduce several new low-level wind shear products. These products include the Microburst Prediction product, the Microburst Trend product, and an improved Microburst Detection Product. The Microburst prediction product will provide estimates of the future location, onset time, and peak intensity of microbursts before their surface effects are evident (Wolfson et. al., 1993). The Microburst Trend product is responsible for warning users about expected increases, over a two minute interval, in wind shear intensity along the approach and departure corridors of a runway. This two minute time period approximates the delay between pilot receipt of an alert and the time of actual encounter with the event. The trend product should serve to improve pilot information when making decisions involving a wind shear event. This is particularly important for currently weak, but rapidly intensifying, wind shears. The Improved Microburst Detection Algorithm being developed under the ITWS program attempts to build on the performance of the TDWR Microburst algorithm by improving POD and PFA and providing fiier localization capabilities. More importantly, enhancements to the TDWR algorithm are necessary in order to 1. provide a consistent input to the microburst trend algorithm. 2. closely relate the microburst alert to the energy loss that the aircraft will actually experience and to alerts from an on-board forward-looking Doppler radar. The TDWR algorithm does a good job detecting the microburst impacted airspace, but makes no attempt to deduce the number and centers of the events. Since the resultant alert shapes are uncorrelated over time, performing a more detailed meteorological analysis, such as location tracking, and size and intensity projections required by the microburst trend product, are compromised. This motivating factor for the improved Microburst Detection Algorithm is discussed in more detail in other works (Dasey. 1993a. Dasey, 1993b). The focus of this paper is on the second motivating factor listed above: relating the microburst alert more closely with actual aircraft performance. Much of this understanding has evolved from the analysis of data from instrumented aircraft penetrations of microbursts within the Orlando terminal area, coincident withTDWR testbed operation (Matthews and Berke, 1993.Campbell et. al., 1992). The microburst penetration flights were conducted by NASA Langley, the University of North Dakota (UND), and several manufacturers of forward-looking wind shear detection systems, including Bendix, Rockwell-Collins, and Westinghouse. Use of this data has allowed comparison of the alert representation from the TDWR Microburst algorithm with that of the initial ITWS algorithm in terms of its relationship with aircraft performance. Section 2. describes a wind shear hazard index, called the F Factor, and its estimation from a ground-based Doppler radar. The estimated F Factors from the TDWR alert shapes are described in section 3. Direct use of TDWR base data for computing shear is explored in section 4, as is the correlation of that data with aircraft F Factor measurements. Estimation of the F Factor from alert shapes output from the initial ITWS detection algorithm is explored in section 5. Section 6 examines the results and emphasizes future research.

We present an overview of the product development strategy and discuss some of the technical considerations. It will be necessary to overcome significant scientific challenges in order to be successful. Our optimism comes from the improved operational meteorological data in the terminal area, from the ability to access and to process these data rapidly, and from ongoing advances in data assimilation for mesoscale models. Our role is to coordinate the fusion of these technical and scientific advances into operational aviation weather products and to evaluate the effectiveness of these products. Major scientific contributions are anticipated from the Forecast Systems Laboratory (FSL), the National Center for Atmospheric Research (NCAR), Pennsylvania State University, and Colorado State University.

Inadvertent engine ingestion of volcanic ash has caused expensive damage to a number of aircraft recently and could have caused accidents in at least two cases [Casadevall, 1993]. Consequently, there is great interest in a real-time air traffic control (ATC) volcanic ash advisory system which could provide timely warnings of operationally significant ash concentrations to planes in flight as well as information for flight planning. The current system (see figure 1) is characterized by non-automatic determination of ash eruption characteristics (especially altitudes) with trajectory analysis based on the National Meteorological Center (NMC) forecast winds being used to provide warnings of future locations. SIGNETS and Airport Weather Advisories are the principal means of providing information on the ash locations to pilots and controllers. After one to three days, volcanic ask from Alaska can be transported over major portions of the US aviation system (figure 2) [Heffter, et al. 1990]. The operational use of the ash trajectory predictions which do not provide information on hazard associated with the ask density has resulted in more frequent disruption of air traffic. The most recent example was an incident on 19 September 1992 where a 17 September eruption from Mt. Spurr in Alaska resulted in a significant disruption of air traffic in the Upper Midwest. A workshop in Washington, DC [Machol, 1993] discussed many of these issues associated with the Spurr disruption and the operational response to ash clouds which had been drifting for several days.

The Integrated Terminal Weather System (ITWS) is an FAA-sponsored program (Sankey, 1993; Ducot, 1993) whose objective is to acquire data and products from a variety of weather sensors, integrate the data and create aviation weather products for users, such as Air Traffic (AT) controllers and traffic managers, pilots, and airline and airport operations managers. The goal of ITWS is to increase capacity at airports, reduce controller workload, and enhance safety. The objective of the ITWS Storm Cell Information (StoCel) and Weather Impacted Airspace (WIA) Detection products is to identify storm cell characteristics (echo top, echo bottom, presence of heavy rain, hail, etc.) and airspace that pilots are likely to avoid because it contains hazardous weather. The StoCel/WIA products rely on the integration of pencil-beam data and products and Air Surveillance Radar (ASR-9) Weather Channel data. ASR-9 radars are useful because they cover the entire airspace of interest, perform a volume update at roughly 30-second intervals, and will be the weather representation most widely available to the Air Traffic Control (ATC) community. On the other hand, the ASR-9 has a 4.8° fan beam which results in a vertical integration over the depth of a storm, so information on the vertical structure of storms is lost. In addition, the current ASR-9 Weather Channel may produce false weather regions during ducting or anomalous propagation (AP) conditions. Nearby WSR-88D radars also cover the entire airspace of interest and provide indications of storm vertical structure. However, the volume update rate is typically on the order of 5 to 10 minutes, depending on the scanning strategy. TDWR radars perform volume updates about every 2.5 to 3 minutes, but perform sector scans that do not cover the entire airspace. Integration of the data from these various sensors produces a product that is superior to a product based on any single sensor. Field tests of components of this algorithm were conducted at Dallas-Ft. Worth (DFW) and Orlando (MCO) International Airports during the summer of 1993. The objectives of these tests are to evaluate the technical performance of the algorithm and the validate the operational concept. This paper will describe the algorithm, and discuss the operational concept and functional requirements for the product. A summary of the results and experiences of the Summer 1993 field tests, and a preliminary evaluation of the performance of the algorithm based on off-line and real-time tests will be provided at the conference.

The Federal Aviation Administration (FAA) initiated the Terminal Doppler Weather Radar (TDWR) program in the mid-1980's in response to the need for improved real-time hazardous weather (especially low-altitude wind shear) surveillance in the terminal area (Turnbull, et al., 1989). The initial focus for the TDWR was to provide reliable, fully automated Doppler radar detection of microbursts and gust fronts and 20-minute warning of wind shifts which could effect runway usage. Subsequent operational demonstrations have shown that the overall terminal situational awareness provided by the TDWR color Geographical Situation Display (GSD) depiction of wind shear locations, weather reflectivity and storm motion also yields substantial improvements in terminal operations efficiency for air traffic managers and for airlines. In this paper, we will describe the current status and deployment strategy for the operational systems and recent results from the extensive testing of the radar system concept and of the weather information dissemination approach.

In order to efficiently use available runway capacity while avoiding undue congestion within terminal airspace, systems of flow control and en route metering have been implemented. Recent work in automation has attempted to extend traffic flow planning to provide precise scheduling of traffic flow within the terminal area itself (from the metering fixes to the runways). The goal of this more detailed terminal scheduling is more efficient runway utilization. This article addresses an important practical question regarding the degree of precision required from the en route portion of such systems in order to allow the terminal scheduler to achieve its throughput benefits. The answer to this question determines the sophistication and rigidity required of en route automation and addresses the question of whether the success of new terminal automation is contingent upon improvements in en route metering. The method of analysis is mathematical modeling and fast-time computer simulation. A crucial parameter is controllability, which expresses the largest flight delay that the terminal scheduling can impose within the airspace available to it. The analysis reveals that achievable run-way utilization depends upon the type of metering employed, the available controllability within the terminal, and the extent to which controllers can be expected to intervene to handle transient peaks in arrival rates that cannot be handled by the automation. The major conclusion of the study is that in order to fully utilize a runway, the standard deviation of the errors in arrival time at the metering fixes should be kept to about half the terminal controllability. For the airports studied, there seems to be sufficient controllability available to allow a terminal scheduler to operate the runways at essentially full capacity when a metering system, even with modest delivery precision, is operating in the en route area.

Mode-S is an enhancement of the ATCRBS secondary surveillance radar (SSR) system which adds selective interrogation of individual aircraft, monopulse processing of the replies and a digital data link between the ground station and the aircraft. These features result in greatly improved surveillance accuracy, virtual elimination of synchronous garble of the replies from closely spaced aircraft, and provide a high capacity digital communication link for a wide variety of ground/air/ground messages.

Pattern-classification and clustering algorithms are key components of modern information processing systems used to perform tasks such as speech and image recognition, printed-character recognition, medical diagnosis, fault detection, process control, and financial decision making. To simplify the task of applying these types of algorithms in new application areas, we have developed LNKnet-a software package that provides access to more than 20 pattern-classification, clustering, and feature-selection algorithms. Included are the most important algorithms from the fields of neural networks, statistics, machine learning, and artificial intelligence. The algorithms can be trained and tested on separate data or tested with automatic cross-validation. LNKnet runs under the UNM operating system and access to the different algorithms is provided through a graphical point-and-click user interface. Graphical outputs include two-dimensional (2-D) scatter and decision-region plots and 1-D plots of data histograms, classifier outputs, and error rates during training. Parameters of trained classifiers are stored in files from which the parameters can be translated into source-code subroutines (written in the C programming language) that can then be embedded in a user application program. Lincoln Laboratory and other research laboratories have used LNKnet successfully for many diverse applications.

This report demonstrates, through data and analysis, how the airport environment can affect ATCRBS surveillance. The Lincoln Laboratory ATCRBS Monopulse Processing Subsystem was used to collect reply data at Memphis International Airport during March 1991. These data show a correlation between aircraft density, potential reflectors, and ATCRBS reply integrity. The number of replies has been shown to be directly related to multipath from reflecting surface, including taxiing aircraft. Additionally, it is shown that conditions can exist during which not all of the replies from ATCRBS equipped aircraft can be processed when forming target report measurements. Finally, it is shown that the bunching of replies in both time and space can introduce reply decoder overloading.